Datasets:

zekeZZ

/

wmdp-bio-retain-corpus-datamodel-sorted

like 0

The authors confirm that all data underlying the findings are fully available without restriction. The sequence data generated in this study are available in the Gene Expression Omnibus database at NCBI under the series accession number GSE57326. Whole C. roseus transcriptome sequence along with functional annotation, expression profiling and identified SSRs are available at the Catharanthus Transcriptome web page (<http://nipgr.res.in/mjain.html?page=catharanthus>). Introduction {#s1} ============ *Catharanthus roseus*, popularly known as Madagascar periwinkle, is a medicinal plant which belongs to family Apocyanaceae. The plant is diploid (2n = 16) and native to islands of Madagascar, but now grown in many tropical countries as ornamental plant [@pone.0103583-Magnotta1]. *C. roseus* is well known for its pharmacological importance as it produces more than 130 terpenoid indole alkaloids (TIAs) including vinblastine and vincristine, which are widely used in anti-cancer chemotherapies [@pone.0103583-VanderHeijden1], [@pone.0103583-Guimaraes1]. Most tissues of *C. roseus* are known to produce alkaloids and no other single plant is known to produce such a wide spectrum of alkaloids [@pone.0103583-Blasko1]. The plant is known to treat diabetes also, due to hypoglycemic properties in its tissue extracts [@pone.0103583-Nammi1]. Moreover, roots of *C. roseus* are known to accumulate ajmalicine and serpentine which help controlling blood pressure and cardio-vascular disorders [@pone.0103583-Svoboda1]. Alkaloid biosynthetic pathways are highly branched and complex, with wide differences in alkaloid composition between underground and aerial tissues. TIAs have high commercial value because they are produced by plants in very low amounts and its infusion is very difficult. The common precursor of TIAs, strictosidine, is the central intermediate formed by the condensation of tryptamine (product of shikimate pathway) and secologanin (product of non-mevalonate pathway) involving strictosidine synthase (STR). Pharmacologically important alkaloids, vinblastine and vincristine (found only in aerial tissues) are synthesized *in vivo* by the condensation of vindoline and catharanthine, both of which are obtained from branch-point intermediate cathenamine. Biochemical pathway resulting in formation of vindoline is specifically present in well differentiated aerial tissues of the plant, but not in roots and cell cultures, thereby marking the presence of tissue-specific TIA pathway in *C. roseus* [@pone.0103583-Rischer1]--[@pone.0103583-Jaggi1]. In recent years, next generation sequencing has become the method of choice for fast and cost-effective transcriptome characterization for non-model plants [@pone.0103583-Morozova1]--[@pone.0103583-Garg1]. Earlier, a large-scale transcriptomic resource from three medicinal plants (*Camptotheca acuminate*, *Catharanthus roseus* and *Rauvolfia serpentina*) have been developed for elucidating monoterpene indole alkaloid (MIA) pathways [@pone.0103583-GongoraCastillo1]. Recently, Van Moerkercke et al. [@pone.0103583-VanMoerkercke1] used RNA-seq approach to construct metabolic pathway database, CathaCyc, for *C. roseus*. Such databases can facilitate identification of key regulator(s) for metabolic pathway engineering. To add on to existing resources, in this study, we generated *C. roseus* transcriptome by assembling RNA-seq data generated from *C. roseus* tissues (leaf, flower and root) and merged it with previously reported *C. roseus* transcripts. The updated comprehensive *C. roseus* transcriptome was screened for simple sequence repeats (SSRs) which might be helpful in development of functional molecular markers. We also identified transcription factor (TF) encoding transcripts in *C. roseus* transcriptome as only few TFs are known, which regulate TIA pathway genes. Expression analysis of genes involved in TIA pathways was also undertaken to reveal their tissue-specific expression. Gene ontology (GO) enrichment analysis highlighted the tissue-preferential/specific expression of transcripts in various biological processes. These data will provide a framework for further functional analysis of genes involved in biosynthesis of important alkaloids. Results {#s2} ======= Transcriptome sequencing and preprocessing of data {#s2a} -------------------------------------------------- To generate the transcriptome of *C. roseus*, three tissue samples (leaf, root and flower) were subjected to next generation sequencing using Illumina platform. Of the total ∼347 million reads generated from all the tissue samples, about 343 million reads were found to be of high quality after filtering with NGS QC Toolkit ([Table S1](#pone.0103583.s008){ref-type="supplementary-material"}), having average Phred quality score of at least 30 at each base position. As the short reads obtained may be redundant (due to PCR amplification at library preparation step) and their assembly needs a high-end server with high random access memory (RAM). Therefore, duplicate reads from each sample were removed and about 230 million non-redundant (NR) reads were obtained ([Table S1](#pone.0103583.s008){ref-type="supplementary-material"}). Optimization and validation of transcriptome assembly {#s2b} ----------------------------------------------------- To generate an optimal transcriptome assembly of *C. roseus*, we systematically compared the performance of various *de novo* short read assembly tools, including Velvet, Oases and ABySS. *De novo* assembly of total (343,384,084) and NR (230,715,698) high quality reads was performed employing a two-step approach. In the first step, primary assembly (best k-mer assembly) was generated using Velvet, Oases and ABySS at different k-mer lengths ranging from 31 to 95 ([Table S2](#pone.0103583.s009){ref-type="supplementary-material"} and [S3](#pone.0103583.s010){ref-type="supplementary-material"}). On the basis of several parameters described earlier [@pone.0103583-Garg2], assemblies obtained from respective assemblers at different k-mers were compared. Assemblies generated by Velvet showed a gradual increase in N50 and average read length with the k-mer length, best being at k-93 ([Table S2A and S3A](#pone.0103583.s009){ref-type="supplementary-material"}). Similarly, assembly at k-93 (from total reads; [Table S2C](#pone.0103583.s009){ref-type="supplementary-material"}) and k-87 (from NR reads; [Table S3C](#pone.0103583.s010){ref-type="supplementary-material"}) had higher N50 and average read lengths, and were considered to be the best assembly generated from ABySS. On the other hand, choosing the best assembly generated from Oases (using NR dataset) was a tricky task, as there was not much difference in N50 and average lengths at different k-mers ([Table S2B](#pone.0103583.s009){ref-type="supplementary-material"}). Finally, assembly at k-57 was selected, which had an optimal assembly size (number of contigs; [Table S3B](#pone.0103583.s010){ref-type="supplementary-material"}) and minimum redundant unigenes. Whereas, assembly of Oases at k-61 generated from total dataset had higher N50 and average lengths ([Table S2B](#pone.0103583.s009){ref-type="supplementary-material"}). By taking all the assembly parameters into consideration along with BLAST results, assembly generated by Oases at k-57 using NR short read dataset (NR-Oases k-57), was considered to be the best, which generated 42909 contigs (≥250 bp) of 1161 bp average length and 1990 bp N50 length ([Table 1](#pone-0103583-t001){ref-type="table"}). 10.1371/journal.pone.0103583.t001 ###### Assembly optimization/validation of Illumina data of *C. roseus*. {#pone-0103583-t001-1} Total high quality reads (best k-mer) Non-redundant high quality reads (best k-mer) MPGR assembly Merged assembly[2](#nt102){ref-type="table-fn"} -------------------------------------------------------------------------- --------------------------------------- ----------------------------------------------- --------------- ------------------------------------------------- ------- ------- ------- -------- -------- -------- **Number of contigs** 133650 39010 128716 106736 71190 39276 53017 42909 86726 59220 **Total size (Mb)** 72.78 25.38 165.01 160.92 57.62 25.62 51.86 49.81 107.79 76.03 **Minimum length (bp)** 100 185 100 250 100 185 107 250 251 250 **Maximum length (bp)** 15524 10893 17071 17071 15524 10913 17112 17112 12048 17141 **Average length (bp)** 544.6 650.7 1282 1507.8 809.5 652.3 978.2 1160.8 1243 1283.9 **N50 length (bp)** 1087 848 2161 2205 1400 841 1911 1990 1683 2115 **Contigs with significant similarity** [1](#nt101){ref-type="table-fn"} 63706 28142 76866 73292 41643 28477 22593 21275 66788 32666 Similarity search was done against TAIR10 proteome. Assembly generated by merging best k-mer with MPGR transcriptome using TGICL. Gongora-Castillo et al. [@pone.0103583-GongoraCastillo1] followed a robust approach to generate a comprehensive *C. roseus* transcriptome from different tissues and treatments. Our second step involved merging of primary assembly (best k-mer; NR-Oases k-57) with previously existing *C. roseus* assembly (MPGR de novo assembly; [@pone.0103583-GongoraCastillo1]). Our earlier studies have shown that TGICL software generates optimal merged assemblies [@pone.0103583-Garg1], [@pone.0103583-Agarwal1]. Unigenes from both the assemblies were size selected (≥250 bp) and transcript isoforms from MPGR assembly were removed (retaining the longest isoform) before subjecting to assembly using TGICL program. The merged assembly resulted in a total of 59220 contigs with improved average length (1284 bp) and increased N50 read length (2115 bp; [Table 1](#pone-0103583-t001){ref-type="table"}). Overall, the merged assembly was found to be much better than those reported previously [@pone.0103583-GongoraCastillo1], [@pone.0103583-Kumar2]. To assess the quality of *C. roseus* transcriptome thus obtained, we checked for the presence of publicly available *C. roseus* sequences in recently assembled transcriptome. Out of 287 full-length protein sequences (downloaded from NCBI), 220 (77%) were found to be present in the assembled transcriptome. Moreover, we also checked for the genes involved in various biochemical pathways and found that all the 108 genes previously reported by Van Moerkercke et al. [@pone.0103583-VanMoerkercke1] were represented in our transcriptome data. Similarly, all the 30 enzymes (genes) known to be involved in TIA bio-synthesis [@pone.0103583-VanMoerkercke1], were also present in the *C. roseus* transcriptome generated in this study. As compared with earlier reported transcriptome assemblies of *C. roseus*, we obtained nearly 19% (MPGR assembly) [@pone.0103583-GongoraCastillo1] and 42% (CathaCyc) [@pone.0103583-VanMoerkercke1] novel transcripts in our assembly. *C. roseus* belongs to the clade Asterids, and genome of three plant *species* of this clade, including *Solanum tuberosum* (potato), *Solanum lycopersicum* (tomato) and *Sesamum indicum* (sesame) have been sequenced so far. A BLAST analysis of *C. roseus* transcriptome against proteomes of tomato, potato, cucumber, grapevine and Arabidopsis, and transcriptomes of six known alkaloid producing plants (*Atropa belladonna*, *C. acuminata*, *Cannabis sativa*, *R. serpentine*, *Rosmarinus officinalis* and *Valeriana officinalis*) revealed higher similarity of *C. roseus* transcripts with *R. officinalis* (60.9%) followed by tomato (58.7%), potato (56.3%), cucumber (56.2%) and grapevine (56.1%) ([Fig. S1](#pone.0103583.s001){ref-type="supplementary-material"}). Further, reciprocal BLAST analysis with annotated protein sequences from closest reference genomes (tomato, potato, cucumber and grapevine) and Arabidopsis showed that, although Arabidopsis happens to be distantly related to *C. roseus* in phylogenetic tree but had the highest number of orthologs (15252) as compared to tomato (12118) and potato (11263), which belong to the same clade (Asterids) as that of *C. roseus* ([Fig. S2](#pone.0103583.s002){ref-type="supplementary-material"}). This may be due to availability of better genome annotation of the model plant Arabidopsis. Cucumber and grapevine had the least number of orthologs in *C. roseus*, because they belong to different clades. *C. roseus* transcripts generated above were designated as *C. roseus* tentative consensus (Cr_TC) and were assigned a unique identifier number from Cr_TC00001 to Cr_TC59220. The whole transcriptome sequence is available at Catharanthus Transcriptome Sequence web page (<http://nipgr.res.in/mjain.html?page=catharanthus>). The total size of transcriptome is ∼76 Mb with nearly 65% of the transcripts longer than 500 bp and more than 40% transcripts larger than 1000 bp ([Fig. S3](#pone.0103583.s003){ref-type="supplementary-material"}). Average GC content of *C. roseus* transcriptome was little lower (40.65%) than Arabidopsis (42.5%; [Fig. S4](#pone.0103583.s004){ref-type="supplementary-material"}), and comparable with that of soybean (40.9%) and chickpea (40.3%) [@pone.0103583-Garg2]. Whereas, average GC content of rice was much higher (55%) with respect to *C. roseus* and other dicot plant species analyzed. Functional annotation of *C. roseus* transcriptome {#s2c} -------------------------------------------------- For comprehensive annotation of *C. roseus* transcripts, similarity search was performed against several public databases sequentially. We were able to annotate 38380 (65%) unigenes with confidence (e-value≤1E-05), while others were considered to be *C. roseus* - specific which may be involved in various important biochemical pathways, whose intermediates and enzymes involved have not been catalogued in public repositories as of now. The putative function assigned to the transcripts is available at Catharanthus Transcriptome Sequence web page. Based on their similarity with Arabidopsis genes, *C. roseus* transcripts were assigned GOSlim terms under biological process, molecular function and cellular components categories ([Fig. S5A](#pone.0103583.s005){ref-type="supplementary-material"}). Among the biological process category, maximum number of *C. roseus* transcripts were assigned with the specific term, protein metabolism (13.17%) followed by response to stress (12.02%). GOSlim terms, nucleotide binding (9.81%), hydrolase activity (8.32%), transferase activity (8.03%) and protein binding (7.36%) were most represented under molecular function category. Among the cellular component category, nucleus (19%) followed by other cytoplasmic component (18.8%) were most represented ([Fig. S5A](#pone.0103583.s005){ref-type="supplementary-material"}). Based on COG (cluster of orthologous groups) classification, at least 33422 (56.43%) transcripts could be classified into 25 COG categories. Among the 25 COG categories, the cluster for general function prediction represented the largest group (7124; 21.31%), followed by post-translational modification, protein turnover, chaperones (3809; 11.40%) and signal transduction mechanisms (3380; 10.11%). In addition, 1941 (5.8%) of *C. roseus* transcripts were assigned into the cluster of unknown function ([Fig. S5B](#pone.0103583.s005){ref-type="supplementary-material"}). To elucidate various biochemical pathways represented in the transcriptome, *C. roseus* transcripts were searched against Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database, which aids in studying the role of gene(s) in complex metabolic pathways. In total, 4436 genes (4738 transcripts) were found to be involved in one of the 318 different KEGG metabolic pathways. Few of the metabolic pathways represented with higher number of genes were, ribosome (117 genes), spliceosome (98 genes), biosynthesis of amino acids (96 genes), RNA transport (87 genes), purine metabolism (82 genes), carbon metabolism (80 genes), oxidative phosphorylation (73 genes), pyrimidine metabolism (71 genes) and protein processing in endoplasmic reticulum (71 genes). Along with plant hormone signal transduction (38 genes) pathway, we found genes involved in various alkaloid biosynthesis pathways also, like terpenoid backbone biosynthesis (26 genes), phenylalanine tyrosine and tryptophan biosynthesis (23 genes), ubiquinone and other terpenoid-quinone biosynthesis (17 genes), tryptophan metabolism (10 genes), tropane piperidine and pyridine alkaloid biosynthesis (eight genes), diterpenoid biosynthesis (11 genes), isoquinoline alkaloid biosynthesis (seven genes), sesquiterpenoid and triterpenoid biosynthesis (five genes), monoterpenoid biosynthesis (two genes) and indole alkaloid biosynthesis (one gene). We also identified TF-encoding genes in *C. roseus* transcriptome and found 1820 transcripts representing 73 TF families. Among the 73 families represented, the MYB-domain (116) family TFs were the most abundant followed by AP2-EREBP- (106), WRKY- (78), bHLH- (75) and HB- (75) domain TFs ([Fig. 1](#pone-0103583-g001){ref-type="fig"}). TFs play an important role in secondary metabolite and TIA accumulation [@pone.0103583-Jaggi1], [@pone.0103583-VanMoerkercke1]. TFs known to be involved in TIA pathway like ORCA2, ORCA3, WRKY, MYC2 and zinc finger DNA-binding protein 1 and 2, were represented in the assembled transcriptome of *C. roseus*. {#pone-0103583-g001} Identification of simple sequence repeats {#s2d} ----------------------------------------- Transcriptome resources have been harnessed for mining of SSRs in several plant species. EST-SSRs provide an insight on the density of SSRs in the transcribed region of genome and have higher rates of transferability across species [@pone.0103583-Varshney1]. *C. roseus* transcriptome was screened for SSRs using MISA search tool and a total of 11620 SSRs were identified in 8644 (14.6%) *C. roseus* transcripts. Among the identified EST-SSRs, di-nucleotide repeats were most represented (55.55%), followed by tri-nucleotide repeats with 41.67% (4842; [Fig. 2A](#pone-0103583-g002){ref-type="fig"}). Among di-nucleotide repeats, AG/CT showed highest occurrence (60.84%), followed by AT/AT (31.73%), AC/GT (7.41%) and CG/CG (0.03%). In case of tri-nucleotide repeats, occurrence of various motifs was uniform except for AAG/CTT, which showed highest frequency (34%) and CCG/CGG being the least abundant (2.15%) ([Fig. 2B](#pone-0103583-g002){ref-type="fig"}). Further, we developed a comprehensive SSR marker resource for *C. roseus* by designing forward and reverse primers from their flanking sequences. In total, we could design primers for 7158 (61.6%) SSR repeat-motifs identified, which can be used for the generation of functionally relevant markers in *C. roseus*. The complete list of SSRs identified in *C. roseus* along with primer sequences are available at Catharanthus Transcriptome Sequence web page. {#pone-0103583-g002} Differential gene expression and gene ontology enrichment analysis {#s2e} ------------------------------------------------------------------ RNA-Seq has been considered to be the method of choice for differential gene expression studies at whole genome level [@pone.0103583-Ozsolak1], [@pone.0103583-Jain2]. In total, approximately 87--90% short reads mapped onto *C. roseus* transcriptome and nearly 84--87% mapped uniquely ([Table 2](#pone-0103583-t002){ref-type="table"}). DESeq package, was used to identify the genes differential expressed in different tissue samples [@pone.0103583-Anders1]. We identified differentially expressed genes among different tissues of *C. roseus* via pairwise comparisons ([Fig. 3A](#pone-0103583-g003){ref-type="fig"}). Leaf and root being the primary site for alkaloid production, differential gene expression in these tissues as compared to flower was analyzed in more detail. A total of 2443 and 2153 genes were differentially expressed in leaf and root, respectively, as compared to flower tissue ([Fig. 3A, B](#pone-0103583-g003){ref-type="fig"}). In leaves, higher number of genes (1635) were down-regulated than 808 up-regulated genes, whereas in roots, there were nearly equal number of up- (1125) and down-regulated (1028) genes ([Fig. 3A](#pone-0103583-g003){ref-type="fig"}). Out of the total 4596 differentially expressed genes in leaf and roots, 679 were common. Among 679 genes commonly differentially expressed in roots and leaves, 72 and 552 genes were up- and down-regulated, respectively ([Fig. 3B](#pone-0103583-g003){ref-type="fig"}). Fifty five genes were found to be up-regulated in one tissue and down-regulated in another tissue. The genes related to photosynthesis were up-regulated in leaf, whereas genes annotated as DNA binding proteins (TFs), disease-resistant proteins and wound response proteins were up-regulated in roots. A heat-map of 1861 up-regulated genes in at least one tissue is represented in [Fig. S6](#pone.0103583.s006){ref-type="supplementary-material"}. Among 1861 up-regulated genes, at least 148 were found to encode for TFs, which are key regulatory proteins. TFs are known to play an important role in accumulation of secondary metabolites in plants [@pone.0103583-Broun1]--[@pone.0103583-Patra1]. Among 148 TFs exhibiting significant differential expression in leaf and root tissues, majority of the transcription factors belonging to AP2-EREBP, HB, MYB, NAC, Tify and WRKY families were up-regulated in root tissues ([Fig. 3C](#pone-0103583-g003){ref-type="fig"}). {#pone-0103583-g003} 10.1371/journal.pone.0103583.t002 ###### Mapping of non-redundant high-quality reads on *C. roseus* transcriptome. {#pone-0103583-t002-2} Tissue samples High quality reads Total mapped reads (%) Uniquely mapped reads (%) ---------------- -------------------- ------------------------ --------------------------- **Leaf** 79025564 70974531 (89.81) 68635709 (86.85) **Flower** 78728416 69955703 (88.86) 67774853 (86.09) **Root** 72961718 63571877 (87.13) 61747179 (84.63) We performed GO enrichment analysis to explore the major functional categories in up-regulated genes in leaves and roots. GO terms associated with various biological processes, such as metabolic process, nucleic acid metabolic process and cellular metabolic process were found to be enriched in up-regulated genes of leaf ([Fig. S7A](#pone.0103583.s007){ref-type="supplementary-material"}). Leaves being actively participating in photosynthesis, GO terms associated with photosynthetic process were also significantly enriched in leaves. Apart from these, biological process GO terms, like cellular response to jasmonic acid stimulus, ion homeostasis, carotenoid, isoprenoid and tertraterpenoid metabolic process were significantly enriched in genes up-regulated in leaf ([Fig. 3D](#pone-0103583-g003){ref-type="fig"}). Likewise, response to jasmonic acid stimulus was also significantly enriched in up-regulated genes of root. Thus, supporting the previous remarks that plant hormone jasmonic acid is one of the main drivers of TIA synthesis in *C. roseus* and plant secondary metabolism in general [@pone.0103583-Rischer1], [@pone.0103583-DeGeyter1]. Roots being an underground tissue is subjected to various biotic stresses present in the rhizosphere. GO terms like response to stress, response to biotic stimulus, defense response and response to fungus were significantly enriched in up-regulated root genes under biological process category. Regulation of transcription, cellular amino-acid derivative metabolic process, jasmonic acid biosynthesis, response to chemical stimulus, response to endogenous stimulus and response to salicylic acid were few other biological process GO terms, which were significantly enriched in up-regulated genes of roots ([Fig. S7B](#pone.0103583.s007){ref-type="supplementary-material"}). Expression profiling and validation of genes involved in TIA pathway {#s2f} -------------------------------------------------------------------- For expression analysis, we mapped the short reads of individual sample from our study and previous study [@pone.0103583-GongoraCastillo1] onto our *C. roseus* transcriptome and analyzed expression profile using DESeq software. Complete expression analysis of genes is available at Catharanthus Transcriptome Sequence web page. This resource can be utilized by researchers to look for expression profile of their gene(s) of interest. *C. roseus* (L.) var. Prabal is well known for its high alkaloid content [@pone.0103583-Dwivedi1] and we were interested in expression analysis of important alkaloids (TIA) biosynthesis pathway genes. We used CathaCyc database [@pone.0103583-VanMoerkercke1] for the analysis of these pathways. The common precursor of TIAs, strictosidine, is the central intermediate formed by the coupling of tryptamine (shikimate pathway) and monoterpene secologanin (methyl erythritol phosphate pathway) as shown in [Fig. 4A](#pone-0103583-g004){ref-type="fig"}. The alkaloid, vinblastine, is synthesized by coupling of vindoline and catharanthine, both of which are obtained from branch-point intermediate cathenamine ([Fig. 4A](#pone-0103583-g004){ref-type="fig"}). We identified *C. roseus* transcripts encoding for most of the enzymes catalyzing different reactions involved in these pathways ([Fig. 4A](#pone-0103583-g004){ref-type="fig"}). {#pone-0103583-g004} We identified 30 genes well-known to be involved in TIA biosynthetic pathways. The BLAST analysis showed that all of these genes were conserved in the sequenced genomes from Asterid clade (tomato and potato) and Arabidopsis at the protein level. However, only 17 (∼57%) and 22 (∼73%) of them exhibited significant similarity with annotated coding region sequences of Arabidopsis and Asterids (tomato and potato), respectively, at the nucleotide level. Further, we analyzed the expression of TIA biosynthetic pathway genes using RNA-Seq data in different tissue samples and treatments reported in our study and previous studies [@pone.0103583-GongoraCastillo1], [@pone.0103583-VanMoerkercke1]. As shown in the [Fig. 4B](#pone-0103583-g004){ref-type="fig"}, majority of the genes of TIA pathway were up-regulated in leaf and root tissues implying that these alkaloids are synthesized mainly in leaves and roots. Similar pattern of expression is also visible in developmental tissues used in the study by Gongora-Castillo et al. [@pone.0103583-GongoraCastillo1]. Both leaf and root tissues share more or less a common gene expression pattern for most TIA pathway genes, except for tabersonine 16-hydroxylase (Cr_TC35206) and deacetylvindoline 4-O-acetyltransferase (Cr_TC35622), which are highly down-regulated in roots and tabersonine 19-hydroxylase (Cr_TC04217), which is highly down-regulated in leaves ([Fig. 4B](#pone-0103583-g004){ref-type="fig"}). Apart from tabersonine 16-hydroxylase and deacetylvindoline 4-O-acetyltransferase, decreased expression of desacetoxyvindoline 4-hydroxylase was seen in root and hairy root cultures ([Fig. 4B](#pone-0103583-g004){ref-type="fig"}). Down-regulation of these enzymes in root tissues and hairy root cultures is in agreement with previous studies [@pone.0103583-Shukla1], [@pone.0103583-StPierre1], as they participate in terminal reactions for vindoline biosynthesis, which is restricted to aerial tissues. On the other hand, tabersonine 19-hydroxylase, which was found to be up-regulated in roots, further endorses previous finding that it helps to operate an alternate mechanism for tabersonine metabolism in roots by side-chain hydroxylation [@pone.0103583-Giddings1]. Gene expression analysis using RNA-seq data revealed that many of the genes involved in TIA pathway were differentially expressed in root and leaf tissues. To validate these findings, quantitative RT-PCR was performed for at least 10 genes of TIA pathway detected to be differentially expressed in root and leaf tissues. Real-time RT-PCR analysis revealed similar expression patterns of all the selected genes as observed in RNA-seq data. Moreover, the statistical analysis also showed a very good correspondence (correlation coefficient of 0.80) among the results of real time RT-PCR and RNA-seq data analysis as shown in [Fig. 4C](#pone-0103583-g004){ref-type="fig"}. Suspension culture supplemented with yeast extract does not seem to be an attractive approach for TIA production due to lower gene expression as seen in gene expression profile ([Fig. 4B](#pone-0103583-g004){ref-type="fig"}). On the other hand, seedlings treated with MeJa showed increased expression, after exposure for longer duration i.e. 5 and 12 days. Expression of TIA genes was relatively higher in hairy root cultures, however it escalated when subjected to MeJa treatment ([Fig. 4B](#pone-0103583-g004){ref-type="fig"}). Expression profiling of TFs known to be involved in TIA pathway across different tissues and treatments revealed that their expression is highly up-regulated in hairy root culture, stem and root ([Fig. 4D](#pone-0103583-g004){ref-type="fig"}). Similar to earlier observations, there was a growth related decrease in TIA transcripts in *C. roseus* and accumulation of bisindole alkaloid content depends on tissue maturity [@pone.0103583-Shukla1], [@pone.0103583-StPierre1], [@pone.0103583-Naaranlahti1], we also found that the expression of TIA genes diminished in the mature leaf tissue. Discussion {#s3} ========== *C. roseus* is widely known for its pharmaceutical potential and has become one of the extensively studied medicinal plants. It is considered to be single biological source of the anti-cancer compounds, vinblastine and vincristine [@pone.0103583-VanderHeijden1], [@pone.0103583-ElSayed1]. Although many good efforts have been made to elucidate the complete pathway of TIA biosynthesis, but few complex steps and intermediate compounds are still unknown. Transcriptome studies, with the advent of next generation sequencing technologies, can help addressing few of these problems via gene discovery. Here, we performed high-throughput sequencing of transcriptome from different tissues of *C. roseus* and used short read assembly tools (Velvet, Oases and ABySS) for *de novo* assembly optimization. A two-step strategy involving merging of best k-mer assembly (from out data) with earlier reported MPGR assembly [@pone.0103583-GongoraCastillo1] was employed to obtain a robust *C. roseus* transcriptome. Based upon various parameters [@pone.0103583-Garg2], assembly generated from Oases at k-mer length of 57 taking NR short reads was considered to be the best. This is in agreement with a previous study by Ghangal et al. [@pone.0103583-Ghangal1] who also reported that assembly generated from NR reads was better than total reads. Merging of best k-mer assembly (NR-Oases-k-57) with MPGR assembly using TGICL further improved the assembly assessment parameters, such as N50 length (2115 bp), average transcript length (1283 bp) and sequence similarity with closely related species. As sequence similarity also marks the completeness of the transcriptome, 77% of the known full-length *C. roseus* proteins were found to be present in our *C. roseus* transcriptome. We also found all the previously reported [@pone.0103583-VanMoerkercke1] genes involved in TIA bio-synthesis (30 genes) represented in our assembled transcriptome. The presence of already reported important alkaloid biosynthetic genes and full-length proteins marks the quality of *C. roseus* transcriptome. Moreover, BLAST analysis with earlier transcriptome sequences of *C. roseus* [@pone.0103583-GongoraCastillo1], [@pone.0103583-VanMoerkercke1] and other related plant proteome and transcriptome sequences revealed a better transcriptome assembly presented in our study. For comprehensive annotation, *C. roseus* transcriptome was subjected to similarity search against various known protein databases. We were able to annotate about 65% of *C. roseus* transcripts. Recently discovered genes encoding geraniol synthase (GES) [@pone.0103583-Simkin1] and iridoid synthase (IS) [@pone.0103583-GeuFlores1], known to be involved in biosynthesis of secologanin (a monoterpenoid alkaloid) from geranyl pyrophosphate were also present in our *C. roseus* transcriptome. Overall, more than 56% of transcripts were classified into 25 COG categories, which is quite higher than other studies [@pone.0103583-Lai1]--[@pone.0103583-Wei1]. We found the category "general function prediction" to be the most represented in COG classification accounting for its need for basic physiological and metabolic functions. The GC content of *C. roseus* transcriptome was found to be very much similar to other dicot plants. Our results concord with earlier findings that there is only marginal variation in average GC content between dicots like Arabidopsis, soybean, tomato, potato, pea and tobacco [@pone.0103583-Carels1]. Many studies have been undertaken to characterize and differentiate different *C. roseus* cultivars using various molecular markers, such as AFLP [@pone.0103583-Kim1], [@pone.0103583-ElDomyati1], RAPD [@pone.0103583-Kim1], [@pone.0103583-Shaw1], ISSR [@pone.0103583-ElDomyati1] and SSRs [@pone.0103583-Shokeen1]--[@pone.0103583-Mishra1]. Microsatellites are co-dominant molecular markers used for marker-assisted selection studies and their identification from high-throughput transcriptome studies have been reported in large number of plant species. A total of 11620 SSRs of 2--6 nucleotides were predicted in *C. roseus* transcripts with di-nucleotides repeats being most abundant followed by tri-nucleotide repeat. This is in accordance with previous studies on *C. roseus*, who also observed more di-nucleotide repeats than tri-nucleotide repeats in the EST datasets [@pone.0103583-Mishra1]. The availability of a large number of SSRs with primer sequences can help large-scale genotyping studies for various applications. Existence of genetic diversity in *C. roseus* have been demonstrated by developing STMS markers [@pone.0103583-Shokeen1], [@pone.0103583-Shokeen2]. Thus, availability of transcriptome for screening of SSRs hold an immense potential for high-throughput genotyping applications in *C. roseus*. TFs are key regulators that can alter the gene expression of several target genes, thereby can regulate metabolic flux. Members of some TF families, such as MYB, AP2-EREBP, WRKY, MYB-related and bHLH, are known to regulate secondary metabolism in plants [@pone.0103583-Kato1]--[@pone.0103583-Zhang1]. Members of the plant-specific AP2-EREBP TF family, namely octadecanoid-derivative responsive Catharanthus AP2-domain protein (ORCA2 and ORCA3), which are known to activate expression of several genes (enzymes) involved in TIA biosynthesis, were identified in our *C. roseus* transcriptome. Recently, Suttipanta et al. [@pone.0103583-Suttipanta1] characterized CrWRKY1 and reported its involvement in the transcriptional regulation of TIA pathway in *C. roseus*. Apart from AP2-domain and WRKY proteins, we also found MYC2, zinc-finger DNA binding protein 1 and 2 in the *C. roseus* transcriptome, which were reported by Van Moerkercke et al. [@pone.0103583-VanMoerkercke1] to be involved in regulation of TIA biosynthesis. Digital expression profiling, a powerful and efficient approach for *in-silico* analysis of gene expression, was employed to determine the expression of genes involved in TIA biosynthesis. When compared with flower, genes involved in TIA biosynthesis were highly active in leaf and root tissues. Except few genes, majority of the TIA pathway genes were up-regulated in leaves and roots, which are the prime source of anticancer and antihypertensive alkaloids, respectively. Our study further confirms previous findings that some of the enzymes involved in late reactions of vindoline biosynthesis are not expressed in cell cultures or in tissues unable to produce vindoline [@pone.0103583-DeLuca1], [@pone.0103583-De1]. As reported earlier by Goklany et al. [@pone.0103583-Goklany1], TIA pathway genes are up-regulated in hairy root cultures elicited with MeJa [@pone.0103583-Kumar1], [@pone.0103583-Raina1], we also observed an increase in gene expression of TIA pathway genes in MeJa induced hairy root cultures. Elicitor, like MeJa are compounds, which induce plant stress response and thereby increasing gene expression and alkaloid biosynthesis. Looking at overall digital expression profiling of TIA pathway genes, we conclude that expression of genes are dependent on plant maturity and are highly expressed in MeJa-elicited hairy root cultures. Nearly equal number of genes were differentially expressed in leaf and roots. Enrichment of GO terms, performed on differentially expressed genes showed that photosynthesis related genes were up-regulated in leaves. GO terms, like jasmonic acid biosynthesis, response to jasmonic acid stimulus, isoprenoid and tetraprenoid metabolic process were also enriched in differentially expressed genes. This further adds on to the findings that MeJa induces expression of TIA pathway genes. In conclusion, we assembled and annotated *C. roseus* transcriptome. Many transcripts harboring microsatellite repeats were identified, which can be used for marker-assisted breeding in *C. roseus*. Differential gene expression and GO enrichment analyses revealed the enrichment of genes involved in secondary metabolite production in leaf tissues, which is the prime source of bisindole alkaloids. Further, expression profiling of TIA genes determined that vindoline exclusively accumulates in aerial tissue of *C. roseus* and exposure to MeJa increases its production. However, deeper understanding of regulatory network governing TIA biosynthesis could help in successful metabolic engineering of alkaloid biosynthesis. The transcriptome resource generated in this study can facilitate understanding of regulatory and metabolic pathways underlying the biosynthesis of alkaloids. Materials and Methods {#s4} ===================== RNA isolation, sequencing and quality filtering {#s4a} ----------------------------------------------- Leaf, root and flower tissues of *C. roseus* L. var. Prabal were harvested from the adult plants grown in field. The tissues were harvested from the plants grown under natural environmental conditions in the experimentation field (28°31′55.3′′N 77°09′54.9′′E) of the National Institute of Plant Genome Research, New Delhi. The field experiments conducted in this study did not involve endangered or protected species and no specific permission was required for these location/activities. The tissues were snap frozen in liquid nitrogen and stored in −80°C until further use. RNA was isolated from tissue samples using TRI reagent (Sigma Life Science, USA). Quantity and quality of RNA samples were measured using Nanodrop (Thermo Fisher Scientific) and Agilent Bioanalyzer (Agilent technologies, Singapore). Sequencing was performed using HiSeq 2000 platform generating paired-end reads of 100 bp length. Stringent quality check was performed on short read datasets by using NGS QC Toolkit v2.3 [@pone.0103583-Patel1] to remove the low quality reads and those having primer/adaptor contamination. Duplicate reads from the dataset were removed using CLC Genomic Workbench (v4.7.2, <http://maq.sourceforge.net/index.shtml>) to obtain NR dataset. *De novo* short read assembly and validation {#s4b} -------------------------------------------- *De novo* transcriptome assembly was performed by using three commonly used short read assemblers, Velvet (v1.2.01) [@pone.0103583-Zerbino1], Oases (v0.2.04) [@pone.0103583-Schulz1] and ABySS (v1.2.6) [@pone.0103583-Simpson1]. All the three assemblers used in this study were run at different k-mer lengths, ranging from 31--95. We employed a two-step approach, the best k-mer assembly obtained from different assemblers was merged with MPGR *C. roseus* transcriptome [@pone.0103583-GongoraCastillo1] and subjected to second round of assembly using TGICL (v2.0) [@pone.0103583-Pertea1] with minimum and maximum overlap length of 40 and 90, respectively. Various assembly parameters kept in consideration for marking the best assembly has been described previously by Garg et al. [@pone.0103583-Garg2]. GC content analysis was done using in-house perl script. To validate the quality of assembled transcriptome, we performed simple and reciprocal BLAST searches at an *E*-value cut-off of ≤1e-05 for identification of best significant match. The proteome sequences of tomato, potato, cucumber, grapevine and Arabidopsis were downloaded from phytozome v9.1 ([www.phytozome.net](http://www.phytozome.net)) and transcriptome sequences of alkaloid producing plants (*A. belladonna*, *C. acuminata*, *C. sativa*, *R. serpentine*, *R. officinalis* and *V. officinalis*) were downloaded from Medicinal plant genomics resource ([www.medicinalplantgenomics.msu.edu](http://www.medicinalplantgenomics.msu.edu)). Functional annotation {#s4c} --------------------- One of the most common approach for annotating transcriptome assembly is similarity search via BLAST. *C. roseus* transcripts were searched against TAIR10 proteome, Uniref90, Uniref100 and non-redundant protein (NCBI-nr) data sets at an *E*-value cut-off of ≤1e-05 for identification of best significant match. GOSlim terms for molecular function, biological process and cellular component were assigned to each *C. roseus* transcripts on the basis of their best match Arabidopsis protein. Similarity search against COG database classified *C. roseus* transcripts among different categories of COG classification system. To look for the genes involved in various pathways, assignment of KEGG Orthology (KO) terms and KEGG pathway construction was performed using KAAS (KEGG Automatic Annotation Server) [@pone.0103583-Moriya1] at default parameters. Read mapping and gene expression analysis {#s4d} ----------------------------------------- For gene expression analysis, high-quality short reads were mapped on to *C. roseus* transcriptome assembly using RNA-seq analysis utility of CLC Genomics Workbench. A maximum of two mismatches were permitted for alignments. Unique read counts for each tissue sample were normalized by calculating the read per kilo-base per million (RPKM) for each transcript. DESeq (v1.10.1) [@pone.0103583-Anders1], a software of R package, was used for differential gene expression analysis. It measures gene expression based on the negative binomial distribution with variance and mean linked by local regression. We calculated the size factor for each sample for normalization of read count data using DESeq. A p-value cut-off of ≤0.05 and at least two-fold change in gene expression was used to identify differentially expressed genes. RPKM values were log~2~ transformed and heat-map showing expression profiles for genes involved in TIA pathway were generated using MultiExperiment Viewer (MeV, v4.8). Hierarchical clustering was performed using Pearson correlation metrics and average linkage rule using MeV. Real-time PCR analysis {#s4e} ---------------------- For real-time PCR analysis, gene-specific primers ([Table S4](#pone.0103583.s011){ref-type="supplementary-material"}) were designed using Primer Express (v3.0) software (Applied Biosystems, USA). Actin was used as an internal control. At least three independent biological replicates with three technical replicates of each biological replicate for each tissue sample were used for analysis. Real-time PCR reactions were carried out essentially following the protocol described previously [@pone.0103583-Garg3]. The correlation between expression profiles of selected genes obtained from real-time RT-PCR and RNA-seq data analysis was determined in R program. Identification of SSR and transcription factors {#s4f} ----------------------------------------------- *C. roseus* transcriptome was screened for the presence of microsatellites (SSRs) using MISA [@pone.0103583-Thiel1]. The number of repeating units considered in this study was, six for di-nucleotides, and five for tri-, tetra-, penta- and hexa-nucleotides. We did not consider mono-nucleotide repeats in this study. Primers for all the identified SSRs were designed using BatchPrimer3 v1.0 (probes.pw.usda.gov/batchprimer3). TFs encoding *C. roseus* transcripts were identified based on the Hidden Markov Model (HMM) profile search of conserved domain present in each TF family as described earlier [@pone.0103583-Garg2]. GO enrichment analysis {#s4g} ---------------------- For GO enrichment analysis, similarity search (BLASTX) was carried out against Arabidopsis proteome and the best hit corresponding to each *C. roseus* transcripts was identified. GO enrichment of different sets of genes was performed using BiNGO tool [@pone.0103583-Maere1] as described previously [@pone.0103583-Singh1]. Supporting Information {#s5} ====================== ###### **Number of** ***C. roseus*** **transcripts showing significant similarity with proteome/transcriptome sequences of closely related/alkaloid producing plants.** (PDF) ###### Click here for additional data file. ###### **Reciprocal BLAST analysis of** ***C. roseus*** **transcripts showing number of orthologous genes in closely related plant species.** (PDF) ###### Click here for additional data file. ###### **Length distribution of transcripts in the** ***C. roseus*** **transcriptome.** (PDF) ###### Click here for additional data file. ###### **GC content distribution in the** ***C. roseus*** **and** ***A. thaliana*** **transcripts.** (PDF) ###### Click here for additional data file. ###### **Functional annotation of** ***C. roseus*** **transcripts.** (A) GOSlim term assignment to the *C. roseus* transcripts in different categories of biological process, molecular function and cellular component. (B) COG function classification of *C. roseus* transcripts. (PDF) ###### Click here for additional data file. ###### **Heat-map showing expression patterns of differentially up-regulated genes in different tissues of** ***C. roseus*** **analyzed in this study.** The scale at the bottom represents log~2~ fold change. (PDF) ###### Click here for additional data file. ###### **Graphical view showing biological process gene ontology term enrichment in up-regulated genes in (A) leaf and (B) root.** The GO enrichment was performed using BiNGO. Node size is proportional to the number of genes in each category and shades represent the scale denoting significance level (white- no significant difference). (PDF) ###### Click here for additional data file. ###### Quality control and duplicate read removal statistics of *C. roseus* libraries. (PDF) ###### Click here for additional data file. ###### *De novo* assembly statistics by different assemblers at different k-mer length using total high-quality reads (a) Velvet (b) Oases (c) ABySS. (PDF) ###### Click here for additional data file. ###### *De novo* assembly statistics by different assemblers at different k-mer length using NR high-quality reads (a) Velvet (b) Oases (c) ABySS. (PDF) ###### Click here for additional data file. ###### Primer sequences used for real-time PCR analysis. (PDF) ###### Click here for additional data file. Authors are thankful to Dr. Rohini Garg for transcription factor analysis. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: MJ AKS. Performed the experiments: MV RS. Analyzed the data: MV RG RS. Contributed to the writing of the manuscript: RG MV MJ.

Introduction ============ Manganese (Mn) is an essential element for living organisms. It is the twelfth most abundant element in earth´s crust and is present in rocks, water, soil and food, normally associated with other elements (Santamaria, 2008\[[@R54]\]; Farina et al., 2013\[[@R19]\]). Environmental and occupational exposure to Mn may occur by contact with fungicides, such as Maneb, Manconzeb, methylcyclopentadienyl manganese tricarbonyl (MTT)-an anti-knock agent in gasoline, Mn-ore mining, Mn alloy production and dry alkaline battery manufacture (Mergler and Baldwin, 1997\[[@R42]\]; Mergler, 1999\[[@R41]\]). Dietary ingestion is the main source of Mn for humans and Mn absorption takes place mostly in the gastrointestinal tract where it is homeostatically controlled in the intestinal wall (Au et al., 2008\[[@R4]\]). The brain is especially susceptible to metal intoxication during embryonic development, when it is known that Mn is able to cross the placenta and to be excreted in the maternal milk (Betharia and Maher, 2012\[[@R7]\]). Mn absorption is increased during the neonatal period, when biliary excretion is poorly developed, leading to elevated concentrations of Mn in the brain and other tissues (Aschner and Aschner, 2005\[[@R2]\]). In children, Mn exposure is associated with alterations in psychomotor and cognitive development; furthermore a positive correlation exists between Mn exposure and hyperactivity (Menezes-Filho et al., 2011\[[@R40]\]; Roels et al., 2012\[[@R52]\]; Torres-Agustín et al., 2013\[[@R61]\]). Exposure to high levels of Mn can lead to pathological conditions, including neurodegeneration (Mergler et al., 1994\[[@R43]\]). The mechanisms mediating Mn toxicity are complex and not completely understood. Some of them include: \(1\) Mn accumulation in astrocytes leading to disruption of their ability to promote neuronal differentiation and decreasing glutamate uptake by astrocytes (Erikson and Aschner, 2003\[[@R16]\]; Giordano et al., 2009\[[@R21]\]); \(2\) Mn induced loss of dopaminergic neurons (Stanwood et al., 2009\[[@R57]\]); \(3\) Inhibition of respiratory chain complexes and induction of reactive oxygen species (ROS) (Zhang et al., 2004\[[@R64]\]; Sriram et al., 2010\[[@R56]\]). The use of alternative models in toxicological studies has been growing over the years. The fruit fly *Drosophila melanogaster* has served as a unique and powerful model for studies on human genetics and diseases. Although humans and flies are only distantly related, almost 75 % of disease related genes in humans have functional orthologs in the fly (Deepa et al., 2009\[[@R13]\]; Pandey and Nichols, 2011\[[@R46]\]). Moreover, the fast and external developmental cycle of this organism enable the study of toxicological effects of compounds during the developmental period. All these advantages make flies an appropriate model for studies related with metal toxicity (Bonilla-Ramirez et al., 2011\[[@R10]\]; Paula et al., 2012\[[@R47]\]) and human neurodegeneration (Hirth, 2010). As the embryonic development period is particularly sensitive to Mn exposure, in this paper we aimed to investigate the behavior and biochemical alterations caused by Mn exposure during the embryonic development of *Drosophila melanogaster*, focusing on adaptations in the antioxidant systems and MAPK signaling pathways. The levels of Mn and major essential elements were also determined. Materials and Methods ===================== Reagents -------- Anti-phospho-p38^MAPK^, anti-phospho JNK, anti-phospho ERK, anti ERK and ß-actin antibodies were purchased from Cell Signaling Technology (Danvers, MA, United States). EDTA (CAS 60-00-4), glycine (CAS 56-40-6), tris(hydroxymethyl) aminomethane (CAS 77-86-1) and ammonium persulfate (CAS 7727-54-0) were purchased from Serva (Heidelberg, Germany). L-Glutathione reduced (CAS 70-18-8), 1-chloro-2,4-dinitrobenzene (CAS 97-00-7), sodium orthovanadate (CAS 13721-39-6), manganese (II) chloride tetrahydrate (CAS 13446-34-9), ß-mercaptoethanol (CAS 60-24-2), methanol (CAS 67-56-1), tween 20 (CAS 9005-64-5), potassium phosphate dibasic (CAS 7758-11-4), potassium phosphate monobasic (CAS 7778-77-0), potassium bicarbonate (CAS 298-14-6) and anti-rabbit imunoglobulin antibody, *N,N,N\',N\'*-Tetramethylethylenediamine (CAS 110-18-9), quercetin (CAS 117-39-5), protease inhibitor cocktail for use with mammalian cell and tissue extracts, 5,5´dithiobis(2-nitrobenzoic acid) (CAS 69-78-3), 2´,7´-dichlorofluorescein diacetate (CAS 2044-85-1), glycerol (CAS 56-81-5), resazurin sodium salt (CAS 62758-13-8), triton x-100 (CAS 9002-93-1), sodium chloride (CAS 7647-14-5), albumin from bovine serum (CAS 9048-46-8), HEPES (CAS 7365-45-9), ß-nicotinamide adenine dinucleotide 2´-phosphate reduced tetrasodium salt were obtained from Sigma Aldrich (St. Louis, MO, United States). Bis-acrylamide, hybond nitrocellulose, acrylamide (CAS 79-06-1), sodium dodecyl sulfate (CAS 151-21-3), boric acid (CAS 10043-35-3) were purchased from GE Healthcare Bio-Sciences AB (Uppsala, Sweden). All other reagents were commercial products of the highest purity grade available. Animals ------- *Drosophila melanogaster*(Harwich strain) was obtained from the National Species Stock Center, Bowling Green, OH, USA. The flies were maintained at 25 °C on 12 h light/dark cycle in glass bottles containing 10 mL of standard medium (mixture of 39 % coarse and 32 % medium corn flour, 10 % wheat germ, 14 % sugar, 2 % milk powder, 1 % salt, 1 % soybean flour, 1 % rye flour, a pinch of methyl paraben (99-76-3) and lyophilized yeast. All experiments were performed with the same strain, and both genders were used at random. Animal treatment ---------------- Adults flies were placed in 10 mL of standard medium supplemented with 3 mL of a fresh solution (0.1 mM, 0.5 mM or 1 mM) of manganese chloride (MnCl~2~). In the control group the standard medium were supplemented with 3 mL of ultrapure water. After ten days laying eggs the adult flies were removed. When eggs were newly ecloded, 1 to 3 day old flies were used for all analyses. The MnCl~2~ concentrations were chosen based on previous studies (Bonilla-Ramirez et al., 2011\[[@R10]\]). Locomotor assay --------------- Locomotor activity was determined using the negative geotaxis assay as described by Bland et al. (2009\[[@R9]\]) with minor modifications. Briefly, for each assay, individual flies (1-3 days old) were immobilized on ice and placed separately in a glass tube; this method of immobilization does not affect fly neurology (Deepa et al., 2009\[[@R13]\]). After 15 minutes the flies were gently tapped to the bottom of the tube and the time required to climb up 8 cm of the tube wall was recorded. Each fly was tested 4 times at 1 minute intervals. For each experiment, the climbing mean was calculated. Metal content ------------- Two hundred flies per group were washed three times in ultrapure water and then dried on a filter paper in the incubator at 37 °C for 90 minutes. Flies were digested in closed vessels according to the procedure described previously by Bizzi et al. (2010\[[@R8]\]). Flies (\~ 70 mg) were transferred to quartz vessels with 6 mL of nitric acid 3 mol L^-1^. After closing and capping the rotor, the vessels were pressurized with 7.5 bar of oxygen using the valve originally designed for pressure release after conventional acid sample digestion. Then, the rotor was placed inside a microwave oven (Multiwave 3000 Microwave Sample Preparation System, Anton Paar, Graz, Austria). The system was equipped with eight high-pressure quartz vessels (volume of 80 mL, maximum operational temperature and pressure of 280 °C and 80 bar, respectively). Pressure was monitored in each vessel during all the runs. Microwave heating program was as follows: \(1\) 1000 W, with a ramp of 5 min; \(2\) 1000 W for 10 min; and \(3\) 0 W for 20 min (cooling step). After digestion, the pressure in each vessel was carefully released. The resulting solutions were transferred to polypropylene vials and diluted to 25 mL with water. Determination of calcium (Ca), iron (Fe), potassium (K), magnesium (Mg), manganese (Mn), sodium (Na), phosphorus (P), sulfur (S), and zinc (Zn) was performed using an inductively coupled plasma optical emission spectrometer (Optima 4300 DV, PerkinElmer, Shelton, USA) with axial view configuration. A concentric nebulizer and cyclonic spray chamber were used. Argon 99.996 % (White Martins, São Paulo, Brazil) was used for plasma generation, nebulization and as auxiliary gas. The instrumental parameters were carried out in according with previous work (Pereira et al., 2013\[[@R48]\]). Two readings were averaged to give one value per biological replicate and expressed as a mean (±) standard deviation of the mean (SD). Metals levels were expressed relative to the weight of flies used for analysis (µg metal/g of dried weight tissue). Cellular viability ------------------ Cellular viability was measured by two different methods. Firstly, cellular viability was measured using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) reduction assay as described by Sudati et al. (2013\[[@R58]\]) with minor modifications. The analysis was performed on the whole body of female flies. The flies were incubated in MTT for 60 min (37 °C), after MTT was removed the sample was incubated in DMSO for 30 min (37 °C). The absorbance from formazan dissolution by addition of DMSO was monitored in an EnsPireR multimode plate reader (PerkinElmer, USA) at 540 nm. The second method used was the resazurin reduction assay. The method uses the indicator resazurin to measure the metabolic capacity of cells. Viable cells reduce resazurin into resorufin, a fluorescent compound (Franco et al., 2009\[[@R20]\]). Groups of 40 flies were mechanically homogenized in 1 mL 20 mM Tris buffer (pH 7.0) and centrifuged at 1,000 RPM for 10 min at 4 °C. The supernatant was incubated in Elisa plates with 20 mM Tris buffer (pH 7.0) and resazurin for two hours. The fluorescence was recorded using EnsPire^R^ multimode plate reader (Perkin Elmer, USA) at~ex~579nm and ~em~584nm. DCF-DA oxidation assay ---------------------- Groups of 20 flies were mechanically homogenized in 1 mL 20 mM Tris buffer (pH 7.0), and centrifuged at 1,000 RPM for 10 min, 4 °C. The supernatant was used to quantify 2'-7'-dichlorofluorescein diacetate (DCF-DA) oxidation as a general index of oxidative stress as described by Perez-Severiano et al. (2004\[[@R49]\]). The fluorescence emission of DCF resulting from DCF-DA oxidation was monitored at regular intervals (~ex~488nm and ~em~530nm) in an EnsPire^R^ multimode plate reader (PerkinElmer, USA). Determination of gene expression by real-time quantitative PCR (qPCR) --------------------------------------------------------------------- Real-time quantitative PCR (qRT-PCR) was performed according to the method described by Paula et al. (2012\[[@R47]\]). The primers utilized are shown in Table 1[(Tab. 1)](#T1){ref-type="fig"}. All samples were analyzed as technical and biological triplicates with a negative control. Threshold and baselines were manually determined using the StepOne Software v2.0 (Applied Biosystems, NY). SYBR fluorescence was analyzed by StepOne software version 2.0 (Applied Biosystems, NY), and the CT (cycle threshold) value for each sample was calculated and reported using the 2^-ΔΔCT^ method (Livak and Schmittgen, 2001\[[@R38]\]). The GPDH gene was used as an endogenous reference showing no alterations in response to the treatment. For each well, analyzed in quadruplicate, a ΔC~T~ value was obtained by subtracting the GPDH C~T~ value from the C~T~ value of the interest gene (sequences of tested genes are represented in Table 1). The ΔC~T~ mean value obtained from the control group of each gene was used to calculate the ΔΔCT of the respective gene (2^-ΔΔCT^). Enzyme assays ------------- For enzyme activity measurements, groups of 40 flies were mechanically homogenized in 1 mL 20 mM HEPES buffer (pH 7.0), and centrifuged at 14.000 RPM for 30 min at 4 °C (Franco et al., 2009\[[@R20]\]). The supernatant was used for determination of Glutathione S-Transferase (GST), Catalase (CAT), Superoxide Dismutase (SOD) and Thioredoxin Reductase (TrxR). The GST activity was assayed following the procedure of Jakoby and Habig (1981\[[@R29]\]) using 1-chloro-2,4-dinitrobenzene (CDNB) as substrate. The assay is based on the formation of the conjugated complex of CDNB and GSH at 340 nm. The reaction was conducted in a mix consisting of 100 mM phosphate buffer (pH 7.0), 1 mM EDTA, 1 mM GSH and 2.5 mM CDNB. CAT activity was assayed following the clearance of H~2~O~2~ at 240 nm in a reaction media containing 50 mM phosphate buffer (pH 7.0), 0.5 mM EDTA, 10 mM H~2~O~2~, 0.012 % TRITON x100 as described by Aebi (1984\[[@R1]\]). SOD activity assay was performed as described by Kostyuk and Potapovich (1989\[[@R35]\]). The assay consists in the inhibition of superoxide-driven oxidation of quercetin by SOD at 406 nm. The complete reaction system consisted of 25 mM phosphate buffer (pH 10), 0.25 mM EDTA, 0.8 mM TEMED and 0.05 µM quercetin. TrxR activity was assayed as described by Holmgren and Björnstedt (1995\[[@R26]\]). The test is based on the reduction of oxidized thioredoxin (Trx-S~2~) to reduced thioredoxin \[Trx-(SH)~2~\], using NADPH at 412 nm in a reaction media containing 0.1 M phosphate buffer (pH 7.0), 10 mM EDTA, 5 mM DTNB, 0.2 mg/mL BSA, 0.2 mM NADPH. All enzyme activities were performed at room temperature (25 ± 1 °C) using a Thermo Scientific Evolution 60s UV-Vis spectrophotometer. The enzyme activities were expressed in milliunits per milligram of total protein content, which was quantified following Bradford (1976\[[@R11]\]). Western blotting ---------------- Quantification of of the phosphorylation of mitogen-activated protein kinases (MAPKs) was performed by Western blotting as described by Posser et al. (2009\[[@R50]\]) with minor modifications. Groups of 40 flies were mechanically homogenized at 4 °C in 200 µL of buffer (pH 7.0) containing 50 mM Tris, 1 mM EDTA, 20 mM Na~3~VO~4~, 100 mM sodium fluoride and protease inhibitor cocktail. The homogenate were centrifuged at 4000 RPM for 10 min at 4 °C and the supernatants collected. After protein determination following Bradford (1976\[[@R11]\]), 4 % SDS solution, ß-mercaptoethanol and glycerol was added to samples to a final concentration of 100, 8 and 25 %, respectively and the samples frozen for further analysis. Proteins were separated using SDS-PAGE with 10 % gels, and then electrotransferred to nitrocellulose membranes (Paula et al., 2012\[[@R47]\]). Membranes were washed in tris-buffered saline with Tween (100 mM tris-HCl, 0.9 % NaCl and 0.1 % Tween-20, pH 7.5) and incubated overnight at 4 °C with specific primary antibodies (anti-phospho-p38^MAPK^, anti-phospho JNK, anti-phospho ERK, anti ERK and anti ß-actin). Following incubation, membranes were washed in tris-buffered saline with Tween and incubated for 1 h at 25 °C with anti rabbit-IgG secondary specific antibodies. Antibody binding was visualized using the ECL Western Blotting substrate Kit (Promega). Band staining density was quantified using the Scion Image software (Scion Image for Windows) and expressed as the percentage (%) of the control group (mean ± standard deviation of the mean). The values were normalized using total proteins (total ERK and ß-actin). Statistical analysis -------------------- Statistical analysis was performed using one-way ANOVA followed by Tukey´s post hoc test. Pearson's correlation test was applied for detection of significant statistical differences among the metals. Differences were considered statistically significant when p \< 0.05. GraphPad Prism 5 Software was used for artwork creation. Results ======= Exposure to Mn causes hyperactive behaviors and alters metal levels in Drosophila melanogaster ---------------------------------------------------------------------------------------------- The evaluation of climbing behavior performance by negative geotaxis showed that flies exposed to 0.5 mM and 1 mM of Mn reached the limit of columns significantly (p \< 0.005) faster than controls (Figure 1[(Fig. 1)](#F1){ref-type="fig"}). Levels of Mn and other essential metals were measured in *D. melanogaster* exposed to Mn. At concentrations of 0.5 and 1 mM, the levels of Mn in flies increased almost two and three fold respectively when compared with the control group (Table 2[(Tab. 2)](#T2){ref-type="fig"}), whereas Ca, Cu, Na and Zn levels decreased significantly in flies treated with Mn. Statistically, a significant negative relationship between Mn uptake and levels of Ca (r = -0,7966), Fe (r = -0,6635), Cu (r = -0,8028), S (r = - 0,6018) and Zn (r = -0,9802) occurred in response to Mn treatment (Table 3[(Tab. 3)](#T3){ref-type="fig"}). Mn exposure during embryonic development decreased cell viability and increased ROS production in flies ------------------------------------------------------------------------------------------------------- Cell viability was evaluated through two different tests, MTT and Resazurin. Both procedures showed a significant drop in cell viability at higher concentrations of Mn, confirming its toxicity at the cellular level (Figure 2A and 2B[(Fig. 2)](#F2){ref-type="fig"}). Many factors have been implicated in Mn-induced neurotoxicity, among them the oxidative stress caused by dopamine oxidation, or its ability in interfering with cellular respiration (Aschner and Aschner, 2005\[[@R2]\]). In this study, exposure to Mn leads to an increase in DCF-DA oxidation, a general index of oxidative stress from 0.5 mM (Figure 3[(Fig. 3)](#F3){ref-type="fig"}). Mn increased CAT and SOD mRNA expression, without altering their enzymatic activity ----------------------------------------------------------------------------------- Expression of mRNA for Cat, Sod and Hsp83 (an homolog of mammalian HSP90) (Bandura et al., 2013\[[@R5]\]) was analyzed by qRT-PCR using specific primers (Table 1[(Tab. 1)](#T1){ref-type="fig"}) in flies treated with 1 mM Mn. CAT and SOD mRNA expression was 250 % higher in treated flies whereas Hsp83 expression was 1000 % higher (Figure 4[(Fig. 4)](#F4){ref-type="fig"}). The levels of the antioxidant enzymes activity TrxR, GST, SOD and CAT were determined TrxR and GST activity were increased at concentrations of 0.5 mM and 1 mM (Figure 5[(Fig. 5)](#F5){ref-type="fig"}), while CAT and SOD activity showed no significant differences. Mn exposure inhibited p38MAPK phosphorylation --------------------------------------------- MAPKs phosphorylation levels were investigated in flies exposed to Mn. There was a 40 % inhibition of p38^MAPK^ phosphorylation in flies exposed to Mn at 1 mM, while the phosphorylation level of extracellular signal-regulated kinases (ERK) was unaltered. C-Jun-N-terminal Kinases 2 (JNK2) phosphorylation was not statistically different from controls (Figure 6[(Fig. 6)](#F6){ref-type="fig"}). Discussion ========== In this study, we determined the biochemical and behavioral alterations in *Drosophila melanogaster* in response to Mn exposure during embryonic development. Mn is an essential element required in key biological processes; however, high levels of Mn are associated with neurological and neuropsychiatric disorders (Mergler, 1999\[[@R41]\]). The risk of Mn overexposure comes from both occupational and environmental sources (Mergler and Baldwin, 1997\[[@R42]\]). Mn intoxication, a syndrome known as Manganism, is characterized by an extrapyramidal dysfunction and neuropsychiatric symptomatology and is associated with prolonged occupational exposure to high concentrations of this metal. Classical symptoms include irritability, intellectual deficits, compulsive behaviors, tremors and cock-like walk (Mergler, 1999\[[@R41]\]; Roth, 2006\[[@R53]\]). In rodents, Krishna et al. (2014\[[@R36]\]) showed that adult mice exposed to Mn through the drinking water presented neurobehavioral deficits and glial activation related with Mn deposition in brain. Moreover, others studies demonstrated that Mn toxicity in rats is accompanied by increased cholesterol biosynthesis and impairments in neuronal function of the hippocampus, which is involved in learning and memory (Öner and Sentürk, 1995\[[@R45]\]; Sentürk and Öner, 1996\[[@R55]\]). It has been shown that Mn supplementation during the neonatal period of rats resulted in increased Mn concentrations in tissues leading to adverse effects on motor development and behavior (Tran et al., 2002\[[@R62]\]). Mn uptake is increased during the neonatal period as biliary excretion, which has been suggested as a pathway for Mn elimination from the body, is poorly developed at this stage (Aschner and Aschner, 2005\[[@R2]\]). Exposure to Mn during the embryonic and early postnatal periods may result in increased levels of Mn in the brain and other tissues including bone, liver, pancreas and kidney (Aschner and Aschner, 2005\[[@R2]\]; Roels et al., 2012\[[@R52]\]). Higher levels of Mn retention *in utero* may affect children´s psychomotor development (Takser et al., 2003\[[@R59]\]). Possible adverse effects of Mn exposure on children´s health include cognitive deficits and hyperactive behaviors (Menezes-Filho et al., 2009\[[@R39]\]; Torres-Agustín et al., 2013\[[@R61]\]). Children exposed to high levels of Mn during the fetal period were more impulsive, inattentive, agressive, defiant, disobedient, destructive and hyperactive (Ericson et al., 2007\[[@R15]\]). It is recognized that factors such as the source and the duration of exposure, as well as nutritional status, can interfere in the intensity and incidence of neurological symptoms associated with Mn exposure in humans. Chronic consumption of drinking-water containing Mn at levels ranging from 81 to 2300 µg/l was associated with progressively higher prevalence of neurological symtoms (Kondakis et al., 1989\[[@R33]\]). The concentrations used in this study were from 0.1 mM of Mn in food (corresponding to 19 mg/L in the medium). Despite the use of relatively elevated concentrations, body levels of Mn were not altered at 0.1 mM. In previous studies, adult flies were acutely exposed to Mn (0.5-20 mM) diluted in sucrose, as the only source of food and liquid, which lead to significant locomotor deficits (Bonilla-Ramirez et al., 2011\[[@R10]\]). Our study is the first where Mn was provided as a cereal based diet over all the embryonic period. Thus, more studies are necessary to understand the rate of Mn uptake from diet in flies and how it may affect neurological behaviour. In our study, flies exposed to Mn at 0.5 mM and 1 mM showed increased locomotor speed in the locomotor behavior test (assayed as negative geotaxis behavior), pointing to a hyperactive-like behavior in *Drosophila melanogaster*. Furthermore, Mn levels were substantially increased in treated flies, while Ca, Cu, Zn, Fe and S levels were all decreased. This relationship may be in part associated with a competition of the metals for the same mechanism of transport into the flies cells. Facilitated diffusion, active transport, divalent metal transport 1 (DMT1), ZIP8 and transferrin (Tf)-dependent transport mechanisms are all involved in cellular Mn transportation (Aschner et al., 2007\[[@R3]\]). Among these metal transport systems, DMT1 has a very broad substrate specificity and is likely to be the major transmembrane protein responsible for the uptake of a variety of divalent cations, including Mn^2+^, Cd^2+^, Zn^2+^, Co^2+^, Ni^2+^, Cu^2+^ and Pb^2+^ (Gunshin et al., 1997\[[@R22]\]). In the flies, many proteins involved in the metabolism of biometals such as ferritin, transferrin, iron regulatory proteins, divalent metal transporter are expressed (Bonilla-Ramirez et al., 2011\[[@R10]\]). In this context, Mn uptake is less frequently studied in comparison with other metals and the mechanisms related to Mn transport are considerably more complex, occurring in most part in the divalent (II) and oxide forms (Tebo et al., 2004\[[@R60]\]). Mn has the capacity to interact and /or compete with Ca (Dittman and Buchwalter, 2010\[[@R14]\]). In a study performed in the aquatic insect *Hydropsyche sparna*, Mn exposure decreased cadmium (Cd) and Zn accumulation. Furthermore, increased Ca concentrations significantly reduced Mn accumulation in the insect (Poteat et al., 2012\[[@R51]\]). Dittmanand and Buchwalter (2010\[[@R14]\]) suggested that Mn is also absorbed by Ca transporters in aquatic insects, where increasing ambient Ca concentrations decrease Mn accumulation. There was also a negative correlation between Mn and Fe levels. Iron deficiency has been suggested as a possible contributing cause of attention deficit and hyperactivity disorder (ADHD) in children (Konofal et al., 2008\[[@R34]\]). Concomitantly, children with ADHD showed elevated serum Mn concentrations (Konofal et al., 2008\[[@R34]\]). Recent studies have suggested that Mn accumulates in dopaminergic neurons via the presynaptic dopamine transporter (DAT) and an altered functioning of the dopaminergic system has been well established in the etiology of ADHD (Farias et al., 2010\[[@R18]\]). Our results showed decreased cell viability using two different methodologies and increased ROS generation in flies exposed to Mn during development. Previous work by our group in PC12 cells, demonstrated that Mn leads to increased production of hydrogen peroxide (H~2~O~2~) (Posser et al., 2009\[[@R50]\]). H~2~O~2~ is a highly permeable and reactive molecule being able to react with metals such as Fe, thus generating hydroxyl radicals (Jiménez Del Río M and Vélez-Pardo, 2004\[[@R30]\]; Barbosa et al., 2010\[[@R6]\]) resulting in a propagation of oxidative damage in cells. Tissues can respond to oxidative stress by modulating antioxidant defenses (Halliwell and Gutteridge, 2007\[[@R23]\]). We measured the gene expression of Hsp83, CAT and SOD in *Drosophila melangaster*. Earlier studies showed that cellular stress may induce heat shock proteins in parallel with ROS production (Kim et al., 2004\[[@R32]\]; Paula et al., 2012\[[@R47]\]). Our results showed a significant increase in Hsp83 mRNA levels in Mn treated flies. Previously, our group demonstrated that exposure of flies to heavy metals such as mercury causes increases in the expression of Hsp83 (Paula et al., 2012\[[@R47]\]). CAT and SOD mRNA levels were significantly increased by Mn, but the enzymatic activity of these proteins was unchanged. The antioxidant ezyme SOD converts superoxide radicals (O~2~^º-^) to H~2~O~2~ and CAT catalyzes the conversion of H~2~O~2~ to oxygen (O~2~) and water (Barbosa et al., 2010\[[@R6]\]), thus neutralizing these reactive species. Considering that both SOD and CAT are crucial in the cell defense against oxidative stress (Halliwell and Gutteridge, 2007\[[@R23]\]), it might be expected that a posttranscriptional regulation mechanisms could maintain adequate levels of these proteins, however, further studies are necessary to elucidate this. Our results also showed that the activity of TrxR and GST was enhanced in Mn exposed flies. These two enzymes play an important role in protection against oxidative stress (Mustacich and Powis, 2000\[[@R44]\]). GST is a complex group of phase II detoxifying enzymes that participate in the metabolism of electrophilic substances, including carcinogenic, mutagenic and toxic compounds (Hayes et al., 2005\[[@R24]\]). TrxR is a dimeric FAD-containing enzyme that catalyzes the NADPH-dependent reduction of the active-site disulfide in Trx-S~2~ to give a dithiol in Trx-(SH)~2~ (Zhao et al., 2002\[[@R65]\]). Thioredoxin consists in one of the major redox-regulating proteins displaying a number of biological activities, including cytoprotection against ROS, protein repairing and protein disulfide reduction and modulation of signaling pathways (Yan et al., 2012\[[@R63]\]). Our data suggest that the increase in the levels of Mn and TrxR activity could represent a response to oxidation of thioredoxin in response to Mn-induced oxidative stress. MAPKs regulate the activity of a range of transcription factors thereby controlling gene expression and cellular function. The three most-studied MAPKs are ERK1/2, JNK1/2 and p38^MAPK^ (Ichijo, 1999\[[@R27]\]). ASK1 (Apoptosis Signaling Kinase 1) is a member of mitogen activated protein kinase kinase family (MAPKK) and an upstream activator of MAPK signaling pathway (Yan et al., 2012\[[@R63]\]). The redox state of thioredoxin regulates ASK1. Under normal conditions, ASK1 is bound to and inhibited by thioredoxin and when thioredoxin is oxidized, ASK1 can be activated and apoptotic signaling through the p38 ^MAPK^/JNK1/2 MAPKs initiated (Ichijo et al., 1997\[[@R28]\]). Studies conducted by Yan et al. (2012\[[@R63]\]) in a pancreatic carcinoma cell line, showed inhibition of TrxR by indolequinones resulting in a change of Thioredoxin-1 redox state to an oxidized form and activation of p38^MAPK^/JNK1/2. Similarly, Cd treatment activated ASK1 and its downstream MAPK in neuronal cells (Kim et al., 2005\[[@R31]\]), and inhibits components of thioredoxin system (Chrestensen et al., 2000\[[@R12]\]), while that Liedhegner et al. (2011\[[@R37]\]) demonstrated that knockdown of ASK1 as well as chemical inhibition of p38^MAPK^ and JNK played protective effects against L-DOPA induced apoptosis. We show that Mn induced TrxR activity while p38^MAPK^ /JNK1/2 phosphorylation were inhibited. This suggests an involvement of thioredoxin system in the mechanism of Mn induced toxicity. Augmented TrxR activity may represent a cellular response to high levels of ROS induced by Mn exposure, thus preventing the oxidation of thioredoxin and its dissociation of ASK1. This could contribute to diminished activation of p38^MAPK^ pathway upstream kinases resulting in lower levels of phosphorylation of this MAPK thus minimizing apoptotic cell death. In summary, our study demonstrate for a first time that developmental exposure to Mn leads to hyperactive-like behavior and accumulation of this metal in *Drosophila melanogaster*. The observed raise in Mn levels is negatively correlated with levels of other essential metals. This result fits with previous studies showing that Mn accumulation and Fe deficiency are associated with hyperactive behavior in children (Ericson et al., 2007\[[@R15]\]; Konofal et al., 2008\[[@R34]\]). The induction of stress responsive genes and antioxidant enzyme activity associated with inhibition of p38^MAPK^ phosphorylation at higher concentrations of Mn may represent an adaptive response to oxidative stress generated by this metal, in an attempt to avoid exacerbated cellular damage. Acknowledgements ================ The authors thank to Universidade Federal do Pampa, Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq 482313/2013-7), Fundação de Amparo à Pesquisa do Estado do Rio Grande do Sul (Fapergs 1954/2551/13-7), FAPERGS/PRONEM, FINEP and INCT-APA for financial support. The authors thank Professor Peter R. Dunkley for critical review of the manuscript. Conflict of interest ==================== The authors declare that they have no conflict of interest. {#T1} {#T2} {#T3} {#F1} {#F2} {#F3} {#F4} {#F5} {#F6}

1. Introduction {#s1} =============== Consider a game in which every decision maker is faced with a finite set of choices such that one specific choice always brings him higher monetary payoff than other choices, irrespective of the choices made by other players. In this situation, the individual choice boils down to going for either a higher or a lower monetary payoff. The straightforward response of a decision maker who cares about his monetary payoff is to disregard dominated actions---i.e., actions that may only deteriorate payoff relative to other actions. This dominance principle is the most basic solution concept of game theory (Camerer, [@B12]). It becomes very powerful when embedded in a strategic reasoning as a stepwise process. In each step, the dominance principle implies that dominated strategies should be eliminated from an agent\'s strategy space. In an important class of games---known as dominance-solvable games---this iterated elimination of dominated strategies leads to a unique solution. Strikingly, the data collected from numerous experiments on dominance-solvable games raise important questions about the empirical accuracy of predictions derived from this principle. Subjects tend to display less strategic sophistication than is needed to justify many applications of iterated dominance (and related refinements) to model human decision making in strategic environments (Crawford, [@B19]). The beauty contest game is one of the textbook examples of this issue[^1^](#fn0001){ref-type="fn"}. A given set of players is asked to choose a number in the range \[0, 100\]. To win the game, a player should choose a number that is the closest to *p* = 2/3 of the average of all chosen numbers. Any number above 2/3 × 100 ≈ 66.7 violates first-order dominance, because the average has to be lower than 100. Knowing this, players should all choose numbers no greater than 66.7, meaning that their average may not exceed 2/3 × 66.7 ≈ 44.5. This reasoning lowers the target as the number of iterations increases, eventually leading to the unique Nash equilibrium in which all players choose 0. In many experimental studies of this game, the numbers chosen by players are used as a proxy of the depth of iterated reasoning.[^2^](#fn0002){ref-type="fn"} A well replicated stylized fact is to observe 1/3 of subjects choosing a number higher than 67, and at least 1/3---a number between 44 and 67. This paper focuses on one of the earliest and simplest example of such an empirical inaccuracy of dominance solvability, adapted from a 2 × 2 game discussed in Rosenthal ([@B43]) and first brought to the laboratory by Beard and Beil ([@B5])[^3^](#fn0003){ref-type="fn"}. The normal-form representation of this game is given in Table [1](#T1){ref-type="table"}. With *L* \< *S* \< *H*, *m* \< *h*, and *s* \< *h*, the game is one-step dominance solvable: the elimination of player B\'s weakly dominated strategy *l* immediately leads to the Pareto-Nash equilibrium (*R, r*)[^4^](#fn0004){ref-type="fn"}. ###### **Generic form of the normal representation of Rosenthal ([@B43]) dominance solvable game**. **Player B** -------------- ----- -------------- -------- **Player A** *L* (S; s) (S; s) *R* (L; m) (H; h) In line with observed behavior in other dominance solvable games, numerous studies (summarized in Table [2](#T2){ref-type="table"}) find frequent failures to achieve the Pareto-Nash equilibrium. In spite of variations in the design (described in the table), deviations from the standard theoretical predictions are systematic and sizable. First, dominance is frequently violated by player Bs. Depending on the exact experimental setup, up to 27% column players choose a strictly dominated action. Second, player As violate iterated dominance, even in those cases in which player Bs commonly obey dominance. As an example, while only 6% of player Bs violate dominance in Jacquemet and Zylbersztejn ([@B35])-ET2 and BT2, 26% of row players still contradict the predictions of dominance solvability by choosing *L* (and this figure may even attain 86% in other instances, see Beard, Beil -- Tr. 5 in Table [2](#T2){ref-type="table"}). As shown in the three middle columns of the table, both the absolute and the relative size of the stakes vary a great deal from one study to the other. Several lessons emerge from this accumulated evidence. First, both players react to their own monetary incentives. Second, in some cases player As also adjust their behavior to player Bs\' incentives. Finally, as shown by Jacquemet and Zylbersztejn ([@B35]), players\' inefficient behavior does not fade away with repetition and cannot be explained by inequality aversion (as framed by Fehr and Schmidt, [@B23]). ###### **Overview of existing experimental evidence**. **Experiment** **Form** **Payoff** **Outcomes (%)** ------------------------- ---------- --------------- ------------------ ------------ ---- ----- ----- ----- ----- Beard, Beil--Tr.1 Seq (9.75; 3.0) (10; 5.0) (3; 4.75) 66 29 6 83 --- Beard, Beil--Tr.2 Seq (9.00; 3.0) (10; 5.0) (3; 4.75) 65 35 0 100 --- Beard, Beil--Tr.3 Seq (7.00; 3.0) (10; 5.0) (3; 4.75) 20 80 0 100 --- Beard, Beil--Tr.4 Seq (9.75; 3.0) (10; 5.0) (3; 3.00) 47 53 0 100 --- Beard, Beil--Tr.5 Seq (9.75; 6.0) (10; 5.0) (3; 3.00) 86 14 0 100 --- Beard, Beil--Tr.7 Seq (58.50; 18.0) (18.0; 28.50) (60; 30.0) 67 33 0 100 --- Beard et al.--Tr.1 Seq (1450; 450) (1500; 750) (450; 700) 79 18 3 83 --- Beard et al.--Tr.2 Seq (1050; 450) (1500; 750) (450; 700) 50 32 18 64 --- Goeree, Holt--Tr.1 Ext (80; 50) (90; 70) (20; 10) 16 84 0 100 --- Goeree, Holt--Tr.2 Ext (80; 50) (90; 70) (20; 68) 52 36 12 75 --- Goeree, Holt--Tr.3 Ext (400; 250) (450; 350) (100; 348) 80 16 4 80 --- Cooper, Van Huyck--Tr.9 Str (4; 1) (6; 5) (2; 4) 27 --- --- --- 86 Cooper, Van Huyck--Tr.9 Ext (4; 1) (6; 5) (2; 4) 21 --- --- --- 84 JZ, 2014--BT1 Str (9.75; 3.0) (3.0; 4.75) (10; 5.0) 51 41 8 84 81 JZ, 2014--ET1 Str (9.75; 5.0) (5.0; 9.75) (10; 10.0) 54 33 13 72 73 JZ, 2014--ET3 Str (9.75; 5.5) (5.5; 8.50) (10; 10.0) 39 48 13 79 76 JZ, 2014--ET4 Str (8.50; 5.5) (5.5; 8.50) (10; 10.0) 25 61 14 82 82 JZ, 2014--ET2 Str (8.50; 8.5) (6.5; 8.50) (10; 10.0) 26 70 4 94 94 JZ, 2014--BT2 Str (8.50; 7.0) (6.5; 7.00) (10; 8.5) 26 70 4 94 94 *For each implementation in row, the first column describes the actual design of the experiment: simultaneous-move strategic-form game (Str), simultaneous-move extensive-form game (Ext), sequential-move game (Seq). The monetary payoffsof each outcome, displayed in columns 2--4, are in USD in Beard and Beil ([@B5]) and Cooper and Van Huyck ([@B14]), in cents of USD in Goeree and Holt ([@B27]), in Yens in Beard et al. ([@B4]), and in Euros in Jacquemet and Zylbersztejn ([@B35]). The game is repeated ten times in changing pairs in Jacquemet and Zylbersztejn ([@B35]), and one-shot in all other instances*. The aim of the present paper is to explore whether this empirical puzzle is related to players\' cognitive skills. In this sense, our investigation belongs to a recent and growing body of experimental studies in both psychology and economics which investigate the relationship between strategic behavior and cognitive skills[^5^](#fn0005){ref-type="fn"}. The main conclusion that can be drawn from these studies is that high cognitive skills predict strategic sophistication and efficient decision making. First, people with high cognitive skills make more accurate predictions about other people\'s intentions. Recent evidence from psychological research reveals the relationship between cognitive skills and the theory of mind. Using the "Reading the Mind in the Eyes" test (RMET, Baron-Cohen et al., [@B3]) to measure one\'s theory of mind, Ibanez et al. ([@B32]) find that people with higher cognitive skills are better at infering the internal emotional states of others[^6^](#fn0006){ref-type="fn"}. Relatedly, the results of a neuroeconomic experiment on the *p*-beauty contest game by Coricelli and Nagel ([@B17]) suggest that strategic thinking about other players\' thoughts and behavior is implemented by medial prefrontal cortex (mPFC) -- one of the brain areas commonly associated with theory of mind[^7^](#fn0007){ref-type="fn"}. An economic experiment by Carpenter et al. ([@B13]) also shows that people with higher cognitive ability make more accurate predictions of others\' choices in a 20-player beauty contest game. Second, people with higher cognitive skills apply more sophisticated reasoning and are more apt in strategic adaptation. Burks et al. ([@B10]) report that subjects with higher cognitive skills more accurately predict others\' behavior in a sequential prisoners\' dilemma game, and adapt their own behavior more strongly. In the context of the *p*-beauty contest game, subjects with higher cognitive skills are not only found to carry out more steps of reasoning on the equilibrium path (Burnham et al., [@B11]; Brañas-Garza et al., [@B8]), but also to adapt their behavior to their opponents\' cognitive skills (Gill and Prowse, [@B26]) as well as to their beliefs about their opponents\' cognitive skills (Fehr and Huck, [@B22]). Third, cognitive skills may be associated with the economic efficiency of outcomes of both individual and group activities. Corgnet et al. ([@B16]) find that higher cognitive skills predict better performance and less shirking in an experimental labor task (summing up tables of 36 numbers without using a pen). Jones ([@B36]), Al-Ubaydli et al. ([@B2]), and Proto et al. ([@B40]) report that groups with higher cognitive skills attain higher cooperation rates in repeated prisoner\'s dilemma games. On the other hand, Al-Ubaydli et al. ([@B1]) do not find a relationship between group members\' average cognitive skills and the efficiency of outcomes in a stag hunt coordination[^8^](#fn0008){ref-type="fn"}. Our contribution is two-fold. First, we provide new evidence on the relationship between strategic behavior and cognitive skills. We show that systematic mismatches between theoretical predictions and actual behavior in a classic 2 × 2 dominance-solvable game have cognitive underpinnings. Subjects with higher cognitive skills are found to be more likely to play dominant strategy and to best respond to other\'s strategy. Furthermore, cognitive skills predict strategic sophistication: only those players with sufficiently high cognitive ability are found to display sensitivity to the presence of uncertainty about others\' behavior. Our second contribution lies in experimental methodology. We extend the recent body of laboratory experiments comparing the performance of different measures of cognitive skills in predicting economic behavior. Notwithstanding the previous results (see e.g., Brañas-Garza et al., [@B8]; Corgnet et al., [@B15]), we report that the Raven\'s test score is a more general predictor of strategic behavior than the Cognitive Reflection Test score. 2. Experimental design {#s2} ====================== Our experiment is based on a 2 × 2 factorial design that varies the payoff matrix and the nature of player B. Each of the four resulting experimental treatments is implemented through a between-subject procedure---each subject participates in only one experimental condition. This data come from a large dataset, part of which has been previously used by Hanaki et al. ([@B30]). The main focus of that study is player As\' behavior under strategic uncertainty and its relation to monetary incentives and fluid intelligence. Certain elements of their design (such as the use of Human and Robot conditions and interest in players\' cognitive skills) inevitably needed to be adopted in the present study in order to address a much more general question of the empirical validity of the solution concept of dominance solvability. More precisely, we are interested in both players\' behavior (so as to measure the use of dominance by player Bs and the use of iterated dominance by player As under different information structures). We also make a methodological contribution, since in this paper we associate players\' behavior with multiple facets of cognitive skills: fluid intelligence (measured by Raven\'s test) and cognitive reflection (measured by CRT). Our first treatment variable is the size of the stakes, as represented by Game 1 and Game 2 in Table [3](#T3){ref-type="table"}. Although they have the same strategic properties, these two game matrices differ in terms of the saliency of monetary incentives to use (iterated) dominance. In Game 2, player As may earn a surplus of only 0.25 when moving from *L* to (*R, r*) (with payoff going from 9.75 to 10), while ending up in (*R, l*) is relatively costly (yielding only 3). In Game 1, the potential gains and losses from action *R* relative to *L* are more balanced: the gain from moving from *L* to (*R, r*) increases to 1.5 (with payoff moving from 8.5 to 10), while the outcome (*R, l*) becomes less costly (now yielding 6.5). The incentives of player Bs, in turn, go in the opposite direction: the gain from using the dominant strategy *r* (and conditional on player As\' choice *R*) is lower in Game 1 \[with payoff increasing from 4.75 to 5 between (*R, l*) and (*R, r*)\] than in Game 2 (where payoff increases from 8.5 to 10). In line with Jacquemet and Zylbersztejn ([@B35]) and Hanaki et al. ([@B30]) (who report that both players only react to their own monetary incentives) and as discussed in Section 3.1, each of these games generates sizable yet diverse empirical violations of dominance solvability. These two games together thus provide a wide range of monetary incentives to use dominance solvability within a common strategic environment[^9^](#fn0009){ref-type="fn"}. ###### **The experimental games**. **GAME 1** ------------ ----- --------------- ----------------- **A** *L* (8.50 ; 3.00) (8.50 ; 3.00) *R* (6.50 ; 4.75) (10.00 ; 5.00) **GAME 2** **B** ****l**** ****r**** **A** *L* (9.75 ; 8.50) (9.75 ; 8.50) *R* (3.00 ; 8.50) (10.00 ; 10.00) Our second treatment variable is related to the nature of player B (the column player) who may be represented either by a human subject (Human condition) or a pre-programmed computer (Robot condition). The Human condition enables us to capture two cardinal breaches of dominance solvability: the failure to use the dominant strategy (player Bs\' behavior) and the failure to best respond to others\' dominant actions (player As\' behavior). However, the latter behavior occurs under strategic uncertainty and thus might stem from two distinct sources: bounded rationality and rational behavior under uncertainty. More precisely, player As may simply have a limited capability of best responding to dominant strategy, but may also intentionally refrain from best responding when in doubt about player Bs\' use of dominant strategy. To separate these two effects, we introduce the Robot condition in which a human subject acting as player A interacts with a computerized player B who is pre-programmed to always choose *r*. We clearly inform the subjects in the Robot condition that they are interacting with a pre-programmed computer: "**the computer chooses** *r* **at each round, without exception**" (bold in the original instruction sheet). This is the only difference in the rules and procedures between Human and Robot conditions[^10^](#fn0010){ref-type="fn"}. Thus, the key property of the Robot condition as compared to the Human condition is neutralizing strategic uncertainty player As face, while maintaining space for boundedly rational behavior. The design of the experiment is otherwise the same in all four experimental conditions. We explore whether behavior is sensitive to learning by considering ten uniform, one-shot interactions. In order to homogenize incentives across rounds, the following rules are implemented: all games are played in strict anonymity, roles are fixed, and subjects\' payoffs are computed based one randomly drawn round. In the Human condition, players are matched into pairs using a perfect stranger, round-robin scheme, which guarantees that subjects are involved in a series of one-shot interactions despite the repetition of the game[^11^](#fn0011){ref-type="fn"}. Our control variables also include two measures of cognitive skills. Both of them are introduced as part of a post-experimental supplementary task. Subjects\' participation is rewarded with extra five Euros; otherwise, their answers are not incentivized[^12^](#fn0012){ref-type="fn"}. The supplementary task starts with a debriefing question, where subjects are asked to "report any information they find relevant about how their decisions has been made." Then, we implement the following measures of cognitive skills. The first task is the standard Cognitive Reflection Test based on Frederick ([@B25]) which "*measures cognitive reflectiveness or impulsiveness, respondents\' automatic response versus more elaborate and deliberative thought*" (Brañas-Garza et al., [@B8], p. 255). It contains three questions: 1. A notebook and a pencil cost 1.10 Euros in total. The notebook costs 1 Euro more than the pencil. How much does the pencil cost? 2. If it takes 5 machines 5 min to make 5 widgets, how long would it take 100 machines to make 100 widgets? 3. In a lake, there is a patch of lily pads. Every day, the patch doubles in size. If it takes 48 days for the patch to cover the entire lake, how long would it take for the patch to cover half of the lake? Subjects are informed that this set of three questions should be answered within 30 s (although we allow them to provide answers even after this time has elapsed). In this way, subjects can be classified according to their overall score (that is, the total number of correct answers) which can range from 0 to 3. The second task is Raven\'s progressive matrix test (often called Raven\'s test), a picture based, non-verbal measure of fluid intelligence, that is "*the capacity to think logically, analyze and solve novel problems, independent of background knowledge*" (Mullainathan and Shafir, [@B38], p. 48). It is widely used by, e.g., psychologists, educators, and the military (Raven, [@B41]). It consists of a series of tasks to be solved within a fixed amount of time. In each task, a subject should pick a single element (among eight options) that best fits a set of eight pictures. The level of difficulty increases from one question to the other[^13^](#fn0013){ref-type="fn"}. In our experiment, each participant is given a series of 16 tasks to be solved within 10 min. Individual scores in Raven\' test are computed as the number of correct answers to the 16 items of the test. 2.1. Experimental procedures ---------------------------- For each game matrix, we run three Human sessions (involving 20 subjects per session: 10 player As interacting with 10 player Bs), and two Robot sessions (involving 20 player As per session interacting with automated player Bs). Subjects are given a fixed fee equal to five euros to compensate participation to the experiment. Upon arrival to the laboratory, participants are randomly assigned to their computers and asked to fill in a short administrative questionnaire containing basic questions about their age, gender, education, etc. Experimental instructions are then read aloud: subjects are informed that they will play multiple rounds of the same game, each round with a different partner, and that their own role will remain unchanged throughout the experiment. Finally, subjects are asked to answer a short comprehension quiz. Once the quiz and any questions from participants are answered, the experiment begins. After each of the ten rounds of the game, subjects are only informed of their own payoffs. Information about past choices and payoffs is updated after each round and displayed at the bottom of the screen. Take-home earnings correspond to the outcome of a single round that is randomly drawn at the end of each experimental session. In addition, the experimental game is followed by supplementary tasks. An additional five euros fee is paid to each subject for completing this part. Immediately after the end of the experimental game, participants are provided with a brief round-by-round summary of their decisions and outcomes, and are asked to provide in a blank space on their computer screens any relevant comments in particular about what might have affected their decisions during the experiment. Subjects are also asked to solve the CRT test and a reduced-form Raven\'s test described above. All the sessions were conducted in February and March 2014. Out of the 200 participants (94 males), 155 were students with various fields of specialization. The majority of subjects (65%) had already taken part in economic experiments. Participants\' average age was 25.6 (st. dev. is 7.5). All sessions took place at the *Laboratoire d\'Economie Experimentale de Paris* (LEEP) at Paris School of Economics. Subjects were recruited via an on-line registration system based on O[rsee]{.smallcaps} (Greiner, [@B28]) and the experiment was computerized through software developed under R[egate]{.smallcaps} (Zeiliger, [@B47]) and z-Tree (Fischbacher, [@B24]). Sessions lasted about 45--60 min, with an average payoff of roughly 18.83 euros (including a five euros show-up fee and five euros for completing the post-experimental tasks). 3. Results {#s3} ========== Our main experimental results can be summarized as follows. First, in line with the existing literature, we observe systematic and sizable deviations from standard predictions based on the principle of dominance solvability. This phenomenon persists across game matrices and despite repetition. Second, we associate strategic behavior with cognitive skills. We find that Raven\'s test score is a more reliable predictor of strategic behavior than CRT score: whenever the latter predicts behavior, the former does too, but not *vice versa*. Subjects with higher Raven\'s test scores are more likely to use the dominant strategy and to best respond to other player\'s dominant strategy. Unlike those with low Raven\'s test score, they also react to the presence of strategic uncertainty. 3.1. Aggregate behavior in experimental games --------------------------------------------- Table [4](#T4){ref-type="table"} outlines the main patterns of behavior in our experimental games. The statistical significance of the changes observed in this table is tested by Models 1--3 in Table [5](#T5){ref-type="table"}. We first focus on the aggregate frequency of Pareto-Nash equilibrium (*R, r*) -- the sole outcome that survives the iterated elimination of (weakly) dominated strategies---found in the Human condition. In both games, we observe substantial deviations from the predictions of this solution concept: overall, players attain the (*R, r*) outcome 58% of times in Game 1 and 43% in Game 2 (Model 1, *H*~0~ : β~1~ = 0, *p* = 0.318). We also observe that efficiency increases over time: in both games, we observe the lowest frequency of (*R, r*) in the initial round (0.333 in Game 1 and 0.200 in Game 2), whereas the highest frequency of (*R, r*) occurs in the final round (0.700 in Game 1 and 0.533 in Game 2). ###### **Aggregate results**. **Round** **Overall** --------------------------------------------------------------- ----------- ------------- ------- ------- ------- ------- ------- ------- ------- ------- ------- ***Pr*** **(*****R,r*****) in the Human condition** Game 1 0.333 0.600 0.667 0.700 0.567 0.600 0.433 0.633 0.567 0.700 0.580 Game 2 0.200 0.333 0.400 0.400 0.433 0.500 0.500 0.433 0.467 0.533 0.420 ***Pr*** **(*****r*****) by player B in the Human condition** Game 1 0.767 0.800 0.867 0.900 0.800 0.800 0.700 0.833 0.867 0.800 0.813 Game 2 0.833 0.933 0.900 0.933 1.000 0.933 0.933 0.900 0.900 0.933 0.920 ***Pr*** **(*****R*****) by player A in the Human condition** Game 1 0.500 0.733 0.700 0.767 0.767 0.800 0.700 0.767 0.700 0.867 0.730 Game 2 0.300 0.333 0.400 0.400 0.433 0.533 0.533 0.500 0.500 0.533 0.447 ***Pr*** **(*****R*****) by player A in the Robot condition** Game 1 0.700 0.750 0.750 0.725 0.800 0.800 0.800 0.825 0.800 0.775 0.773 Game 2 0.500 0.575 0.725 0.575 0.800 0.700 0.700 0.775 0.775 0.775 0.690 *Columns 1--10 summarize the frequencies of outcomes (defined in rows) as % of all outcomes observed in each round of a given experimental treatment. The last column provides overall results*. ###### **Aggregate results: statistical support**. **Model 1** **Model 2** **Model 3** ------------------------------------- -------------------------------------------- -------------------------------------------- -------------------------------------------- Constant (β~0~) 0.580[^\*\*\*^](#TN1){ref-type="table-fn"} 0.813[^\*\*\*^](#TN1){ref-type="table-fn"} 0.730[^\*\*\*^](#TN1){ref-type="table-fn"} (0.144) (0.032) (0.084) 1\[*Game* 2\] (β~1~) --0.160 0.107[^\*^](#TN1){ref-type="table-fn"} --0.283[^\*\*^](#TN1){ref-type="table-fn"} (0.103) (0.044) (0.140) 1\[*Robot*\] (β~2~) 0.043 (0.103) 1\[*Robot*\] × 1\[*Game* 2\] (β~3~) 0.201 (0.161) *N* 600 600 1400 *R*^2^ 0.026 0.025 0.066 Estimates of linear probability models on outcome (R, r) (Model 1), decision r by player B (Model 2) and decision R by player A (Model 3). Standard errors (in parentheses) are clustered at the session level in Human treatments (three clusters per game matrix, six in total) and individual level in the Robot condition (40 clusters per game matrix, 80 in total) and computed using the delete-one jackknife procedure. All models contain a dummy variable set to 1 for game matrix 2 (and 0 for game matrix 1). In Model 3, we also introduce an additional dummy variable set to 1 for Robot condition (and 0 for Human condition) and well as the interaction between these two variables. *indicate significance at the 10/5/1% level*. To further explore the roots of these deviations, we turn to the aggregate patterns of both players\' behavior in Human and Robot conditions. We focus on three behavioral dimensions of dominance solvability: the use of dominant strategy (captured by player Bs\' behavior in the Human condition) and the ability to best respond to other player\'s dominant action with and without bearing the uncertainty about the latter (which is captured by player As\' behavior in the Human and Robot conditions, respectively). Inefficiency is caused by both players, although their roles differ from one game to another: the scope of inefficient behavior is similar for both players in Game 1, and highly asymmetric in Game 2. Overall, player As select action *R* with probability 0.730 in Game 1 and 0.447 in Game 2 (Model 3, *H*~0~ : β~1~ = 0, *p* = 0.047). However, player As\' behavior happens to be misaligned with player Bs\' actual decisions which follow the opposite trend: the total frequency of action *r* increases from 0.813 in Game 1 to 0.920 in Game 2 (Model 2, *H*~0~ : β~1~ = 0, *p* = 0.060). Importantly, the data from Robot sessions suggest that the uncertainty about player Bs\' behavior is not the only driver of player As\' choices. Player As frequently and systematically fail to best respond to player Bs\' dominant action even when the latter comes with certainty in the Robot condition, although their willingness to select action *R* increases in both games as compared to the Human condition (to 0.773 in Game 1 and 0.690 in Game 2)[^14^](#fn0014){ref-type="fn"}. The fact that inefficient actions from player As prevail in the absence of strategic uncertainty may suggest that at least some of them are boundedly rational decision makers. In the next section, we analyze how these three behavioral components of dominance solvability vary as a function of players\' cognitive skills. 3.2. Cognitive skills and strategic behavior -------------------------------------------- The average score in Raven\'s test (CRT) is 8.679 out of 16 with SD 3.117 (0.479 out of 3 with SD 0.852). Our experimental sample is properly randomized across treatments regarding both measures. We do not reject the null hypothesis that Raven\'s test scores have the same distributions in all treatments (*p* = 0.275, Kruskal-Wallis test). A Kruskal-Wallis test applied to the CRT scores leads to the same conclusion (*p* = 0.502). We also replicate several results from previous studies combining Raven\'s test and CRT regarding the relationship between both scores as well as gender differences (Brañas-Garza et al., [@B8]; Corgnet et al., [@B15]). There is a moderate, yet highly significant correlation between Raven and CRT scores (Spearman\'s ρ = 0.306, *p* \< 0.001) which suggests that they may have a common source, but do not capture the same cognitive skills. Furthermore, the average score of males is significantly higher than the average score of females (Raven\'s test: 9.382 with SD 0.341 vs. 8.014 with SD 0.384, *p* = 0.009; CRT: 0.676 with SD 0.111 vs. 0.291 with SD 0.087, *p* = 0.007; two-sided *t*-tests)[^15^](#fn0015){ref-type="fn"}. We also observe that many subjects (70%) of our 200 participants fail to provide at least one correct answer in our standard CRT. 16% provide exactly one, 8% -- two, and 6% -- three correct answers. This stands in line with Brañas-Garza et al. ([@B8]) who report the respective frequencies of 67, 23, 9, and 1% for a similar sample size (*N* = 191), and echoes the scores in the least performant sample reported in a seminal study by Frederick ([@B25]): out of 138 students of the University of Toledo, 64% provide no correct answer, 21% provide one, 10% provide two, and 5% provide three corrects answers. ### 3.2.1. Cognitive predictors of strategic behavior: aggregate results In this part, we study the cognitive correlates of strategic behavior. Figures [1](#F1){ref-type="fig"}, [2](#F2){ref-type="fig"} present the aggregate evolution of behavior as a function of cognitive skills, measured either by CRT score or by Raven\'s test score across roles (player A or player B) and experimental conditions (Human or Robot). {#F1} {#F2} In Figure [1](#F1){ref-type="fig"}, the sample is divided into two subsamples: subjects who provided at least one correct answer to CRT (referred to as *CRT* \> 0) and those who did not (referred to as *CRT* = 0). The aggregate patterns of behavior weakly differ between the two subsamples. Bootstrap proportion tests fail to reject the null hypothesis that the overall proportions of decision *R* are the same for both CRT categories in the Human condition (*p* = 0.126) and in the Robot condition (*p* = 0.235)[^16^](#fn0016){ref-type="fn"}. The aggregate proportions of decision *r*, in turn, are found to be statistically different (*p* = 0.037), subjects with a CRT score zero being less likely to play *r* than subjects who gave at least one correct answer. In Figure [2](#F2){ref-type="fig"}, we split our sample into three subsamples based on Raven\'s test score (1st tertile: less than 8 correct answers, 2nd tertile: between 8 and 10 correct answers, 3rd tertile: more than 10 correct answers). Although, bootstrap proportion tests suggest that player As\' behavior in the Human condition does not vary significantly between these three subsamples (1st tertile vs. 2nd tertile: *p* = 0.255, 2nd vs. 3rd: *p* = 0.580, 1st vs. 3rd: *p* = 0.565), significant differences arise for both player As in the Robot condition (*p* = 0.001, *p* = 0.735, *p* \< 0.001, respectively) and for player Bs (*p* = 0.064, *p* = 0.057, *p* \< 0.001, respectively). Raven\'s test score seems to have a more systematic association with players\' behavior than CRT score, although both measures fail to predict behavior under strategic uncertainty. ### 3.2.2. Cognitive skills and dominance solvability: regression analysis In what follows, we provide further econometric insights into these preliminary results. Following Brañas-Garza et al. ([@B8]); Corgnet et al. ([@B15]), we use three individual characteristics discussed in the previous section -- gender, Raven\'s test score and CRT score (kept as a dummy variable with value 1 if the subject gave at least one correct answer at the CRT test and 0 otherwise) -- to explain behavior in our experimental games[^17^](#fn0017){ref-type="fn"}. The econometric specification is based on the linear probability model and the estimation procedure is outlined in Jacquemet and Zylbersztejn ([@B35]). We also control for payoff scheme and repetition effects by including game matrix and round dummies. We consider three different outcome variables: player As\' behavior in the Human and the Robot treatment, and player Bs\' behavior in the Human treatment. Given the correlation between CRT and Raven\'s test scores, including both variables in the model might result in multicollinearity and lead to the under-rejection of the nullity of respective coefficients. For each outcome, we first include these two measures separately in Models 1 and 2, while Model 3 includes both variables. This evidence is summarized in Table [6](#T6){ref-type="table"}. ###### **Cognitive predictors of strategic behavior: regression analysis**. ******Pr****** **(********R********) by player A** **Pr (********r********) by player B** --------------- ---------------------------------------------------- ------------------------------------------ ------------------------------------------ -------------------------------------------- --------------------------------------------- --------------------------------------------- -------------------------------------------- -------------------------------------------- -------------------------------------------- Const. 0.423[^\*\*\*^](#TN2){ref-type="table-fn"} 0.552[^\*\*^](#TN2){ref-type="table-fn"} 0.563[^\*\*^](#TN2){ref-type="table-fn"} 0.573[^\*\*\*^](#TN2){ref-type="table-fn"} 0.242[^\*^](#TN2){ref-type="table-fn"} 0.240[^\*^](#TN2){ref-type="table-fn"} 0.705[^\*\*\*^](#TN2){ref-type="table-fn"} 0.430[^\*\*\*^](#TN2){ref-type="table-fn"} 0.444[^\*\*\*^](#TN2){ref-type="table-fn"} (0.080) (0.176) (0.197) (0.088) (0.135) (0.135) (0.027) (0.103) (0.099) 1\[CRT\>0\] 0.131 0.152 0.062 (0.024) 0.109[^\*^](#TN2){ref-type="table-fn"} 0.046 (0.095) (0.121) (0.102) (0.102) (0.047) (0.036) Raven 0.013 0.018 0.0426[^\*\*\*^](#TN2){ref-type="table-fn"} 0.0434[^\*\*\*^](#TN2){ref-type="table-fn"} 0.0313[^\*\*^](#TN2){ref-type="table-fn"} 0.0287[^\*\*^](#TN2){ref-type="table-fn"} (0.025) (0.027) (0.013) (0.012) (0.009) (0.008) 1\[*Game* 2\] −0.270[^\*^](#TN2){ref-type="table-fn"} 0.263 0.266 0.068 0.054 0.056 0.100 0.132[^\*^](#TN2){ref-type="table-fn"} 0.129[^\*^](#TN2){ref-type="table-fn"} (0.129) (0.139) (0.136) (0.083) (0.076) (0.079) (0.052) (0.056) (0.056) 1\[Male\] 0.132 0.187 0.158 0.096 0.072 0.077 0.025 0.024 0.017 (0.126) (0.100) (0.107) (0.090) (0.076) (0.089) (0.046) (0.046) (0.047) Round:      2 0.133 0.133 0.133 0.063 0.063 0.063 0.067 0.067 0.067 (0.092) (0.092) (0.092) (0.048) (0.048) (0.048) (0.042) (0.042) (0.042)      3 0.150 0.150 0.150 0.138[^\*\*\*^](#TN2){ref-type="table-fn"} 0.138[^\*\*\*^](#TN2){ref-type="table-fn"} 0.138[^\*\*\*^](#TN2){ref-type="table-fn"} 0.083 0.083 0.083 (0.109) (0.109) (0.109) (0.050) (0.050) (0.050) (0.048) (0.048) (0.048)      4 0.183[^\*\*^](#TN2){ref-type="table-fn"} 0.183[^\*\*^](#TN2){ref-type="table-fn"} 0.183[^\*\*^](#TN2){ref-type="table-fn"} 0.050 0.050 0.050 0.117[^\*^](#TN2){ref-type="table-fn"} 0.117[^\*^](#TN2){ref-type="table-fn"} 0.117[^\*^](#TN2){ref-type="table-fn"} (0.070) (0.070) (0.070) (0.047) (0.047) (0.047) (0.048) (0.048) (0.048)      5 0.200[^\*^](#TN2){ref-type="table-fn"} 0.200[^\*^](#TN2){ref-type="table-fn"} 0.200[^\*^](#TN2){ref-type="table-fn"} 0.200[^\*\*\*^](#TN2){ref-type="table-fn"} 0.200[^\*\*\*^](#TN2){ref-type="table-fn"} 0.200[^\*\*\*^](#TN2){ref-type="table-fn"} 0.100 0.100 0.100 (0.089) (0.089) (0.089) (0.045) (0.045) (0.045) (0.052) (0.052) (0.052)      6 0.267[^\*\*^](#TN2){ref-type="table-fn"} 0.267[^\*\*^](#TN2){ref-type="table-fn"} 0.267[^\*\*^](#TN2){ref-type="table-fn"} 0.150[^\*\*\*^](#TN2){ref-type="table-fn"} 0.150[^\*\*\*^](#TN2){ref-type="table-fn"} 0.150[^\*\*\*^](#TN2){ref-type="table-fn"} 0.067 0.067 0.067 (0.088) (0.088) (0.088) (0.054) (0.054) (0.054) (0.049) (0.049) (0.049)      7 0.217[^\*^](#TN2){ref-type="table-fn"} 0.217[^\*^](#TN2){ref-type="table-fn"} 0.217[^\*^](#TN2){ref-type="table-fn"} 0.150[^\*\*\*^](#TN2){ref-type="table-fn"} 0.150[^\*\*\*^](#TN2){ref-type="table-fn"} 0.150[^\*\*\*^](#TN2){ref-type="table-fn"} 0.017 0.017 0.017 (0.098) (0.098) (0.098) (0.047) (0.047) (0.047) (0.048) (0.048) (0.048)      8 0.233[^\*\*^](#TN2){ref-type="table-fn"} 0.233[^\*\*^](#TN2){ref-type="table-fn"} 0.233[^\*\*^](#TN2){ref-type="table-fn"} 0.200[^\*\*\*^](#TN2){ref-type="table-fn"} 0.200[^\*\*\*^](#TN2){ref-type="table-fn"} 0.200[^\*\*\*^](#TN2){ref-type="table-fn"} 0.067 0.067 0.067 (0.088) (0.088) (0.088) (0.057) (0.057) (0.057) (0.056) (0.056) (0.056)      9 0.200 0.200 0.200 0.188[^\*\*\*^](#TN2){ref-type="table-fn"} 0.188[^\*\*\*^](#TN2){ref-type="table-fn"} 0.188[^\*\*\*^](#TN2){ref-type="table-fn"} 0.083 0.083 0.083 (0.113) (0.113) (0.113) (0.047) (0.047) (0.047) (0.060) (0.060) (0.060)      10 0.300[^\*\*^](#TN2){ref-type="table-fn"} 0.300[^\*\*^](#TN2){ref-type="table-fn"} 0.300[^\*\*^](#TN2){ref-type="table-fn"} 0.175[^\*\*\*^](#TN2){ref-type="table-fn"} 0.175[^\*\*\*^](#TN2){ref-type="table-fn"} 0.175[^\*\*\*^](#TN2){ref-type="table-fn"} 0.067 0.067 0.067 (0.115) (0.115) (0.115) (0.056) (0.056) (0.056) (0.049) (0.049) (0.049) *R*^2^ 0.151 0.141 0.160 0.050 0.139 0.140 0.060 0.108 0.111 Estimates of linear probability models explaining the likelihood of decision R by player A and decision r by player B. Standard errors (in parantheses) are clustered at the session level in the Human condition (three clusters per game matrix, six in total) and individual level in the Robot condition (40 clusters per game matrix, 80 in total) and computed using the delete-one jackknife procedure. Models 1 and 2 include a single measure of cognitive skills (a dummy set to 1 for a positive CRT score, or Raven\'s test score), while Model 3 combines both variables. Other independent variables include gender, game matrix and round dummies. The number of observations is N = 600 for Human and N = 800 for Robot conditions. *indicate significance at the 10/5/1% level*. We first turn to player Bs\' behavior. Models 1 and 2 suggest that both the coefficient of *CRT* \> 0 dummy and the coefficient of Raven\'s test score are positive and significant (*p* = 0.067 for *CRT* \> 0 and *p* = 0.015 for Raven). In Model 3, the coefficient of Raven\'s test score remains highly significant (*p* = 0.014), while the coefficient of CRT becomes insignificant (*p* = 0.253). Their joint significance (*p* = 0.034) implies that cognitive skills predict the use of dominant strategy. We now turn to player As\' behavior in the Human condition. Notwithstanding the previous set of results, cognitive skills are not found to explain player As\' choices. The coefficient of *CRT* \> 0 dummy is insignificant (*p* = 0.226) in Model 1, and so is the coefficient of Raven\'s test score (*p* = 0.633) in Model 2. If we account for both, Model 3 reveals that the coefficients of both scores are neither individually (*p* = 0.226 for *CRT* \> 0 and *p* = 0.550 for Raven\'s test score) nor jointly significant (*p* = 0.503). Finally, the behavior of player As in the Robot condition is only predicted by Raven\'s test score: unlike *CRT* \> 0 dummy, its coefficient remains positive and highly significant across models (*p* ≤ 0.001). Unsurprisingly, the joint insignificance of both coefficients in Model 3 is also rejected (*p* = 0.003). Altogether, the results presented in Table [6](#T6){ref-type="table"} suggest that cognitive skills predict certain components of strategic behavior: the use of dominant strategy (reflected in player Bs\' behavior), as well as the ability to best respond to other player\'s dominant strategy (reflected in player As\' behavior in the Robot condition). Moreover, in both cases Raven\'s test score is a more reliable predictor of behavior than CRT score. However, we also observe that Raven\'s test score fails to predict player As\' behavior once player Bs\' behavior becomes uncertain, that is once we move from Robot to Human condition. This, in turn, points toward an interplay between the degree of strategic uncertainty, behavior in the experimental games, and individual cognitive skills. Importantly, the existence of such an interplay is also supported by Figure [2](#F2){ref-type="fig"} which shows that the aggregate levels of efficiency shift upwards between the Human condition and the Robot condition for the 2nd and 3rd Raven\'s score tertile, but not the 1st tertile. In order to formally test this conjecture, we now look at the reaction of player As with different cognitive skills to the disappearance of strategic uncertainty. Splitting the data according to Raven\'s score tertile, for each of the three subsamples we compare player As\' behavior in the Human condition to their behavior in the Robot condition by regressing player As\' choice on the Robot dummy (set to 1 for the Robot and to 0 for the Human condition). We also include the previous set of independent variables (except for Raven\'s test score itself). These results are summarized in Table [7](#T7){ref-type="table"}. The coefficient of the Robot dummy captures the effect of eliminating strategic uncertainty on player As\' behavior for each of the three subsamples. This suggests that only player As with high enough cognitive skills are sensitive to the uncertainty about player Bs\' behavior. The behavior of players with low Raven\'s test score (1st tertile) is unresponsive to the degree of strategic uncertainty: the coefficient of the Robot dummy is close to zero and insignificant (*p* = 0.822). For players with medium scores (2nd tertile), we find a positive yet weakly significant effect (*p* = 0.087) which becomes amplified and highly significant for those player As whose Raven\'s test score belongs to the 3rd tertile of the experimental sample (*p* = 0.012). ###### **The effect of strategic uncertainty and cognitive skills: evidence from player As\' behavior in Human and Robot conditions**. **Raven\'s test score tertile** --------------- ---------------------------------------- -------------------------------------------- ------------------------------------------ Constant 0.277[^\*^](#TN3){ref-type="table-fn"} 0.592[^\*\*\*^](#TN3){ref-type="table-fn"} 0.330[^\*\*^](#TN3){ref-type="table-fn"} (0.147) (0.065) (0.135) 1\[Robot\] 0.044 0.158[^\*^](#TN3){ref-type="table-fn"} 0.428[^\*\*^](#TN3){ref-type="table-fn"} (0.195) (0.090) (0.155) 1\[CRT\>0\] 0.002 0.038 0.016 (0.262) (0.066) (0.188) 1\[Male\] 0.144 0.033 0.212 (0.145) (0.063) (0.179) 1\[Game 2\] 0.034 −0.245[^\*\*^](#TN3){ref-type="table-fn"} −0.176 (0.146) (0.092) (0.155) Round dummies Yes Yes Yes *N* 480 610 310 *R*^2^ 0.048 0.173 0.298 Estimates of linear probability models on decision R by player A. Standard errors (in parentheses) are clustered at the session level in the Human condition (three clusters per game matrix, six in total) and individual level in the Robot condition (40 clusters per game matrix, 80 in total) and computed using the delete-one jackknife procedure. Data from Human and Robot conditions are pooled and split into three subsamples based on Raven\'s test score tertiles. Other independent variables include a dummy set to 1 for a positive CRT score, as well as gender, game matrix and round dummies (omitted from the table). *indicate significance at the 10/5/1% level*. Finally, it is also worth noting that player As\' reaction to the payoff scheme also varies as a function of Raven\'s test score. The coefficient of the Game 2 dummy is close to zero and highly insignificant in the 1st tertile regression (*p* = 0.890). Then, it becomes negative in 2nd and 3rd tertile models (although it is only statistically significant in the former with *p* = 0.012 and *p* = 0.271, respectively). This, in turn, stands in line with the previous finding that player As\' willingness to play *R* increases as the safe choice *L* becomes less attractive relative to outcome (*R, r*). It also seems that the magnitude of this effect is mediated by player As\' cognitive skills, although not in a monotone way. 4. Conclusion {#s4} ============= This paper studies the relationship between strategic behavior and cognitive skills---cognitive reflection and fluid intelligence---in a classic 2 × 2 dominance-solvable game. Our results show that subjects with higher fluid intelligence (measured by Raven\'s progressive matrices test) are more likely to play dominant strategy, and also more likely to best respond to other\'s strategy. Furthermore, fluid intelligence predicts strategic sophistication: only those players with sufficiently high Raven\'s test score are found to display sensitivity to the presence of uncertainty about others\' behavior. Cognitive reflection (measured by CRT), in turn, lacks the power to predict behavior in our experimental setting. We see three main conclusions that stem from these findings. First, these results contribute to the ongoing debate on the relationship between rationality and intelligence (see Stanovich, [@B45], for a critical review). For instance, Stanovich and West ([@B46]) distinguish between two aspects of rational behavior: instrumental rationality which is understood as the "ability to take appropriate action given one\'s goals and beliefs," and epistemic rationality which enables agents to hold "beliefs that are commensurate with available evidence." In the strategic environment investigated in this paper, instrumental rationality can be associated with the ability to solve the game, while epistemic rationality---with the ability to play it with others. Our experimental data suggest an important relationship between fluid intelligence (rather than reflective thinking) and both of these facets of rationality in strategic settings. Both the ability to use dominance and iterated dominance to efficiently solve the game, as well as the responsiveness to the availability of strategic information, is found to be predicted by Raven\'s test score (but not by CRT score). The second contribution is related to the experimental methodology. Despite the fact that CRT and Raven\'s test are both commonly used to measure cognitive skills in experimental subject pools, still very little is known about their relative performance in predicting different types of behavior. Therefore, the choice of one test over the other may happen to be at least as intuitive as evidence-based. As mentioned before, to the best of our knowledge only two experiments address this issue. Brañas-Garza et al. ([@B8]) do so in a strategic environment (*p*-beauty contest game), while Corgnet et al. ([@B15])---in a non-strategic one (individual choices on wealth distribution). Both studies find that CRT performs better than Raven\'s test in predicting subjects\' behavior. The result of the present experiment points the to the opposite conclusion. We believe that this difference is driven by the very nature of the experimental tasks which may involve different types of cognitive effort. In our view, this issue deserves attention in future research. Finally, although we find evidence that behaving in accordance with dominance solvability is positively correlated with cognitive skills, we also substantiate that most of the variance in individual decision making cannot be explained by such skills. Thus, exploring factors alongside cognitive skills that generate strategic behavior remains an open and important empirical question. An interesting avenue is to disentangle individual determinants, e.g., personal characteristics (such as cognitive skills) that are associated with appropriate behavior, from environmental determinants, that is, those features of the decision making environment that lead decision makers to take certain types of actions. 5. Author contributions {#s5} ======================= NH, NJ, SL, and AZ all contributed equally to this work. Authors are listed in an alphabetical order. 6. Funding ========== This project has received funding from JSPS-ANR bilateral research grant BECOA (ANR-11-FRJA-0002), as well as the LABEX CORTEX (ANR-11-LABX-0042) of Université de Lyon, and LABEX OSE of the Paris School of Economics (ANR-10-LABX_93-01), both within the program "Investissements d\'Avenir" (ANR-11-IDEX-007) operated by the French National Research Agency (ANR). Ivan Ouss provided efficient research assistance. We thank Juergen Bracht, Colin Camerer, Guillaume Fréchette, Haoran He, Asen Ivanov, Frédéric Koessler, Rosemarie Nagel, Ariel Rubinstein, Jason F. Shogren, Jean-Marc Tallon, Antoine Terracol, and Marie Claire Villeval for their comments. NH and NJ gratefully acknowledge the *Institut Universitaire de France*. SL thanks the School of Business at the University of Western Australia for hospitality and support. A major part of this work was conducted while NH was affiliated with Aix-Marseille University (Aix-Marseille School of Economics, AMSE) and NJ was affiliated with Université de Lorraine (BETA). NH and NJ thank both institutions for their various supports. Conflict of interest statement ------------------------------ The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ^1^This class of games has been first introduced by Moulin ([@B37]) as the *p*−beauty contest games, where *p* (often equal 2/3) stands for the target fraction of all numbers\' average. ^2^See Nagel ([@B39]) and Ho et al. ([@B31]) for early evidence from the laboratory, Costa-Gomes and Crawford ([@B18]) for a laboratory experiment supporting a behavioral model of bounded rationality, and Bosch-Domenech et al. ([@B7]) for related evidence from the field. ^3^Both Camerer ([@B12]) and Crawford ([@B19]) consider this game as a basic example of a dominance-solvable game, and a glaring case of a mismatch between theoretical predictions and actual behavior. ^4^If the game is played sequentially (so that player A moves first), the same solution can be obtained through backward induction. Note that if *s* \> *h*, the solution does not change (since *l* remains player B\'s weakly dominated strategy), but the outcomes are no longer Pareto-rankable. Beard and Beil ([@B5]), Schotter et al. ([@B44]), and Goeree and Holt ([@B27]) find that this environment also generates important violations of standard theoretical predictions. ^5^Cognitive skills are often measured using (amongst others) the Cognitive Reflection Test (CRT, Frederick, [@B25]), the Raven\'s progressive matrices test (Raven, [@B42]), or both (like in this study). The details of these two measures are presented in Section 2. ^6^RMET consists of a series of photos of the area of the face involving the eyes. Subjects are asked to choose one of the four words that best describes what the person in the photo is thinking or feeling. ^7^See Hampton et al. ([@B29]) for related evidence. ^8^Al-Ubaydli et al. ([@B1], [@B2]) also report that individual cognitive skills do not predict individual willingness to reach efficient outcomes in these two game. ^9^Herein, we restrict our design to these two game matrices and do not seek to further investigate the effects of monetary incentives on both players\' behavior. These effects are analyzed in detail in Jacquemet and Zylbersztejn ([@B35]) and Hanaki et al. ([@B30]). ^10^An English translation of the original instructions in French is provided as supplementary material. ^11^See Jacquemet and Zylbersztejn ([@B34]) for a detailed motivation and description of this design. ^12^Absence of monetary incentives for providing corrects answers is a standard procedure for both CRT and Raven\'s tests. Recent evidence on both tests suggests that monetary incentives do not *per se* affect people\'s performance. See Brañas Garza et al. ([@B9]) for a metastudy on the determinants of CRT scores and Eckartz et al. ([@B21]) and Dessi and Rustichini ([@B20]) for experimental evidence on the role of monetary incentives in Raven\'s test. ^13^See Raven ([@B42]) for an overview. ^14^Model 3 suggests that these two proportions are not significantly different: testing *H*~0~ : β~1~ + β~3~ = 0 yields *p* = 0.303. The increase in the proportion of decisions *R* between Human and Robot conditions is insignificant for Game 1 (*H*~0~ : β~2~ = 0, *p* = 0.679) and significant for Game 2 (*H*~0~ : β~2~ + β~3~ = 0, *p* = 0.054). ^15^See also Frederick ([@B25]) and Bosch-Domènech et al. ([@B6]) for related evidence. ^16^We test the difference in proportion of a given outcome between two experimental conditions by carrying out a bootstrap proportion test that accounts for within-subject correlation, i.e., the fact that the same individual takes 10 decisions. The procedure consists of bootstrapping subjects and their corresponding decisions over all 10 rounds instead of bootstrapping decisions as independent observations (see e.g., Jacquemet et al., [@B33], for a detailed description of the procedure). ^17^Given that most CRT scores in our sample are null and the higher the score, the less frequent it gets, dichotomizing the CRT score variable limits the impact of the outliers on the overall results. Supplementary material {#s6} ====================== The Supplementary Material for this article can be found online at: <http://journal.frontiersin.org/article/10.3389/fpsyg.2016.01188> ###### Click here for additional data file. [^1]: Edited by: Nikolaos Georgantzis, University of Reading, UK [^2]: Reviewed by: Adriana Breaban, Tilburg University, Netherlands; Gerardo Sabater-Grande, Jaume I University, Spain [^3]: This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

***Background.*** VRE is an endemic hospital-acquired organism in the US. There are no proven methods or means of decreasing or eliminating carriage of VRE which is a risk factor for invasive disease. Two small clinical studies showed reduction or elimination of VRE colonization using the probiotic LGG but neither study cultured stool for the presence of LGG, raising questions about whether effects on VRE colonization were due to LGG or other factors. ***Methods.*** To determine whether LGG can eliminate colonization by VRE, we undertook a randomized, double blind, placebo controlled trial in which subjects with VRE received 2 weeks of LGG or placebo. Stool samples were collected and cultured for the presence of VRE and LGG at baseline and days 7, 14, 21, 28 and 56. Presence of LGG was also assessed by PCR on day 14 samples from a subset of 7 subjects. Subjects were monitored closely for adverse events. ***Results.*** Of 694 patients screened, 11 were randomized. Five subjects received LGG and 6 received placebo. VRE was not eliminated during or after treatment in any subject. No significant decreases in VRE colony counts occurred in the subjects receiving LGG (table). LGG was recovered from stool in 2/5 subjects who received LGG. These 2 subjects were among 3/5 who received LGG and were also treated with antibiotics. LGG was detected by PCR in 4/4 samples from subjects who received LGG and 0/3 subjects in the placebo group. Adverse events were similar in the 2 groups. Median Stool VRE Counts by Study Day -------------------------------------- ------ ------ ------ ----------- ------ -------- ------ Day 0 6.97 4.48 7.72 Day 0 7.18 4.36 9.53 Day 7^a^ 7.49 5.72 8.46 Day 7^b^ 8.11 1.04 8.65 Day 14 6.26 4.51 9.54 Day 14 7.86 2.79 8.56 Day 21 6.90 4.60 7.43 Day 21^b^ 7.28 2.54 8.45 Day 28 6.59 4.18 7.38 Day 28 8.28 6.08 8.83 Day 56^b^ 5.26 3.81 7.52 Day 56^b^ 6.74 \<1^c^ 7.20 ^a^ one positive swab included, no CFU available ^b^ one sample not received ^c^ lowest level of detection based on 1ml plated ***Conclusion.*** Although the probiotic LGG is safe in patients with comorbid conditions, we were unable to demonstrate that it resulted in elimination of VRE in an RCT, likely because of continued administration of antibiotics in this patient population. ***Disclosures.*** **All authors:** No reported disclosures. [^1]: **Session:** 108. Clinical - Enteric Infections [^2]: Friday, October 10, 2014: 12:30 PM

A stressful life event may affect prognosis of breast cancer directly through stress-induced alterations of the immune and neuroendocrine system and indirectly through changes in health behaviour, such as physical activity, consumption of alcohol, compliance to therapy and coping with disease. In all, six previous studies have addressed the association between stressful life events and breast cancer prognosis: two of which found a significantly increased risk for recurrence ([@bib12]; [@bib10]), two studies found no association ([@bib6]; [@bib8]), whereas two studies found, contrary to what was expected, a significantly lower risk of recurrence ([@bib2]; [@bib5]). These inconsistent findings may be due to the use of different measures of exposure and methodological weaknesses including self-reported measure of exposure, small samples (*N*=94--665) and selection or recall bias (refer to web appendix: [Supplementary Table A1](#sup1){ref-type="supplementary-material"}). In this large population-based study, of more than 20 000 breast cancer cases, we use objective information from population-based registers and clinical databases to examine the association between a single life event stressor, loss of a partner and breast cancer prognosis. Loss of a partner is a common and also very stressful life event, implying considerable changes to everyday life ([@bib7]). Materials and methods ===================== Linkage of registry data ------------------------ Information on sex, date of birth, current and historical addresses, emigration, disappearance and death with date of these incidences were obtained from the Central Population Register in which all Danish residents since 1968 have been registered with a personal identification number allowing linkage of information between national registers ([@bib11]). Breast cancer ------------- We obtained information on date of breast cancer diagnosis (defined as date of primary surgery), date of recurrence (which is reported up to 10 years after diagnosis), tumour size (in mm), number of tumour-positive lymph nodes, malignancy grade, hormone receptor status and menopausal status, from Danish Breast Cancer Cooperative Group, which contain information on nearly 95% of all breast cancer cases in Denmark since 1977 ([@bib9]). Other cancers ------------- Information on first primary cancer, excluding nonmelanoma skin cancer, was obtained from the Danish Cancer Registry, which since 1943 have registered all cases of cancer (ICD-7) in Denmark ([@bib13]). Stressful life event -------------------- A stressful life event was defined as the death of a cohabiting partner either in the 4 years before breast cancer diagnosis or in subsequent years. 'Cohabitation\' was defined as two persons of the opposite sex over the age of 16, with a maximum age difference of 15 years, living at the same address with no other adults in the residence. Socioeconomic status and comorbidity ------------------------------------ Information on educational level and disposable income (categorised in [Table 1](#tbl1){ref-type="table"}) was obtained 2 years before breast cancer diagnosis from the population-based Integrated Database for Labour Market Research in Statistics Denmark with data on sociodemographic factors in Denmark since 1980 ([@bib14]). Information on comorbidity categorised according to the Charlson comorbidity index (scores 0, 1 and ⩾2), excluding cancers ([@bib3]) was obtained from the Danish National Patient Register with information on all somatic diseases leading to hospitalisation since 1977, and from 1995 also information on all outpatient visits ([@bib1]). Analysed cohort --------------- To attain accurate information from the included registers our base population is restricted to all 3.4 million Danish residents born between 1925 and 1973 who resided in Denmark 1994--2006 and entered the cohort at age 30 (for more details [@bib4]). From this population we identified 22 366 women diagnosed with breast cancer between 1 January 1994 and 31 December 2006 with no previous history of cancer (except nonmelanoma skin cancer), who resided in Denmark 2 years before diagnosis, and had a cohabiting partner up to 4 years before their breast cancer diagnosis. We excluded 1153 women with missing values on one or more covariates or who had resided in Denmark for less than 2 years. In all, 21 213 eligible women were followed for recurrence and death. Statistical analyses -------------------- We used Cox regression to assess hazard ratios (HRs) with 95% confidence intervals (CIs) for all-cause mortality and breast cancer recurrence, respectively, according to the vital status of the partner. For all-cause mortality follow-up time was counted from the date of diagnosis until death, emigration or 31 December 2010, whichever came first. For breast cancer recurrence follow-up time was counted from the date of diagnosis until death, emigration, 10 years of follow-up or 31 December 2006, whichever came first. The exposure, death of partner (after diagnosis), was included as a time-dependent variable, so that person--time before the death of the partner was counted as unexposed, whereas person--time after the date of death of the partner was counted as exposed. In all analyses, time since breast cancer diagnosis was used as the underlying time scale, and baseline hazards were allowed to vary across age at breast cancer diagnosis in 1-year intervals. The HRs were first adjusted for educational level and income, both considered as potential confounders. Subsequently, we adjusted for comorbidity, period of diagnosis, tumour size, number of positive lymph nodes, receptor status and malignancy grade (I--IV), as these factors are strongly associated with outcome, however, not obviously associated with the exposure. We estimated HRs in the intervals \[0--1\],\]1--2\],\]3--4\] years for exposure before diagnosis and for latencies of: \[0--2\],\] 2--5\],\]5--17\] years for exposure after diagnosis. We investigated whether a change in cohabitation status influenced the estimated association, by censoring at 1 January in the year in which the cohabitation status changed. Further, we estimated the association for women diagnosed with hormone-receptor-positive breast cancer and post-menopausal women only. Results ======= In the analysis of the association between loss of partner and all-cause mortality 172 773 person--years of follow-up were accrued, with a median follow-up of 7.7 years (ranging 0--17). During follow-up, 5660 women died, 762 lost their partner in the 4 years before their breast cancer diagnosis and 2259 lost their partner during follow-up. As expected, women who lost their partner were older and had a lower education and income compared with those who did not ([Table 1](#tbl1){ref-type="table"}). After adjustment for education and income as well as period of diagnosis, comorbidity and severity of breast cancer, women who had lost a partner were not at a significantly higher risk for recurrence or all-cause death from that of women who did not lose a partner, no matter if the event happened in the 4 years before diagnosis or in subsequent years ([Table 2](#tbl2){ref-type="table"}) or at different latencies ([Table 3](#tbl3){ref-type="table"}). We found only minor changes to the estimates when censoring at change in cohabitation status and when measuring the association only among women diagnosed with hormone-receptor-positive breast cancer or post-menopausal women (results not shown). Discussion ========== Our results do not support the concern that experiencing a major stressful life event, loss of a partner, negatively affects breast cancer recurrence or all-cause mortality, whether the event occurs in the 4 years before or 0--17 years after diagnosis. Of six previous studies four support this finding ([@bib6]; [@bib2]; [@bib8]; [@bib5]). Two previous studies found a higher risk of recurrence among women reporting stressful life events, however, both were of retrospective design and the observed estimate may reflect differential recall and reporting of events (Ramirez *et al*, 1989; [@bib10]). We addressed several methodological limitations of the previous studies. First, we included more than 30 times as many cancer patients as the largest study published so far. Second, the use of national registers and databases to identify the study population ensures high representativeness and minimal risk for selection bias. Third, the exposure (death of partner) was measured independently of the participants, eliminating recall bias. Fourth, the register-based cohort design ensures temporality and minimises loss to follow-up and misclassification of outcomes. Finally, we estimated the effect of the death of partner before and after diagnosis on both recurrence and all-cause mortality. Still, loss of partner is relatively rare among newly diagnosed breast cancer patients, and 73% of the cohort was alive and recurrence free at exit, resulting in small number of events in certain subgroups. The observed estimates of all-cause mortality may represent overestimates of breast cancer-specific mortality. We adjusted, however, the analyses for comorbidity and investigated also recurrence as outcome. Finally, an association may be present among women with poor coping resources or accumulated stressful life events. We were unable to take these into account. The death of a partner is a common major life event, and our finding of no association with breast cancer recurrence or all-cause mortality may provide reassurance for women confronting breast cancer. This work was funded by the Health Insurance Foundation; and the Danish Cancer Society. For their support in data management we thank Visti Birk Larsen, MD, and Marianne Steding-Jessen MSci, Survivorship, Danish Cancer Society Research Center. **Author contributions** All authors have contributed to the conception and design or analysis and interpretation of data and approved the final version of the report. [Supplementary Information](#sup1){ref-type="supplementary-material"} accompanies the paper on British Journal of Cancer website (http://www.nature.com/bjc) The authors declare no conflict of interest. Supplementary Material {#sup1} ====================== ###### Click here for additional data file. ###### Descriptive characteristics at entry of 21 213 women with breast cancer diagnosed in 1994--2006, Denmark, by vital-status of the partner at exit   **Partner\'s death before diagnosis (*n*=762), No. (%)** **Partner\'s death after diagnosis (*n*=2259), No. (%)** **Partner alive at exit (*n*=18 192), No. (%)** -------------------------------------------------------- ---------------------------------------------------------- ---------------------------------------------------------- ------------------------------------------------- *Sociodemographic characteristics*  Age at time of diagnosis (years)   30--39 2 (1) 25 (1) 1266 (7)   40--49 40 (5) 219 (10) 4640 (26)   50--59 154 (20) 670 (30) 6926 (38)   60--69 385 (51) 1061 (47) 4501 (25)   ⩾70 181 (24) 284 (13) 859 (5)   Mean/median 64/65 60/61 54/54          Level of education[a](#t1-fn1){ref-type="fn"}         Basic or high school 440 (58) 1204 (53) 7127 (39)   Vocational education 227 (30) 703 (31) 6551 (36)   Higher education 95 (12) 352 (16) 4514 (25)          Disposable income[b](#t1-fn2){ref-type="fn"}   Lowest (1st quartile) 340 (45) 536 (24) 2489 (14)   Middle (2nd--3rd quartile) 308 (40) 1107 (49) 8468 (47)   Highest (4th quartile) 114 (15) 616 (27) 7235 (40)         *Medical characteristics*  Tumour size (mm)   0--10 98 (13) 415 (18) 2851 (16)   11--20 297 (39) 963 (43) 7352 (40)   21--50 311 (41) 780 (35) 6560 (36)   ⩾51 22 (3) 53 (2) 737 (4)   Unknown 34 (4) 48 (2) 692 (4)          No. of positive lymph nodes   0 393 (52) 1323 (59) 8979 (49)   1--3 204 (27) 644 (29) 5380 (30)   ⩾4 140 (18) 274 (12) 3434 (19)   Unknown 25 (3) 18 (1) 399 (2)          Malignancy grade   Grade I 186 (24) 660 (29) 4197 (23)   Grade II 261 (34) 738 (33) 6248 (34)   Grade III 130 (17) 321 (14) 3795 (21)   Grade IV 49 (6) 85 (4) 861 (5)   Unknown 136 (18) 455 (20) 3091 (17)          ER and PR Status[c](#t1-fn3){ref-type="fn"}   Negative 129 (17) 380 (17) 3882 (21)   Positive 588 (77) 1733 (77) 13 254 (73)   Unknown 45 (6) 146 (6) 1056 (6)          Period of diagnosis   1994--1996 49 (6) 608 (27) 3100 (17)   1997--1999 134 (18) 557 (25) 3792 (21)   2000--2002 217 (28) 580 (26) 4527 (25)   2003--2006 362 (48) 514 (23) 6773 (37)          Charlson comorbidity index[d](#t1-fn4){ref-type="fn"}   0 603 (79) 1976 (87) 16 334 (90)   1 100 (13) 176 (8) 1201 (7)   ⩾2 59 (8) 107 (5) 657 (4) Highest-education attained 2 years before diagnosis: basic school/high school (7--12 years of primary, secondary and grammar school); vocational training (10--12 years of education); higher education (⩾13 years of education). Disposable income extracted 2 years before diagnosis; income quartiles calculated according to the entire population in the investigated age group, after taxation and interest per person, adjusted for number of people in the household and deflated according to the 2000 value of the Danish crown, as: deflated household income/(no. of person in household 0.6). Negative indicates that individual was estrogen receptor (ER) and progesterone receptor (PR) negative. Positive indicates ER-positive or PR-positive. Accumulated value of all hospital contacts from 1978 to date of diagnosis; scores weighted by level of severity assigned to 19 conditions (excluding cancers) and grouped on the basis of the cumulated sum of scores of 0, 1 and ⩾2. ###### Hazard ratios (HRs) for death and recurrence with 95% confidence intervals (CIs) according to partner\'s vital status among 21 213 women with breast cancer diagnosed in 1994--2006, Denmark         **Crude**[a](#t2-fn1){ref-type="fn"} **Adjusted HR**[b](#t2-fn2){ref-type="fn"} **Adjusted HR**[c](#t2-fn3){ref-type="fn"} ------------------------------------------ -------- --------- ------ -------------------------------------- -------------------------------------------- -------------------------------------------- -------------- ------ -------------- *All-cause mortality*  Partner alive at exit (ref.) 18 192 144 410 4977 1.00   1.00   1.00    Exposed in the 4 years before diagnosis 762 5216 243 1.14 (1.00, 1.30) 1.07 (0.94, 1.22) 1.10 (0.95, 1.27)  Exposed 0--17 years after diagnosis 2259 23 147 440 1.12 (1.01, 1.25) 1.09 (0.98, 1.20) 1.09 (0.98, 1.22)                     *Recurrence*  Partner alive at exit (ref.) 19 312 86 453 2635 1.00   1.00   1.00    Exposed in the 4 years before diagnosis 762 2839 59 0.83 (0.64, 1.08) 0.82 (0.63, 1.06) 0.82 (0.61, 1.09)  Exposed 0--17 years after diagnosis 1139 7581 85 0.98 (0.78, 1.22) 0.97 (0.78, 1.21) 0.93 (0.73, 1.18) Stratified by age at diagnosis in 1-year intervals. Adjusted for the highest-attained educational level, disposable income, and stratified by age at diagnosis in 1-year intervals (defined in [Table 1](#tbl1){ref-type="table"}). Adjusted for the highest-attained educational level, disposable income, period of diagnosis, comorbidity, tumour size, no. of tumour-positive lymph-nodes, hormone receptor status, malignancy grade and stratified by age at diagnosis in 1-year intervals (*N*=19 186) (defined in [Table 1](#tbl1){ref-type="table"}). ###### Hazard ratios (HRs) for death and their 95% confidence intervals (CIs) according to partner\'s vital status and time between diagnosis, death and death of partner among 21 213 women with breast cancer diagnosed in 1994--2006, Denmark         **Crude**[a](#t3-fn1){ref-type="fn"} **Adjusted HR**[b](#t3-fn2){ref-type="fn"} **Adjusted HR**[c](#t3-fn3){ref-type="fn"} ---------------------------------------------- ----- ------- ----- -------------------------------------- -------------------------------------------- -------------------------------------------- -------------- ------ -------------- *All-cause mortality*  Exposure before diagnosis   Time from bereavement to diagnosis (years)    \[0--1\] 228 1610 76 1.17 (0.93, 1.48) 1.15 (0.91, 1.44) 1.04 (0.82, 1.34)    \]1--2\] 195 1391 65 1.14 (0.89, 1.46) 1.04 (0.81, 1.34) 1.05 (0.80, 1.38)    \]2--4\] 339 2215 102 1.11 (0.91, 1.36) 1.04 (0.85, 1.26) 1.19 (0.96, 1.47)                      Exposure after diagnosis   Time from bereavement to death (years)    \[0--1\] 647 5768 173 1.20 (0.98, 1.48) 1.16 (0.95, 1.43) 1.11 (0.89, 1.39)    \[1--5\] 766 7237 156 1.11 (0.97, 1.28) 1.08 (0.94, 1.23) 1.09 (0.94, 1.26)    \>5 846 10142 111 1.07 (0.88, 1.31) 1.04 (0.85, 1.27) 1.08 (0.87, 1.34)                     *Recurrence*  Exposure before diagnosis   Time from bereavement to diagnosis (years)    \[0--1\] 228 904 15 0.66 (0.40, 1.10) 0.66 (0.39, 1.09) 0.57 (0.32, 1.01)    \[1--2\] 195 800 14 0.71 (0.42, 1.20) 0.69 (0.41, 1.17) 0.74 (0.42, 1.32)    \[2--4\] 339 1135 30 1.05 (0.73, 1.51) 1.03 (0.72, 1.49) 1.11 (0.74, 1.65)                      Exposure after diagnosis   Time from bereavement to death (years)    \[0--1\] 467 2439 48 0.90 (0.61, 1.34) 0.90 (0.60, 1.33) 0.73 (0.46, 1.17)    \[1--5\] 412 2837 28 1.01 (0.76, 1.34) 1.00 (0.75, 1.33) 1.01 (0.75, 1.36)    \>5 260 2306 9 1.02 (0.52, 2.00) 1.02 (0.52, 2.00) 1.16 (0.57, 2.37) Stratified by age at diagnosis in 1-year intervals. Adjusted for the highest-attained educational level, disposable income and stratified by age at diagnosis in 1-year intervals. Adjusted for the highest-attained educational level, disposable income, period of diagnosis, comorbidity, tumour size, no. of tumour-positive lymph-nodes, hormone receptor status, malignancy grade and stratified by age at diagnosis in 1-year intervals (*N*=19 186) (defined in [Table 1](#tbl1){ref-type="table"}). The reference category (HR=1) are not experiencing the death of a cohabiting partner in the given year-interval.

INTRODUCTION {#sec1-1} ============ Heat is a physical hazard that could create problems almost in all working environments, especially during the warm months.\[[@ref1]\] Heat stress has a significant relationship with reduced performance.\[[@ref2][@ref3][@ref4]\] Hence, in recent decades the effects of environmental factors and an individual\'s factors on physiologic strains due to heat (heart rate, body core temperature and the sweat rate) have been attempted to be evaluated, and an index or criterion has been tried to be presented in a numerical frame known as as heat stress index.\[[@ref5]\] Among the empirical indices, the wet bulb globe temperature index is used as the evaluating index for heat stress caused by environment factors. To calculate wet bulb globe temperature index, it is necessary to use wet bulb temperature (*T*~W~), black globe temperature (*T*~g~) and dry bulb temperature (T~a~) (for outdoor).\[[@ref6]\] The physiological strain index was established and evaluated by Moran, which is an evaluating index for physiological strains, considering the input load to the cardiovascular system and body thermoregulation.\[[@ref7]\] Sometimes, the measuring process intervenes with workplace activities by some indices and its application is practically difficult in working fields. Sometimes the measuring process requires rather long time (wet bulb globe temperature index).\[[@ref7][@ref8]\] Although direct indices are more applicable as compared with other indices, they only involve the environmental variables such as wet bulb, dry bulb and black globe temperatures. In this regard, Cheung suggested that not only should physiologic responses be used in occupational confrontation standards, but the thermal perceptions that are perceptual responses to heat stress should also be considered.\[[@ref9]\] In a study performed by Dehghan *et al*.,\[[@ref10]\] for content reliability and validation of the structure of the heat stress score index in the climatic conditions of Iran for the male population, results showed that the heat stress score index scale is a reliable and suitable method for screening heat stress in Iran. But regarding the anatomic, physiologic and emotional differences between men and women, the results could not be generalized and considered for women, since gender could also affect the rate of heart stress that a person could tolerate.\[[@ref11]\] The results showed that although the density of activated sweat glands is greater in women, the sweat rate in young women and also old women is less than that in men. Therefore, sweat production by the glands is considerably low in women.\[[@ref12]\] There are different important thermoregulatory factors both in men and women, including specific physiologic factors in women (sex hormones, body water regulation, exercise capacity, anthropometric specifications (weight and body size), body structure (muscles and body fat content) and social behavioral specification (daily physical activity).\[[@ref13]\] The results obtained at the end of three activities in a study by Gagnon *et al*., showed that the body core temperature in women is 38.12 ± 0.18°C as compared with 37.52 ± 0.37°C in men and the esophagus temperature in women is 38.14 ± 0.20°C as compared with 37.41 ± 0.33°C in men. Thus women get tired sooner than men when performing activities and hence production rate decreases, and they need more time to rest while performing similar working activities.\[[@ref14]\] In a study by Yokota *et al*., of women soldiers, five main factors of physical structure specification (tall--fat, tall--thin, average, short--thin, short, obese) were analyzed by anthropometric multiple considerations. Results under similar heat stress showed different tolerance levels, and that body core temperature (T~c~) in these women was less than that in obese--short or fat--tall women.\[[@ref15]\] Due to a lot of differences between men and women regarding heat exposure and also due to differences in perceptions between men and women with regard to ambient temperature, it is necessary to evaluate the effect of heat stress in employed women. The difference in regulating temperature is usually evidenced by the greater body temperature that is quite harmful for women in very hot environments. Compared with men, women require (a) a considerable amount of fat that acts as an insulator and increases heat or thermal storage; (b) a thermoregulation system for high temperatures; and (c) less aerobic capacity, which increases the relative working load of an activity.\[[@ref16]\] Researchers have found that men have less heart rate than women for a given level of heat stress.\[[@ref17]\] Also the area of body covered is greater in women than men in Islamic communities, which probably affects the heat transfer between the body and the environment. Therefore, the purpose of this study was to validate a questionnaire for heat strain evaluation in women. METHODS {#sec1-2} ======= Subjects {#sec2-1} -------- This cross-sectional research was performed on 96 healthy employed women, during summer in 2012. Mean ± (SD) of age was found to be 31.5 ± 7.48 years, of height 1.61 ± 0.05 m, of weight 61.55 ± 10.35 kg, and of body mass index 23.52 ± 3.75 kg/m^2^ in different workplaces. They worked in different warm workplaces, including greenhouse settings, hospital laundry, and confectionary factories. Subjects were selected by systematic random sampling. The inclusion criteria were non-cold disease during the past week, diabetes, epilepsy, convulsion, hyperthyroidism, respiratory diseases, cardiovascular diseases, and not consuming medicines. The exclusion criteria were non-participation and non-cooperation in measuring heart rate and oral temperature. This study was performed after getting permission from the Ethic Committee in Medicine. Measurements {#sec2-2} ------------ After selecting a subject, her height and weight were measured (weight by a digital scale with 0.1-kg accuracy). The belt-like sensor of the heart-rate monitoring device (Polar Electro RS100, Finland) was fastened to the chest and the monitor was fastened to the participant\'s wrists.\[[@ref18]\] Also, oral temperature was measured by a medical digital thermometer (Digital Thermometer; Omron). Oral temperature was measured for 5 min with a closed mouth and in conditions where environment temperature (ambient temperature) was over 18°C, prohibited from eating, drinking, and smoking for at least 15 min before the measurement, to reduce the effect of environmental conditions.\[[@ref19]\] Heart rate and oral temperature were measured in the resting and working states. After 15 min of rest out of the warm workstation (cool place), heart rate and oral temperature were measured as baseline. Then the subject started to work, and heart rate and oral temperature were measured every 5 min for 120 min. For simultaneous measurement of physiological parameters, a questionnaire, including 45 items, was asked to be filled out in the resting and working states every 30 min for 120 min (four times). This questionnaire included the most important probable factors in onset of heat \[[Table 1](#T1){ref-type="table"}\].\[[@ref20][@ref21][@ref22][@ref23][@ref24]\] ###### Probable effective factors onset heat strain  The value of Cronbach\'s α was calculated for reliability of the questionnaire. For validity of the questionnaire\'s structure, calculations regarding the correlation coefficient of the questionnaire\'s questions, exploratory factor analysis and confirmatory factor analysis were used. The maximum likelihood method was applied for estimating the model, and to consider the fitness of the model, χ^2^ indices, (goodness-of-fit indices (GFI)), comparative fit indices (CFI), root mean-squared error of approximation (RMSEA), and root mean-squared residual (RMR) were used. If GFI and CFI were greater than 90%, it would indicate very appropriate fitness and if the indices were greater than 80%, it would indicate appropriate fitness. If RMR and RMSEA were less than 0.05, it indicated very suitable fitness and if they were less than 0.08, it showed acceptable fitness.\[[@ref25]\] Finally, after calculating regression weight, which was obtained in the three-factor model, it was multiplied by the score on each of the questions selected by individuals and the total showed the total score on the questionnaire. RESULTS {#sec1-3} ======= In this study the total number of participants was 142 and the mean ± SD of age was 31.5. ±7.48 years, of height 1.61 ± 0.05 m, of weight of 61.55 ± 10.35 kg, of working history 2.55 ± 0.62 years, and of body mass index of 23.52 ± 3.78 kg/m^2^. In this study, internal stability of the questions was estimated by using a Cronbach\'s α value of 0.68, on 45 questions. The variable intensity of thermal discomfort had the highest correlation (0.71) and the variable shift work had the lowest correlation as compared with the total correlation. Kaiser Meyer Olkin (KMO) test was used to analyze the adequacy (sufficiency) of the sample volume. The value of KMO for the present research was 0.89, which indicates suitability of the number of samples for performing the factor analysis. For validity and significance of the model, analyses with one, five, four, and confirmatory factor analysis with two and three factors, were used. Variables that had item--total correlation less than 0.24 were eliminated from the study. Finally, the structure was accepted according to filling the indices of a three-factor model with 16 variables. Eleven variables in the first factor (perceptual), three variables in the second factor (environmental), and three variables in the third factor (individual or personal) were considered \[[Table 2](#T2){ref-type="table"}\]. Entered variables in the perceptual factor with a factor loading of 0.20-1.36 were determined. Entered variables in the environmental factor with a loading of 0.35-1.09 were considered and entered variables in the personal factor with a factor loading of 0.30-0.53 were determined. The first factor of 42.4, second factor of 12.9, and third factor of 8.3 were taken from the variance and the total of three factors, with a variance 63.6 being determined. ###### Regression weight of the entered variables in the three-factor questionnaire  As shown in [Table 3](#T3){ref-type="table"}, the RMSEA, GFI and CFI indices indicate suitable fit and confirm the estimated model. ###### Fitness indices in the three-factor questionnaire  Pearson correlation test showed that there was a significant correlation between the questionnaire scores and oral temperature (*r* = 0.414, *P* \< 0.001) and heart rate (*r* = 0.247, *P* = 0.015). Oral temperature was selected as the decision criterion for determining the cut-off point, so that we considered an oral temperature of less than 37.5°C as no or low heat strain and more than 37.5°C as heat strain. Receiver operator characteristic analysis was used to calculate sensitivity, specificity, and cut-off point Area under the receiver operator characteristic curve \[[Figure 1](#F1){ref-type="fig"}\], which is a global summary statistic of diagnostic accuracy, was 0.659 (*P* \< 0.033). The appropriate cut-off point of the questionnaire was obtained to be 17, which will indicate the existence of heat strain in a person \[[Table 4](#T4){ref-type="table"}\]. Sensitivity and specificity were obtained to be 0.789 and 0.584, respectively, using this cut-off point. {#F1} ###### Values of sensitivity and specificity of the questionnaire in the cut-off point domain  DISCUSSION {#sec1-4} ========== Gender is one of the effective factors in heat exchange between the human body and the environment Hence evaluation and assessment of heat stress under different climatic conditions is essential among the male and female populations. Since men and women under heat exposure, in different studies, did not show similar results regarding heart rate and rectal temperature, under various climatic conditions and similar workplace activities,\[[@ref12][@ref13][@ref14][@ref17]\] therefore, the present study was performed with the aim of determining the validity of a questionnaire method to evaluate heat strain in women, in their workplaces. In an perceptual--observational checklist, Bethea and Parsons\[[@ref26]\] selected seven determining parameters for heat stress risk, of which five parameters, ambient temperature, humidity, radiant temperature average, air movement and rate of physical activity, had been entered into our questionnaire. Two variables type and material of the clothes had been eliminated from the questionnaire due to low item--total correlation. The feeling of heat in human beings depends on their heat balance. Heat balance is in turn affected by physical activities, clothes, and climatic parameters such as ambient temperature, radiant temperature average, air velocity, and humidity.\[[@ref27]\] All parameters apart from the variable clothes were accepted in our questionnaire with high factor loading. In a study performed by Maeen Zakaria, the most important parameters introduced in emergence of heat stress were physical activity, clothes, and climatic conditions, and the most important physiology responses to heat stress were heart rate, rectal temperature, and intensity of perspiration,\[[@ref28]\] and all these parameters were considered in this investigation. In a research done by Rodriguez\[[@ref29]\] on the cooling responses to the heat stress, sweat rate was expressed as one of the most important mechanisms in cooling the body via skin. Also parameters ambient temperature, air velocity, and humidity were considered as the determining factors in sweat rate;\[[@ref29]\] all the above parameters were accepted as being significant. Binkley *et al*.,\[[@ref30]\] have recommended the use of the parameters consuming liquids and intensity of thirst during physical activities under warm climatic conditions. Mueller and Diehl\[[@ref31]\] have stated the highest rate of heart stroke and mortality to be among athletes and people performing heavy physical activities.\[[@ref31]\] Gaffin *et al*., recognized sunlight intensity as the greatest source of receiving heat under warm climatic conditions and introduced direct exposure to heat sources and radiation temperature as the next factors for receiving heat, emergence of diseases, and heat exhaustion,\[[@ref32]\] and the above items were entered into our questionnaire. Chengalur *et al*.,\[[@ref33]\] recognized ways of controlling heat stress in workplaces, including reducing air temperature, reducing humidity, increasing air velocity, reducing physical activity intensity, wearing suitable clothes, using a heat shield, work--rest programs, replacing fluids and electrolytes, heat adaptation, and instructing workers. All the parameters were considered in our questionnaire. In a similar study that was performed on the male population, the variables type of clothing, color of clothing, material of clothes, type of personal protective equipments, and clinical symptoms had significant factor loading in a structured questionnaire,\[[@ref10]\] whereas these variables had no significant factor loading in our questionnaire, therefore they were eliminated. Instead, the variables solar radiation, working dimensions, duration of heat exposure, and history of heat stroke replaced them. The present study, which was performed among the population of women in rather warm areas, showed that the rate of heat stress was not so high in this population, the reasons for which may be: (1) due to sensitivity of women to high temperatures, most of the women were doing their heavy load work during the colder times of the day; (2) providing a cold resting area was provided at a distance near workstations; (3) had a rest time for eliminating tiredness and reducing heat stress; (4) existence of cool water sources near the workstations and in short distances, which compensation for body dehydration in the individuals; and (5) evidence show that aware workers who had been encouraged to self-pacing adjusted their work under heat stress conditions to prevent the physiological pressures imposed on them. CONCLUSIONS {#sec1-5} =========== The results of this research indicated that the 16 variables in our questionnaire \[[Appendix 1](#A1){ref-type="app"}\], effective in onset of heat strain, could be measured by questions and observation including key factors in the domain of heat stress. This quantitative questionnaire also has an acceptable reliability and validity, and a cut-off point, and therefore it could be used in the preliminary screening of heat strain in women in warm workplaces, when other heat stress evaluation methods are not available The results of this study have not been under any other climate conditions, hence it is recommended that the prototype should be investigated in other climate conditions. This study was the result of an MSc dissertation approved by the Vice President for research of Isfahan University of Medical Sciences under project number 391056. The authors highly appreciate the women workers who participated in the study. **Source of Support:** Nil **Conflict of Interest:** None declared 

1. Introduction {#sec1} =============== The nervous system has two parts: (1) the CNS and (2) the peripheral nervous system. Neural signals from the CNS have repeatedly been observed to activate specific muscle groups during the performance of motor tasks, referred to as motor primitive (MP). How the CNS makes this selection from a seemingly vast pool of such primitives to attain a behavioral goal is a complicated question in the field of motor control. The task is computationally challenging, and for over 50 years, researchers since Bernstein have investigated how the CNS reduces the degrees of freedom by focusing on a smaller set of variables \[[@B1]\]. The evidence in support of the CNS use of motor primitives seems conclusive. Sherrington, Sharrard, and Ferrier and Yeo concluded that after stimulating the spinal cord (SC), the intricate network of nerves resulted in highly coordinated functional synergies in the musculature of dogs, frogs, cats, rabbits, and monkeys \[[@B2]--[@B4]\]. The root stimulation of various SC segments in humans resulted in different coordinated muscular reflexes in the lower limb \[[@B3]\]. In this review paper, we will analyze evidence supporting the existence of MS among different species resulting in the modular organization of the CNS. Microstimulation is predominantly used to examine natural motor behavior \[[@B5]\]. The combination of MS/encoded modules produces different natural motor behaviors that are task-dependent (i.e., task-specific MS) or task-independent (shared MS) \[[@B6]--[@B8]\]. There is an ongoing debate about the origin of MS whether they have a neural origin or a nonneural origin, that is, whether they are encoded in the CNS or activated because of task constraints. With advances in EMG and functional magnetic resonance imaging (fMRI) data analysis, the current view leans towards the organization of MS as spatiotemporal components in the brain and SC, thus supporting their neural origin. We examine the various techniques and algorithms used for the extraction of MS and present a mathematical model of MS. We also review how the central pattern generators (CPGs) in the absence of peripheral feedback trigger the MS. Further, MS as a physiological marker in stroke patients and its future prospects in neuro-rehabilitation, robotics, and sports science are discussed. 2. The Existence of MS {#sec2} ====================== The presence of MS in motor tasks has been supported in experimental studies for several decades. [Figure 1](#fig1){ref-type="fig"} shows how encoded MS/MP in the CNS activates muscles resulting in force production and limb displacement. 2.1. Force Fields/Motor Primitives {#sec2.1} ---------------------------------- MS have been associated with the stable force produced by the limbs during natural motor behavior. Maton and Bouisset observed that synergistic groups of muscles produced different external forces during supination or pronation and the external force produced was the sum of the forces of various muscles \[[@B9]\]. Further, the EMG signal was equal to the coefficient times the forces produced by the muscles \[[@B9]\]. It is believed that these forces are present in the CNS and are known as motor primitives (MPs). Giszter et al. during SC microstimulation in frogs concluded the existence of MP \[[@B10]\]. These MPs or convergent force fields (CFFs) present in the SC among vertebrates and invertebrates are the building blocks for complex motor behavior, and by their vector combination, a wide repertoire of behaviors can be generated \[[@B10]--[@B15]\]. It has been proposed that the CNS uses these convergent force fields to solve the inverse dynamics problem to reduce the kinematic degrees of freedom \[[@B16]\]. Motor primitives usually develop from a neonatal stage to toddler stage \[[@B17]\]. Some of them are encoded into the spinal cord during skill acquisition. In humans, an internal model of the force field in the CNS is constructed with a given motor task during a robot-guided movement. These newly encoded computational modules represent internal coordinates or muscle/joint coordinates \[[@B18], [@B19]\]. Huesler further demonstrated synchronization and nonsynchronization of motor units on inter- and intramuscular pairs during the production of force \[[@B20]\]. The motor unit synchronization during precision grip varied along different force and muscle activation levels thus validating the presence of predefined modules. Neuromechanical models provide better understanding of superfluity of muscles and their neural control during a behavioral task \[[@B21]\]. The muscle fibers and motor neurons (MN) together constitute a motor unit (MU); the activation of MU results in muscle activation and contraction \[[@B22]\]. The motor neurons carry this encoded information/motor drives for the coordinated activation of specific groups of muscles for specific tasks; thus, MPs/motor drives are also referred to as MS. McKay and Ting gave more insight into MS as the fundamental blocks for movements using a 3D model of the cat\'s hind limb \[[@B23]\]. The premotor drives serve as units for muscle coordination. The drives represent MP in the circuitry of the SC, and the unit bursts are the activation command from the central network \[[@B24]\]. Hart and Giszter observed the complex behavior in the brainstem of spinalized frogs, whereby the extracted drives associated with the unit burst from EMG data were focused on a set of muscles \[[@B25]\]. MPs are classified into kinematic (stable correlation between joint angles), dynamic (stable correlation between joint torques), and neuronal, based on previous studies \[[@B26], [@B27]\]. There is little evidence of kinematic synergies represented in the M1 region (of the motor cortex), but kinematic synergies seem to originate from the MS \[[@B28], [@B29]\]. The CNS, instead of employing synergies at the kinematic and/or dynamic level, controls them as MS, which are the covariation of correlated, less stable EMG activities \[[@B26], [@B28], [@B29]\]. From the current understanding of natural motor behavior, MP can be classified into acquired and adaptive. The acquired MPs are embedded in the genetic design of our nervous system, whereas the adaptive MPs are learned and then get encoded into the spinal circuitry \[[@B14], [@B18], [@B19], [@B30], [@B31]\]. The changes in the CFFs or MPs cause changes in muscle coordination eliciting different natural motor behaviors \[[@B31], [@B32]\]. Various techniques like microstimulation, cutaneous stimulation, and N-methyl-D-aspartate (NMDA) iontophoresis were used to understand the relation between MP and motor behavior \[[@B11], [@B32]--[@B35]\]. The presence of MP is supported by recent developments in the field of computational neuroscience. 2.2. Combination of Motor Modules {#sec2.2} --------------------------------- We have already discussed the presence of CFFs or MPs in the spinal circuitry, which gave us an idea that endpoint forces are dependent on coactivation of muscles. The discussion will now focus on how these primitives are organized in the spinal circuitry and how they are combined to produce various movements. Georgopoulos et al. observed that during reaching movements, the cellular discharge was higher in the rhesus macaque motor cortex when the movement was in the desired direction \[[@B36]\]. It was concluded that the preferred location of the limb in a 3D space is achieved by the vector sum of motor cortex\'s cellular discharge (population coding). Mussa-Ivaldi et al. investigated simultaneous stimulation of two sites in the spinal cord resulting in the endpoint forces being equivalent to vector summation of each site after costimulating them individually in spinalized frogs \[[@B37]\]. The outcome was the same in cats during intracortical microstimulation (ICMS) of the primary motor cortex \[[@B38]\]. Tresch et al. and Kargo and Giszter also found the summation of primitives in vertebrates using ICMS and cutaneous stimulation \[[@B35], [@B39]\]. Lemay et al. used EMG to compute the forces at the muscular level and mechanically at the ankle in hind limbs of spinalized frogs \[[@B40]\]. During intraspinal electrical stimulation in the spinalized frog, the endpoint forces were the vector summation of each stimulated site. Capaday and van Vreeswijk demonstrated a mechanism in which the motor output is a linear summation without being affected by the presence of nonlinearity (intracortical and intraspinal nonlinearity) in the neural circuits \[[@B41]\]. It was concluded that such nonlinear neural circuit elements do not affect the linear summation of the motor output. It is interesting that in vertebrates, summation of the motor unit action potentials (MUAPs) is observed using glutamate iontophoresis or electrical stimulation of neurons \[[@B38], [@B42]\], which is very much similar to MP summation during stimulations of the SC. Electromyographic signals being an algebraic summation of MUAPs are extensively utilized for the extraction of movement primitives or MS by employing decomposition algorithms \[[@B40]--[@B42]\]. This understanding leads us to the neural basis of MS. 2.3. Neuromechanical Origin of MS {#sec2.3} --------------------------------- The modular organization of MS in the SC is linked to their neural origin \[[@B7], [@B12], [@B13], [@B17], [@B25], [@B43], [@B44]\]. The feasible force fields with synergies leading to a volumetric reduction of the forces offer dimensional control criteria \[[@B45], [@B46]\]. Thus, a low-dimensional spatiotemporal structure can explain the muscle activation pattern and the neural origin \[[@B6], [@B23], [@B43], [@B45], [@B47]--[@B50]\]. The low-dimensional space was also consistent between stroke patients and normal patients and within the stroke patient\'s affected and unaffected sites leading to the speculation of the neural emergence of MS \[[@B51]--[@B53]\]. Focal intraspinal N-methyl-D-aspartate (NMDA) also supports the existence of encoded MS in the CNS \[[@B32]\]. d\'Avella et al. extracted MS during different natural behaviors in frogs \[[@B6]\]. Their results suggest that a small number of components were sufficient to explain the high-dimensional space vector of the time-varying muscle pattern. The low-dimensional space observed was independent of the task performed as synergies for kicking in frogs were similar between different motor behaviors and hence called shared MS. The lack of similarities in the synergies during different behaviors is due to task-dependent synergies. Such low dimensionality is dependent on task constraints. d\'Avella and Bizzi extracted synchronous and time-varying synergies in frogs and found that most synergies were shared, but synergies associated with biomechanical constraints also existed \[[@B7]\]. The synergies extracted during postural tasks in humans had a low-dimensional space \[[@B48]\]. That study also suggested that MS from the CNS were changed during postural adjustment and the number of extracted synergies remained consistent suggesting the neural origin of MS in cats, frogs, humans, and primates \[[@B45], [@B48], [@B54]--[@B58]\]. The intrasubject consistency of MS was observed during different tasks, thus concluding modulation of MS recruitment instead of its structure \[[@B59]\]. However, dorsal root transection on frogs altered the temporal pattern and amplitude of activation coefficients (structure) of shared synergies. The fine-tuning of the synergies was not consistent within the frogs because of different musculatures \[[@B60]\]. A low-dimensional space is related to neural control and reduction in degrees of freedom \[[@B6], [@B23], [@B43], [@B61]--[@B63]\]. Valero-Cuevas et al. provided evidence against the lack of dimensionality, thus arguing against the concept of MS by employing a simple task using intramuscular EMG in the index finger \[[@B50]\]. The argument against the synergy hypothesis is that they only exist during open-loop tasks. However, the claim that MS are hard to reveal in the closed-loop tasks because such tasks are not rich in behavior, is not well supported since MS are well suited with open-loop and closed-loop tasks and can be implemented as an optimal feedback controller. A slight variation in the task should display their variability \[[@B7], [@B45], [@B50]\]. Kutch and Valero-Cuevas, using cadaveric and computational models, showed that the biomechanics of the limbs required change in muscle-tendon length to a low dimensionality in different movements. Each muscle group varied the length change individually leading to the emergence of nonneural coupling between muscles and the presence of dimensionality reduction constraints during isometric force production in different directions \[[@B64]\]. The muscles were individually coordinated instead of being grouped together in this experiment. A low-dimensional space originated directly from motor commands (feed-forward MS), biomechanical constraints, and the individual response of each muscle (feed-backward MS) in the experiment. The endpoint forces for feed-forward MS were static as the dimensional space emerged directly from motor commands, contradicting the fact that variation end-point forces with different tasks is commonly observed on a cellular and physical level (cortical discharge) \[[@B15], [@B36], [@B65]--[@B68]\]. This concludes that the CNS does not need to control a group of muscles to observe EMG signal of low dimensionality. De Groote et al. using a musculoskeletal model and minimized muscle effort found that MS existence is dependent on the task constraints rather than on neural control \[[@B69]\]. Criticism of the synergy hypothesis because of lack of dimensionality could be due to the data processing approach and noise. These factors make it harder to reveal the reduced dimensionality in EMG signals \[[@B50], [@B69]\]. Bizzi and Cheung also advocated for the neural origin of MS by discussing the dimensional reduction of the subspace with task constraints \[[@B17]\]. For isometric force production in lower limbs, seven synergies were required but the dimensionality reduced further to 4-5 synergies during locomotion \[[@B17], [@B64], [@B70]\]. The dimensional space was lower than expected from the task constraint, as it is the neural signals that limit motor output required for the task \[[@B21]\]. Countering Kutch\'s experimental model, they added that the organization of muscles around a joint results in muscle coupling as it produces regularities in EMG \[[@B17], [@B64]\]. The question raised by Bizzi and Cheung on the lower limb producing isometric force with similar dimensionality can be understood from Hagio and Kouzaki \[[@B17], [@B43]\]. The latter studies provide strong evidence for the neural basis of MS in the lower limb during isometric force production in a 3D space. There are many factors that affect the dimensional subspace: the data processing approach and noise factor \[[@B69]\], smoothing or regularities in the EMG signal \[[@B71]\], and the change in muscle-tendon length synchronously \[[@B17]\]. During tasks involving 52 postures and recorded EMG from intrinsic and extrinsic hand muscles, Weiss and Flanders concluded that the MUs from the CNS were linked to coactivation of multiple muscles \[[@B72]\]. With these evidences, it would be correct to infer that MS have neural basis. The feedback from the peripheral nervous system assists to acquire MP into the spinal circuitry for a certain task. The stored MP will assist the CNS to adjust biomechanics of the limbs to accomplish the task \[[@B73]\]. In short, the neural control in most cases overcomes the passive dynamics but in some instances utilizes the passive dynamics to achieve the task \[[@B21]\]. In the absence of sensory feedback, a CPG reduces the dimensionality of locomotion \[[@B74]\]. The CPG generates the MS with variation of the task by the process of neuromechanical tuning in rhythmic movements, and thus it is more appropriate to say that MS have a neuromechanical origin \[[@B12], [@B21], [@B61], [@B73]--[@B80]\]. This process is clearly defined in [Figure 2](#fig2){ref-type="fig"}. 3. Brain and MS {#sec3} =============== Recent research has often employed brain fMRI to study MS. When performing a task, the primary motor cortex (M1) region exhibits changes on a cellular level by a high neuronal discharge rate in a preferred direction \[[@B15], [@B36], [@B66]\]. In M1, the axons of Betz cells and cortical neurons descend from the brain to the SC forming corticospinal pathways. Cash and Yuste demonstrated linear summation of excitatory inputs by pyramidal neurons (Betz cells) in the hippocampus (i.e., sliced biopsy) of rats \[[@B81]\]. Like MPs, linear arithmetic of neurons provides independent processing of multiple channels without disruption of information \[[@B41], [@B81]\]. The M1 is subdivided into the rostral and caudal regions in higher primates including humans. The former mediates motoneuronal activity in the SC, and the latter is known for higher skill acquisition by skipping the spinal circuitry \[[@B82]\]. Cherian implanted electrode arrays in the M1 region of rhesus macaques. The outcome was that under a force field, the neuronal discharge was synchronous with muscle dynamics and M1 did not account for motor learning directly. Law et al.\'s research with implanted microelectrode arrays in rhesus macaques while performing a task showed that the extra group of neurons was comodulated from M1, and such flexibility of M1 resulted in enhanced motor skills \[[@B83]\]. The MS encoded into the spinal circuitry are updated by the cortical regions through corticospinal pathways pertaining to specific tasks. The flexibility of these task-specific MS is restrained by the unequal combination of muscle fields \[[@B84]\]. In rhesus macaques, electrical microstimulation of the motor cortex generates synergistic muscular activity for multiple degrees of movements \[[@B33], [@B34]\]. Decomposition algorithms were used on muscle and motor cortex data acquired after microstimulation of the cortical region while performing different tasks. The neuronal firing pattern decomposed and corresponded to the same dimensionality as that from the muscle recordings, which concluded the presence of spatiotemporal synergies in the brain \[[@B34]\]. Using transcranial magnetic stimulation (TMS), it was found that the human cortical region is associated with the activation of different synergistic groups of muscles in the lower portion and this was validated with EMG and fMRI recordings \[[@B85]\]. The cortical regions are the center for muscle coactivation or MS pertinent to functional tasks \[[@B86], [@B87]\]. Optimal coordinated movement is caused by the convergence and divergence of the corticospinal system with central neuromotor noise. The organization of MS in the primate motor cortical region has been studied, and synergies in this area are designated as discrete spatiotemporal synergies \[[@B34]\]. In phylogenetically advanced species, while grasping, instead of synergetic control of muscles, a more fractionated control is superimposed \[[@B88]\]. The fractionated control is due to the new cortico-motor neuronal pathways which bypass the spinal circuitry. The new M1 cells in higher primates are associated with these fractionated control pathways which are also important for skill acquisition \[[@B82], [@B88]\]. This novel approach of the CNS is the foundation for the development of prosthetic arms. The feedback mechanism is important for movement, as in grasping MS are modulated with respect to the variation in the shape of object among monkeys \[[@B56], [@B89]\]. In the primary motor cortex, M1 dictates reaching and grasping objects as a single movement separated more in time than in space \[[@B15], [@B54], [@B90]\]. However, mammals with transected SC produced rhythmic movements revealing the presence of CPGs in the CNS. Brown conducted an experiment with afferented and deafferented cats and found similar MS in both groups of animals \[[@B80]\]. The CPG activates the muscle groups that generate rhythmic patterns in vertebrates without sensory feedback \[[@B75], [@B76], [@B80], [@B91]\]. In the absence of a sensory feedback or during rhythmic movements like running, modulation in CPG by neuromodulators results in the modulation of the timing, duration, and magnitude of MS \[[@B54], [@B70], [@B74], [@B91], [@B92]\]. Neural signals are transmitted from the basal ganglia and thalamus to the midbrain locomotor region (MBLR) and to the SC for CPG generation \[[@B54], [@B76]\] ([Figure 2](#fig2){ref-type="fig"}). Prochazka and colleagues \[[@B78], [@B79]\] and Drew et al. \[[@B86]\] concluded that the motor commands pertaining to movement pace originate from the cortex region, channeled towards the SC via the MBLR. It drives the CPG timer in the SC to produce cadences with the flexor and extensor phase duration. During the change in velocity of the rhythmic movements, the CPG pattern formation network driven by the motor commands modulates the activation of the muscles as per square law relationship \[[@B78]\]. Hence, CPG phase durations and the muscle forces match without the presence of a sensory feedback with biomechanically changing events. Overduin et al. using ICMS provided a better insight into the location of MS \[[@B5]\]. The spatiotemporal patterns rather than being encoded into the cortex region were present in the deeper part of the brain (brainstem or SC) \[[@B5]\]. The alpha-motor neurons in the brainstem and spinal cord were activated during kinetic and kinematic gait events that map spatiotemporal patterns \[[@B93]\]. We could infer that the M1 region dictates the movement through alpha-motor neurons, Betz cells, and other cortical neurons by triggering the MS in the SC or brainstem. 4. Algorithms for Extracting MS {#sec4} =============================== Muscle synergies as a linear combination are usually extracted using matrix factorization algorithms like the independent component analysis (ICA), nonnegative matrix factorization (NNMF), and factor analysis (FA) \[[@B94]\]. The application of dimensional reduction algorithms on EMG data is important to observe voluntary or nonvoluntary movements because the planning of movements happens in a low-dimensional space. The mathematical model of MS is classified into time-varying synergies (spatiotemporal synergies) and time-invariant synergies (spatially fixed or synchronous and temporal pattern) \[[@B95], [@B96]\]. The spatial and temporal synergies are elicited individually by the CNS \[[@B49]\]. The nonnegative matrix factorization multiplicative method is commonly used to extract the MS. Being an optimization algorithm, the linear decomposition minimizes the reconstruction error \[[@B6]\]. It is also important to determine which type of MS needs to be extracted from the EMG. The point process statistics method was used by Hart and Giszter to distinguish between time-variant and time-invariant MS prior to extracting them from the EMG pulse timing or onset timing \[[@B97]\]. Equation ([1](#EEq1){ref-type="disp-formula"}) gives a mathematical model of MS: here, *E* is the EMG signal with *n* sensors (or muscles) and a total of *m* samples, **W** is the synergy matrix with reduced dimensionality, *s* is the number of components or dimensions to be extracted from the EMG, *C* includes coefficients of neural command vectors and *e* is the residual error. $$\begin{matrix} {E_{n \times m} = \mathbf{W}_{n \times s} \times C_{s \times m} + e.} \\ \end{matrix}$$(a)Time-variant synergies: $$\begin{matrix} {E\left( t \right) = \sum\limits_{i = 1}^{N}\mathbf{W}_{i}C_{i}\left( {t - t_{i}} \right),} \\ \end{matrix}$$where *N* is the number of synergies and *E*(*t*) is the activation of muscles at time *t*. The individual synergy vectors are shifted in amplitude and scaled in time by the activation coefficients \[[@B95]\]. (b)Time-invariant synergies: $$\begin{matrix} {E\left( t \right) = \sum\limits_{i = 1}^{N}\mathbf{W}_{i}C_{i}\left( t \right).} \\ \end{matrix}$$ The spatial patterns W are the task-dependent control inputs whereas the activation coefficients C are the task -independent predefined modules \[[@B95]\]. Artificial neural networks (ANN) have been similarly employed to extract MS \[[@B98]\]. The spatial and temporal patterns from EMG were estimated by the algorithms with no unique solution \[[@B99]\]. The performance of PCA is not laudable when compared with that of other algorithms \[[@B45], [@B50], [@B94], [@B100]\]. Tresch et al. compared the performance of different algorithms \[[@B94]\] using a simulated and recorded (EMG) data set with nonnegative values. The performance of each algorithm is listed in [Table 1](#tab1){ref-type="table"}. Principal component analysis sometimes gives negative values in the synergy subspace, which represents the inhibitory response of the spinal circuitry \[[@B94]\]. Krishnamoorthy et al. used an uncontrolled manifold approach (UCM) hypothesis to extract postural MS \[[@B48]\]. An autoencoder performed better as a synergy extractor than other algorithms while reconstructing the EMG data; it extends its performance by providing an agonistic and antagonistic relationship between muscles \[[@B98]\]. Principal component analysis also provides such relations but does not offer good reconstruction efficiency \[[@B98]\]. Algorithms like NNMF and probabilistic independent component analysis (pICA) that utilize the hypothesis of muscle synergies performed with higher classification accuracy to distinguish between the single and multiple degrees of freedoms of upper and lower extremities \[[@B101], [@B102]\]. Electromyographic signals being corrupted by signal-dependent noise could be the reason that these algorithms conveyed better classification accuracy since NNMF and pICA execute well with signal-dependent noise \[[@B25], [@B101]--[@B104]\]. The dimensionality of control commands (neural commands from the CNS) is not exactly the same as the elements of the state space (musculoskeletal structure); therefore, the exact number of synergies needs to be determined prior to extracting them using different computational algorithms \[[@B48]\]. The MS analysis is dependent on many variables including muscles included, algorithms, EMG normalization method, constant or varying synergy vector (SV), output vector normalization method, and synergy comparison method \[[@B105]\]. Bartlett\'s test, Akaike information criteria, Bayesian information criteria, Laplacian information criteria, and Likelihood ratio test can all be used to identify the correct number of synergies for Gaussian noise-corrupted data but fail to do so for signal-dependent noise. For signal-dependent noise, an ad hoc procedure based on log-likelihood curves may identify the correct number of synergies \[[@B94]\]. For estimating the count of MS in comparison to the ad hoc procedure, *R*-squared curve or VAF (variance accounted for) curve gives a more accurate estimate. By definition, VAF and *R*-squared are similar (1 − sum of squared errors/total sum of errors), but in standard Pearson correlation, the total sum of squared errors for VAF is with respect to zero not the same with that for *R*-squared which is related to the mean. 4.1. Procedure (NNMF Multiplicative Update Method) {#sec4.1} -------------------------------------------------- For determining the applicable MS from raw unshuffled EMG data, *N* synergies (1--10) were extracted from EMG data set with *n* muscle and *m* samples ([Figure 3](#fig3){ref-type="fig"}). The algorithm estimates *W* and *C*, the EMG data is shuffled and fed to the algorithm again with earlier extracted synergies as fixed synergies, and the activation coefficient (*C*) is kept unfixed and allowed to be estimated. The EMG data is reconstructed again by multiplying the newly estimated activation coefficient vector and fixed synergy vector *E* = **W** × **C**. For each synergy point, the data is reconstructed; then, *R*-squared or VAF curve is plotted against the *N* number of synergies (1--10) using the expression (1 − sum of squared errors/sum of total squared errors). $$\begin{matrix} {SSE = \sum\sum\left( {E_{nm} - {\hat{E}}_{nm}} \right)^{2},} \\ {SST = \sum\sum\left( {E_{nm} - {\overline{E}}_{nm}} \right)^{2}.} \\ \end{matrix}$$ Here, *n* is the number of muscles and *m* is the number of samples. There are four main ways to calculate the number of MS based on the VAF curve, which are as follows: \(1\) Best linear fit (BLF) method: Cheung presented that moving along with a greater number of synergies or components on the *x*-axis of the graph reduces the mean square error and the curve becomes a straight line. The point at which it attains the plateau is chosen for the correct number of synergies \[[@B60]\]. \(2\) Knee point (KP) method: Cheung et al. presented another method in which the point or the number is chosen where the curve has an increase of smaller than 75% \[[@B51]\]. \(3\) Elbow method: Tresch et al. suggested the point at which the change in the slope of curves is maximum \[[@B94]\]. \(4\) Threshold method: Torres-Oviedo et al. used a threshold basis of 0.9 (90%) on the VAF curve to find the number of synergies for extraction \[[@B58]\]. 4.2. Automated Task Decoding Method {#sec4.2} ----------------------------------- This method uses different decoding algorithms (linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), naive Bayesian (NB) algorithm, and K-NN clustering) and statistical testing. The number of synergies is chosen at a point where there is no statistical significance after adding more synergies to the decoding parameters of the algorithm. The statistical significance is computed between the decoding parameters of *N*, *N-1*, and *N*th synergy which are pseudo-randomly shuffled 100 times \[[@B103]\]. There is no perfect algorithm for MS estimation because with biomechanical constraints, the accuracy of the algorithms gets reduced \[[@B100]\]. In the classical VAF method, the variance causes difficulty in choosing the exact number of synergies. Delis et al. presented a more accurate method to extract a smaller set of synergies from the data \[[@B103]\]. In this method, the performance of synergy computation is dependent on the algorithm type as LDA performed faster and more precise using the dataset in their study. 5. MS as a Physiological Marker for Neurorehabilitation {#sec5} ======================================================= 5.1. Stroke {#sec5.1} ----------- Muscle synergy structures based on motor control outputs are beneficial for recognizing alterations in the brain for various motor tasks. Thus, MS prove to be important in the field of human locomotion and neurorehabilitation as motor impairments can be understood by means of the patterns or structures of MS \[[@B53], [@B106]--[@B108]\]. Cheung in their study in ischemic stroke patients with affected and unaffected arm found MS similarities in both arms with respect to different lesion sizes and locations in frontal cortical areas. This led to the conclusion that the cortical signals activate the muscles at both sites in a similar way \[[@B60]\]. However, due to cortical lesions, altered activations of muscles occur resulting in deficit motor performance. This altered coordination can be emphasized for stroke recovery. Alteration in the MS is dependent on the severity of impairment \[[@B107], [@B109]\]. Recent studies provided additional factors which affect the structure of MS or possibly the similarities in MS. In most cases, alteration in MS is observed, but the results by Cheung showed that preserved synergies could be due to hand movement with the intact sensorimotor cortex \[[@B60]\]. Merged synergies are associated with the abnormal coupling of joints in chronic stroke patients during different movements \[[@B52], [@B110]\]. These abnormal synergies, instead of being eliminated, can be augmented through robot-assisted therapy \[[@B53], [@B108], [@B111], [@B112]\]. We have emphasized the impact of MS with respect to the level of impairment or cortical lesion, but the alteration in MS can also result with time among stroke patients. There are three main stages of stroke categorized on the basis of duration by computed tomography (CT). ### 5.1.1. Subacute Stroke {#sec5.1.1} The time period for the stroke is from 48 hours to several weeks. The upper-limb MS in such stroke is very similar to that in healthy patients, as the previous study also revealed, but the neural drives or activation coefficients displayed alteration because of the presence of a cortical lesion or damage in the cerebral hemisphere \[[@B108]\]. Hashiguchi et al.\'s study on subacute stroke patients revealed both merging and fractionation of MS in the lower limb during gait; however, unlike addition of more synergies for good motor performance \[[@B52]\], the merging of synergies resulted in poor muscle coordination. It is possible that because of the short time span, new synergies were not added to the subspace for flexible control of the limb by stroke patients. ### 5.1.2. Acute Stroke {#sec5.1.2} The time period is very short which is less than 24--48 hours. Any voluntary movement is less likely. ### 5.1.3. Chronic Stroke {#sec5.1.3} The time period ranges from months to years. In chronic stroke, fractionated and merged MS were observed \[[@B109]\]. In such condition, MS are not usually conserved as it appeared in earlier studies. The alteration in proximal MS is more distinct in severe cases than in mild and moderate cases, although some synergies in mild, moderate, and severe stroke were conserved in elbow flexors and extensors but varied in shoulder muscles. The conserved, merged, and fractioned synergies exist together in stroke patients \[[@B52], [@B109]\]. The difference in the similarity of the MS could be due to the neuro-anatomical sites or intact sensorimotor cortex \[[@B106], [@B113]\]. With the intact sensorimotor cortex, MS similarity is poor with the newly generated MS. In contrast to this without the intact sensorimotor cortex, MS similarity is higher in chronic stroke \[[@B106]\]. Mcmorland et al. further added that the degree of similarity and preservation of synergies among individuals with intact sensorimotor cortex was positively associated with hand function \[[@B113]\]. Godlove found that MS after stroke are related to perilesional high gamma observded in the electrocorticography (ECoG) signals \[[@B114]\]. 5.2. Fugl-Meyer Assessment {#sec5.2} -------------------------- The studies discussed so far show consistency to the Brunnstrom approach that includes stages of motor recovery. The MS count remained static in pain with altered spatiotemporal patterns in the upper limb \[[@B115]\]. In stage 3, the pain is high because of muscle stiffness and it can be inferred that with decreasing spasticity stages, emergence of new MS or augmentation of affected MS for rehabilitation is likely to be achieved successfully. The following stages were implicated: (1) flaccid paralysis: there are no reflexes; (2) appearance of spasticity: basic MS start appearing and abnormal movement of limbs will be present; (3) increased spasticity: there is increased stiffness of the muscles and voluntary movements can be attained, but motor control is absent; training of muscles or MS in this stage can lead to preliminary recovery; (4) decreased spasticity: motor control starts appearing and training should be continued; (5) complex movement combination: here, the MS patterns get more coordinated and complex movement can be attained by the limbs; (6) disappearance of spasticity: with the disappearance of spasticity, increasing motor control is achieved; difficulty during rapid complex movements remains and MS are more coordinated; and (7) normal functioning: normal functioning of the limbs as optimal control is attained. Based on the above approach, Brunnstrom and Fugl-Meyer et al. developed the scoring system for the quantitative clinical assessment of five domains for rehabilitation \[[@B116], [@B117]\]. The five domains of the assessment are the motor function, sensory function, joint pain, joint motion, and balance \[[@B116], [@B117]\]. 5.3. Rehabilitation {#sec5.3} ------------------- In patients with a neurological disorder, the imbalance of muscle coordination is compensated by other muscle groups. For example, in chronic stroke, trunk synergies are compensatory to overcome aberrant coupling in the arm for better control \[[@B112], [@B118]\]. Similar compensatory synergies have been observed in lower limbs during gait \[[@B119]\]. These compensatory synergies inhibit the recovery process of specific MS as it becomes hard to train the specific muscle groups to recover. Thus, specific muscle synergies should be trained and any compensatory muscle synergies should be inhibited during the training period depeding on the level of impairment among patients \[[@B112]\]. In contrast, studies also suggest that this compensatory behavior should not be ignored but rather be utilized for better poststroke recovery \[[@B119], [@B120]\]. Fractionated MS are observed in poststroke individuals after several years. It can become the flexible control strategy for movement as synergies are added. The formation of new synergies, which were evident in chronic stroke, can be useful for neurorehabilitation as these synergies were adapted for improved controllability of limbs \[[@B51], [@B121]\]. A minimum of six-week repetitive training can induce changes in the white matter to build new or augment existing synergies \[[@B53], [@B106], [@B121]\]. In the lower limb, while standing, stroke-affected patients and healthy individuals exhibited common synergies, since the stroke patients shifted the center of mass to maintain balance and their MS were similar to those of a healthy person \[[@B48], [@B122], [@B123]\]. In most studies, NNMF (multiplicative method) is used to study the synergies among stroke patients \[[@B51], [@B52], [@B107], [@B109]\]. Different methodological techniques for synergy extraction can provide better information on a trial-by-trial basis or on an individual basis for better recovery from stroke. 6. MS Application in Sports {#sec6} =========================== Muscle synergies have been analyzed thus far during normal physical activities, but during peak physical activities, the role of MS with respect to the task is also of great interest. The distribution of relative weight of the muscles in the synergy space helps in understanding the movements. Modulating compensatory synergies that minimizes the chances of injury by specific training is a better solution in sports rehabilitation and performance. Matsunaga et al. analyzed the MS before and after 10 minutes of running \[[@B124]\]. Similar to previous studies, the number of MS was consistent. The first three modules or MS were similar, but the fourth MS showed the activation of muscles around the pelvis region moving from the trunk region. This could be the reason that during foot strike, we are more prone to injury as the postural control has been shifted from the trunk to the lower limb. Some sports like gymnastics require high postural stability. Trunk muscles play a vital role in human postural control \[[@B112], [@B118], [@B124]\]. Among gymnasts, while performing a giant swing, three synergies were identified \[[@B55]\]. The first two synergies were consistent between subjects, but the third synergy displayed variability. Frère and Hug concluded that the variability in the third synergy could be related to the lower-level neural control rather than to the biomechanical constraints. Here, the synergies associated with the trunk, arms, and shoulder muscles cooperate to limit further the extension of the shoulder joint \[[@B55]\], thus reducing the chances of injury. A recent research showed that after five weeks of strength training (bench press), there was an intrasubject variability of MS \[[@B125]\], whereas those who had not trained themselves and continued with their daily routine showed no such variability \[[@B125]\]. Shaharudin et al. analyzed a group of untrained individuals performing rowing movements on slide and fixed ergometers \[[@B126]\]. The results were consistent with those of previous studies. Muscle synergies were modulated, but their structure remained the same \[[@B59], [@B126]\]. The CNS had distributed the weights of the muscles differently for rowing on a slide ergometer (leg muscles) and a fixed ergometer (back muscles) to minimize injuries. We can say that the nervous system continuously makes adjustments to optimize the selection of MS that are best suited for the specific movement. It is reasonable to assume that this optimization of the MS selection process is based on identifying and utilizing the most economical MS profile for the movement and one that reduces the chances of injury. Barroso\'s research on MS stated that functional motor impairment (spinal cord injury) quantitative assessment was performed well while cycling. Muscle synergy modulation plays a vital role in sports performance and, in the future, can become a vital tool in recovery from injury \[[@B55], [@B124], [@B126]\]. 7. MS Application in Robotics {#sec7} ============================= In this section, we will discuss the application of MS in robotics. A low-dimensional controller controls the dynamics of the limbs without losing the performance. A number of studies have utilized the hypothesis of MS to build artificial limbs \[[@B101], [@B102], [@B127]\]. Berniker et al. built a simple controller with MS and a low-dimensional model whose performance is close to a full-dimensional controller \[[@B61]\]. Alessandro et al. in his review presented the principal foundation of using the hypothesis of MS in robotics \[[@B95]\]. The synergies extracted from the task are combined to form the motor signal for a newtask and were tested further based on the observed task \[[@B95]\]. Including the dynamics of the musculoskeletal model can help better approximate muscular activities. Rasool et al. examined the state space modeling, used the MS matrix extracted from EMG data as an observation matrix, and included the dynamical model of the upper extremity \[[@B102], [@B128]\]. The neural drive was further estimated using an updated state-constrained Kalman filter that has a synergy matrix and the upper-extremity dynamic matrix, which can be used to control upper-extremity myoelectric prostheses \[[@B102], [@B128]\]. Afzal et al. did similar work on the lower limb during overground movements \[[@B101]\]. Direct and pure kinematic modeling of the synergy patterns could result in inconsistency in grasping \[[@B129]\]. Pisa/Italian Institute of Technology modeled a robotic hand with the concept of translating soft synergies into adaptive synergies to solve this issue. To implement such model, underactuated hands with designed ligaments and innovative joints were used. The underactuated hand\'s parameters were selected to mimic the given synergies for consistent grasping \[[@B127], [@B129]\]. There are currently many improvements under way, and dynamical systems have been developed for movements that are more complex. This will lead to the addition of more flexible movements in robots with higher controllability. 8. Conclusions {#sec8} ============== We have presented recent research results related to MS in motor control, neurorehabilitation, robotics, and sports science. This review paper concludes that the modular organization of MS in the CNS and their combination lead to a variety of natural motor behaviors. These predefined encoded primitives reduce the dimensionality of the behavior for better task control \[[@B23], [@B45], [@B53], [@B54], [@B57], [@B58], [@B61], [@B130]\]. The concept of MS is still under study with critical and unbiased views towards it \[[@B13], [@B17], [@B59], [@B64], [@B104], [@B131]\]. We have postulated that the MS are represented as spatiotemporal synergies in the CNS and are triggered individually as spatial and temporal patterns by the CNS \[[@B43], [@B47], [@B49], [@B88]\]. The linear decomposition algorithms (PCA, NNMF, and ICA) are predominantly used to extract spatiotemporal, temporal, and spatial synergies from EMG. In the fields of neuroscience and robotics, the MS hypothesis has proven to be efficient for motor control of the limbs by reducing the degrees of freedom \[[@B101], [@B102], [@B127], [@B132]\]. Temporal synergies are crucial in understanding the neural basis of MS \[[@B43]\]. The case for the neural origin of MS is still a question of debate, but the evidence shows a strong inclination towards its neural-physiological origin as well as spatiotemporal pattern representation in the brain and the SC. We have also concluded that the alteration in MS is dependent on the site of lesions, the severity of impairment, the stage of stroke, and, to a certain extent, the complexity of the task performed by the patients. As MS can be preserved, fractionated, and merged among stroke patients, this makes it a physiological marker for neurorehabilitation. The emergence of new MS and augmentation with robot-assisted therapy and sports and exercise therapy provide a relation to changes in the white matter and neuroplasticity \[[@B133]\]. The MS can also be used to assess the distribution of the muscle weights during movements that are more vigorous. The abnormal shifting of the activation of the muscles can be observed from MS patterns and hence can help us in reducing the chances of injury \[[@B124]\]. Besides stroke, other neurological disorders like cerebral palsy, dystonia, and spinal injury have been investigated by the hypothesis of MS, which allow flexibility in neurorehabilitation or diagnosis in these and other disorders \[[@B8], [@B134]--[@B138]\]. The association of MS and CPG provides a new approach for myoelectric prostheses among arm and leg amputees by reducing the degrees of freedom \[[@B101], [@B102]\]. Instead of considering MS as implemented in software, there are some applications in robotics where a hardware model has been built around an MS concept \[[@B127], [@B129]\]. Most of the MS studies utilize different algorithms, and to this date, no single approach to optimally process the EMG signal and extract MS has been identified. Studies have revealed that different algorithms responded differently to noise in the EMG source and were likely to affect the accuracy or precision of the results. There is still considerable room for further research related to MS in the field of neuroscience, robotics, and sports. This review work was made possible by the funding support from the UA at Little Rock Engineering and Information Technology (EIT) and the UA at Little Rock School of Counseling, Human Performance and Rehabilitation. Conflicts of Interest ===================== There is no conflict of interest from any author for publishing this article. {#fig1} ![CPG timer circuit in the absence of peripheral feedback coordinates the movements. The figure shows the neuromechanical tuning \[[@B78], [@B79]\]. The primary motor cortex region dictates the movement via the basal ganglia and thalamus to the MBLR which is a part of the brainstem where spatial--temporal patterns are encoded. Force and displacement are sensed by the muscle spindle and the Golgi sensory receptor through a feedback.](ABB2018-3615368.002){#fig2} {#fig3} ###### The spectrum of performance of different algorithms for synergy estimation from the EMG signal affected with Gaussian noise and signal-dependent noise performance of algorithm in the identification of the subspace and activation coefficients \[[@B94]\]. Performance Gaussian variance noise (synergy estimation) Signal-dependent noise (synergy estimation) Identifying subspace Activation coefficient ------------- ---------------------------------------------- --------------------------------------------- ---------------------- ------------------------ PCA Low Low High Intermediate ICA High Intermediate Low Low FA High Intermediate High High NNMF Intermediate Intermediate Intermediate Intermediate ICAPCA High High High High pICA High High High High [^1]: Academic Editor: Panagiotis K. Artemiadis

Introduction {#s1} ============ The sequencing of the human genome \[[@pgen-0030099-b001],[@pgen-0030099-b002]\] and subsequent work describing sequence variation amongst human populations \[[@pgen-0030099-b003]\] has provided the necessary resources for large-scale studies of the effects of genetic variation on human gene expression. Identifying functionally important variation has the potential for increasing understanding of gene regulation and for providing efficient markers to study the effects of variation in gene expression on human disease risk \[[@pgen-0030099-b004]\]. Experimentally demonstrating the potential functional effects of DNA polymorphism is difficult, as these effects may be both tissue and stimulus specific. Significant efforts have focused on transcriptional regulation, because of the strong suspicion that the majority of human phenotypic variation is due to regulatory variants \[[@pgen-0030099-b005],[@pgen-0030099-b006]\]. Novel allele-specific transcript quantification approaches to candidate genes \[[@pgen-0030099-b007],[@pgen-0030099-b008]\] have been employed, along with broader approaches to investigate the absolute levels of expression of thousands of genes \[[@pgen-0030099-b009],[@pgen-0030099-b010]\]. Using these methods, several *cis*-acting SNPs that correlate with gene expression have been identified. However, fine mapping these effects and determining the mechanisms underlying the associations has been more difficult \[[@pgen-0030099-b011]\]. In this study, we used a different approach---that of evaluating effects on splicing efficiency---to study the effects of common genetic polymorphism on gene function. The vast majority of human genes are comprised of three or more exons that need to be efficiently spliced together to form mature mRNA. Variation in this process occurs naturally and is thought to be an important mechanism whereby different protein products can be derived from the same gene sequence \[[@pgen-0030099-b012]\]. Single base changes that affect splicing can have dramatic effects on gene function and can cause disease, usually because the splice mutation results in a shift in the amino acid reading frame. Most commonly observed alternative splicing events preserve the reading frame and have more subtle effects on protein function \[[@pgen-0030099-b013]\]. There are an increasing number of examples in which the genetically determined modulation of alternative splicing has been implicated in common complex disease traits, such as the associations between the G protein-coupled receptor *(GPRA)* and asthma susceptibility \[[@pgen-0030099-b014]\]*,* cytotoxic T lymphocyte antigen 4 *(CTLA4)* and autoimmune disease \[[@pgen-0030099-b015]\], and the CD45 (leucocyte common) antigen and infectious and autoimmune diseases \[[@pgen-0030099-b016],[@pgen-0030099-b017]\]. The potential effects of common SNPs on splicing isoforms have been suggested by bioinformatic analysis of expressed sequence tags \[[@pgen-0030099-b018]\]. In a small number of genes, these potential effects have been demonstrated experimentally \[[@pgen-0030099-b019]--[@pgen-0030099-b021]\]. Here, we used lymphoblastoid cell lines (LCLs) from the Centre d\'Etude du Polymorphisme Humain (CEPH) as an experimental model system to investigate the relationship between variation in simple cassette exon splicing events and genotypic diversity. We sought to determine (1) whether individual variation in splicing patterns was commonly observed, (2) if any observed phenotypic variation could be explained by genetic differences among individuals, and (3) whether any genetic differences could be localised and the functional element identified. Results {#s2} ======= Inter-individual Variation in Splice Pattern {#s2a} -------------------------------------------- Our initial aim was to investigate whether there was variation among individual LCLs in simple cassette exon events. These events were defined as the occurrence of complete exon skipping in two or more mRNA isoforms. We used a strategy of exon selection that we believe increased the likelihood of detecting allele-specific effects on alternative splicing. We argue that for genes in which common SNPs affect splicing, at least two mRNA transcript isoforms of that gene will be relatively commonly observed. Conversely, where only one transcript isoform has been observed and documented, the likelihood of a SNP-related splicing event is reduced. We identified 2,281 simple cassette exon events from the European Bioinformatics Institute Alternative Splicing Database (EBI-ASD) in which each transcript isoform had been observed in at least two clone libraries. From these, we selected the 250 genes with the highest expression levels in LCLs as detected by global microarray analysis. We carried out reverse transcriptase PCR (RT-PCR) analysis of these 250 genes and found that in LCLs both transcript isoforms were present in 70 (28%) of the genes. We proceeded to investigate whether the amount of different isoforms varied between 22 different LCLs. Of the 70 events that produced both full-length and exon-skipped products, we found that 18 (26%) showed significant variation among cell lines, in which at least one cell line showed a ratio of PCR products that differed by more than 10% of the mean value for the entire sample set of 22 cell lines (10% difference in relative abundance is the lower limit of sensitivity of the detection assay). These 18 events were retested using RNA derived from an independent round of cell culture. Six events, centered around genes *CASP3, CD46, IFI16, RBM23, SH3YL1,* and *ZDHHC6,* demonstrated repeatable and consistent variation of the splicing pattern among different cell lines. The genes and exons for each of these six events are listed in [Table 1](#pgen-0030099-t001){ref-type="table"}. None of the skipped exons resulted in a shift in the reading frame of the mRNA. We did not investigate the remaining 12 events; these provided inconsistent results, as splicing isoforms were present only at very low intensity or in only one or two cell lines. ###### Details of the Alternatively Spliced Exons and Associated SNPs  SNP Genotype Predicts Splice Pattern {#s2b} ------------------------------------ We next investigated the relationship between DNA sequence variation and observed differences in splice isoforms among LCLs. We looked at the correlation between SNP genotype and splicing pattern over the 500-kb region surrounding each of the six splicing events that showed consistent variation among the LCLs. Two sources of SNP genotyping data were used. First, we analysed SNP genotypes from the International HapMap Project \[[@pgen-0030099-b003]\]. Second, we resequenced the skipped exons and 150 bp of the flanking introns for each event in each of the 22 cell lines. Resequencing did not identify any SNPs that were not already identified on the HapMap resource. Approximately 350 SNPs were available for each gene at an average density of 0.7 SNPs per kb. For each of the six events, highly significant correlations between SNPs and the observed splicing pattern were identified ([Figure 1](#pgen-0030099-g001){ref-type="fig"}). The maximum values for the Pearson\'s statistic were 0.76 (*p* \< 10^−4^) for the *CASP3* event and over 0.86 (*p* \< 10^−6^) for the other five events. For five of the six events, the SNP nearest the intron--exon boundary showed the strongest correlation with splicing pattern. For the *ZDHHC6* event, a slightly higher value for the Pearson\'s statistic was seen for a group of three SNPs lying over 50 kb away from the gene and in very strong linkage disequilibrium (LD) (R^2^ = 0.95) with the SNP nearest the intron--exon boundary. When we studied the *ZDHHC6* SNP nearest the intron--exon boundary in the minigene system (see below), we observed a direct effect of this SNP on splicing efficiency, suggesting that this SNP, rather than the more distant group of three SNPs, was responsible for the observed variation in splice pattern. To test whether the identified SNP accurately predicted the variation in splice pattern, we selected a new set of nine unrelated LCLs in which there was at least one example of each of the possible SNP genotypes. For each of the six genes, the splicing pattern observed was accurately predicted by the SNP genotype ([Figure S1](#pgen-0030099-sg001){ref-type="supplementary-material"}). {ref-type="table"}). Each of the six genes is between 15 and 50 kb in size.](pgen.0030099.g001){#pgen-0030099-g001} The correlations between splice pattern and individual SNPs are highly significant even after allowing for correction for multiple comparisons. If we use a simple Bonferroni correction for the 350 SNPs that were tested for each simple cassette exon event, all results remain significant at the 0.05 level. This level of correction is overly conservative, since the LD relationship among the SNPs means that they are not independent of one another. Furthermore, it is remarkable that for five of the six events it is the SNP closest to the intron--exon boundary that is the strongest predictor of splicing phenotype. When we analysed the effects of the SNP nearest the intron--exon boundary of each event, a clear effect of genotype on relative abundance of each product was found. The measured ratios of the two splice products are plotted by genotype in [Figure 2](#pgen-0030099-g002){ref-type="fig"}. The magnitude of allele-specific effect was similar for each gene and represented an approximately a 2-fold additive effect on the ratio of splice isoforms. In four out of the six events, the minor allele was associated with an increased abundance of mRNA with the exon-skipping event. For the other two (*RMB23* and *ZDHHC6*), the minor allele was associated with an increased abundance of the full length mRNA. These data suggest that *cis*-acting variation is directly modulating the pattern of observed alternative splicing at these loci. {#pgen-0030099-g002} For five of the six events, there is an apparent dose-dependent effect with larger effects seen in homozygotes compared with heterozygotes. For the *CASP3* event, the effect of the SNP on relative abundance of the two transcript isoforms was only seen in the homozygous state. The effect of the *CASP3* SNP in the homozygous state was nevertheless clear-cut and repeatable. We were puzzled as to why we were unable to detect an effect when the SNP was present on only one chromosome since a *cis*-acting mechanism of action seems most likely. One possible explanation is that when only one chromosome carries the splicing SNP, up-regulation of expression from the other chromosome compensates for the loss of the full-length product. To test this hypothesis, we quantified the relative abundance of *CASP3* transcripts derived from each chromosome in LCLs from 16 unrelated CEPH individuals heterozygous for the rs4647603 *CASP3* exonic SNP. These experiments showed that all 16 heterozygous individuals had a higher relative abundance of transcripts containing the rs4647603 G allele compared to those with the A allele (on average 2.7 times more G than A, [Figure S2](#pgen-0030099-sg002){ref-type="supplementary-material"}). In the homozygous state, the A allele is associated with increased exon skipping. In the heterozygous state, the relative proportions of the full-length and exon-skipped products appear unchanged. The higher relative abundance of *CASP3* transcripts containing the rs4647603 G allele suggests increased expression of *CASP3* derived from this chromosome and is consistent with an effect of the rs4647603 A allele on splicing in the heterozygote state. Splice Site Analysis {#s2c} -------------------- Splice site signal scores from the donor and acceptor sites of the test exons predicted to show alternative splicing were compared with those from a genome-wide set of constitutively spliced exons ([Figure 3](#pgen-0030099-g003){ref-type="fig"}). As a group, the test exons had significantly (*p* = 1 × 10^−5^) weaker splice site signal scores than those from constitutive exons. The difference was greatest for the exons in which alternative splicing was experimentally demonstrated. The effect was seen in both donor and acceptor sites and was slightly more pronounced at the donor sites. Although the differences in splice site strength were statistically significant, there was extensive overlap in splice site scores between the groups ([Figure 3](#pgen-0030099-g003){ref-type="fig"}). {#pgen-0030099-g003} The potential effects on exonic splice enhancer strength of the four exonic SNPs shown to correlate with splice pattern were tested using four different prediction algorithms (see [Materials and Methods](#s4){ref-type="sec"}). For two SNPs, no effects were predicted by any of the four models tested. For the other two SNPs, the results were contradictory (different models showed both increased and decreased splice enhancer activity). There were no differences in the number of SNPs in the 50-bp regions around the intron--exon junction for the 180 exons that did not show alternative splicing in our experimental model, compared to the 70 exons that did. This suggests that using the position of known SNPs to select for exons with allele-specific splicing patterns is unlikely to be fruitful. Furthermore, of the six exons that showed allele-specific splicing patterns, three showed splicing patterns that were correlated with SNPs situated more than 50 bases from the intron--exon junctions. Minigene Analysis Confirms Modulation of Splicing by SNP Genotype {#s2d} ----------------------------------------------------------------- To investigate whether SNP genotype directly defined splice isoform pattern, we carried out minigene analysis in two genes. In *ZDHHC6,* the test SNP was situated in the middle of the exon, 99 bp away from the intron--exon boundary. In *SH3YL1,* the test SNP was also exonic, but in this case only 2 bp away from the intron--exon boundary. For each gene, we independently cloned two fragments that differed only by the alleles of the SNP correlated with exon skipping. Each fragment consisted of the alternatively spliced exon plus 180 bp of intronic sequence on each side. The fragments were inserted into a minigene splicing vector that was used to transfect HEK293T cells. After 48 h, mRNA was extracted from the cells and the relative abundance of mRNA (full length and alternative spliced) transcripts derived from the minigene plasmid was determined. For both genes we observed that the SNP allele associated with increased exon skipping in the LCL experiments was also associated with increased exon skipping in the minigene system ([Figure 4](#pgen-0030099-g004){ref-type="fig"}). For four genes (*SH3YL1, ZDHHC6, IFI16,* and *RNPC4*) there were between four and ten SNPs identified on the HAPMAP resource that were in complete LD with the SNP nearest the intron--exon boundary. All but one of these SNPs lay more than 2 kb away from the exon of interest and are not testable using the minigene system. We cannot discount an effect of these more distant SNPs on the observed allele-specific splicing event. {#pgen-0030099-g004} Discussion {#s3} ========== This study describes reproducible phenotypic variation in splicing among individuals, in each case arising from a simple cassette exon event that is associated with genotypic variation in SNPs close to the corresponding intron--exon boundaries. Our starting point was to screen for phenotypic variation in splicing in 22 lymphoblastoid cell lines, and then to identify SNPs associated with this phenotypic variation. Interestingly, the splicing-associated SNPs identified experimentally in this study did not show any clear difference in position or sequence context from other SNPs that were not associated with splicing variation. The mechanisms by which alternative splicing is regulated are poorly understood. Exon recognition and splicing requires the presence of basic "classic" splice sites (the branch point, polypyrimidine tract, and the 3′ and 5′ splice sites). The efficiency of the splicing process can be affected in some exons by the presence of auxiliary or modulating elements ([Figure 5](#pgen-0030099-g005){ref-type="fig"}). The consensus sequences for the known modulating elements are degenerate and frequently found throughout the genome. DNA sequence variation can modulate alternative splicing, and to date attention has focused on disease-causing *cis*-acting mutations affecting the use of constitutive and alternative splice sites, together with *trans*-acting variants that affect the basal splicing machinery and factors regulating splicing \[[@pgen-0030099-b022]\]. In contrast to mechanistic studies of disease process, our study started from the premise of defining simple cassette exon events in which there was significant variation among a panel of LCLs and relating this to genotypic diversity. We found consistent variation in six out of 70 simple cassette exon events, and for each of these six events we found a clear relationship between genotype and splice phenotype. Analysis of SNPs typed by the International HapMap project and those derived experimentally by resequencing showed that the SNPs with the strongest correlation were those closest to the intron--exon boundaries of the splicing events. For two of the SNPs we carried out minigene experiments, and both showed *cis-*acting effects on gene splicing using this system. {#pgen-0030099-g005} It is perhaps not surprising that we were unable to detect any specific patterns in the sequence context of the six SNPs identified in this study, given the apparent degenerate nature of consensus sequences that bind splice modulator proteins. Overall the splice-site strengths of the exons that were predicted to be skipped by the EBI-ASD database were weaker than those of constitutive exons, and those that we were able to demonstrate to have alternative splicing in our experimental system had the weakest splice site strength. However, there was significant overlap among the groups, and splice site strength cannot be used to identify the most likely exons to study. Equally, the presence of SNPs close to the intron--exon boundaries did not differ between those exons that did and did not show alternative splicing, suggesting that selecting exons to study according to whether there is a "splice site SNP" (defined for example on Ensembl as a SNP lying within 10 bp of the intron--exon junction) will not enrich for those SNPs that actually affect the splicing process. Only two out of the six SNPs identified in this study were within 10 bp of an intron--exon junction. The exonic SNPs that correlated with splice pattern in this study showed no consistent effects on splice enhancer strength using four different predictive models. Thus, the sequence context or position of the SNPs would not identify those likely to influence splicing efficiency. A different approach to identify allele-specific alternative splicing events that does not rely on the sequence context or the position of SNPs is to identify allele-specific RNA isoforms from EST databases \[[@pgen-0030099-b018]\]. This approach requires the presence of an exonic SNP not involved in the alternative splicing event to be in high LD with the functional splicing SNP and limits its broad applicability. When applied to our data using HAPMAP SNPs, only the events in *ZDHHC6* and *RBM23* have the potential of being identified. The EST method is prone to false-positive results, particularly for low frequency SNPs, if there are insufficient representative ESTs available in the database. We suggest that an experimental approach, rather than a bioinformatic approach, will be necessary to identify splicing phenotype-associated SNPs, at least until more is learned about how these SNPs exert their functional effects. We believe that identifying alternative splicing events is the essential first step in this experimental approach. While splice enhancers and suppressors are found in constitutive exons and their flanking introns, these exons by definition are not observed to show alternative splicing. This suggests either that no SNPs occur within functionally important splice elements or that the splice enhancer/suppressor signals are not required for the accurate splicing of these exons. The relative positions and sequence context of experimentally identified splicing SNPs can be used to refine predictive algorithms and may provide new insights into which of the many exonic and intronic splice modulator sequences present in every gene are functionally important in regulating the splicing process. Our method of isoform quantification and pooling strategy meant that our ability to detect rare events was limited. Dilution experiments determined that both the full-length and exon-skipped transcript products were detectable even when their starting concentrations differed by 100-fold. Thus, provided that both transcripts were present in at least one of the 22 cell lines, and the minor transcript was present at an abundance of 30% or greater, the event would be detected. If the rare transcript was present in three or more cell lines, the sensitivity increased to a lower abundance of 10%. The method we used is not readily scalable to whole genome analysis. Microarray-based approaches to the analysis of alternative splicing have been published \[[@pgen-0030099-b023],[@pgen-0030099-b024]\]. These approaches can analyse the splicing patterns of many thousands of exons and have been used to distinguish splicing patterns seen in different tissues. Interpretation is complex, and for some arrays sensitivity is low and false positive rates are high. Although it is likely that the technology will improve, these approaches have not yet been shown to have the sensitivity to detect the level of variation we observed in this study, particularly for low-abundance isoforms. The advantage of the system we describe is targeted amplification of the splicing event of interest, which we believe provides greater sensitivity. Nevertheless, use of an array-based approach is likely to become the most efficient method to identify allele-specific splicing effects at a whole genome level. For the splicing phenotypes, our experiments using the minigene system suggest that the SNP closest to the intron--exon boundary that shows correlation with the splicing phenotype is very likely to be the functional element. For four of the genes in this study there were additional SNPs in complete LD with the SNP nearest the intron--exon boundary, and although most were over 2 kb away from the exon-skipping event it is possible that the presence of these SNPs influence the splicing process. Further work is needed to define the consequences of the loss of these exons on the functional activities of the encoded protein isoforms and in the levels of expression. There is already evidence that biological consequences of the alternative splicing event we describe in *CD46* are likely to be important. CD46 is a cell-surface glycoprotein involved in regulation of complement activation and it acts as a receptor for several pathogens including measles virus, *Streptococcus pyogenes, Neisseria gonorrhea,* and Neisseria meningitidis \[[@pgen-0030099-b025]\]. CD46 is known to have two protein isoforms with distinct cytoplasmic tails of 16 or 23 amino acids generated by alternative splicing of exon 8 \[[@pgen-0030099-b026]\]. These different tails have pivotal effects on the intracellular precursor processing of, and signal transduction by, the CD46 protein \[[@pgen-0030099-b026]--[@pgen-0030099-b028]\]. We have demonstrated that the inclusion of exon 8 is strongly associated with the presence of a nearby SNP (rs2724374), and whether or not this SNP is directly functionally responsible for the pattern, rs2724374 is a genetic marker for what appears to be an important functional protein isoform. Variants of *CD46* have been associated with outcome in hemolytic uremic syndrome \[[@pgen-0030099-b029]\], but genetic association studies using the rs2724374 SNP have not been reported. For *CASP3* we have shown that, in individuals who are heterozygous for the splicing SNP rs4647603, there appears to be compensatory upregulation of expression of the full-length *CASP3* isoform derived from the other chromosome. This suggests some functionally important difference between the two *CASP3* isoforms. The consequences of the allele-specific splice events we have defined are summarised in [Table 2](#pgen-0030099-t002){ref-type="table"}. ###### Biological Consequences of Identified Allele-Specific Alternative Splicing Events  In this study we focused on only one form of splicing variation in a relatively small number of genes. Larger-scale whole genome studies investigating additional splicing patterns, such as alternative donor and acceptor sites, will be needed to determine the extent of SNP-associated splicing phenotypes. Our findings raise the possibility that SNP effects on splicing may be at least as prevalent in the genome as those on overall gene expression \[[@pgen-0030099-b011]\]. SNPs that predict splicing phenotypes are likely to be important markers to include in genetic association studies of complex diseases. Materials and Methods {#s4} ===================== Exon selection. {#s4a} --------------- A number of different publicly available databases of observed mRNA transcripts are available. We used the EBI-ASD ([www.ebi.ac.uk/asd](http://www.ebi.ac.uk/asd)), which is a database of computationally delineated alternative splice events derived from alignments of expressed sequence tags or cDNA sequences with the corresponding genomic sequences for each gene. Using this resource, we identified transcripts where at least two isoforms are detected in which complete exons (called simple cassette exon events on the EBI-ASD) are skipped. These events generally result in transcripts that differ sufficiently in size to be readily distinguished by simple agarose electrophoresis. Primers were designed in the flanking exons, and product sizes for the full length and exon-skipped products were calculated. Cell lines. {#s4b} ----------- LCLs from 22 unrelated CEPH individuals selected from the HapMap collection were obtained from the Coriell Institute for Medical Research. Cells were cultured at 37 °C in a 5% CO~2~ environment using RPMI 1640 cell culture medium with 10% fetal calf serum, 200 mM L-glutamine, penicillin, and streptomycin. Cell density was maintained between 200,000 and 800,000 cells/ml. DNA and RNA were each extracted from 10 million cell aliquots. Constitutive expression levels in CEPH cell lines were defined for pooled RNA from four LCLs using an Affymetrix human U133A expression microarray (Affymetrix, <http://www.affymetrix.com>). RNA and cDNA synthesis. {#s4c} ----------------------- RNA was extracted from cell pellets using TRIREAGENT (Sigma-Aldrich, <http://www.sigmaaldrich.com>), chloroform, isopropanol, and ethanol precipitation. Total RNA was quantified using UV spectrophotometry. mRNA was extracted from 20 μg of total RNA aliquots using the Dynabeads mRNA purification kit (Invitrogen, <http://www.invitrogen.com>), and was cDNA synthesised using Stratascript reverse transcriptase (Stratagene, <http://www.stratagene.com>) with oligo(dT) primers. 1 μl of cDNA was derived from 100 ng of total RNA. Parallel reverse transcriptase negative controls were generated in all cDNA syntheses. PCRs were carried out at standard conditions (30 cycles, melting at 94 °C, annealing at 58 °C, and extension at 72 °C, each for 30 s) using BioTaq DNA polymerase (Bioline, <http://www.bioline.com>). Primers were designed in the exons flanking the simple cassette exon event. Two products of different lengths were predicted to be amplified, one including the cassette exon and a shorter product lacking the cassette exon. The products were resolved on 2% agarose gels. Detecting variation among samples. {#s4d} ---------------------------------- Pooled cDNA from all 22 cell lines was used to test each set of primer pairs. Identification of the expected full length product and the shorter product lacking the cassette exon (and no other products) was used to confirm that the predicted alternative splicing event was detectable in our experimental system. Primer sets showing the two expected RT-PCR products were subsequently taken forward to determine if there was variation in the proportion of the two products among different individual cell lines. Detection of variation among cell lines was carried by performing RT-PCR on RNA from each cell line separately. For each cell line, the relative amount of each of the two RT-PCR products (representing the full length and skipped mRNA) was quantified using image analysis of the products visualised on ethidium bromide gels (ImageQuant software; Amersham Biosciences, <http://www4.gelifesciences.com>). Since both RT-PCR products were amplified by the same primer sets, the RT-PCR was truly competitive, allowing the accurate determination of their relative abundance \[[@pgen-0030099-b030]\]. To assess the robustness of the ethidium bromide--based quantification method it was compared with quantification using a fluorescence-based technique. Primer pairs with a 5′ FAM modification were used to amplify exon-skipping events from six different genes using cDNA from nine different LCLs. The amplicons ranged in size from 137 bp to 617 bp. The relative amounts of PCR products for each of the 54 reactions were quantified using GeneScan software (Amersham Biosciences). The ratios of quantified products from this method showed excellent correlation with those derived from image quantification of ethidium bromide--stained gels (correlation coefficient 0.93). We determined the sensitivity of the ethidium bromide quantification system using known starting concentrations of DNA fragments of different lengths, and then quantifying the resulting amplicons. We were able to show that over a range of different product signal intensities, differences in the ratios of the different sized starting material of 10% or greater could be detected reliably ([Figure S3](#pgen-0030099-sg003){ref-type="supplementary-material"}). The sensitivity of this method was independent of cycle number. Each of the 22 cell line samples was assayed in duplicate. The mean ratio of the abundance of the two RT-PCR products from each primer set was calculated for each cell line. When the relative abundance for an individual cell line differed by more than 10% from the average value for the full set of 22 samples, the experiment was repeated using a fresh aliquot of cell culture material. Those events that gave consistent differences in the repeat analysis were then analysed further. SNP identification. {#s4e} ------------------- Genotypes for SNPs positioned within 250 kb on either side of the exon-skipping event were downloaded from the International HapMap Project Web site (<http://www.hapmap.org>) \[[@pgen-0030099-b003]\]. For each event with reproducible variation, we also resequenced the skipped exon and 150 base pairs of the flanking introns to determine if additional SNPs close to the event could be identified, using DNA from each of the 22 LCLs. Sequencing was carried out using purified PCR products generated with M13-tagged primers. Genotype correlation analysis. {#s4f} ------------------------------ For each splicing event with reproducible variation, we calculated Pearson\'s correlation between the ratio of band intensities for the two RT-PCR products and the SNP genotype. In this analysis, we have assumed that any functional SNPs will be *cis*-acting, and thus expect to see an effect that is more pronounced in homozygotes than in heterozygotes. Thus, for the purposes of the correlation analysis, genotypes were coded 1, 2, and 3 to represent the genotypes AA, Ab, and bb, where A represents the major allele and b represents the minor allele. The value of Pearson\'s correlation was determined for each SNP in each 500-kb region. Splice site analysis. {#s4g} --------------------- Splice donor and acceptor sequences were scored using a position specific score matrix (PSSM) method \[[@pgen-0030099-b031]\]. Alignments of mRNA and EST sequences to the reference human genomic assembly (version: hg17) were taken from the University of California Santa Cruz genome database (<http://genome.ucsc.edu>) and used to define a population of well-supported (appearing in more than nine transcripts) constitutive splice sites. These were used to train the PSSMs, considering three exonic, six intronic nucleotides at the splice donor, and three exonic, 18 intronic nucleotides for the splice acceptor PSSM \[[@pgen-0030099-b031]\]. We compared the splice site scores of the 250 exons predicted by EBI-ASD to show exon skipping (subdivided into those that showed exon skipping in our experimental model and those that did not) with the splice site scores of 7,431 exons that were always found in mRNA transcripts (constitutively present) randomly selected from the genome. We sought to determine if splice site strength could predict those exons that were likely to be skipped. We also sought to determine if the SNP density near to the intron--exon boundaries differed in those exons that showed alternative splicing compared to those that did not. Finally, the sequence context of SNPs correlated with specific splice patterns was analysed to determine whether they affected know splice enhancer or silencer elements, using four published algorithms: <http://ast.bioinfo.tau.ac.il/ESR.htm> \[[@pgen-0030099-b032]\], <http://genes.mit.edu/burgelab/rescue-ese> \[[@pgen-0030099-b033]\], <http://cubweb.biology.columbia.edu/pesx> \[[@pgen-0030099-b034]\], and <http://rulai.cshl.edu/tools/ESE> \[[@pgen-0030099-b035]\]. Minigene analysis. {#s4h} ------------------ Both allelic forms of the SNPs showing correlation with splice patterns in the *ZDHHC6* and *SH3YL1* genes were cloned into a minigene splicing vector (pALTER MAX modified splice vector, <http://www.promega.com>). Within this modified splicing vector, the multiple cloning site (MCS) of the conventional pALTER MAX minigene vector was replaced by an insert, so that the MCS falls within an intron instead of being within expressed sequence. The new insert contains the 5′ donor splice site from the human β-globin gene intron 1, the MCS, and the 3′ acceptor splice site from the intron of an immunoglobulin gene. When a PCR product with primers designed within introns is used, all donor and acceptor splice sites are present and thus the construct is spliced correctly. The fragments cloned consisted of the exon plus an average of 180 bases of flanking intron. All inserts were confirmed by fluorescent sequencing. HEK293T cells were transfected following the manufacturer\'s protocol (FuGENETM6, Boehringer Mannheim). Cells (3 × 10^5^) were transfected with 1 μg of DNA and were harvested after 48 h. The relative abundance of the full length and alternatively spliced mRNA derived from the plasmid was analysed using the same methodology as described for the CEPH cell RNA. Allele-specific transcript quantification. {#s4i} ------------------------------------------ Allele-specific differences in *CASP3* expression were determined using a transcribed marker polymorphism (rs4647603) in the exon of interest to distinguish the relative abundance of transcript containing this exon arising from the two alleles. Sixteen unrelated CEPH individuals heterozygous for the transcribed marker were selected from the HapMap collection and obtained from the Coriell repository. RNA and cDNA from each individual were prepared as described above. DNA was extracted using a purification kit (Blood and Cell Culture DNA, Nucleon BACC2; Tepnel, <http://www.tepnel.com>). For each individual, data were obtained from nine replicates from each of two independent cultures. Allele-specific transcript quantification was carried out by single nucleotide primer extension and MALTI-TOF analysis using a SpectroREADER MassArray (Sequenom, <http://www.sequenom.com>) mass spectrometer as described previously \[[@pgen-0030099-b007]\]. RNA ratio values were normalized with the ratios observed for genomic DNA. Supporting Information {#s5} ====================== ###### Relative Transcript Abundance Grouped by SNP Genotype for Nine Additional Unrelated Cell Lines The ratio of transcript abundance (skipped product/full length product) for each of the six alternative splicing events was accurately predicted by the SNP genotype. (355 KB DPF) ###### Click here for additional data file. ###### Allele-Specific Differences in *CASP3* Expression Allelic imbalances were determined using an exonic polymorphism to distinguish the relative abundance of transcript arising from the two alleles (G/A) in 16 unrelated CEPH heterozygous individuals. RNA ratios were normalized with the DNA ratios and the data plots represent the average from two independent experiments. Variability between biological replicas was small (mean of relative difference of 9%) (237 KB DPF) ###### Click here for additional data file. ###### Sensitivity of Detection Assay Relationship between the measured ratios of band intensity of 2 fragments of DNA after amplification using competitive PCR compared with ratios of the two fragments in the starting material. The two DNA templates were themselves PCR products of different sizes (250 and 463 bp) amplified with M13-tagged primers. These PCR products were diluted and quantified using the picogreen system. A range of different ratios of each of the starting templates was then generated by mixing different volumes together. The mixed samples were then amplified in a single reaction using the M13 primer set, generating two products of different lengths. The products were run out on agarose gels stained with ethidium bromide and visualised with ultraviolet light. Digital photographs of the images were quantified using ImageQuant software (Amersham Biosciences). Each point on the graph represents the mean of eight measurements for each ratio; the bars show 95% confidence intervals. The assay is designed to be sensitive to changes in relative abundance rather than to detect actual molar ratios. Thus, for example, an assay result showing a measured ratio of 3:1 compared with a known ratio of 1:1 does not affect the sensitivity of the assay to detect differences in actual starting concentrations. (257 KB DPF) ###### Click here for additional data file. Accession Numbers {#s5a} ----------------- The National Center for Biotechnology Information (NCBI) Entrez Gene (<http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene>) accession numbers for the genes discussed in this paper are *CASP3,* 836; *CD46,* 4179; *IFI16,* 3428; *RBM23,* 55147; *SH3YL1,* 26751; and *ZDHHC6,* 64429. Antonio Velayos-Baeza and Clotilde Levecque provided expert technical assistance for the minigene experiments. **Author contributions.** J. Hull and D. Kwiatkowski conceived and designed the experiments. J. Hull, S. Campino, K. Rowlands, and G. Elvidge performed the experiments. J. Hull, S. Campino, M.-S. Chan, R. R. Copley, M. S. Taylor, and G. Elvidge analyzed the data. J, Hull, S. Campino, K. Rockett, M.-S. Chan, R. R. Copley, M. S. Taylor, K. Rowlands, G. Elvidge, B. Keating, and J. Knight contributed/ reagents/materials/analysis tools. J. Hull, S. Campino, R. R. Copley, J. Knight, and D. Kwiatkowski wrote the paper. **Funding.** Susana Campino was supported by a fellowship from Fundacao para a Ciencia e Tecnologia, Portugal. The study was supported financially by the United Kingdom Medical Research Council. **Competing interests.** The authors have declared that no competing interests exist. CEPH : Centre d\'Etude du Polymorphisme Humain EBI-ASD : European Bioinformatics Institute Alternative Splicing Database LCL : lymphoblastoid cell line LD : linkage disequilibrium RT-PCR : reverse transcriptase PCR SNP : single nucleotide polymorphism

Introduction ============ Bipolar disorder (BP) is among the more reliably diagnosed^[@bib1]^ and most heritable illnesses in psychiatry.^[@bib2]^ Although a recent large-scale genome-wide association study (GWAS) meta-analysis found strong evidence for association of common markers in the calcium channel subunit gene *CACNAC1* and the cell-surface protein gene *ODZ4* with BP, the effect sizes are modest and account for only a small proportion of the phenotypic variance and heritability.^[@bib3]^ Several hypotheses have been proposed to explain why gene identification in psychiatric phenotypes such as BP has met with less success than might have been anticipated. Among these are a higher burden of rare variants,^[@bib4]^ a more polygenic genetic architecture^[@bib5]^ and a more heterogeneous phenotype.^[@bib6]^ If the search for etiology is complicated by phenotypic heterogeneity, a potential method to mitigate this problem might be to delineate clinical subphenotypes that could be more biologically homogenous. Such phenotypes may also help uncover potential susceptibility risk factors for domains of psychopathology that may be shared across diagnoses. In this study, we focus on the Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV subtype of BP with mood-incongruent psychotic features (MICP). This subphenotype was first described a century ago as a potential diagnostic tool to differentiate the more common mood-*congruent* psychotic symptoms of mood disorders from the mood-*incongruent* psychotic symptoms of SZ.^[@bib7]^ More modern conceptions of BP no longer exclude cases with mood-incongruent psychotic symptoms, which are now believed to comprise one-third to one-half of all psychotic BP.^[@bib8],\ [@bib9]^ Mood-incongruent psychotic symptoms in BP appear to be important markers of severity, being associated with worse overall outcome,^[@bib10],\ [@bib11]^ poor lithium responsiveness^[@bib12]^ and higher rates of attempted suicide.^[@bib8],\ [@bib9]^ The presence of MICP symptoms might also be a manifestation of overlapping genetic susceptibility between BP and SZ. Family studies have found a modestly elevated risk of SZ in relatives of cases with mood disorders and MICP,^[@bib13],\ [@bib14]^ and vice versa.^[@bib15]^ More recently, two genome-wide linkage scans of BP with MICP have been performed;^[@bib9],\ [@bib16]^ while there was limited overlap between the two, a number of genome-wide suggestive signals emerged in regions previously implicated in SZ linkage studies, including on chromosomes 1p32, 2q12 and 13q31. We hypothesize that the BP subtype with MICP features may help uncover susceptibility genes for BP not readily detectable by examining the broad BP phenotype. Given the clinical similarities between MICP symptoms in BP and in SZ, we also hypothesize that some of these susceptibility genes would show partial overlap with susceptibility genes for SZ. To test these hypotheses, we have performed a GWAS analyses of the MICP phenotype using several large BP data sets and further sought to determine whether our findings might also influence susceptibility to SZ, by making use of the recently published Psychiatric GWAS Consortium (PGC) SZ results. Materials and methods ===================== Subjects -------- Subjects with BP and MICP were initially drawn from three large, independent samples of BP, here termed the National Institute of Mental Health Bipolar Disorder Consortium (NIMH), the Wellcome Trust Case--Control Consortium (WTCCC) and the German samples. Subject collection and genotyping were conducted separately for each sample. However, quality control steps taken in the analyses of these samples have been identical. NIMH sample ----------- Eligibility, ascertainment and assessment procedures for the NIMH sample have been previously described.^[@bib17]^ All subjects were assessed with the Diagnostic Interview for Genetic Studies (DIGS), and this was combined with family informant data and medical records to assign diagnoses based on DSM-III-R or DSM-IV criteria. Controls were ascertained throughout the United States and were genotyped as part of the Molecular Genetics of Schizophrenia II (MGS2) Collaboration.^[@bib18]^ All control subjects completed a psychiatric questionnaire and those endorsing a history of BP, psychosis or symptoms sufficient for a diagnosis of major depression were excluded. Controls were matched to cases according to ethnicity and sex. Wellcome Trust Case--Control Consortium --------------------------------------- The WTCCC sample comprised the original 1868 participants with BP and 2938 controls from the WTCCC-1 study, and an additional 2725-independent controls from the WTCCC-2 study. Clinical assessment included a semi-structured interview and review of case notes. Ratings of symptom occurrence and course of illness were made using the Bipolar Affective Disorder Dimension Scale^[@bib19]^ and/or the operational criteria (OPCRIT) item check list.^[@bib20]^ Diagnoses were based on all available data using the Research Diagnostic Criteria.^[@bib21]^ Both WTCCC-1 and WTCCC-2 controls were evenly ascertained from the 1958 British Birth Cohort and the UK Blood Service Control Group samples. German sample ------------- The collection and genotyping of the German BP sample has been recently described.^[@bib22]^ In brief, probands with bipolar I disorder were recruited through consecutive hospital admissions and assessed using a structured interview and the OPCRIT checklist. Best- estimate diagnoses were assigned according to DSM-IV criteria. A population-based control sample screened for psychiatric disorders was assembled from the PopGen, KORA and Heinz Nixdorf Recall studies. Mood-incongruent psychosis phenotype ------------------------------------ Considerable attention was given to ensuring comparability between the phenotypes across the three samples. While the NIMH, WTCCC and German samples used different diagnostic instruments (that is, DIGS, Bipolar Affective Disorder Dimension Scale, OPCRIT) to diagnose participants, the psychotic symptoms queried by these interviews are quite similar and, in many cases, identical. In all studies, mood-incongruence was defined according to the DSM-IV definition as psychotic 'content...inconsistent with depressive themes such as guilt, illness, personal inadequacy or catastrophe\... \[or\] inconsistent with manic themes such as inflated worth, power, knowledge, identity, or special relationship to a deity or famous person.\' Further DSM-IV mood-incongruent psychotic symptoms include 'delusions of thought insertion...of thought broadcasting, and...of control\'. For the NIMH samples, the designation of psychotic features was based on a lifetime history of auditory or visual hallucinations or delusions. In the DIGS subjects are asked about the presence of psychotic symptoms during their most severe depression, as well as during a subsequent review of lifetime psychotic symptoms. Interviewers were asked to determine the incongruence of psychotic content following descriptions of hallucinations or delusions in a mood episode. At least one positive response to these questions in either depressive or manic episodes rendered a subject positive for a lifetime history of mood-incongruence. If insufficient information was available to rule in or rule out a lifetime presence of mood-incongruent psychosis, the case was designated as indeterminate. In the genotyped NIMH sample of 2191 BP cases, 1372 (62.6%) were classified as having a lifetime history of psychosis and 960 (43.8%) met criteria for a lifetime history of mood-incongruent psychosis. The WTCCC study used a semi-structured diagnostic instrument similar to the DIGS, which was then used to complete the Bipolar Affective Disorder Dimension Scale and OPCRIT rating scales. The designation of mood-incongruence was made if a case had a rating of \>20 in the Bipolar Affective Disorder Dimension Scale mood-incongruent dimension or if the OPCRIT ratings (questions 54, 58, 59, 61--63, 66--68, 72--77) were positive for any of the DSM-IV mood-incongruent items. Out of the 1868 cases from the WTCCC-1 study, 873 (46.7%) were classified as having had MICP. In the German study, the designation of mood-incongruence was made using the OPCRIT items that reflected the DSM-IV definition. Of the 645 cases with BPI, 363 (56.3%) were classified as having had mood-incongruent psychosis. Genotyping ---------- *National Institute of Mental Health Bipolar Disorder Consortium.* Unrelated cases and controls were genotyped in two separate stages by the Bipolar Genome Study. In all, 1001 cases and 1034 controls selected for the Genetic Association Information Network study were genotyped by the Broad Institute using the Affymetrix 6.0 platform (Santa Clara, CA, USA), while an additional 1190 cases and 401 controls were genotyped by the Translational Genomics Research Institute, also on the Affymetrix 6.0 platform. We included the final cleaned data set from the primary case--control analysis of each sample. The Genetic Association Information Network-BP sample included genotype data on 724 067 single-nucleotide polymorphisms (SNPs). The Translational Genomics Research Institute data set included genotype data on 728 187 SNPs. *Wellcome Trust Case--Control Consortium.* Genotyping was conducted in two stages, with cases and controls available from WTCCC-1 and controls only from WTCCC-2. The combined WTCCC-1 and 2 samples consisted of 1868 cases and 5682 controls (2957 from WTCCC-1 and 2725 from WTCCC-2). There were 382 370 SNPs generated using Affymetrix 5.0 from stage 1 and 696 889 SNPs generated from the Affymetrix 6.0 platform in stage 2. *German sample.* The genotyping of the German BP sample has been recently described.^[@bib22]^ Genotyping was performed on the Illumina HumanHap550 array (Illumina Inc, San Diego, CA, USA). We utilized the final cleaned data set, which included 645 cases and 1310 controls with genotype data on 516 024 SNPs. Data management --------------- *Quality control.* Each data set underwent identical quality control steps. We dropped subjects with a missing data rate⩾3%, and SNPs with a missing data rate⩾5%. SNPs were also dropped if their frequency was\<1% or if they had a Hardy--Weinberg equilibrium *P*\<0.001. The quality control steps were applied to each study individually before and after imputation, as well as to the combined mega-analysis sample. To assess for cryptic relatedness, genotyped SNPs for each study were combined and identity by descent analysis was performed using a linkage disequilibrium-pruned data set. We removed one of any pair of individuals with identity by descent score sharing\>0.2. *Principal component analysis.* To account for population stratification in the GWAS sample, we performed principal component analysis within each data set using markers in linkage equilibrium (*R*^2^\<0.2) with a minor allele frequency \>5%. We also excluded all AT/GC SNPs and all SNPs within the major histocompatibility complex (chr6: 25--23 Mb) and chromosome 8 inversion (chr8: 7--13 Mb) regions. Principal components were selected based on visualization of a screen plot (graphs of the principal components are shown in [Supplementary Figure 1](#sup1){ref-type="supplementary-material"}). Plotting of the principal components showed two, 11, and one clear outlier(s) in the NIMH, WTCCC and German data sets, respectively, that were removed. *Imputation.* Imputation was performed to facilitate a meta-analysis of the three data sets, which were originally genotyped on Affymetrix 5.0 (WTCCC-1), Affymetrix 6.0 (all NIMH and the WTCCC-2 control samples) and Illumina HumanHap550v3 microarray chips. We used BEAGLE to flip orientation to the positive strand and to impute allelic dosages for autosomal SNPs in the cases and controls. Each data set was imputed separately using phased haplotype data from HapMap I & II release 24 (<http://hapmap.ncbi.nlm.nih.gov/>) as the reference panel. The WTCCC, NIMH, and German samples were imputed separately and subsequently combined. Quality control filters were applied to the individual studies and to the combined data set. SNPs were removed according to the following criteria: minor allele frequency\<0.01, *R*^2^\<0.3, Hardy-Weinberg equilibrium *P*\<0.001 and a PLINK INFO score\<0.8. The final imputed files contained 2 488 722 (NIMH), 2 404 996 (WTCCC) and 2 472 086 (German) markers. Data analysis ------------- *Association analysis.* We used the imputed allelic data from the NIMH, WTCCC and German data sets and performed test of association in logistic regression analysis for each dataset, comparing cases with MICP versus controls. Tests of association were performed in PLINK using allelic dosages in an additive logistic regression model including covariates for the top principal components. As a measure of potential residual population stratification, we used WGAviewer (Duke University, Durham, NC, USA) to calculate lambda (*λ*) as the ratio of the median observed versus expected *χ*^2^ statistic. To facilitate comparison with other samples, we also calculated a standardized lambda (*λ*~1000~) for a sample of 1000 cases and 1000 controls. *Case-only analysis.* We performed a case-only analysis of our top findings. Principal components and association studies were performed in each individual sample using all cases with MICP versus all other cases without MICP. As several variables can lead to the designation of MICP, we included as negative only those subjects without any missing data for those variables. *Meta-analysis.* Association analysis of each data set was performed with data set-specific principal components. Meta-analysis was subsequently performed in PLINK under a fixed effect model for both the case--control and case-only analysis. *Polygenic analysis.* Using the PGC SZ results as a training set, we performed a polygenic analysis of each of the three samples in cases with and without MICP. The PGC SZ results were clumped into a data set of 85 284 markers in linkage equilibrium (*r*^2^\<0.25) using the CEU HapMapII samples as a reference. We used PLINK to calculate polygenic scores using the SZ risk alleles weighted by their log(OR) as the initial training set sample. Polygenic scores were derived in each individual MICP samples using several *P*-value thresholds. A case-only logistic regression was subsequently performed for each data set using the polygenic scores as the dependent variable and each study\'s principal components as covariates. Fixed effect meta-analysis of each association was subsequently performed in STATA (Stata Corp, College Station, TX, USA) using its 'meta\' function. As a majority of the controls in the MICP analysis (5022 out of 8148) overlapped with the controls in the PGC schizophrenia (SZ) analysis, we restricted our MICP analysis to a case-only analysis. Results ======= We first performed association analyses of each individual data set. The number of cases with mood-incongruent psychosis across the three data sets was 960 (43.8%) in the NIMH, 873 (46.7%) in the WTCCC and 363 (56.3%) in the German samples. Quantile-quantile (QQ) plots for each sample are shown in [Supplementary Figure 1](#sup1){ref-type="supplementary-material"}. The *λ* estimates were 1.01, 1.04 and 1.02 for the NIMH, WTCCC and German samples, respectively. There were no genome-wide significant findings within each sample, leading us to proceed to a fixed effect meta-analysis of the three data sets. The meta-analysis included a total of 2196 cases with BP with MICP and 8418 controls. As shown in [Figure 1](#fig1){ref-type="fig"} and [Table 1](#tbl1){ref-type="table"}, the most highly associated marker in the meta-analysis was rs1171113, with the C allele being overrepresented in cases (odds ratio (OR)=1.23, *P=*9.67 × 10^--8^). Figure 2a highlights this association in more detail, showing that rs1171113 tags an ∼150--200 kb region of chromosome 6q14.2 that encompasses the genes *PRSS35* (protease, serine 35) and *SNAP91* (synaptosomal-associated protein, 91 kDa homolog, also known as clathrin coat assembly protein AP180). The second strongest association was found on chromosome 3p22.2 with the rs9834970 marker (risk allele=T; OR=0.82; *P=*9.71 × 10^--8^). As shown in [Figure 2b](#fig2){ref-type="fig"}, this marker is ∼15 kb downstream of *TRANK1* (tetratricopeptide repeat and ankyrin repeat containing 1 gene), also known as *LBA1* (lupus brain antigen 1). Finally, [Figure 2c](#fig2){ref-type="fig"} shows the third strongest finding with a *P*\<10^−6^, on chromosome 14q24.2 (rs2333194; *P=*7.0 × 10^--7^), which is in the third intron of *NUMB.* To determine whether these top findings were specifically associated with MICP, we performed a logistic regression analysis comparing BP cases with MICP versus all other BP cases. We found evidence of association of the risk alleles with MICP cases in two of our top findings: rs1171113 (OR=1.14, *P=*0.014) and rs2333194 (OR=1.12, *P=*0.02), but not in rs9834970 (OR=1.04, *P=*0.42). To test the hypothesis that the MICP subset of BP may be associated with SZ, we performed a polygenic analysis of the MICP data set using a case-only design, as most of the controls in this study overlap with those of the PGC SZ study. First, we selected common markers between the PGC SZ and our MICP data sets and subsequently performed linkage disequilibrium-pruning to obtain markers in approximate linkage disequilibrium (*r*^2^\<0.25). We derived a polygenic score from the PGC SZ analysis using several association *P*-value thresholds, ranging from 0.05--0.5, and performed a logistic regression analysis in each case-only sample, testing whether an individual\'s polygenic score was associated with the presence or absence of MICP. Each sample was corrected for by its study-specific principal components. In a meta-analysis of the three samples, we found significant evidence for polygenic overrepresentation of SZ risk alleles in BP cases with MICP ([Table 2](#tbl2){ref-type="table"}), with the strongest evidence of association found at the *P*\<0.2 threshold (*P*~meta~=0.003). Evidence of association was found across all *P*-value thresholds, suggesting that this association was unlikely to be spurious. Discussion ========== In this study we have combined three large BP GWAS samples to test the hypotheses that (1) a subphenotype of BP, marked by MICP, would show association in loci not found in analyses of the more broadly defined phenotype; and (2) that the MICP BP subphenotype would show evidence at a broader level for a relationship with SZ. Our meta-analysis of 2196 cases and 8418 controls found association of several markers with *P*\<1 × 10^--6^, though none reached genome-wide significance. Prior family studies suggested that mood disorders with MICP might be etiologically related to SZ.^[@bib14],\ [@bib23]^ In support of this hypothesis, our polygenic analysis using risk alleles from the PGC SZ GWAS study found an association, *en masse*, between large numbers of SZ variants and the presence of MICP in BP cases, suggesting that genetic overlap across syndromes is potentially due to shared polygenic variation with individual effects that are likely too small to be detected in single marker GWAS analyses. The marker with the strongest association (rs1171113), which was enriched in the case-only analysis, resides in a region of high linkage disequilibrium that includes the genes *PRSS35* and *SNAP91*. The *PRSS35* gene encodes an inactive serine protease homolog of no known function (reference uniprot database^[@bib24]^). *SNAP91* codes for a 91 kDa synaptosomal-associated protein, also known as clathrin assembly protein 180 (AP180), which is enriched in the presynaptic terminal and expressed only in neurons.^[@bib25]^ *SNAP91* is essential for the formation and function of clathrin coated vesicles, which are the major means of recycling vesicles at the presynaptic membrane.^[@bib26]^ Interestingly, *SNAP91* has also been proposed to have a role in calcium signaling and the Wnt pathway. Studies have found that the *SNAP91* protein can inhibit two important intracellular signaling molecules, phospholipase C-γ1 and phospholipase D1.^[@bib27],\ [@bib28]^ Both molecules have been implicated in the Wnt pathway^[@bib29]^ and phospholipase C-γ1 has a pivotal role in intracellular inositol signaling, which is hypothesized to have a role in lithium\'s mechanism of action. As some studies have suggested that BP with mood-incongruent psychosis is less responsive to lithium treatment,^[@bib12],\ [@bib30]^ it is intriguing to speculate that *SNAP91*, via its effect on phospholipase C-γ1, may be involved in lithium response. Our second strongest association signal (rs9834970) is ≈12 kb downstream of the *LBA1* gene, which is also known as *TRANK1* (tetratricopeptide repeat and ankyrin repeat containing 1), and was originally identified as a brain-specific antigen in a murine model of systemic lupus erythematosus, a human disorder with frequent neuropsychiatric symptoms.^[@bib31]^ This marker was recently found to have genome-wide significant evidence for association with BP in a large meta-analysis of mixed ancestry samples.^[@bib32]^ Notably, this marker also showed suggestive evidence for association with SZ in the PGC SZ meta-analysis (*p*=3.6 × 10^--5^).^[@bib33]^ Our association with this marker was found with far smaller sample sizes; however, we did not find evidence that the rs9834970 risk allele was specifically associated with MICP in a case-only analysis, suggesting that this marker is likely to be representative of a risk allele for the broader diagnosis of BP. Our third strongest association (rs2333194) did show modest overrepresentation in MICP in our case-only analysis and was found in *NUMB*, an important modulator of the notch signaling pathway, which among other functions, has been shown to regulate neuronal differentiation *in vitro*.^[@bib34]^ Psychiatric phenotypes are likely amalgams of multiple genetic, development and environmental influences. Although single gene findings can illuminate important pathophysiological mechanisms, phenotypic considerations are likely to be more correlated with markers of aggregate genetic effects, such as familiality or 'global\' genetic analysis like polygenic risk profiling.^[@bib35]^ In this study such a polygenic analysis has found evidence for association of the MICP BP phenotype with SZ, which is consistent with prior family studies, and provides further support for etiological ties between the major mood and psychotic disorders. A prior study found a similar association between SZ and a phenotype that is closely related to BP with MICP, that of schizoaffective disorder as diagnosed by the Research Diagnostic Criteria.^[@bib36]^ Although cases with RDC schizoaffective disorder are included within our diagnosis of BP with MICP, we were not able to derive this RDC-based diagnosis from the DSM-based DIGS interview. The results of this study should be interpreted in light of several limitations. Perhaps most important is the limited power to detect risk variants of small effect sizes, particularly as loss of statistical power is inherent in the subphenotype approach, which splits a broader case sample into smaller units. Even with the three large BP samples, our study (assuming a common and fully informative marker allele) only had sufficient power to detect ORs in the ≈1.3 range, which is slightly greater than that of our top finding (OR=1.23). Although larger samples will provide increased statistical power, they will also create the challenge that arises from the need to harmonize subphenotype definitions. As patients with psychosis may fail to recall specific psychotic symptoms,^[@bib37]^ differential misclassification of mood-incongruence is an important concern, particularly in case-only analysis, which prompted us to focus our primary analysis on a case--control design. In summary, we have performed a large GWAS of the MICP subphenotype and found no associations meeting genome-wide significance criteria, although several associations at the 'suggestive\' level warrant further investigation. More globally, our polygenic analysis has found modest evidence for association of the MICP BP phenotype with SZ, which is consistent with prior family studies, and provides further support for etiological ties between the major mood and psychotic disorders. Supported by NARSAD, NIMH (K99/R00MH086049) and the Wellcome Trust (award 076113). We express our appreciation to the families and individuals who participated in this project. We thank those involved in the BP Research Network (bdrn.org) and the MDF-The Bipolar Organization for the help of its staff and members. Funding for recruitment and phenotype assessment has been provided by the Wellcome Trust and the Medical Research Council. This study makes use of data generated by the Wellcome Trust Case--Control Consortium. A full list of the investigators who contributed to the generation of the data is available from <http://www.wtccc.org.uk>. The NIMH samples were genotyped under the direction of Bipolar Genome Study. The Principal Investigators and Co-Investigators were: John R Kelsoe, Tiffany A Greenwood, Caroline M Nievergelt, Rebecca McKinney, Paul D Shilling, Nicholas Schork, Erin N Smith, Cinnamon Bloss, John Nurnberger, Howard J Edenberg, Tatiana Foroud, Daniel L Koller, Elliot Gershon, Chunyu Liu, Judith A Badner, William A Scheftner, William B Lawson, Evaristus A Nwulia, Maria Hipolito, William Coryell, John Rice, William Byerley, Francis McMahon, Thomas G Schulze, David T Chen, Wade Berrettini, James B Potash, Peter P Zandi, Pamela B Mahon, Melvin G McInnis, Sebastian Zöllner, Peng Zhang, David W Craig, Szabolcs Szelinger, Thomas B Barrett. [Supplementary Information](#sup1){ref-type="supplementary-material"} accompanies the paper on the Translational Psychiatry website (http://www.nature.com/tp) The authors declare no conflict of interest. Supplementary Material {#sup1} ====================== ###### Click here for additional data file. ###### Click here for additional data file. {#fig1} {#fig2} ###### Meta-analysis results for the comparison of bipolar disorder cases with mood-incongruent psychotic features to normal controls, showing markers associated with a *P*\<1 × 10^--5^ *Chromosome* *Location (bp)* *Marker* *Relation to gene* *Distance to gene (bp)* *Gene* *Risk allele* *Odds ratio* P*(meta)* -------------- ----------------- ------------ -------------------- ------------------------- ------------------------- --------------- -------------- --------------- 6 84283314 rs1171113 Intronic 0 *PRSS35/SNAP91* C 1.23 9.67 × 10^−8^ 3 36831034 rs9834970 Upstream 12278 *TRANK1/LBA* T 0.82 9.71 × 10^−8^ 14 72836967 rs2333194 Intronic 0 *NUMB* A 0.83 7.03 × 10^−7^ 3 5680052 rs11710433 Intergenic   N/A C 0.81 1.16 × 10^−6^ 3 13712001 rs4450798 Intergenic 57078 *FBLN2* T 1.26 1.54 × 10^−6^ 8 54717753 rs11773966 Intronic 0 *ATP6V1H* A 2.17 1.92 × 10^−6^ 3 52693320 rs10865974 Intronic 0 *PBRM1/GNL3* T 0.84 2.18 × 10^−6^ 17 6034675 rs3744728 Intergenic 66204 WSC domain containing 1 C 1.51 2.80 × 10^−6^ 1 30298301 rs2860031 Intergenic   N/A G 0.82 2.81 × 10^−6^ 1 24304543 rs3934861 Intronic 0 *MYOM3* A 0.8 3.76 × 10^−6^ 7 137509306 rs10255295 Intergenic 55716 *AKR1D1* G 1.31 4.73 × 10^−6^ 10 32349011 rs1775715 Intronic 0 *KIF5B* G 1.18 5.13 × 10^−6^ 5 17268444 rs2962370 Upstream 2306 *BASP1* G 1.19 5.14 × 10^−6^ 6 37282723 rs1680005 Upstream 5209 *TMEM217* A 1.18 7.45 × 10^−6^ 2 98450324 rs12617721 Intronic 0 *INPP4A* C 1.2 7.45 × 10^−6^ 22 24205265 rs1930961 Intergenic 17620 *CRYBB2P1* C 0.72 9.49 × 10^−6^ ###### Association of PGC schizophrenia polygenic scores with mood-incongruent psychosis in cases with bipolar disorder *Threshold* P*-value (PGC schizophrenia training data set)* *Regression Z-statistic* *MICP case-only* P*-value* ------------------------------------------------------------- -------------------------- ---------------------------- *P*\<0.05 2.59 0.01 *P*\<0.1 2.59 0.01 *P*\<0.2 3.00 0.003 *P*\<0.3 2.91 0.004 *P*\<0.4 2.86 0.004 *P*\<0.5 2.76 0.006 Abbreviations: MICP, mood-incongruent psychotic features; PGC, Psychiatric GWAS Consortium. [^1]: These authors contributed equally to this work. [^2]: These authors contributed equally to this work.

Introduction ============ The metabolic properties of cancer cells are different from those of normal cells. Most of the malignant cancer cells prefer aerobic glycolysis to mitochondrial oxidative phosphorylation (OXPHOS) to satisfy energy and biomass synthesis requirement to survive, grow and proliferate.[@b1-ijn-12-5701]--[@b6-ijn-12-5701] During aerobic glycolysis, an abnormally high rate of glucose uptake was implemented by tumor cells, and most incoming glucose was then metabolized into lactate.[@b7-ijn-12-5701] It was shown that glucose deprivation selectively induces cell injury in transformed human cells via metabolic oxidative stress.[@b8-ijn-12-5701] 2-Deoxy-[d]{.smallcaps}-glucose (2-DG), an analog of glucose, was phosphorylated by hexokinase to 2-DG-phosphate (2-DG-6P) that cannot undergo further glycolysis and thus was used to mimic glucose deprivation and inhibit glycolysis in vivo.[@b9-ijn-12-5701] Accumulation of 2-DG or 2-DG-6P can cause ATP depletion, cell cycle inhibition and cell death.[@b10-ijn-12-5701] If aerobic glycolysis is inhibited in cancer cells, the function of mitochondrial OXPHOS could be restored to compensate for reduced ATP.[@b11-ijn-12-5701] Thus, dual targeting of mitochondrial and glycolytic pathways was suggested as a promising anti-tumor strategy. At present, a group of compounds targeting mitochondria to induce apoptosis and suppress cancer, mitocans, have been defined.[@b12-ijn-12-5701] As shown as an example, α-tocopheryl succinate (α-TOS), the vitamin E derivative, exerts pro-apoptotic effects in a wide range of tumors and is well tolerated by normal tissues.[@b13-ijn-12-5701] It targets mitochondrial complex II to produce reactive oxygen species (ROS) that then trigger apoptosis.[@b14-ijn-12-5701] More specifically, ROS promote phosphorylation of the Mst1 kinase, resulting in phosphorylation of the transcription factor FoxO1 that translocates into the nucleus. This is followed by increased expression of Noxa that diverts Mcl-1 from Bak, causing the formation of pores in the mitochondrial outer membrane, whereby promoting the apoptotic cascade downstream of mitochondria.[@b15-ijn-12-5701]--[@b17-ijn-12-5701] Notably, at the molecular level, α-TOS acts as a Bcl-2 homology domain 3 (BH3) mimetic, effectively sensitizing cancer cells to other drugs. In vitro, α-TOS showed significant anti-tumor effect. Interestingly, in vivo, α-TOS can be easily converted into α-tocopherol (α-TOH) that boosts tumor surveillance in the liver.[@b18-ijn-12-5701] The main obstacle for the successful application of α-TOS-based clinical treatments is the hydrophobic nature of this drug that significantly reduces its bioavailability and therapeutic activity.[@b19-ijn-12-5701] In addition, the intravenous or intraperitoneal administration of α-TOS in an organic phase or oil emulsion can cause important side effects, such as the formation of aggregates as a result of drug mixing with blood plasma, inflammation and embolization processes.[@b20-ijn-12-5701] In recent years, the drug-loaded polymeric nanoparticles (NPs) offer bright promise for cancer therapy.[@b21-ijn-12-5701]--[@b24-ijn-12-5701] To potentially maximize therapeutic activity while minimizing negative side effects of 2-DG and α-TOS, an amphiphilic nanocarrier targeting folate receptor (FR) was designed and used in our study to deliver both chemoagents. It is well known that FR was frequently overexpressed in malignant tissue and cells, which provides unique opportunity to specifically target cancer cells by virtue of its high affinity for folic acid and folate analogs.[@b25-ijn-12-5701] Until now, FR targeting has been exemplified using folate conjugates with a wide variety of diagnostic and therapeutic probes. In this study, we found that mitochondria-targeted drug α-TOS synergizes with 2-DG to trigger various cancer cell death in vitro; the amphiphilic nanocarrier targeting FR on cancer cells and loaded with both showed enhanced anti-tumor effects with less toxicity in vivo. Thus, our study provided a multifunctional nanocarrier with improved anti-tumor therapeutic effects and less toxicity. Materials and methods ===================== Materials --------- 2-DG, α-TOS (PHR1029), coumarin-6, bovine serum albumin (BSA), dimethyl sulfoxide (DMSO), 3-(4,5-dimethylthi azolyl-2)-2,5-diphenyltetrazolium bromide (MTT) and folate acid (FA) were purchased from Sigma-Aldrich Co. (St Louis, MO, USA). Soluble starch, *N*-(3-dimethylaminopropyl)-*N*′-ethylcarbodiimide hydrochloride (EDC), *N*-hydroxy succinimide (NHS), *N*,*N*′-dicyclohexylcarbodiimide (DCC), 4-dimethylaminopyridine (DMAP), stearic acid, glycidylt-rimethylammonium chloride (GTAC), potassium bromide and DMSO-*d*~6~ were purchased from Aladdin (Shanghai, China). DiR dye was purchased from Caliper Life Sciences (Waltham, MA, USA). Annexin V-fluorescein isothiocyanate (FITC)/propidium iodide (PI) was purchased from eBioscience (San Diego, CA, USA). McCoy's 5A Medium and Dulbecco's Modified Eagle's Medium (DMEM) were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Matrigel was purchased from BD (Franklin Lakes, NJ, USA). Besides these, other chemical reagents that were of analytical grade were purchased from Sinopharm Chemical Reagent Co., Ltd (Xi'an, China). Methods ------- ### Cell lines and culture conditions The research had ethical review board approval of the Fourth Military Medical University. HT29 colon adenocarcinoma cells were purchased from American Type Culture Collection (ATCC); HeLa cervical carcinoma cells and A549 lung adenocarcinoma cells were routinely used in our laboratory. Cells were cultured in McCoy's 5A Medium or DMEM with 10% fetal bovine serum (FBS) (Thermo Fisher Scientific) following the instructions from ATCC. ### Synthesis of encapsulation materials Synthesis procedure of cationic amphiphilic starch (CStSa, also (+)StSa) was published in *Biomaterials*.[@b26-ijn-12-5701] Synthesis procedure of BSA-FA followed the literature of Meng et al.[@b27-ijn-12-5701] Briefly, 80 mg of BSA, 40 mg of EDC, 35 mg of NHS and 50 mg of FA were added to DMSO, and the mixture was stirred for 12 h in the dark at 25°C. Then, the sample was centrifuged at 10,000 rpm for 30 min to remove the excess FA and EDC/NHS. In this reaction, EDC and NHS were used to form FA-NHS, which could conjugate with the amino group of BSA. The product was assayed by ^1^H-nuclear magnetic resonance (^1^H-NMR). ### Preparation of α-TOS-2-DG-loaded and FR-targeted nanoparticles (TDF NPs) In the first step, we prepared α-TOS-loaded CStSa micelles. A total of 3 mg of α-TOS and 5 mg of CStSa were co-dissolved in 500 μL of DMSO and then dropwise added to 7 mL of distilled water under stirring. After stirring for 10 min, 5 mg of 2-DG was added into the solution and kept under stirring for 10 min. Then, 2 mL of BSA-FA solution (10 mg/mL) was added into the mixture solution under stirring slowly. Finally, TDF NP solution was transferred into a dialysis membrane (3,500 Da molecular weight cut-off \[MWCO\]) and dialyzed against distilled water for 2 h during which the medium was replaced by clean water thrice, and at the end of the procedure, the volume was set to 10 mL. TDF NPs were further concentrated by treatment with PEG-8000, dialysis and ultrasound redispersion. No drug-loaded NPs with BSA-FA (F NPs) were also synthesized to be a control. ### Characterization of TDF NPs The hydrodynamic diameters and zeta potentials of the TDF NPs were measured by Malvern Instruments (Nano-ZS90, Malvern, UK). The morphology of TDF NPs was observed by a transmission electron microscope (TEM) (H-800; Hitachi, Tokyo, Japan). The solution of TDF NPs was dropped on a copper grid that was coated with ultra-thin carbon support film. When the grid dried, the samples were observed by TEM. The formulas for the calculation of encapsulation efficiency (EE) and drug loading (DL) of TDF NPs are as follows: $$\text{Encapsulation~efficiency}_{\text{drug}}(\%) = \frac{C_{\text{drug~remain}}}{C_{\text{drug~input}}} \times 100\%$$ $$\text{Drug~loading}_{\text{drug}}(\%) = \frac{C_{\text{drug~remain}}}{C_{\text{NPs~input}}} \times 100\%$$where the C~drug\ input~ was the concentration of α-TOS or 2-DG input before encapsulation, C~drug\ remain~ was the concentration of α-TOS or 2-DG after lyophilization and C~NPs\ input~ was the total amount of NPs. ### In vitro drug release profiles of TDF NPs Drug release from TDF NPs was measured by a dialysis method; the experiment was conducted using dialysis membrane (MWCO 3500) in phosphate-buffered saline (PBS) (pH 7.4) under stirring at 30 rpm and 37°C for 72 h. The amount of 2-DG in the sample solutions was determined by spectrophotometry. As the control, the same amount of 2-DG- and α-TOS-loaded CStSa micelles (TD NPs without FA-BSA compared with TDF NPs) and free 2-DG were dispersed in PBS (pH 7.4), and drug release profiles were analyzed at the same condition. The amount of α-TOS is water insoluble, and hence it would not release into the PBS. ### In vitro anti-tumor activity assay To demonstrate the anti-tumor activity, MTT assay, colony formation assay and cell death were performed, respectively. For MTT assay, 150 μL of HT29, HeLa and A549 cells were seeded at 6×10^4^/mL density per well in 96-well plates and then treated with 2-DG or α-TOS, or a combination of 2-DG or α-TOS, or TDF NPs. The dose was referred to the previous report.[@b28-ijn-12-5701],[@b29-ijn-12-5701] For colony formation assay, HT29 or HeLa cells were seeded at 300 cells per dish in 6 cm cell culture dishes and treated with 2-DG and α-TOS for 6 h, then grew for 7--14 days, and the number of colonies formed was counted. For cell death, annexin V-FITC/PI was used according to the kit manual and analyzed on a flowcytometer Coulter Epics XL/MCL (Beckman Coulter Inc, Brea, CA, USA). ### In vivo biodistribution studies All the animal experiments were approved by and performed in accordance with the permitted guidelines of the Animal Ethics Committee of the Fourth Military Medical University. FR-positive HeLa cells (5×10^6^ cells in 100 μL of medium) were mixed with 100 μL of Matrigel and subcutaneously injected into the right flank of 8-week-old healthy nude mice. To monitor FR nanocarrier targeting to tumor, the dye DiR, a near-infrared radiation (IR) fluorescent dye, was used to encapsulate DiR-loaded FR-targeted NPs. Free DiR was used as the control. Six tumor-bearing nude mice were randomly divided into two groups, each mouse of one group was injected intravenously with 100 μL of DiR-loaded FR-targeted NPs, while each mouse of another group was injected intravenously with 100 μL of DiR (2 mg/mL). Then, imaging was taken at 0.5 h, 24 h, 48 h and 72 h after injection with IVIS Lumina II In Vivo Imaging System (Caliper Life Sciences). ### In vivo tumor suppression experiments To investigate the anti-tumor efficacy of TPF NPs in vivo, 24 HeLa tumor-bearing mice were randomly divided into four groups (n=6). The administration was started when the tumor volume was \~50 mm^3^. A total of 100 μL of TDF NPs were intravenously injected at 3-day intervals for six times, an equal volume (100 μL) of normal saline or blank NPs were intravenously administrated as the controls, while the last group was injected with 50 μL of 3 mM free 2-DG intravenously and 50 μL of 30 μM α-TOS in corn oil intraperitoneally, respectively. The body weight of mice and the lengths of the longest tumor axis (a(t), mm) and the vertical axis (b(t), mm) were measured every other day, and the tumor volume (v(t), mm^3^) was calculated using the following equation: V =0.5× a × b^2^. The mice were sacrificed on the 21th day after the first administration, and the tumor of each mouse was weighed. ### Histological assessment The excised tumor and liver tissues were fixed with 4% paraformaldehyde solution for 1 day and then embedded in paraffin and cut into 5 mm thick sections. Hematoxylin and eosin (HE) staining was performed, and hepatocyte or tumor cell necrosis was observed to assess the liver toxicity and anti-tumor effect under the microscope (OLYMPUS IX71; Olympus Corporation, Tokyo, Japan). ### Statistical analysis The data on in vivo tumor suppression assay were given as mean ± SD of independent experiments. The normality of body weight and tumor size was tested by Kolmogorov--Smirnov test on SPSS Version 16.0 (SPSS Inc., Chicago, IL, USA), then one-way analysis of variance (ANOVA) or Student's *t*-test was performed on Prism 5.0 software to determine the difference among the samples (GraphPad Software, Inc., La Jolla, CA, USA). *P*-values \<0.05 were considered as statistically significant. Results ======= A synergistic anti-tumor effect of α-TOS and 2-DG in vitro ---------------------------------------------------------- To determine the anti-tumor effect of α-TOS and 2-DG in combination, we detected the inhibitory effects on cancer cells, including colonic cancer HT29 cell highly favoring glycolytic metabolism, cervix adenocarcinoma HeLa cell preferring to OXPHOS and lung carcinoma A549 cell in a dose- and time-dependent model. MTT assay could reflect the growth of cancer cells by detecting live cells. Considering that the maximum dose of outer encapsulated 2-DG was no \>3.5 mM (data not shown), the dose of 3 mM of 2-DG was used in our experiment. Based on 3 mM of 2-DG, 15 μM, 30 μM and 45 μM of α-TOS were combined with 2-DG, respectively. Our results showed that 30 μM, even 15 μM, of α-TOS alone could significantly inhibit the growth of cancer cells in 48 h, and 3 mM of 2-DG by itself also could lead to the inhibition of tumor cells. However, the cell death was no \<50% when administrated with single drug. In contrast, a combined treatment of 15 μM of α-TOS and 3 mM of 2-DG could remarkably reduce the survival of cancer cells compared to treatment with α-TOS or 2-DG alone. Expectedly, the survival of cancer cells dramatically decreased with the increased doses of α-TOS to 30 μM or 45 μM ([Figure 1A](#f1-ijn-12-5701){ref-type="fig"}; \**P*\<0.05), which indicated that the combined treatment was dose-dependent. The MTT assay also showed that 45 μM of α-TOS did not significantly reduce the cell variability of HeLa cell compared to 30 μM of α-TOS; meanwhile, the combination treatment of 30 μM of α-TOS and 3 mM of 2-DG induced \>50% of cell death in HT29 and A549. Thus, the combination dose of 30 μM of α-TOS and 3 mM of 2-DG was used for drug-loaded dose of nanocarrier next performed. Further, time-dependent effects of the combined treatment were detected in 24 h, 48 h and 72 h after administration. [Figure 1B](#f1-ijn-12-5701){ref-type="fig"} shows that the survival of tumor cells with combined treatment decreased in 24 h; with the treatment prolonged to 48 h, even 72 h, the survival continued to decrease with a mild trend ([Figure 1B](#f1-ijn-12-5701){ref-type="fig"}; \**P*\<0.05), indicating that the anti-tumor effects of combined treatment were superior to that of single drug, and the inhibition to cancer cells was also time-dependent. Even early in 12 h, our data on flow cytometry showed that cell apoptosis and death were induced by combination drug treatment. In HT29, HeLa and A549 cells, the percentage of the early apoptosis and late apoptosis in combination treatment group was higher than that of single drug administration, especially for HeLa cells ([Figure 1C](#f1-ijn-12-5701){ref-type="fig"}). To further know the proliferation of a single cancer cell, we performed a plate colony formation assay. Our results demonstrated that the colony number was decreased when α-TOS or 2-DG was treated alone, especially α-TOS, while it remarkably reduced, even no colony was present on plate when treatment of α-TOS and 2-DG was combined, which showed that a low dose of α-TOS and 2-DG in combination mostly inhibited the proliferation of HT29 cancer cells ([Figure 1D](#f1-ijn-12-5701){ref-type="fig"}). Taken together, our results showed that the combined treatment of α-TOS and 2-DG could exhibit significantly superior anti-tumor effects on cancer cells to treatment of α-TOS or 2-DG alone, indicating that the treatment in combination by lower doses or shorter time could have a comparable anti-tumor effect. Preparation and characterization of TDF NPs ------------------------------------------- In order to overcome the administration of liposoluble agent α-TOS and further reduce the potential toxicity of 2-DG, tumor-targeting nanocarriers were designed. With consideration of the advantages of FR and its ligand as a targeting delivery system, we synthesized BSA-grafted FA and then covered them onto dual drug-loaded micelles, which absorbed 2-DG on surface and encapsulated α-TOS in the core, to prepare targeting dual drug nanocarriers ([Scheme 1](#f6-ijn-12-5701){ref-type="fig"}). FR is frequently overexpressed on the vast majority of cancer tissues such as ovarian, colorectal and breast cancer, while its expression is limited in healthy tissues and organs. The ligand, folic acid, is small, stable over a broad range of temperatures and pH values, inexpensive, and non-immunogenic, and it retains its ability to bind to the FR after conjugation with drugs or diagnostic markers. After folate attaches to its receptors located within caveolae, it is internalized through the endocytotic pathway. As the pH of the endosome approaches five, the BSA-FA was decomposed and the drug released. [Figure S1](#SD1-ijn-12-5701){ref-type="supplementary-material"} shows that folate was correctly grafted on BSA. The concentrations of drugs in TDF NPs were adjustable in limited range. Dialysis membrane (3,500 Da MWCO) and PEG-8000 were used in enrichment of drugs. After dewatering, ultrasound was employed for redispersion. Finally, the actual usage dose of α-TOS and 2-DG could match the concentrations of free drugs. [Table 1](#t1-ijn-12-5701){ref-type="table"} displays the EE and drug loading capacity (DLC) of TDF NPs. The data were comparable to those drawn from other reports, which were shown to efficiently prevent a burst release of NPs. Indeed, dispersion of free 2-DG and 2-DG- and α-TOS-loaded CStSa micelles without BSA-FA (TD NPs) into the outer dialysis chamber showed different speeds ([Figure 2A](#f2-ijn-12-5701){ref-type="fig"}). Within 5 h, free 2-DG completely dispersed into outer dialysis solutions; within 24 h, 80% of 2-DG was released from TD NPs into outer solutions. While within 48 h, 50% of 2-DG was released from TDF NPs into outer solutions, indicating that TDF NPs covered by BSA-FA have the potential to control drug burst release. For TDF NPs, almost no α-TOS release was found within 80 h, and release of 2-DG was sustained and slow within 72 h. Further, we detected the magnifications of TDF NPs through TEM. [Figure 2B and C](#f2-ijn-12-5701){ref-type="fig"} shows that the NPs have a clear spherical shape with a core--shell structure and the size was smaller than 100 nm. The TDF NPs were owing to the electrostatic interaction between negative BSA-FA and positive α-TOS-loaded CStSa micelles. The zeta potential of α-TOS-loaded CStSa micelles and TDF NPs was +28.2±2.9 mV and −22.3±2.1 mV, respectively ([Figure 2D](#f2-ijn-12-5701){ref-type="fig"}). The change of zeta potential in samples showed that TDF NPs comprised TD NPs and BSA-FA, and BSA-FA was covered on the surface of TD NPs factually. In vitro anti-tumor effect of TDF NPs ------------------------------------- To detect the anti-tumor activity of TDF NPs, HT29, HeLa and A549 cells mentioned earlier were tested. The results of MTT assay demonstrated that drug-loaded TDF NPs have remarkable inhibition to the growth of HeLa cells, while no significant suppression to that of HT29 and A549 cells, compared with F NPs without drug loading ([Figure 3A](#f3-ijn-12-5701){ref-type="fig"}). Notably, we found that only F NPs slightly promoted the growth of HeLa cells, which indicated that the nanocarrier was less cytotoxic to tumor cells, as expected low or no cytotoxicity to normal cells. Further detection of FR expression on these tumor cells, we found that HeLa cells expressed highest level of FR, whereas low or undetectable expression in HT29 and A549 cells ([Figure S2](#SD2-ijn-12-5701){ref-type="supplementary-material"}). The data on FR expression level was consistent with previous reports.[@b30-ijn-12-5701] Our results indicated that α-TOS- and 2-DG-encapsulated TDF NP was effective against tumor cells, and the suppression to tumor cells was dependent on FR-mediated endocytosis. To further confirm the tumor-specific cellular uptake of TDF NPs, HeLa cells were incubated with or without 1% FA after coumarin 6-labeled F NPs were added, then cellular uptake behavior was observed using fluorescent microscopy. [Figure 3B](#f3-ijn-12-5701){ref-type="fig"} exhibits the behavior pattern of HeLa cells after incubating with coumarin 6-labeled F NPs with or without FA for 2 h. Strong green fluorescence was observed in the cytoplasm of HeLa cells incubated without FA, while few fluorescence was seen in the cytoplasm of HeLa cells and most appeared in the proximal outer membrane region when FA was added. It seemed that FA added into medium bound to F NPs, thus preventing the occurrence of endocytosis mediated by the binding of F NPs to FR on the surface of HeLa cells. In vivo distribution of TDF NPs ------------------------------- Further, in vivo the distribution of TDF NPs was studied using near infrared (NIR) fluorescence imaging. In order to facilitate the observation of biodistribution of TDF NPs in vivo, DiR dye was encapsulated by CSaSt to generate DiR-loaded nanoparticles (D-NPs), which comprised CSaSt, BSA-FA and DiR. Then, D-NPs (test group) and free DiR (control group) were injected i.v. into HeLa tumor-bearing nude mice, respectively. The red fluorescence became gradually strong in the tumor-bearing right hindlimb from 0.5 h until 24 h after the injection of D-NPs and was extremely remarkable to 72 h, while no apparent fluorescence accumulated on the tumor of mouse until 24 h after injection with free-DiR ([Figure 4](#f4-ijn-12-5701){ref-type="fig"}), which suggested the tumor-specific distribution of TDF NPs. Besides, considerable fluorescence was only seen around the spleen of mouse test group, which may be caused by phagocytosis of immune cells such as macrophage. Notably, the fluorescence intensity was reduced faster in control group than in test group within 48 h, indicating a slower clearance of TDF NPs in the body. All the earlier results demonstrated that TDF NPs could specifically load drug to FR-positive tumors and prolong the retention in the body, thus improving the anti-tumor efficiency of loaded drugs and reducing the risk of systemic toxicity. In vivo tumor suppression effect of TDF NPs ------------------------------------------- Finally, we studied the tumor suppression effect of TDF NPs in vivo. The weight loss, tumor volume and histologic section of tumor and liver could fully reflect the efficiency and toxicity of treatment to mouse. In [Figure 5A](#f5-ijn-12-5701){ref-type="fig"}, the body weight of TDF NP-treated group grew slowly, suggesting the efficiency of TDF NPs treatment, while that of PBS-treated group lost significantly in the late phase of treatment, indicating the occurrence of cancer cachexia for PBS-treated group. Meanwhile, TDF NP-treated group showed a lowest growth of tumor size compared to groups treated with PBS, F NPs and α-TOS plus 2-DG ([Figure 5B](#f5-ijn-12-5701){ref-type="fig"}). All the tumors were excised and scaled with a ruler ([Figure 5C](#f5-ijn-12-5701){ref-type="fig"}). For PBS group, four mice died before the end point for the cancer cachexia. As expected, the tumor volume of TDF NPs group was remarkably smaller than F NPs, suggesting that the combination treatment took a good anti-tumor effect. Meanwhile, the anti-tumor effect of TDF NPs was better than the non-nano group, indicating that the two drug-loaded NPs had an effective biological function in vivo. HE staining of pathological sections of tumor demonstrated notable necrosis and fewest tumor cells infiltration in TDF NPs group. In addition, the liver tissue of TDF NPs group showed less damage than that of free α-TOS plus 2-DG group ([Figure 5D](#f5-ijn-12-5701){ref-type="fig"}). All these data suggest an effective tumor suppression and less liver cytotoxicity of TDF NPs. Discussion ========== In this study, we synthesized a novel anti-tumor nano-medicine, TDF NPs, which was proved to be effective and less cytotoxic compared to free drug 2-DG or α-TOS, thus providing a promising cancer treatment. Several reports showed that metabolic profiles of cancer cells divert significantly from those of non-neoplastic cells; thus, tumors rely on these metabolic alterations for growth, metastasis and survival, the atypical pathways might be potential targets of antineoplastic drugs. The glucose metabolism in cancerous cells is remarkably different from that in their normal counterparts. Proliferating cancer cells prefer to obtain ATP via aerobic glycolysis rather than OXPHOS even in the presence of ample oxygen. We thus use 2-DG, non-metabolizable glucose analog, which targets glucose metabolism to deplete cancer cells of energy. Our results showed that 3 mM of 2-DG alone could inhibit the growth and proliferation of cancer cells in vitro. Previous work showed that the ratio of 3 mM of 2-DG to glucose in the culture media was 1:3.3 and the ratio proved to inhibit glycolysis. Our results were consistent with previous cell culture studies.[@b28-ijn-12-5701] However, the suppression effect of 2-DG by itself was limited since 6 mM of 2-DG did not significantly increase the inhibition ratio. Previous studies showed that 2-DG combined with other therapeutic agents or radiotherapy could exhibit a synergistic anticancer effect. Since OXPHOS is another pathway to contribute to energy production in cancers, targeting mitochondrial metabolism has been proposed for cancer therapy. α-TOS, the vitamin E analogs, showed great promise for future clinical applications as a potential anticancer agent. However, the low solubility in physiological media limited the use of α-TOS. By designing an amphiphilic nanocarrier encapsulating α-TOS and 2-DG, we overcome poor bioavailability of α-TOS, thus dramatically improving the anti-tumor effect of 2-DG treatment alone. In our study, both α-TOS and 2-DG, when administered as single agents, were limited to inhibit the proliferation of HT29, HeLa and A549 tumor cells, and when administered in combination, their effect on such tumor growth was synergistic, thus resulting in significant cell death, as demonstrated in vitro by MTT and Annexin V/PI staining. The advantage of combined administration was proved by other such studies. Goldberg et al[@b31-ijn-12-5701] combined Ras inhibitor farnesylthiosalicylic acid (FTS) and 2-DG to treat pancreatic tumors, Cheong et al[@b32-ijn-12-5701] used 2-DG and metformin to treat breast tumor, Zhang et al[@b33-ijn-12-5701] combined α-TOS and doxorubicin to treat gastric cancer and so on. Co-delivery nanocarrier means two or more therapeutic drugs are co-encapsulated into one nanocarrier. Many reports have proved that it is an efficient tool to simultaneous administration of multiple therapeutics, including cytotoxic agents, chemosensitizers, small interfering RNA (siRNA) and anti-angiogenic agents.[@b26-ijn-12-5701] In our previous work, we have designed and synthesized PgS and CSaSt micelles.[@b26-ijn-12-5701],[@b34-ijn-12-5701] For CSaSt micelles, we successfully encapsulated insoluble apogossypolone and soluble doxorubicin into one nanocarrier targeting hyaluronic acid receptor CD44. Further, the EE and DL were measured; the drug-loaded NP was characterized by IR spectra, TEM and drug release assay. As expected, the dual drug nanocarriers showed remarkable accumulation in tumor region and promoted the anti-tumor efficacy. Notably, in vivo acute toxicity experiment showed that LD~50~ of drug-loaded NP was remarkably lower compared with the combination group (apogossypolone and doxorubicin administrated simultaneously without encapsulation into one nanocarrier). All these work indicated that we have established a safe and effective nanocarrier system. For α-TOS and 2-DG, the main problem is to set a proper dose ratio of α-TOS and 2-DG for FR-positive HeLa cells. Dong et al[@b35-ijn-12-5701] tested for the sensitivity of several tumor cells to α-TOS and found IC~50~ value varied greatly for various tumor cells. We also found the difference of live cell percentage among HT29, HeLa and A549 cells when treated with α-TOS. Compared with HT29 cells, HeLa cells had a slightly lower IC~50~ value ([Table S1](#SD3-ijn-12-5701){ref-type="supplementary-material"}). Based on the data of our experiment, a combination dose of 30 μM of α-TOS and 3 mM of 2-DG was determined, which was suitable for HeLa cells. For breast carcinomas and Ras overexpression-induced tumor, the in vivo dose of α-TOS varied from 6 μM to 15 μM.[@b29-ijn-12-5701],[@b35-ijn-12-5701] Thus, for different tumors, an optimized in vivo dose of TDF NPs should be explored. For FR highly expressed HeLa cells, the TDF NPs showed a significant anti-tumor effect compared with the combination of the two free drugs ([Figure 5C and D](#f5-ijn-12-5701){ref-type="fig"}). Although the combination of the two free drugs did not show significant toxicity to mice ([Figure 5A and D](#f5-ijn-12-5701){ref-type="fig"}), it did not lead to a parallel anti-tumor effect to the TDF NPs. In turn, the combination of the two free drugs needs to increase dose to show the notable anti-tumor effect, which may cause significant toxicity in vivo. Thus, the TDF NPs showed a great potential in tumor therapy. Conclusion ========== In this study, we targeted the tumor cell metabolism pathway to inhibit their growth and proliferation with the combination of two chemical agents α-TOS and 2-DG. Our results showed that in vitro α-TOS and 2-DG had a synergistic anti-tumor effect on tumor cells, especially on HeLa cells; TDF NPs encapsulating α-TOS and 2-DG also had a significant inhibition to the growth of FR highly expressed tumor cells. In vivo TDF NPs could target to FR-positive tumor and displayed a remarkable anti-tumor effect with less toxicity to liver. Supplementary materials ======================= Synthesis and characterization of encapsulation materials and bovine serum albumin (BSA)-folate acid (FA) --------------------------------------------------------------------------------------------------------- Synthesis procedure of cationic amphiphilic starch was published in *Biomaterials*.[@b36-ijn-12-5701] Synthesis procedure of BSA-FA followed literature of Meng et al.[@b37-ijn-12-5701] Briefly, 80 mg of BSA, 40 mg of *N*-(3-dimethylaminopropyl)-*N*′-ethylcarbodiimide hydrochloride (EDC), 35 mg of *N*-hydroxysuccinimide (NHS) and 50 mg of FA were added to dimethyl sulfoxide (DMSO), and the mixture was stirred for 12 h in the dark at 25°C. Then, the sample was centrifuged at 10,000 rpm for 30 min to remove the excess FA and EDC/NHS. In this reaction, EDC and NHS were used to form FA-NHS, which could conjugate with amino group of BSA. The product was assayed by ^1^H-nuclear magnetic resonance (^1^H-NMR). The tumor cell lines and their folate receptor (FR) expression -------------------------------------------------------------- RNA extraction, complementary DNA (cDNA) reverse transcription and real-time polymerase chain reaction (PCR) were performed according to our previous work.[@b38-ijn-12-5701] ###### The ^1^H-NMR result of BSA-FA. **Note:** As shown by the ^1^H-NMR spectra of BSA-FA dissolved in DMSO-*d*~6~, peaks assignment for the BSA and FA were straightforward, where the peaks at 6.5--8.5 ppm from FA. **Abbreviations:** BSA, bovine serum albumin; FA, folate acid; ppm, parts per million. ###### FR expression in HT29, HeLa and A549 tumor cell lines. **Abbreviation:** FR, folate receptor. ###### LC~50~ values of α-TOS for apoptosis in tumor cell lines Cell type α-TOS ----------- ------- HT29 37±6 HeLa 30±4 A549 35±5 **Notes:** Tumor cell lines were treated at \~60% confl uency. The IC~50~ values were derived from viability curves using the MTT viability assay and are expressed in μM. **Abbreviations:** α-TOS, α-tocopheryl succinate; LC~50~, lethal concentration 50. This study was supported by the Innovation Project of Shaanxi Science and Technology (2015KTCL03-13) and National Natural Science Foundation of China (NSFC No 31571215). We acknowledge the technician Jintao Hu for flow cytometry analysis. **Disclosure** The authors report no conflicts of interest in this work. ###### A synergistic anti-tumor effect of α-TOS and 2-DG was observed in vitro. **Notes:** HT29, HeLa and A549 tumor cells were treated with α-TOS, 2-DG alone or α-TOS plus 2-DG. The dose-dependent growth of tumor cells was tested by MTT assay in (**A**) and time-dependent growth in (**B**). The death of tumor cells was confirmed by flow cytometry in (**C**), and the proliferation of tumor cells was detected by clone forming assay in (**D**). Data points represent the mean ± SD of three separate experiments (n=3), each performed in triplicate. \**P*\<0.05 compared with α-TOS or 2-DG treatment alone. **Abbreviations:** α-TOS, α-tocopheryl succinate; 2-DG, 2-deoxy-[d]{.smallcaps}-glucose; MTT, 3-(4,5-dimethylthiazolyl-2)-2,5-diphenyltetrazolium bromide; PI, propidium iodide.   ![The synthesis and characterization of TDF NPs.\ **Notes:** Drug release from TDF NPs was measured by a dialysis method in (**A**); (**B**) TEM image, (**C**) size distribution and (**D**) zeta potential of TDF NPs.\ **Abbreviations:** α-TOS, α-tocopheryl succinate; 2-DG, 2-deoxy-[d]{.smallcaps}-glucose; TDF NPs, α-TOS-2-DG-loaded and folate receptor-targeted nanoparticles; TEM, transmission electron microscope; TD NPs, 2-DG- and α-TOS-loaded CStSa micelles without FA-BSA grated.](ijn-12-5701Fig2){#f2-ijn-12-5701} ![In vitro TDF NPs cellular function and behavior.\ **Notes:** The effects of TDF NPs on the growth of tumor cells in (**A**); cellular uptake behavior of TDF NPs (**B**). Data points represent the mean ± SD of three separate experiments (n=3), each performed in triplicate. \**P*\<0.05 compared with F NPs (blank nanocontrol). Scale bar 50 μm; magnification ×200.\ **Abbreviations:** α-TOS, α-tocopheryl succinate; 2-DG, 2-deoxy-[d]{.smallcaps}-glucose; TDF NPs, α-TOS-2-DG-loaded and folate receptor-targeted nanoparticles; F NPs, BSA-FA grafted nanoparticles without drug loaded; BSA, bovine serum albumin; FA, folate acid; NS, no significance.](ijn-12-5701Fig3){#f3-ijn-12-5701} ![In vivo distribution of TDF NPs in tumor-bearing mice.\ **Notes:** Fluorescence images of DiR-loaded nanoparticles in vivo. HeLa tumor-bearing nude mouse was injected free DiR dye or D-NPs intravenously. In vivo DiR fluorescence images of the mice at three time points 24 h, 48 h and 72 h were shown. The sites of tumors were indicated by yellow circles.\ **Abbreviations:** D-NPs, DiR-loaded nanoparticles; α-TOS, α-tocopheryl succinate; 2-DG, 2-deoxy-[d]{.smallcaps}-glucose; TDF NPs, α-TOS-2-DG-loaded and folate receptor-targeted nanoparticles.](ijn-12-5701Fig4){#f4-ijn-12-5701} ###### In vivo tumor suppression effects of TDF NPs. **Notes:** (**A**) The body weight curves of tumor-bearing mice treated with PBS, nanocontrol, TDF NPs and a combination of α-TOS and 2-DG, respectively; (**B**) the tumor growth curves, (**C**) tumors and (**D**) pathological section images of the tumor and liver tissue. Circles indicated necrosis region, and rectangle indicated the 10×10 zoom region. \**P*\<0.05 compared with PBS or F NPs treatment (nanocontrol). The 4×10 magnification scale bar 500 μm; 10×10 magnification scale bar 200 μm. **Abbreviations:** TDF NPs, α-TOS-2-DG-loaded and folate receptor-targeted nanoparticles; F NPs, BSA-FA grafted nanoparticles without drug loaded; PBS, phosphate buffer saline; α-TOS, α-tocopheryl succinate; 2-DG, 2-deoxy-[d]{.smallcaps}-glucose.   ![Schematic representation of the assembling process of TDF NPs.\ **Note:** The process of TDF NP synthesis was schematically represented.\ **Abbreviations:** α-TOS, α-tocopheryl succinate; 2-DG, 2-deoxy-[d]{.smallcaps}-glucose; TDF NPs, α-TOS-2-DG-loaded and folate receptor-targeted nanoparticles; (+StSa), cationic amphiphilic starch; BSA, bovine serum albumin; FA, folate acid; FR, folate receptor.](ijn-12-5701Fig6){#f6-ijn-12-5701} ###### Summary of TDF NPs EE and DLC Drug EE DLC ------- ------------ ------------ α-TOS 94.3%±1.3% 8.9%±0.83% 2-DG 61.7%±7.7% 13.2%±2.6% **Note:** Values stated as mean ± SD. **Abbreviations:** α-TOS, α-tocopheryl succinate; 2-DG, 2-deoxy-[d]{.smallcaps}-glucose; TDF NPs, α-TOS-2-DG-loaded and folate receptor-targeted nanoparticles; EE, encapsulation efficiency; DLC, drug-loading capacity. [^1]: These authors contributed equally to this work

INTRODUCTION {#sec1-1} ============ The World Health Organization (WHO) defined health in its broader sense in 1946 as "a state of complete physical, mental, and social well-being and not merely the absence of disease or infirmity."\[[@ref1][@ref2]\] According to the WHO, the main determinants of health include the social and economic environment, the physical environment, and the person\'s individual characteristics, and behaviors. More specifically, key factors that have been found to influence whether people are healthy or unhealthy include\[[@ref3][@ref4][@ref5]\] Income and social status, social support networks," education and literacy' employment/working conditions, social environments, physical environments, personal health practices and coping skills, healthy child development, biology and genetics, health-care services, gender, and culture.\[[@ref6]\] The aim of this hypothesis is to bring to you and all of WHO directors and specialist attention to ask whether the health of individuals has a mathematical code or not? If so, the new view must be considered in regard with the health of the world population, which will be discussed. Hypothesis {#sec2-1} ---------- ### Does the health of individuals have a mathematical code? {#sec3-1} It seems that each society in the World has particular features, which can be: (1) Geographic characteristics of hometown, (2) Geographic of place of residence, (3) heredity, (4) amount of income, (5) access to health services, (6) culture and custom, (7) life expectancy, (8) possibility of contracting diverse diseases, (9) natural and unnatural disasters, (10) number of people living in particular family or country, (11) extent of literacy, (12) extent of support to other members of society, (13) extent of beliefs in spiritual issues, (14) access to appropriate accommodation, (15) national and international journeys, (16) extent of education, (17) violence and ethnic group aggressions, (18) war, (19) economic sanctions, (20) adequate health behaviors, (21) access to mass media, (22) access to and essential food calories, (23) access to safe water, and (24) mental health aspects. Since all the mentioned aspects and other relevant factors vary from person-to-person around the world, it is possible to support the theory that an individual\'s, health may have mathematical codes. What is the mathematical code for health? The mathematical code for good health could be defined as all the effective consequences for individuals to be healthy[1](#fn1){ref-type="fn"} that can be defined as the followings: All health aspect of people, which could be calculated and measurable (i.e., which I have listed in 24 categories as mentioned as above). Can the health code in each person change? When someone is born, he or she gets one or some of the above mathematical code features. After a while, this mathematical code can changed positively as a result of the effectual health factors or negatively because of different diseases or harsh occurrences. Example 1: A child is born with genetic problems. Her or his mathematical code is the sum of the above features minus his or her genetic problems that can continue for the rest of his or her life or can affect his or her mathematical code until the genetic problems are cured. Therefore, the mathematical code of each is the entire positive and negative occurrences that are effective to his or her health.Example 2: A child is born with no genetic problems. However, since he or she was not vaccinated against poliomyelitis, he or she gets that disease. Thus, because poliomyelitis is a lifetime illness, it is excluded from his or her positive health code.Example 3: Child is born with malnutrition. When that the problem is solved, his or her mathematical code will improve.Example 4: Healthy individual lives in an area with air pollution. Therefore, the air pollution causes different respiratory diseases for that individual. Unless this person does not change the area of residence, the mathematical code will continue to decrease negatively.Example 5: After contracting different diseases, a healthy individual is not able to continue daily life normally. Because of the total days of illness and costs of treatment, the mathematical code will change the worse. Why do we need the mathematical health code? We have said that health is a quality and includes physical, psychological, and social features. However, nobody has ever declared that we can measure health. Measuring the health code will assist us to understand {#sec2-2} ------------------------------------------------------ To what extent the person is healthyTo what extent is the society' healthy and this includes members of the society\'s healthWhether normal distribution of society\'s health exists in quantitative manners?What factors lead to an increase individuals' health code?Can individuals have some of these factors?Which one of these factors is unchangeable?What factors cause a decrease in people\'s code?Can individuals change some of these factors?Whether that are individuals' responsibility regarding their mathematical health code?Whether measuring the health code enhances enthusiasm and the interest of people in increasing their mathematical health code?Whether measuring the health code enhances enthusiasm and interest of Governments to increase the societies' mathematical health code?Whether the measuring health code demonstrates health insurances companies' responsibilities regarding the maintenance of good health code in individualsMeasuring health code demonstrates international organizations (WHO, United Nation, food agriculter organization etc.,) responsibilities regarding maintenance of good health code in the WorldHealth insurance cost is less for people with a high score in mathematical codeHealth insurance cost is higher for people with a low score in mathematical code.Example: If someone is in good health, the cost of health insurance is only for medical check-ups.It illustrates imperative factors for good health (for governments)It illustrates imperative factors for good health (for families). CONCLUSIONS {#sec1-2} =========== I believe that each individual has mathematical code for their health. Because different thing, which are the effect the health of humans are well described as mathematical phenomenon (i.e., number of bones, teeth, cells, nerve, gene, and etc., which have been calculated with mathematical rules). In addition, in my opinion each factors in this hypothesis could be considered as a numerical issue by using definitions and ask from international organizations (i.e., Air pollution could be defined by WHO. as Air without pollution (with score -2) and Air with pollution (with score-2 and −1 for heavy and light air pollution respectively. The same for an individual without genetic disorder could be considered as 1 score and with a genetic disorder could be considered as −1. All factors could be calculated and divide over 100. The individual mathematical code could be obtained; on the other hand, the health of individual score is going to 100 and the unhealthy sore far away that. Of course, may there are several interactions between aspects of health in various of conditions, for overcome this problem as mentioned before the mathematical code is not constant that depend on attention on of health either individually or by governments. In other words, the people and governments trying to increase their score time by time. The Governments have the responsibility to calculate mathematical code for individual at the time birth and re-calculation could be done each 5 year by governments and Ministry of Health with supervision of international health organizations and statists for whole the nations could be published formally by United Nation and other organizations to encourage governments to prepare the best quality of health. As our organs parts follow mathematical rules, we can utilize these rules for the benefit of our health\'s mathematical code. Therefore, it is possible to calculate all the effective factors on individuals' health, including current and future factors, by 100. All the individuals, families, societies and international organizations should attempt to reach this indicator. The definition of health in the early 1900s was synonymous with good hygiene, not having an infectious disease, wellness or absence of illness. **Source of Support:** Nil **Conflict of Interest:** None declared

Introduction ============ Emergency department (ED) crowding has been identified as a significant patient safety issue as it has been associated with increased patient mortality, delayed resuscitation efforts, and increased length of stay in hospital.[@b1-oaem-7-079] Since ED triage processes influence the time to assessment, extent of patient workup, and length of stay, they play an important role in ED crowding and patient safety.[@b1-oaem-7-079] Despite growing numbers of mental health patients presenting to general hospital EDs for care,[@b2-oaem-7-079] triage nurses' accuracy for rating urgency levels of mental health patients is low.[@b3-oaem-7-079] Moreover, studies testing Canadian Triage and Acuity Scale (CTAS) have included few mental health presentations in their designs.[@b3-oaem-7-079]--[@b5-oaem-7-079] The CTAS is used by the majority of EDs in Canada.[@b6-oaem-7-079],[@b7-oaem-7-079] Agreement and accuracy study designs testing the CTAS have varied. While studies of the CTAS have found that inter-observer reliability is 'high' or 'significant',[@b7-oaem-7-079],[@b8-oaem-7-079] moderate or good levels of agreement are not uncommon.[@b9-oaem-7-079],[@b10-oaem-7-079] In 2004, CTAS was revised to include first- and second-order modifiers.[@b6-oaem-7-079] These modifiers are objective observations that can alter the CTAS acuity level.[@b6-oaem-7-079] As an example, first-order modifiers can include vital signs or pain scales, whereas second-order modifiers are more specific to the presenting complaint and help to determine the patient's risk.[@b6-oaem-7-079] In 2008, significant revisions to the CTAS mental health category and the corresponding second-order modifiers were released.[@b11-oaem-7-079] This study contributes to the CTAS literature by testing these 2008 revisions. In particular, the study suggests that the consistent use of second-order modifiers may increase the accuracy of urgency ratings in mental health presentations.[@b12-oaem-7-079] Purpose ======= The purpose of this study was to test the inter-rater reliability among triage nurses and their accuracy of assigning urgency ratings for standardized mental health patient scenarios, based on the 2008 CTAS guidelines.[@b11-oaem-7-079] Research questions ------------------ The following research questions were developed for this study: What is the inter-rater/inter-observer reliability among triage nurses assigning levels of urgency to mental health patient scenarios, based on the 2008 CTAS guidelines, using a standardized triage tool (Emergency Department Information System \[EDIS\])?How accurate are triage nurses in assigning levels of urgency to mental health patient scenarios, based on the 2008 CTAS guidelines, using a computerized tool (EDIS)? Methodology =========== Following a review of previous work on inter-rater reliability involving triage nurses, a consistent approach to research design was not found.[@b9-oaem-7-079],[@b13-oaem-7-079] Although 'real-time' design was considered, the limits on time and variation among patient presentations across the participating sites were considered significant challenges. While paper-based scenarios have raised methodological concerns,[@b14-oaem-7-079] they offer the ability to obtain responses from all participants based on the same information, thus reducing the variability of the patient's presentation that occurs in real-time design.[@b13-oaem-7-079],[@b15-oaem-7-079] Accordingly, paper-based scenarios were used as a 'suitable estimate' of inter-rater reliability with live triage cases.[@b16-oaem-7-079] Sample ------ The sample of 18 triage nurses was drawn from three participating sites, including one community and two tertiary acute care hospitals. Each participant completed a demographic questionnaire that asked about years of triage experience, specific education or training, and overall comfort level and confidence with triaging specific mental health presentations. Scenario development -------------------- The patient scenarios used in this study were developed by referencing triage notes from mental health patients' presentations from a teaching hospital ED. The abstracted data for the patient scenarios included: a) Canadian Emergency Department Information System (CEDIS) category, b) CTAS chief complaint, c) age, d) mode of arrival, e) vital signs, and f) details of the triage note. The abstracted data also indicated whether or not the patient attended the department alone as well as the patient's involuntary status under a provincial 'Mental Health Act', prior history of mental health diagnoses, level of risk, and displays of agitation and/or aggressive behavior. The abstracted data were used to form a template for the mental health patient scenarios and served as the first stage of scenario development. The data's primary use was to determine the components of mental health triage assessments, for example, 'Mental Health Act' status. A total of 38 scenarios were distributed to an expert panel comprised of emergency physicians and academic researchers with expertise on the topic who reviewed the scenarios for both face and content validity. The expert panel directed revisions and established the accurate triage ratings (that is, 'correct' urgency level). Following the process of expert review, 20 scenarios were chosen to represent the revised guidelines and a typical distribution of presentations. Included were: two CTAS 1 (immediate), five CTAS 2 (emergent), six CTAS 3 (urgent), five CTAS 4 (less urgent), and two CTAS 5 (not urgent) scenarios. Methods ------- Ethical approval for this study was obtained from the University of Manitoba, Education/Nursing Research Ethics Board. All study participants utilized the EDIS, a computerized tool that facilitates application of CTAS guidelines during the triage process. Although the EDIS prompts the nurse through a series of decisions, the triage nurse maintains the ability to override the final CTAS generated score and choose a CTAS rating that best matches their clinical assessment. The CTAS guidelines include the use of CEDIS entrance complaints and second-order modifiers. Each participant rated the 20 scenarios in randomized order. As the participants rated each of the scenarios, they had the opportunity to select a second-order modifier to assist in their decision making. In circumstances where the participant chose a second-order modifier, the modifier chosen was recorded in the EDIS and data sheets were printed for each scenario triaged. At the beginning of each session, participants were provided scripted information, including the assumption that the presenting patient was medically stable. Participants were asked to enter their triage assessments as true to a real-life scenario as possible. Statistical analysis -------------------- The inter-rater reliability of the CTAS has been measured in previous studies using written patient scenarios and analyzed using an unweighted and weighted kappa statistic.[@b7-oaem-7-079],[@b8-oaem-7-079] Unlike previous studies,[@b7-oaem-7-079],[@b8-oaem-7-079] which compared raters between groups the current study tested the agreement within one group. This distinction necessitated the use of a statistical analysis method capable of measuring agreement among multiple raters. The present study used Fleiss' kappa statistic and Kendall's coefficient of concordance. Fleiss' kappa is bounded between 1 and −1 where 0 represents no agreement and perfect agreement is considered to be at values over 0.80.[@b18-oaem-7-079]--[@b20-oaem-7-079] Kendall's coefficient calculation ranges between 0--1, where 0 represents no agreement and 1 represents complete agreement.[@b21-oaem-7-079] Since Kendall's coefficient partially accounts for the degree of closeness of responses, overall agreement scores calculated using Kendall's statistic may be higher than those using the Fleiss calculation. Results ======= The participants were experienced triage nurses; over 40% of the sample reported more than 10 years of triage experience, with high levels of education in emergency nursing (see [Table 1](#t1-oaem-7-079){ref-type="table"}) and confidence in mental health triage (see [Table 2](#t2-oaem-7-079){ref-type="table"}). In contrast to earlier findings, participants did not report lower rates of confidence with aggressive behaviors.[@b22-oaem-7-079],[@b23-oaem-7-079] No relationship was found between level of confidence and years of experience. Data analysis ------------- The overall Fleiss' kappa, the measure of inter-rater reliability for this sample of triage nurses (n=18), was 0.312. Based on guidelines developed by Landis and Koch,[@b20-oaem-7-079] this kappa represents only fair agreement, but is nonetheless statistically significant (*P*\<0.0001). Additional analysis showed the agreement among triage nurses based on CTAS level. Moderate agreement was shown for CTAS level 1 (kappa =0.459) and 4 (kappa =0.500). The kappa statistic for CTAS level 2 was 0.107, CTAS level 3 was 0.218, and CTAS level 5 was 0.022. Kendall's coefficient for this sample was 0.680 (*P*\<0.0001), indicating moderate agreement. In order to calculate the accuracy of the urgency ratings, a custom code in SAS was used to compute the 'Light' statistic[@b24-oaem-7-079] to determine whether the raters agree with correct responses significantly more than by chance alone. A calculated *P*-value \<0.001 demonstrated that triage nurses' agreement with the correct level was not purely random (Dufault B, University of Manitoba, personal communication, February, 2011). Accuracy has also been presented as a correct response by urgency level in earlier studies.[@b17-oaem-7-079] A similar calculation was conducted in this study to determine which CTAS levels produced the highest reported accuracy (see [Table 4](#t4-oaem-7-079){ref-type="table"}). Discussion ========== Since the overall agreement in this study was calculated statistically to be 'fair', significant variability in the urgency ratings existed. In this sample, the highest rate of agreement occurred at the highest (CTAS level 1) and lower ends of the scale (CTAS level 4) (see [Table 3](#t3-oaem-7-079){ref-type="table"}). Despite the significant variability in urgency ratings, the influence of the second-order modifiers was an important finding. Triage scales like CTAS are designed to support decision making, guiding the triage nurse to a correct decision.[@b25-oaem-7-079] The addition of second-order modifiers to the CEDIS entrance complaints in the mental health category are intended to further aid in the triage nurses' decision making. As an example, for the CEDIS entrance complaint of 'depression, suicidal, or deliberate self harm', the second-order modifiers are: 1) attempted suicide or clear suicide plan, 2) active suicidal intent, 3) uncertain flight or safety risk, 4) suicidal ideation, no plan, and 5) depressed, no suicidal ideation.[@b11-oaem-7-079] The first three second-order modifiers listed above would result in a CTAS score of 2, whereas the fourth modifier would result in a CTAS score of 3, and the fifth modifier would result in a CTAS score of 4.[@b11-oaem-7-079] The additional cues provided by these modifiers may help to increase the structure of the tasks at triage, thereby allowing for greater analysis by the nurse while making their decision. Arguably, not utilizing second-order modifiers places greater emphasis on intuition or gut feelings, which may impact triage accuracy. If more participants in this study had relied on the second-order modifiers, the overall rate of agreement and accuracy may have been higher. Over-triage and under-triage are important considerations in triage literature. Over-triage occurs when a patient is seen faster than is required, while under-triaging results in a patient waiting longer than is considered appropriate.[@b13-oaem-7-079] The influence of the nurse's experience level on over- or under-triaging has been reported with mixed results.[@b13-oaem-7-079],[@b26-oaem-7-079] Nurses in the present study were more likely to over-triage. The slight tendency toward over-triaging may have been influenced by the atypical conditions, including: more time to gather data, absence of department pressures, and the knowledge that they were participating in a study. Moreover, specific scenarios influenced the rate of over- and under-triaging in this study. In particular, a high percentage of participants over-triaged scenarios 16 and 20 (see [Table 4](#t4-oaem-7-079){ref-type="table"}). Furthermore, nearly 80% of the participants assigned CTAS level 3 to scenario 20, which may suggest the content of the scenario more accurately represented that level of urgency. Half of the participants assigned CTAS level 3 rather than the 'correct' score of 2 for both scenarios 13 and 18. This finding may suggest distinguishing between CTAS level 2 and 3 can pose challenges to nurses when triaging mental health presentations. Based on this sample, the influence of education, confidence, and comfort level on the accuracy or inter-rater reliability of urgency ratings among triage nurses is not yet clear. Overall, the inter-rater reliability was fair to moderate despite high levels of reported confidence and educational preparation with triaging mental health presentations. The influence of second-order modifiers on the accuracy or urgency ratings may be significant. While it is possible that the scenarios in this study lacked the specific triage cues that triage nurses rely on when triaging mental health scenarios, participants who consistently used second-order modifiers had higher accuracy ratings. Nurses who assigned urgency ratings that matched the 'correct' response more than 60% of the time used second-order modifiers for the majority, and in some instances, all of the 20 scenarios (see [Table 5](#t5-oaem-7-079){ref-type="table"}). Nurses in this group also refrained from or avoided entirely the use of override to change the computer-generated CTAS score. In contrast, nurses that assigned the correct score less often (less than 40% of the time) were less likely to use second-order modifiers or avoided their use altogether. Some individuals in this second group of nurses also utilized override more often than triage nurses who had higher frequencies of 'correct' responses. This provides some support for the notion that the use of heuristics and intuitive decision making in the emergency setting[@b27-oaem-7-079] may introduce a cognitive bias that could compromise patient outcomes.[@b27-oaem-7-079] Limitations ----------- Several limitations of the present study exist. Participants used paper-based scenarios to assign their levels of urgency and conducted their ratings in an environment that was in stark contrast to their typical work settings. These paper-based scenarios were reviewed by an expert committee but were not piloted with a sample of ED triage nurses. As an example, the fact that nearly 80% of the participants assigned a rating of 3 to scenario 20 suggests that if the scenario was piloted by practicing triage nurses, it may have undergone revision. While a pilot of the scenarios may have led to additional revisions, a pilot may have reduced the sample of participants available for the study. The sample and research design were also limitations of the study and must be considered when reviewing these findings. The participants in this study identified themselves as confident and comfortable with mental health presentations. Future studies should consider mixing mental health scenarios in amongst medical and trauma presentations to reduce the focus on mental health, potentially reducing the effect of self-selection. Conclusion ========== Overcrowding, rising patient acuity, and longer lengths of stay have increased the pressure on health care systems to devise and implement triage systems that are both fast and accurate.[@b28-oaem-7-079] Interestingly, although mental health patients are presenting to general hospital EDs for care in greater numbers, inter-rater reliability and the accuracy of triage ratings may be lower than for medical presentations.[@b2-oaem-7-079],[@b3-oaem-7-079],[@b28-oaem-7-079] Although lack of confidence may influence ratings, the participants in this study were highly experienced and rated themselves as comfortable with mental health presentations. Despite this, inter-rater reliability was low, particularly in the mid ranges of the CTAS score (level 2 and 3) when second-order modifiers were not used as a triage aid. Specific focus on the use of second-order modifiers in the orientation and ongoing education of triage nurses may improve the reliability and validity of the CTAS when used to assign urgency ratings to mental health presentations. **Disclosure** The authors report no conflicts of interest in this work. ###### Specialized mental health training or experience Educational courses completed Number of triage nurses Percentage of nurses who completed the specified educational course ---------------------------------- ------------------------- --------------------------------------------------------------------- None 2 11.1 CTAS training 1 5.6 Regional triage orientation 1 5.6 Advanced emergency course 1 5.6 Other educational course 1 5.6 More than one educational course 12 66.7 Total 18 100.0 **Abbreviation:** CTAS, Canadian Triage and Acuity Scale. ###### Triage nurses' comfort level with mental health patient presentations Presentation Not at all comfortable Mildly confident Moderately confident Very confident Total ---------------------------------- ------------------------ ------------------ ---------------------- ---------------- ------- Triage of mental health patients 0 4 7 7 18 Psychotic symptoms 0 6 8 4 18 Manic symptoms 0 4 9 5 18 Anxiety 0 0 9 9 18 Depression 0 1 9 8 18 Suicidal ideation 0 1 10 7 18 Aggressive behaviors 0 4 9 5 18 Behavior/personality disorders 2 7 5 4 18 ###### Percentage correct by CTASurgency level CTAS urgency level Number of scenarios Number of correct responses Number of responses Percentage correct -------------------- --------------------- ----------------------------- --------------------- -------------------- 1 2 27 36 **75%** 2 5 29 90 32.2% 3 6 71 108 65.7% 4 5 63 90 **70%** 5 2 4 36 11.1% **Note:** The figures in bold represent the highest percentage correct by CTAS urgency level. **Abbreviation:** CTAS, Canadian Triage and Acuity Scale. ###### Accuracy, over-triage, and under-triage by one level Scenario number Correct answer Percentage of correct responses (accuracy) n=18 Number of nurses who over-triaged by one level Number of nurses who under-triaged by one level Total n=18 ----------------- ---------------- ------------------------------------------------- ------------------------------------------------ ------------------------------------------------- ------------ 1 3 72.2% 1 3 17 2 4 94% 1 0 18 3 2 22.2% 8 6 18 4 4 77.7% 4 0 18 5 2 22.2% 6 6 14 6 3 72.2% 3 2 18 7 4 94% 0 1 18 8 1 77.7% 0 4 18 9 2 38.8% 6 4 17 10 3 72.2% 3 1 17 11 3 61.1% 2 5 18 12 1 72.2% 0 2 15 13 2 44.4% 1 9 18 14 3 66.6% 0 6 18 15 4 77.7% 3 1 18 16 5 11.1% 11 0 13 17 5 11.1% 2 0 4 18 2 33.3% 0 9 15 19 3 50% 8 1 18 20 4 0.05% 14 0 15 ###### Use of second-order modifiers Rater Frequency of use of second-order modifiers (N=20) Use of override when rating scenarios (N=20) Percentage of correct responses ------- --------------------------------------------------- ---------------------------------------------- --------------------------------- 1 20 (100%) 0 70% (n=14) 2 20 (100%) 0 55% (n=11) 3 1 (5%) 8 (40%) 40% (n=8) 4 16 (80%) 0 60% (n=12) 5 19 (95%) 2 (10%) 50% (n=10) 6 14 (70%) 0 55% (n=11) 7 20 (100%) 0 55% (n=11) 8 15 (75%) 2 (10%) 65% (n=13) 9 18 (90%) 0 60% (n=12) 10 0 0 35% (n=7) 11 19 (95%) 0 55% (n=11) 12 17 (85%) 0 65% (n=13) 13 20 (100%) 0 55% (n=11) 14 12 (60%) 3 (15%) 50% (n=10) 15 19 (95%) 7 (35%) 35% (n=7) 16 20 (100%) 0 65% (n=13) 17 20 (100%) 6 (30%) 55% (n=11) 18 20 (100%) 0 45% (n=9)

Introduction ============ Hepatocellular carcinoma (HCC) is the eighth most common malignancy worldwide and the most common malignant neoplasm of the liver. Epidemiologically, HCC is most common in Asia and sub-Saharan Africa \[[@b1], [@b2]\]. The most common risk factors identified include cirrhosis and hepatitis B and C. Additional risk factors include: hemochromatosis, α~1~-antitripsin deficiency, exposure to aflatoxins, Thorotrast administration, oral contraceptives, and vinyl chloride exposure \[[@b1], [@b3]\]. Hepatitis B is considered the primary cause of approximately 80% of cases worldwide. The peak age group for development is in the 6th to 8th decades with a definite male predominance of 4:1. The median survival rate remains dismal being less than 1 year and a 5-year survival rate of only 2%--5%. \[[@b4]\] There has been some improvement in 1-year survival over the past 5 years, thought to be a reflection of earlier detection of smaller resectable tumors and more aggressive surgical approaches. Clinical presentation ===================== The most common symptomatology includes: abdominal pain, malaise, fatigue, and weight loss \[[@b5]\]. Findings on physical examination often include an enlarged, irregular, and nodular liver as a result of cirrhotic change. Jaundice and abnormal liver function tests may only be evident later in the course because of the significant functional reserve of the liver parenchyma. Paraneoplastic syndromes are not uncommon including hypercalcemia and hyperglycemia as well as polycythemia which occurs in less than 10% of patients \[[@b6]\]. Pathology ========= HCC is the result of a series of events that cause the development of frank malignancy from smaller lesions. A number of stages have been identified including: cirrhosis and regenerative nodules, dysplastic nodules, early HCC, and advanced HCC \[[@b7], [@b8]\]. The earliest phase i.e., regenerative nodules are areas of hepatocyte proliferation surrounded by fibrous septa that develop as a non-specific response to liver injury. In contrast to regenerative nodules, the dysplastic nodule is pre-malignant and represents proliferation of hepatocytes with nuclear variation that can be identified as low-grade to high-grade dysplasia. Early HCC is those lesions that are less than 2 cm in diameter with well-differentiated cells and a preserved trabecular appearance \[[@b8]\]. These early tumors usually have a well-defined margin with a thin fibrous capsule but may demonstrate extra-nodular growth into surrounding liver sinusoids. Large HCCs, those greater than 2 cm, are classified as advanced HCC. These may be nodular, massive, and diffuse. Diffuse HCC results in tumor nodules dispersed within the liver parenchyma. Staging ======= Staging of HCC is critical to clinical management and is assessed through the TNM system. The primary lesion is defined by tumor size, number, and location of lesions, invasion of local vascular structures, and extension into the biliary system. Additionally, the staging system evaluates the location of regional, nodal disease and the presence or absence of distant metastases ([Table 1](#T1){ref-type="table"}) \[[@b5], [@b9]--[@b11]\]. T1 lesions are those that are solitary tumors without vascular invasion, T2 lesions are solitary tumors with vascular invasion or multiple tumors, none larger than 5 cm; T3 represents multiple tumors larger than 5 cm, tumor invading major portal or hepatic venous structures, and T4 lesions are tumors with direct invasion of adjacent organs. Using a binary method, lymph node staging consists of N0 without regional metastases and N1 with regional lymph node metastasis. Distant metastases are defined with M0 being no distant metastasis and M1 being the presence of distant metastasis. Table 1TNM staging system devised by the American Joint Committee on CancerClassificationMeaning*Tumor*TXTumor cannot be assessedT0No evidence of primary tumorT1Solitary tumor $\documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts,amssymb,amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n}{<-> linotext}{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document} $ \le $ \end{document}$ 2 cm without vascular invasionT2aSolitary tumor $\documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts,amssymb,amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n}{<-> linotext}{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document} $ \le $ \end{document}$ 2 cm with vascular invasionT2bMultiple tumors \<2 cm limited to one lobe without vascular invasionT2cSolitary tumor \>2 cm without vascular invasionT3aSolitary tumor \>2 cm with vascular invasionT3bMultiple tumors \<2 cm limited to one lobe with vascular invasionT3cMultiple tumors \>2 cm limited to one lobe with or without vascular invasionT4aMultiple tumors in more than one lobeT4bTumors involving a major branch of the portal vein or hepatic vein(s)T4cInvasion of adjacent organs other than the gallbladder*Nodes*NXRegional lymph nodes cannot be assessedN0No regional lymph nodesN1Regional lymph node metastasis*Metastases*MXDistant metastasis cannot be assessedM0No distant metastasisM1Distant metastasis present (most common to lung and bones)*Stage*IT1 N0 M0IIT2 N0 M0IIIAT3 N0 M0IIIBT1 N1 M0T2 N1 M0T3 N1 M0IVAT4 N \[any\] M0T \[any\] N0 M1IVBT \[any\] N \[any\] M1 Imaging findings ================ Computed tomography (CT) and magnetic resonance (MR) are the primary modalities used to obtain imaging studies for the diagnosis and staging of HCC \[[@b12]--[@b24]\]. In this manuscript we specifically address the use of multidetector CT (MDCT) in the assessment of hepatocellular carcinoma \[[@b5], [@b10]\]. Although MR imaging may be more sensitive in detecting small lesions, MDCT is highly effective in the detection and staging of HCC. This is, however, significantly dependent upon technique, which must be meticulous with new MDCT scanners. There are three CT-defined phases of contrast enhancement that should be exploited for the accurate detection and staging of HCC. The first phase, the hepatic arterial phase (HAP) occurs at 20--30 s following the administration of contrast material. Contrast material should be given with a bolus injection and power injector and rates should be fast in the range of 4--6ml/s. Scanning in the second phase should ideally be performed as an early parenchymal phase in the range of 40--55 s, and the third phase, now termed the hepatic venous phase (HVP), previously termed the portal venous phase (PVP) on single detector CT occurs from 65 to 80 s following contrast administration ([Fig. 1](#F1){ref-type="fig"}). Hypervascular lesions are best imaged during the early arterial phase when they appear as areas of increased enhancement relative to the unenhanced liver parenchyma \[[@b5], [@b11], [@b12]\]. Although hypervascular lesions are usually best seen during this phase, they are often also seen during the early parenchymal phase and are usually less well seen during the standard HVP/PVP phases. MDCT also allows for high-quality, thin-section imaging. Scans are ideally performed at less than 1 cm, specifically 5 mm sections. The early phase images can be used to develop a three-dimensional (3D) vascular map. When a capsule is present, it is usually hypodense on the HAP or mixed density on the PVP and showing some enhancement on delayed images. The use of MDCT has essentially replaced CT arterial portography and allows a non-invasive approach to imaging HCC. Multiphasic imaging allows the detection of tumor thrombus within vascular structures most efficiently on the later phase images. One may be able to distinguish between bland tumor thrombus and tumor thrombus by identifying the enhancement of clot in earlier phases. The sensitivity and specificity of MDCT are still evolving with changes in the equipment. This should erode the previously documented limitations of detection although in the background of the cirrhosis, there still are significant limitations in detection. The following images demonstrate the appearance of HCC on MDCT for the various stages using the TNM system as referenced in [Table 1](#T1){ref-type="table"} ([Figs 2--8](#F2 F3 F4 F5 F6 F7 F8){ref-type="fig"}). The variables of detectability of HCC depend on the phase of contrast administration and each individual tumor is important to appreciate ([Figs 9--11](#F9 F10 F11){ref-type="fig"}). Figure 1Demonstration of optimal image appearance in the three major phases: (a) hepatic arterial phase (HAP), (b) late arterial phase (LAP), (c) hepatic venous phase (HVP). MDCT allows the ability to acquire images within each of these three distinct phases for optimal imaging of hypervascular lesions. Figure 2Hepatocellular carcinoma TNM system---T1. (a) Color diagram demonstrating a T1 lesion: solitary \<2 cm without the presence of vascular invasion. (b) MDCT at the dome of the liver in the hepatic arterial phase (HAP) demonstrating a 2 cm hypervascular lesion in the dome of the liver. Figure 3Hepatocellular carcinoma TNM system---T2. (a) A T2a lesion represents a solitary tumor \<2 cm with vascular invasion. (b) A T2b lesion is multiple tumors \<2 cm limited to one lobe without vascular invasion. (c) A T2c lesion is a solitary tumor \>2 cm without vascular invasion. (d) CT scan showing multiple phases with T2b lesions: HAP, LAP, HVP. (e) Example of T2c lesion in left lobe. Figure 4Hepatocellular carcinoma TNM system---T3. Color diagrams demonstrating T3 lesion staging. (a) T3a is a solitary tumor \>2 cm with vascular invasions. (b) A T3b lesion consists of multiple tumors limited to one lobe \<2 cm with vascular invasion. (c) A T3c lesion consists of multiple tumors limited to one lobe of the liver measuring \>2 cm with or without vascular invasion. (d) Multiphasic CT scan showing T3c lesions: HAP, LAP, HVP. Figure 5Hepatocellular carcinoma TNM system---T4. (a) T4a demonstrating multiple tumors in more than one lobe of the liver. (b) T4b multiple tumors involving a major branch of the portal venous or hepatic venous system. (c) CT scan at the level of the portal vein during the late arterial phase (LAT) demonstrating multiple tumors in a bilobar distribution. Figure 6MDCT of the liver demonstrating a T4b lesion with involvement of the main portal vein as well as multiple lesions in the liver. (a) Early phase (arrow) indicating a clot in the portal vein. (b) Later phase (arrow) with a clot in the portal vein. Figure 7(a) An MDCT of a HCC T4 lesion demonstrating an extensive clot (arrow) within the main portal vein. (b) HCC, T4 lesion in a patient with a clot in the main portal vein (arrow) and extensive collaterals, three phases: HAP, LAP, HVP. Figure 8Hepatocellular carcinoma TNM system demonstrating nodal disease. (a) Color diagram demonstrating nodes along the region of the porta hepatis. (b) MDCT scan at the level of the porta hepatis demonstrating nodal disease in the porta hepatis. Figure 9Multifocal HCC demonstrating optimal enhancement noted in the HAP (a) and LAP (b). Lesions are seen as hypervascular against a low-density liver in the early hepatic arterial phase HAP and continue to enhance in the LAP. Note that the lesions are very poorly seen and have almost disappeared during the hepatic venous phase (HVP) (c). Figure 10Hepatocellular carcinoma demonstrating best visibility in the late arterial phase (LAP) but also being able to be seen in the hepatic venous phase (HVP). Lesion not seen on HAP. (a) HCC not seen on HAP. (b) Hypervascular lesion in the dome of the liver during the LAP. (c) The lesion is still seen very well while in the HVP but now the low-attenuation necrosis is evident. Figure 11Hepatocellular carcinoma with satellite lesion only seen on the late arterial phase (LAP). (a) Early hepatic arterial phase (HAP) imaging demonstrates the large lesion is only subtly seen but the satellite lesion is not clearly evident. (b) The primary large mass in the liver is now enhancing and the satellite lesion (arrow) can be seen in the left lobe of the liver (LAP). (c) Dominant lesion poorly seen and satellite lesion no longer seen. Treatment options ================= The treatment of HCC includes surgery, chemotherapy, radiation therapy, and, most importantly, combination therapies. Data from the National Cancer Data Bank (NCDB) reported that approximately 17.7% of patients were treated with chemotherapy alone, 17% with surgery alone, and 3.2% with radiation therapy \[[@b25]--[@b29]\]. Surgery ------- Surgery is considered the primary treatment option for candidates in stages 1, 2, or 3a. In these cases, surgical removal can be performed by either immediate tumor resection or by orthotopic liver transplantation. The 5-year survival rates for these patients following liver transplantation ranges from 58% to 75% \[[@b3]--[@b7], [@b24]--[@b26]\]. For resection, 5-year survival ranges from 35% to 51% \[[@b25]\]. The recurrence rate after resection has been reported to be as high as 1/3 of patients with a recurrence rate of 3%--17% for orthotopic liver transplantation \[[@b30]\]. The 5-year survival rates for single lesions smaller than 5 cm has been reported as 63% \[[@b31]\]. Larger tumors, however, are associated with poorer outcomes especially in the absence of a tumor capsule, the presence of multiple nodules, satellite lesions, or the presence of vascular invasion. Other treatment options ----------------------- Percutaneous therapies such as ethanol injection, transarterial catheter embolization, cryoablation, and radiofrequency ablation (RFA), are secondary treatment options. RFA is a common therapeutic option for tumors smaller than 3 cm and results in 5-year survival rates of just less than 50% \[[@b32]\] ([Fig. 12](#F12){ref-type="fig"}). Figure 12Demonstration of radiofrequency ablation (RFA) of HCC. (a) The low attenuation peripherally in the right lobe is the result of RFA. There is some mild enhancement peripherally (arrow). (b) The low attention peripheral RFA defect is now slightly smaller and the area of enhancement has resolved, indicating it did not represent recurrent tumor but rather benign enhancement related to the RFA procedure on the scan done 4 months previously. Post-chemoembolization demonstrates the focal high-density area posteriorly in the right lobe (arrowhead). (c) MDCT 9 months later demonstrates the continued decrease in the size of the RFA defect without recurrence and the chemoembolized tumor is also smaller (arrowhead). Adjuvent systemic chemotherapy has been used with single and multiple agents including 5-fluoro uracil, interferon, cisplatinum, thalidomide, octreotide, and tamoxifen.

Introduction {#jcmm13229-sec-0001} ============ Bladder cancer (BCa) is the second most common cancer arising in the genitourinary tract, and it is characterized by its unpredictable nature of progression [1](#jcmm13229-bib-0001){ref-type="ref"}. Considering that nearly 33--75% of patients failed to respond to therapy due to the disease relapse and metastasis [2](#jcmm13229-bib-0002){ref-type="ref"}, there is still an urgent need for further investigation of the carcinogenesis and development of BCa. The Mediator is a multiprotein complex that regulates transcription at the level of RNA polymerase II (polII) assembly and links to developmental diseases and cancer [3](#jcmm13229-bib-0003){ref-type="ref"}, [4](#jcmm13229-bib-0004){ref-type="ref"}. Mediator complex subunit 19 (Med19) is one of components of the Mediator complex, which was originally identified in a screening for mutants with increased aerobic expression of the CYC7 gene in yeast [5](#jcmm13229-bib-0005){ref-type="ref"}. An increasing number of studies have identified that Med19 is involved in the progression and metastasis of various cancers [6](#jcmm13229-bib-0006){ref-type="ref"}, [7](#jcmm13229-bib-0007){ref-type="ref"}, [8](#jcmm13229-bib-0008){ref-type="ref"}, [9](#jcmm13229-bib-0009){ref-type="ref"}, [10](#jcmm13229-bib-0010){ref-type="ref"}. Zhang *et al*. [9](#jcmm13229-bib-0009){ref-type="ref"} first reported the causation of Med19 high expression could facilitate BCa cell growth. In addition, Wen and colleagues suggested that Med19 might promote invasive behaviour and bone metastasis of BCa cells by stimulating the expression of bone morphogenetic protein‐2 (BMP‐2) [11](#jcmm13229-bib-0011){ref-type="ref"}. Although it established the important role of Med19 in BCa, the exact mechanism remains elusive. Recently, it was reported that the Wnt/β‐catenin signalling pathway is involved in the cell proliferation and differentiation in BCa [12](#jcmm13229-bib-0012){ref-type="ref"}. Mediator complex has been proposed to function as a signal transducer coupled with its genetic links to the Wnt/β‐catenin signalling pathway [13](#jcmm13229-bib-0013){ref-type="ref"}. Therefore, we further performed experiments to test whether Med19 subunit in Mediator plays an important role in BCa cell growth and tumour metastasis *via* regulating Wnt/β‐catenin signalling pathway. Materials and methods {#jcmm13229-sec-0002} ===================== Patient and tissue sample {#jcmm13229-sec-0003} ------------------------- This study was approved by the Ethics Committee of the Affiliated Yantai Yuhuangding Hospital of Qingdao University (Permit Number: 2013‐46). Each patient provided signed consent to permit the use of samples in our study. We collected 15 fresh BCa tissues paired with corresponding adjacent non‐cancerous tissues from patients who underwent surgery between March 2015 and April 2015. During surgery, fresh tumour tissue and paired non‐cancerous tissue isolated from at least 2 cm away from the tumour border were collected in the operating room and processed immediately in liquid nitrogen within 15 min. None of these patients received neoadjuvant or adjuvant chemotherapy before the operation. In addition, 167 paraffin‐embedded archived BCa samples between July 2013 and February 2015 were obtained from our hospital for immunohistochemistry (IHC). The criteria for enrolment were histopathological identification of bladder urothelial carcinoma, newly diagnosed without preoperative chemotherapy or radiotherapy, and no history of other tumours. All pathology slides were thoroughly re‐evaluated by two senior uropathologists, who were blind to patient clinical outcome. Patients were stratified by gender, and by tumour number, grade, stage and recurrence. Immunohistochemical staining and evaluation criteria {#jcmm13229-sec-0004} ---------------------------------------------------- All tumour sections were dewaxed and rehydrated by routine methods and incubated in 3% H~2~O~2~ for 30 min. Slides were incubated with rabbit polyclonal primary antibodies against Med19 at a dilution of 1:100 in a humidified chamber 4°C overnight. Sections were stained with 3,3′‐diaminobenzidine (DAB) and counterstained with haematoxylin according to the manufacturer\'s protocol. Bile duct tissue samples served as negative controls. Sections with confirmed positive expression of Med19 were used as positive controls. Based on the percentage for Med19 immune‐positive tumour cells, a score of one was given when ≤5% of cells were positive; two when 6--25%, three when 26--50% and four when ≥50% of cells were positive. Staining intensity was scored as 0 (negative), 1 (weak), 2 (moderate) and 3 (strong). Both scores were multiplied and the resulting score was used to dichotomize Med19 expression as low (≤6) and high (\>6). Cell culture and transfection {#jcmm13229-sec-0005} ----------------------------- The human bladder cancer cell lines T24, UM‐UC3 and 5637 were obtained from the Institute of Biochemistry and Cell Biology, Shanghai, China. Cells were grown in RPMI1640 (Gibco BRL, Grand Island, NY, USA) supplemented with 10% foetal bovine serum (Gibco BRL) at 37°C in a humidified incubator with 5% CO~2~. One day prior to infection, cells were plated at a density of 20--30%. Recombinant lentivirus expressing short‐hairpin RNA (shRNA) targeting Med19 (target sequence: shRNA \#1, 5′‐GGTGAAGGAGAAGCTAAGT‐3′; shRNA \#2, 5′‐GTAGCTCTTTCAATCCTAT‐3′) and a non‐silencing control were constructed by GeneChem, Shanghai, China, and cells were also transfected with the empty vector control. Cells were harvested for analysis of mRNA and protein levels 3 days after infection. Cell proliferation assay {#jcmm13229-sec-0006} ------------------------ Cells were seeded in 96‐well culture plates (3 × 10^3^ cells/well) in triplicates and were examined at 0, 1, 2, 3 and 4 days after incubation. At indicated time‐points, 10 μl (5 mg/ml) of 3‐(4,5‐Dimethylthiazol‐2‐yl)‐2,5‐diphenyltetrazolium bromide (MTT) (Sigma‐Aldrich, St. Louis, USA) was added to each well. After 1‐hr incubation, 150 μl of dimethyl sulfoxide (DMSO) was added for formazan crystals dissolution with a 15‐min incubation time at 37°C. The optical density (OD) was recorded at 490 nm using a microplate reader (Bio‐Rad, Hercules, CA, USA). Cells were seeded into 96‐well plate with 3000 cells/well in triplicate for cell counting at indicated time‐points using Countess II FL Automated Cell Counters (Invitrogen, Carlsbad, CA, USA). Wound‐healing assay {#jcmm13229-sec-0007} ------------------- Cells (5 × 10^5^) were seeded on six‐well plates and scraped firmly with a plastic pipette tip. The cells were washed once to remove cell debris, and fresh serum‐free medium was added. The wound‐healing process was captured at the beginning, 12 and 24 hrs after scratching. Experiments were carried out in triplicate and repeated three times. Transwell migration assay {#jcmm13229-sec-0008} ------------------------- Polycarbonate membrane inserts with 8‐μm pores (Corning Life Sciences, Bedford, MA, USA) were placed in 24‐well cell culture plates. Cells were suspended at a concentration of 1 × 10^5^ cells/ml in 100 μl of serum‐free medium and were plated in the uncoated upper chamber. Foetal bovine serum (10%), used as a chemoattractant, was added to the bottom chamber. After 24 hrs of incubation, those that had migrated to the bottom surface were fixed, stained and scored visually in five random fields under a microscope. Each experiment was performed in replicate, and the mean value was calculated from three independent experiments. Quantitative real‐time reverse transcription PCR (qRT‐PCR) {#jcmm13229-sec-0009} ---------------------------------------------------------- Total RNA was extracted with TRIzol reagent (Takara, Carlsbad, CA, USA) according to the manufacturer\'s protocol. The cDNA was synthesized using the Revert Aid First‐Strand cDNA Synthesis Kit (TransGene, Beijing, China). For qRT‐PCR, each sample was analysed in triplicate. The PCR amplification conditions were as follows: 1 cycle at 95°C for 10 min, 45 cycles at 95°C for 15 s and 60°C for 60 s. The relative expression of mRNA was calculated using the formula 2^−∆∆CT^ [12](#jcmm13229-bib-0012){ref-type="ref"}. Table [S1](#jcmm13229-sup-0001){ref-type="supplementary-material"} lists the primers. Luciferase assay {#jcmm13229-sec-0010} ---------------- Wnt signalling activation was measured by transfection with TOPflash/FOPflash reporter plasmids (Addgene, Cambridge, MA, USA). Cells (5 × 10^4^) were seeded in 24‐well plates and cultured for 12 hrs. Cells were transfected with 100 ng of TOPflash/FOPflash reporter luciferase plasmid and 5 ng of the internal control plasmid pRL‐TK (Progema, Madison, WI, USA) using Lipofectamine 2000 (Invitrogen). The luciferase activity was measured 36 hrs after transfection by Luciferase Assay System (Promega) according to the manufacturer\'s protocol, and the firefly luciferase activity was normalized to the renilla luciferase activity. All experiments were performed three times in triplicate. Western blot {#jcmm13229-sec-0011} ------------ Protein lysates were loaded to each lane for analysis by 12% sodium dodecyl sulphate--polyacrylamide gel electrophoresis (SDS‐PAGE) and were subsequently transferred to nitrocellulose (NC) membranes (Millipore, Darmstadt, Germany). After blocking with 5% non‐fat milk in Tris‐buffered saline with Tween for 1 hr, the membrane was incubated with primary antibody at 4°C overnight. Additional information about the primary antibodies and their dilutions is provided in Table [1](#jcmm13229-tbl-0001){ref-type="table-wrap"}. After three times of 10‐min washes with Tris‐buffered saline with Tween, the membrane was incubated with HRP‐conjugated secondary antibodies (1:5000) (Santa Cruz, CA, USA) for 1 hr at room temperature. Immunoreactive bands were detected by a chemiluminescent detection system (ECL, Pierce, Rockford, IL, USA). The relative protein levels were calculated based on β‐actin protein. ###### Antibody and dilution Antibody Biological source Clone Dilution for Western blot Company ----------------- ------------------- ------------ --------------------------- ------------------------------- Anti‐Med19 Rabbit Polyclonal 1:200 Sigma‐Aldrich, St. Louis, USA Anti‐Wnt2 Rabbit Polyclonal 1:200 Santa Cruz, CA, USA Anti‐β‐catenin Mouse Monoclonal 1:1000 Millipore, Darmstadt, Germany Anti‐E‐cadherin Mouse Monoclonal 1:200 Abcam, Cambridge, UK Anti‐Gsk3β Rabbit Monoclonal 1:5000 Abcam, Cambridge, UK Anti‐Cyclin‐D1 Rabbit Polyclonal 1:200 Santa Cruz, CA, USA Anti‐MMP‐9 Rabbit Polyclonal 1:500 Santa Cruz, CA, USA Anti‐β‐actin Mouse Monoclonal 1:2000 Santa Cruz, CA, USA John Wiley & Sons, Ltd Xenograft model of tumour growth *in vivo* {#jcmm13229-sec-0012} ------------------------------------------ The experimental procedures were approved by the hospital institutional animal care and use committee. Male immune‐deficient BALB/c nude mice (4--6 weeks old) were purchased from Beijing Wei‐tong Li‐hua Laboratory Animals and Technology Ltd, Beijing, China. The cells (1 × 10^6^) were suspended in Matrigel (BD Biosciences, Franklin Lakes, NJ, USA) and subcutaneously implanted into the left flank (control cells) and right flank (shMed19‐transfected cells) of nude mice. Fifteen days after implantation, tumour volumes were measured every 5 days until the mice were killed at day 30. Tumour mass was weighted and tumour volume was calculated using the following formula: *V* (mm^3^) = 1/2 length^2^ × width^2^. Proliferative index was investigated using IHC staining for Ki‐67, and the experimental procedure was performed as above. Statistical analysis {#jcmm13229-sec-0013} -------------------- All statistical analyses were performed by SPSS 19.0 (Chicago, IL, USA). Data are presented as mean ± SD of three independent experiments. Differences between groups were analysed with Student\'s *t*‐test for unpaired observations. Chi‐square test was used to evaluate the results of IHC staining. Mann--Whitney *U*‐test was used to compare the Western blot results of Med19 expression in 15 paired BCa and the adjacent normal tissues. Statistical significance was considered if *P* \< 0.05. Results {#jcmm13229-sec-0014} ======= Correlation between Med19 expression and clinicopathological parameters {#jcmm13229-sec-0015} ----------------------------------------------------------------------- We examined Med19 protein expression in 15 paired BCa tissues and the adjacent normal tissues by Western blot analysis. Med19 has higher expression level in BCa tissues than in the adjacent normal tissues (*P* \< 0.01, Fig. [1](#jcmm13229-fig-0001){ref-type="fig"}A and B), and similar result was observed on the mRNA level of Med19 by qRT‐PCR (*P* \< 0.01, Fig. [1](#jcmm13229-fig-0001){ref-type="fig"}C). In normal tissues, Med19 protein staining was very weak (Fig. [1](#jcmm13229-fig-0001){ref-type="fig"}D and G). However, the staining of Med19 was moderate in low‐grade tumour tissues (Fig. [1](#jcmm13229-fig-0001){ref-type="fig"}E and H) and strong in high‐grade ones (Fig. [1](#jcmm13229-fig-0001){ref-type="fig"}F and I). Table [2](#jcmm13229-tbl-0002){ref-type="table-wrap"} summarizes the association between Med19 protein expression and clinicopathological characteristics. Med19 overexpression was detected in 88 of 167 cases (52.7%). The expression level of Med19, particularly localized in the nucleus and cytoplasm of tumour cells, correlates with advanced stage (*P* = 0.003) and high grade (*P* = 0.001), but did not show a significant association with gender (*P* = 0.738), tumour number (*P* = 0.071) or recurrence (*P* = 0.101). {#jcmm13229-fig-0001} ###### Association of Med19 expression with clinical pathologic characteristics of patients Parameters Number Med19 expression *P* value --------------- -------- ------------------ ----------- ------- Patients 167 79 (47.3) 88 (52.7) Gender 0.738 Male 127 61 (48.0) 66 (52.0) Female 40 18 (45.0) 22 (55.0) Stage 0.003 ≤T1 96 55 (57.3) 41 (42.7) \>T1 71 24 (33.8) 47 (66.2) Grade 0.001 1 42 30 (71.4) 12 (28.6) 2 65 28 (43.1) 37 (56.9) 3 60 21 (35.0) 39 (65.0) Tumour number 0.071 Unifocal 120 62 (51.7) 58 (48.3) Multifocal 47 17 (36.2) 30 (63.8) Recurrence 0.101 Positive 116 50 (43.1) 66 (56.9) Negative 51 29 (56.9) 22 (43.1) John Wiley & Sons, Ltd Knockdown of Med19 inhibited bladder cancer cell growth *in vitro* and *in vivo* {#jcmm13229-sec-0016} -------------------------------------------------------------------------------- To examine the functional role of Med19 in BCa, shRNA‐mediated knockdown of Med19 was achieved by lentiviral transduction of T24, UM‐UC3 and 5637 cells, which was demonstrated by green fluorescent protein (GFP) expression under fluorescence microscopy. The qRT‐PCR results confirmed two specific sequences against Med19 (shMed19, \#1 and \#2) significantly inhibited Med19 mRNA expression in T24, UM‐UC3 and 5637 cells (Fig. [2](#jcmm13229-fig-0002){ref-type="fig"}A). We further confirmed the knockdown effect of Med19 at protein level by Western blot (Fig. [2](#jcmm13229-fig-0002){ref-type="fig"}B). These results suggested successful knockdown of endogenous Med19 expression by Med19 shRNAs in T24, UM‐UC3 and 5637 cells. {#jcmm13229-fig-0002} To evaluate the viability of BCa cells *in vitro*, MTT assay was employed and a strong inhibition of cell proliferation, reflected by the optical density, was observed in the Med19 shRNA‐treated groups compared to the control groups (*P* \< 0.05, Fig. [2](#jcmm13229-fig-0002){ref-type="fig"}C). Comparing the growth curves of BCa cells, we also found that Med19 knockdown by shRNA significantly reduced cell proliferation rates over 4 days *in vitro* (*P* \< 0.05, Fig. [2](#jcmm13229-fig-0002){ref-type="fig"}D). To test the functional role of Med19 in tumour growth *in vivo*, T24 and UM‐UC3 cells expressing shRNA targeting Med19 or control shRNA were subcutaneously inoculated into nude mice. Thirty days later, tumours were harvested, and measurement of tumour volume showed that knocking‐down of Med19 protein leads to lower tumour volume (both *P* \< 0.05, Fig. [2](#jcmm13229-fig-0002){ref-type="fig"}E) and reduced tumour weight compared to control groups (both *P* \< 0.01, Fig. [2](#jcmm13229-fig-0002){ref-type="fig"}F). Furthermore, IHC analysis showed that down‐regulation of Med19 suppressed the expression of proliferation marker Ki‐67 in tumour tissues compared with control shRNA‐treated groups (Fig. [2](#jcmm13229-fig-0002){ref-type="fig"}H). Results of quantitative analysis of Ki‐67‐positive cells (index) in the tumour are summarized in Figure [2](#jcmm13229-fig-0002){ref-type="fig"}H. As compared with the controls, the Ki‐67 index in the tumour tissues was significantly decreased after the Med19 shRNA treatment (both *P* \< 0.01). These findings were consistent with the results of the cell proliferation assay *in vitro*. Knockdown of Med19 inhibited bladder cancer cell migration *in vitro* {#jcmm13229-sec-0017} --------------------------------------------------------------------- Cell migration ability is a critical parameter for metastatic potential of cancer cells. To test whether Med19 contributes to BCa cell migration, wound‐healing and transwell migration assays were performed for cells expressing Med19 shRNA or control shRNA. For wound‐healing assay, 24 hrs after the scratch, the gaps in the control groups were apparently smaller than the ones in the Med19 shRNA groups (Fig. [3](#jcmm13229-fig-0003){ref-type="fig"}A), indicating the role of Med19 in BCa cell migration. Suppression of cell migration in Med19 shRNA‐expressing cells was also observed by the transwell assay, where knockdown of Med19 reduced the number of T24, UM‐UC3 and 5637 cells that migrated into the lower filter (all *P* \< 0.01, Fig. [3](#jcmm13229-fig-0003){ref-type="fig"}B). {#jcmm13229-fig-0003} Med19 knockdown stimulated E‐cadherin expression and suppressed activity of Wnt/β‐catenin pathway {#jcmm13229-sec-0018} ------------------------------------------------------------------------------------------------- Given the effects of Med19 on proliferation and migration of BCa cells, we looked into E‐cadherin and Wnt/β‐catenin pathway, which plays key roles in cell proliferation and migration, and we analysed the expression of E‐cadherin and the activity of Wnt/β‐catenin pathway. We observed an increased protein level of E‐cadherin after Med19 knockdown (Fig. [4](#jcmm13229-fig-0004){ref-type="fig"}A). TOP/FOPflash luciferase reporter assay was used to examine the effect of Med19 on β‐catenin/TCF‐dependent transcriptional activity, as a canonical experiment for the detection of Wnt/β‐catenin signalling activity [14](#jcmm13229-bib-0014){ref-type="ref"}. The TOP/FOPflash reporter activities in T24, UM‐UC3 and 5637 cells with Med19 knockdown by shRNA were significantly inhibited (all *P* \< 0.05, Fig. [4](#jcmm13229-fig-0004){ref-type="fig"}B). In addition, depletion of Med19 reduced the protein expression of Wnt2 and β‐catenin. However, protein level of Gsk3β was elevated (*P* \< 0.05, Fig. [4](#jcmm13229-fig-0004){ref-type="fig"}C). We then examined the protein expression levels of matrix metalloproteinase‐9 (MMP‐9) and Cyclin‐D1, which are classic downstream targets of the Wnt/β‐catenin signalling pathway [15](#jcmm13229-bib-0015){ref-type="ref"}, [16](#jcmm13229-bib-0016){ref-type="ref"}, [17](#jcmm13229-bib-0017){ref-type="ref"}, [18](#jcmm13229-bib-0018){ref-type="ref"}. Indeed, depletion of Med19 had a potent inhibitory effect on the expression of MMP‐9 and Cyclin‐D1 (*P* \< 0.05, Fig. [4](#jcmm13229-fig-0004){ref-type="fig"}D). We also confirmed these data at mRNA level (Fig. [4](#jcmm13229-fig-0004){ref-type="fig"}A, C and D). Therefore, these results suggest that Med19 knockdown stimulated E‐cadherin expression and decreased the activity of Wnt/β‐catenin signalling pathway, possibly leading to the suppression of BCa cell proliferation and migration. {#jcmm13229-fig-0004} Inhibition of BCa cell proliferation and migration by Med19 knockdown can be rescued by activating Wnt/β‐catenin signalling {#jcmm13229-sec-0019} --------------------------------------------------------------------------------------------------------------------------- To further confirm the role of Med19 in regulating the activity of the Wnt/β‐catenin signalling, lithium chloride (LiCl, sigma) was used to activate the Wnt/β‐catenin pathway. As demonstrated in Fig. [5](#jcmm13229-fig-0005){ref-type="fig"}A, LiCl significantly increased protein level of β‐catenin in Med19 shRNA‐treated cells. As expected, LiCl markedly increased the expression of Cyclin‐D1 and MMP‐9 in BCa cells along with Med19 knockdown (Fig. [5](#jcmm13229-fig-0005){ref-type="fig"}B). Furthermore, we examined the biological mechanism of Med19 in BCa cells. Through MTT and transwell assay, we found that activation of Wnt/β‐catenin signalling by LiCl promoted cell proliferation and migration in Med19 shRNA‐treated BCa cells (Fig. [5](#jcmm13229-fig-0005){ref-type="fig"}C and D). Collectively, our results indicate that Wnt/β‐catenin signalling is a functional mediator of Med19‐related proliferation and migration in BCa cells. {#jcmm13229-fig-0005} Discussion {#jcmm13229-sec-0020} ========== Med19 is a key subunit of the mammalian Mediator complex. Due to its critical role in stabilizing the entire Mediator complex, Med19 is indispensable for the regulation of gene transcription. In the absence of Med19, the binding affinity of the Mediator with RNA polymerase II is decreased and phosphorylation of the carboxyl terminal domain (CTD) *via* the transcription factor II H fails to occur [19](#jcmm13229-bib-0019){ref-type="ref"}, [20](#jcmm13229-bib-0020){ref-type="ref"}. Med19 also plays an important role in tumour formation and progression. Wen and coworkers suggested that Med19 might promote invasive behaviour and bone metastasis of BCa cells by stimulating the expression of BMP‐2 [11](#jcmm13229-bib-0011){ref-type="ref"}. In breast cancer, osteosarcoma and hepatocellular carcinoma, cell proliferation capability was significantly reduced *in vitro* and *in vivo* following knockdown of Med19 [6](#jcmm13229-bib-0006){ref-type="ref"}, [8](#jcmm13229-bib-0008){ref-type="ref"}, [10](#jcmm13229-bib-0010){ref-type="ref"}. These results indicate that Med19 is a human oncogene that might represent a promising new target for therapeutic intervention. The present study demonstrates that Med19 is highly expressed in BCa tissues and Med19 expression levels are positively correlated with clinical stage and pathological grade of BCa (Table [2](#jcmm13229-tbl-0002){ref-type="table-wrap"}). It is concluded that increased expression of Med19 might promote BCa development. We then applied a lentivirus‐mediated shRNA targeting assay to knockdown Med19 in BCa cells. The results showed that BCa cell proliferation capacity was remarkably decreased *in vitro* and *in vivo* following Med19 knockdown, which might be due to the down‐regulation of Cyclin‐D1 signalling, an important regulator for cell cycle progression [21](#jcmm13229-bib-0021){ref-type="ref"}. In addition, Med19 inhibition can significantly suppress BCa cell migration by wound‐healing and transwell migration assays (Fig. [3](#jcmm13229-fig-0003){ref-type="fig"}A and B). Interestingly, the down‐regulation of MMP‐9 and up‐regulation of E‐cadherin in our results may explain why depletion of Med19 significantly impaired BCa cell migration capacity. In line with previous findings, both increased expression of MMP‐9 and decreased expression of E‐cadherin correlate with bladder cancer metastasis [22](#jcmm13229-bib-0022){ref-type="ref"}, [23](#jcmm13229-bib-0023){ref-type="ref"}. The Wnt signalling pathway regulates cell proliferation and differentiation, and activation of Wnt pathway is involved in the pathogenesis of BCa [23](#jcmm13229-bib-0023){ref-type="ref"}. Mediator complex is commonly seen as a molecular bridge that connects DNA‐binding transcription factors to RNA polymerase II, which regulate different signalling events into the appropriate response at level of RNA transcription [3](#jcmm13229-bib-0003){ref-type="ref"}, [4](#jcmm13229-bib-0004){ref-type="ref"}, [24](#jcmm13229-bib-0024){ref-type="ref"}. Some subunits of Mediator complex have been shown to play an essential role in several important signalling pathways related to carcinogenesis, including TGF‐β/SMAD, AKT/MAPK and NF‐κB signalling [25](#jcmm13229-bib-0025){ref-type="ref"}, [26](#jcmm13229-bib-0026){ref-type="ref"}. Subunit Med12 interacts with β‐catenin, and it has been identified as an important transducer of Wnt/β‐catenin signalling in the development of colorectal tumours [27](#jcmm13229-bib-0027){ref-type="ref"}. Imberg‐Kazdan *et al*. [7](#jcmm13229-bib-0007){ref-type="ref"} recently identified Med19 as a regulator that affects androgen receptor (AR) signals and proliferation of AR‐expressing prostate cancer. However, the mechanism Med19 interacts with other signalling pathways leads to cancer development remains poorly understood, and further investigation is needed to determine the function of Med19 in BCa initiation and progression. Aberrant activation of the Wnt/β‐catenin signalling pathway, resulting in uncontrolled cell proliferation and impaired migration, is a common event in human cancer development [23](#jcmm13229-bib-0023){ref-type="ref"}, [28](#jcmm13229-bib-0028){ref-type="ref"}. Wnt/β‐catenin signalling activation inhibits glycogen synthase kinase‐3β (GSK3β) phosphorylation of β‐catenin, leading to its cytosolic accumulation [28](#jcmm13229-bib-0028){ref-type="ref"}. It has been shown that E‐cadherin inhibits the Wnt/β‐catenin pathway through its binding at C‐terminal domain of β‐catenin on the cellar surface, thereby sequestering β‐catenin from the cytoplasmic pool [29](#jcmm13229-bib-0029){ref-type="ref"}. We found the suppression of Med19 enhanced expression of E‐cadherin by both qRT‐PCR and Western blot assays. In the Wnt pathway, β‐catenin acts as a key transcriptional coactivator and transmits extracellular signals for the activation of important downstream target genes such as MMP‐9 and Cyclin‐D1 [15](#jcmm13229-bib-0015){ref-type="ref"}, [16](#jcmm13229-bib-0016){ref-type="ref"}, [17](#jcmm13229-bib-0017){ref-type="ref"}, [18](#jcmm13229-bib-0018){ref-type="ref"}, [30](#jcmm13229-bib-0030){ref-type="ref"}, [31](#jcmm13229-bib-0031){ref-type="ref"}. Our results showed that the expression of Wnt2 and active β‐catenin, core members of the Wnt/β‐catenin signalling pathway, were decreased following Med19 knockdown, and the expression of GSK3β, a molecule indispensable for maintaining low levels of β‐catenin, was significantly increased. Furthermore, MMP‐9 and Cyclin‐D1 significantly decreased after attenuation of Med19. To further confirm the role of Med19 in regulating the activity of the Wnt/β‐catenin pathway, LiCl was used to activate the Wnt/β‐catenin pathway. Our data showed that the inhibition of BCa cell proliferation and migration by Med19 knockdown can be rescued by cells treated with LiCl. TOP/FOPflash luciferase reporter assay also confirmed that Med19 knockdown inhibited the activation of Wnt/β‐catenin pathway. Together, these data indicate that Med19 participates in BCa cell survival through modulating the Wnt/β‐catenin signalling pathway. Mediator complex can be divided into four different modules termed the head, middle, tail and cyclin‐dependent kinase 8 (CDK8) module [3](#jcmm13229-bib-0003){ref-type="ref"}, [4](#jcmm13229-bib-0004){ref-type="ref"}. Med19 is a component of head module and plays an important role in regulating gene expression as a transcription coactivator. Our finding shows that Med19 expression affects the BCa cell proliferation and migration, which correlates with the Wnt/β‐catenin pathway. To the best of our knowledge, this is the first report to present effects of Med19 on cancer progression induced by Wnt/β‐catenin pathway. However, the mechanism of how Med19 interacts with Wnt/β‐catenin pathway is still unclear; thus, it will be necessary to perform further investigations in future studies. In conclusion, our study suggests that high expression of Med19 is closely associated with aggressive characteristics of BCa, and provide good evidence to show the involvement of Med19 in both proliferation and migration of BCa cells, which directly regulates Wnt/β‐catenin signalling pathway. These findings strongly indicate that Med19 may act as an oncogene in BCa development, thereby highlighting Med19 as a potential therapeutic target for BCa treatment. Conflict of interest {#jcmm13229-sec-0022} ==================== The authors confirm that there are no conflict of interests. Author contribution {#jcmm13229-sec-0023} =================== Hejia Yuan and Jitao Wu designed and conducted the experiments and analysed data. Shengqiang Yu was involved in the design, animal experiment and drafting of the manuscript. Yuanshan Cui and Changping Men collected the data of patients with bladder cancer. Diandong Yang performed Western blot and participated in the animal experiment. Hejia Yuan and Zhenli Gao were responsible for cell culture. Zhe Zhu and Jitao Wu wrote and revised the manuscript. Supporting information ====================== ###### **Table S1** Primer sequence for qRT‐PCR ###### Click here for additional data file. This work was supported by grants from the National Nature Science Foundation of China (Nos. 81572835; 81302234), Nature Science Foundation of Shandong Province (No. ZR2015HM021). [^1]: Equal study contribution.

Globally, respiratory syncytial virus (RSV) is a leading cause of acute lower respiratory infection (ALRI) including bronchiolitis and pneumonia among young children \[[@R1],[@R2]\]. Studies indicate that most children are infected by RSV in the first two years of life with infants bearing the highest rates of RSV--associated lower respiratory illness \[[@R3],[@R4]\]. A recent meta--analysis estimated 3.4 million hospitalizations and 66 000--199 000 RSV--associated deaths among children \<5 years of age with ALRI, with 99% of the deaths occurring in developing countries \[[@R5]\]. While there are some studies on burden of pneumonia and viral etiology in India and other developing countries \[[@R2],[@R6]\], recent data from a community--based and hospital--based studies have further emphasized the importance of RSV among children \<5 years of age \[[@R7]-[@R9]\]. Systematic data are needed to better understand seasonality, burden and mortality associated with RSV infection in children. RSV infection has a different clinical presentation, age distribution, risk factors, and seasonality compared to influenza infection and requires studies specifically designed to detect and evaluate RSV \[[@R3],[@R4]\]. For example, RSV illness may present without fever particularly among infants, whereas influenza is more likely to present as a febrile illness and thus, fever may be included in the case definitions \[[@R10]\]. However, existing surveillance networks for influenza, with protocols and case definitions designed for influenza have also often been used to generate burden estimations for RSV \[[@R11]\]. Thus, there is a need to identify appropriate case--definitions for epidemiologic field studies to accurately estimate the RSV burden among children. The presence of a broad platform to estimate the rates of hospitalized influenza which captured all--cause hospitalization in a well--defined population with health demographic surveillance system \[[@R12]\] enabled us to evaluate the sensitivity and specificity of different case definitions and RSV burden among children \<5 years in a rural setting in northern India. The current paper utilizes data from this hospital--based surveillance study to evaluate case definitions for RSV detection and the impact of choice of case definitions on RSV hospitalization rate estimates among children aged \<5 years in a rural setting in India. METHODS ======= Study site ---------- The Ballabgarh Health and Demographic Surveillance System (HDSS) site is about 40 km south of Delhi and comprised a population of about 90 000 in June 2011 including 9500 children 0--59 months of age in 28 villages \[[@R13]\]. Based on health utilization survey of the site population conducted in April 2009, three public hospitals and 30 private facilities (ranging in size from 5--35 beds) in Ballabgarh and Faridabad towns were included for daily surveillance for patients from the catchment area seeking inpatient care \[[@R12],[@R14]\]. Immunization coverage for EPI vaccines (BCG, DPT, OPV, Hepatitis B and Measles) provided through public health facilities was \>95% in the study villages \[[@R13]\]; Hib vaccine was introduced into public health program in 2012--2013. Coverage for pneumococcal and rota vaccines are not known but likely to be low as they are available only in private facilities. Enrolment and data collection ----------------------------- During July 2009 -- December 2012, hospital--based daily surveillance--enrolled children aged 0--59 months from Ballabgarh--HDSS area who were hospitalized overnight with any acute medical illness or acute exacerbation of chronic illness at participating medical facilities \[[@R8],[@R12]\]. Trained study physicians collected data using a standardized form on demographics, medical history and clinical symptoms by interview of the caregivers, and extracted data on clinical signs at admission from the medical record followed by clinical examination of cases for additional clinical information. Study definitions ----------------- The presence of fever or key respiratory signs or symptoms was determined among all hospitalized children aged 0--59 months. Fever was defined as either measured temperature \>38.0°C at admission or parental report of fever because antipyretic use is known to be common in the study community \[[@R15]\]. Key respiratory symptoms or signs were defined as parental report of cough or fast breathing or physician exam findings of tachypnea, crepitation, wheezing, nasal flaring, chest in--drawing, grunting, or stridor. Tachypnea was defined based on the definition used by the Indian Integrated Management of Neonatal and Childhood Illness (IMNCI) as ≥60 breaths/min in children aged 0--2 months of age, ≥50 breaths/min in children aged 2--12 months of age, and ≥40 breaths/min in children aged 12 months --5 years of age \[[@R16]\]. Data on clinical signs and symptoms were used *post hoc* to classify each patient using standard case definitions specified by WHO (May 2011, see [**Box 1**](#B1){ref-type="boxed-text"}), ie, acute respiratory infection (ARI), severe acute respiratory infection (SARI), and influenza--like illness (ILI), and evaluated them for RSV positivity. ###### **Case definitions for respiratory syndromes \[**[@R17]**,**[@R18]**\]** **ARI** -- acute respiratory infection (WHO, 2011): Acute onset of at least one of the following four respiratory symptoms: cough or sore throat or shortness of breath or coryza and a clinician's judgment that illness is due to infection. **ILI** -- influenza--like illness (WHO, 2011): An acute respiratory illness with onset during the last 7 days with measured temperature ≥38°C, AND cough. (Dropped sore throat). **ILI** (old): Sudden onset of fever (\>38°C) with cough or sore throat, in absence of other diagnoses. **SARI** -- severe ARI (WHO, 2011): An acute respiratory illness with onset during the previous 7 days requiring overnight hospitalization that includes history of fever or measured fever of ≥38°C, AND cough, AND shortness of breath or difficulty breathing. **SARI** (old): Meets ILI (old) definition (sudden onset of fever (\>38°C) with cough or sore throat) and has shortness of breath or difficulty breathing and requires overnight hospitalization. **Pneumonia** (IMCI): Fast breathing or chest indrawing **Severe pneumonia** (IMCI): General danger signs -- Not able to drink, persistent vomiting, convulsions, lethargic or unconscious, stridor, or severe malnutrition Specimen collection and laboratory methods ------------------------------------------ Nasal and throat samples were collected by a study nurse from all enrolled patients within 24 hours of admission to the hospital using polyester swabs; in infants only nasal swabs were collected. The swabs were placed immediately into viral transport media and transported on ice to laboratory on the same day for processing. Specimens were tested for RSV and influenza using US Centers for Disease Control and Prevention (CDC) real--time reverse transcription polymerase chain reaction (rRT--PCR) protocols, as described previously \[[@R8],[@R12]\]. Data analysis ------------- We assessed different signs and symptoms associated with RSV positivity using bivariate analysis for different age--groups and backward stepwise logistic regression adjusted for age--groups. We also assessed the ability of standard case definitions (ARI, SARI, ILI) for respiratory illness to capture RSV--associated hospitalizations by calculating sensitivity and specificity for each case definition using all RSV positive hospitalized patients as the gold standard. We assessed the impact of standard case definitions on average annual incidences of RSV--associated hospitalizations using available 3 calendar years' data from 2010 to 2012. The population of June 2011 in the HDSS was considered the mid--term population denominator for calculations. Annual health utilization surveys were used to estimate the average proportion of hospitalization in enrolled facilities \[[@R14]\]. The annual hospitalization rates based on enrollment were adjusted for missed hospitalizations in non--study facilities and were multiplied by the positivity rate of the viruses to get an estimate of virus specific hospitalization rates. We calculated incidence rates for four age groups (0--5 months, 6--11 months, 12--23 months and 24--59 months) for RSV and influenza because clinical manifestations, as well as viral etiologies are likely to be different in these age groups. The average annual incidence was calculated separately for each case definition among all medical and respiratory admissions, to evaluate the effect of using different screening definitions on the estimations of RSV--associated hospitalizations. The data analysis was done using STATA 12 (College Station, Texas, USA) \[[@R17]\] and Micosoft Excel. A *P*--value of \<0 · 05 was considered statistically significant for all analyses. The 95% confidence intervals (CI) for odds ratios, sensitivity and specificity were calculated. RESULTS ======= Background characteristics -------------------------- During the study period, 505 children aged 0--59 months from the HDSS area hospitalized with acute medical illness in the health facilities under surveillance were enrolled; of these 79.6% (402/505) were aged 0--24 months and 71.7% (362/505) were males ([**Table 1**](#T1){ref-type="table"}). RSV was detected in 82 (16%) hospitalized patients with 89% (73/82) of detections among children \<2 years old (*P* \< 0.001). There was no significant difference in gender, time from symptom onset to specimen collection or any underlying medical condition among RSV negative and positive children (with the exception of chronic diarrhea observed among RSV negative children, data not shown). The RSV detections among hospitalized children occurred with seasonal peaks between September--October and then again in January--February of each year (data not shown). ###### Characteristics of children \<5 years of age enrolled for all cause hospitalization in Ballabgarh, Haryana, India from July 2009 -- December 2012 Characteristics (n, %) Total (n = 505) RSV+ (n = 82) RSV-- (n = 423) *P--*value ------------------------------------------------------- ----------------- --------------- ----------------- ------------ **Age group (months):** 0--5 114 (22.6) 35 (42.7) 79 (18.7) \<0.001 6--11 146 (28.9) 16 (19.5) 130 (30.7) 12--23 142 (28.1) 22 (26.8) 120 (28.4) 24--35 43 (8.5) 4 (4.9) 39 (9.2) 36--59 60 (11.9) 5 (6.1) 55 (13) **Sex:** Female 143 (28.3) 24 (29.3) 119 (28.1) 0.834 Male 362 (71.7) 58 (70.7) 304 (71.9) **Time from symptom onset to specimen collection:\*** 0--2 days 127/468 (27.1) 16/79 (20.2) 111/389 (28.5) 0.497 3--4 days 135/468 (28.8) 26/79 (32.9) 109/389 (28.0) 5--7 days 116/468 (24.8) 21/79 (26.6) 95/389 (24.4) 8--10 days 39/468 (8.3) 9/79 (11.4) 30/389 (7.7) ≥11 days 51/468 (10.9) 7/79 (8.9) 44/389 (11.3) **\***Data on 37 cases, including 3 RSV--positive cases, were missing. Clinical characteristics ------------------------ Among the enrolled children, 347 (68.7%) had some respiratory illness. Further, those with symptoms of cough (OR = 5.3, 95% CI 2.8--10.1) and fast breathing (OR = 3.9, 95% CI 2.4--6.3) were more likely to test positive than negative for RSV ([**Table 2**](#T2){ref-type="table"}). The presence of signs of wheeze, chest in--drawing, tachypnea, and crepitation had significantly higher odds of being RSV positive vs RSV--negative based on bivariate comparisons. Other less commonly seen signs of respiratory distress, ie, nasal flaring, grunting, accessory muscle usage were also significantly associated with being RSV positive. Stepwise backward logistic regression ([**Table 3**](#T3){ref-type="table"}) analysis of all clinical features identified the presence of cough, fast--breathing, crepitation and hypoxia as significant predictors for RSV--associated hospitalization, while history of fever and diarrhea were significantly associated with non--RSV--associated hospitalization. ###### Age--specific clinical signs and symptoms significantly associated with laboratory--confirmed respiratory syncytial virus infection among hospitalized patients \<5 y of age, in Ballabgarh, India, July 2009 -- December 2012; (n = 505)\* 0--5 months 6--23 months 24--59 months 0--59 months ------------------------ ------------------- ----------------- ------------------- ----------------- ------------------ ----------------- ------------------- ----------------- **RSV+ (n = 35)** **OR (95% CI)** **RSV+ (n = 38)** **OR (95% CI)** **RSV+ (n = 9)** **OR (95% CI)** **RSV+ (n = 82)** **OR (95% CI)** **Symptoms:** Fever 27 (77.1) 0.6 (0.2--1.6) 32 (84.2) 0.8 (0.3--2.2) 9 (100.0) -- 68 (82.9) 0.8 (0.4--1.5) Cough 32 (91.4) 4.4 (1.2--15.7) 32 (84.2) 6.5 (2.6--16.0) 6 (66.7) 1.6 (0.4--6.8) 70 (85.4) 5.3 (2.8--10.1) Breathing difficulty 20 (57.1) 3.5 (1.5--7.9) 15 (39.5) 3.3 (1.6--6.9) 2 (22.2) 1.5 (0.3--8) 37 (45.1) 1.8 (1.1--2.7) Nasal discharge 13 (37.1) 1.4 (0.6--3.1) 20 (52.6) 2.3 (1.1--4.5) 4 (44.4) 2 (0.5--8) 37 (45.1) 1.8 (1.1--2.9) Sore throat (\>2years) 0 (0) -- 0 (0) -- 2 (22.2) 1.5 (0.3--8) 2 (16.7) 1.4 (0.3--7.1) Ear discharge 1 (2.9) 1.1 (0.1--12.9) 4 (10.5) 2.3 (0.7--7.6) 0 (0) -- 5 (6.1) 1 (0.4--2.7) Fast breathing 18 (51.4) 2.9 (1.3--6.7) 18 (47.4) 6.1 (2.9--12.8) 3 (33.3) 3.8 (0.8--17.3) 39 (47.6) 3.9 (2.4--6.3) Lethargy 9 (25.7) 1.5 (0.6--3.9) 3 (7.9) 0.3 (0.1--0.9) 3 (33.3) 1.2 (0.3--5) 15 (18.3) 0.7 (0.4--1.2) Refusal to feed 15 (42.9) 1 (0.5--2.3) 10 (26.3) 0.6 (0.3--1.3) 3 (33.3) 0.8 (0.2--3.3) 28 (34.2) 0.8 (0.5--1.3) Seizure 2 (5.7) 0.9 (0.2--4.9) 0 (0) -- 2 (22.2) 6.4 (1--41.4) 4 (4.9) 1.6 (0.5--4.7) Unconsciousness 2 (5.7) 0 (0--0) 0 (0) -- 0 (0) -- 2 (2.4) 1.5 (0.2--11) Vomiting 10 (28.6) 0.4 (0.2--0.9) 22 (57.9) 0.4 (0.2--0.8) 4 (44.4) 0.5 (0.1--2) 36 (43.9) 0.4 (0.2--0.6) Diarrhea 13 (37.1) 0.5 (0.2--1.2) 16 (42.1) 0.2 (0.1--0.4) 1 (11.1) 0.2 (0--1.7) 30 (36.6) 0.3 (0.2--0.5) Rash 1 (2.9) 2.3 (0.1--37.8) 0 (0) -- 0 (0) -- 1 (1.2) 0.3 (0--2.6) Jaundice 1 (2.9) 1.1 (0.1--12.9) 0 (0) -- 0 (0) -- 1 (1.2) 0.7 (0.1--4.2) **Signs:** Fever 1 (2.9) 0.2 (0--1.3) 10 (26.3) 1.8 (0.8--4) 6 (66.7) 5 (1.2--21.3) 17 (20.7) 1.1 (0.6--2) Stridor 4 (11.4) 2.4 (0.6--10.3) 5 (13.2) 9.3 (2.4--36.5) 0 (0) 0 (0--0) 9 (11) 5.7 (2.2--14.8) Nasal flaring 5 (14.3) 2.4 (0.6--8.9) 4 (10.5) 9.7 (2.1--45.2) 0 (0) 0 (0--0) 9 (11) 5.1 (2--12.9) Chest in--drawing 13 (37.1) 5.1 (1.2--13.9) 6 (15.8) 3.7 (1.3--10.6) 0 (0) 0 (0--0) 19 (23.2) 5.7 (2.9--11.3) Grunting 5 (14.3) 3 (0.8--12.1) 4 (10.5) 7.2 (1.7--30.3) 0 (0) 0 (0--0) 9 (11) 5.6 (2.2--14.7) Accessory muscle use 5 (14.3) 4.2 (0.9--18.8) 1 (2.6) 0.8 (0.1--6.7) 0 (0) 0 (0--0) 6 (7.3) 2.5 (0.9--6.8) Crepitation 21 (60) 2.9 (1.3--6.6) 17 (44.7) 5.0 (2.4--10.3) 4 (44.4) 5.5 (1.3--23.2) 42 (51.2) 5 (3--8.2) Wheeze 24 (68.6) 5 (2.1--11.8) 10 (26.3) 2.8 (1.2--6.4) 2 (22.2) 2.2 (0.4--11.7) 36 (43.9) 4.5 (2.7--7.5) Tachypnoea 25 (71.0) 1.4 (0.6--3.3) 16 (42.1) 1.9 (0.9--3.8) 1 (11.1) 0.6 (0.1--4.8) 42 (51.2) 2.2 (1.4--3.5) Hypoxia\* 8 (22.9) 1.5 (0.6--4) 8 (22.9) 2.6 (1.0--6.6) 1 (11.1) 2.8 (0.3--28.3) 17 (20.7) 2.5 (1.3--4.8) OR -- odds ratio, CI -- confidence interval \*Defined as oxygen saturation \<90% on room air or \<95% on oxygen therapy. ###### Clinical predictors of respiratory syncytial virus associated hospitalization among children aged \<5 y by stepwise backward logistic regression adjusted for age--groups\* Symptoms and signs Odds ratio (95% CI) *P* \> z --------------------------- --------------------- ---------- Cough 1.85 (1.13--0.46) 0.013 History of fever 0.4 (0.23--0.11) 0.001 History of fast breathing 3.44 (2.26--0.73) \<0.001 Diarrhea 0.62 (0.41--0.12) 0.021 Inability/refusal to feed 0.57 (0.39--0.1) 0.004 Unconsciousness 0.07 (0.01--0.07) 0.005 Nasal flaring 3.13 (1.41--1.27) 0.005 Stridor 2.73 (1.22--1.12) 0.014 Accessory muscle use 0.22 (0.08--0.11) 0.003 Crepitation 2.79 (1.81--0.61) \<0.001 Hypoxia 2.71 (1.58--0.74) \<0.001 CI -- confidence interval \*Age groups: 0--5 months, 6--11 months, 12--23 months and 24--59 months. Case definitions ---------------- We then examined the sensitivity and specificity of standard case definitions (ARI, SARI, ILI) for detection of RSV--associated hospitalization ([**Figure 1**](#F1){ref-type="fig"}). Among the standard case definitions, ARI had the highest sensitivity (87.8%) and specificity (40%) based on receiver--operating characteristics. All other case definitions (ILI, and SARI, (both old and revised 2011 versions) had lower sensitivity and variable specificity, with older the definition of SARI showing higher specificity). We also evaluated different clinical syndromes including IMCI definitions of pneumonia/severe pneumonia \[[@R18]\] and other syndromes. The sensitivity of IMCI either pneumonia or severe pneumonia was high (75.6%) but specificity was low (31.9%). Among other combination of symptoms and signs, we found history of cough or crepitation along with presence of any one of following-- history of fast--breathing, breathing difficulty, nasal discharge, sore--throat, chest--in--drawing, wheeze or hypoxia, had high sensitivity (81 · 7%) and specificity (60%). {ref-type="boxed-text"} for case definitions details of each syndrome.](jogh-05-020419-F1){#F1} Burden of RSV in children ------------------------- The average annual incidence of RSV--associated hospitalizations was found to be higher at 7 · 4 (95% CI: 4.9--10 · 5) per 1000 child--years in those 0--23 months as compared to 0.5 (95% CI 0.1--1.5) among 24--59 months population signifying that most of burden of RSV--associated hospitalization is among children under two years of age. Further breakdown of the age--specific annual incidence of RSV--associated hospitalization per 1000 children revealed that the highest rate occurred in young infants 0--5 months (OR = 15.2, 95% CI 8.3--26.8), followed by 6--23 months (OR = 5.3, 95% CI 3.2--8.7) with comparable rates for the 6--11 months and 12--23 months age groups ([**Figure 2**](#F2){ref-type="fig"}). Incidence rates for influenza were lower across all age groups. {#F2} Impact of case definitions on RSV burden ---------------------------------------- We assessed the impact of the use of different standard case definitions on RSV burden estimates by comparing the definitions with RSV hospitalization based on all--cause hospitalization ([**Table 4**](#T4){ref-type="table"}). Use of the ARI case definition among hospitalized children would have detected 90% and 86% of the RSV--associated hospitalization rates in children aged \<2 years and \<5 years, respectively. In contrast, use of definitions which require presence of fever and most commonly used for influenza surveillance platforms, ie, SARI or ILI definitions, (both old and revised) would have under estimated RSV burden by as much as 50--85% in both \<2 as well as \<5--year age groups ([**Table 4**](#T4){ref-type="table"}). ###### Effect of screening case definition on respiratory syncytial virus (RSV) associated annual hospitalization rates (2010--2012)\* Population under surveillance Under--2 years (N = 3956) Under--5 years (N = 9740) ------------------------------- ----------------------------- ----------------------------------- -------------------------- ------------------------------ ----------------------------- ----------------------------------- --------- ------------------------------ **Case definitions** **No. met case definition** **No. of RSV positive cases (%)** **Incidence Rate (IR)†** **IR under--estimation (%)** **No. met case definition** **No. of RSV positive cases (%)** **IR†** **Under--estimation IR (%)** All Medical 386 67 (17) 7.4 NA 484 74 (15) 3.2 NA ARI (WHO) 239 60 (25) 6.6 --10% 299 64 (21) 2.8 --14% SARI (WHO) 90 34 (38) 3.8 --49% 107 35 (33) 1.5 --53% SARI (Old) 35 10 (29) 1.1 --85% 41 11 (27) 0.5 --85% ILI (WHO) 37 10 (27) 1.1 --85% 54 14 (26) 0.6 --81% ILI (Old) 106 26 (25) 2.9 --61% 139 32 (23) 1.4 --57% ARI -- acute respiratory illness, SARI -- severe acute respiratory illness, ILI -- influenza--like illness, WHO -- World Health Organization \*For incidence rate calculations data for full calendar years were used. Data imputed for 6 weeks surveillance gap between 31 January 2012 and 13 March 2012. †Per 1000 age--specific population. DISCUSSION ========== The uniqueness of this study based on a comprehensive surveillance system designed to capture all--cause hospitalization in a well--defined population, together with the availability of highly sensitive molecular testing for RSV allowed us to estimate RSV--associated the hospitalization rates among children aged \<5 years in rural northern India \[[@R8],[@R12]\]. Most studies estimating RSV--associated burden rely on existing surveillance platforms for influenza in developing countries \[[@R11],[@R19]\]. A very important aspect of our study was that we were able to evaluate the impact of using different case definitions on burden estimation. We provided evidence that the WHO--defined ARI case definition has the highest sensitivity for RSV--associated hospitalization and demonstrate the limitations of definitions like ILI and SARI commonly used for influenza surveillance. We found that testing only children meeting the SARI and ILI definitions would have under--estimated the burden of RSV--associated hospitalization by almost 50--85%, although specificity would have been significantly higher with the latter case definition. Several case definitions, including ARI, have been used in studies for RSV burden estimation in many countries \[[@R11],[@R19]-[@R22]\]. This highlights the importance of the use of a sensitive case definition for surveillance of RSV to avoid underestimation of the burden. The all--cause hospitalization surveillance also allowed us to compare the symptoms and signs of hospitalized children with or without RSV and thereby identify the clinical predictors for RSV--associated hospitalization in children; this would not have been possible if only standard case definitions were used to identify potential RSV patients. We found that the presence of cough, fast--breathing, crepitation and hypoxia are independent predictors of RSV infection. Of note, Durani et al. (2008) in their study among children hospitalized with ARI found the combination of cough, wheezing and retractions to be good clinical predictor for RSV infection \[[@R23]\]. We found that even though fever is a common presenting symptom and sign among children being hospitalized, it is not a good predictor of RSV--associated hospitalization in this population. The use of history or presence of fever in screening case definitions lowers the sensitivity of the definition. We also observed that two--thirds of hospitalized patients had some respiratory symptoms, suggesting that a very high proportion of hospitalizations are due to respiratory symptoms in rural India. This observation corroborates previous findings that ARI is a significant cause of morbidity in the developing world \[[@R4],[@R24],[@R25]\]. We found substantial incidence of RSV--associated hospitalization in the study community especially among \<2--year old children. The RSV--associated incidence of hospitalization per 1000 child--years was 3.2 among \<5--year children, and 7.4 among \<2--year children, with highest incidence rate of 15.2 per 1000 child years among infants 0--5 months, which is similar to findings of an earlier community--based study from this area \[[@R7]\]. RSV--associated hospitalization rates among children \<5years observed in our study were also comparable to what has been observed in Kenya (2.9/1000 child--years) and Guatemala (2--13.7/1000 child--years), although lower than in Thailand (9.8/1000 child--years), Indonesia (34/1000 child--years), Nigeria (94/1000 child--years) \[[@R20]-[@R22],[@R26]\]. The highest risk group for RSV--associated burden was infants 0-- to 5--month old, which was also observed in Thailand (15.4/1000 child--years), Indonesia (41/1000 child--years), Hong Kong (\<6m, 23.4--31.1/1000 child--years), Guatemala (5.9--45.9/1000 child--years), Kenya (11.0/1000 child--years) and Nigeria (116/1000 child--years). Studies from Brazil, USA and Korea have established that infants are at high risk of RSV--associated burden both in terms of incidence in community and proportion of hospitalization \[[@R19]-[@R22],[@R27]-[@R31]\]. Even though the rates of RSV--associated hospitalization vary in different countries, most of the burden is observed among children aged \<2 years, therefore studies focusing on RSV--associated morbidity and mortality or high--risk group for RSV infections may consider children aged \<2 years. Also, prioritizing this age--group for any preventive measure would likely have profound effect on prevention of RSV--associated hospitalization and deaths in India and other developing countries \[[@R32],[@R33]\]. The study's limitations include first that it was designed to address influenza--associated burden, so data on variables such as gestation at birth were not collected; this data might have allowed us to also understand some of the risk--factors for RSV infection. Second, the active surveillance at health facilities (which were almost 30 plus facilities) did not capture all hospitalizations for the denominator population, and we had to make adjustments in rates of hospitalization using HDSS survey results. It is plausible that children seeking care at participating hospitals may be different from those who did not seek care in these hospitals, thus biasing the incidence rates for RSV. Third, there is a possibility that we have under--estimated the burden of RSV--associated hospitalization as some children might not have been hospitalized in spite of being diagnosed with severe respiratory illness \[[@R11],[@R24]\]. Despite these limitations, we believe that this study enabled us to understand the effect of surveillance case definitions on population--based rates of RSV hospitalization in northern rural India. However, due to the study design where we captured all cause medical hospitalization instead of ILI or SARI, we are in unique position to not only address the burden of RSV, but also assess various screening case definition for RSV--associated hospitalization. Analysis of this broad platform for RSV case definition assessment allowed us to recommend the WHO--defined ARI case definition as the most appropriate screening case definition for RSV among those considered. CONCLUSION ========== In conclusion, we observed that RSV is a substantial cause of hospitalization among children aged \<2 years, and especially among infants aged \<6 months. This is true regardless of the screening definition used, even though rates may be underestimated if an insensitive screening definition is used. These data will help support public health strategies and interventions, targeting young children to reduce the overall RSV--associated morbidity and mortality among children in the developing world. We thank Marika Iwane, Eduardo Azziz--Baumgartner, and Marc--Alain Widdowson for their comments and suggestions for design of the study. We also thank Drs. Akhilesh C Mishra, Marc--Alain Widdowson, Josh Mott, Fatimah Dawood, Marika Iwane, Vivek Gupta, and Debjani Ram Purakayastha for their contribution to this study. The authors would also like to thank the study participants, their caregivers, the health facilities and the community of the Comprehensive Rural Health Services Project (CRHSP) villages for their cooperation for this study. **Funding**: This study was funded through cooperative agreement 1U01IP000206 between the US Centers for Disease Control and the National Institute of Virology, Pune. **Disclaimer:** The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention. **Ethics approval:** Written informed consent was obtained from the parent/legal guardian of each participant prior to enrollment. The study protocol was reviewed and approved by the Institutional Review Boards of the Indian Council of Medical Research, All India Institute of Medical Sciences and US CDC. **Authorship declaration:** SS, RBL, AK, and SB conceptualized the study design. SKR, and SB managed the surveillance platform. BG, AC, and SB performed sample processing and viral testing. SS and RBL analyzed the data and wrote the first draft. AK, MKI, SIG, PS, MSC reviewed the results and helped writing the manuscript. All authors reviewed the manuscript. **Competing interests:** All authors have completed the Unified Competing Interest form at [www.icme.org/coi_disclosure.pdf](http://www.icme.org/coi_disclosure.pdf) (available on request from the corresponding author). The authors declare no conflict of interest.

INTRODUCTION {#sec1-1} ============ The World Health Organization estimates that more than 180 million people worldwide have diabetes, and by 2030 it is expected that this number will have doubled.\[[@ref1]\] There is an alarming increase in the incidence and prevalence of diabetes mellitus (DM) in Asian Indians.\[[@ref2]\] Diabetes is a micro-macrovascular disorder with debilitating effects on many organs. Pulmonary complications of DM have been poorly characterized with conflicting results. The alveolar capillary network in the lung is a large micro-vascular unit and may be affected by microangiopathy.\[[@ref3]\] However, because of its large reserve, substantial loss of the microvascular bed can be tolerated without developing dyspnoea. As a result, pulmonary diabetic micro-angiopathy may be under-recognized clinically. In DM pulmonary functions have been studied frequently in countries other than India,\[[@ref4]\] while in our country there are few studies concerning these abnormalities and their relationship with glycosylated hemoglobin (HbA1c) and duration of the disease. Reduced elastic recoil, reduced lung volume, diminished respiratory muscle performance, chronic low grade inflammation,\[[@ref5][@ref6]\] decrease in pulmonary diffusion capacity for carbon monoxide,\[[@ref7]\] autonomic neuropathy involving respiratory muscles\[[@ref8]\] are some of the important changes occurring in DM. Despite the unclear nature, the relationship between DM and pulmonary function tests (PFTs) remains important because of potential epidemiological and clinical implications. The loss of pulmonary reserve may become clinically important. Hence, we hypothesized that PFTs are affected in DM in Indian population and the changes may correlate with HbA1c and the duration of the disease. Objectives {#sec1-1-1} ---------- - The study aims to evaluate the PFTs in type 2 DM patients and compare them with the age and gender matched healthy controls. - We also determine the co-relation of the HbA1c and duration of the disease with PFTs in type 2 DM patients. MATERIALS AND METHODS {#sec1-2} ===================== The study was carried out in collaboration with Diabetes Outpatient Department of Sassoon General Hospitals. The Institutional Ethics Committee approved the study protocol. Sixty male patients of type 2 DM diagnosed by the treating physician, of the age group 40--60 years taking oral hypoglycemics, were randomly selected from the Diabetes Outpatient Department. Exclusion criteria {#sec1-2-1} ------------------ 1. Patients having complaints of cough, sputum, or dyspnoea. 2. Smokers and patients with any cardio-respiratory illnesses or major diseases. Sixty normal healthy males of the same age group and socioeconomic status from patient′s relatives were selected as control group. The controls were also thoroughly examined clinically. Those with cardio-respiratory, musculoskeletal, or endocrine diseases were excluded from the study. Fasting and postprandial blood glucose levels were measured by glucose oxidase method to rule out type 2 DM in them. Informed written consent was taken from patients as well as from controls. All the patients were handed a questionnaire that contained a detailed personal and medical history. PFTs of the patients as well as of the controls were performed with turbine flow sensor-based 702 Helios - Spirometer (Chandigarh, India) between 11 am and 12 pm. All the tests were conducted according to American Thoracic Society/European Respiratory Society (ATS/ERS guidelines) in a quiet room in sitting position by the trained personnel.\[[@ref9]\] The controls and patients performed spirometry three times at the interval of 15 minutes and the best of the three was taken into account. Parameters recorded were - forced vital capacity (FVC) in liters, forced expiratory volume in 1 second (FEV~1~), FEV~1~/FVC in percentage (%), forced expiratory flow during 25% of FVC (FEF~25~), forced expiratory flow during 50% of FVC (FEF~50~), forced expiratory flow during 75% of FVC (FEF~75~), forced expiratory flow during 25--75% of FVC (FEF~25--75~), forced expiratory flow during 0.2--1.2 liters of FVC (FEF~0.2--1.2~), and peak expiratory flow rate (PEFR). For all these parameters percentage of predicted values for the respective age, height, and weight were taken into consideration. Nearly 2 ml of venous blood was collected in ethylenediamine tetra acetic acid (EDTA) bulb in all the diabetic patients with aseptic precautions. HbA1c of all the patients was estimated by ion exchange resin method by the diagnostic glycohaemoglobin kits of Asritha Diatech as per the guidelines provided.\[[@ref10]\] All data were collected in a data collection form and then transferred to an Excel sheet by two independent data entry operators. Discrepant values were corrected by checking the data collection form. Clean data was then analyzed statistically. PFTs of diabetic patients and controls were compared by applying Student′s unpaired ′*t*′ test. Correlations between FVC and FEV~1~ and HbA1c and duration of illness in diabetic patients were analyzed by applying Pearson′s coefficient. Statistical analysis was done by using SSPS version 11 and Graphic Prism Pad version 5 (Statistician, B.J. Medical College). RESULTS {#sec1-3} ======= [Table 1](#T1){ref-type="table"} depicts the physical characteristics of the normal controls as well as the patients of DM. Age, height, and weight of both the groups were comparable as statistically there was no difference between them (*P* \> 0.05). Our study showed that all the pulmonary parameters, that is, FVC, FEV~1~, FEF~25~, FEF~50~, FEF~75~, FEF~25--75~, FEF~0.2--1.2~, and PEFR were significantly reduced except FEV~1~/FVC in patients of type 2 DM as compared with the healthy controls (*P* \< 0.05). The ratio FEV~1~/FVC is almost equal in normal controls and diabetic patients (*P* \> 0.05) \[[Table 2](#T2){ref-type="table"}, [Figure 2](#F2){ref-type="fig"}\]. On correlating the FVC and FEV~1~ with duration of illness and HbA1c, we found that there was no significant correlation between them (*P* \> 0.05) \[[Table 3](#T3){ref-type="table"}, Figures [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}\]. ###### Physical characteristics of subjects  ###### Comparison of PFTs in patients with type 2 DM and healthy controls  {#F1} ###### Correlation of HbA1c and duration of DM with PFTs  {#F2} DISCUSSION {#sec1-4} ========== Our study showed that all the pulmonary parameters, that is, FVC, FEV~1~, FEF~25~, FEF~50~, FEF~75~, FEF~25--75~, FEF~0.2--1.2~, and PEFR were significantly reduced except FEV~1~/FVC in patients of type 2 DM as compared with the healthy controls. This is in accordance with previous studies.\[[@ref11]--[@ref15]\] Some of the prospective and cross sectional studies have shown low vital capacity or restrictive pattern in type 2 DM.\[[@ref16][@ref17]\] Meta-analysis by van den Borst, *et al*. showed that DM is associated with statistically significant, impaired pulmonary function in a restrictive pattern. Moreover, these results were irrespective of body mass index (BMI), smoking, diabetes duration, and HbA1c levels.\[[@ref18]\] Uchida, *et al*. found that there was decreased pulmonary diffusing capacity in patients with diabetes with perfusion defect on ventilation perfusion scintigrams.\[[@ref19]\] It was not possible for us to analyze the pulmonary diffusing capacity because of practical difficulties. Davis, *et al*. conducted a study in Western Australia in large number of patients of type 2 DM. They found that VC, FVC, FEV~1~, and PEFR decreased at an average of between 1.1% and 3.1% of predicted values/year in type 2 DM patients.\[[@ref11]\] Ehrlich, *et al*. showed that patients with type 2 DM were at increased risk of several pulmonary condition like - asthma, Chronic Obstructive Pulmonary Disease (COPD), fibrosis, and pneumonia.\[[@ref20]\] Few studies have mentioned that no significant differences were observed in patients of type 2 DM.\[[@ref21]--[@ref23]\] Probably the small sample size is the reason behind these findings. Pathophysiology of reduced lung function is still an interesting research issue. Normal lung mechanics and gas exchange are influenced by the integrity of the pulmonary connective tissue and microvaculature. Acceleration of aging process in connective tissue cross links and presence of nonenzymatic glycosylation and modification of alveolar surfactant action causes reduction in PFTs.\[[@ref3]\] There have been reports of histopathological changes in the diabetic patients. In the study by Weynand *et al*.,\[[@ref24]\] it was found that alveolar epithelium, endothelium capillary, and basal laminaes were thickened in lungs on electron microscopy, when compared with the controls. In addition, the thickening of basal laminae was of the same magnitude in lung and kidney. Diabetic microangiopathy might be existing in the pulmonary vascular bed. Moreover, reduced pulmonary capillary blood volume was found, favoring the evidence of microangiopathy. This could lead to redistribution of the pulmonary circulation, resulting in well ventilated areas to become underperfused.\[[@ref25]\] The thorax and lungs are rich in collagen and elastin. Stiffening of thorax and lung parenchyma can occur because of nonenzymatic glycosytion of these structural compounds. This may lead to restrictive pattern.\[[@ref3]\] In our studies, since the FVC/FEV~1~ ratio is statistically not significantly different in DM patients as compared with normal controls, other PFT values are lower in DM patients; this strongly suggests restrictive pattern in DM patients. Studies have even shown diabetic polyneuropathy, which affects respiratory neuromuscular function and thus reducing pulmonary volumes.\[[@ref26]\] On correlating the FVC and FEV~1~ with duration of illness and HbA1c, we found that there was no significant correlation between them \[[Table 3](#T3){ref-type="table"}, Figures [2](#F2){ref-type="fig"} and 3\]. There are certain studies showing no correlationship between HbA1c and PFTs.\[[@ref7][@ref14][@ref22]\] They argued that HbA1c levels are indicators of glycemic control for a short period of 1--2 months, it was not adequate to conclude that the plasma glucose level was not related to decreased PFTs. While some studies have shown that the decline in PFTs was negatively correlated with HbA1c.\[[@ref11][@ref13]\] There are studies that have reported no significant correlation between PFTs and duration of diseases,\[[@ref22]\] while some of the studies have reported a strong negative correlation of PFTs with duration.\[[@ref14][@ref15]\] Since DM is a disease, which involves multiple organs randomly, the study of the effect of duration of the disease on them requires further research. Several studies have analyzed the association between impaired lung function and death and found that a 10% decrease in FEV~1~ was associated with a 12% increase in all cause mortality in type 2 DM.\[[@ref27]\] Clinical implications {#sec1-4-1} --------------------- Pulmonary dysfunction should be regarded as a specific derangement induced by DM. Further studies may clarify whether this should be included as a long-term complication of diabetes. The role of strict glycemic control on pulmonary function in diabetic patients is another interesting aspect and needs further studies. The impairment in PFTs can lower the threshold for clinical manifestations of acute or chronic lung disease. Patients with DM admitted with pneumonia have increased risk of complications and mortality.\[[@ref28]\] Limitations and scope {#sec1-4-2} --------------------- It seems to be necessary to repeat PFTs and to assess the changes of pulmonary functions among the same subjects. Over a long observation course, the relationship between the plasma glucose concentration and the PFTs can be elucidated. SUMMARY AND CONCLUSION {#sec1-5} ====================== DM being a systemic disease, also affects lungs causing restrictive type of ventilatory changes probably because of glycosylation of connective tissues, reduced pulmonary elastic recoil, and inflammatory changes in lungs. We found that glycemic levels and duration of disease are probably not the major determinants of lung pathology, which requires further research. **Source of Support:** Nil **Conflict of Interest:** None declared.

Introduction {#sec1-1} ============ Subdural hematoma, one of the complications of shunt insertion, is usually caused by rapid excessive drainage of cerebrospinal fluid. The subdural hematoma generally appears as a crescent-shaped mass over the convexity adjacent to the inner table of the skull on computed tomography (CT) scans or adjacent to the dura matter on magnetic resonance images (MRI). We describe the rare case of an adult with untreated compensated congenital hydrocephalus with unusual shaped huge hemispheric subdural hematoma. Case Report {#sec1-2} =========== We admitted an interesting case of compensated congenital hydrocephalus. Patient was a 39-year-old female with history of enlarged head size and delayed developmental mile stones since birth, who presented with intermittent headache and vomiting for 1.5 months, right hemiparesis for five days, urinary incontinence, and drowsiness for one day. There was history of multiple episodes of minor head injury due to frequent falls. There was no history of coagulation disorder, seizure, shunt surgery, or any drug intake. On examination, patient was conscious, oriented, but drowsy, head circumference was 74 cm, pupils were bilateral normal size and reactive, bilateral papilledema was present and right side hemiparesis (power: MRC grade 1 in upper limb and 3 in lower limb) was present. Hemoglobin was 12 gm% and other routine blood investigations including coagulation profile were normal. CT scan was showing huge left hemispheric chronic subdural hematoma with unusual shape occupying whole left cerebral hemisphere and compressing the falx with almost complete obliteration of left lateral ventricle with dilated right lateral ventricle \[[Figure 1](#F1){ref-type="fig"}\]. MRI was showing huge subdural hematoma hyperintense on both T1 and T2 weighted image with additional evidence of multiple septations in subdural hematoma \[Figure [2](#F2){ref-type="fig"} and [3](#F3){ref-type="fig"}\]. Patient was operated by double burr hole craniostomy. Large amount of thick motor oil colored subdural fluid was drained. Subdural space was irrigated with saline solution till the effluent became clear. Brain did not reach to surface completely at the end of procedure. Subgaleal close drainage system was put. {#F1} {#F2} {#F3} Postoperatively, patient became alert. The patient did not develop any problems in the early postoperative period and her subgaleal drain was removed on the second postoperative day. Right hemiparesis improved to power grade 4 in both upper and lower limbs. Postoperative CT scan showed almost complete evacuation of hematoma \[[Figure 4](#F4){ref-type="fig"}\]. {#F4} Discussion {#sec1-3} ========== Chronic subdural hematoma is one of the most frequent types of intracranial hemorrhage that is still associated with significant morbidity. Although minor previous head injury is sometimes unrecognized, a mild traumatic event has preceded the hemorrhage in 60 to 80% of reported cases.\[[@ref1]\] Very few cases of chronic subdural hematoma occupying almost entire hemisphere have been reported. Although the cause of chronic subdural hematoma varies, it usually spreads out over the cerebral convexity and generally appears as a crescent-shaped lesion on neuroradiological images.\[[@ref1][@ref2]\] Unlike epidural hematomas, subdural hematomas are not restricted by dural tethering at the cranial sutures; they can cross suture lines and continue along the falx and tentorium. However, they do not cross the midline because of the meningeal reflections,\[[@ref3]\] but in our case, contralateral extension of chronic subdural hematoma was another unique finding. When hemorrhages of differing ages exist within a subdural collection, septae may separate the different blood products as happened in our case. The cause of chronic subdural hematoma in our patient is probably repeated head trauma, but the exact mechanism for this unusual shaped hematoma in our patient is unclear. Chronic subdural hematoma is probably caused by tearing of the bridging veins or bleeding from an enlarged vein in the subdural zone. Due to long standing compensated hydrocephalus with a large head and little brain parenchyma, our patient tolerated this huge chronic subdural hematoma for long duration before it became symptomatic. Decrease of elastence of brain due to chronic compensated congenital hydrocephalus may have allowed hematoma to grow unexpectedly. **Source of Support:** Nil **Conflict of Interest:** None declared.

The content published in Cureus is the result of clinical experience and/or research by independent individuals or organizations. Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein. All content published within Cureus is intended only for educational, research and reference purposes. Additionally, articles published within Cureus should not be deemed a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus. Introduction ============ Obesity (Body Mass Index (BMI) \> 30 kg/m^2^) and associated poor health-related quality of life (HRQoL) affect nearly one-third of American men over 60 years old \[[@REF1]\]. HRQoL in cancer patients is dynamic and may be adversely impacted by obesity. Excess abdominal fat is associated with an increased risk of urinary incontinence and sexual dysfunction \[[@REF2]\]. Total serum testosterone levels are inversely associated with BMI \[[@REF3]\], and low levels may contribute to greater incidences of fatigue in obese patients \[[@REF4]\]. In the United States, approximately 220,000 men are newly diagnosed with prostate cancer (PCa) each year \[[@REF5]\]. Obesity may have a multifaceted impact on a PCa diagnosis and management \[[@REF6]\]. Men with a high BMI may have greater incidences of aggressive PCa \[[@REF7]\]. Due to inherent technical difficulties associated with increased abdominal adipose tissue distribution, such as setup inconsistencies and increased prostatic movement during treatment \[[@REF8]\], obesity can have a negative impact on PCa radiation therapy (RT) outcomes, with obese patients experiencing higher rates of biochemical recurrence and PCa specific mortality \[[@REF9]\]. Cancer control outcomes following brachytherapy are not affected by obesity presumably due to the image-guided placement of radioactive sources directly within the prostate \[[@REF10]-[@REF11]\]. Obesity may also have a negative impact on post-RT HRQoL, due to greater radiation exposure to the rectum, bladder, and sexual organs \[[@REF12]-[@REF13]\]. Robotic stereotactic body radiation therapy (SBRT) involves conformal dose delivery of a few hundred, non-coplanar treatment beams from a linear accelerator mounted on a flexible robotic arm. It employs real-time image guidance to track implanted fiducials in the prostate, accounting for prostatic movements in six dimensions \[[@REF14]\]. This allows delivery of the therapeutic beam to the prostate with less than 1 mm error, potentially minimizing the volume of critical structures receiving radiation \[[@REF15]\]. Increased treatment accuracy in obese patients may further reduce critical organ scatter and ultimately improve HRQoL. This study reports HRQoL in obese men after SBRT for PCa by examining the relationship between BMI and commonly associated urinary symptoms, bowel symptoms, sexual function, and hormonal symptoms after SBRT. Materials and methods ===================== Patient selection ----------------- Patients eligible for study inclusion had clinically localized PCa treated with SBRT at Georgetown University Hospital. Patients who received androgen deprivation therapy (ADT) were excluded from this analysis due to its known negative impact on HRQoL \[[@REF16]\]. This retrospective review was approved by the Georgetown University Internal Review Board (IRB 2009-510). Patient BMIs were calculated from the baseline weight and height \[[@REF17]\]. Obesity was defined as a BMI ­\>30 kg/m^2^\[[@REF18]-[@REF19]\]. PCa risk groups were defined using the D\'Amico criteria \[[@REF20]\]. Other patient and treatment characteristics such as age, race, Charleson comorbidity index (CCI) \[[@REF21]\], prostate volume, pretreatment prostate-specific antigen (PSA) and testosterone, Gleason score, use of sexual aid, and SBRT dose were acquired from the medical records. SBRT treatment planning and delivery ------------------------------------ Accuray's CyberKnife was employed to treat the prostate as previously described \[[@REF22]\]. Treatment planning involved fusion of thin-cut CT images and high-resolution MR images, after 4-6 gold fiducials were placed in the prostate. The clinical target volume (CTV) included the prostate and proximal seminal vesicles. The planning target volume (PTV) was defined as the CTV and 3 mm in the posterior direction and 5 mm in all other directions. The rectum, bladder, testes, and penile bulb were contoured and further evaluated with dose-volume histogram (DVH) analysis, using multiplan inverse treatment planning. The PTV received 35-36.25 Gy in five fractions of 7-7.25 Gy over one to two weeks. Follow-up and statistical analysis ---------------------------------- Patients completed the expanded prostate cancer index composite (EPIC)-26 questionnaire at baseline (one hour prior to first SBRT fraction) and during routine follow-up visit every six months after completion of SBRT, for two years. The EPIC-26 urinary domain was divided into two functional sub-domains, incontinence and irritative/obstructive domains \[[@REF23]\]. In addition, one question assessed associated overall bother. The EPIC-26 bowel domain included five questions related to individual symptoms and one question related to overall bother. The EPIC-26 sexual function domain utilized five questions regarding sexual function and one question regarding sexual bother. Lastly, the EPIC-26 hormonal domain had five questions with one item assessing lack of energy. EPIC scores for each domain and the individual questions ranged from 0-100, with lower values representing worsening symptoms. To statistically compare responses between the two BMI groups, the responses were assigned a score, and the significance of the scores was assessed using Mann-Whitney U test. Clinically significant change was assessed by the minimally important difference (MID) in the EPIC score. This was defined as a change of one-half standard deviation (SD) from the baseline \[[@REF24]\]. The Wilcoxon signed-rank test was conducted to determine the statistical significance of the average score of the cohort at each time point. Results ======= Between February 2008 and April 2012, 88 obese and 178 non-obese prostate cancer patients were treated on an institutional SBRT protocol. Characteristics of both obese and non-obese groups were similar prior to SBRT, with a few important differences (Table [1](#TAB1){ref-type="table"}). The median patient age was 68 years for obese and 70 years for non-obese. The obese and non-obese cohorts were composed of 56.8% and 55.6% Caucasian, 41.0% and 36.0% African ancestry, respectively. The median prostate volume in both groups was 38 cc. Pre-treatment PSA values were similar, but baseline pre-treatment total serum testosterone levels varied; 360.5 ng/dL in non-obese and 265.5 ng/dL in obese patients. Additionally, significant comorbidities were more common in obese patients. The D'Amico classification shows a majority, 60.9% of obese and 52.5% of non-obese patients, were intermediate-risk. Seventy-eight percent of both cohorts were treated with 36.25 Gy in five 7.25 Gy fractions. Table 1Patient Baseline Characteristics*PSA =*Prostate Specific Antigen *SBRT =*Stereotactic Body Radiation Therapy  All (N =266)Non-obese (N= 178)Obese (N= 88)Age (Years)Median Age (Range)69 (44-94)7068RaceWhite56.0%55.6%56.8% Black37.6%36.0%41.0% Other6.4%8.4%2.3%Charlson Comorbidity IndexCCI=066.9%73.0%54.6% CCI=122.6%16.3%35.2% CCI[\>]{.ul}210.5%10.7%10.2%Body Mass Index (kg/m2)18.5- 24.9 (n)4827.0%\-- 25-29.9 (n)13073.0%\-- 30-34.9 (n)61\--69.3% 35-39.9 (n)23\--26.1% 40.0-44.9 (n)4\--4.6%Prostate Volume (cc)Median Volume (Range)38.0 (9.3-138.7)38 (9.3-138.7)38 (17.6-86.2)Pre-Treatment PSAMedian PSA(Range)6.0 (0.8-32.5)6.1 (0.8-32.5)5.8 (1.5-18.6)Pre-Treatment Testosterone (ng/dL)Median Testosterone (Range)320 (71-1149)360.5 (106-980)262.5 (71-114)Risk Groups (D'Amico's)Low Risk39.5%41.0%36.4% Intermediate Risk55.3%52.3%61.4% High Risk5.3%6.74%2.3%Sexual AidNone63.0%63.3%62.5% Any Aid37.0%36.7%37.5%SBRT Dose36.25 Gy78.2%78.2%78.2% 35 Gy21.2%21.2%20.7% Other0.8%0.6%1.2% The baseline summary scores of both BMI groups were comparable (Table [2](#TAB2){ref-type="table"}). There were no differences in urinary incontinence scores and urinary irritation/obstruction scores between the obese and non-obese cohorts. Urinary bother scores were similar between BMI groups. Bowel function at baseline was 94.44 in obese and 95.02 in non-obese patients, while associated bowel bother was 90.12 and 91.20, respectively. Patients in both BMI groups had low but similar sexual function (*p* = 0.305) and bother (*p* = 0.487). Lastly, summary of hormonal symptoms between BMI groups remained consistent, with 91.55 in obese and 92.49 in non-obese. Table 2Pre-Treatment Quality of Life (QoL) EPIC-26 scores*EPIC =*Expanded Prostate Cancer Index Composite *Std. Dev* = Standard Deviation All (N =266)Not Obese (N= 178)Obese (N= 88)   MeanStd. DevMIDMeanStd. DevMeanStd. Dev*p*-valueUrinary Incontinence86.3113.466.7386.2513.9886.4512.400.764 Irritative/Obstructive87.3212.615.5286.3713.4789.2910.440.129 Bother78.1025.7412.8776.4026.4681.6123.950.083Bowel Function94.839.394.6995.028.9794.4410.240.765 Bother90.8519.069.5391.2018.4490.1220.390.27Sexual Function53.2132.4316.2254.9331.9649.6733.290.305 Bother64.7635.8317.9266.0634.9162.0737.710.487Hormonal Summary92.1811.445.7292.4910.8191.5512.670.934 Table [3](#TAB3){ref-type="table"} shows the EPIC summary scores linearly from baseline to 24 months after SBRT. Collected scores of urinary incontinence, irritative/obstructive symptoms, and bother were comparable between obese and non-obese groups at all time points. Bowel function and associated bother between the two cohorts were also statistically similar. Summary score of sexual function seems to decline in both obese and non-obese men, but remains similar over the 24 months (Figure [1a](#FIG1){ref-type="fig"}). Sexual bother score in obese men (48.05) was significantly lower (p= 0.0076) than that reported by non-obese men (62.26), only at 24 months (Figure [1b](#FIG1){ref-type="fig"}).  Table 3Obese and Non-Obese Patient Urinary, Bowel, Sexual, and Hormonal Domain EPIC-26 Scores\*Sexual bother between obese and non-obese patients at 24 months is statistically significant (*p*-value of 0.01).*EPIC =*Expanded Prostate Cancer Index Composite *Std. Dev* = Standard Deviation Baseline6 Month12 Month18 Month24 Month Mean*p*-valueMean*p*-valueMean*p*-valueMean*p*-valueMean*p*-value Not ObeseObese Not ObeseObese Not ObeseObese Not ObeseObeseNot ObeseObese Urinary Incontinence86.386.50.7690.090.30.4087.785.70.9185.484.70.9586.487.20.97 Irritative/ Obstructive86.489.30.1385.886.50.3882.884.30.4884.785.20.7785.989.00.18 Bother76.481.60.0877.476.60.8668.468.70.8875.572.00.6075.876.60.83Bowel Function95.094.40.7791.490.80.4291.390.00.7391.691.20.4192.893.30.70 Bother91.290.20.386.789.70.2684.484.80.5787.289.50.2689.290.30.44Sexual Function54.750.10.3149.544.80.2545.740.40.2743.339.10.3843.237.20.17 Bother65.962.50.4964.858.20.1662.757.30.2859.553.00.2062.348.10.01\*Hormonal Function92.491.70.9392.389.60.1491.888.40.0892.588.50.0791.689.80.41 Figure 1Non-Obese (red) and Obese (blue) Patients Reported Mean Quality of Life (QoL) EPIC-26 Sexual Domain Scores at Baseline and Following SBRT for Prostate Cancer.Shown are plots for: (a) EPIC sexual overall functional summary, (b) EPIC sexual bother.  The EPIC-26 inquired on lack of energy after SBRT, at each follow-up. Compared to baseline, all time points except at 18 months had similar fatigue scores as shown in Figure [2a](#FIG2){ref-type="fig"}. However, this difference was not clinically insignificant. At 18 months, the average fatigue score of all patients in the study (78.52) was less than that at baseline (81.92) (*p*= 0.036). Fatigue was relatively constant, except at 18 months when it decreased. Lack-of-energy scores were statistically different (p = 0.042) in obese (72.03) and non-obese patients (81.79) at 18 months, as seen in Figure [2b](#FIG2){ref-type="fig"}. At other times points, lack-of-energy scores between both BMI groups were similar. Figure 2EPIC-26 Loss of Energy ScoresShown is plot for: (a) EPIC loss of energy score at baseline and following SBRT for prostate cancer, (b) EPIC loss of energy score in obese (blue) and non-obese (red) patients with prostate cancer. The thresholds for clinically significant changes in scores (½ standard deviation above, *green;* and below the baseline, *purple*) are marked with dashed lines. EPIC scores range from 0--100 with higher values representing a more favorable health-related QoL. Discussion ========== Consideration of urinary, bowel, sexual, and hormonal side effects are critical in an individual patient's choice of treatment for prostate cancer \[[@REF12]\]. Excess abdominal adipose tissue in obese men may hinder the accuracy of therapeutic beams, thereby diminishing their efficacy, and the increased critical organ scattered dose may significantly compromise the patients' HRQoL. The comparison of urinary and bowel functions and bother post-SBRT between obese and non-obese patients demonstrates only limited differences. Utilization of fiducial markers with inter and intrafraction image guidance may have reduced potential differences in HRQoL between BMI groups. Our results appear similar to obese and non-obese patients having undergone brachytherapy for PCa \[[@REF11], [@REF25]\]. Both sexual function and bother are affected by biological, psychological, and sociological factors (e.g., serum testosterone levels, age, confidence, marital status, and partner satisfaction). Sexual function is clinically decreased in obese men and associated bother is greater than that in non-obese men 24 months after SBRT. While radiation may have certainly contributed to declining sexual function in both BMI groups, it is unlikely to have uniquely affected obese patients. A low baseline serum testosterone level in obese men may be a causative factor for increased late sexual dysfunction, bother, and lack of energy. Fatigue is a common problem in obese patients and may also be a side effect of RT \[[@REF26]\]. The specific etiology of RT-related fatigue is poorly understood and most likely multi-factorial. Fatigue levels were similar between obese and non-obese patients at baseline and most follow-ups after SBRT. Although clinically insignificant, obesity seems to play a role in patient-reported fatigue at the 18-month follow-up. Obese men reported greater levels of fatigue compared to non-obese men only at 18 months after SBRT; at all other time points, obesity does not seem to enhance fatigue in a PCa patient following SBRT. The present study has several limitations. This is a retrospective study of prospectively collected data from a single institution cohort. This limits the translational generalizability to institutions whose patient population and SBRT protocols are not similar. Only a small number of our patients were morbidly obese (BMI \>40), limiting our ability to access the impact of morbid obesity on post-SBRT HRQoL.  Conclusions =========== Prostate SBRT affects obese and non-obese patients similarly in a majority of HRQoL domains. Minimal differences in HRQoL were identified between obese and non-obese patients post-SBRT. A longer follow-up is required to determine the impact of obesity on cancer control. The Department of Radiation Medicine at Georgetown University Hospital receives an educational grant from Accuray to support a research coordinator. Dr. Sean Collins and Dr. Brian Collins are clinical consultants for Accuray. Georgetown University Internal Review Board issued approval IRB 2009-510 **Animal subjects:** This study did not involve animal subjects or tissue.

Introduction {#s1} ============ As of December 2011, over 8 million people infected with HIV were receiving antiretroviral therapy (ART) in low- and middle-income countries which represents a 26-fold increase since 2003 [@pone.0065653-UNAIDS1]. Due to HIV's error-prone replication, high mutation rate and viral recombination, development of some HIV drug resistance (HIVDR) is inevitable, even with appropriate ART prescribing and adherence [@pone.0065653-Coffin1]--[@pone.0065653-Bennett1]. HIVDR has significant human and financial implications: it limits treatment options; second-line ART regimens involve more long-term toxicity; and annual costs of second-line regimens are 4--8 times higher than that of currently recommended first-line regimens [@pone.0065653-Bennett2]. As the number of people on treatment increases, the emergence of meaningful population-level HIVDR becomes a greater risk which has the potential to undermine the dramatic gains that ART programs have had in reducing the morbidity and mortality of HIV-infected people in resource-limited settings [@pone.0065653-Jordan1]. Monitoring of ART program factors known to be associated with the emergence of HIVDR for the purpose of improving programmatic functioning, may minimize the emergence of preventable HIVDR, especially at ART sites where viral load and HIVDR testing is not routinely available. For example: HIVDR testing is not required to predict the emergence of drug-resistant HIV in settings where inappropriate prescribing practices (mono- or dual-therapy), treatment interruptions due to suboptimal patient adherence, poor patient retention on ART, or ART supply shortages or stock-outs occur at unacceptably high levels. These factors have been shown to be associated with the development of HIVDR [@pone.0065653-Vergne1]--[@pone.0065653-VanOosterhout1]; thus, their monitoring may alert national ART program planners to issues which may be adjusted to minimize the emergence of HIVDR. HIVDR Early Warning Indicators {#s1a} ------------------------------ The foundation of the World Health Organization's (WHO) global HIVDR prevention and assessment strategy [@pone.0065653-Bennett3], which includes laboratory-based surveys of acquired [@pone.0065653-Jordan2] and transmitted [@pone.0065653-Bennett4] HIVDR, is the monitoring of HIVDR Early Warning Indicators (EWI). EWIs assess ART site and program factors potentially associated with HIVDR [@pone.0065653-World1]. Utilizing data routinely collected in patients' medical and pharmacy records, EWI monitoring is a minimum-resource strategy designed to be integrated into national monitoring and evaluation programs. EWIs survey factors related to patient care, patient behavior, and clinic-level and program management, all of which are associated with the emergence of HIVDR. When monitored annually at all or a large number of representative ART sites, EWIs provide countries with evidence to make programmatic adjustments at the level of an individual site or the country, when necessary. WHO updated their EWI definitions in 2012 [@pone.0065653-WHO1]. HIV in Namibia {#s1b} -------------- Namibia is a resource-limited country in sub-Saharan Africa that has been severely affected by the HIV epidemic. In Namibia, there are approximately 200,000 people living with HIV in a population of 2.1 million [@pone.0065653-Central1]. Among 15--49 year olds, approximately 18.2% are infected with HIV-1 [@pone.0065653-The1]. The epidemic is predominantly spread via heterosexual contact, and prevalence estimates vary by region with up to 37.7% infected with HIV-1 in the most heavily-affected areas in the north [@pone.0065653-The1]. ART rollout {#s1c} ----------- ART has been available in Namibia's private sector since 1997 and in the public sector since 2003. At 90%, Namibia has one of the highest ART coverage rates in Sub-Saharan Africa with 88,717 eligible patients on ART as of December 2010 [@pone.0065653-WHO2]. At present, ART is available at all 40 public hospitals and at an additional 111 satellite/outreach service points, as well as 30 Integrated Management of Adolescent and Adult Illness (IMAI) sites [@pone.0065653-Republic1]. Of these sites, the national ART program considers 35 to be main public ART sites. Main public ART sites are sites that dispense ART independently of other ART sites and dispense ART to patients at IMAI and satellite/outreach sites. In the public sector, ART is provided free of charge following a population-based model of care with one primary first-line regimen and three alternate first-line regimens consisting of two nucleoside reverse transcriptase inhibitors (NRTI) combined with a non-nucleoside reverse transcriptase inhibitor (NNRTI). The recommended second-line regimen consists of 2 NRTIs with a ritonavir-boosted protease inhibitor (PI). ART initiation is based on WHO clinical staging and/or CD4 cell count ≤350 cells/mm^3^. All public ART sites have access to first and second-line ART regimens. At all public ART sites, viral load testing is performed six months after ART initiation and targeted viral load testing is performed to confirm clinical or immunological failure [@pone.0065653-Republic1]. With support from Management Sciences for Health (MSH) Namibia, a standardized pharmacy record system, the Electronic Dispensing Tool (EDT), is used to dispense all ART. In the private sector, ART is provided utilizing an individual model of care with ART regimens selected based on results of drug resistance testing. Minimizing HIVDR {#s1d} ---------------- The Namibia Ministry of Health and Social Services (MoHSS) has been proactive in minimizing preventable HIVDR. Together with its national and international partners, the MoHSS publishes annual reports and national plans, following WHO recommendations, for the prevention and assessment of HIVDR. The national ART program mandates the use of standardized national ART prescribing practices, WHO pre-qualified drugs, and standardized medical and pharmacy record-keeping systems for national surveillance [@pone.0065653-Republic1]. In 2009, Namibia piloted EWI monitoring in nine sites, which led to adjustments in the existing national data capture tools, as well as feedback and trainings for staff at each individual site [@pone.0065653-Hong1]. The 2010 EWI exercise scaled up the monitoring efforts to all main ART sites. Methods {#s2} ======= Early Warning Indicators Selection {#s2a} ---------------------------------- Namibia chose the following five indicators based on relevance to anticipated program interventions and availability of data: *ART Prescribing Practices*, *Patients lost to follow-up 12 months after ART initiation*, *Patients switched to second-line ART during first 12 months*, *On-time ARV drug pick-up*, and *ARV drug supply continuity* [@pone.0065653-World1]. Definitions (numerator/denominator) for these selected EWIs and their respective recommended targets are summarized in [Table 1](#pone-0065653-t001){ref-type="table"}. EWI definitions were based on previous (2010) WHO-EWI guidance [@pone.0065653-World1] and not on the latest (2012) WHO EWI guidance [@pone.0065653-WHO1]. 10.1371/journal.pone.0065653.t001 ###### Selected 2010 WHO Early Warning Indicator definitions (Numerator/Denominator) and targets. {#pone-0065653-t001-1} Indicator Definition (Numerator/Denominator) ----------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ART prescribing practices [Numerator:]{.ul} Number of adult patients initiating ART at the site who initially pick up from the pharmacy, an appropriate first-line ART regimen.[\*](#nt105){ref-type="table-fn"} [Denominator]{.ul}: Number of adult patients initiating ART at the site on or after the designated EWI sample start date.[†](#nt108){ref-type="table-fn"} Sampling continues until the full sample size is reached. Target: 100% Patients lost to follow-up at 12 months[\#](#nt113){ref-type="table-fn"} [Numerator:]{.ul} Number of adult patients initiating ART at the site who are lost to follow-up[§](#nt109){ref-type="table-fn"} during the 12 months after starting ART. [Denominator:]{.ul} Number of adult patients initiating ART at the site on or after the designated EWI sample start date. Sampling continues until the full sample size is reached. Target: ≤20% Patients switched to a second-line regimen at12 months[\#](#nt113){ref-type="table-fn"} [Numerator:]{.ul} Number of adult patients initiating ART at the site whose initial ART regimen was changed to a regimen that includes a different drug class during the first 12 months after ART initiation (including switches for regimen failure and substitutions for toxicity). [Denominator:]{.ul} Number of patients initiating ART at the site on or after the designated EWI sample start date who are retained on ART 12 months after start. Sampling continues until the full sample size is reached.[∞](#nt110){ref-type="table-fn"} Target: 0% On-time ARV drug pick-up[\#](#nt113){ref-type="table-fn"} [Numerator:]{.ul} Number of adult patients who have picked up all their prescribed ARV drugs on time[‡](#nt111){ref-type="table-fn"} for two consecutive drug pick-ups after a baseline pick-up. [Denominator:]{.ul} Number of patients who picked up ARV drugs on or after the designated EWI sample start date. Sampling continues until the full sample size is reached. Target: ≥90%[!](#nt112){ref-type="table-fn"} ARV drug-supply continuity [Numerator:]{.ul} Number of months in the designated year in which there were no stock-out∧ days of any adult ARV drug routinely used at the site. [Denominator:]{.ul} 12 months. Target: 100% ART - Antiretroviral therapy. ARV - Antiretrovirals. LTFU - Lost to follow-up. EWI - Early Warning Indicator. Appropriate first-line ART regimen: An ART regimen that meets one or both of the following definitions: • Standard regimen listed in national ART guidelines and used according to those guidelines. • Regimen recommended in the WHO treatment guidelines. EWI sample start date: The date designated as the start of the sampling. The sample start date is fixed by the HIVDR Working Group. Lost to follow-up: Patients who had not returned to the pharmacy or clinic ≤90 days after the last ART run-out date during the 12-months after the date of ART initiation were classified as LTFU. Transfers of care to another site and deaths were excluded from the numerator and denominator. Stopping therapy without restarting was classified as not LTFU if the patient continued to attend clinic appointments. Transfers of care to another site, deaths, and ART stop without a restart were excluded from the denominator. On-time pick-up of ARV drugs: A patient pick-up of ARV drugs on or before the date the previously dispensed drugs would have run out if they had been taken according to schedule. The following patients were excluded from the denominator: Patients who transferred out, died, or stopped ART without a restart between baseline pick-up date and baseline pick-up run-out date. For EWIs: *Patients LTFU at 12 month*, *Patients switched to a second-line regimen at 12 months*, and *On-time ARV drug pick-up,* if no ARV pick-up date, regimen, and number of pills dispensed was not recorded in the records, it was assumed that no pick-up had occurred, resulting in the most conservative estimate of each indicator. ∧Stock-out: Any occurrence of zero stock of a routinely-used ARV drug at the site at which the patient routinely picks up ARVs. Table adapted from WHO HIV Drug Resistance EWI guidance document [@pone.0065653-World1]. Ethics Statement {#s2b} ---------------- Ethical review was not required as this data was public health surveillance data abstracted from existing routinely collected ministry of health medical records. Only anonymised data were abstracted from the medical records for public health surveillance purposes. Names, dates of birth, addresses, and unique patient identifier numbers were not abstracted from records. After discussion with the Tufts Medical Center institutional review board, it was determined that because this was routine public health de-identified data analyzed within the Ministry of Health and Social Services in Namibia, no formal written waiver was necessary. The data used for this study was obtained from and analyzed by the MoHSS of Namibia. Site Selection and Data Abstraction {#s2c} ----------------------------------- All 35 main public ART delivery sites were selected for EWI abstraction in 2010, which included the nine sites that participated in EWI abstraction in 2009. Only main ART sites were selected for EWIs because satellite/outreach and IMAI sites are considered part of the main ART sites (utilizing same ART staff, pharmacists and record systems). Private ART sites were not selected for inclusion because they do not utilize a population-based model of care and have routine HIVDR testing available. Data abstraction was conducted in October 2010 centrally by a data abstraction team formed by the HIVDR technical working group (TWG) in collaboration with the WHO and the MoHSS. The team consisted of four members who had been trained on the WHO methodology of EWI data abstraction. Primary data were centrally queried from the EDT national database through automatic queries into an Excel tool provided by the WHO, which calculated results for each indicator. Data Quality Assessment {#s2d} ----------------------- Data quality assessments were implemented throughout the EWI process. Three elements of data quality were considered in the assessments: data reliability, data completeness, and data consistency [@pone.0065653-World1]. Data reliability, which is an assessment of the quality of the abstraction, was assessed by confirming 10% of the centrally-queried data to the existing data in the EDT. Data completeness was assessed from the centrally-queried data; and sites with a large percentage of data missing were removed from EWI analyses. Finally, assessment of data consistency was initially performed during the pilot of EWIs [@pone.0065653-Hong1] and the most optimal source for each variable was determined. EDT data were considered the gold standard for pharmacy pick-up dates and ART regimens dispensed, while the national electronic patient management system (ePMS) and paper records (Patient Care Booklets) were considered gold standard for information about patient status such as dates of transfer in and transfer out, dates of death, and dates of stop. Therefore, EDT data for patients who had incomplete pill pick-ups were validated and corrected by comparing records in ePMS and Patient Care Booklets, looking for dates of transfer out, death or ART stop. Centrally-queried EDT data for patients who had inappropriate ART regimens at start or at 12 months were validated and corrected with the site-specific EDT system to ensure accuracy of the queries. Validation with ePMS and paper medical records was performed by the individual ART sites that were trained on EWI methodology at a national EWI conference. Sample Size {#s2e} ----------- In order to make the results generalizable to the patient population at the ART site, the sampling strategy was based on calculating a minimum sample size for each indicator at each site, based on the number of eligible patients for each EWI. Two different cohorts of 'eligible patients' were formed: 1) Patients consecutively initiating ART for the first time on or after the EWI sample start date of July 1, 2008 (*ART prescribing practices*, *Patients LTFU at 12 month*, and *Patients switched to a second-line regimen at 12 months*), and 2) Patients consecutively picking up ART on or after the EWI sample start date of January 1, 2010, regardless of duration of regimen (*On-time ARV drug pick-up*). Sample size calculations were performed to provide a 95% CI of ±7%, assuming a true prevalence of 50%; this provided the most conservative estimate of the sample size required [@pone.0065653-World1]. For ARV drug-supply continuity, data were abstracted on stock-outs of each ARV drug in routine use from January 1, 2009 until December 31, 2009 using a stock-take module in EDT, which records the stocks in the central store and dispensary at each site. Results {#s3} ======= Namibia abstracted data on three EWIs: *ART prescribing practices, Patients LTFU at 12 months,* and *Patients switched to a second-line regimen at 12 month* at 35 main ART sites ([Figure 1](#pone-0065653-g001){ref-type="fig"}). Two ART sites did not have complete data available so were removed from EWI analyses. Data from 3,875 patients were abstracted and analyzed. Site-specific EWI results are presented in [Table 2](#pone-0065653-t002){ref-type="table"}, and the national EWI summary is presented in [Table 3](#pone-0065653-t003){ref-type="table"}. *On-time ARV drug pick-up* and *ARV drug-supply continuity* could not be monitored because information entered into existing patient records was found to be incomplete or inaccurate. ART site profiles are presented in [Table 4](#pone-0065653-t004){ref-type="table"}. {#pone-0065653-g001} 10.1371/journal.pone.0065653.t002 ###### Namibia site-specific EWI results. {#pone-0065653-t002-2} ART site Percent appropriate initial ART regimen prescriptions Percent starting first-line ART lost to follow-up at 12 months Percent starting first-line ARTwhose regimen was switched to second line during first 12 months ---------- ------------------------------------------------------- ---------------------------------------------------------------- ------------------------------------------------------------------------------------------------- 1 89/89 (100) **23/98 (23)** 0/93 (0) 2 120/120 (100) 16/111 (14) 0/119 (0) 3 74/74 (100) **21/67 (31)** 0/72 (0) 4 127/127 (100) **35/127 (28)** **4/128 (3)** 5 **109/110 (99)** **38/108 (35)** **2/106 (2)** 6 22/22 (100) **7/19 (37)** 0/19 (0) 7 153/153 (100) **50/155 (32)** **3/100 (3)** 8 **173/175 (99)** **53/168 (32)** 0/150 (0) 9 **99/100 (99)** 17/92 (19) **3/92 (3)** 10 144/144 (100) **32/139 (24)** **1/108 (1)** 11 120/120 (100) **29/115 (25)** **1/112 (1)** 12 98/98 (100) 9/95 (10) 0/93 (0) 13 100/100 (100) 13/97 (13) **1/93 (1)** 14 **99/100 (99)** **21/95 (22)** **1/90 (1)** 15 **99/100 (99)** 13/98 (13) 0/99 (0) 16 133/133 (100) 9/119 (8) 0/122 (0) 17 61/61 (100) **13/48 (27)** **1/56 (2)** 18 100/100 (100) 3/95 (3) 0/95 (0) 19 **74/75 (99)** 9/69 (13) 0/74 (0) 20 180/180 (100) 7/175 (4) 0/82 (0) 21 100/100 (100) 6/99 (6) 0/94 (0) 22 **174/176 (99)** **37/175 (21)** **2/172 (1)** 23 155/155 (100) 19/148 (13) **1/132 (1)** 24 **144/145 (99)** 20/134 (15) **2/128 (2)** 25 160/160 (100) 17/159 (11) 0/92 (0) 26 100/100 (100) **35/93 (38)** **2/93 (2)** 27 **99/100 (99)** 18/99 (18) **3/97 (3)** 28 175/175 (100) 30/175 (17) **1/165 (1)** 29 **142/145 (98)** **41/124 (33)** **6/107 (6)** 30 110/110 (100) 7/107 (7) **2/95 (2)** 31 **117/120 (98)** **44/119 (37)** **9/108 (8)** 32 42/42 (100) **26/38 (68)** 0/30 (0) 33 154/154 (100) 13/150 (9) 0/133 (0) Bold data indicates a site has not achieved the WHO-recommended target for that EWI. ART - Antiretroviral therapy. LTFU - Lost to follow-up. EWI - Early Warning Indicator. Cohort of patients initiating ART at the site on or after the sample start date (July 1, 2008). 10.1371/journal.pone.0065653.t003 ###### National EWI summary. {#pone-0065653-t003-3} EWI EWI target for all sites (time period) Number of sites meeting EWI target(% of sites meeting target) ------------------------------------------------------------------------------- ---------------------------------------- --------------------------------------------------------------- Percent appropriate initial ART regimen prescriptions Target: 100% (7/1/08--6/30/09) 22 of 33 (67%) Percent starting first-line ART lost to follow-upat 12 months Target: ≤20% (7/1/08--6/30/09) 17 of 33 (52%) Percent starting first-line ART switched to second lineduring first 12 months Target: 0% (7/1/08--6/30/09) 15 of 33 (45%) ART - Antiretroviral therapy. EWI - Early Warning Indicator. 10.1371/journal.pone.0065653.t004 ###### ART site profiles. {#pone-0065653-t004-4} Site Year ARTstarted Geographical location (Region) Patients on ART Pre-ART patients ART starters in 12 month ART prescriber: patient ratio ------ ----------------- -------------------------------- ----------------- ------------------ -------------------------- ------------------------------- 1 2004 Kavango 821 1927 239 1∶205 2 2004 Ohangwena 3600 3157 489 1∶900 3 2004 Ohangwena 6157 4757 815 1∶1231 4 2004 Omaheke 1051 2149 396 1∶263 5 2004 Otjozondjupa 2115 91 408 1∶529 6 2004 Karas 685 164 152 1∶685 7 2004 Caprivi 5780 1894 1089 1∶723 8 2004 Khomas 3697 4498 1746 1∶616 9 2005 Kunene 328 142 115 1∶82 10 2003 Khomas 5500 6500 1800 1∶917 11 2004 Karas 2296 1154 450 1∶574 12 2005 Hardap 716 370 250 1∶179 13 2004 Kavango 1999 688 422 1∶666 14 2004 Kavango 1114 1731 941 1∶371 15 2005 Otjozondjupa 2000 NA 228 1∶1000 16 2004 Omusati 2247 1843 535 1∶562 17 2005 Otjozondjupa 338 49 86 1∶169 18 2004 Eenhana 1438 973 319 1∶360 19 2005 Erongo 700 528 156 1∶140 20 2000 Oshikoto 15971 3000 1897 1∶1597 21 2004 Kunene 851 589 200 1∶106 22 2003 Oshana 13967 4485 1746 1∶2327 23 2003 Omusati 2872 6709 792 1∶319 24 2004 Otjozondjupa 1577 1519 473 1∶1577 25 2003 Omusati 4534 4136 2000 1∶2267 26 2004 Kunene 567 814 279 1∶567 27 2003 Hardap 708 NA 290 1∶64 28 2003 Kavango 5052 2648 1120 1∶1684 29 2004 Erongo 2600 1417 445 1∶520 30 2004 Omusati 1326 922 1048 1∶1326 31 2004 Oshikoto 1487 1513 324 1∶1487 32 2005 Erongo 376 108 166 1∶376 33 2003 Erongo 4063 2559 945 1∶677 ART - Antiretroviral therapy. NA - Data not available. ART Prescribing Practices {#s3a} ------------------------- Twenty-two of 33 (67%) sites met the target of 100% appropriate initial ART regimen prescriptions, according to national ART guidelines [@pone.0065653-Republic1]. ([Table 2](#pone-0065653-t002){ref-type="table"}--[3](#pone-0065653-t003){ref-type="table"}) Of the sites not meeting the target of 100%, all eleven sites achieved appropriate prescribing of 98--99%. No patient at any site was prescribed mono-or dual-therapy. Every patient who was prescribed an inappropriate first-line regimen was on an appropriate PI-based regimen. There were significant differences between pre-validated and validated data. Pre-validated data showed 8/33 (24%) sites met the target for *ART prescribing practices*. The validation process discovered that EDT automatic queries had included not only patients consecutively initiating ART for the first time, but some patients who were transferred in on ART from another site and patients who were on pre-exposure prophylaxis. EDT automatic queries also generated some incorrect starting regimens, which were revealed when validated with the dispensing record on EDT. Patients Lost to Follow-up at 12 Months {#s3b} --------------------------------------- Seventeen of 33 (52%) sites met the target of ≤20% LTFU 12 months after ART initiation. ([Table 2](#pone-0065653-t002){ref-type="table"}--[3](#pone-0065653-t003){ref-type="table"}) The range of LTFU rates between ART sites was large from 3% to 68%. Pre-validated data showed 13/33 (42%) sites met the target for LTFU. The validation process discovered discrepancies between EDT and Patient Care Booklets, including differences in dates of death, dates of transfer out, patients in-transit, and ART start dates. Patients Switched to Second-line during first 12 Months {#s3c} ------------------------------------------------------- Fifteen of 33 (45%) sites met the target of 0% of ART initiators switched to a second-line regimen during the first 12 months. ([Table 2](#pone-0065653-t002){ref-type="table"}--[3](#pone-0065653-t003){ref-type="table"}) In sites not achieving the target of 0%, very few patients had been switched to second-line regimens (range 1%--8%). Pre-validated data showed 0/33 (0%) sites met the target of 0% switch to second line. The validation process revealed discrepancies between the automatic query for regimen at 12 months when compared with that in the EDT dispensing record. On-time ARV Drug Pick-up {#s3d} ------------------------ Existing record keeping systems did not allow for monitoring of on-time pill pick-up at all ART sites. Following the 2009 EWI exercise, national record keeping tools were adapted to capture data necessary to monitor this EWI; however, these adjustments would not be expected to be available for the abstraction of these 2010 EWIs. ARV Drug-supply Continuity {#s3e} -------------------------- Existing record keeping systems did not allow for monitoring of ARV drug stock-outs for all sites because drug stock-outs were not adequately recorded. Discussion {#s4} ========== The purpose of implementing routine HIVDR EWI monitoring is to assess the extent to which ART programs are functioning to optimize prevention of HIVDR. EWI monitoring provides countries with an evidence base that can optimize programmatic functioning--both at the site and national level--thereby, leading to public health action to optimize patient care and minimize the emergence of preventable HIVDR. In 2010, Namibia implemented their second round of integrating EWI monitoring into routine national ART program functioning. With technical support from the WHO, Namibia adapted the WHO generic EWI abstraction guidance [@pone.0065653-World1] with its existing human capacity and health systems infrastructure. Based on clinical relevance and availability of data, Namibia selected five recommended WHO EWIs and scaled-up monitoring to 33 of 35 main ART sites throughout the country from the previous nine sites in 2009: *ART Prescribing Practices*, *Patients lost to follow-up 12 months after ART initiation*, *Patients switched to second-line ART during first 12 months*, *On-time ARV drug pick-up*, and *ARV drug supply continuity*. These 33 ART sites are meant to be representative of all public ART sites in the country. Overall, it was found that the existing pharmacy and medical record database (EDT, ePMS and Patient Care Booklets) allowed for valid monitoring of three EWIs: *ART Prescribing Practices*, *Patients lost to follow-up 12 months after ART initiation*, and *Patients switched to second-line ART during first 12 months*. Though the existing pharmacy and medical record database did not allow for adequate monitoring of the remaining two EWIs--*On-time ARV drug pick-up* and *ARV drug supply continuity*--the abstraction exercise itself served to inform the TWG about the limitations in the existing record-keeping system and give guidance on the necessary modifications needed for future EWI monitoring. The previous 2009 EWI pilot [@pone.0065653-Hong1] resulted in modifications to the data capture tools such that all five EWIs could be abstracted. However, there was insufficient time between the 2009 EWI pilot and 2010 scale-up for the effects of the 2009 program implementations to be reflected in the scale-up results. In 2008 to 2009, only 67% of ART sites met the WHO target of 100% for *ART Prescribing Practices*, which is somewhat lower than recently published data from other African settings (Bennett et al in 907 sites (74%) [@pone.0065653-Bennett2], Billong et al in 40 sites (90%) [@pone.0065653-Billong1], Dzangare et al in 81 sites (88%) [@pone.0065653-Dzangare1], and Sigaloff et al in 13 sites (85%) [@pone.0065653-Sigaloff1]). Importantly, although the percentage of sites meeting the target was low in Namibia, sites that did not meet the 100% target still had very few patients prescribed an inappropriate first-line regimen, and no patient were prescribed dual- or mono-therapy. All the inappropriate first-line regimens were in fact appropriate PI-based regimens which would not be expected to increase the risk of the emergence of HIVDR. Therefore, these data demonstrate overall successful implementation of national prescribing guidelines and training of ART staff in appropriate ART prescribing. Only 52% of sites met the WHO target of ≤20% for *Patients lost to follow-up 12 months after ART initiation*, which is not unlike recently published data from other African countries (Bennett et al in 794 sites (59%) [@pone.0065653-Bennett2], Billong et al in 40 sites (20%) [@pone.0065653-Billong1], Dzangare et al in 81 sites (50%) [@pone.0065653-Dzangare1], and Sigaloff et al in 13 sites (92.3%) [@pone.0065653-Sigaloff1]). These data suggest that many patients are being lost to care within their first 12 months of ART and/or many patients are transferring out to alternate ART sites without informing their ART site. This patient population may be at high risk for experiencing ART treatment interruptions and developing HIVDR. The broad range of LTFU rates between sites suggests that there may be factors at the site-level that may be influencing a site's ability to retain patients in care. These data have prompted the MoHSS to engage in public health action, specifically operational research to assess predictors of and reasons for LTFU and of intensifying defaulter tracing with the goal of re-engaging patients back into care within 48 hours of running out of pills. This intensification of defaulter tracing will include the following elements: 1) routinely updated patient locator information; 2) dedicated patient tracer; 3) generation of daily lists of defaulters; 4) daily tracing (phone and physical) of defaulters; and 5) standardized system of tracing deaths and transfers out. EWI results were also used in Papua New Guinea to strengthen their ART services by: 1) establishing a formalized referral system that documents patient transfers between clinics; and 2) regular review of patient clinic attendance and drug pickups to identify patients at risk for suboptimal adherence and ways to remove barriers to on-time pill pickup by providing subsidies for transportation and food for those in need [@pone.0065653-Daoni1]. Only 45% of sites achieved the target of 0% of *Patients switched to second-line ART during first 12 month*. However, in sites not meeting the target, very few patients were switched to second-line regimens during the first 12 months, suggesting appropriate physician prescribing practices and success in managing ARV toxicity and side effects through in-class substitutions. These data suggest that patients who were retained in care at 12 months had good clinical outcomes and were not failing therapy. Similarly to previously reported EWI data in Namibia [@pone.0065653-Hong1], data for *On-time ARV drug pick-up* were considered not to be a true reflection of population-level adherence in Namibia's ART sites. In Namibia, routine pharmacy dispensing practice is meant to include the routine counting of remnant pills (number of pills left over from the previous prescription) and the dispensing of a specified number of days of pills. However, the number of remnant pills was not routinely recorded, thus it was not possible to calculate the actual pill run-out date, necessary to monitor this EWI. Instead, the importance of these results lie in the operational lessons learned which can be applied to the monitoring of the EWI in future years. Similar to trends in other countries monitoring EWIs, Namibia has begun to build capacity to monitor medicine possession ratio (MPR defined as numbers of days of pills dispensed/number of days in the interval), as MPR may be better than on-time ARV pickup for identifying patients at risk for HIVDR [@pone.0065653-Sigaloff1], [@pone.0065653-Ma1]. MPR is not affected by variances in pill dispensing practices and next appointment dates by pharmacists, therefore, may allow for a more valid measurement of maximum ARV adherence. These data have prompted the MoHSS to engage in public health action, specifically operational research which showed MPR to be associated with short-term virologic suppression 6 months after starting ART in Namibia [@pone.0065653-Hong2]. Similarly to previously reported EWI data in Namibia [@pone.0065653-Hong1], it was not possible to assess *ARV drug supply continuity.* In the previous EWI pilot [@pone.0065653-Hong1], existing pharmacy records did not capture stock at the level of the site dispensary but rather at a more central level which limited the country's ability to assess drug stock at the dispensing point. Based on the pilot EWI exercise [@pone.0065653-Hong1], programmatic changes were implemented to capture stock at the ARV dispensing point. However, pharmacists did not enter in stock into the electronic system in real-time and were allowed to dispense to patients on "0" stock. Therefore, it appeared in the record that many drug stock-outs occurred even when ARV drugs were available to the patients. Recognizing the importance of monitoring this EWI, as ARV drug stock-outs can be a cause of treatment interruptions and HIVDR [@pone.0065653-Oyugi1]--[@pone.0065653-Eholi1], [@pone.0065653-VanOosterhout1], Namibia is in process of applying these operational lessons learned to the monitoring of this important EWI in future years. The successful 2010 EWI exercise and integration plan will provide Namibia a solid evidence base that can be used to make statements about national and site-specific programmatic functioning and potential HIVDR. This evidence base will serve to contextualize results from Namibia's surveys of acquired HIVDR in patients starting ART and from surveys of transmitted HIVDR in specific geographic regions. The EWI data training and monitoring process has mobilized the national ART program and its partners to institute adjustments in existing databases, which will facilitate monitoring of WHO recommended EWIs in the future and which will yield a more accurate assessment of overall programmatic functioning. Additionally, EWI monitoring has prompted the national program in further public health action, specifically engaging sites in dialogue regarding the need for standard pharmacy dispensing practices and the application of standardized application of national clinical outcome definitions. An EWI national meeting took place with representatives of all ART sites present. Because data quality assessment proved to be a critical component of EWI monitoring, this meeting was conducted with the goal of training ART sites how to validate their own EWI electronic data with their paper medical records. EWI data quality assessments revealed differences between pre-validated and validated data due to inconsistencies between record systems and lack of cross-linkages. Data quality assessment is paramount to appropriate reporting and, when routinely practiced, serves to strengthen performance of staff and highlight areas of weakness leading to stronger health care systems. Importantly, validated data gives program managers confidence in the data for use in operations, planning and decision-making. Thus, the results of data quality assessments can help inform changes that will improve patient monitoring systems and clinic practice in order to minimize the emergence of HIVDR. One important limitation of this exercise is that Namibia could evaluate only three out of the five selected EWIs. Strengthening ART record systems, as mentioned above, will help in alleviating this limitation in future EWI rounds. An additional limitation is that although data for satellite/outreach and IMAI sites were included within the main ART sites, they could not be disaggregated and analyzed separately. Plans are being made to strengthen record keeping systems to allow routine capture of data necessary to disaggregate these sites for analyses. An additional limitation of this exercise is that EWIs were not conducted at private ART sites which limit our ability to make broader statements on ART delivery as a whole in Namibia. Although private ART sites deliver ART at an individual level with regimen selection based on HIVDR testing, programmatic data on these ART-experienced sites would be valuable. Plans are currently being made to pilot EWIs at the private ART sites in Namibia in 2013. An additional limitation of this report is that pediatric patients were not included; however, pediatric EWI monitoring assessing the use and availability of pediatric formulations and weight-based dosing will be implemented. Despite lessons learned from the EWI pilot in Namibia (2009) [@pone.0065653-Hong1], ART site performance seem not to have improved in this second round of EWIs (2010); subsequent rounds of EWI monitoring would provide meaningful evaluation in terms of the EWI trends over time. Importantly, results from the first two years of EWI monitoring in Namibia have resulted in public health action, specifically performance feedback to individual ART sites, as well as to the national ART program. These data have resulted in important programmatic changes and operational research which will optimize patient care and minimize preventable HIVDR. Over the next five years, as capacity to monitor EWIs builds throughout Namibia utilizing WHOs updated EWI guidance [@pone.0065653-WHO1], the MoHSS will expand EWI monitoring to engage IMAI and satellite/outreach sites, the private sector, and pediatric patient populations. The Republic of Namibia Ministry of Health and Social Services (Andrew Ndishishi, Norbert P. Forster, Ella Shihepo, Anna-Maria Nitschke), WHO-Namibia (Magda Robalo, Tiruneh Desta), Management Sciences for Health/Systems for Improved Access to Pharmaceuticals and Services funded by USAID (David Mabirizi), Namibia Institute of Pathology (Boniface Makumbi), Tufts University School of Medicine (Christine Wanke, Alice Tang), and WHO-Geneva (Silvia Bertagnolio). [^1]: **Competing Interests:**All authors above have no conflicts of interests except one author is employed by 'Abt Associate Inc'. This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials. [^2]: Conceived and designed the experiments: SYH AJ MRJ. Performed the experiments: SYH MS MD MG A. Badi VS DP A. Blom SM LJ KL. Analyzed the data: SYH MS MD MG VS DP A. Badi SM LJ KL. Wrote the paper: AJ JG MS MD MG A. Badi VS DP A. Blom SM LJ KL MRJ SYH.

1. Introduction {#sec1} =============== Although over the last decades, epidemiological data reveal a significant decline in caries prevalence in children and adolescents from industrialized Western countries \[[@B1], [@B2]\], dental caries is still a common oral disease, with a prevalence ranging around 60--90% \[[@B3], [@B4]\], even diffused among preschool children. Dental caries negatively affects children\'s quality of life by their psychological and social relations, also causing feelings of discomfort from an early age \[[@B5], [@B6]\]. Furthermore, dental caries leads to considerable cost in the short and long term \[[@B7]\]. For these reasons, a reduction in the prevalence of dental caries is desirable \[[@B8]\] and identifying high-risk populations for developing dental caries can help in achieving this goal by targeting the right audience with preventive strategies \[[@B7], [@B9]\]. Several studies and a recent meta-analysis have shown that education level and socioeconomic status (SES) of children\'s parents significantly influence their oral health status \[[@B10]--[@B12]\]. This could probably be related with the strong influence of these factors on the dietary habits \[[@B13], [@B14]\], the consumption of sweets being an important cause in the aetiology of caries, together with teeth form, salivary flow \[[@B15]\], and oral hygiene \[[@B16], [@B17]\]. There is a convincing evidence of an association between the amount and frequency of sugary snacks and an increased risk of caries \[[@B18]\], but different predictors could play a role in determining the final influence of the family\'s SES on children\'s dental health. Additionally, the influence of SES on dental health has been shown to be stronger for preschool children than older children \[[@B19], [@B20]\], because the preschool period is the time when wrong oral health habits, caries patterns, and risk factors are being established, and thus, it is the ideal time to act with preventive strategies to establish healthy trends which can have a lifelong influence \[[@B21], [@B22]\]. Thus, identifying the risk predictors in preschool children can be useful to set a primary prevention in this population and maintain a low caries risk status \[[@B23]\]. In the literature, there are few studies based on European samples assessing that SES significantly influence the development of dental caries in preschool children \[[@B3], [@B10], [@B19]\]. However, in Italy, this kind of studies did not involve the entire national territory, but only samples from Italian islands, reflecting the conditions of a particular population \[[@B2], [@B20]\]. They stated an overall prevalence of dental caries of 38.56% among 5-year olds \[[@B2]\] and of 42.5--57.5% among 6-year olds \[[@B20]\], and the maternal educational level was found to be a protective factor for the presence of caries \[[@B2]\] and also associated with children\'s caries severity \[[@B20]\]. Furthermore, this aspect was insufficiently analysed in Southern Europe, limiting the development of preventive strategies in this area. Thus, the aim of this cross-sectional study was to assess the association among oral health-related behaviors, socioeconomic factors, and dental caries in Italian preschool children. 2. Materials and Methods {#sec2} ======================== The present survey was performed in 10 public schools belonging to "Unione dei Comuni Vallata del Tronto," in the province of Ascoli Piceno (Marche, Italy) in central Italy. The criteria of selection for the 10 schools were them belonging to the same region of central Italy, the availability of the school director to carry out the project, and the possibility of conducting clinical examinations observing the appropriate health rules. The project consisted of a clinical examination in a classroom and the distribution of questionnaires to be completed by children\'s parents at home. Parents were preliminary informed about the aims and methodologies and encouraged to submit their children to the project. They were also asked to sign a proper informed consent form, after which their children were enrolled in the study. A total sample of 514 preschool children from 3 to 6 years (280 males and 233 females; mean age 4.58 years) were enrolled in the study, all attending the 1st, the 2nd, or the 3rd year of nursery school, among which 484 children were native Italians, while 30 children were from other ethnic groups. The survey lasted complessively eight days. First, the children underwent the clinical examination in comfortable classrooms, with good lighting. Disposable kits, consisting of disposable gloves, a dental probe, a mouth mirror, and a pair of tweezers, were used. Dental caries diagnosis was assessed according to the visual examination method by the British Association for the Study of Community Dentistry (BASCD): all surfaces of each eligible tooth were examined and scored. Caries was diagnosed visually at the "caries into dentine" level. In cases of doubt, the visual examination was followed, if needed, by a dental probing (by using a dental explorer *no*. 23). \[[@B24]\] At the end of the clinical examination, each child was categorized on the basis of his/her caries experience in the primary dentition ("Yes" when dmf-t was \>0 and "No" for caries-free children). The clinical examinations were conducted by one calibrated examiner: the examiner calibration has been previously realized over 30 children (equivalent to the 36.5% of the whole sample to examine), recruited in the Paediatric Dentistry Department of the University of L\'Aquila (Italy) where the entire survey was projected. The results over 30 preschool children were that there was a 100% agreement between the diagnoses made by the operator and those made by 2 expert colleagues after a customary dental examination on the professional chair and with appropriate light, with full sensitivity and specificity. In the second phase of the project, after the clinical examination in the classrooms, a questionnaire about mouth hygiene, eating habits, and socioeconomic factors were given to the parents of the involved children asking them to complete it honestly and appropriately, at home. Data of the questionnaire were recorded and reported in a descriptive analysis in order to illustrate the characteristics of the two groups: children with caries experience in the primary dentition, *versus* caries-free children. Frequency and percentages were calculated for discrete and nominal variables, while quantitative variables were evaluated through the median and the interquartile range (IQR). Since the data have a nonnormal distribution according to the Shapiro--Wilk test, a nonparametric analysis was conducted in order to determine the differences between the average values, and the *χ*^2^ test or Fisher\'s exact test was applied to evaluate the frequency differences between the two groups. Univariate and multivariate logistic regression models were used. In these models, the caries experience ("Yes" or "No") was considered as a dependent variable, and the contribution to this condition of demographic, socioeconomic, eating and oral hygiene habits was considered as explanatory variables. In order to study the relationship between the dependent variable and the explanatory variables, a cross-tabulation analysis was initially performed among groups and the *χ*^2^ test was used. Then, only the variables with a statistically significant association with caries experience were introduced to the logistic regression model, with the presence/absence ("Yes" or "No") of caries experience as dependent variable. For each analysis, the threshold for statistical significance was set at *p* \< 0.05. The data were processed through the statistical package STATA/IC 12.0. The study was conducted in accordance with the Declaration of Helsinki (1964) and approved by the ethical committee of the University of L\'Aquila, Italy. 3. Results {#sec3} ========== Out of the whole sample of 514 children, 419 participants (228 males, 191 females) were caries free, and 95 children had caries experience in the primary dentition (53 males, 42 females) corresponding to the 18.4% of participants. The *χ*^2^ test assessed that age and gender were not significantly associated with the dependent variable (respectively, *p*=0.109 and *p*=0.874). The most affected teeth were the second lower primary molars, followed by the first lower primary molars and by the first upper primary molars. The characteristics of the whole sample were as follows: the majority of children had Italian parents with high school level of education and came from families with an average income between €12.000 and €24.000 and home ownership ([Table 1](#tab1){ref-type="table"}). Regarding the eating habits, the great part of the sample reported to have 2-3 meals a day, without eating sugars before going to bed ([Table 2](#tab2){ref-type="table"}). The majority of the sample had received breastfeeding for a period of about 10 months ([Table 2](#tab2){ref-type="table"}). Regarding oral health, the great part of the sample reported brushing teeth about 2 times a day, even when tired; they have not been visited by a dentist before the present project, and they have never received any fluoride application on teeth, although they used to brush his/her teeth with a fluoride toothpaste ([Table 2](#tab2){ref-type="table"}). In addition, the majority of children has not received teeth sealings yet ([Table 2](#tab2){ref-type="table"}). [Table 3](#tab3){ref-type="table"} shows the results of univariate and multivariate logistic regression models. The unadjusted analysis evidenced a number of factors that, taken individually, can affect the caries experience in the primary dentition, which are the mother\'s non-Italian nationality (*p*=0.001); a family income lower than 12000 Euro a year (*p*=0.002); the absence of a family own residence (*p*=0.003); a number of 2 times that the patient habitually eats between the two meals (*p*=0.002), a number of dental visits in a year (*p*=0.004); the consumption of candies and sweets between meals (*p*=0.003); the consumption of sweet drinks or sweet food before going to bed (*p*=0.042). Finally, the mothers\' perception of children\'s caries experience was also significantly associated with the caries experience (*p*=0.001). The adjusted multivariate step revealed that factors significantly affecting the caries experience were the mother\'s non-Italian nationality; the number of dental visits in a year; and concerns of caries expressed by the mother. Mothers with non-Italian nationality were 6.24 times (C.I. 11.12--18.22; *p*=*p*.004) more exposed to caries experience of their children, compared to Italian mothers. And a personal impression of the mothers about the caries experience of their children was also a significant predictor. In addition, patients who received almost one visit for year were less exposed to caries experience compared to children who did not. 4. Discussion {#sec4} ============= The present study analysed a sample of preschool children in central Italy, demonstrating that some socioeconomic factors and habits resulted as risk factors for dental caries development in preschool children. Among the socioeconomic factors, the maternal foreign nationality was identified as the main risk factor. A similar result was also reported in northern Europe (the Netherlands) by van der Tas et al. \[[@B3], [@B7]\] that observed a similar result in a greater sample than the present one. Those authors stated that the maternal education level and their foreign nationality are the most influential risk factors for caries development in preschool children. Thus, both in northern and southern Europe, differently from other continents, there seems to exist a negative association between the immigrant status and oral health, as also reported by Noymer and Lee \[[@B25]\] for the general health. In Italy, Carta et al. \[[@B20]\] and Pizzo et al. \[[@B2]\], respectively, from Sardinia and Sicily islands, observed a strongest negative influence of the maternal education level on their preschool children\'s oral health, rather than economic factors, data which agree with the present study. Educational level is an important marker of socioeconomic condition and maternal educational level was previously identified as one of the best predictors in all countries for children\'s health including oral health \[[@B2], [@B10], [@B26]\], especially when the disease prevalence is high \[[@B13]\]. Previous studies enlightened a gap between Italy and other EU countries, where there is a screening program across the population (infantile, adolescent, and adult) at regular intervals of ten years, which allowed, from 1973, to reduce drastically the prevalence of caries, especially in preschool age \[[@B27]--[@B29]\]. On the contrary, the preschool caries risk pattern in Italy appears to be similar to that reported by van der Tas et al. \[[@B3]\] and the caries-free participants percentage even higher in Italy (81.6% vs 77.1%). This is still far from what was asked by the World Health Organization (WHO) targets for 2010, which were to provide for a reduction of caries, with an average of caries-free participants equal to approximately 90% among preschool children \[[@B30]\]. This difference from the reported gap may be due to the variation of several heterogeneous factors, which could be represented by disparities in socioeconomic development, dietary habits, parental attention and attitudes in child oral hygiene, and access to public dental health service \[[@B31], [@B32]\]. The last aspect, as demonstrated in the international scientific literature, can play an important role in influencing caries distribution: the presence of public dental care services and of national preventive programs can contribute to reduce caries in populations with least social, economical, and cultural levels and can also balance the disparities between local and immigrant people \[[@B33]\]. In particular, analysing the dietary habits and parental attention to children\'s dental health, the present study showed a predominance of the reduced assiduousness to the dental specialist as a primary caries risk factor in preschool children. Central Italy, analysed in this study, is particularly representative of the whole nation being observed as an intermediate situation between the northern and southern parts of the country. A marked discrepancy between the various ethnic groups, with a notable presence of caries in children of foreign origin, was also previously observed \[[@B33]\] and confirmed by the findings of the present study. This study showed the necessity to carry out more epidemiological surveys fairly distributed across the national territory and consequently to implement an ever-increasing number of prevention campaigns among the populations that may involve dental practitioners, paediatricians, teachers, parents, and children in order to reduce the incidence of the disease. Disparities in socioeconomic development, dietary habits, parental attention and attitudes in child oral hygiene, and access to public dental health services represent important risk factors for the development of dental caries in preschool children. In particular, the maternal foreign origin seems to be related to a higher caries risk among other factors. Data analysis suggests monitoring of the preschool children population over time through future preventive surveys and planning local preventive actions, aimed to increase the caries-free participants percentage in the local population considering socioeconomic and daily habits predictive factors, especially in relationship with the increasing immigration in southern Europe during the last decades. The data obtained from the examined sample are acceptably in line with what were observed in northern Europe. On the basis of the present results, an appropriate strategy to reduce the discrepancy in the oral health of children consists in implementing the educational programs aimed to instruct their mothers (in particular those with foreign nationality) in order to raise their attention to their children\'s oral health. These information and prevention programs could be implemented in kindergartens, which are a point of socialization and integration of families. This would further improve the educational offer of the schools that carry them out. This study was limited because the sample was from a single region of Central Italy; thus, the results cannot be generalized to the whole population from southern Europe. In addition, another limit of the present investigation is that the data evidences caries experience, but caries severity (number of decayed, filled, or missed teeth) was not evaluated. 5. Conclusions {#sec5} ============== Out of the whole sample of 514 preschool children from Central Italy (the province of Ascoli Piceno) (aged from 3 to 6 years; mean age: 4.58 years), 419 participants (228 males, 191 females) were caries free, and 95 children showed caries experience in the primary dentition (53 males, 42 females), corresponding to the 18.4% of participants. The multivariate logistic model revealed that factors significantly affecting the presence of caries were the mother\'s nationality, the number of dental visits in a year, and caries concerns expressed by the mother. Data Availability ================= The data used to support the findings of this study are included within the article. Disclosure ========== This research has been presented as a poster at the 14th EAPD Congress Palazzo dei Congressi Lugano, Switzerland 2018. Conflicts of Interest ===================== The authors declare that there are no conflicts of interest regarding the publication of this paper. Authors\' Contributions ======================= Simona Tecco and Gianmaria F. Ferrazzano contributed equally to the study. ###### Socioeconomic factors associated with caries experience in the primary dentition (*N* = 514). Socioeconomic factors *N* Caries experience in the primary dentition *p* value --------------------------------------- ----- -------------------------------------------- ------------ ------------- Mother\'s nationality  Italian, *n* (%) 259 228 (88.0%) 31 (11.0%) 0.001^*∗∗*^  Non-Italian, *n* (%) 30 13 (43.3%) 17 (56.7%) Father\'s nationality  Italian, *n* (%) 264 230 (87.2%) 34 (12.8%) 0.001^*∗∗*^  Non-Italian, *n* (%) 24 10 (41.7%) 14 (58.3%) Income  Higher than €36.000, *n* (%) 39 35 (55.5%) 4 (44.5%) 0.001^*∗∗*^  Between €24.000 and €36.000, *n* (%) 80 70 (84.0%) 10 (16.0%)  Between €12.000 and €24.000, *n* (%) 106 89 (87.5%) 17 (12.5%)  Lower than €12.000, *n* (%) 36 20 (89.7%) 16 (10.3%) Own residence  Yes, *n* (%) 221 192 (86.8%) 29 (13.2%) 0.002^*∗∗*^  No, *n* (%) 65 46 (70.8%) 19 (29.2%) Mother\'s title of study  University degree, *n* (%) 48 42 (87.5%) 6 (12.5%) 0.028^*∗*^  High school diploma, *n* (%) 158 137 (86.7%) 21 (13.3%)  Middle and primary school, *n* (%) 80 59 (73.7%) 21 (26.3%) Father\'s title of study  University degree, *n* (%) 27 25 (92.6%) 2 (7.4%) 0.157 ns  High school diploma, *n* (%) 152 129 (84.9%) 23 (15.1%)  Middle and primary school, *n* (%) 107 84 (78.5%) 23 (21.5%) ^*∗∗*^ *p* \< 0.01 (*χ*^2^ test or Fisher\'s exact test); ^*∗*^*p* \< 0.05 (*χ*^2^ test or Fisher\'s exact test); ns: not significant. ###### Association among caries experience in the primary dentition and the answers to the questionnaire concerning oral health-related behaviors. Questions *N* (*N* = 513) Caries experience in the primary dentition *p* value ----------------------------------------------------------------------- ----------------- -------------------------------------------- ------------ --------------- How often does the patient brush his teeth? 2 (1-2) 2 (1-2) 2 (1-2) \^0.077 ns How often does the patient eat between meals?          Median (IQR) 1 (1-2) 1 (1-2) 2 (1-3) \^0.002^*∗∗*^ Does the patient often have candies and sweets between meals?          No 147 132 (89.8%) 15 (10.2%) °0.003^*∗∗*^  Yes 141 108 (76.6%) 33 (23.4%) If yes, how often?       \^0.671 ns  Median (IQR) 1 (1-2) 1 (1-2) 1 (1-2) Has the patient ever been visited by a dentist?          Yes 105 83 (79.0%) 22 (21.0%) °0.139 ns  No 183 157 (85.8%) 26 (14.2%) If yes, how many times a year?       \^0.001^*∗∗*^  Median (IQR) 1 (1-2) 1 (1-2) 1 (1-2) Topical fluoride application?          Yes 47 39 (82.98) 8 (17.02) °0.924 ns  No 237 198 (83.54) 39 (16.46) Has the patient taken or is he still taking fluoride?          Yes 116 100 (86.21) 16 (13.79) °0.273 ns  No 171 139 (81.29) 32 (18.71) Is the patient using a fluoride toothpaste?          Yes 228 189 (82.89) 39 (17.11) °0.734 ns  No 59 50 (84.75) 9 (15.25) Does the patient have sealings?          Yes 11 7 (63.64) 4 (36.36) °0.075 ns  No 276 232 (84.06) 44 (15.94) Breastfeeding?          Yes 242 199 (82.23) 43 (17.77) °0.139 ns  No 45 41 (91.11) 4 (8.89) If yes, how long?          \<10 months 161 137 (85.09) 24 (14.91) °0.101 ns  ≥10 months 81 62 (76.54) 19 (23.46) Do you think your child may have caries?          No 244 223 (91.39) 21 (8.61) °0.001^*∗∗*^  Yes 43 17 (39.53) 26 (60.47) Does the patient have sweet drinks or sweet food before going to bed?          No 201 174 (86.57) 27 (13.43) °0.039*∗*  Yes 86 66 (76.74) 20 (23.26) When (age in months) did you start brushing his/her teeth?          Median (IQR) 24 (15 36) 24 (12 36) 29 (24 36) \^0.010^*∗∗*^ Do you always brush your child\'s teeth even when tired?          Always 80 66 (82.50) 14 (17.50) °0.891 ns  Often 111 94 (84.68) 17 (15.32)  Sometimes 83 70 (84.34) 13 (15.66)  Never 13 10 (76.92) 3 (23.08) Do you think children\'s teeth are important?          Of course 225 188 (83.56) 37 (16.44) °0.989 ns  Yes 55 46 (83.64) 9 (16.36)  No 7 6 (85.71) 1 (14.29) °*χ*^2^ test or Fisher\'s exact test; \^Mann--Whitney test; ^*∗∗*^*p* \< 0.01; ^*∗*^*p* \< 0.05; ns: not significant. ###### Results from the logistic regression analysis considering the caries experience as dependent variable. Variable Association between caries experience in primary dentition and the considered variables using a univariate logistic regression model. Multivariate logistic regression model ----------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------- ------------- ------ ------------- ------------- Mother\'s nationality  Italian° 1     1   0.004^*∗∗*^  Non-Italian 8.64 4.11--22.9 0.001^*∗∗*^ 6.24 1.12--18.22 Income  Higher than €36.000° 1     1      Between €24.000 and €36.000 1.25 0.37--4.27 0.722 ns 1.32 0.37--4.72 0.668 ns  Between €12.000 and €24.000 1.67 0.53--5.32 0.384 ns 1.23 0.37--4.15 0.735 ns  Lower than €12.000 7 2.05--23.85 0.002^*∗∗*^ 3.56 0.88--14.45 0.075 ns Own residence  Yes° 1     1      No 2.73 1.41--5.30 0.003^*∗∗*^ 1.71 0.74--3.98 0.211 ns Number of times the patient eats between the meals 1.74 1.23--2.47 0.002^*∗∗*^ 0.49 0.21--1.18 0.112 ns Candies and sweets between meals:  No° 1 1.39--5.21 0.003^*∗∗*^ 1 1.01--3.22 0.124 ns  Yes 2.69 1.32 Number of dental visits every year 2.36 1.32--4.20 0.004^*∗∗*^ 3.43 1.18--9.95 0.024^*∗*^ Do you think your child may have caries?  No° 1     1      Yes 16.24 7.61--34.64 0.001^*∗∗*^ 6.59 1.19--36.48 0.031^*∗*^ Does the patient have sweet drinks or sweet food before going to bed?  No° 1     1      Yes 1.95 1.03--3.72 0.042^*∗*^ 3.54 0.01--45.92 0.143 ns ^*∗∗*^ *p* \< 0.01; ^*∗*^*p* \< 0.05. [^1]: Academic Editor: Adam Reich

1. Introduction =============== Diets rich in soybean products are associated with reduced incidence of cardiovascular disease, osteoporosis and certain human cancers \[[@b1-ijms-08-00651]--[@b3-ijms-08-00651]\]. Animal studies have shown that soy diets block oxidative degradation of lipoproteins, prevent atherosclerosis of blood vessels and inhibit radiation-and carcinogen-induced tumors of various tissues. Soybeans contain ingredients with potential biological effects and isoflavone is one of the most extensively studied ingredients. Twelve isoflavones are found in soybean and are present in four chemical forms: malonylglucosides (malonyldaidzin, malonylgenistin, and malonylglycitin), acetylglucosides (acetyldaidzin, acetylgenistin, and acetylglycitin), glucosides (daidzin, genistin, and glycitin) and aglycones (daidzein, genistein, and glycitein). Among these, the glucoside is the predominate form in soybean \[[@b4-ijms-08-00651]--[@b5-ijms-08-00651]\]. *In vitro* and *in vivo* studies have shown that isoflavone aglycones such as genistein inhibit ultraviolet light-induced damage \[[@b6-ijms-08-00651]--[@b8-ijms-08-00651]\]. However, it has also been shown that isoflavone glycosides such as genistin, daizin and glycitin have biological activities on melanoma cells and myoblasts \[[@b7-ijms-08-00651],[@b9-ijms-08-00651]\]. Therefore, it is possible that aglycones and glycosides can show some physiological relevance. Soybean cake is a by-product during processing of soybean oil. Defatted soybean contains a high amount (2121.9 μg/g) of isoflavones \[[@b4-ijms-08-00651]\]. Kao and coworkers have isolated four isoflavone extracts, namely malonylglucoside, acetylglucoside, glucoside and aglycone, from soybean cake by chromatography \[[@b5-ijms-08-00651],[@b10-ijms-08-00651]\]. They found these isoflavone extracts possess differential antioxidant activities. For example, the acetylglucoside extract exhibits the highest efficiency in 1,1-diphenyl-2-picrylhydrazyl (DPPH) scavenging assay and the glucoside extract is the most efficient for chelating metal ions \[[@b5-ijms-08-00651]\]. Overall, the isoflavone extracts show better antioxidant activities than isoflavone standards \[[@b5-ijms-08-00651]\]. Therefore, the purpose of this study was to determine the protective effects of these four isoflavone extracts on Ultraviolet B (UVB)-induced keratinocyte damage, including their effects on UVB-induced hydrogen peroxide (H~2~O~2~) generation and mitogen-activated protein kinase (MAPK) signaling \[extracellular-regulated kinase (ERK1/2), p38 and c-jun N-terminal kinase (JNK)\] in keratinocytes. 2. Materials and Methods ======================== 2.1. Materials -------------- 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), aprotinin, leupeptin, phenylmethylsulfonyl fluoride (PMSF), sodium fluoride (NaF) and sodium orthovanadate were purchased from Sigma Chemical Co. (St Louis, MO). Antibodies raised against p38 and p-JNK were from Cell Signaling Technology (Beverly, MA). Antibodies raised against JNK, ERK1/2 and p-p38 were from R&D system, Inc. (Minneapolis, MN). Antibody raised against p-ERK1/2 was from Santa Cruz Biotechnology (Santa Cruz, CA) 2.2. Preparation of Isoflavone extracts --------------------------------------- Isoflavone extracts (group I\~IV) were prepared by a method as previously described by Kao *et al*. \[[@b5-ijms-08-00651]\]. A sample (50 g) was mixed with deionized water-ethanol (150 ml, 1:1, v/v) and the mixture was shaken for 2 h. After centrifuging at 6000 rpm for 20 min (25 °C), the supernatant was collected and filtered through a glass filter paper. Then soybean cake extract (80 ml) was poured onto the top of a glass column (375 × 45 mm I.D.) containing Diaion HP-20 adsorbent (200 g) which was prewetted with ethanol (1 l) and deionized water (1 l). The water-soluble impurities were removed with deionized water (400 ml), followed by water-ethanol (900 ml, 85:15, v/v) to elute malonylglucosides and water-ethanol (3300 ml, 73:27, v/v) to elute glucosides. The residual isoflavones (acetylglucosides and aglycones) were eluted with water-ethanol (200 ml, 30:70, v/v) and water-ethanol (400 ml, 5:95, v/v), respectively. The two fractions were mixed, then evaporated to dryness under vacuum and dissolved in isopropanol. The aglycone and acetylglucoside fractions were separated respectively by injection of isopropanol solution (20 ml) into a Yamazen Hi-Flash^TM^ silica gel column, with the mobile phase changed to n-hexane-isopropanol-ethanol (8:9:1, v/v/v) and flow rate adjusted to 20 ml/min. Each fraction was injected onto a High Pressure Liquid Chromatography (HPLC) system to monitor the composition and concentration of isoflavone. Contents in aglycone, glucoside, acetylglucoside and malonylglucoside isoflavone extracts are shown in [Table 1](#t1-ijms-08-00651){ref-type="table"}. 2.3. Cell Culture ----------------- Human immortalized keratinocytes (HaCaT cells) were maintained in Dulbecco's modified Eagle's medium (DMEM) with 10 % fetal calf serum (GibcoBRL, Invitrogen Life Technologies, Carlsbad, CA), 100 units/ml penicillin, and 100 μg/ml streptomycin (Sigma Chemical Co., St. Louis, MO). The cells were cultured in a humidified incubator at 37 °C and 5l % CO~2~. For most experiments, cells reaching a 90 % -- 95 % of confluency were starved in DMEM at 37 °C for 24 h. 2.4. Drug treatment and UVB irradiation --------------------------------------- Keratinocytes cultured on 1.5-cm or 3-cm culture dish (Costar, Cambridge, MA) were pretreated with isoflavone extracts for the indicated time. After two washes with phosphate-buffered saline (PBS) cells were incubated with DMEM (1 ml). Then, cells were irradiated by UVB in a Bio-Sun system illuminator from VL (Vilber Lourmat, France) with a UV peak at 312 nm. Ultraviolet B was supplied by a closely spaced array of two UVB lamps, which delivered uniform irradiation at a distance of 10 cm. UVB irradiation dose was 40 mJ/cm^2^. After UVB exposure, cells were fed with fresh DMEM containing isoflavones, incubated for the indicated time, and collected for further analysis. 2.5. Cell viability assay (MTT assay) ------------------------------------- The viability of cells was determined by the MTT assay as previously described \[[@b11-ijms-08-00651]\] with a minor modification. Briefly, PBS-or isoflavone-pretreated cells were exposed to UVB and incubated for an additional 24 h. After a brief wash with medium, MTT (0.5 mg/mL in DMEM) was used for the quantification of living metabolically active cells \[[@b12-ijms-08-00651]\]. Mitochondrial dehydrogenases metabolized MTT to a purple formazan dye, which was measured photometrically at 550 nm. Cell viability is proportional to the absorbance measured \[[@b13-ijms-08-00651]\]. 2.6. Flow cytometric analysis of intracellular H~2~O~2~ ------------------------------------------------------- Intracellular production of H~2~O~2~ was assayed as previously described \[[@b14-ijms-08-00651]\] with a minor modification. Briefly, confluent keratinocytes starved with DMEM were preteated with various concentrations of isoflavones for 12 h. Cells were washed with PBS and DMEM, and then treated with dihydrorhodamine 123 (10 μg/mL) in DMEM for 30 mins. After a brief wash, cells were irradiated by UVB and then were collected by scraping and centrifugation. The cell pellets were resuspended in 1 mL PBS and then analyzed immediately by flow cytometer (Partech GmBH, Munster, Germany) at excitation and emission wavelengths of 488 and 525 nm, respectively. Fluorescence signals of 10,000 cells were collected to calculate mean fluorescence intensity of a single cell. 2.7. Cell lysate preparation and Western blot analysis of JNK, ERK1/2 and p38 ----------------------------------------------------------------------------- Keratinocytes treated with or without UVB were washed with PBS twice. Cells were lysed in lysis buffer \[17 mM Tris--HCl, pH 7.4, 50 mM NaCl, 5 mM EDTA. 1 mM sodium fluoride, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 1 mM sodium orthovanadate, 1 mM PMSF, and 1 μg/mL aprotinin and leupeptin (freshly prepared)\]. After sonication, the lysate was centrifuged (14,000 *g* for 10 min at 4°C), and supernatant was removed. The protein content was quantified by Pierce protein assay kit (Pierce, Rockford, IL). Total protein was separated by electrophoresis on 10% SDS--polyacrylamide gels and the proteins were electroblotted onto PVDF membranes and then probed using the indicated specific antibodies. Immunoblots were detected by enhanced chemiluminescence (Chemiluminescence Reagent Plus from NEN, Boston, MA). 2.8. Statistical analysis ------------------------- Otherwise where indicated, data were expressed as mean ± standard errors (SE). Comparison of means of two groups of data was made by using the unpaired, two-tailed Student *t*-test. All data are analyzed by SigmaPlot 2002 for Windows Version 8.0. 3. Results and Discussion ========================= 3.1. Isoflavone extracts inhibited UVB-induced cell death --------------------------------------------------------- To determine protective effects of isoflavone extracts obtained from soybeans (group I\~IV, [Table 1](#t1-ijms-08-00651){ref-type="table"}) on UVB-irradiated keratinocytes, cell viability was determined by the MTT assay. [Figure 1](#f1-ijms-08-00651){ref-type="fig"} shows that isoflavone extracts alone didn't affect keratinocyte viability but keratinocytes viability was decreased after UVB exposure. However, UVB-induced cell death was inhibited by the treatment of group I\~IV isoflavone extracts at concentrations in 0.5 and 1 % ([Figure 1](#f1-ijms-08-00651){ref-type="fig"}). 0.5 % of isoflavone extracts was sufficient to exert maximum protective effect in UVB-irradiated keratinocytes. These results indicate that aglycone, glucoside, acetylglucoside and malonylglucoside isoflavone extracts are able to prevent UVB-induced human skin damage. 3.2. Isoflavone extracts inhibit UVB-induced H~2~O~2~ generation in keratinocytes --------------------------------------------------------------------------------- It has been shown that H~2~O~2~ is generated in cultured human skin cells during UVB irradiation \[[@b15-ijms-08-00651],[@b16-ijms-08-00651]\]. In addition, group I\~IV isoflavone extracts have been shown to possess antioxidant activity \[[@b5-ijms-08-00651]\]. Therfore, we determined whether these extracts affect UVB-induced intracellular H~2~O~2~ production. Intracellular H~2~O~2~ in keratinocytes exposed to UVB was measured by dihydrorhodamine 123 (DHR 123), which has been shown to react with H~2~O~2~ in the presence of peroxidase and is extensively used as a probe for the detection of intracellular H~2~O~2~ \[[@b16-ijms-08-00651];[@b17-ijms-08-00651]\]. As shown in [Figure 2](#f2-ijms-08-00651){ref-type="fig"}, flowcytometric analysis showed that mean fluorescence, i.e. H~2~O~2~ production, was increased in UVB-treated cells ([Figure 2A](#f2-ijms-08-00651){ref-type="fig"}, panels a and b). The increase of intracellular H~2~O~2~ by UVB irradiation was decreased by the treatment of these isoflavone extracts ([Figure 2](#f2-ijms-08-00651){ref-type="fig"}), while the basal level of intracellular H~2~O~2~ was not affected (data not shown). The result directly demonstrated that all isoflavone extracts have potent scavenging activity which can prevent UV induced intracellular H~2~O~2~ production. 3.3. Isoflavone extracts differentially inhibited UVB-induced MAP kinase signaling pathway ------------------------------------------------------------------------------------------ Ultraviolet B irradiation has been shown to activate ERK1/2, JNK, and p38 kinase \[[@b14-ijms-08-00651];[@b16-ijms-08-00651];[@b18-ijms-08-00651]\], which may lead to skin cell damage. Thus, we examined if these isoflavone extracts could affect UVB-induced MAPK activation. We observed that ERK1/2, JNK, and p38 phosphorylation were apparently increased in UVB-irradiated keratinocytes. The basal level of JNK phosphorylation was not affected by isoflavone extracts, but UVB-induced JNK activation was inhibited by group I\~III but not by group IV isoflavone extracts. Group III isoflavone extract elicited more inhibitory effect than group I and II ([Figure 3A](#f3-ijms-08-00651){ref-type="fig"}, upper panels). However, UVB-induced ERK1/2 and p38 activation in keratinocytes was not significantly affected by group I\~IV isoflavone extracts ([Figure 3B and 3C](#f3-ijms-08-00651){ref-type="fig"}, upper panels), suggesting that aglycone, glucoside, acetylglucoside and malonylglucoside isoflavones differentially affect MAPK activation. The inhibition by these isoflavones was not due to uneven loading because reprobing of the immunoblots with antibodies raised against total JNK, ERK1/2, and p38 showed equal loading of each sample ([Figure 3A, B, C](#f3-ijms-08-00651){ref-type="fig"}, lower panels). Discussion ========== In the present study we examined the protective effects in four groups of isoflavone extracts from soybean cake on UVB-induced keratinocyte damage. Treatment with isoflavone extracts (groups I\~IV) significantly increased cell viability in UVB-irradiated keratinocytes. Both glucoside (group II) and acetylglucoside (group III) extracts exhibited higher efficacy in preventing UVB-induced cell death than malonylglucoside and aglycone extracts ([Figure 1](#f1-ijms-08-00651){ref-type="fig"}). This is consistent with the results reported by Kao and coworkers \[[@b5-ijms-08-00651]\] stating that among these four isoflavone extracts, the glucoside and acetylglucoside extracts are relatively stronger antioxidants in scavenging DPPH free radical and chelating metal ions \[[@b5-ijms-08-00651]\]. This suggests that UVB protective effect of these isoflvones is related with their antioxidant activities. However, group I\~IV isoflavone extracts exerted similar inhibitory effects on UVB-induced H~2~O~2~ production ([Figure 2](#f2-ijms-08-00651){ref-type="fig"}). Since H~2~O~2~ and other reactive oxygen species play an important role in causing UVB-induced skin cell damage \[[@b19-ijms-08-00651]\], the scavenging effects on free radicals other than H~2~O~2~ by these isoflavone extracts may also contribute to their protective effect on UVB-induced cell death. UV irradiation has been reported to upregulate expression of transcription factors \[[@b20-ijms-08-00651];[@b21-ijms-08-00651]\], which is mediated by the sequential activation of cytoplasmic protein kinases. To date, three structurally related but biochemically and functionally distinct MAPK signal pathways have been identified, including ERK1/2, JNK, and p38 \[[@b22-ijms-08-00651]\]. Peus *et al.*\[[@b17-ijms-08-00651],[@b23-ijms-08-00651]\] have demonstrated that ERK1/2, JNK, and p38 are activated in human epidermal keratinocytes following exposure to UVB irradiation. Several lines of evidence have also supported that the activation of these kinases correlates with skin cell death in response to UVB \[[@b24-ijms-08-00651]\]. Assefa *et al*. \[[@b25-ijms-08-00651]\] have shown that UVB induce keratinocyte apoptosis and produce a sustained p38 activation, which in turn leads to cytochrome c release and procaspase-3 activation. Gene disruption studies have shown that JNK mediates UV-stimulated apoptosis via the mitochondrial pathway by a Bax/Bak-dependent mechanism \[[@b26-ijms-08-00651]\]. These studies indicate that at least p38 and/or JNK are involved in UVB irradiation-induced keratinocyte death. In the present study, we found that these four isoflavones differentially affected UVB irradiation-induced MAPK activation. Groups I\~III exhibited a pronounced inhibitory effect on UVB irradiation-induced JNK phosphorylation, whereas group IV did not. Moreover, these isoflavone extracts did not have inhibitory effects on UVB irradiation-induced p38 and ERK1/2 activation ([Figure 3](#f3-ijms-08-00651){ref-type="fig"}). However, all these isoflavone extracts reversed UVB irradiation-induced keratinocyte death ([Figure 1](#f1-ijms-08-00651){ref-type="fig"}), suggesting group IV may act through affecting other signaling pathways. Although group I\~IV isoflavone extracts can reverse UVB irradiation-induced keratinocyte death, it was found they have differential reversal effects ([Figure 1](#f1-ijms-08-00651){ref-type="fig"}). It has been reported that UVB irradiation dose at 40 mJ/cm^2^ induces not only H~2~O~2~ but also many DNA damages such as cyclobutane pyrimidine dimers and 6--4 photoproducts in keratinocytes \[[@b27-ijms-08-00651],[@b28-ijms-08-00651]\]. Therefore, an agent that can activate the DNA repair system may also have an anti-cell death ability. Based on this hypothesis and the mentions discussed above, group II and III isoflavone extracts possibly have relatively stronger antioxidant activity on ROS, inhibitory effect on JNK activation, and inductory effect on DNA repairing. However, whether they can directly activate the DNA repair system in human keratinocytes needs to be further determined. In conclusion, our study demonstrated that isoflavone extracts from soybean cake can inhibit UVB-induced intracellular H~2~O~2~ production, JNK activation, and consequently protect human keratinocytes against UVB-induced death. Since these isoflavone extracts are relative stable and obtained easily than isoflavone standards, soybean cake may be the source for developing effective agents in skin care against photoaging. This work was supported by the research grants from the National Science Council and from Fu-Jen Catholic University, Taipei, Taiwan. {#f1-ijms-08-00651} ![Group I\~IV isoflavone extracts inhibited UVB irradiation-induced intracellular H~2~O~2~ production in human keratinocytes. (A) Intracellular H~2~O~2~ production \[denoted by mean fluorescence (MF)\] was expressed as histogram and (B) the results from four independent experiments was analyzed. \**P* \< 0.05 vs. UVB-exposed cells without isoflavone treatment.](ijms-08-00651f2){#f2-ijms-08-00651} {#f3-ijms-08-00651} ###### Contents in group I\~IV isoflavone extracts from soybean. Extracts Isoflavone name Concentration (μg/ml) ----------------------------------------- ----------------- ----------------------- aglycone extract (**Group I)** daidzein 47.3 ± 0.5 genistein 36.4 ± 0.1 glycitein 16.3 ± 0.2 glucoside extract **(Group II)** daidzin 25.3 ± 0.1 genistin 45.9 ± 1.6 glycitin 29.2 ± 0.2 acetylglucoside extract **(Group III)** acetyldaidzin 19.7 ± 0.9 acetylgenistin 69.0 ± 1.6 acetylglycitin 12.8 ± 0.7 malonylglucoside extract **(Group IV)** malonyldaidzin 195.2 ± 1.2 malonylgenistin 140.4 ± 4.8 malonylglycitin 64.0 ± 0.8

Introduction ============ Around one billion adults worldwide smoke,[@ref1] with high prevalence in developing countries, where 49% of men and 11% of women use tobacco.[@ref2] Although the prevalence of current smokers has decreased over time in several countries, the global absolute number of smokers has increased owing to population growth.[@ref3] Policies have successfully encouraged people to quit, using aids such as nicotine replacement therapy and electronic cigarettes (e-cigarettes).[@ref4] In the Health Survey for England (2013 and 2014), 26% of current smokers reported that they wanted to cut consumption down but were not trying to stop, and 40-41% said that they smoked less than in the previous year.[@ref5] The percentage of smokers who consume one to five cigarettes per day has steadily risen (from 18.2% to 23.6% between 2009 and 2014[@ref5]), with a similar pattern in the US, where the proportion of smokers who consume less than 10 cigarettes per day increased from 16% to 27% between 2005 and 2014.[@ref6] A recent Cochrane review discussed the evidence for ways of helping smokers who wish to reduce their consumption.[@ref7] Smoking few cigarettes is generally believed to be relatively safe, as has been incorrectly assumed for light/low nicotine cigarettes.[@ref8] Among 24 658 US adolescents, 10% thought that light smoking was not harmful, and only 35% of light smokers considered their habits to be associated with "a lot of harm."[@ref9] Reducing consumption might be expected to reduce harm in a proportionate way---that is, that smoking one instead of 20 cigarettes per day has about one twentieth (5%) of the risk. This seems to be the case for lung cancer, for which the large American Cancer Society Prevention Study II showed an approximately linear relation between risk of lung cancer and number of cigarettes smoked per day, but the dose-response for cardiovascular disease is steep at low consumption and then levels off,[@ref10] consistent with the shape reported previously.[@ref11] In a seminal systematic review of second-hand smoke and coronary heart disease among never smokers published in the *BMJ* 20 years ago, Law and colleagues drew attention to the 1.30 risk ratio being relatively large compared with the 2-3 typically seen in studies of active smokers.[@ref12] Their conclusions on second-hand smoke were supported by a meta-analysis of active cigarette smoking and risk of coronary heart disease from five cohort studies, in which the modelled relative risk for smoking one cigarette per day (1.39) was consistent with that for exposure to second-hand smoke. Although the non-linear relation between coronary heart disease and low cigarette consumption has been reported before (individual studies, as well as official reports from the US Surgeon General), it still is still not commonly known by the general public or health professionals, particularly those not involved in tobacco and health. We thus aimed to extend the previous work on coronary heart disease,[@ref12] by using a systematic review to provide a major body of evidence. We also aimed to show that a similar non-linear relation exists between stroke and low cigarette consumption. Methods ======= Data sources and searches ------------------------- We did a systematic literature review of English language articles published between 1946 and May 2015 in Medline (MOOSE guidelines[@ref13]) that reported the association between cigarette consumption and coronary heart disease and stroke. Supplementary figure A shows the search terms and flowchart: 13 861 abstracts were reviewed (by DM and SB), and any selected for consideration had their reference list manually checked for additional studies. Several study reports were based on combining data from at least two separately conducted cohort studies. Study selection and data extraction ----------------------------------- We included prospective cohorts with at least 50 cardiovascular disease events (mortality, morbidity, or both) to minimise the potential for reporting bias, in which large but unreliable effects might be seen in small studies. Reports had to give hazard ratios from a Cox proportional hazards regression or relative risks based on incidence/mortality, which must have been adjusted by at least age, or incidence reported in age groups. Results had to be available in at least three smoking categories, not including the reference group of never smokers. The populations of the cohorts had to be generally healthy; we excluded studies based only on people at high risk (for example, taking drugs for cardiac related disorders). Results had to be given separately for men and women, or, if they were based on both combined, the hazard ratios must be adjusted for age and sex. We excluded six studies spuriously showing that the hazard ratio or relative risk decreased with increasing consumption (justification in supplementary figure A). Study characteristics extracted were country, time period, sex, smoking categories, incidence, hazard ratio or relative risk, number of participants, number of events, and confounding factors adjusted for. In the few instances in which only age adjusted incidence/mortality results were available, we calculated the relative risk in each smoking category. Most studies reported hazard ratios, and we always used hazard ratios adjusted for multiple factors when provided (supplementary table A); 30 of the 55 publications made allowance for multiple (at least two) factors in addition to age and sex when providing hazard ratios. We extracted hazard ratios and relative risks separately for coronary heart disease, stroke, or cardiovascular disease (coronary heart disease and stroke combined). Statistical methods ------------------- Hereafter, we refer to hazard ratio or relative risk as relative risk (consistent with many studies included). Instead of modelling risk with consumption for each study (which is non-linear), we modelled the logarithm of risk, using similar methods as before.[@ref12] [@ref14] This involved fitting a log-linear variance weighted regression model between incidence or relative risk and cigarette consumption (using all reported smoking categories in the publications). Although this approach makes the relation more linear (when examined on a log scale), it might still underestimate the increase in risk at very low consumption levels. We obtained a regression model for each study report separately (Stata software). For consumption, we used the midpoint of the reported number of cigarettes per day---for example, three cigarettes per day if the category was one to five cigarettes per day---which we then adjusted for carboxyhaemoglobin and cotinine because this allows for lower inhalation with increasing cigarette consumption as previously established.[@ref14] For studies that reported relative risks adjusted for age (or for additional factors), the model contained the logarithm of the relative risk (dependent variable) and consumption (independent variable) using only the midpoint of the cigarettes per day categories. For studies that reported incidence in each age category, we fitted log-linear model that contained incidence (dependent variable) and consumption (independent variable) with age as a covariate (median age in each age category), and we estimated the relative risk by using an interaction term between age and consumption. This provided estimates in each age category (45, 55, and 65 years) because the risk of cardiovascular disease changes with age.[@ref15] The reference value of 1.0 (never smokers) was not included in the regression to avoid forcing the model through the origin and unduly affecting the dose-response relation (also because we were ultimately interested only in comparing between high and low consumption). We used the standard error of the logarithm of the relative risk, or the number of events if the standard error was unavailable, as weights in the regression; if both were unavailable, we did an unweighted log-linear regression for the study. The reference group was lifelong never smokers, although in seven reports it was unclear whether former smokers might have been included. The main quantitative measure was the percentage change in risk (excess relative risk) associated with smoking one (or five) cigarette(s) per day, expressed as a proportion of the percentage change for smoking 20 cigarettes per day. For example, if the relative risks were 1.4 and 1.9 for smoking one and 20 cigarettes per day, respectively, the proportion of excess relative risk associated with one cigarette per day is 44%: (1.4−1)/(1.9−1)×100. One or five cigarettes per day reflect typical levels of low consumption. We did three different types of analyses, to check for consistency. Firstly, from each regression analysis for each study, we used the model to estimate the relative risk for smoking one cigarette per day compared with never smokers, and also for smoking five and 20 cigarettes per day. We then calculated the excess relative risks for one and five cigarettes per day (compared with 20) and took the median value of each of these across studies. We did multiple separate analyses according to combinations of sex and disease type ("within study" analyses). Secondly, we obtained a single regression model across all studies (again done separately for each combination of sex and disorder) by using the individual cigarettes per day values and reported relative risk estimates (log scale) in a random effects meta-regression (SAS Proc Mixed). We then used the pooled coefficients to estimate the relative risk for one, five, and 20 cigarettes per day (another "within study" analysis). We also used these regressions to examine whether a quadratic trend might be better than a linear trend but found no evidence of this (the quadratic coefficients were negligible and not statistically significant). Thirdly, from the log-linear regression model in each study, we estimated the relative risk for smoking one cigarette per day and then combined these across studies in a random effects meta-analysis, fitted separately for each disease group and sex, using RevMan; we repeated this for smoking five and 20 cigarettes per day. These results (and corresponding diagrams) indicate the variability in relative risk in each smoking group across studies, but they do not directly reflect the within study correlation between risk and consumption (as in the first and second analyses above). The results are examined in relation to assuming that smoking one cigarette per day is associated with about 5% of the excess relative risk when smoking 20 cigarettes per day. Our regressions used a logarithmic scale, so smoking one cigarette per day would actually have 3.5% or 5.5% of the excess risk if the relative risk for 20 cigarettes per day was 2.0 or 3.0, respectively, values typically seen in the studies (log(relative risk for 20 cigarettes per day)=20×log(relative risk for one cigarette per day)). Patient involvement ------------------- No patients were involved in setting the research question or the outcome measures, nor were they involved in developing plans for design or implementation of the study. No patients were asked to advise on interpretation or writing up of results. There are no plans to disseminate the results of the research to study participants or the relevant patient community. We did not evaluated whether the studies included in the meta-analysis had any patient involvement. Results ======= The meta-analyses were based on 141 separately conducted cohort studies contained in 55 study reports (several involved the pooling of multiple studies),[@ref16] [@ref17] [@ref18] [@ref19] [@ref20] [@ref21] [@ref22] [@ref23] [@ref24] [@ref25] [@ref26] [@ref27] [@ref28] [@ref29] [@ref30] [@ref31] [@ref32] [@ref33] [@ref34] [@ref35] [@ref36] [@ref37] [@ref38] [@ref39] [@ref40] [@ref41] [@ref42] [@ref43] [@ref44] [@ref45] [@ref46] [@ref47] [@ref48] [@ref49] [@ref50] [@ref51] [@ref52] [@ref53] [@ref54] [@ref55] [@ref56] [@ref57] [@ref58] [@ref59] [@ref60] [@ref61] [@ref62] [@ref63] [@ref64] [@ref65] [@ref66] [@ref67] [@ref68] [@ref69] [@ref70] [@ref71] and two other study reports are referred to later on.[@ref72] [@ref73] [Table 1](#tbl1){ref-type="table"} shows all summary results. ###### Relative risk of cardiovascular disease for smoking one, five, or 20 cigarettes per day (CPD): summary results from meta-analyses Cohort No of study reports Approximate No of participants Approximate No of events Pooled relative risk (95% CI) for smoking (compared with never smokers)[\*](#t1n1){ref-type="table-fn"} Excess relative risk, as % of that for 20 CPD[†](#t1n2){ref-type="table-fn"} ---------------------------------------------------------------------------------------- --------------------- -------------------------------- -------------------------- --------------------------------------------------------------------------------------------------------- ------------------------------------------------------------ ------------------------------------------------------------------------------ ---- ------------------------------------------------------------------------------- ------------------------------------------------------------------------------- **Coronary heart disease** Men 26 2.31 million 57 152 1.48 (1.30 to 1.69); (1.45)[‡](#t1n3){ref-type="table-fn"} 1.58 (1.39 to 1.80); (1.56)[‡](#t1n3){ref-type="table-fn"} 2.04 (1.86 to 2.24); (2.06)[‡](#t1n3){ref-type="table-fn"} 46; (46)[\*](#t1n1){ref-type="table-fn"}; (42)[‡](#t1n3){ref-type="table-fn"} 57; (56)[\*](#t1n1){ref-type="table-fn"}; (53)[‡](#t1n3){ref-type="table-fn"} Women 18 2.34 million 29 870 1.57 (1.29 to 1.91); (1.59)[‡](#t1n3){ref-type="table-fn"} 1.76 (1.46 to 2.13); (1.79)[‡](#t1n3){ref-type="table-fn"} 2.84 (2.21 to 3.64); (2.81)[‡](#t1n3){ref-type="table-fn"} 31; (31)[\*](#t1n1){ref-type="table-fn"}; (33)[‡](#t1n3){ref-type="table-fn"} 43; (41)[\*](#t1n1){ref-type="table-fn"}; (44) Combined 5 1.01 million 15 153 1.65 (1.53 to 1.78); (1.67)[‡](#t1n3){ref-type="table-fn"} 1.72 (1.62 to 1.83); (1.81)[‡](#t1n3){ref-type="table-fn"} 2.34 (1.96 to 2.79); (2.44)[‡](#t1n3){ref-type="table-fn"} 53; (49)[\*](#t1n1){ref-type="table-fn"}; (47)[‡](#t1n3){ref-type="table-fn"} 61; (54)[\*](#t1n1){ref-type="table-fn"}; (56)[‡](#t1n3){ref-type="table-fn"} Men aged:  45 years 8 938 000 27 697 1.65 (1.26 to 2.16) 1.81 (1.40 to 2.33) 2.72 (2.16 to 3.43) 35 46  55 years 8 1.41 (1.17 to 1.70) 1.51 (1.27 to 1.80) 2.03 (1.74 to 2.36) 33 44  65 years 8 1.17 (0.96 to 1.43) 1.24 (1.03 to 1.48) 1.49 (1.28 to 1.74) 20 36 Women aged:  45 years 3 555 000 14 665 1.26 (0.98 to 1.62) 1.34 (0.92 to 1.96) 2.19 (1.11 to 4.32) 11 26  55 years 3 1.21 (1.05 to 1.39) 1.26 (0.98 to 1.62) 1.77 (1.00 to 3.11) 15 28  65 years 3 1.15 (1.06 to 1.25) 1.24 (1.11 to 1.40) 1.47 (0.94 to 2.29) 36 45 **Stroke** Men 17 3.40 million 71 173 1.25 (1.13 to 1.38); (1.37)[‡](#t1n3){ref-type="table-fn"} 1.30 (1.18 to 1.43); (1.42)[‡](#t1n3){ref-type="table-fn"} 1.64 (1.48 to 1.82); (1.62)[‡](#t1n3){ref-type="table-fn"} 41; (39)[\*](#t1n1){ref-type="table-fn"}; (60)[‡](#t1n3){ref-type="table-fn"} 52; (47)[\*](#t1n1){ref-type="table-fn"}; (68)[‡](#t1n3){ref-type="table-fn"} Women 10 3.59 million 60 520 1.31 (1.13 to 1.52); (1.35)[‡](#t1n3){ref-type="table-fn"} 1.44 (1.22 to 1.70); (1.48)[‡](#t1n3){ref-type="table-fn"} 2.16 (1.69 to 2.75); (2.13)[‡](#t1n3){ref-type="table-fn"} 34; (27)[\*](#t1n1){ref-type="table-fn"}; (31)[‡](#t1n3){ref-type="table-fn"} 44; (38); (42)[‡](#t1n3){ref-type="table-fn"} Combined 2 228 000 2874 1.52 (1.10 to 2.10); (1.56)[‡](#t1n3){ref-type="table-fn"} 1.63 (1.19 to 2.21); (1.65)[‡](#t1n3){ref-type="table-fn"} 1.90 (1.54 to 2.35); (2.03)[‡](#t1n3){ref-type="table-fn"} 58; (58)[\*](#t1n1){ref-type="table-fn"}; (54)[‡](#t1n3){ref-type="table-fn"} 66; (70)[\*](#t1n1){ref-type="table-fn"}; (63)[‡](#t1n3){ref-type="table-fn"} Men aged:  45 years 2 315 000 4456 1.41 (1.03 to 1.94) 1.62 (1.26 to 2.09) 2.89 (2.31 to 3.62) 22 35  55 years 2 1.27 (1.02 to 1.57) 1.39 (1.09 to 1.75) 2.01 (1.46 to 2.76) 25 43  65 years 2 1.18 (0.90 to 1.54) 1.21 (0.89 to 1.64) 1.44 (0.96 to 2.15) 15 30 Women aged:  45 years 1 534 000 5512 1.40 (0.93 to 2.11) 1.60 (1.14 to 2.24) 2.64 (2.20 to 3.17) 24 37  55 years 1 1.25 (0.95 to 1.64) 1.41 (1.13 to 1.76) 2.22 (1.97 to 2.51) 20 34  65 years 1 1.12 (0.85 to 1.47) 1.25 (1.00 to 1.56) 1.87 (1.66 to 2.11) 14 29 **Cardiovascular disease (coronary heart disease and stroke not reported separately)** Men 7 111 000 3480 1.45 (1.00 to 2.11); (1.61)[‡](#t1n3){ref-type="table-fn"} 1.59 (1.11 to 2.26); (1.70)[‡](#t1n3){ref-type="table-fn"} 2.19 (1.56 to 3.09); (2.10)[‡](#t1n3){ref-type="table-fn"} 20; (38)[\*](#t1n1){ref-type="table-fn"}; (55)[‡](#t1n3){ref-type="table-fn"} 34; (50)[\*](#t1n1){ref-type="table-fn"}; (64)[‡](#t1n3){ref-type="table-fn"} Women 1 153 000 2768 1.65 (1.13 to 2.40) 1.74 (1.30 to 2.34) 2.16 (1.69 to 2.76) 56; (56)[\*](#t1n1){ref-type="table-fn"} 64; (64)[\*](#t1n1){ref-type="table-fn"} Combined 4 1.00 million 36 525 1.63 (1.53 to 1.73); (1.64)[‡](#t1n3){ref-type="table-fn"} 1.71 (1.63 to 1.80); (1.75)[‡](#t1n3){ref-type="table-fn"} 2.27 (1.96 to 2.62); (2.25)[‡](#t1n3){ref-type="table-fn"} 50; (50); (51)[‡](#t1n3){ref-type="table-fn"} 60; (56)[\*](#t1n1){ref-type="table-fn"}; (60)[‡](#t1n3){ref-type="table-fn"} From combining relative risk for one CPD across all studies (and again, separately, for five and 20 CPD). Although they do not reflect within study correlations, in most cases they are close to those obtained from [fig 2](#f2){ref-type="fig"} and also meta-regressions (both of which are based on within study analyses). From within study analyses ([fig 2](#f2){ref-type="fig"}); they represent median values across studies. Estimates obtained from single meta-regression model across all studies (for men and women separately and for each disorder). Coronary heart disease ---------------------- The pooled relative risk from 26 study reports was 1.48 (95% confidence interval 1.30 to 1.69) for men who smoked, on average, one cigarette per day and 1.58 (1.39 to 1.80) for those who smoked five cigarettes per day; the relative risk for smoking 20 cigarettes per day was 2.04 (1.86 to 2.24) ([fig 1](#f1){ref-type="fig"}; supplementary figure B). (Excluding three studies that might have included former smokers in the reference group increased the relative risks for one and 20 cigarettes per day to 1.53 and 2.09, as expected.) [Figure 2](#f2){ref-type="fig"} shows the distribution of the excess relative risks; most had values of at least 25%. Using within study comparisons, smoking one cigarette per day had 46% (interquartile range 24-56%) of the excess relative risk for that when smoking 20 cigarettes per day, and the corresponding estimate for five cigarettes per day was 57% (36-64%). {#f1} ![Distribution of excess relative risk for smoking one or five cigarettes per day, each in relation to smoking 20 per day, using within study results (horizontal dashes show median). For example, in Lawlor et al (2008),[@ref49] estimated relative risk for coronary heart disease (CHD) was 1.83 or 2.63 for those smoking one or 20 per day, respectively (from regression analysis of this study). Proportion of excess relative risk associated with one cigarette per day is therefore 51%: (1.83−1)/(2.63−1), which is plotted. (A negative value is when relative risk for one (or five) per day is \<1.0.) For CHD in men, one study (Wen et al 2004)[@ref67] reported decreasing relative risks for increasing consumption for ≥65 age group, which appears as excess relative risk percentage of \>100% (for completeness these are kept in, but do not affect median value)](haca040906.f2){#f2} The 18 reports of women showed that one cigarette per day had 31% (interquartile range 2-46%) of the excess risk of 20 cigarettes per day (pooled relative risks 1.57 *v* 2.84), and smoking five cigarettes per day had 43% (14-55%) the excess risk (relative risk 1.76) ([fig 3](#f3){ref-type="fig"}; supplementary figure C. (Excluding one study that might have included former smokers in the reference group increased the relative risks for one and 20 cigarettes per day to 1.63 and 2.87.) {#f3} All of these estimates were similar to those obtained from the meta-regression (using a single model across studies) ([table 1](#tbl1){ref-type="table"}). Also, the relative risk estimates for one, five, and 20 cigarettes per day were mostly similar when produced by pooling these separately across studies (not within study analysis) to those from the meta-regressions (within study analysis). There was a suggestion that the relative risks at low consumption might be higher for women than for men (1.57 *v* 1.48 for one cigarette per day; 1.76 *v* 1.58 for five cigarettes per day), consistent with a higher risk of coronary heart disease in women reported by others.[@ref74] A comparison between sexes could also be examined directly within the same study cohort, where a higher relative risk was seen, without modelling: Hirayama et al (relative risk 1.61 for women versus 1.50 for men, for smoking one to four cigarettes per day),[@ref30] Nilsson et al (1.47 *v* 1.24, for smoking one to seven cigarettes per day),[@ref54] Prescott et al (2.14 *v* 1.03, for smoking three to five cigarettes per day),[@ref73] and Bjartveit et al (2.94 *v* 2.74, for smoking one to four cigarettes per day).[@ref17] Supplementary figure D shows the forest plots for the age and sex adjusted relative risks in five studies for which results were not reported separately by sex: consuming one or five cigarettes per day had 53% or 61% of the excess risk, compared with 20 cigarettes per day ([table 1](#tbl1){ref-type="table"}). Supplementary figures E and F are the forest plots for coronary heart disease and smoking consumption in men and women separately for people aged 45, 55, and 65 years. The individual relative risks among men reflect the decreasing strength of association between coronary heart disease and smoking as people get older. The excess risk for smoking one cigarette per day expressed as a percentage of that for 20 cigarettes per day remained high throughout ([fig 2](#f2){ref-type="fig"}): 35%, 33%, and 20% for a man aged 45, 55, and 65 years, respectively; the corresponding figures for women were 11%, 15%, and 36% (in which the older age group seems to have a larger estimate, but there were only three studies here). [Table 1](#tbl1){ref-type="table"} shows the results for five cigarettes per day. All estimates (men, women, and both together) are much higher than the expected 5% had a linear or log-linear relation existed between consumption and risk. Stroke ------ [Figure 4](#f4){ref-type="fig"} and supplementary figures G and H show the relative risks for stroke. Among men who smoked one cigarette per day, the relative risk was 1.25 (1.13 to 1.38); for women, it was 1.31 (1.13 to 1.52). The corresponding estimates for smoking 20 cigarettes per day were 1.64 (1.48 to 1.82) and 2.16 (1.69 to 2.75). These are again consistent with a slightly larger effect of smoking in women at the lowest smoking levels but more so at higher consumption, compared with men (1.44 *v* 1.30 for five cigarettes per day; 2.16 *v* 1.64 for 20 cigarettes per day), as seen elsewhere.[@ref74] {#f4} From the within study analyses ([fig 2](#f2){ref-type="fig"}), the distribution of excess relative risks again showed that most exceeded 25%. Smoking one cigarette per day had an estimated 41% (interquartile range −7-62%) of the excess relative risk of men who smoked 20 cigarettes per day (from 17 studies), and the corresponding figure for five cigarettes per day was 52% (9-70%). These were similar to the findings in women (10 studies), in whom one cigarette per day had 34% (3-51%) of the excess risk of 20 cigarettes per day and five cigarettes per day had 44% (16-60%). Supplementary figure I shows the forest plots for the age and sex adjusted relative risks. Supplementary figure J shows the forest plots for stroke and cigarette consumption in men according to age. The excess risk for smoking one cigarette per day expressed as a percentage of that for 20 cigarettes per day was 22%, 25%, and 15% for a man aged 45, 55, and 65 years (two studies); the corresponding figures for women were 24%, 20%, and 14% (although these were based on only one study). As with coronary heart disease, all estimates for stroke (men, women, and both together) were much higher than the 5% value expected with a linear or log-linear relation. All cardiovascular disease -------------------------- Supplementary figures K and L are forest plots for cardiovascular disease (coronary heart disease and stroke reported together), showing adjusted relative risks in men or women. Again, results were consistent with those seen for each disorder separately. Heterogeneity and bias ---------------------- The heterogeneity seen in some meta-analyses is largely due to statistically significant relative risk estimates that differ from each other, and several reasons for this may exist (for example, with or without adjustment for multiple confounders). In [figure 1](#f1){ref-type="fig"}, 15 estimates for one cigarette per day were each statistically significant, ranging between 1.19 and 2.48. However, even the lowest relative risk of 1.19 is a significant increase in risk of coronary heart disease (representing 25% of the excess risk compared with its corresponding estimate for 20 cigarettes per day: relative risk=1.77). We explored the possibility that some heavy smokers reduced to light smoking during the course of the study, which in turn might substantially reduce the relative risks in the high consumption categories, moving them closer to that for light smokers, when using baseline consumption to produce relative risks. This could overestimate the excess relative risk for one to five cigarettes per day when compared with that for 20 cigarettes per day. Such changes in smoking habits are expected to have largely occurred in the later years, so we examined only studies that had follow-up to 1995, to see whether the relative risks were much higher than those based on all studies. This was not the case. The pooled relative risks for coronary heart disease associated with smoking 20 cigarettes per day were 1.8 (1.6 to 2.0) for men and 2.5 (2.0 to 3.1) for women, a modest reduction compared with 2.0 and 2.8 from all studies in [table 1](#tbl1){ref-type="table"}. Also, we found no evidence of a negative trend between size of relative risk for smoking 20 cigarettes per day and last calendar year of follow-up (which might suggest many heavy smokers cutting down, and whether this increases over time): Spearman's correlations were positive: 0.30 (P=0.15) for men and 0.33 (P=0.20) for women (coronary heart disease studies). Three large studies (from different countries: Denmark, Norway, and South Korea) specifically examined the effect of reduced smoking on risk of cardiovascular disease. In one study (19 423 adults), only 7.2% of "heavy" smokers (at least 15 cigarettes per day) reduced their consumption by at least 50% but continued to smoke when assessed five to 10 years after baseline (verified by carbon monoxide or cotinine concentrations). There was no clear risk reduction for coronary heart disease compared with continuing heavy smokers after 14 years' follow-up (adjusted relative risk 1.06), in contrast to a relative risk of 0.67 for quitters.[@ref75] However, a large reduction in risk of lung cancer was seen in the group who reduced consumption (relative risk 0.44).[@ref76] In the second study (51 210 adults), 4.2% of heavy smokers (at least 15 cigarettes per day) reduced their consumption by at least 50% but continued to smoke when recorded three to 13 years after baseline. The adjusted relative risk for cardiovascular disease after 21 years' follow-up was 1.02 (compared with continuing heavy smokers), unlike the benefit seen in quitters (relative risk 0.46) or the positive effect on risk of lung cancer in those who reduced (relative risk 0.66).[@ref77] In the third study (475 734 adults), 5.2% of heavy smokers (at least 20 cigarettes per day) reduced to less than 10 cigarettes per day two years later, with little risk reduction after nine years' follow-up (adjusted relative risk 0.85 for stroke and 0.92 for coronary heart disease, compared with continuing heavy smokers), in contrast to the beneficial effect in quitters (relative risk 0.70 for stroke and 0.43 for coronary heart disease)[@ref78] and the effect on lung cancer in those who reduced (relative risk 0.66).[@ref79] These studies indicate that a substantial bias is unlikely to be produced by heavy smokers cutting down, because only a small proportion did so, and that those who reduced consumption did not seem to have much benefit in terms of cardiovascular disease risk. Model reliability ----------------- We checked the reliability of the regression models by comparing the estimated relative risks for smoking one, five, and 20 cigarettes per day with those seen in several individual studies that reported results specifically for low consumption (one to seven cigarettes per day). Our modelled estimates were close to those observed (supplementary table B). A high excess relative risk (in comparison with 20 cigarettes per day) was seen in 17 of 20 estimates (median 57% all estimates; 49% for coronary heart disease and 62% for stroke, comparable to those from the meta-analyses). Supplementary figure M shows examples of individual studies of coronary heart disease or stroke, plotting the observed (reported) relative risks with the ones we estimated using the log-linear model; the fit was generally good. Confounding ----------- We explored the influence of confounding factors by doing meta-analyses according to whether studies made allowance for three or more factors (which in addition to age included cholesterol for studies of coronary heart disease and cholesterol or blood pressure for studies of stroke) ([table 2](#tbl2){ref-type="table"}). One study did not adjust for either cholesterol or blood pressure but made allowance for multiple other confounders, so we also included it with the "adjusted" group.[@ref56] Additional factors often included body mass index, education, history of diabetes, and physical activity (see supplementary table A). ###### Meta-analyses according to whether studies made allowance for multiple confounding factors Cohort and analysis[\*](#t2n1){ref-type="table-fn"} No of studies From pooling results for 1 and 20 CPD separately across studies From meta-regressions (uses within study analyses) ----------------------------------------------------- --------------- ----------------------------------------------------------------- --------------------- ---------------------------------------------------- -- ------ ------ --------- **Coronary heart disease** Men:  Adjusted 11 1.74 (1.50 to 2.03) 2.27 (1.90 to 2.72) 58 1.65 2.22 53 (54)  Unadjusted 15 1.36 (1.18 to 1.56) 1.89 (1.71 to 2.08) 40 1.33 1.91 36 (38) Women:  Adjusted 9 2.19 (1.84 to 2.61) 3.95 (3.34 to 4.67) 40 2.12 3.98 38 (34)  Unadjusted 9 1.26 (1.07 to 1.49) 2.11 (1.91 to 2.34) 23 1.28 2.12 25 (23) **Stroke** Men:  Adjusted 6 1.30 (1.11 to 1.53) 1.56 (1.31 to 1.86) 54 1.35 1.55 64 (62)  Unadjusted 11 1.20 (1.07 to 1.35) 1.68 (1.45 to 1.95) 29 1.26 1.68 38 (34) Women:  Adjusted 5 1.46 (1.20 to 1.78) 2.42 (1.67 to 3.52) 32 1.50 2.39 36 (33)  Unadjusted 5 1.15 (0.98 to 1.35) 1.94 (1.44 to 2.61) 16 1.14 1.91 15 (34) CPD=cigarettes per day; RR=relative risk compared with never smokers. Adjusted includes only studies that reported RRs after allowance for ≥3 multiple confounders (which includes cholesterol for coronary heart disease studies and cholesterol or blood pressure for stroke studies), plus another study that made multi-factor adjustments.[@ref60] Unadjusted includes all other studies (although all allowed for age and occasionally one more factor). Percentage excess RR for smoking 1 CPD as percentage of that for 20 CPD. Numbers in parentheses are from same type of analyses as in [fig 2](#f2){ref-type="fig"} (that is, median value from within study comparisons). Among men, 11 studies of coronary heart disease had multivariable adjusted relative risks,[@ref17] [@ref23] [@ref33] [@ref37] [@ref47] [@ref48] [@ref49] [@ref59] [@ref61] [@ref68] [@ref71] and the pooled relative risks were 1.74 and 2.27 for smoking one and 20 cigarettes per day ([table 2](#tbl2){ref-type="table"}). From the meta-regressions, one cigarette per day has 53% of the excess relative risk of 20 cigarettes per day. These adjusted relative risks were higher than those obtained from the 15 other studies that did not allow for multiple confounders: 1.36 and 1.89 for one and 20 cigarettes per day, and the excess relative risk for one cigarette per day is 36% (lower than the estimate when we used adjusted relative risks). Among women (nine studies),[@ref17] [@ref33] [@ref37] [@ref40] [@ref48] [@ref56] [@ref59] [@ref68] [@ref71] the pooled adjusted relative risks were 2.19 and 3.95 for one and 20 cigarettes per day; and one cigarette per day represents 38% of the excess relative risk for 20 cigarettes per day. The pooled relative risks for the other nine studies that did not allow for multiple confounders were 1.26 and 2.11 for one and 20 cigarettes per day, and the excess relative risk for one cigarette per day was 25% (again, lower than the estimate when we used adjusted relative risks). Among men, there were six studies of stroke,[@ref29] [@ref41] [@ref43] [@ref49] [@ref52] [@ref61] and the pooled adjusted relative risks were 1.30 and 1.56 for smoking one and 20 cigarettes per day, with one cigarette per day representing 64% of the excess relative risk for 20 cigarettes per day. In the other 11 studies that did not allow for multiple confounders, the pooled relative risks were 1.20 and 1.68 for one and 20 cigarettes per day, and one cigarette per day had 38% of the excess relative risk for 20 cigarettes per day. Among women (five studies),[@ref29] [@ref39] [@ref41] [@ref48] [@ref56] the relative risks for stroke were 1.46 and 2.42 for one and 20 cigarettes per day, and one cigarette per day had 36% of the excess relative risk for 20 cigarettes per day. In the other five studies without multiple adjustment, the relative risks were 1.15 and 1.94 (15% of the excess relative risk). All of the studies that reported results for men and women combined had relative risks adjusted for multiple confounders. Estimates of excess relative risk associated with one cigarette per day were 47% (coronary heart disease), 54% (stroke), and 51% (cardiovascular disease), from the meta-regressions in [table 1](#tbl1){ref-type="table"}. As with previous analyses, the adjusted relative risks among women for smoking one cigarette per day were higher than for men (2.19 *v* 1.74 for coronary heart disease and 1.46 *v* 1.30 for stroke) ([table 2](#tbl2){ref-type="table"}). Study quality ------------- Study quality is difficult to assess, particularly when examining old studies, because "positive" design attributes were often not reported in publications. Our aim was not to examine a new association between a risk factor and a disorder but rather to use a feature of an already established causal relation, so the question of study quality is not so relevant. However, the variability in different observational study designs is the reason why we focused only on prospective cohort studies. Nevertheless, we examined study quality with the Newcastle-Ottawa assessment scale for cohort studies,[@ref80] using the largest set (that is, the 26 studies of coronary heart disease in men). Of these, we considered 15 to be "good quality," and the pooled relative risk for smoking one cigarette per day was 1.62 (1.45 to 1.82), higher than that based on all studies (relative risk 1.48); our interest was in whether it would be substantially lower. Discussion ========== We have shown that a large proportion of the risk of coronary heart disease and stroke comes from smoking only a few cigarettes. This has important consequences for smokers who believe that light smoking carries little or no harm. Our estimates for people who smoke one or five cigarettes per day represent light smoking, given that the daily habits of such smokers typically vary between one and five cigarettes per day. We have also indicated that the relative risk for smoking either one or five cigarettes per day seemed to be higher among women than men. Smoking one cigarette per day carries around 40-50% of the excess risk for developing coronary heart disease and stroke of smoking 20 cigarettes per day, and smoking five cigarettes per day has around 55-65% of the excess risk (particularly when we focused on studies that reported relative risks adjusted for multiple confounders). The high relative risk associated with low smoking levels is seen clearly in individual cohort studies (supplementary table B). For example, in one study (42 722 people), the relative risk for coronary heart disease among men was 2.74 (one to four cigarettes per day), representing 63% of the excess relative risk for smoking 20-24 cigarettes per day (relative risk 3.75).[@ref17] This contrasts with the effects observed for lung cancer in the same study, with relative risks of 2.79 versus 31.69,[@ref17] representing 6% of the excess relative risk, consistent with a linear relation between cigarette consumption and risk---that is, 5% of the consumption associated with about 5% of the excess risk, which has also been shown in other large studies.[@ref10] [@ref56] A recent study (290 215 US adults) showed that consistent light smoking throughout a lifetime also has a large excess risk for cardiovascular disease mortality: hazard ratio 2.78 for smoking less than one cigarette per day and 1.50 for one to 10 cigarettes per day, compared with 2.77 and 3.16 for smoking 21-30 and more than 30 cigarettes per day, respectively.[@ref81] We have also confirmed that low cigarette consumption is associated with a high risk of stroke. This evidence is further supported by studies of second-hand smoke in never smokers,[@ref82] [@ref83] [@ref84] [@ref85] in the same way as for coronary heart disease.[@ref12] [@ref84] In a meta-analysis of seven studies of never-smokers,[@ref83] the relative risks for developing stroke associated with second-hand smoke, compared with unexposed never smokers, were 1.35 (95% confidence interval 1.22 to 1.50) in all participants, 1.40 (1.09 to 1.81) among men, and 1.43 (1.28 to 1.61) among women, consistent with our results for actively smoking one cigarette per day. Potential confounding is worth considering. Different studies adjusted for different factors, but always for at least age and sex (when men and women were analysed together), which are two important confounders for cardiovascular disease. However, heavy smokers tend to have more adverse cardiovascular risk factors than light smokers (such as higher body mass index and central adiposity and poorer diet).[@ref86] [@ref87] [@ref88] Therefore, light smokers should have characteristics that are more protective against cardiovascular disease, compared with heavier smokers. Adjusting for these other risk factors should attenuate differences in cardiovascular disease risk between light and heavy smokers, not dilute them, such that when these factors are allowed for the estimates of excess risk for one or five cigarettes per day, in relation to 20, should be even larger than when based on all studies together. This is what we found when focusing only on studies that had adjusted for multiple confounding factors ([table 2](#tbl2){ref-type="table"}). The relative risks for coronary heart disease and stroke in our analyses are in line with that for all current smokers reported by Thun et al 2013 using several cohort studies,[@ref63] and they also suggest that the association between smoking and these disorders has got stronger over time. For coronary heart disease, an earlier estimate of relative risk was 1.78 among men compared with 2.50 in more recent cohort studies, with similar figures for women (2.0 previously and now 2.86). However, some of this effect could be due to decreasing exposure to second-hand smoke in the reference group (never smokers) after the introduction of smoke-free legislation. If the effect is becoming stronger, the relative risk for light smokers could now be even higher than we report, with a potentially greater percentage of excess risk in relation to heavier smokers. Although we had only summary data (hence limited ability to show trends reliably), we saw some suggestion of a positive trend between the size of the relative risk for smoking one cigarette per day and the last calendar year of follow-up for each study: Spearman's correlation 0.51 (P=0.008) for men and 0.21 (P=0.42) for women when we used studies of coronary heart disease, and 0.23 (P=0.39) and 0.56 (P=0.11) among men and women for studies of stroke. Owing to the large effect of tobacco smoke at low doses, exposure to second-hand smoke in the reference group (never smokers) might lead to underestimation of the relative risk for one and 20 cigarettes per day and consequently dilute the percentage effect of one compared with 20 cigarettes per day. The extent of this depends on the degree of contamination (particularly for women who have never smoked, who might be more likely to be exposed to second-hand smoke from their husbands in earlier studies than men who never smoked) and the reliability of measuring exposure to second-hand smoke. Many of the studies started before smoke-free laws were implemented. Only one study adjusted for second-hand smoke,[@ref33] and the reported relative risks for coronary heart disease associated with one versus 20 cigarettes per day were 1.45 versus 1.82 in men and 2.03 versus 2.63 in women, in line with those from the meta-analyses. Strengths of study ------------------ Strengths of our analyses include that we combined data from 55 cohort study reports (which together contained 141 separate cohort studies), many of which were large. For example, the studies of coronary heart disease in men were together based on approximately 3.07 million participants, including more than 75 000 cases of coronary heart disease; for stroke, the total was approximately 3.53 million men, including at least 73 000 cases. Similarly, for women, the combined studies contained around 2.56 million participants, including at least 36 000 cases of coronary heart disease, with corresponding numbers of 3.78 million and 62 000 cases in studies of stroke. The meta-analyses should therefore provide sufficiently reliable estimates of relative risks associated with low and high cigarette consumption. By using only prospective cohort studies, in which smoking consumption is recorded before development of cardiovascular disease, we avoid biases associated with retrospective designs, such as case-control studies. We report results separately for three disease groups (coronary heart disease, stroke, and cardiovascular disease), each according to sex and age. We also did three types of statistical analyses. Importantly, results showed consistency between men and women, between the disease groups, and between the different forms of analysis. Limitations of study -------------------- Our analyses also had some limitations. Firstly, we did not have individual level data for study participants (many studies are old). A few datasets of cardiovascular disease and smoking are publicly available, but our aim was to be comprehensive and not restrict ourselves to having only a few studies. Furthermore, cigarette consumption is often recorded in categories (such as one to five and six to 10 cigarettes per day), not a specific number, so the ability to do regression modelling using whole numbers of cigarettes (rather than categories) is limited. Also, smokers are not expected to consume the same number of cigarettes each day, so using categories probably better reflects their intake. Having raw data would allow more sophisticated models between risk and consumption to be examined (with increased power for these analyses), compared with using a log-linear regression of summary data (based on only several smoking categories). However, our aim was to get sufficiently good approximate estimates of the excess risks in relation to the primary comparison: between the lowest and about 20 cigarettes per day group, rather than describe the whole dose-response range. As such, our estimates are supported by two sources of evidence (several individual studies and a potentially more sensitive dose-response model from a large study). The first source comprises the effects reported in individual studies (supplementary table B), showing a consistently high observed relative risk of coronary heart disease/stroke at the lowest cigarette consumption, relative to the highest consumption group, without using a fitted model,[@ref17] [@ref30] [@ref34] [@ref40] [@ref51] [@ref54] [@ref56] [@ref58] [@ref64] [@ref72] [@ref73] which are in line with our modelled estimates. The second source comprises the results from one of the largest studies (Cancer Prevention Study II),[@ref10] in which the authors fitted a non-linear model between a measure of tobacco smoke (particular matter: PM~2.5~) and the relative risk for cardiovascular disease, using their raw data. The model was: relative risk=1+(0.2685×PM~2.5~ dose^0.2730^). An inhaled PM~2.5~ dose of about 12 mg corresponds to about one cigarette per day, which produces a relative risk of 1.53 (both sexes combined), reassuringly in between our estimate of 1.48 for men and 1.57 for women (coronary heart disease) using our simpler log-linear model (and close to 1.63 for cardiovascular disease and both sexes combined). The relative risk estimate for 20 cigarettes per day from the more sophisticated model is 2.20, so one cigarette per day represents 44% of the excess relative risk ((1.53−1)/(2.20−1)), close to our estimate of 50% (cardiovascular disease both sexes; [table 1](#tbl1){ref-type="table"}). Furthermore, the Cancer Prevention Study II showed that there was no low threshold associated with a safe level of smoking in relation to cardiovascular disease risk, for which even an inhaled PM~2.5~ dose of 1 mg (one twelfth of a cigarette per day) has an expected relative risk of 1.25.[@ref10] Secondly, methods are available for estimating dose-response associations for meta-analyses that take into account that relative risk estimates across smoking categories are expected to be correlated within a study because they use the same reference group (never smokers in our case). One such method requires frequency counts in each exposure group and assumes that adjusted relative risks are similar to unadjusted ones.[@ref89] However, frequency data were not reported for many studies, and it is essential to use age adjusted relative risks because age is an important confounder for cardiovascular disease; and ideally other known confounders should also be accounted for. One main consequence of using methods such as this is that they produce wider 95% confidence intervals, which is unlikely to change our conclusions. Thirdly, we used number of cigarettes per day, which is the most commonly reported measure, including in high profile studies.[@ref56] Although duration of smoking is also important when considering risk, it is highly correlated with age, which itself is a risk factor, so separating their effects can be difficult[@ref90]; however, large studies tend to show a relation between duration and risk.[@ref90] Because light smoking seems to have dramatic effects on cardiovascular disease, shorter duration might also be associated with a higher than expected risk. This was confirmed in three cohort studies that reported duration,[@ref39] [@ref51] [@ref91] and Pope et al 2011 concluded that the steep association with cigarettes per day did not materially change when duration was allowed for in the Cancer Prevention Study II study.[@ref10] In another study,[@ref51] the relative risk for less than 10 years' smoking duration was 1.73, compared with 2.51 for 30-40 years' duration, representing 48% of the excess relative risk (and these relative risks had been adjusted for number of cigarettes smoked per day). Similarly, the relative risk for smoking one to five cigarettes per day was 1.88, representing 40% of the excess relative risk for smoking 15-20 cigarettes per day (3.20), and these relative risks had been adjusted for duration (years) of smoking. Although long duration has persistent cumulative effects, a large proportion of the risk seems to occur in the short term.[@ref92] Fourthly, some heavy smokers could misreport as light smokers at baseline (or vice versa, although few like this are expected), but if this represented a substantial proportion there would probably be non-linear associations between consumption and the risk of other disorders (for example, lung cancer), which is generally not seen in large studies.[@ref10] [@ref17] [@ref56] However, self reported smoking status has been shown to be acceptable, at least in older observational studies.[@ref93] Even if we assumed that misclassification was so extreme that it halved the excess risk for coronary heart disease for one cigarette per day (from [table 1](#tbl1){ref-type="table"}, 24% for men where relative risk=1.48 and 29% for women where relative risk=1.57), these estimates would still be substantially higher than the 5% expected if assuming a linear relation with risk. Supporting biological mechanisms -------------------------------- Substantial biological evidence shows that components of cigarette smoke lead to endothelial injury, cell dysfunction, atherosclerosis and acute thrombosis, and decreased ability of the blood to carry oxygen.[@ref85] [@ref90] Several such studies were summarised previously with regards to increased platelet aggregation and increased carotid arterial wall thickening at low cigarette consumption, and coronary heart disease and stroke may have common underlying pathways.[@ref12] [@ref85] Harmful effects at low doses are further supported by studies of second-hand smoke that show adverse actions on subclinical vascular disease and thickening of carotid artery walls.[@ref90] Barnoya and Glantz describe a wide range of potential mechanisms by using a comprehensive literature review to purport that platelet and endothelial function, arterial stiffness, atherosclerosis, oxidative stress, inflammation, heart rate variability, energy metabolism, and increased infarct size are all sensitive to second-hand smoke.[@ref94] They also noted that even brief exposure to second-hand smoke has notable adverse effects on these mechanisms, compared with that in active smokers. Three recent experimental studies focused on low consumption/exposure.[@ref95] [@ref96] [@ref97] In one study, 29 smokers each consumed a single cigarette, immediately after which they had a significant decrease in blood vessel output power and significant increase in blood vessel ageing level and remaining blood volume 25 minutes later, as markers of atherosclerosis.[@ref95] In another study, human coronary artery endothelial cells were exposed to the smoke equivalent to one cigarette, which led to activation of oxidant stress sensing transcription factor NFR2 and up-regulation of cytochrome p450, considered to have a role in the development of heart disease.[@ref96] These effects were not seen when heart cells were exposed to the vapour from one e-cigarette.[@ref96] A study exposed adult mice to low intensity tobacco smoke (two cigarettes) for one to two months and found adverse histopathological effects on brain cells.[@ref97] Indirect evidence for large harmful effects seen at low consumption also comes from studies reporting significantly reduced hospital admissions for cardiovascular disease shortly after the introduction of smoke-free legislation in various countries,[@ref98] [@ref99] [@ref100] [@ref101] [@ref102] including systematic reviews.[@ref84] [@ref103] [@ref104] One such review, based on 45 studies, showed that the risk of hospital admission was reduced by 15% for all coronary events and 16% for cerebrovascular events.[@ref105] The authors reported that the benefit remained with longer follow-up after the legislation was implemented, and greater risk reductions were seen with more comprehensive laws. Occasional smokers and reduced smoking -------------------------------------- Limited data exist on the increase in risk among occasional or non-daily smokers. A previous study found a 50% increased risk of cardiovascular disease mortality among men in Finland who smoked occasionally.[@ref106] Of those who reported smoking daily or occasionally in the Smoking Toolkit Study in England, only 2% smoked less than one cigarette per day ("very light"),[@ref107] but just over 10% smoked on a non-daily basis.[@ref108] The non-daily smokers in the Smoking Toolkit Study smoked on average 5.2 cigarettes a day,[@ref108] so their risk is probably similar to that reported in our review. In the results section, we outlined three large studies that reported little benefit on the risk of cardiovascular disease among heavy smokers who significantly reduced their consumption (unlike the large risk reduction for lung cancer), further supportive of a substantial effect of light smoking on cardiovascular disease. More evidence exists on the beliefs about health and reduced smoking (as opposed to quitting), in addition to the large US study mentioned in the introduction.[@ref9] One survey among 12-15 year old students showed that almost 60% of regular smokers believed that occasional smoking carried little or no health risks,[@ref109] and in another study 60% of e-cigarette users said that the reason for using e-cigarettes was to reduce cigarette consumption in order to reduce health risks.[@ref110] Even in a recent survey of 1602 people in France in 2014-15 (51% were former or current smokers), 34% thought that smoking up to 10 cigarettes per day carried no risk of lung cancer, and only half of respondents believed that there was no safe cigarette.[@ref111] Other surveys indicate that smokers perceive harm reduction associated with cigarettes marketed as "light" or "low tar,"[@ref112] [@ref113] [@ref114] [@ref115] even though the scientific evidence shows no benefit. Although cutting down has clear benefits, particularly for risk of cancer, the reduction in cardiovascular disease risk is not as large as smokers might expect. Policy implications and future research --------------------------------------- Individual research studies on the effects of light smoking have occasionally appeared in the media. Examples include "Even a cigarette a day is bad for your health" in the *New York Times* in December 2016 and the BBC's "Light smoking doubles sudden death risk in women" in December 2012; governmental reports have also referred to this question.[@ref90] However, our paper is the first to combine results across many studies covering both coronary heart disease and stroke, making it a valuable reference that can be used to strengthen public health campaigns (including those on smoking cessation services) and to provide a strong health incentive for smokers to quit (particularly women), rather than cut down. We also hope to raise more awareness of the subject among cardiovascular health professionals, primary care physicians, and smoking cessation specialists. Heart disease and stroke are common disorders and causes of death. In the UK, about 73 000 deaths due to coronary heart disease and 41 000 due to stroke occur each year (compared with 36 000 for lung cancer),[@ref116] and this is after the decline in mortality over time, mainly due to prevention and better treatments. However, the number of deaths is greatly over-shadowed by the number of events: more than 493 000 inpatient hospital episodes for coronary heart disease and 236 000 for stroke each year.[@ref116] This means that many more people are living with cardiovascular disease, with a major effect on their social and physical functioning, as well as time off work and use of local health services. The situation is similar in the US, with 370 000 deaths from coronary heart disease and 140 000 from stroke each year (compared with 155 000 for lung cancer), but the number of first heart attacks is 525 000 and that of first strokes is 610 000.[@ref117] [@ref118] Fifteen to 20% of all cardiovascular disease events might be attributable to smoking, representing a substantial number of people that require care and treatment, but many events are avoidable. Thun et al, using recent US cohort study data (beginning 2000-10), indicated that given the increasing relative risks for coronary heart disease over time, about two thirds of the coronary heart disease deaths that occur in smokers could be attributable to their habit.[@ref63] The impact of smoking in places like China is of major interest. Although smoking prevalence in China has decreased in recent years, the absolute number of smokers is high, with an estimated 1 million deaths (all causes) due to tobacco in 2010.[@ref119] In a nationally representative survey in 2010, only 17% of current smokers said that they intended to quit, indicating that if Chinese smokers follow similar patterns to those in Western countries, many active smokers could be more inclined to reduce consumption rather than quit completely.[@ref120] The relatively low overall smoking prevalence among all Chinese women (\<2%) might mask differences between those in rural and urban areas, as well as habits in younger women. In a 2008 survey of girls and women aged 14-24 years at high school or college, 4.2% of those in urban areas were current smokers, double the 1.9% seen in rural areas; and 38% of those surveyed in the urban locations did not believe that smoking increases the risk of cardiovascular disease (compared with 6% when asked about lung cancer).[@ref121] Quitting smoking greatly reduces the risk of cardiovascular disease, with important benefits gained soon after stopping (quicker than for cancer).[@ref53] [@ref56] [@ref85] [@ref90] [@ref122] Smokers can use nicotine containing products such as gum, patches, and electronic cigarettes. Although e-cigarettes have had much attention, they are considered by several experts to be significantly safer than cigarettes,[@ref123] [@ref124] and they are believed to be partly responsible for the decline in smoking prevalence in the UK,[@ref125] findings that are in contrast to the claim that e-cigarettes help to maintain smoking rates. Therefore, they are an important component of harm reduction that can help people to quit completely,[@ref4] [@ref85] which is necessary to significantly reduce the risk of cardiovascular disease. Although specific adverse effects of e-cigarettes on the cardiovascular system could be investigated further,[@ref126] [@ref127] such effects, if they exist, are unlikely to be as harmful as the high risk of cardiovascular disease associated with light smoking that we show here. Conclusions ----------- Smokers who cut down the number of cigarettes they use can benefit from large reductions in the risk of cancer and some benefits on cardiovascular disease. However, smoking only one to five cigarettes per day is associated with a risk of coronary heart disease and stroke that is substantially higher than many health professionals or smokers recognise (as much as half the risk of smoking 20 per day). We show clearly that no safe level of smoking exists for cardiovascular disease at which light smokers can assume that continuing to smoke does not lead to harm. Smokers need to quit completely rather than cut down if they wish to avoid most of the risk associated with heart disease and stroke, two common and major disorders caused by smoking. ### What is already known on this topic 1. Smoking increases the risk of developing coronary heart disease and stroke 2. Many smokers believe that cutting down the number of cigarettes they smoke substantially reduces their risk of developing tobacco related disorders 3. Occasional individual studies and a meta-analysis of five studies in 1997 reported that light cigarette smoking (less than five per day) is associated with a higher than expected risk of coronary heart disease ### What this study adds 1. Men who smoke about one cigarette per day have a 48% higher risk of heart disease than never smokers and a 25% higher risk of stroke (or 74% and 30%, respectively, when allowing for confounding factors) 2. The estimates are higher in women: 57% for heart disease and 31% for stroke (or 119% and 46% when allowing for multiple confounders), again compared with never smokers. 3. People who smoke about one cigarette each day have about 40-50% of the excess risk associated with smoking 20 per day (coronary heart disease and stroke) Extra material supplied by the author ###### Supplementary figures and tables Contributors: AH developed the study concept. SB and DM did the literature search and data extraction. DM did the statistical analyses, with assistance from JKM and AH. All authors were involved in drafting and finalising the manuscript. AH and DM contributed equally to the project. AH is the guarantor. Funding: This study was supported by a core grant from Cancer Research UK (C444/A15953). Competing interests: All authors have completed the ICMJE uniform disclosure form at [www.icmje.org/coi_disclosure.pdf](http://www.icmje.org/coi_disclosure.pdf) (available on request from the corresponding author) and declare: no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work. Ethical approval: Not needed. Data sharing: No additional data available. Transparency declaration: AH affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

Introduction ============ Migraine is a common neurologic disorder that has become a major public health concern worldwide.[@b1-jpr-11-763] This disorder is characterized by recurrent, moderate-to-severe headaches. Clinical manifestations typically include unilateral headache with nausea, emesis, phonophobia, and visual sensory disturbance.[@b2-jpr-11-763] The two major types of migraine are migraine with aura (MA) and migraine without aura (MO), as defined by the International Headache Society.[@b3-jpr-11-763] Migraine is a complex disorder, and involves the influence of various environmental and genetic factors.[@b4-jpr-11-763],[@b5-jpr-11-763] While scientists have obtained valuable information concerning the development and clinical treatment of migraine, the mechanisms underlying the pathogenesis of migraine remain unclear.[@b6-jpr-11-763] Dopamine has been suggested to be the second putative protagonist in headache.[@b7-jpr-11-763] Among the proposed mechanisms, the dopaminergic system is responsible for maintaining the dopamine--norepinephrine ratio.[@b8-jpr-11-763] An imbalance in this ratio makes individuals susceptible to migraine.[@b9-jpr-11-763] Dopamine receptors are believed to play specific roles in the pathogenesis of migraine.[@b10-jpr-11-763] Dopamine receptor D2 (DRD2) has been widely studied in central nervous system disorders, including schizophrenia, posttraumatic stress disorder, movement disorders, and migraine.[@b11-jpr-11-763] Both *DRD2* genotyping and monitoring of plasma drug concentrations may be useful for alleviating clinically dominant symptoms in patients.[@b12-jpr-11-763] Previous studies suggested that single-nucleotide polymorphisms (SNPs) in *DRD2* are susceptibility loci for migraine.[@b13-jpr-11-763] *DRD2* polymorphism was observed to be significantly associated with the risk of migraine in an Indian population.[@b14-jpr-11-763] *DRD2* haplotype may be a risk factor in migraine susceptibility. In Japanese subjects, Onaya et al showed that *DRD2* polymorphism significantly and independently contributed to the development of migraine from medication overuse headache.[@b15-jpr-11-763] However, other studies did not support the contribution of *DRD2* toward the genetic predisposition to migraine in a Spanish population and a North Indian population.[@b13-jpr-11-763],[@b16-jpr-11-763] The SNP rs1800497 on chromosome 11q23.2 is located in exon 8 of the ankyrin repeat domain containing one gene downstream of *DRD2*.[@b17-jpr-11-763] *DRD2* rs1800497 was observed to be associated with different conditions such as autism,[@b18-jpr-11-763] cramps,[@b19-jpr-11-763] migraine,[@b14-jpr-11-763] memory loss,[@b20-jpr-11-763] and schizophrenia.[@b21-jpr-11-763] *DRD2* rs1800497-T-carriers faced a significantly reduced risk of schizophrenia susceptibility in Asian populations.[@b22-jpr-11-763] Gluskin and Mickey suggested that *DRD2* rs1800497 probably contributes to important individual differences in human striatal function, neuropsychiatric disease risk, and pharmacologic response.[@b23-jpr-11-763] *DRD2* rs1800497 has been consistently reported to be linked with reduced striatal D2 receptor expression in both postmortem expression studies and in vivo radioligand binding studies.[@b24-jpr-11-763],[@b25-jpr-11-763] However, this link between rs1800497 and D2 receptor expression was not apparent in patients with neuropsychiatric disorders.[@b23-jpr-11-763] The D2 receptor expression in these subjects tended to be higher among rs1800497-T-carriers than among C homozygotes.[@b23-jpr-11-763] Significant associations between *DRD2* rs1800497 and migraine risk were observed in a primary cohort study (208 patients with migraine and 200 healthy control subjects).[@b14-jpr-11-763] However, no significant associations were observed in a secondary cohort (127 patients with migraine and 200 healthy control individuals) of Indian subjects.[@b14-jpr-11-763] Importantly, no studies have investigated the putative relationship between *DRD2* rs1800497 and migraine risk in Han Chinese populations. In this study, we enrolled 500 age- and sex-matched volunteers and conducted a case--control study to validate the relationship between the *DRD2* polymorphism rs1800497 and migraine risk in a Chinese Han population. Additionally, we examined the role of DRD2 in the pathogenesis of migraine by determining plasma DRD2 levels. Materials and methods ===================== Case--control subjects ---------------------- This study was approved by the Ethics Committees of the Affiliated Hospital of Ningbo University, and written informed consent was obtained from all participants. This study included 250 unrelated patients with migraine and 250 age- and sex-matched healthy control subjects. All subjects were enrolled randomly between January 2014 and October 2017 from the Affiliated Hospital of Ningbo University, Ningbo city, Zhejiang province, China. Patients were diagnosed by at least two independent pain specialists and neurologists based on their responses provided in a validated medical questionnaire prepared according to the International Headache Society.[@b26-jpr-11-763] Patients with migraine were assigned to two subgroups: MA and MO. The control subjects had no history of personal or familial migraine, and were selected randomly while undergoing a health examination in the hospital. Individuals with obesity, hypertension, diabetes, and other neurologic or psychiatric disorders were excluded from this study. Laboratory analysis ------------------- A venous blood sample was collected from each subject within 12 h after fasting in an EDTA-coated vacutainer. Biochemical indices, including levels of triglyceride (TG), total cholesterol (TC), high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, apolipoprotein A-I, and apolipoprotein B, were measured using an automated clinical chemistry analyzer (Beckman Coulter, Brea, CA, USA). Plasma was separated from the blood samples by centrifugation at 1000×*g* for 10 min. Totally, 168 sex- and age-matched individuals, including 56 patients with MA, 56 patients with MO, and 56 control individuals, were enrolled for protein expression analysis. Plasma DRD2 concentration was determined using enzyme-linked immunosorbent assay kits (R&D Systems, Minneapolis, MN, USA). Genotype analysis ----------------- Lymphocyte DNA was isolated from peripheral blood using the salting-out procedure. The TaqMan Allelic Discrimination Assay System (assay ID: C_7486676_10) was used for *DRD2* rs1800497 genotyping. All assays were optimized to a total volume of 10 μL, containing 10 ng of genomic DNA, 0.125 μL of primer-TaqMan Probe mixture, and 5 μL of TaqMan Universal PCR Master Mix 2X (Thermo Fisher Scientific, Waltham, MA, USA). Polymerase chain reaction amplification was conducted using a 7500 Real-Time PCR System (Thermo Fisher Scientific). Thermal cycling conditions included 95°C for 10 min, followed by 50 cycles at 95°C for 15 s, and 60°C for 1 min. Genotypes were determined using the algorithm and software supplied by the manufacturer (Thermo Fisher Scientific). To determine the random genotyping error rate, 5% DNA samples were examined in duplicate, showing 100% consistency in genotype. Statistical analysis -------------------- Continuous variables were expressed as the mean and SD and compared using the independent samples *t*-test. Categorical variables were compared using Pearson's chi-squared test. Quantitative data were compared using one-way analysis of variance or the Kruskal--Wallis test. The odds ratio (OR) and 95% CI were calculated by logistic regression to determine the relative risk of rs1800497 genotypes for migraine development. All statistical analyses were conducted with SPSS statistical software (version 16.0; SPSS, Inc., Chicago, IL, USA). Power analysis was performed using Power and Sample Size Calculation software (version 3.0.43).[@b27-jpr-11-763] Statistical significance was defined as a *p*-value \<0.05. Results ======= Clinical characteristics of subjects ------------------------------------ The clinical characteristics of both groups are listed in [Table 1](#t1-jpr-11-763){ref-type="table"}. The case group comprised 250 patients (125 males and 125 females) experiencing migraine with a mean age of 43.02±16.70 years, while the control group comprised 250 healthy subjects (125 males and 125 females) with a mean age of 41.98±11.23 years. Among the patients, 68 had MA (39 males and 29 females) with a mean age of 44.00±17.33 years and 182 had MO (86 males and 96 females) with a mean age of 41.04±9.70 years. No significant differences were observed in the biochemical indices (levels of TG, TC, high-density lipoprotein, low-density lipoprotein, apolipoprotein A-I, and apolipoprotein B) between the patients with migraine and controls (*p*\>0.05). Comparison of genotype and allelic distributions ------------------------------------------------ The genotype distributions of *DRD2* rs1800497 in the control subjects and patients were within the Hardy--Weinberg equilibrium model (*p*\>0.05). As shown in [Table 2](#t2-jpr-11-763){ref-type="table"}, significant differences were observed in the genotype (*c*^2^=6.37, *p*=0.041) and allele (*c*^2^=4.69, *p*=0.03; OR=1.33, 95% CI=1.03--1.72, power=58%) frequencies between the migraine and the control groups. Sex analysis indicated a positive association for rs1800497 between female patients with migraine and control subjects (genotype: *c*^2^=7.84, *p*=0.019; allele: *c*^2^=6.60, *p*=0.010; OR=1.61, 95% CI=1.12--2.30, power=73.4%). Furthermore, a significant association was observed only in female patients with MO (genotype: *c*^2^=6.88, *p*=0.032; allele: *c*^2^=5.65, *p*=0.017; OR=1.59, 95% CI=1.08--2.36, power=65.1%). However, no significant difference was observed in the male groups (*p*\>0.05). The genetic models analyzed are listed in [Table 3](#t3-jpr-11-763){ref-type="table"}, and an increased risk was observed between the case and control groups under the recessive model (CC+CT vs TT: *c*^2^=4.69, *p*=0.03; OR=1.99, 95% CI=1.06--3.76, power=57.9%). In the sex analysis, positive associations were observed in female patients with migraine under the dominant (*c*^2^=5.15, *p*=0.023; OR=1.84, 95% CI=1.08--3.14, power=61.9%) and recessive models (*c*^2^=4.72, *p*=0.029; OR=2.45, 95% CI=1.07--5.63, power=58.3%). Significant collections were likely to be present in the patients with MO under the dominant (*c*^2^=3.82, *p*=0.05; OR=1.76, 95% CI=1.00--3.11, power=49.7%) and recessive models (*c*^2^=4.85, *p*=0.027; OR=2.58, 95% CI=1.09--6.12, power=59.3%). DRD2 expression and genotype ---------------------------- The mean plasma DRD2 levels in the control individuals (mean±SD: 24.20±2.78 ng/L) were significantly lower than those in the patients with MA (30.86±3.69 ng/L, *p*\<0.0001) and MO (31.88±4.99 ng/L, *p*\<0.0001; [Figure 1](#f1-jpr-11-763){ref-type="fig"}). Additionally, there was a sex-based difference in DRD2 expression in the patients with MA (male vs female: 29.46±3.59 vs 32.27±3.27 ng/L, *p*\<0.01) and MO (male vs female: 29.18±3.50 vs 34.58±4.84 ng/L, *p*\<0.0001; [Figure 1](#f1-jpr-11-763){ref-type="fig"}). No significant difference was observed between the male (24.45±3.14 ng/L) and female (23.96±2.39 ng/L, *p*\>0.05; [Figure 1](#f1-jpr-11-763){ref-type="fig"}) control individuals. Plasma DRD2 levels for different genotypes are shown in [Figure 2](#f2-jpr-11-763){ref-type="fig"}. There were no differences in plasma DRD2 levels among three genotypes in the control group (CC vs CT vs TT: 23.41±2.40 vs 24.51±2.69 vs 25.52±3.99 ng/L, *p*\>0.05). The DRD2 concentration in patients with migraine with genotype CC (24.76±3.76 ng/L) was much lower than that in patients with genotypes CT (30.93±3.85 ng/L, *p*\<0.0001) and TT (37.06±3.95 ng/L, *p*\<0.0001). In the subsequent subgroup analysis, similar results were obtained for the patients with MA (CC vs CT vs TT: 25.09±3.84 vs 28.57±2.84 vs 33.37±1.58 ng/L, *p*\<0.0001) and MO (CC vs CT vs TT: 24.65±3.79 vs 31.65±3.86 vs 38.29±3.74 ng/L, *p*\<0.0001). Discussion ========== In this case--control study, 250 patients with migraine (68 patients with MA and 182 with MO) and 250 unrelated healthy control individuals were analyzed. Our results showed that the *DRD2* polymorphism rs1800497 was strongly related to the risk of MO in Han Chinese subjects. Furthermore, through sex- stratified comparison, rs1800497 was observed to be associated with the risk of MO in females. DRD2 was highly expressed in the patients with MA and MO, and in the patients with migraine, its expression was regulated by the *DRD2* rs1800497 genotypes. Growing evidence suggests that high lipid concentration constitutes a risk factor for the development of migraine.[@b28-jpr-11-763],[@b29-jpr-11-763] Migraine frequency was shown to be associated with dyslipidemia in women.[@b30-jpr-11-763],[@b31-jpr-11-763] Plasma TC and TG levels may influence migraine severity.[@b30-jpr-11-763] Cinzia et al suggested that lipoprotein (a) levels differed significantly between female Italian patients and female Italian control subjects.[@b32-jpr-11-763] However, Goulart et al reported that different lipid and lipoprotein subfractions were positively associated with MO in both sexes in Brazilians.[@b33-jpr-11-763] Rist et al showed that TG and TC levels were associated with the risk of MA development, although not with other forms of headache in the French elderly.[@b29-jpr-11-763] Chorazka et al observed that the levels of TC and low-density lipoprotein cholesterol may contribute to an increased risk of migraine in Polish subjects.[@b34-jpr-11-763] In this study, we observed no differences in lipid concentrations between the patients with migraine and control individuals. These conflicting findings may be related to the different dietary habits of the Chinese population. Genome-wide association studies have revealed numerous genetic variants responsible for the risk factors for migraine development.[@b35-jpr-11-763],[@b36-jpr-11-763] Dopaminergic gene variants are the most studied entities in migraine studies.[@b37-jpr-11-763] For example, the dopamine beta-hydroxylase gene polymorphism rs6271 was suggested to be a genetic factor influencing migraine susceptibility in the Turkish population.[@b38-jpr-11-763] The dopamine trans- porter gene variant *SLC6A3* rs40184 contributed to the risk of MA development in Germans.[@b13-jpr-11-763] An imbalance in dopamine receptors D2/D3 was shown to be associated with migraine attacks and ictal cutaneous allodynia.[@b39-jpr-11-763] *DRD2* polymorphisms may be the genetic determinants of detoxification outcome in patients with medication overuse headache.[@b15-jpr-11-763],[@b40-jpr-11-763] Although several studies reported that *DRD2* polymorphisms are associated with the risk of migraine, many other studies showed both positive and negative relationships with *DRD2*. Significant associations of a functional polymorphism at the dopamine beta-hydroxylase locus with migraine risk were observed in Australian and German subjects,[@b13-jpr-11-763],[@b41-jpr-11-763] although not in Spanish subjects.[@b42-jpr-11-763] The variant rs1800497 was located 9.5 kb downstream of *DRD2*. Ghosh et al observed a significant association between rs1800497 and migraine risk in a meta-analysis of an Indian population.[@b14-jpr-11-763] However, the results of the meta-analysis involved in the two independent studies were contradictory.[@b14-jpr-11-763] Our study including 500 age- and sex-matched individuals revealed that *DRD2* rs1800497 contributed to the risk of migraine in Han Chinese. This could be partly explained by the specific genetic background and dietary habits of the Chinese. The SNP rs1800497 was reported to mediate protein--protein interactions.[@b17-jpr-11-763] Previous studies suggested that rs1800497 robustly influenced the availability of striatal D2 dopamine receptor.[@b23-jpr-11-763] rs1800497 is located in the 3′ untranslated region of *DRD2*, near the promoter, which is an inhibitory domain for the binding of nuclear factor-kappa B (NF-κB)-regulated genes. There are two NF-κB binding sites in the *DRD2* promoter, and *DRD2* expression is regulated by NF-κB.[@b43-jpr-11-763] rs1800497 variants may influence NF-κB--mediated regulation of *DRD2* and upregulate *DRD2* transcription.[@b14-jpr-11-763] *DRD2* expression may be involved in the mechanism of migraine pathogenesis, and DRD2 antagonists have been proposed as antimigraine drugs.[@b44-jpr-11-763] Our study showed similar results. Plasma DRD2 levels were significantly higher in the patients with migraine than in the control subjects. Additionally, the plasma DRD2 level was regulated by different rs1800497 genotypes. Sexual dimorphism has been observed to influence the prevalence and severity of neurologic disorders.[@b45-jpr-11-763] Females show a much higher prevalence of migraine worldwide than males.[@b46-jpr-11-763] Previous studies revealed that female sex hormones are the major factors determining the risk and characteristics of migraine, accounting for the sex-based differences observed in this study; however, there was evidence supporting the underlying genetic variance as well.[@b19-jpr-11-763] Sazci et al reported a significant association of the *NNMT* rs694539 variant with female patients with migraine, but no association with male patients.[@b47-jpr-11-763] CoSkun et al suggested that the genotypes rs2229741-GG and rs726281-GG are responsible for reduced risk of migraine in female Turkish patients.[@b48-jpr-11-763] Dopamine signaling was demonstrated to be linked to the female hormones, namely, progesterone and prolactin, in the migraineurs.[@b49-jpr-11-763] Serum prolactin level was considerably high in the female migraineurs.[@b50-jpr-11-763] Our results showed that rs1800497 was associated with the risk of migraine in female patients among the three genotypes. Plasma DRD2 levels were significantly higher in female patients than in male patients. Our findings provide insight into the prediction of the risk of headache in females. Conclusion ========== Our case--control study showed that the *DRD2* polymorphism rs1800497 was significantly associated with the risk of migraine in Han Chinese females. Additionally, plasma DRD2 level was higher in patients with migraine than in control subjects. Female patients with migraine had much higher DRD2 levels than male patients with migraine. DRD2 expression may be regulated by *DRD2* rs1800497 genotypes in patients with migraine. **Author contributions** All authors contributed toward data analysis, drafting and revising the paper, and agree to be accountable for all aspects of the work. **Disclosure** The authors report no conflicts of interest in this work. {#f1-jpr-11-763} {#f2-jpr-11-763} ###### Comparison of characteristics between cases and controls Characteristics Control (250) Migraine (250) MA (68) MO (182) *p*~1~ *p*~2~ *p*~3~ ----------------------- --------------- ---------------- ------------- ------------ -------- -------- -------- Age, years, mean±SD 41.98±11.23 43.02±16.70 44.00±17.33 41.04±9.70 0.414 0.248 0.364 Male (n) 125 125 39 86 1.000 0.549 0.741 TG (mmol/L), mean±SD 1.48±0.88 1.50±1.11 1.58±0.99 1.48±1.04 0.823 0.419 1.000 TC (mmol/L), mean±SD 4.38±1.01 4.41±1.18 4.49±1.08 4.39±1.04 0.760 0.433 0.920 HDL (mmol/L), mean±SD 1.12±0.31 1.16±0.69 1.10±0.78 1.18±0.39 0.403 0.747 0.076 LDL (mmol/L), mean±SD 2.55±0.86 2.44±0.89 2.49±0.91 2.39±0.99 0.160 0.615 0.074 ApoA (g/L), mean±SD 1.09±0.21 1.07±0.22 1.09±0.52 1.06±0.72 0.299 1.000 0.533 ApoB (g/L), mean±SD 0.81±0.41 0.79±0.25 0.89±0.23 0.75±0.45 0.511 0.124 0.150 **Note:** *p*~1~, the *p*-value for control vs migraine; *p*~2~, the *p*-value for control vs MA; *p*~3~, the *p*-value for control vs MO. **Abbreviations:** ApoA, apolipoprotein A; ApoB, apolipoprotein B; HDL, high-density lipoprotein; LDL, low-density lipoprotein; MA, migraine with aura; MO, migraine without aura; TC, total cholesterol; TG, triglyceride. ###### Distribution of rs1800497 genotypes and alleles in both groups Gender Group Genotype *c*^2^ *p* (*df*=2) Allele *c*^2^ *p* (*df*=1) OR (95% CI) -------- ----------- ------------ ------------ -------------- ----------- ------------ -------------- ------------- ----------- ------------------- -- All Control 99 (0.40) 135 (0.54) 16 (0.06) 333 (0.67) 167 (0.33) Case 80 (0.32) 140 (0.56) 30 (0.12) 6.37 **0.041** 300 (0.60) 200 (0.40) 4.69 **0.03** 1.33 (1.03--1.72) MO 59 (0.32) 102 (0.56) 21 (0.12) 4.81 0.09 220 (0.60) 144 (0.40) 3.47 0.062 1.31 (0.98--1.73) MA 21 (0.31) 38 (0.56) 9 (0.13) 4.29 0.117 80 (0.59) 56 (0.41) 2.84 0.092 1.39 (0.95--2.06) Male Control 48 (0.38) 70 (0.56) 7 (0.06) 166 (0.66) 84 (0.34) Case 46 (0.37) 69 (0.55) 10 (0.08) 0.58 0.748 161 (0.64) 89 (0.36) 0.22 0.639 1.09 (0.76--1.58) MO 32 (0.37) 49 (0.57) 5 (0.06) 0.03 0.984 113 (0.66) 59 (0.34) 0.02 0.887 1.03 (0.68--1.55) MA 14 (0.36) 20 (0.51) 5 (0.13) 2.29 0.318 48 (0.62) 30 (0.39) 0.62 0.431 1.23 (0.73--2.09) Female Control 51 (0.41) 65 (0.52) 9 (0.07) 167 (0.67) 83 (0.33) Case 34 (0.27) 71 (0.57) 20 (0.16) 7.84 **0.019** 139 (0.56) 111 (0.44) 6.60 **0.01** 1.61 (1.12--2.30) MO 27 (0.28) 53 (0.55) 16 (0.17) 6.88 **0.032** 107 (0.56) 85 (0.44) 5.65 **0.017** 1.59 (1.08--2.36) MA 7 (0.24) 18 (0.62) 4 (0.14) 3.39 0.184 32 (0.55) 26 (0.45) 2.78 0.095 1.63 (0.91--2.92) **Note:** Results with *p*-values less than 0.05 are shown in bold font. The minor allele frequency of rs1800497-T was 0.19 in HapMap-CEU; 0.41 in -JPT; 0.35 in -CHB; 0.44 in -CHD. **Abbreviations:** *df*, degrees of freedom; MA, migraine with aura; MO, migraine without aura; OR, odds ratio. ###### Comparison of the dominant model and recessive model in rs1800497 between cases and controls by gender Gender Group Dominant model *c*^2^ *p* OR (95% CI) Recessive model *c*^2^ *p* OR (95% CI) -------- ----------- ---------------- ------------ ----------- ------------------- ----------------- ------------ ----------- ------------- ------------------- -- All Control 99 (0.40) 151 (0.60) 234 (0.94) 16 (0.06) Case 80 (0.32) 170 (0.68) 3.14 0.076 1.39 (0.97--2.01) 220 (0.88) 30 (0.12) 4.69 **0.03** 1.99 (1.06--3.76) MO 59 (0.32) 123 (0.68) 2.34 0.126 1.37 (0.92--2.04) 161 (0.88) 21 (0.12) 3.55 0.059 1.91 (0.96--3.77) MA 21 (0.31) 47 (0.69) 1.73 0.188 1.47 (0.83--2.60) 59 (0.87) 9 (0.13) 3.45 0.063 2.23 (0.94--5.29) Male Control 48 (0.38) 77 (0.62) 118 (0.94) 7 (0.06) Case 46 (0.37) 79 (0.63) 0.07 0.791 1.07 (6.64--1.78) 115 (0.92) 10 (0.08) 0.57 0.45 1.47 (0.54--3.98) MO 32 (0.37) 54 (0.63) 0.03 0.862 1.05 (0.59--1.85) 81 (0.94) 5 (0.06) NA NA 1.04 (0.32--3.39) MA 14 (0.36) 25 (0.64) 0.08 0.777 1.11 (0.53--2.35) 34 (0.87) 5 (0.13) NA NA 2.48 (0.74--8.31) Female Control 51 (0.41) 74 (0.59) 116 (0.92) 9 (0.08) Case 34 (0.27) 91 (0.73) 5.15 **0.023** 1.84 (1.08--3.14) 105 (0.84) 20 (0.16) 4.72 **0.029** 2.45 (1.07--5.63) MO 27 (0.28) 69 (0.62) 3.82 **0.05** 1.76 (1.00--3.11) 80 (0.83) 16 (0.17) 4.85 **0.027** 2.58 (1.09--6.12) MA 7 (0.24) 22 (0.76) 2.78 0.095 2.17 (0.86--5.45) 25 (0.86) 4 (0.14) NA NA 2.06 (0.59--7.23) **Abbreviations:** MA, migraine with aura; MO, migraine without aura; OR, odds ratio; NA, not available.

INTRODUCTION {#s1} ============ Hepatocellular carcinoma (HCC) is one of the most common tumors worldwide. Surgical resection is recognized as the most effective method for treatment of HCC. Unfortunately, only a minority of patients currently diagnosed with HCC may benefit from this radical option as most patients are diagnosed in advanced stages[@b1],[@b2]. The melanoma differentiation-associated gene-7 (*MDA-7*) is a novel tumor suppressor gene belonging to the IL-10 cytokine superfamily. It was first identified in 1995 by subtraction hybridization of a cDNA library derived from a human HO-1 melanoma cell line treated with interferon (IFN)-β and protein kinase C activator, mezerin[@b3]. Overexpression of MDA-7 has been shown to induce tumor cell apoptosis *in vitro* in a wide variety of cancer cells including melanoma, breast and colorectal cancer[@b4]-[@b6]. Furthermore, MDA-7-induced apoptosis is tumor selective with minimal cytotoxicity observed in normal human cells. Adenovirus can be used to mediate gene transfer in a wide range of cell types and high level transgene expression can be obtained at high titers[@b7]. In this study, we constructed a recombinant adenovirus expressing green fluorescent protein (GFP) and tumor suppressor MDA-7 (AdvGFP/MDA-7). We observed that AdvGFP/MDA-7 down-regulated vessel expression of vascular endothelial growth factor (VEGF) and CD34, reduced vessel density in tumors, and inhibited tumor growth in a nude mouse tumor model. MATERIALS AND METHODS {#s2} ===================== Cell line, reagent and mice {#s2a} --------------------------- Human embryonic kidney cell line (QBI-293A) was cultured in Dulbecco\'s Modified Eagle\'s Medium (DMEM; Sigma, China) supplemented with 10% fetal calf serum. Human pancreatic carcinoma cell line HepG2 was maintained in RP1640 medium. Biotin-conjugated anti-CD34, VEGF antibodies and anti-mouse IgG antibody were obtained from Santa Cruz Biotechnology (Shanghai, China). BALB/c nude mice were obtained from Shanghai Laboratory Animal Center, Chinese Academy of Sciences (Shanghai, China) and maintained according to the guidelines for animal experimentation of the Institutional Animal Care and Use Committee at Qingdao University. The study protocol was approved by the local institutional review board. Construction of recombinant adenoviral vectors {#s2b} ---------------------------------------------- A pair of primers was designed to clone the human MDA-7 gene directly from RNA of IFN-γ-treated human peripheral blood mononuclear cells by using reverse transcription-polymerase chain reaction (RT-PCR). The sense primer sequence was: 5′-ATGGATATCATGCAGGGCCAAGAATTCCACTT-3′, and the antisense primer sequence 5′-GCACTCGAGTCAGAGCTTGTAGAATTTC-3′. The cloned *MDA-7* cDNA fragment was ligated into pAd-Track-GFP vector expressing GFP to form pAdTrack-GFP/MDA-7 expressing both GFP and human MDA-7. The resultant pAdGFP/MDA-7 and pAdGFP plasmids were purified from transfected BJ5183 cells, and then linearized by *Pac* I digestion. The resulting digest was transfected into human embryonic kidney 293 cells by Lipofectamine, leading to formation of the recombinant adenoviruses AdvGFP and AdvGFP/MDA-7. The AdvGFP/MDA-7 vectors were amplified in QBI-293A cells, and purified by cesium chloride ultracentrifugation, and stored at -80°C. Transfection of HepG2 cells with AdvGFP/MDA-7 {#s2c} --------------------------------------------- Exponentially growing human HCC cell line HepG2 was seeded onto a 6-well plate at 1×10^5^ cells per well and incubated for 48 h. The cells were allowed to grow till 80% confluence and the medium was removed. The mixtures were incubated with viral solutions for 12 h. Transfections were terminated by replacing viral solutions with fresh medium. Two days after culturing, fluorescence microscopy was used to observe the expression of GFP in infected HepG2 cells. *MDA-7* gene expression in HepG2 cells {#s2d} -------------------------------------- Two weeks after transfection, total RNA was extracted from AdvGFP/MDA-7-transfected, AdvGFP-transfected and non-transfected HepG2 cells, respectively. Expression of *MDA-7* mRNA was detected by RT-PCR, and *β-actin* served as internal control. The protocol of RT-PCR was as follows: denaturation at 93°C for 2 min; 30 cycles of denaturation at 94°C for 55 sec, annealing at 55°C for 55 sec and extension at 72°C for 50 sec; final extension at 72°C for 10 min. The reaction products were electrophoresed on 15% agarose gel and photographed. The calculated data represented the expression of *MDA-7* mRNA in each group. Western blotting analysis {#s2e} ------------------------- HepG2 cells were cultured, infected and harvested. Total cell lysates were resolved by SDS-PAGE and transferred to a nitrocellulose membrane. The membrane was blocked in PBS containing 3% (W/V) nonfat dry milk at room temperature for 2 h. The membrane was incubated with primary anti-MDA-7 (1:2,000) antibody in blocking solution at room temperature for 2 h. All membranes were then washed and incubated for 1 h at room temperature, and incubated overnight with alkaline phosphatase-conjugated secondary antibody (1:10,000, Sigma, Oakville, Ontario, Canada). Detection of apoptotic cell by flow cytometry {#s2f} --------------------------------------------- HepG2 cells (5×10^5^) reaching 70% confluence were infected with AdvGFP and AdvGFP/MDA-7. Two d later, cells were harvested after digestion by 0.25% trypsin, washed with PBS, and fixed in 70% ethanol overnight. Cells were resuspended in 1 mL of PBS containing 1 mg/mL RNaseA and 0.5 mg/mL propidium iodine (Sigma, Oakville, Ontario, Canada). After 30 min of incubation, cells were analyzed by flow cytometry. *In vivo* experiment {#s2g} -------------------- Male BALB/c nude mice (10 mice each group) were inoculated subcutaneously in the flank with 1×10^7^ HepG2 cells. Two weeks later, when the tumors were 0.2-0.3 cm in size, the mice were injected intratumorally with PBS or AdvGFP or AdvGFP/MDA-7 (1×10^7^ pfu/50 µL) each day for 5 d and tumor growth was monitored daily. Tumor size was calculated using the following formula: *V* = *ab*^2^/2, where *V* represents the tumor size, *a* is the larger and *b* is the smaller of the two dimensions. For immunohistochemistry, VEGF and CD34 were measured with immunohistochemical kits. Brown staining in or around the nucleus was taken as positive immunoreactivity for VEGF and CD34. One lumen of blood vessels was assessed as a new blood capillary. Statistical analysis {#s2h} -------------------- All data were expressed as mean±SD, and all statistical analyses were performed using the statistical software SPSS 13.0. Statistically significant differences were tested by Student\'s *t*-test and one-way analysis of variance (ANOVA) as appropriate. A *P* value \< 0.05 was considered significant. RESULTS {#s3} ======= Transfection of HepG2 cells with AdvGFP/MDA-7 {#s3a} --------------------------------------------- The recombinant adenoviral vector was validated by restriction enzyme digestion and DNA sequencing. The results demonstrated that the *MDA-7* cDNA sequence was consistent with that in the GenBank, which indicated that the recombinant adenoviral vector AdvGFP/MDA-7 was constructed successfully. To determine the optimal multiplicity of infection (MOI) for a maximal transgene expression, HepG2 cells were infected with AdvGFP/MDA-7 at various MOIs and examined by fluorescence microscopy. At MOI of 10 and higher, more than 95% of the HepG2 cells transfected with AdvGFP/MDA-7 were GFP positive ([***Fig. 1***](#jbr-26-01-053-g001){ref-type="fig"}). Therefore, an MOI of 10 was selected as the optimal dose for transfection of HepG2 cells. {#jbr-26-01-053-g001} MDA-7 expression in HepG2 cells {#s3b} ------------------------------- RT-PCR analysis demonstrated that *MDA-7* mRNA was stably expressed in HepG2 cells infected with AdvGFP/MDA-7 ([***Fig. 2***](#jbr-26-01-053-g002){ref-type="fig"}), while no expression of *MDA-7* mRNA was detected in HepG2 cells infected with AdvGFP or PBS. The expression of β-actin was not observably altered with any of the transfections. To investigate the expression of *MDA-7* in AdvGFP/MDA-7-transfected HepG2 cells, we examined the expression of human *MDA-7* examined by Western blot analysis. We found that human *MDA-7* was only expressed in AdvGFP/MDA-7-transfected HepG2 cells ([***Fig. 3***](#jbr-26-01-053-g003){ref-type="fig"}), demonstrating AdvGFP/MDA-7 mediated effective infection and specific expression of the human *MDA-7* gene in HepG2 cells. {#jbr-26-01-053-g002} {#jbr-26-01-053-g003} MDA-7 expression induces apoptosis of HepG2 cells {#s3c} ------------------------------------------------- As shown in [***Fig. 4***](#jbr-26-01-053-g004){ref-type="fig"}, AdVGFP/MDA-7 transfection induced alterations in cell cycle distribution *in vitro* with a reduction in the number of cells in the S-phase and caused a G2/M-phase arrest. However, our flow cytometric studies revealed no alteration of cell cycle in HepG2 cells transfected with AdVGFP or PBS. {#jbr-26-01-053-g004} AdvGFP/MDA-7 suppresses tumor growth *in vivo* {#s3d} ---------------------------------------------- Treatment of nude mice bearing xenografts of HepG2 tumor cells infected with AdvGFP/MDA-7 induced a significant reduction in tumor growth compared with those treated with AdvGFP \[AdvGFP/MDA-7, (763.23±16.11) mm^3^ *vs* AdvGFP (1563.43±19.26) mm^3^ or PBS (1 795.62±19.70) mm^3^, *P* \< 0.05 in both\]. The results indicated that AdvGFP/MDA-7 could effectively suppress tumor growth *in vivo* ([***Fig. 5***](#jbr-26-01-053-g005){ref-type="fig"}). {#jbr-26-01-053-g005} AdvGFP/MDA-7 down-regulates the expression of VEGF and CD34 {#s3e} ----------------------------------------------------------- As a possible mechanistic explanation for AdvGFP/MDA-7-induced reduction in tumor growth, the ability of AdvGFP/MDA-7 to alter the expression of VEGF and CD34 in tumors was assessed. AdvGFP/MDA-7 treatment significantly down-regulated the expression of VEGF and CD34 in tumors infected with AdvGFP/MDA-7 compared with those infected with AdvGFP and PBS ([***Fig. 6***](#jbr-26-01-053-g006){ref-type="fig"}, *P* \< 0.05), suggesting that down-regulation of VEGF and CD34 expression was a possible mechanism whereby AdvGFP/MDA-7 inhibited tumor growth *in vivo*. In addition, the intratumoral microvascular density (MVD) decreased significantly in tumors treated with AdvGFP/MDA-7 (5.5±0.8) compared with those treated with AdvGFP (18.3±1.2) or PBS (16.5±1.5) (*P* \< 0.05 in both). {#jbr-26-01-053-g006} DISCUSSION {#s4} ========== As one of the world\'s most common malignancies, HCC is the third most common cause of cancer-related deaths, and in recent years the incidence of the disease has been increasing. Difficulty in early diagnosis of HCC results in poor prognosis. Non-surgical therapy, such as radiotherapy, chemotherapy and gene therapy, plays important roles in the treatment of patients with HCC[@b8]. However, the effect of adenovirus-mediated human MDA-7 on the disease remains unknown till now. MDA-7 was discovered in 1995, and identified by subtractive hybridization of melanoma cells following treatment with IFN-β and mezerein, which caused terminal differentiation and growth arrest. MDA-7 is a secreted protein with cytokine-like properties and belongs to the IL-10 cytokine family, which includes IL-10, IL-19, IL-20, IL-22 and IL-26. AdVMDA-7-mediated cancer gene therapy has been demonstrated to be cytotoxic *in vitro* to various tumor cells. The *in-vivo* tumor suppression effect of AdVMDA-7 has been proved in various animal tumor models. This anti-tumor effect of MDA-7 is independent of classical tumor suppressor genes, such as *p53*, *Rb* and *p16*[@b9]-[@b11]. In this study, we constructed an adenovirus vector containing both GFP and MDA-7. Our findings demonstrated that tumor cells infected with AdvGFP/MDA-7 at an MOI of 10 resulted in GFP expression in more than 98% of the cells, indicating that this construct exhibited high infectivity. Expression of human *MDA-7* was observed in AdvGFP/MDA-7-transfected HepG2 cells, demonstrating the specific expression of the human *MDA-7* gene in HepG2 cells. The decision of cells to differentiate is commonly made in the G1 phase of the cell cycle\[20\]. The regulation of cells entering from the G1 phase of the cell cycle into the S phase is particularly important, as normal cells must pass through a checkpoint in late G1 to progress to the S phase. In this study, AdvGFP/MDA-7 was demonstrated to induce apoptosis of HepG2 cells *in vitro*. Treatment of HepG2 cells with AdvGFP/MDA-7, but not AdvGFP or PBS, resulted in a significant reduction in the number of cells in the S phase concomitant with an increase in the number of cells in the G2/M phase, which was consistent with findings of a previous study[@b13]. Tumor growth relies on angiogenesis, the formation of new blood vessels, to receive an adequate supply of oxygen and nutrients[@b14]. VEGF is an important regulator in the growth of many solid tumors conferring survival advantage by inducing vascular formation via stimulation of endothelial cells[@b15]. Angiogenesis in HCC is based on the same fundamental principles of activation, proliferation and migration of endothelial cells. Overexpression of angiogenic genes such as *VEGF* and *CD34* has been shown to be associated with enhanced tumorigenicity and tumor metastatic potential[@b16],[@b17]. CD34 is a cell surface marker of progenitor cells, and is frequently used as an indicator of newly formed vessels[@b18]. In this study, we demonstrated that *in-vivo* expression of angiogenesis-associated molecules CD34 and VEGF was significantly down-regulated in tumors treated with AdvGFP/MDA-7 compared with those in AdvGFP-treated tumors. Moreover, AdvGFP/MDA-7 treatment resulted in reduced MVD within tumors. Our findings indicate that down-regulation of VEGF and CD34 expression is a possible mechanism underlying AdvGFP/MDA-7-mediated inhibition of HCC growth *in vivo*. In summary, AdvGFP/MDA-7 can replicate in HepG2 cells and induce cellular apoptosis. It also down-regulates expression of angigenic genes *VEGF* and *CD34*, resulting in reduced tumor vessel formation. In an animal tumor model, AdvGFP/MDA-7 gene therapy significantly inhibited HCC growth and prolonged the survival of tumor-bearing nude mice. It is therefore concluded that adenovirus-mediated human MDA-7 gene therapy may serve as a novel therapeutic approach for HCC. ☆This study was supported by grants from Shandong Province Postdoctoral Innovation Project Special Foundation (No. 201003048) and the Postdoctoral Science Foundation of China (No. 2011M500697). ▵These authors contributed equally to the study.

Uterine serous carcinoma (USC) represents a clinically aggressive and highly malignant endometrial cancer. Initially distinguished from other types of primary endometrial adenocarcinoma in 1982, Hendrickson et al. \[[@B1]\] performed a pathologic review of 256 stage I endometrial cancers treated at Stanford University, with 26 of the cases revealing USC. Of these cases, 40% had deep myometrial invasion compared to 12% of typical adenocarcinomas. With deep invasion, relapse rates for USC were 63% compared to only 30% for adenocarcinomas. As a whole, 50% of the USC patients recurred, with most recurrences occurring on peritoneal surfaces in the upper abdomen. The authors concluded that this type of pattern of spread was similar to ovarian cancer and should be treated as such with either upper abdominal and pelvic radiotherapy or chemotherapy \[[@B1]\]. Several retrospective studies have suggested a benefit to adjuvant chemotherapy specifically in USC, although in the absence of randomized evidence, the impact of chemotherapy remains widely unknown. In a retrospective, multi-institutional study of 142 stage I USC patients treated with platinum and taxane based adjuvant chemotherapy, Fader et al. \[[@B2]\] found five year progression free and cause specific survival benefits compared to adjuvant radiation alone or observation. Similarly, Viswanathan et al. \[[@B3]\] reported improvement in overall survival with taxane based adjuvant chemotherapy compared to those who did not receive chemotherapy in a retrospective review of 135 patients of all stages with USC. In this same review, recurrence free survival was improved with radiation therapy compared to those who did not receive radiation therapy, suggesting that both chemotherapy and radiation play a critical role in the overall management of such patients \[[@B3]\]. To further evaluate the benefits of adjuvant radiation and systemic therapy, the Albert Einstein group conducted a phase 2 trial of pelvic radiation "sandwiched" between taxane and platinum based chemotherapy. Eighty-one patients with USC were enrolled in the study, in which patients were treated with 3 cycles of paclitaxel and carboplatin, followed by pelvic external beam radiation therapy and vaginal brachytherapy, followed by 3 additional cycles of chemotherapy. For stages I--II, overall survival at 3 years was 84%, compared to 50% for stages II--IV. Authors concluded that this method was well tolerated; however, overall survival outcomes were similar to historical controls \[[@B4]\]. In this issue of Journal of Gynecologic Oncology, Mahdi et al. \[[@B5]\] report their results of their review of the Surveillance, Epidemiology, and End Results (SEER) program 2000--2009, in which the authors investigated the impact on survival of external beam radiation therapy in stages I--IV USC patients who also received adjuvant chemotherapy. The authors are to be commended for this work, in which a large dataset of 1,838 patients were analyzed with appropriate statistics in order to reach their endpoint. In an era in which it is becoming more difficult to obtain funding to conduct large phase 3 randomized trials, particularly for relatively rare entities such as USC, we often must rely more on high-quality phase 2 studies and large database reviews such as the one presented by Mahdi et al. \[[@B5]\], to guide treatment decisions. In their study, Mahdi et al. \[[@B5]\] evaluated the benefit of external beam irradiation. Patients who received brachytherapy as their only form of radiation therapy, plus chemotherapy, were excluded. The authors found that patients who received external beam radiation had a significantly better overall survival and disease specific survival than those who did not receive external beam radiation; however, on multivariate analysis controlling for age, stage, race, and extent of pelvic lymph node dissection, the survival benefit persisted only for stage III patients. The authors conclude that combined modality treatment with both chemotherapy and radiation therapy should always be considered for patients with stage III USC. Although there have been no randomized trials dedicated to patients with USC, several trials have included patients with this variant. These have suggested that the benefit of chemotherapy alone is modest. Gynecologic Oncology Group (GOG) 249 study was a phase 3 randomized trial comparing adjuvant pelvic radiation therapy versus vaginal cuff brachytherapy followed by taxane/platinum chemotherapy. This study included 601 patients with high risk, early stage endometrial cancer, in which 15% of the cohort consisted of stages I--II USC. At 24 months, preliminary data presented in 2014 reported no difference in recurrence free or overall survival between the two adjuvant treatment strategies \[[@B6]\]. These data suggest, although with short follow-up thus far, that chemotherapy was not superior to adjuvant pelvic radiation. Similarly, the Nordic Society of Gynecologic Oncology/European Organization for the Research and Treatment of Cancer (NSOG-9501/EORTC 55991) trial randomized patients to adjuvant pelvic radiation therapy plus or minus sequential chemotherapy \[[@B7]\]. This study included 141 stages I--III USC and clear cell carcinoma patients, and for this subset of patients with high risk histologies, progression free survival at 5 years was equivalent in both study arms, suggesting a lack of benefit of sequential chemotherapy to adjuvant radiation therapy. Each of the patients analyzed in this issue's SEER program analysis of 1,838 USC patients received adjuvant chemotherapy. However, patients who also received radiation therapy either with or without a brachytherapy boost, appeared to have a better outcome particularly if they had stage III disease. This is not surprising given the known local control and progression free survival benefit from adjuvant radiation. It is unclear whether radiation therapy alone, with the omission of chemotherapy, would confer the same improvement in survival. In their introduction, Mahdi et al. \[[@B5]\] report that "given the aggressive nature of this disease and its propensity for systemic spread, adjuvant chemotherapy has become the standard adjuvant therapy in patients with USC." We agree that chemotherapy is the most often utilized adjuvant treatment for patients in USC, however, data, particularly randomized data, does not support such a claim that chemotherapy is the standard of care as the most appropriate adjuvant treatment. In summary, we agree with the conclusion drawn from the authors' well written and well analyzed SEER database review of adjuvant treatment strategies for USC―this review validates the findings of previous studies and suggests that the improved local control achieved with external beam radiation may translate into improved survival for some patients. USC remains an uncommon, but highly malignant and clinically aggressive form of endometrial cancer. The adjuvant treatment of this entity remains controversial, and individual cases should be routinely discussed in a multidisciplinary setting in order to guide patient care. Limited randomized data do not confirm the superiority of chemotherapy over radiation therapy for patients with locoregionally-confined USC. Phase 2 series and retrospective reviews discussed previously, including this large SEER review by Mahdi et al. \[[@B5]\] suggest a continued benefit to external beam radiation therapy in the adjuvant treatment of USC. **Conflict of Interest:** No potential conflict of interest relevant to this article was reported.

Introduction {#s1} ============ A wealth of data has demonstrated that a variety of non-human animals are able to solve some numerical tasks [@pone.0065262-Vallortigara1], but little is known about how salient numbers are relative to other properties [@pone.0065262-Cantlon1]. Some authors have argued that non-human animals can learn to master abstract numerical competence outside of their natural environment only after undergoing extensive laboratory training, which occurs whenever the stimuli employed do not offer quantitative non-numerical information [@pone.0065262-Breukelaar1], [@pone.0065262-Davis1], [@pone.0065262-Davis2], [@pone.0065262-Davis3], [@pone.0065262-Seron1]. After being trained to respond to ordinal relationships linked to six Arabic numerical symbols, when tested, squirrel monkeys chose the larger number in all the previously experienced combinations as well as in new ones [@pone.0065262-Olthof1]. Trained to sort in ascending order stimuli representing from 1 to 4 elements, rhesus monkeys [@pone.0065262-Brannon1], hamadryas baboons, squirrel monkeys [@pone.0065262-Smith1] and brown capuchin monkeys [@pone.0065262-Judge1] were able to generalize to new numbers --from 5 to 9- and new stimuli. Monkeys trained to respond (in ascending or descending order) to pairs of numerosities (1--9) spontaneously ordered in that same direction new pairs of larger values (i.e., 10,15,20,30) [@pone.0065262-Cantlon2]. When trained to place values (6-5-4) in descending order, rhesus macaques were able to apply the descending rule to novel values (1-2-3) [@pone.0065262-Cantlon2]. Lemurs (*Lemur catta* and *Eulemur mongoz*), when trained to select between two images the higher-ranking one, could learn the ordinal relationship between seven stimuli, showing a transitive inference reasoning [@pone.0065262-McLean1]. But abstraction is not just a prerogative of primates: an African grey parrot (*Psittacus erithacus*) was, for example, able to use labels to order numbers from 1 to 8 [@pone.0065262-Pepperberg1]. Adult Clarks\' Nutcrackers could identify the 4^th^ or the 6^th^ positions in a series of 16 identical ones [@pone.0065262-Rugani1]. When all other spatial cues were being controlled, day-old domestic chicks (*Gallus gallus*) could identify an element within a series of identical ones solely on the basis of ordinal information [@pone.0065262-Rugani2], [@pone.0065262-Rugani3]. While these and other studies prove that some animals possess some abstract numerical competence, it cannot be excluded that pure numerical ability emerges only following long laboratory training. Studies which were, instead, carried out to specifically investigate spontaneous, i.e. in the absence of training, numerical discrimination have been unable to clarify if and when non-verbal subjects (non-human animals and pre-verbal infants) rely on number or on other cues. Many of those studies were performed considering quantitative variables (usually either cumulative surface area or total volume) one at a time [@pone.0065262-Boysen1], [@pone.0065262-Beran1], [@pone.0065262-Beran2]. If indeed the quantitative cues are not specifically controlled it becomes impossible to verify if the subjects are relying on quantitative cues or on the number itself [@pone.0065262-Vallortigara1], [@pone.0065262-Vallortigara2], [@pone.0065262-Backer1]. Using a preferential looking method, six-month-old infants discriminated large numerosities that differ by a ratio of 0.5 (8vs.16 or 16vs.32), but not 0.667 (8vs.12 or 16vs.24), when presented with visual arrays in which many quantitative variables were controlled [@pone.0065262-Xu1], [@pone.0065262-Xu2]. At the same age infants discriminated 8vs.16 and 4vs.8 sounds but failed in discriminating 8vs.12 and 4vs.6 sounds, providing evidence that the same ratio (0.5) limits numerosity discrimination in auditory-temporal sequences and visuo-spatial arrays [@pone.0065262-Lipton1], [@pone.0065262-Lipton2]. When monkeys (*Macaca mulatta*) observed an experimenter hiding a number or apple pieces, one at a time, in an opaque container and a different number of apple pieces in another opaque container, they approached the larger quantity when the following pairs were presented: 1vs.2, 2vs.3, 3vs.4 and 3vs.5. In order to examine the possibility that the monkeys were focusing on volume rather than on number, in one control condition the experimenters placed 3 pieces of apple in one opaque container and 1 piece of apple, equal in volume to the three pieces grouped together, in another one. The monkeys once again chose the larger number, showing that the numerical cue was considered more important than the quantitative one [@pone.0065262-Hauser1]. Horses (*Equus caballus*), likewise, selected the larger of two quantities when presented with small numerical contrasts (1vs.2 and 2vs.3) even when the total volume of the two sets was equal [@pone.0065262-Uller1]. Using heterogeneous elements in experimental paradigms seemed to be the best way to effectively control for quantitative variables and consequently to test the abstraction of numerical values. Until now heterogeneous items have been used to test abstract numerical competence in experiments characterized by long training procedures [@pone.0065262-Brannon1], [@pone.0065262-Merritt1], [@pone.0065262-Scarf1]. The study being outlined here describes experiments in which chicks were reared in an environment where food was available only behind a screen picturing a particular number (positive stimulus, i.e. 5) of heterogeneous elements (differing in colour, size and shape) and not behind another screen picturing a different number (neutral stimulus, i.e. 10) of heterogeneous elements. We were interested in investigating the chicks\' spontaneous encoding of numerical information during rearing and to evaluating their ability to discriminate between large numbers of heterogeneous items solely on the basis of numerical cues when quantitative variables (area and perimeter) were being controlled. During testing, the chicks could freely choose to approach the positive or the neutral stimulus. The former pictured the same number of elements as did the positive stimulus but differing for colour, size and shape from the rearing ones and the latter represented the same number of elements as in the neutral stimulus during the rearing period but different again for colour, size and shape. If the animals were spontaneously encoding numerical information, we expected them to move towards the stimulus associated with food both when quantitative cues were missing (due to the use of randomly sized heterogeneous stimuli) and when the cues were not the same as those used during rearing. Materials and Methods {#s2} ===================== Ethics Statement {#s2a} ---------------- The experiments complied with all applicable national and European laws concerning the use of animals in research and were approved by the Italian Ministry of Health (permit number: 5/2012 emitted on 10/1/2012). All procedures employed in the experiments included in this study were examined and approved by the Ethical Committee of the University of Padua (Comitato Etico di Ateneo per la Sperimentazione Animale -- C.E.A.S.A.) as well as by the Italian National Institute of Health (N.I.H). Experiment 1 {#s2b} ------------ The goal of the first experiment was to investigate the chicks\' ability to discriminate between large numbers of heterogeneous items (5 *vs.*10) solely on the basis of numerical cues (when quantitative variables were unavailable), or when quantitative variables (area and perimeter) are being controlled. Since we were interested in spontaneous encoding of numerical information, the chicks were exposed for about two days to the contingent presentation of food with a certain number (i.e. 5) of items and not with another (i.e. 10). A similar procedure had been used to demonstrate chicks\' spontaneous discrimination of possible and impossible objects [@pone.0065262-Regolin1] and their sensitivity to the Ebbinghaus illusion [@pone.0065262-RosaSalva1] as well as other types of numerical discrimination [@pone.0065262-Rugani4]. ### Subjects {#s2b1} "Hybro" domestic chicks (*Gallus gallus*), a local variety of the White Leghorn breed, were used. These were obtained weekly, every Monday morning when they were a few hours old, from a local commercial hatchery (Agricola Berica, Montegalda, Vicenza, Italy). On arrival, the chicks were housed individually in standard metal cages (28×32×40 cm). Chicks were housed individually as this procedure allowed to employ half of the animals and it allowed to obtain more informative data. In fact data obtained from individual chicks that have been reared in pairs are not independent. Moreover, individual testing would be distressful to pair-reared chicks. The rearing room was constantly monitored for temperature (28--31°C) and humidity (68%) and was illuminated continuously by fluorescent lamps (36 W) located 45 cm above each cage. Water, placed in transparent glass jars (5 cm in diameter, 5 cm high) in the centre of the cages, was available *ad libitum*. During the three days, rearing period (from Monday to Wednesday morning), the chicks found food behind two of four vertical plastic screens (10×14 cm) located approximately 10 cm in front of each of the cage\'s four corners. The two screens hiding food were decorated with identical pictures representing a certain number of elements (*Positive Stimulus*, *S^p^*) while the other two screens not associated with food were decorated with identical pictures of a different numerousness (*Neutral Stimulus*, *S^n^*). All of the screens were covered with static 2D images picturing a certain number of elements whose images were randomly selected from sets of patterns of different shapes (10 different), colours (10 different) and sizes (10 different, ranging between 0.5 cm and 2 cm) and printed on identical white rectangular plastic boards (screens) (11.5×9 cm). During the rearing period four screens were always present in each cage: two representing *S^p^* and two representing *S^n^* ([Fig. 1](#pone-0065262-g001){ref-type="fig"}). To prevent the chicks from learning to identify the stimuli on the basis of the spatial disposition of the elements depicted on the screens, six different pairs of stimuli were used. In each, the elements\' disposition on the screens was randomly determined in such a way that the distance between elements varied from 0.3 to 3.8 cm. Three times a day the stimuli were replaced, in such a way that each chick was exposed, for about 8 hours, to each pair of stimuli. Every time the stimuli were replaced, the screens were also rotated from corner to corner in order to avoid positional learning. An artificial imprinting object (a red capsule measuring 2×3 cm) was suspended (at chick\'s height) in each rearing cage to prevent social isolation. Artificial imprinting objects are effective social substitutes of real social mates. After about one-two hours of exposure the chick responds to the artificial object with a range of behavioural responses which are clearly identifiable as social-affiliative [@pone.0065262-Bolhuis1], [@pone.0065262-Bateson1], [@pone.0065262-Regolin2], [@pone.0065262-Regolin3], [@pone.0065262-Fontanari1]. {#pone-0065262-g001} On the morning of the third day (testing day) each chick underwent a single test to verify how the rearing period had affected their numerical discrimination. Numerosities used during testing were the same as those utilized during the rearing period, but all the elements appearing on the screens were new and presented in different spatial dispositions. Three *Testing Conditions*, depending on the dimensions of the stimuli used, were employed. In the Random Size Group (RSG, N = 20, with group-5E: N = 10 and group-10E: N = 10), the dimensions of the elements were randomly presented; in the Perimeter Control Group (PCG, N = 30, with group-5E: N = 15, and group-10E: N = 15) the overall perimeter of the two stimuli was equated; and in the Area Control Group (ACG, N = 40 with group-5E: N = 20 and group-10E: N = 20) the overall areas of the two stimuli were equated. ### The Apparatus and Procedure {#s2b2} Testing took place in an experimental room, adjacent to the rearing room, in which temperature and humidity were controlled (25°C and 70%, respectively) and which was kept dark except for light shining from two lamps (40W) placed at a height of 25 cm at either end of the apparatus. The apparatus ([Fig. 2](#pone-0065262-g002){ref-type="fig"}) consisted of a runway (45 cm long, 20 cm wide and 30 cm high). One of the two stimuli was placed at the far end of each side of the apparatus, at a height of 2 cm, so that it was entirely visible to a chick placed in the central area of the apparatus. The positions of the two testing stimuli and the bird\'s starting position (i.e. with the *S^p^* either to the left or to the right) were balanced across the experiments. The apparatus was made up of three compartments (each 15 cm long): a central, starting area considered a no-choice zone and two side compartments (choice zones). {#pone-0065262-g002} On testing day, each chick was individually placed at the starting area with the two (positive and neutral) stimuli positioned at the end of the two arms of the apparatus. Choosing one of the two compartments (indicative of its preference for that stimulus) meant that at least ¾ of the chick\'s body entered the area as the subject was looking at the screen. No choice was instead scored whenever chicks entered a choice zone but looked at the opposite screen [@pone.0065262-Rugani1], [@pone.0065262-Rugani4]. At starting time each chick was placed in the starting position and its behaviour was videorecorded throughout the duration of the experiment, i.e. six minutes. Placed above the apparatus and connected to a monitor, a video camera enabled the experimenter to track the chicks\' behaviour during the test without being seen using a computer operated device. This was activated every time a chick entered a choice zone, and registered the amount of time each chick spent near to either stimulus. An index of choice was calculated for every chick according to the formula used to analyse choice behaviour [@pone.0065262-Andrew1]: The times spent approaching the *S^p^*/(Time spent near *S^n^* + Time spent near *S^p^*) were calculated. Values at about 0.5 indicated no preference for either stimulus; values \>0.5 indicated a preference for *S^p^* and values \<0.5 indicated a preference for *S^n^*. Significant differences with respect to chance level (0.5) were calculated by one-sample two tailed t-tests. ### Results and Discussion {#s2b3} An ANOVA was used to analyze *S^p^* (5E or 10E) and the *Testing Condition* (RSG, PCG or ACG) as well as the independent variables. The dependent variable was the *Index of Choice* for *S^p^*. The ANOVA did not uncover a significant main effect for *S^p^* (F(1,84) = 0.192, p = 0.763; group-5E: N = 45; Mean = 0.657, SEM  = 0.031; group-10E: N = 45; Mean  = 0.643, SEM  = 0.023) or for *Testing Condition* (F(2,84) = 0.092; p = 0.763: RSG: N = 20; Mean  = 0.642, SEM  = 0.036; PCG: N = 30; Mean  = 0.673, SEM  = 0.029; ACG: N = 40; Mean  = 0.646, SEM  = 0.028). As the interaction (*Testing Condition* × *S^p^*) was not significant (F(2,84) = 0.270, p = 0.764), the data were merged and the resulting mean (N = 90; Mean  = 0.654, SEM  = 0.031) was found to be significantly above chance level (one-sample t-test, t(89) = 4.968; p\<0.001), see [Fig. 3](#pone-0065262-g003){ref-type="fig"}. {#pone-0065262-g003} These results demonstrate that 3-day-old chicks spontaneously discriminate between the numbers of heterogeneous elements even when quantitative variables (area and perimeter) are being controlled. This behaviour seems to be explained by Analogue Magnitude System (AMS) processing, a non-verbal numerical system according to which encoding of numerosities is only approximate [@pone.0065262-Gallistel1]. Depending on the ratio between numbers to be discriminated and in accordance with Weber\'s law, as the ratio between the numerosities to be discriminated becomes larger, the response times decrease and accuracy increases [@pone.0065262-Gallistel1]. Experiment 2 {#s2c} ------------ Experiment 1 demonstrated that numerical discrimination of large (5vs.10) sets of heterogeneous elements is possible solely on a numerical basis. In experiment 2 we used a comparison (i.e. 6vs.9) with a 0.67 ratio which is more difficult to discriminate than the previous one (0.50). ### Subjects, Apparatus and Procedure {#s2c1} A new group of 40 chicks was used. All the chicks were tested using the same numerical comparison: 6vs.9. The rearing conditions were the same as those described above. For one subgroup of chicks (group-6E: N = 20) *S^p^* pictured 6 elements (6E) and for the second (group-9E: N = 20) *S^p^* pictured 9 elements (9E). Six pairs of rearing stimuli differing from one another with regard to the spatial disposition of the elements pictured were used. The testing stimuli were composed of heterogeneous elements positioned in a different way with respect to the rearing situation. ### Results and Discussion {#s2c2} A t-test comparing the *Index of Choice* registered by the two groups did not reveal any significant difference (t(38) = 1.022; p = 0.313; group-6E: N = 20; Mean  = 0.463, SEM  = 0.034; group-9E: N = 20; Mean  = 0.508, SEM  = 0.028). The data from the two groups were therefore merged, and the resulting mean (N = 40; Mean  = 0.486, SEM  = 0.022) did not result different from chance level (one-sample t-test, t(39) = 0.636; p = 0.528), see [Fig. 3](#pone-0065262-g003){ref-type="fig"}. The first two experiments showed that the chicks were unable to discriminate between two sets of elements characterized by a 0.67 (6vs.9) ratio while they were able to distinguish between two sets characterized by a 0.50 (5vs.10) ratio. Experiment 3 {#s2d} ------------ The aim of this experiment was to evaluate if there is an absolute upper limit in numerical discrimination. As the 0.50 ratio was discriminated when the 5vs.10 comparison was employed, we arbitrarily decided to duplicate each set: the comparison tested here was therefore 10vs.20. ### Subjects, Apparatus and Procedure {#s2d1} A new group of 66 chicks was used. The rearing conditions were the same as in the previous experiments. All the chicks were tested using the same numerical comparison: 10vs.20. Rearing stimuli consisted of six pairs each of 10 or 20 heterogeneous elements. The testing stimuli consisted once again of heterogeneous elements, but with different spatial dispositions, colours, shapes and sizes with respect to the rearing stimuli and which varied in the three *Testing Conditions*. In the Random Size Group (RSG, N = 29, with group-10E: N = 14 and group-20E: N = 15) the dimensions of the elements were randomly selected in both the 10E and the 20E. In the Perimeter Control Group (PCG, N = 18, with group-10E: N = 9, and group-20E: N = 9) the overall perimeter of the two stimuli was equated. While in the Area Control Group (ACG, N = 19 with group-10E: N = 10 and group-20E: N = 9) the overall areas of the two stimuli were equated. ### Results and Discussion {#s2d2} An ANOVA was used to analyse the *S^p^* (10E or 20E) and *Testing Condition* (RSG, PCG or ACG) as well as the independent variables. The dependent variable was the *Index of Choice* for *S^p^*. The ANOVA did not uncover a significant main effect for *S^p^* (F(1,60) = 0.017, p = 0.900; group-10G: N = 33; Mean  = 0.639, SEM  = 0.025; group-20G: N = 33; Mean  = 0.626, SEM  = 0.021) or for *Testing Condition* (F(2,60) = 0.049; p = 0.952: RSG: N = 29; Mean  = 0.627, SEM  = 0.029; PCG: N = 18; Mean  = 0.641, SEM  = 0.025; ACG: N = 18; Mean  = 0.633, SEM = 0.027). As the interaction (*Testing Condition* × *S^p^*) was not significant (F(2,60) = 0.610, p = 0.547), the data were merged and the resulting mean (N = 66; Mean  = 0.633, SEM  = 0.026) was found to be significantly above chance level (one-sample t-test, t(65) = 5.115; p\<0.001) ([Fig. 3](#pone-0065262-g003){ref-type="fig"}). Conclusions {#s3} =========== The aim of this study was to test 3-day old chicks\' capacity to discriminate between stimuli representing different numerosities and, in particular, to assess their ability to process those stimuli entirely on the basis of numerical cues. The results demonstrate that chicks discriminate between numbers, even when quantitative variables are unavailable, if the numbers compared have a 0.50 ratio (either 5vs.10 or 10vs.20), but not when the ratio between numbers is 0.67 (i.e., 6vs.9). Performance, therefore, seems to be affected not by the overall absolute number of items (15 elements are discernible in the comparisons 5vs.10 but not in the 6vs.9), but by the ratio between the numbers to be discriminated. This suggests that processing in the cases described is carried out using the Analogue Magnitude System (AMS). Interestingly, these results differ from those outlined in precedent study [@pone.0065262-Rugani5] focusing on the same species but using a different experimental setting; in that case the chicks were able to discriminate both between 5vs.10 and 6vs.9, but only when quantitative as well as numerical cues were available. Why were the chicks investigated during the current study able to discriminate numerosities on the basis of numerical cues alone? In order to answer that question we need to go back and re-examine the 2011 study. During the experiments carried out at that time, the chicks were reared with a set of artificial social objects upon which they became imprinted. Social objects were also employed during the experiments carried out during our latest study, but the ones used (two-dimensional red squares) were all identical to one another during the rearing period while at testing each of the two sets was composed of homogeneous elements (although elements could differ in size to control for quantitative variables). At testing during the 2011 study, the chicks watched the imprinted objects disappearing one at a time (i.e. sequentially) behind screens located in different positions in the test area. After all the elements had appeared and disappeared the chicks were expected to look for the larger set associated to one of the two screens. If we want to compare the two studies, we need to examine how they differed in terms of: i. the characteristics of the stimuli used (homogeneous vs. heterogeneous sets, respectively, in the precedent and in the current study), and ii. the modality in which the chicks experienced the stimuli during the test. i. Stimuli characteristics (homogeneous vs. heterogeneous). Some investigators have suggested that using sets of objects made up of elements identical to one other (i.e., a homogenous set) favours computation of continuous variables [@pone.0065262-Clearfield1], [@pone.0065262-Clearfield2], [@pone.0065262-Feigenson1], [@pone.0065262-Feigenson2], while heterogeneity of objects within the same set favours computation of exact numbers [@pone.0065262-Feigenson3], [@pone.0065262-Feigenson4], [@pone.0065262-Feigenson5]. In accordance with these hypotheses, our precedent results indicated that when the chicks were required to discriminate between homogeneous sets of items in a 2vs.3 comparison, they processed quantity information, but when heterogeneous stimuli were presented, the chicks coded the numerousness [@pone.0065262-Rugani6]. ii. Stimuli presentation at testing. When stimuli disappear one after another, as in our sequential presentation paradigm [@pone.0065262-Merritt1], higher demands are posed on the working memory since they are no longer perceptually available at the moment of choice; a lower performance is thus expected. The modality of stimulus presentation could, according to a recently proposed hypothesis [@pone.0065262-Hyde1] and verified by different brain activation patterns, favours diversified kinds of elaboration. According to that hypothesis, the simultaneous presentation of whole sets of elements directs attention to the entire collection, thus activating AMS processing. On the contrary, presenting the elements one after another, focuses attention on each object, thus activating the Object File System (OFS) processing which concentrates on single objects at the expense of the overall set, which in any case cannot contain more than 3 elements [@pone.0065262-Rugani7]. These considerations appear relevant to the results emerging from our two studies; in fact in the present study the whole sets of elements were visibly accessible to the chicks during testing, while in the 2011 paradigm the stimuli used at testing were presented sequentially. It is important to underline that the difference in the two studies with regard to stimuli presentation was limited to the testing phase. During the rearing stage the chicks in both studies were exposed to the entire sets of stimuli, probably triggering AMS processing. There may have been interference in the cognitive systems activated during the 2011 study when sequential presentation (linked to OFS) was used at testing. It would be interesting in future studies to directly test how stimuli presentation during rearing with respect to testing stages can affect OFS or AMS processing. Some considerations can also been drawn regarding the paradigm used in the current study that made it possible to highlight the chicks\' spontaneous learning and therefore to better assess their spontaneous behaviour. After rearing, the chicks, in fact, associated some numerical patterns with food. In contrast with previous studies [@pone.0065262-Andrew1], learning did not require any explicit training (i.e. through shaping) during which an operant behavioural response (i.e. pecking at the positive stimulus) was reinforced as in conventional discrimination learning tasks. Our paradigm offers an advantage with respect to tasks requiring animals to choose/directly respond to two or more food options. In fact, animals\' choices in those cases are expected to maximize possibility of survival according to the optimal foraging theory [@pone.0065262-Krebs1]. Researchers carrying out numerical discrimination tasks [@pone.0065262-Uller2], [@pone.0065262-Call1], [@pone.0065262-Ward1], [@pone.0065262-IrieSugimoto1], [@pone.0065262-Baker1], [@pone.0065262-Garland1] consider preference/choice directed towards the larger quantities of food a strategy directed at exploiting food resources. Although this is often the case, the best choice could also depend on a variety of factors -- usually neglected -- but not necessarily related to overall amounts such as optimal size of food for catching or ingesting, optimal density, etc. Differences in spontaneous discrimination in diverse species could then be related to their foraging strategy rather than to numerical cognitive skills. Direct comparison between food quantities generally cannot guarantee simultaneous control over quantitative variables, such as volume, surface, etc. of food items. In those cases data on spontaneous foraging demonstrate spontaneous proto-numerical discrimination, but do/can not test purely numerical abilities. The paradigm used in the current study offers an advantage also with respect to conventional operant conditioning, which is known to be linked to a behavioural response between the stimulus and the reinforcement. It is well known that when an auto-shaping paradigm is used, pigeons auto-reinforced with water manifest a drinking behaviour to "previous neutral stimulus" (a key), while those auto-reinforced with food show an eating behaviour [@pone.0065262-Jenkins1]. The concept of positive behavioural reinforcement could be extended to paradigms using shaping to test numerical competences and help explain some results. For example, pigeons trained to discriminate between two numerousness of non-edible elements to obtain a food reinforcement performed better when reinforced to respond to the larger rather than to the smaller numerousness [@pone.0065262-Emmerton1]. By contrast, the performance of the chicks\' studied here did not differ when food was associated with a larger or a smaller number of objects, and this could be explained by the fact that no conditioned pecking response was required for the stimulus to be associated with food. Under these conditions the characteristics of the stimulus (numerousness of objects) are not affected by the features of the attractor (food), thus making it possible to manipulate the latter as required and to better investigate numerical cognitive abilities. The same paradigm could easily be employed using a different kind of attractor, i.e., a social (i.e. an imprinting) object in order to further reduce any possible association between food and stimulus. To conclude, this work demonstrated how an abstract recognition of large [@pone.0065262-Rugani2] numbers could be precociously available in one animal species, disproving the 'last resort' theory, according to which animals rely on numerical information solely when quantitative cues (considered more salient) are not available [@pone.0065262-Davis4]. The data outlined here support the hypothesis that animals naturally extrapolate and use numerical cues, suggesting that numerical information constitutes a crucial cue for animal survival. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: RR LR. Performed the experiments: RR. Analyzed the data: RR. Contributed reagents/materials/analysis tools: RR. Wrote the paper: RR LR GV.

1. Introduction {#sec1-molecules-23-03330} =============== Tomato (*Solanum lycopersicum*) is one of the most widespread vegetables worldwide, with more than 160 million tons produced in 2013 alone \[[@B1-molecules-23-03330]\]. Tomatoes are highly consumed, mainly as fresh fruits or processed products, due to their nutritional and beneficial properties to human health \[[@B2-molecules-23-03330],[@B3-molecules-23-03330],[@B4-molecules-23-03330]\]. These crops, however, suffer attacks of various pathogens such, as viruses, bacteria, fungi and nematodes, causing substantial production losses. Late blight, caused by the phytopathogenic oomycete *Phytophthora infestans*, is one of the most devastating tomato diseases, demanding high chemical input for disease control worldwide \[[@B5-molecules-23-03330],[@B6-molecules-23-03330],[@B7-molecules-23-03330],[@B8-molecules-23-03330]\]. On farms with conditions favorable to pathogens, such as a high humidity and temperature, late blight can cause severe epidemics and destroy the entire tomato crop production \[[@B9-molecules-23-03330]\]. The expenses due to tomato late blight and control measures are estimated to exceed \$5 billion annually worldwide \[[@B10-molecules-23-03330]\]. The development of resistant cultivars is an important alternative for pathogen management and control, but as yet there are no highly resistant cultivars. The early detection of asymptomatic infected *S. lycopersicum* with *P. infestans* is therefore fundamental to establish effective strategies for pathogen management and control. *P. infestans* exhibits a two-phases life cycle. Post infection in the host is characterized as an asymptomatic biotrophic phase, and a late necrotrophic stage is characterized by tissue degradation and disease \[[@B11-molecules-23-03330],[@B12-molecules-23-03330]\]. The identification of *P. infestans* is realized by traditional techniques that require isolation from the plant tissue followed by culture-based morphological evaluation, but this protocol requires a high level of taxonomic expertise \[[@B13-molecules-23-03330]\]. Recently, new techniques of plant pathogen diagnosis, including monoclonal antibodies, enzyme-linked immune sorbent assay (ELISA) and DNA-based techniques, have been introduced, which are far more specific, sensitive and accurate \[[@B14-molecules-23-03330],[@B15-molecules-23-03330],[@B16-molecules-23-03330],[@B17-molecules-23-03330]\] than the traditional techniques. These techniques efficiently identify the pathogens, but they fail to provide information on the pathogen-host interaction at the molecular level or the biochemical alterations caused by late blight. Mass spectrometry techniques are known to provide rapid and accurate target and untargeted analysis of metabolites, allowing the detection of pathogens in plants and detection of the metabolites of the pathogen-host interaction. MALDI-MS has been applied to evaluate metabolic alterations due to different phytopathogen infestations. MALDI-MS was able to establish metabolic interactions between rice−bacterium (*Oryza sativa* infected with *Xanthomonas oryzae*) and soybean−aphid (*Glycine max* colonized with *Aphis glycines*) \[[@B18-molecules-23-03330]\]. Additionally, MALDI-MS profiles of proteins, lipids and metabolites have also revealed plant-pathogen interactions \[[@B14-molecules-23-03330],[@B19-molecules-23-03330],[@B20-molecules-23-03330]\], specifically, the protein profile of sugarcane after infection by *Sporisorium scitamineum*, \[[@B21-molecules-23-03330]\] the identification of differential proteins of rice leaves infected with the fungus *Cochliobolus miyabeanus* \[[@B22-molecules-23-03330]\], and the proteins that may lead to the resistance of tomato plants to *P. infestans* \[[@B7-molecules-23-03330]\]. Metabolomics based on LC-MS and GC-MS approaches reveal changes in the metabolic profiles of the plants as a response to the attack of pathogens \[[@B11-molecules-23-03330],[@B23-molecules-23-03330],[@B24-molecules-23-03330],[@B25-molecules-23-03330],[@B26-molecules-23-03330],[@B27-molecules-23-03330],[@B28-molecules-23-03330],[@B29-molecules-23-03330]\]. Metabolomics has been applied to investigate the biochemical responses of tomato to the attacks of single and multiple pathogen infestations; the discrimination of different times post infection; the differentiation between non-infected and infected plants; the susceptibility/resistance to several pathogens; and even the beneficial interactions with mycorrhizal fungi \[[@B30-molecules-23-03330]\]. All of these approaches have provided a deeper insight into biological processes and supported the discovery of potential biomarkers \[[@B29-molecules-23-03330],[@B31-molecules-23-03330],[@B32-molecules-23-03330]\]. The metabolomics studies of infected tomatoes have focused on the identification of metabolic pathways and those metabolites that are up- and downregulated by infection with tomato mosaic virus (ToMV) \[[@B33-molecules-23-03330]\]; tomato yellow leaf curl virus (TYLCV) \[[@B34-molecules-23-03330]\]; potato spindle tuber viroid (PSTVd) \[[@B35-molecules-23-03330]\]; root-knot nematode (RKN) *Meloidogyne incognita* \[[@B36-molecules-23-03330]\]; *Pseudomonas syringae* and *Botrytis cinerea* \[[@B37-molecules-23-03330]\], and infestation with spider mites (*Tetranychus urticae*) and aphids (*Myzus persicae*) \[[@B38-molecules-23-03330],[@B39-molecules-23-03330]\]. There are, however, only a few metabolomics studies about the metabolic interactions between tomato cultivars and *P. infestans* that have focused on the early detection of infected plants. Here, we applied metabolomic analysis of LC-MS and MALDI-MS profiles associated with multivariate data analysis in the early detection of tomato infection by *P. infestans* and differentiated between the infection times. 2. Results and Discussion {#sec2-molecules-23-03330} ========================= 2.1. Multivariate Statistical Analysis of LC-MS Metabolomics Data {#sec2dot1-molecules-23-03330} ----------------------------------------------------------------- To investigate the comprehensive metabolic changes that occur in response to infection by *P. infestans*, tomato plants (Santa Cruz Kada cultivar) were inoculated with a sporangial solution of *P. infestans* and collected at 4, 12, 24, 36, 48, 60, 72 and 96 h post inoculation (hpi). At 96 hpi, the infected asymptomatic leaves were collected and examined under optical and stereoscopic microscope to confirm the sporangia structures of *P. infestans*. In the LC-MS untargeted metabolomics analysis, the ion chromatograms obtained in the ESI positive and negative ion modes (see [Figures S1a and S2a](#app1-molecules-23-03330){ref-type="app"}, respectively) showed no apparent differences between the infected and non-infected samples, but the Cloud Plot univariate analysis showed 5582 and 4790 features for the positive and negative ion modes, respectively, with statistical significance (ANOVA *p* ≤ 0.01, see [Figures S1b and S2b](#app1-molecules-23-03330){ref-type="app"}). In the first step of the statistical processing, the principal component analysis (PCA), a technique used for dimensionality reduction of multivariate data whilst preserving most of the variance \[[@B40-molecules-23-03330]\], was applied to LC-MS metabolic profiles data in order to find cluster of samples. The PCA score plot ([Figures S3 and S4](#app1-molecules-23-03330){ref-type="app"}) clearly shows three clusters of the samples according to the different infection times, corresponding to early asymptomatic infection (4, 12, 24 and 36 hpi), late asymptomatic infection (48, 60, 72 and 96 hpi) and the initial control. Secondly, multivariate analysis of OPLS-DA was then applied, and it revealed significant differences between the control and early- and late-infected asymptomatic plants, along with additional information regarding significant molecular features of the infection stages. In the OPLS-DA plots ([Figure 1](#molecules-23-03330-f001){ref-type="fig"}a,b), three different clusters in both the positive and negative ion modes corresponding to the initial control, early infection (4 to 36 hpi, cluster I) and late infection (48 to 96 hpi, cluster II) were delineated. Additionally, the heat map plots based on HCA showed additional information about the similarities between samples and clusters \[[@B41-molecules-23-03330]\]. The variable importance in projection (VIP) scores, which estimate the importance of each variable in the projection used in a PLS model \[[@B42-molecules-23-03330]\], show potential discriminant metabolites with high score values and high discriminator power ([Figure 1](#molecules-23-03330-f001){ref-type="fig"}c,d). The annotated metabolites isocoumarin (M301T17); the diterpene lactone (M579T17) ([Figure 1](#molecules-23-03330-f001){ref-type="fig"}c) and the triterpene saponin (M984T24) ([Figure 1](#molecules-23-03330-f001){ref-type="fig"}d) increase in intensity with the progression of the infection, whereas the intensity of the triterpenoid (M685T19); the peonidin 3-(4-sinapoylgentiobioside) (M832T22) ([Figure 1](#molecules-23-03330-f001){ref-type="fig"}c) and the sulfoquinovosyldiacylglycerol (M820T23) ([Figure 1](#molecules-23-03330-f001){ref-type="fig"}d), decrease. The level of annotation and additional information on these metabolites are summarized in [Supplementary Table S1](#app1-molecules-23-03330){ref-type="app"}. Subsequently, partial least squares-discriminant analysis (PLS-DA) that is a supervised chemometric method used to optimize the separation between different groups \[[@B40-molecules-23-03330]\], was applied to investigate specific metabolic changes in the early asymptomatic infection stage. In [Figure S7a,b](#app1-molecules-23-03330){ref-type="app"}, the differentiation of the infection times (4, 12, 24 and 36 hpi) based on metabolic profiles is now apparent. The VIP score-plots derived from PLS-DA ([Figure 2](#molecules-23-03330-f002){ref-type="fig"}a,b) can also discriminate other possible metabolites, such as the triterpene saponin (M798T23), the demissine (M529T12) ([Figure 2](#molecules-23-03330-f002){ref-type="fig"}a) and the triterpenoid saponin (M791T21) ([Figure 2](#molecules-23-03330-f002){ref-type="fig"}b), which increase at the early asymptomatic infection stage. For level of annotation see the [Table S1](#app1-molecules-23-03330){ref-type="app"}. Similarly, PLS-DA analysis was applied to investigate the specific metabolic changes at the late infection stage (48 to 96 hpi, cluster II). The differentiation of the 48, 60, 72 and 96 hpi is now also apparent. The PLS-DA plots of [Figure S8a,b](#app1-molecules-23-03330){ref-type="app"} confirm the closeness between the post inoculation time points of the cluster II. The gradient variation of the component between 48 and 96 for the LC-MS data in both positive and negative ion modes is notable. The VIP plot (VIP \> 5) of the late infection stage allows the identification of late discriminant metabolites. The polyacetylene (M277T16); the isoprene (M367T19); the carboxylic acid esters (M207T14) and the macrolide (M253T13) ([Figure 3](#molecules-23-03330-f003){ref-type="fig"}b) features show a gradual increase in relative abundance from 48 to 96 hpi. Other features are also present in the s-plot ([Figure S8c,d](#app1-molecules-23-03330){ref-type="app"}), supporting the idea that the bulk of the potential biomarkers increase as a function of infection time. These discriminant molecular features of the [Figure 2](#molecules-23-03330-f002){ref-type="fig"}a,b and [Figure 3](#molecules-23-03330-f003){ref-type="fig"}a,b are in concordance with the ions depicted in the s-plot derived from the OPLS-DA model ([Figures S7c,d and S8c,d](#app1-molecules-23-03330){ref-type="app"}), where the p(corr) axis represents the reliability of each variable and has a value of between + 1 and − 1. Variables in the extreme lower left and upper right quadrants are reliable for the extraction of putative biomarkers \[[@B28-molecules-23-03330]\]. The permutation test applied to the LC-ESI (+)-MS data from the early and late asymptomatic infection stages (4--96 hpi and control) revealed a Q2 value of 0.691 and an R2Y of 0.995 (*p* \< 5 × 10^−4^) ([Figure S9a](#app1-molecules-23-03330){ref-type="app"}). Likewise, the LC-ESI (−)-MS data from the early and late asymptomatic infection stages (4--96 hpi and control) revealed a Q2 value of 0.893 and an R2Y of 0.977 (*p* \< 0.003) ([Figure S10a](#app1-molecules-23-03330){ref-type="app"}). The permutation test applied to the multivariate LC-ESI (+)-MS data from only at the early asymptomatic infection stage (4--36 hpi) revealed a Q2 value of 0.85 and an R2Y of 0.999 (*p* \< 5 × 10^−4^), and multivariate LC-ESI (+)-MS data from only at the late asymptomatic infection stage (48--96 hpi) revealed a Q2 value of 0.741 and an R2Y of 0.998 (*p* \< 5 × 10^−4^). Similarly, the LC-ESI (−)-MS data (4--36 hpi) revealed a Q2 value of 0.943 and an R2Y of 0.997 (*p* \< 0.01), and the LC-ESI (−)-MS data (48--96 hpi) revealed a Q2 value of 0.908 and an R2Y of 0.984 (*p* \< 0.01). These results indicate that the OPLS-DA model has good fitness and prediction power. Cross validation (CV) of the PLS-DA models by leave-one-out (LOO) or 10-fold revealed five latent variables (components) for the optimal performance of the multivariate data ([Figures S9b--e and S10b--e](#app1-molecules-23-03330){ref-type="app"}). The discriminant molecular features ([Table 1](#molecules-23-03330-t001){ref-type="table"}) M845T24 (*m*/*z* 845.4816), M984T24 (*m*/*z* 983.5392) and M798T23 (*m*/*z* 797.5049) were annotated as triterpene saponins using MS and MS/MS spectra in the MS-Finder database. Similarly, the features M529T12 (*m*/*z* 528.7643) and M737T20 (*m*/*z* 737.4186) were annotated as the steroidal saponins demisine and tuberoside J, respectively. The triterpene and steroidal saponins are usually reported to have important roles in the defense response of plants against pathogens, pests and herbivores due to their antimicrobial, antifungal, antiparasitic, insecticidal and anti-feed properties \[[@B43-molecules-23-03330],[@B44-molecules-23-03330],[@B45-molecules-23-03330],[@B46-molecules-23-03330]\]. The saponins are produced by tomatoes and have been studied in detail in relation to their potential role in the defense response of plants against phytopathogenic fungi \[[@B45-molecules-23-03330],[@B47-molecules-23-03330]\]. The biosynthesis of triterpenoid saponins is induced in the roots in response to *Phytophthora cactorum* attack in roots of *Fragaria vesca* \[[@B48-molecules-23-03330]\]. Using the MS and MS/MS spectra obtained from the LC-MS analysis, the following compounds were successfully annotated: steroidal glycoalkaloids α-tomatine (M1079T12; *m*/*z* 1078.5436), dehydrotomatine (M1077T12; *m*/*z* 1076.527), hydroxytomatine isomer I (M1095T10; *m*/*z* 1094.5382), the unknown glycoalkaloids UGA 11 (M1109T12; *m*/*z* 1108.5541) and UGA 28 (M1121T12; *m*/*z* 1120.5120) and the aglycon tomatidine (M414T20; *m*/*z* 414.3385) ([Table 1](#molecules-23-03330-t001){ref-type="table"}), which occur naturally in tomatoes \[[@B49-molecules-23-03330],[@B50-molecules-23-03330],[@B51-molecules-23-03330]\]. However, some studies of glycoalkaloids have shown that these molecules have antibiotic properties against a variety of fungi \[[@B52-molecules-23-03330],[@B53-molecules-23-03330]\], suggesting that tomatine (α-tomatine and dehydrotomatine) may play a major role in disease resistance in the tomato plants \[[@B49-molecules-23-03330],[@B51-molecules-23-03330],[@B54-molecules-23-03330]\]. Tomatidine is an important biomarker of infection, because bacterial and fungal pathogens secrete various types of tomatinase enzymes that can detoxify α-tomatine by removing one or more sugar residue \[[@B55-molecules-23-03330],[@B56-molecules-23-03330]\]. Thus, it has also been suggested that products resulting from tomatinase activity play an indirect role in the virulence of pathogens against tomato plants by suppressing plant defense responses \[[@B55-molecules-23-03330]\]. However, the sensitivity to saponins might be correlated with the type of sterols present in the membranes of the potential pathogen. The oomycetes have been shown to be insensitive to saponins, probably because their membranes lack 3β-hydroxy sterols \[[@B55-molecules-23-03330],[@B57-molecules-23-03330]\], similar to plant cell membranes. In *P. infestans* there is evidence of genes encoding certain glycoside hydrolases with potential activity against glycoalkaloids, but there is no evidence that deglycosylation takes place \[[@B52-molecules-23-03330]\]. Even though the glycoalkaloids were annotated, the LC-MS data does not allow for determining whether the *P. infestans* infection of the tomato plants produces a significant effect on these metabolites. On the other hand, the discriminant features M293T15 (*m*/*z* 293.1774) and M221T16 (*m*/*z* 221.1561) are attributed to phytuberin and rishitin, respectively ([Table 1](#molecules-23-03330-t001){ref-type="table"}). These sesquiterpenoids are antimicrobial phytoalexins produced by plants in response to biotic and abiotic stress \[[@B58-molecules-23-03330],[@B59-molecules-23-03330]\]. Phytuberin and rishitin are reported to be present during the infection of potato plants with *P. infestans* \[[@B60-molecules-23-03330],[@B61-molecules-23-03330],[@B62-molecules-23-03330]\]. Furthermore, the feature M267T14 (*m*/*z* 267.1617) ([Table 1](#molecules-23-03330-t001){ref-type="table"}) was also annotated as the toxic furanosesquiterpene dihydro-7-hydroxymyoporone, which has been isolated from sweet potato (*Ipomoea batatas*) infected with *Ceratocystis fimbriata* \[[@B63-molecules-23-03330]\]. Some of the significant features annotated as flavonoids are apigenin 7-\[rhamnosyl-(1-\>2)-galacturonide\] (M610T24; *m*/*z* 610.1758; \[M + NH~3~\]^+^) and peonidin 3-(4-sinapoylgentiobioside) (M832T22; *m*/*z* 832.2387; \[M + H\]^+^). Additionally, naringenin-hexose I (M433T13; *m*/*z* 433.1115) and the catechin 7,4′-dimethyl ether (M317T13; *m*/*z* 317.1020) were annotated ([Table 1](#molecules-23-03330-t001){ref-type="table"}). The defensive role of flavonoids is less known; however, they are thought to be beneficial to the plant itself as physiologically active compounds, principally to stress protecting agents and play a significant role in plant resistance \[[@B4-molecules-23-03330],[@B64-molecules-23-03330],[@B65-molecules-23-03330]\]. Some metabolomics studies of tomatoes infected with *Botrytis cinerea* showed higher concentrations of flavonoids, such as rutin and quercetin-3-galactoside \[[@B37-molecules-23-03330]\]. Similarly, rutin, saponarin and several kaempferol and related compounds were identified in potato leaves resistant to late blight \[[@B66-molecules-23-03330]\]. Additionally, NMR-based metabolic profiling showed that rutin was the flavonoid that was most expressed in response to *Pseudomonas syringae* infection of tomato \[[@B67-molecules-23-03330]\], and higher levels of rutin are associated with late blight resistance of different cultivars of potato plants \[[@B68-molecules-23-03330]\]. Regarding the specific interaction tomato-*P. infestans*, tomato resistance to phytopathogen was associated with genes involved with reactive oxygen species (ROS) scavenging systems \[[@B69-molecules-23-03330]\]. Besides, the overexpression of the SpWRKY1 gene in tomato regulates antioxidants as flavonoids to reduce ROS accumulation and alleviate cell membrane injury after *P. infestans* infection \[[@B70-molecules-23-03330],[@B71-molecules-23-03330]\]. The other 67 metabolites that were annotated as organic acids, macrolides, alkaloids, and gibberellins are summarized in the [supplementary information, Table S1](#app1-molecules-23-03330){ref-type="app"}. The annotation data includes retention time, molecular formulas, exact masses, ion products and metabolite identification levels according to the MSI guidelines \[[@B72-molecules-23-03330],[@B73-molecules-23-03330]\]. 2.2. MALDI-MS Protocol for Analysis of Infected Tomato {#sec2dot2-molecules-23-03330} ------------------------------------------------------ Similar to the LC-MS analysis, we aimed at differentiating the early and late asymptomatic stages and the infection times of the tomato plants infected by *P. infestans* through MALDI-MS metabolic profiles. MALDI-MS is a fast and direct analysis that requires no chromatographic separation \[[@B21-molecules-23-03330],[@B75-molecules-23-03330],[@B76-molecules-23-03330]\]. To optimize the analysis protocol, seven different MALDI matrices were tested. DHB, 9-AA and *trans*-ferulic acid were tested in the positive ion mode, and DMAN, DHB, 4-NA, MBT and ATT were tested in the negative ion mode. The DHB and MBT matrices provided the best MALDI spectra, according to the number and abundance of the ions, over the mass range (600--1500 Da) where there was no matrix effect. Using the selected matrices, characteristic MALDI-MS profiles were obtained from the control, early asymptomatic infection (12 hpi) and late asymptomatic infection stages (96 hpi). [Figure S11](#app1-molecules-23-03330){ref-type="app"} shows that *m*/*z* 871.4, 909.3, 1034.2 and 1072.1 were the major ions in the positive ion mode. [Figure S12](#app1-molecules-23-03330){ref-type="app"} shows *m*/*z* 629.8, 661.8, 709.4, 793.4, 815.4 and 837.4 as major ions in the negative ion mode. Due to quite similar MALDI profiles, multivariate analysis was applied in an attempt to differentiate the infected plants into three clusters, similar to what was observed with the LC-MS analysis of the control, early asymptomatic infection (cluster I, 4--36 hpi) and late asymptomatic infection stage (cluster II, 48--96 hpi). 2.3. Multivariate Data Analysis of MALDI-MS {#sec2dot3-molecules-23-03330} ------------------------------------------- As shown in [Figure 4](#molecules-23-03330-f004){ref-type="fig"}a,b and [Figure 5](#molecules-23-03330-f005){ref-type="fig"}a,b, the MALDI-MS metabolic profiles of early and late asymptomatic infected plants (clusters I and II) and control plants could be differentiated using the PLS-DA analysis. Therefore, such fast MALDI-MS analysis could be used to identify late blight in tomato plants in the absence of symptoms on the basis of their corresponding metabolic profiles. The VIP score plots ([Figure 4](#molecules-23-03330-f004){ref-type="fig"}c,d) show discriminating features for the control, early and late asymptomatic infection clusters that were detected in the positive ion mode (VIP scores \> 1). The discriminating ion of *m*/*z* 675.4 decreases throughout the infection, but the discriminating ions of *m*/*z* 1034.2 and 1072.1 decrease in the early asymptomatic infection ([Figure S11](#app1-molecules-23-03330){ref-type="app"}) and then increase in the late asymptomatic infection stage (see *m*/*z* 1034.2; 1056.2 and 1072.1 ion, [Figure S11](#app1-molecules-23-03330){ref-type="app"}). These ions are attributed to the protonated molecule and adducts of the glycoalkaloid α-tomatine, that is, \[M + H\]^+^ of *m*/*z* 1034.2; \[M + Na\]^+^ of *m*/*z* 1056.2; and \[M + K\]^+^ of *m*/*z* 1072.1, which is also detected in LC-MS as the ion at *m*/*z* 1078.5436 \[M + FA − H\]^−^. The decrease in the relative abundance of α-tomatine adducts in early asymptomatic infection in relation to the control should therefore be related to the successful infection of asymptomatic tomato plants within the first hours. However, there was a subsequent increase of α-tomatine adduct intensities at the late infection stage compared to control, which may be associated with the posterior glycosylation of tomatidine to α-tomatine. This glycosylation appears to be crucial in protecting the cell from the toxic effect of steroidal alkaloids, such as tomatidine, which causes marked developmental defects, including the growth retardation of tomato plants \[[@B51-molecules-23-03330]\]. The most downregulated metabolite in the late asymptomatic infection stage ([Figure S11](#app1-molecules-23-03330){ref-type="app"}) was detected as the ion of *m*/*z* 871.5 \[M + H\]^+^ (VIP \> 2.5), attributed to pheophytin α. The ions of *m*/*z* 893.3 and *m*/*z* 909.4 correspond to the adducts \[M + Na\]^+^ and \[M + K\]^+^, respectively, of pheophytin α ([Figure S11](#app1-molecules-23-03330){ref-type="app"}). This same ion at *m*/*z* 871.5609 was detected by LC-MS, and its fragmentation pattern allowed us to attribute it as pheophytin α. Pheophytin α is a chlorophyll derivative involved in the electron transfer pathway of photosystem II \[[@B77-molecules-23-03330],[@B78-molecules-23-03330]\]. It has been found that genes involved in photosynthesis and chlorophyll biosynthesis are downregulated upon challenge by virulent and avirulent pathogens \[[@B79-molecules-23-03330],[@B80-molecules-23-03330],[@B81-molecules-23-03330]\], along with the upregulation of defense-related pathways \[[@B82-molecules-23-03330]\]. Likewise, a few metabolites detected in the negative ion mode (score \> 0.5) have similar behavior throughout the infection. The intensities of the ions of *m*/*z* 837.4; 839.4 and 831.4 increased in the early stage in relation to control, but those of the ions of *m*/*z* 709.3 and 725.3 are most abundant in the control in relation to the early and late asymptomatic infection stages ([Figure 5](#molecules-23-03330-f005){ref-type="fig"}c,d). The upregulated ions of *m*/*z* 793.4 (\[M − H\]^−^), *m*/*z* 815.4 (\[M -- Na − 2H\]^−^) and *m*/*z* 837.4 (\[M + FA − H\]^−^) ([Figure S12](#app1-molecules-23-03330){ref-type="app"} and [Figure 5](#molecules-23-03330-f005){ref-type="fig"}c) are attributed to the sulfolipid known as 1,2-di-*O*-palmitoyl-3-*O*-(6-sulfoquinovopyranosyl)glycerol (*m*/*z* 793.5109) isolated from *Byrsonima crassifolia* \[[@B83-molecules-23-03330]\]. Sulfoquinovosyl diacylglycerols (SQDGs) and phosphatidylglycerols (PGs) are major classes of the thylakoid membrane lipids in plants \[[@B84-molecules-23-03330],[@B85-molecules-23-03330],[@B86-molecules-23-03330]\]. The increase in SDQGs by *P. infenstans* inoculation might be indicating compositional and/or structural changes of the chloroplast membrane, as reported in tobacco plants inoculated with *Phytophthora parasitica* \[[@B87-molecules-23-03330]\]. The score of the VIP plots ([Figure 4](#molecules-23-03330-f004){ref-type="fig"}c,d and [Figure 5](#molecules-23-03330-f005){ref-type="fig"}c,d ) shows that the discrimination power of metabolites is lower in MALDI-MS than in LC-MS, revealing that LC-MS metabolomics provides more significant potential biomarkers. It is likely that MALDI-MS suffers from extensive ion suppression \[[@B76-molecules-23-03330],[@B88-molecules-23-03330]\] due to a lack of chromatographic separation and to interferences of the low molecular weight (\<500 Da) ions of the matrix \[[@B75-molecules-23-03330]\]. The MALDI-MS approach could be relevant to future applications for detecting late blight directly on the leaf material through imaging analysis. Although the irregularity of the leaf surface is a limitation for some MSI techniques \[[@B89-molecules-23-03330],[@B90-molecules-23-03330]\], there are some strategies, such as imprinting with several adsorbent materials, that allow the selective adsorption of specific metabolites and that would enable the rapid detection of infection in tomato plants. 3. Materials and Methods {#sec3-molecules-23-03330} ======================== 3.1. Chemicals {#sec3dot1-molecules-23-03330} -------------- The LC-MS-grade solvents used were acetonitrile, isopropanol (Sigma-Aldrich, St. Louis, MO, USA) was used as an additive for the mobile phase and MALDI matrix solution preparation. The ultrapure water was purified by a Direct-Q water system (Millipore, Bedford, MA, USA). The MALDI matrices were 2,5-dihydroxybenzoic acid (DHB), 2-mercaptobenzothiazole (MBT), 6-aza-2-thiothymine (ATT), 4-nitroaniline (4-NA), N,N,N′,N′-tetramethyl-1,8-naphthalenediamine (DMAN), 9-aminoacridine (9-AA) and *trans*-ferulic acid (Sigma-Aldrich). The lipid standards used for TOF mass calibration were sphingomyelin, 1,2-dipalmitoyl-*sn*-glycero-3-phosphocholine, 2-oleoyl-1-palmitoyl-glycero-3-phosphocholine, and L-α-phosphatidylcholine (Sigma-Aldrich). 3.2. Tomato Plant Samples {#sec3dot2-molecules-23-03330} ------------------------- The seeds of the Santa Cruz Kada tomato cultivar were obtained from a local market. The seeds were germinated in a germination chamber (120 mm Petri dishes containing wet germination paper) maintained at 18 °C with a 16 h photoperiod for three days. The germinated seeds were planted in 16 cm diameter pots containing a 1:1 mixture of soil and vermiculite and subsequently subirrigated once per day. The 45 plants were maintained at 18 °C, with a 16 h photoperiod, and approximately 60--70% relative humidity (RH). 3.3. Pathogen Strain {#sec3dot3-molecules-23-03330} -------------------- A culture of *P. infestans*, named as IBSP-34, was obtained from the microorganism collection of the Biological Institute of Sao Paulo, SP, Brazil. The isolate was subcultured on V8 media \[[@B91-molecules-23-03330]\] in 90 mm Petri dishes at 18 °C. After 2--3 weeks, a sporangial suspension was prepared by scraping the surfaces of the colonies with a sterile scalpel, and the mycelia were suspended in sterilized water to produce the infecting solution. The concentration of sporangia in the suspension was adjusted to 1.0 × 10^5^ sporangia mL^−1^ using a Neubauer chamber. 3.4. Infection of Tomato Plants with Phytophthora Infestans {#sec3dot4-molecules-23-03330} ----------------------------------------------------------- After 5--6 weeks, 40 plants were inoculated at the same time with 10 μL of the sporangial suspension of *P. infestans* at four different sites, two on each side of the midrib of the leaf. The infection was carried out by carefully depositing a drop of the sporangial suspension on the leaf with the aid of a micropipette, without mechanically wounding the leaf. The 5 control plants were inoculated with ultrapure water using the same procedure. The experimental design for the infection consisted of two sets of plants that were the initial control (CN) and the inoculated (IN) plants. At each time point after inoculation (4, 12, 24, 36, 48, 60, 72 and 96 hpi), five infected plants were randomly collected and separately analyzed as replicates. 3.5. Sample Preparation {#sec3dot5-molecules-23-03330} ----------------------- All the tomato leaves were excised with sterilized scissors and were immediately macerated in liquid nitrogen. A 100 mg aliquot of the crushed powder was vortexed (Multi Reax, Heidolph, Schwabach, Germany), extracted with 1 mL of methanol for 10 min at room temperature, and then centrifuged for 5 min at 12,000× *g* at 20 °C (Centrifuge 5418, Eppendorf, Hamburg, Germany). The supernatants were stored at −20 °C until analysis. For LC-MS analysis, 50 µL of the supernatant was diluted with 950 µL of methanol in a vial. For MALDI-MS analysis, 1 µL of the supernatant was deposited on the MALDI plate. 3.6. Untargeted Analysis of Metabolites {#sec3dot6-molecules-23-03330} --------------------------------------- ### 3.6.1. UHPLC-Q-TOF-MS Analysis {#sec3dot6dot1-molecules-23-03330} The extracts of infected and noninfected leaves were analyzed with a 1290 Infinity UHPLC coupled to an Agilent 6550 iFunnel Q-TOF LC-MS system with Agilent Dual Jet Stream electrospray ionization technology (ESI, Agilent Technologies, Santa Clara, CA, USA). For metabolite separation, a Kinetex XB-C18 Core-Shell column (2.1 × 150 mm, 1.7 μm, 100 Å; Phenomenex Inc., Torrance, CA, USA) was used and maintained at 40 °C. The mobile phases were A (0.1% aqueous formic acid) and B (0.1% formic acid in methanol). The chromatographic gradient of B was increased from 5% to 95% over 18 min and maintained for 7 min at 95% with a flow rate of 0.35 mL·min^−1^. Subsequently, the initial conditions were reached in 8 min, and the column was equilibrated for 7 min. The injection volume was 2 µL. To avoid degradation, the samples were maintained at −20 °C in a freezer, and prior to analysis they were placed in an autosampler maintained at a room temperature of approximately 21 °C. Prior to injection, the needle was washed for 20 s with a mixture of H~2~O:ACN:IPA (4:4:2) using the flush port. The external needle wash volume was 150 µL. The ESI source was used in the positive or negative ion mode in the following conditions: drying gas temperature 250 °C, drying gas flow rate 14.0 L·min^−1^, sheath gas temperature 250 °C, sheath gas flow rate 10.0 L·min^−1^, nebulizer gas 45 psig, and capillary voltage +3.5 kV and −3.5 kV for the positive and negative ionization mode, respectively. The Q-TOF parameters were acquisition rate of 1.0 spectra/s over the *m*/*z* 100−1700 amu range, skimmer voltage 65 V, octopole RF 750 V, fragmentor 150 V and nozzle voltage 350 V. The multichannel plate detector voltage was 650 V, and the photomultiplier tube voltage was 700 V. The pulser was set to a pulse width of 125 counts/pulse, with a pulse width of 25 counts. A second reference sprayer orthogonal to the sample sprayer in the electrospray source was used to introduce a reference solution for accurate mass determinations. The reference mass ions were of *m*/*z* 121.0509 and *m*/*z* 922.0098 in ESI (+) and *m*/*z* 119.0363 and *m*/*z* 966.0007 in ESI (−). The detection window for the reference masses was set to 10 ppm, with an average of 10 scans and a minimum peak height of 100 counts/s. A mass calibration was performed with an Agilent tune mix from 100 to 1600 Da. The data were acquired in profile mode using high resolution mode (2 GHz). First, an experiment in full scan mode over the entire mass range was performed, and then two fragmentation experiments were carried out. The auto MS/MS experiments were performed with fixed energies of 30, 40 or 50 eV, in which a first step was to select the 10 most intense ions per scan to produce the MS^2^ spectra. The auto MS/MS experiments that were performed in the variable energy mode used scan speeds that varied based on the precursor abundance and used energies based on the precursor mass-to-charge. The collision energies applied were 6.5 V/100 Da and were offset by 2.0 V. The QC samples consisted of a pool of all the different inoculation times and controls and were analyzed at the beginning and at the end of each batch and after every 10 injections. ### 3.6.2. MALDI-MS Profile Analysis {#sec3dot6dot2-molecules-23-03330} The extracts (1 μL) were directly smeared onto a 384-position MALDI target plate (Bruker Daltonics, Bremen, Germany). After drying, the spots were immediately overlaid with 1 μL of the matrix solution: 2,5-dihydroxybenzoic acid (DHB) for the positive ion mode and 2-mercaptobenzothiazole (MBT) for the negative ion mode. All MALDI matrices used for method optimization (DHB, ATT, 4-NA, MBT, 9AA and *trans*-ferulic acid) were prepared at a concentration of 10 mg·mL^−1^ in methanol/water (80:20 *v*/*v*). The DMAN matrix solution only was prepared at a concentration of 10 mg·mL^−1^ in methanol/water (90:10 *v*/*v*) because of its low solubility. Tuning of the smart beam laser position and energies, as well as electronic adjustment of the lens, was performed by Bruker when necessary. The external TOF calibration was made using 1 μL of a lipids mixture composed of sphingomyelin, 1,2-dipalmitoyl-*sn*-glycero-3-phosphocholine, 2-oleoyl-1-palmitoyl-glycero-3-phosphocholine and [l]{.smallcaps}-α-phosphatidylcholine, followed by 1 μL of DHB applied in the same conditions as the samples \[[@B92-molecules-23-03330]\]. The MALDI-MS analyses were performed on a Bruker Autoflex III MALDI--TOF/TOF mass spectrometer equipped with a 334 nm smart beam laser. The metabolite profiles were acquired in the TOF reflector mode, using the positive or negative ion mode. The accelerating voltage was +20 kV for positive and −20 kV for the negative ion mode, with delayed extraction of 260 ns for both. Each spectrum was manually collected as an average of 5000 laser shots (1000 laser shots at five different positions on the same spot). The laser energy was set at 70% for spectra acquisition. The *m*/*z* 600--1500 range was used for metabolite profile acquisition of both positive and negative ion modes. The spectra were acquired in triplicates via the AutoExecute tool of the Flexcontrol acquisition software (version 2.4; Bruker-Daltonik GmbH, Bremen, Germany). 3.7. Multivariate Data Analysis {#sec3dot7-molecules-23-03330} ------------------------------- ### 3.7.1. LC-MS-Based Metabolomics Data {#sec3dot7dot1-molecules-23-03330} The LC-MS data processing was performed using MassHunter Qualitative Analysis (Agilent Software B07.00, Santa Clara, CA, USA), in which raw original data was converted to the *mzData* format. The *mzData* files were uploaded to XCMS online for further data processing (<https://xcmsonline.scripps.edu/>). The XCMS software was used for feature detection, retention time correction, feature alignment and univariate statistical analysis \[[@B93-molecules-23-03330]\]. The data were analyzed as pairwise jobs, with the following settings: centWave feature detection with 5 ppm of maximal tolerated *m*/*z* deviation; minimum peak width 5 s; maximum peak width 20 s; signal/noise threshold 6; noise filter abundance 0; prefilter abundance 100; mzdiff 0.01, and integration method type 1. Obiwarp retention time correction with 1 *m*/*z* step size (profStep) was used to generate the profiles. Other parameters were alignment: mzwid 0.015; minfrac 0.5; mz width: 0.015; bw: 5. An unpaired parametric t test (Welch test) was performed to identify significant features with a p value threshold of 0.05 and a fold change threshold (highly significant features) of 1.5 \[[@B94-molecules-23-03330]\]. The *.csv* file from the XCMS processing was uploaded to MetaboAnalyst 3.0 (<http://www.metaboanalyst.ca/>) for multivariate statistical data analysis. The file comprised a list of features (*m*/*z*, retention times and intensities) for all samples from infected and noninfected leaves. The data processing applied an integrity check, missing value check, data filter, and normalization before statistical analysis \[[@B95-molecules-23-03330],[@B96-molecules-23-03330]\]. The presence of missing values or features with constant values (i.e., all zeros) was checked, and data filtering using the interquantile range (IQR) was applied to remove variables close to the baseline or detection limits and variables with near-constant value \[[@B97-molecules-23-03330]\]. ### 3.7.2. MALDI-MS Profile Data {#sec3dot7dot2-molecules-23-03330} The analysis of MALDI data was conducted in three distinct steps: (1) preprocessing, (2) processing and (3) statistical analysis. The raw spectra were preprocessed in the FlexAnalysis software (Bruker-Daltonik GmbH, Bremen, Germany) after baseline subtraction for background removal, alignment of the spectra scale, ion selection with an S/N ratio greater than 3 and normalization of intensities. Data processing was performed before multivariate analysis for the metabolite profiles in MetaboAnalyst 3.0. The uploaded files (.*csv* format) comprised a list of features (*m*/*z* and relative intensities). The ions were realigned within a tolerance of *m*/*z* 0.4 (0.4 Da) to remove ions that appeared in less than half of the samples in each group \[[@B92-molecules-23-03330]\]. The presence of missing values or features with constant values (i.e., all zeros) was checked, and data filtering using relative standard deviation (RSD) was applied to remove variables close to baseline or detection limit and variables with near-constant values. Both LC-MS and MALDI-MS data were normalized by sum for adjustment of the differences among samples and Pareto scaling (mean-centered and divided by the square root of the standard deviation of each variable) was used to make individual features more comparable \[[@B92-molecules-23-03330]\]. To discriminate the infection times of the tomato plants with *P. infestans* based on their LC-MS and MALDI-MS metabolite profiles, principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA) and orthogonal partial least squares discriminant analysis (OPLS-DA) was performed on the data using MetaboAnalyst 3.0. To identify the molecular features related to variation between groups of samples, the corresponding loading plots and the variable importance in projection (VIP) were applied \[[@B98-molecules-23-03330]\]. The validation of the classification models obtained by multivariate analysis was made to confirm the capability of classification and prediction of the models. The data were permuted 100 times, and Q2 and R2Y were used as quality-of-fit criterion \[[@B99-molecules-23-03330]\]. 3.8. Annotation of Metabolites {#sec3dot8-molecules-23-03330} ------------------------------ The major discriminant features of the LC-MS analyses were selected out of the multivariate analysis, and their annotation was made according to the exact mass (*m*/*z*) of the protonated or deprotonated ion and their fragmentation spectra. The *.txt* data archives with *m*/*z* and relative abundance of the ions obtained from MS and MS/MS spectra were uploaded to MS-FINDER software ver. 2.26 (<http://prime.psc.riken.jp/Metabolomics_Software/MS-FINDER/>), which provides molecular formulas of the precursor ions based on accurate mass, isotope ratio, and product ion spectra \[[@B100-molecules-23-03330]\]. The Metlin (<http://metlin.scripps.edu>), ChemSpider ([www.chemspider.com](www.chemspider.com)), MassBank (<http://www.massbank.jp/>), HMDB (<http://www.hmdb.ca/>), and LipidMaps (<http://www.lipidmaps.org/>) spectra databases and a comparison with fragmentation profiles of previously reported metabolites from tomato infection were also used for confirmation of the metabolite annotation \[[@B1-molecules-23-03330],[@B38-molecules-23-03330],[@B51-molecules-23-03330],[@B101-molecules-23-03330],[@B102-molecules-23-03330],[@B103-molecules-23-03330]\]. 4. Conclusions {#sec4-molecules-23-03330} ============== The LC-MS metabolites profiles discriminate between early and late asymptomatic infection, and between each infection time in the infected tomato plants and identified major metabolites that are altered in late blight. Metabolites detected via LC-MS operating in the negative ion mode provided more discriminant clusters compared to those detected in the positive ion mode. The annotated metabolites correspond to tricarboxylic acid (TCA) cycle secondary metabolites and include terpenoids, flavonoids, alkaloids, saponins, sesquiterpenes and glycoalkaloids. We found the metabolite tomatidine to be an important biomarker of infection because it is produced by the action of the fungal pathogen enzymes. Also, we found that saponins might be early infection metabolite markers because their abundance increases between 4 and 36 hpi as specific response to the type of sterols present in the pathogen membrane. We found the metabolite isocoumarin (M301T17) as a good infection marker because its abundance increases linearly along the post infection time. These metabolites could be relevant in future applications to detect late blight directly on the asymptomatic leaf material through imaging analysis using DESI-MS or MALDI imaging. Additionally, the metabolite profiles obtained by MALDI (±)-MS, associated with multivariate analysis, have provided late blight detection of tomato plants in early and late asymptomatic infection. The major discriminant metabolites are α-tomatine, pheophytin α and 1,2-di-*O*-palmitoyl-3-*O*-(6-sulfoquinovopyranosyl)glycerol, but α-tomatine has an important role in infection control because it decreases within the first hours and increases in the late asymptomatic infection stage. MALDI (±)-MS seems to offer a rapid and effective method to detect late blight in asymptomatic tomato plants and therefore it could function as a suitable guide for the management of sanitary defense approaches. **Sample Availability:** Not available. The following are available online, Figures S1 and S2: (**a**) Chromatographic profiles and (**b**) Cloud plot of metabolites detected in infected tomato samples, LC-ESI (+)-MS and LC-ESI (−)-MS, respectively; Figures S3 and S4: PCA, LC-ESI (+)-MS and LC-ESI (−)-MS, respectively; Figures S5 and S6: LC-ESI (+)-MS and LC-ESI (−)-MS, respectively; Figures S7--S10: Validation; Figures S11--S12: MALDI-MS profiles. Table S1: Metabolites in tomato plants infected with *P. infestans*. ###### Click here for additional data file. P.G.G., F.N.d.S. and C.C. developed the idea. P.G.G. performed most of the experimental work, data analysis and wrote the first draft of the manuscript. F.N.d.S. helped with the multivariate data analysis and the discussion. S.Z. contributed to the inoculum preparation. F.N.d.S., M.N.E. and C.C. helped with the proofreading and revision of the manuscript. This research was funded by Facultad de Ciencias, Universidad de los Andes, COLCIENCIAS supportedthe doctoral fellowship \[6172\] to Paula Galeano, the National Council for Scientific and Technology (CNPq) supported the fellowship \[140743/2013-8\] to Fabio Neves and general financial support \[447708/2014-7\] to ThoMSon Mass Spectrometry Laboratory. The authors declare no conflicts of interest. ###### OPLS-DA score plots for the eight post-infection time points (4, 12, 24, 36, 48, 60, 72 and 96 hpi) and the initial control: (**a**) LC-ESI (+)-MS and (**b**) LC-ESI (−)-MS. VIP score-plot derivate of PLS-DA analysis (**c**) LC-ESI (+)-MS and (**d**) LC-ESI (−)-MS. The coding M and T after the number indicates nominal mass and retention times, respectively, of each feature. Cluster I includes the infection times of 4--36 hpi. Cluster II includes the infection times of 48--94 hpi. Coding 4-hpi indicates positive ionization mode analysis for the samples collected at four hours post infection. Coding 4-hpi_neg means negative ionization mode analysis of the samples collected at four hours post infection. Coding C and C_neg means the analysis of the controls in positive and negative ionization modes, respectively. Variable Importance in Projection (VIP), is a weighted sum of squares of the PLS loadings taking into account the amount of explained Y-variation, in each dimension. VIP scores are calculated for each component and were used average of two components to calculate the feature importance. The color scale depends on the range intensity of the metabolites in all samples and the media intensity of the samples the same time.   {#molecules-23-03330-f002} {#molecules-23-03330-f003} ###### Multivariate partial least squares discriminant analysis applied to the MALDI (+)-MS data: (**a**) Early asymptomatic infection and control and (**b**) Late asymptomatic infection and control. The VIP plots (**c**) Early asymptomatic infection and control and (**d**) Late asymptomatic infection and control.   {#molecules-23-03330-f005} molecules-23-03330-t001_Table 1 ###### Principal metabolites associated with tomato late blight annotated with LC-MS/MS data. Annotation level: identified metabolites (level 1), putatively annotated compounds (level 2), putatively characterized compound classes (level 3), and unknown compounds (level 4) \[[@B73-molecules-23-03330],[@B74-molecules-23-03330]\]. Molecular Feature RT (min) Metabolite Molecular Formula Annotation Level Theor. (*m*/*z*) Found (*m*/*z*) AME (ppm) Adducts ------------------- ---------- ------------------------------------------------------------- ------------------- ------------------ ------------------ ----------------- ----------- ------------------- M609T9 9.06 Quercetin 7-(rhamnosylglucoside) C~27~H~30~O~16~ 2 609.14555 609.1464 1.4 \[M − H\]^−^ M1095T10 9.85 Hydroxytomatine isomer I C~50~H~83~NO~22~ 3 1094.53832 1094.5382 0.1 \[M + FA − H\]^−^ M1077T12 11.52 Dehydrotomatine C~50~H~81~NO~21~ 2 1076.52776 1076.5277 0.1 \[M + FA − H\]^−^ M1109T12 11.58 UGA11 C~52~H~87~NO~24~ 3 1108.55397 1108.5541 0.1 \[M − H\]^−^ M1121T12_1 11.72 UGA28 C~52~H~83~NO~25~ 3 1120.51759 1120.5120 5.0 \[M − H\]^−^ M1079T12 11.94 Tomatine C~50~H~83~NO~21~ 2 1078.5440 1078.5436 0.4 \[M + FA − H\]^−^ M433T13_2 12.86 Naringenin-hexose I C~21~H~22~O~10~ 3 433.11347 433.1115 4.5 \[M − H\]^−^ M317T13_2 13.09 Catechin 7,4′-dimethyl ether C~17~H~18~O~6~ 2 317.10251 317.1020 1.6 \[M − H\]^−^ M293T15_1 14.51 Phytuberin C~17~H~26~O~4~ 2 293.17528 293.1774 7.2 \[M − H\]^−^ M221T16_2 16.30 Rishitin C~14~H~22~O~2~ 2 221.15415 221.1561 8.8 \[M − H\]^−^ M737T20 19.55 Tuberoside J C~39~H~64~O~14~ 2 737.41121 737.4186 10.0 \[M − H\]^−^ M414T20 19.59 Tomatidine C~27~H~45~NO~2~ 2 414.3372 414.3385 3.1 \[M − H\]^−^ M791T21 21.46 Triterpenoid saponins- Sapimukoside J C~44~H~72~O~12~ 2 791.49455 791.4955 1.2 \[M − H\]^−^ M832T22_2 22.21 Peonidin 3-(4-sinapoylgentiobioside) C~39~H~43~O~20~ 2 832.24259 832.2387 4.7 \[M + H\]^+^ M794T22_1 22.24 1,2-Di-O-palmitoyl-3-O-(6-sulfoquinovopyranosyl)glycerol C~41~H~78~O~12~S 2 793.51357 793.5109 3.4 \[M − H\]^−^ M798T23_2 22.58 Triterpene saponins- UNPD101109 C~43~H~72~O~13~ 2 797.50511 797.5049 0.3 \[M + H\]^+^ M820T23 22.69 Sulfoquinovosyldiacylglycerols C~43~H~80~O~12~S 3 819.5289 819.5269 2.4 \[M − H\]^−^ M743T23_3 22.83 3-*O*-a-[l]{.smallcaps}-Arabinopyranosylproanthocyanidin A5 C~35~H~32~O~16~ 2 743.13788 743.1395 2.2 \[M + Cl\]^−^ M610T24 24.07 Apigenin 7-\[rhamnosyl-(1-\>2)-galacturonide\] C~27~H~28~O~15~ 2 610.17719 610.1758 2.3 \[M + NH~4~\]^+^ M845T24 24.40 Triterpene saponins- Cyclopassifloside III C~43~H~72~O~16~ 2 845.48986 845.4816 9.8 \[M + H\]^+^ M984T24_1 24.49 Triterpene saponins C~48~H~84~O~18~ 3 983.53461 983.5393 4.8 \[M + Cl\]^−^ M984T25 24.87 Triterpene saponins C~48~H~84~O~18~ 3 983.53461 983.5392 4.7 \[M + Cl\]^−^

Background {#Sec1} ========== *Enterobacter* is a genus of the family Enterobacteriaceae, consisting of common Gram-negative, facultative anaerobic, rod-shaped, non-spore-forming bacteria. *E. aerogenes* is recognized as an important bacterial pathogen in hospital acquired infections (Jarvis and Martone [@CR6]). In this study, we examined the transmission of *E. aerogenes* infections in two healthcare workers; by isolating the pathogen from the blood cultures of both the patients and complete genomic sequence analysis and pulsed-field gel electrophoresis (PFGE) revealing that both the isolates were identical. *Enterobacter* is the eighth most common pathogen in healthcare-associated infections in the United States (Hidron et al. [@CR5]) and constitutes 2.9 % of healthcare-associated bloodstream infections in Korea (Son et al. [@CR12]). We encountered two patients who were otherwise healthy nurses with no underlying conditions and had nearly identical clinical signs and symptoms 2 h apart, transferred to our Emergency department from the same hospital. The transmission of *E. aerogenes* in health care workers has not been reported anywhere until now; thus this report describes the first transmission of *E. aerogenes* in healthcare workers. Methods {#Sec2} ======= Case description {#Sec3} ---------------- We encountered two patients in our Emergency department on 17th October, 2013 with nearly identical clinical signs and symptoms who were otherwise healthy nurses with no underlying conditions. Both the patients were admitted 2 h apart with complaints of multiple episodes of vomiting and loose stool. They also complained of pain in the epigastric and right upper quadrant (RUQ) regions. On examination, the abdomen was soft but tender in the RUQ region, blood pressure (BP) was below the normal range, and heart rate and respiratory rate were elevated. The findings of the physical examinations of both patients were unremarkable. The laboratory findings of both patients are presented in Table [1](#Tab1){ref-type="table"}. Blood culture was performed; meanwhile, initial treatment was started with intravenous fluids, metronidazole and ceftriaxone. Arterial blood gas (ABG) analysis revealed metabolic acidosis in one patient (Case 1). The blood culture turned positive on the 2nd hospital day.Table 1Laboratory findings of the two casesCaseAge/sexLaboratory investigation findingsCo-infectionWBC (counts/mm^3^)Hemoglobin (g/dL)Platelet counts (mm^3^)Liver function testSerum creatinine (mg/dL)T. Bil (mg/dL)AST (U/L)ALT (U/L)S. Alb (g/dL)Case 141/F396011.660,0002.551222783.102.61Hepatitis C virus\ *Burkholderia cepacia*Case 237/F348012.953,0003.972562102.881.52--*WBC* white blood cell, *T. Bil* total bilirubin, *AST* aspartate aminotransferase, *ALT* alanine aminotransferase, *S. Alb* serum albumin After receiving the blood culture reports, which were positive for *E. aerogenes* in both patients, we investigated the two cases in detail, as we suspected a common source of *E. aerogenes* infection. The drug resistance patterns of the two isolated *E. aerogenes* strains were identical; resistant to ampicillin/sulbactam, intermediately resistant to imipenem. The two patients were not aware that they had experienced similar symptoms until after they were admitted to our hospital. During the investigation, we found that the two patients had used the same saline bottle to prepare ketorolac injections that they then self-administered, and that neither patient was aware of the other. Metronidazole was stopped after completion of a 5-day course, and ceftriaxone was replaced with cefepime on the 8th hospital day after we got culture positive result. The general condition of both patients improved gradually, and all biochemical parameters returned to within their normal ranges before the patients were discharged, which was after 2 weeks of hospitalization. Informed written consent was obtained from both the patients for participation in this study. Antimicrobial susceptibility testing {#Sec4} ------------------------------------ The clinical isolates were identified with a VITEK II automated system (bioMe´rieux, Marcy l'Etoile, France). Antimicrobial susceptibility determinations including the MICs were performed automatically with the VITEK II system. The MIC was interpreted as susceptible or resistant according to the guidelines of the Clinical and Laboratory Standards Institute (CLSI) MIC interpretive standards for *E. aerogenes* where applicable (Gouby et al. [@CR3]; National Committee for Clinical Laboratory Standards [@CR10], [@CR11]). Bacterial isolates and media {#Sec5} ---------------------------- *Enterobacter aerogenes* was isolated from the blood cultures of both patients with sepsis. The organisms were cultivated in LB broth (Difco Laboratories, Detroit, MI, USA) at 37 °C for 12-18 h with agitation. Nutrient agar (Eiken Chemical, Tokyo) and 5 % sheep blood agar (Becton--Dickinson, Tokyo) were the solid media used for *E. aerogenes* culture. Strains of *E. aerogenes* were stored at −80 °C in 3 % skim milk (Difco) supplemented with 5 % glucose (Difco). Polymerase chain reaction (PCR) and sequencing {#Sec6} ---------------------------------------------- Molecular identification of two *E. aerogenes* isolates was performed using conventional PCR (C-PCR) targeting the 16S rRNA gene. Genomic DNA was extracted from bacterial cultures using the QIAamp DNA mini kit (Qiagen, Westburg, Netherlands) according to the manufacturer's instructions. C-PCR was performed in a 20 µL reaction volume using the primers 27F (5′-AGAGTTTGATCCTGGCTCAG-3′; *Escherichia coli* 16S ribosomal DNA base pair positions 8--27) and 1492R (5′-TACGGHTACCTTGTTACGACTT-3′, positions 1492--1507). Each reaction mixture contained AmpliTaq Gold^®^ 360 Master Mix (Applied Biosystems, Waltham, MA, USA), 1 µL each of the 5 µM forward and reverse primers, and 2 µL of genomic DNA. The cycling conditions consisted of the following steps: 2 min at 95 °C; 30 cycles of 1 min at 95 °C, 30 s at 55 °C, and 45 s at 72 °C; and a 10 min extension at 72 °C. The PCR products were visualized by electrophoresis on an ethidium bromide-stained 1.5 % agarose gel. A Biosystems Veriti™ 96-Well Thermal Cycler (Applied Biosystems, Foster City, CA) was used for this experiment. Amplified and purified DNA was prepared for direct sequencing using a QIAquick PCR Purification Kit (Qiagen, Westburg, Netherlands) and was sequenced by dideoxy termination with an automatic sequencer (ABI Prism 3730XL DNA analyzer). Sequence homology analysis was performed by the National Center for Biotechnology Information (National Institutes of Health) BLAST network service, and both of our sample species had 99 % pairwise similarity with the complete genome of *E. aerogenes* KCTC 2190 (accession no. CP002824). Pulsed-field gel electrophoresis of isolates {#Sec7} -------------------------------------------- We also performed pulsed-field gel electrophoresis (PFGE) to confirm that the two *Enterobacter* species isolated from the patients were the same strain of the bacterium. The two *E. aerogenes* isolates were subjected to DNA restriction analysis with 10 U/µl of the SmaI enzyme in appropriate buffer. The DNA fragments were separated by pulsed-field gel electrophoresis through a 1.2 % agarose gel as described previously (Murchan et al. [@CR9]). We could document the identical DNA banding patterns based on typing results (Fig. [1](#Fig1){ref-type="fig"}).Fig. 1Pulsed-field gel electrophoresis (PFGE) profiles of *E. aerogenes* from humans after Xba-I digestion. *Lane M*: markers (*Salmonella braenderup* ATCC BAA-664), *Lanes B, C*: case 1 and *Lanes D, E*: case 2 Results {#Sec8} ======= *Enterobacter aerogenes* was isolated from the blood cultures of both patients. Complete genomic sequencing analysis revealed that both the isolates were identical. PFGE also showed that both the strains were indistinguishable (Fig. [1](#Fig1){ref-type="fig"}). The drug resistance patterns of the two isolated *E. aerogenes* strains were identical. They were resistant to ampicillin/sulbactam (MIC ≥ 2 µg/mL) and were intermediately resistant to imipenem (MIC ≥ 2 µg/mL). They were susceptible to amikacin (MIC \< 2 µg/mL), ceftriaxone (MIC \< 1 µg/mL), tigecycline (MIC \< 0.5 µg/mL), gentamicin (MIC \< 1 µg/mL), piperacillin (MIC \< 4 µg/mL) colistin (MIC \< 2 µg/mL), cefepime (MIC \< 1 µg/mL), ceftazidime (MIC \< 2 µg/mL), ciprofloxacin (MIC 0.25 µg/mL), meropenem (MIC \< 0.25 µg/mL), minocycline (MIC \< 1 µg/mL) and piperacillin/tazobactam (MIC \< 4 µg/mL). Both patients were managed successfully, and all biochemical parameters returned to within their normal ranges within 2 weeks of hospitalization, at which time the patients were discharged. Discussion {#Sec9} ========== *Enterobacter aerogenes*, a component of the normal flora of the human gastrointestinal tract, is a significant nosocomial pathogen and a common cause of iatrogenic bacteremia (Hidron et al. [@CR5]). The incidence of bacteremia due to *E. aerogenes* has increased gradually, accounting for nearly 11 % of nosocomial infections in some series (Acolet et al. [@CR1]; Lin et al. [@CR8]). Although community-acquired infections are occasionally observed, nosocomial infections are, the most frequent by far. Patients most susceptible to acquire *Enterobacter* infections are those who stay in the hospital, especially in the intensive care unit (ICU) for prolonged periods. Other major risk factors for *Enterobacter* infections include the prior use of antimicrobial agents; concomitant malignancy (especially hemopoietic and solid organ malignancies); hepatobiliary disease; ulcers of the upper gastrointestinal tract; diabetes mellitus; chronic renal failure; and immunosuppression (Lin et al. [@CR8]). The gastrointestinal tract is a common endogenous reservoir for *E. aerogenes,* and spread of infection from the gastrointestinal tract is difficult to ascertain, which may explain why the portal of entry of *E. aerogenes* often cannot be identified (Kanemitsu et al. [@CR7]). Outbreaks have been traced to various common sources, including total parenteral nutrition solutions, isotonic saline solutions, albumin, digital thermometers, intravenous catheters, mechanical ventilator and dialysis equipment. In both of our patients, sepsis was severe and had initial symptoms of gastro-enteritis, and the same *E. aerogenes* strain was isolated in the blood of both patients. Our first patient also had a co-infection with *Burkholderia cepacia*, a member of a bacterial group known as the *B. cepacia* complex. This infection mainly occurs in patients with underlying lung disease, such as cystic fibrosis and chronic granulomatous disease, and in immunocompromised individuals, hospitalized patients and drug addicts (Govan et al. [@CR4]). A limitation of this study is that we could not find the residual saline solution, used for diluting ketorolac tromethamine prior to intravenous injection, as it had already been discarded before we began investigating the source of *E. aerogenes* in these two cases. A peculiarity of our cases is that both patients developed severe sepsis preceded by gastroenteritis symptoms after self-administration of intravenous ketorolac mixed with normal saline taken from a common source at the hospital where they worked. We do not know exactly how the normal saline solution was contaminated with *Enterobacter.* Although it is evident that *Enterobacter* strains commonly arise from endogenous intestinal flora of hospitalized patients and that a nurse's hand can be contaminated with *Enterobacter* while taking care of patients, this incident highlights the possibility of hospital-acquired sepsis due to traditional nosocomial microorganisms in drug abusers who are also healthcare workers and have access to hospital supplies. Neither of these patients had any history of any severe underlying diseases, prior antimicrobial use or any previous hospital admission. Conclusion {#Sec10} ========== Here, we conclude that in addition to the general population, healthcare workers, especially those who are also intravenous drug abusers, should be considered as subjects who could be a source of transmission of pathogens like *E. aerogenes*. Piyush Jha and Choon-Mee Kim contributed equally to this paper PJ, CMK and JHC were involved in data analysis and writing of this manuscript. DMK and SJJ planned this study and reviewed the final manuscript. NRY, BJ and SWK did data collection and experimental analysis. YJA, JKC and DYJ reviewed the literature. All authors read and approved the final manuscript. Acknowledgements {#FPar1} ================ None. Competing interests {#FPar2} =================== The authors declare that they have no competing interests.

Enteropathy-associated T-cell lymphoma (EATL) is defined as an intestinal lymphoma of intraepithelial T lymphocytes. EATL is further classified into two distinct types: type I (classical) EATL, which comprises 80%--90% of all cases; and type II EATL, a monomorphic variant of the disease. Type I EATL occurs at a higher frequency in northern Europe, where celiac disease is more common, and is characterized by the presence of large tumor cells with a CD3^+^CD4^--^CD8^--^CD56^--^ immunophonotype \[[@b1-kjpathol-48-6-426]\]. Conversely, type II EATL, originally described as CD56^+^ intestinal lymphoma, consists of monomorphic small- to medium-sized tumor cells, typically with a CD3^+^CD4^--^CD8^+^CD56^+^ immunophenotype, with weak or no association with celiac disease \[[@b1-kjpathol-48-6-426],[@b2-kjpathol-48-6-426]\]. Recently, a number of reports have defined type II EATL as a distinct T-cell neoplasm predominant in patients of Asian ethnicity with no history of enteropathy or Epstein-Barr virus (EBV) association. In addition, type II EATL is characterized by frequent expression of gamma-delta T-cell receptors (γδ TCR). EBV-positive cases were suggested to represent extranodal natural killer (NK)/T-cell lymphoma rather than type II EATL. Whether EBV-positive cases with similar morphology and phenotype should be included in the definition of EATL and whether a proportion of these cases are γδ T-cell lymphomas remain debatable. Herein, we report a case of T-cell lymphoma of the jejunum with a CD3^+^CD4^--^CD8^+^CD56^--^betaF1^+^ phenotype, rearranged γ TCR genes, and diffuse EBV-encoded RNA (EBER) positivity. CASE REPORT =========== A 54-year-old Korean man was referred to our institute for evaluation of recurrent hematochezia. In the previous three years, the patient had experienced five episodes of hematochezia, the last occurring three days prior to presentation, and all occasions required embolization of the superior mesenteric artery. He appeared chronically ill but was mentally alert. Initial vital signs were within normal limits: blood pressure, 112/75 mm Hg; heart rate, 82/min; respiration, 16/min; and body temperature, 36.4°C. Complete blood counts were as follows: white blood cell count, 6,100/µL; hemoglobin, 10.4 g/dL; and platelet count, 187×10^3^/µL with normocytic normochromic peripheral blood morphology. Abdominopelvic computed tomography showed diffuse thickening of the proximal jejunal wall without obstruction and multiple enlarged mesenteric lymph nodes ([Fig. 1A](#f1-kjpathol-48-6-426){ref-type="fig"}). Endoscopy showed multiple discrete geographic ulcers oozing fresh blood in the proximal jejunum ([Fig. 1B](#f1-kjpathol-48-6-426){ref-type="fig"}). The patient underwent a segmental resection of the small intestine. On macroscopic examination, multiple ill-demarcated geographic ulcers covered with hemorrhagic exudate were identified in the jejunum. Obstruction or perforation was not observed. The intestinal mucosa was diffusely edematous ([Fig. 1C](#f1-kjpathol-48-6-426){ref-type="fig"}). Microscopically, the central zone of the ulcer showed dense lymphocytic infiltration in the lamina propria and submucosa. The infiltrate was composed of monotonous, small- to medium-sized lymphocytes with scanty cytoplasm and round, hyperchromatic nuclei ([Fig. 2A](#f2-kjpathol-48-6-426){ref-type="fig"}, [B](#f2-kjpathol-48-6-426){ref-type="fig"}). Angiocentricity, angioinvasion, or coagulative necrosis was not observed. On immunophenotyping, the tumor cells were CD3^+^CD5^+^CD8^+^TIA-1^+^TCR-βF1^+^CD20^--^CD56^--^ and CD4^--^ ([Fig. 3](#f3-kjpathol-48-6-426){ref-type="fig"}). *In situ* hybridization for EBER showed strong signals in the majority of tumor cells, which were CD3^+^ T-cells. The adjacent jejunal mucosa showed extensive intraepithelial lymphocytosis in the epithelium, mild villous atrophy, and crypt hyperplasia ([Fig. 2C](#f2-kjpathol-48-6-426){ref-type="fig"}, [D](#f2-kjpathol-48-6-426){ref-type="fig"}). These lymphocytes revealed cytological and immunologic features identical to those of the central zone. Polymerase chain reaction analysis using Biomed primers showed clonally rearranged TCR γ genes. Based on these findings, the lesion was diagnosed as type II EATL with EBV positivity. Chemotherapy with cyclophosphamide, doxorubicin, and vincristine was started, and the patient was alive with no evidence of recurrence at 23 months after the surgery. DISCUSSION ========== Type II EATL is a newly recognized entity that requires further characterization and exhibits distinctive clinicopathological features. This lymphoma is the predominant type of EATL in Asian populations and is not associated with celiac disease. Histologically, type II EATL reveals monotonous small- to medium-sized tumor cells and lack of EBV infection of the tumor cells. The immunophenotype of this tumor is typically CD3^+^CD8^+^CD56^+^CD4^--^ and frequently γδ TCR^+^ \[[@b2-kjpathol-48-6-426]\]. On histological and immunohistochemical analyses, the present case is compatible with type II EATL as shown by enteropathic features (villous atrophy, crypt hyperplasia, and diffuse intraepithelial lymphocytosis) and a CD3^+^CD8^+^TIA-1^+^TCR-βF1^+^CD4^--^ immunophenotype. Although type II EATL typically expresses CD56, approximately 9% of cases are CD56- as in the present case \[[@b3-kjpathol-48-6-426]\]. However, EBV positivity is not a typical feature of type II EATL \[[@b1-kjpathol-48-6-426],[@b2-kjpathol-48-6-426],[@b4-kjpathol-48-6-426],[@b5-kjpathol-48-6-426]\], and most authors report a complete absence of EBV in type II EATL \[[@b2-kjpathol-48-6-426],[@b4-kjpathol-48-6-426],[@b5-kjpathol-48-6-426]\]. Several authors have suggested that diffuse positivity of EBER is an indication that the tumor in question should be diagnosed as extranodal NK/T-cell lymphoma \[[@b4-kjpathol-48-6-426]\]. However, Tse *et al*. \[[@b6-kjpathol-48-6-426]\] reported three cases with extensive EBER expression but without histological features of NK/T-cell lymphoma in a study of 34 type II EATLs in an Asian population. Furthermore, two of the EBER-positive cases in that study had clonal TCR gene rearrangement, implying T-cell lineage. In addition, other authors have also reported EBV-positive EATL ([Table 1](#t1-kjpathol-48-6-426){ref-type="table"}). Although the CD56 status was not specified in that study, except for one case positive for CD56, our case resembles those of Tse *et al*. \[[@b6-kjpathol-48-6-426]\], suggesting that a small minority of type II EATL cases may be EBV-positive. These observations suggest that type II EATL is a heterogeneous disease with distinct phenotypic and genotypic characteristics. No potential conflict of interest relevant to this article was reported. {#f1-kjpathol-48-6-426} {#f2-kjpathol-48-6-426} {#f3-kjpathol-48-6-426} ###### Summary of previously published reports including a patient with EBV-positive EATL Reference Country No. of cases No. of EBV-posi- tive cases EBV-positive cell population ------------------------------------------------------------------ ------------------------------------------------- -------------- ----------------------------- ---------------------------------------- Tse *et al.* (2012) \[[@b6-kjpathol-48-6-426]\] Multinational (Asia) 34 3 Cytotoxic T cells Ng (2012) \[[@b7-kjpathol-48-6-426]\] Singapore 1 1 CD3+, CD4--, CD8--, CD56-- tumor cells Delabie *et al.* (2011) \[[@b1-kjpathol-48-6-426]\] Multinational (North America, Europe, and Asia) 58 4 Subpopulation of likely reactive cells Takeshita *et al.* (2011) \[[@b8-kjpathol-48-6-426]\] Japan 24 2 CD56--, CD8-- T cells 1 CD56+, CD8+ T cells Quintanilla-Martinez *et al.* (1997) \[[@b9-kjpathol-48-6-426]\] Multinational (Mexico, Austria, and Germany) 6 1 Pleomorphic medium and large cells Pan *et al.* (1993) \[[@b10-kjpathol-48-6-426]\] UK 11 3 Large CD3+ tumor cells EBV, Epstein-Barr virus; EATL, enteropathy-associated T-cell lymphoma.

This paper is part of the Special Issue: *Gender and Health Inequality - intersections with other relevant axes of oppression.* More papers from this issue can be found at [www.globalhealthaction.net](http://www.globalhealthaction.net/index.php/gha/issue/view/1717#Special%20Issue:%20Gender%20and%20Health%20Inequality%20-%20intersections%20with%20other%20relevant%20axes%20of%20oppression) Introduction {#S0001} ============ Social inequalities in health are a pressing concern for the Swedish society, and socioeconomically disadvantaged women seem to fare the worst when it comes to health ([@CIT0001]). Research on health inequalities has, however, been criticized for failing to capture such intersecting inequalities, and for concentrating on demonstrating the existence of health inequalities rather than studying the underlying processes ([@CIT0002]). This study uses an intersectional approach as the basis for decomposing intersecting gender and economic inequalities in mental health in northern Sweden. The concept of intersectionality, developed by the legal scholar Kimberlé Crenshaw ([@CIT0003], [@CIT0004]), has gained increased attention within population health research as a theoretical framework of multiple intertwined axes of inequality ([@CIT0002], [@CIT0005]--[@CIT0007]). The adoption of intersectionality by population health research has not been frictionless, however, and a particular issue of contention has been how to translate intersectionality theory into quantitative methodology ([@CIT0002], [@CIT0005], [@CIT0008]). It has further been argued that an intersectional approach for population health should distinguish between *social positions* -- which comprise potentials for privilege, oppression, and marginalization on the one hand -- and *social processes*, which (re)produce or counteract health inequalities across social positions on the other; and ideally both positions and processes should be examined ([@CIT0002]). The reflection of intersectional positions in population patterns of health can here be understood as a process of embodiment ([@CIT0009], [@CIT0010]), whereby social inequalities, through pathways of embodiment, become expressed in individual bodies and thereby create health inequalities. Intersectionality research within population health has so far mostly focused on the overall pattern of health outcomes across contrasting intersectional positions ([@CIT0011], [@CIT0012]) or emphasized the 'extreme groups' of multiple disadvantage or advantage, for example, high-income men or low-income women ([@CIT0005], [@CIT0013]). A range of measures capturing the consequences of intersectional positions have also been formulated, for example, see the review by Jackson et al. ([@CIT0014]). Less attention has been paid to the health consequences of people belonging to mixed locations ([@CIT0002], [@CIT0008], [@CIT0015]), that is, people who are occupying positions of mixed advantage or disadvantage, for example, financially well-off women or low-income men. As argued by Sen and Iyer ([@CIT0015]), such intersectional 'middle groups' differ from the 'extreme groups' in ways which may be particularly illuminating to understand the processes of intersectionality. They, furthermore, describe how multiple middle groups can be understood as comprising *dominant* and *subordinate* middle groups ([@CIT0015]). Here, dominant middle groups are those that have a structural advantage along one axis relative to a subordinate middle group, such as middle-income men having a structural gender advantage over middle-income women, as well as an economic advantage over low-income men ([@CIT0015]). Middle groups are structurally adjacent, and either dominant or subordinate, to at least one other middle group, and they may therefore be involved in the direct struggle for resources and entitlements in different arenas, such as at home, in the public space, and at work. In contrast to the extreme groups, holding a mixed position of simultaneous advantage and disadvantage also enables *leveraging* of structural advantages along one axis to counteract the disadvantages along another axis, in order to gain a secondary advantage. Moreover, although any social and health inequalities between intersectional middle groups are expected to be smaller than between extreme groups, the middle groups commonly represent a considerable portion of the population. Sen et al. ([@CIT0015], [@CIT0016]) have proposed a methodological approach to highlight intersectional middle groups, and have illustrated how leveraging is used between economic and gender middle groups to secure entitlements to health care in India ([@CIT0015]). In this study, leveraging is understood as processes whereby middle groups can gain relative advantages in terms of mental health, and leveraging can therefore be construed as a pathway of embodiment. The specific arrangements of a society\'s welfare systems, health system, and gender order influence the possibilities of effective leveraging. Sweden enjoys a comparatively gender and economic equitable society ([@CIT0017]), with universal health care, well-developed social protections systems, and progressive gender equality policies, for example, with respect to parental leave ([@CIT0018]). These contextual characteristics of Sweden would be expected to translate into relatively small gender and economic inequalities in health, and little room for intersectional middle groups to leverage a relative advantage to counteract a relative disadvantage. Nevertheless, processes of oppression, marginalization, and discrimination are still operating in daily life across multiple spheres in Sweden. For example, despite the social welfare systems in place, economical resources are still strongly shaping the mental health of Swedish women and men ([@CIT0019]). Moreover, despite high labor market participation for women in Sweden, the gendered division of labor is expressed in women taking an undue share of unpaid domestic work ([@CIT0020]). Both women and those of socioeconomic disadvantage may also be engaged in jobs characterized by a poorer psychosocial work environment ([@CIT0021]) and less secure contracts ([@CIT0022]), compared with men and those of socioeconomic advantage. Exposure to gender violence is another extreme expression of gender inequalities that can impact mental health ([@CIT0023]), and other forms of degrading or humiliating treatment in everyday life have also shown to be important for health and healthcare seeking in both Swedish women and men, but particularly among those socioeconomically disadvantaged ([@CIT0024], [@CIT0025]). Further processes of marginalization involve discrimination in the health system based on gender and/or socioeconomic disadvantage ([@CIT0025], [@CIT0026]), or refraining from seeking necessary health care due to shortage of funds ([@CIT0027]). As such, there are multiple processes which may be relevant for leveraging between middle groups of economic and gender intersections in Sweden, and which thereby can act as pathways of embodiment upholding mental health inequalities. This article aims to employ the approach suggested by Sen and Iyer ([@CIT0015]) by 1) examining mental health inequalities between intersectional groups reflecting structural *positions* of gender and economic affluence; and to develop the approach by 2) decomposing any observed health inequalities between dominant and subordinate middle groups to experiences and conditions representing processes of privilege and oppression, using Blinder-Oaxaca decomposition analysis. Methods {#S0002} ======= Study population and procedures {#S0002-S20001} ------------------------------- The study population comprised the participants of the cross-sectional population-based survey 'Health on Equal Terms' carried out in spring 2014 by the County Councils of the four northernmost counties of Sweden: Norrbotten, Västerbotten, Jämtland/Härjedalen, and Västernorrland. The target population included all residents in the four counties aged 16--84 years. The sample frame consisted of 704,099 individuals of the target population, identified through the Total Population Register of Statistics Sweden on November 30, 2013. Sampling was done in two steps. First, as part of the national 'Health on Equal Terms' survey, a small national sample (*N*=1,789; 3.4%) was randomly selected without stratification. Second, as part of the regional expanded sample, a larger regional random sample (*N*=50,300; 96.6%) stratified into 276 strata by county, municipality, gender, and age was selected ([@CIT0028], [@CIT0029]). The overall participation rate was 49%, resulting in a sample size of *N*=25,667. No additional inclusion or exclusion criteria were applied to the study, and all participants with valid responses on gender and income were eligible for inclusion in the analysis. In total, 25,585 individuals were included: 14,988 of which belonged to middle groups (see Measures, Exposure: intersectional positions by gender and income, below) and therefore included for the multiple analyses. Due to item non-response, effective *N* was 24,580--25,585 in descriptive/bivariate analyses on the total sample, and *N*=13,385 in multiple analyses on the subsample comprising the middle groups. The survey was implemented through postal questionnaire covering areas of health and well-being, drug use and health care contacts, health behaviors, and working and social conditions. In addition, sociodemographic individual-level data, such as annual income (for 2012) and country of birth, were retrieved from the Total Population Register of Statistics Sweden and linked to the survey data through the unique Swedish Personal Identity Number. The use of the 'Health on Equal Terms' survey in this study was reviewed and approved by the Regional Ethical Review Board in Umeå (approval no. 2015/134--31Ö). Measures {#S0002-S20002} -------- ### Outcome: mental health Mental health symptoms were self-reported through the General Health Questionnaire (GHQ)-12 ([@CIT0030]), a commonly used screening instrument for mental health. The GHQ-12 has displayed good psychometric properties in the Swedish setting, including good internal consistency and excellent validity when it comes to detecting depressive disorders ([@CIT0031]). The GHQ-12 consists of 12 items covering symptoms during the previous weeks, including lack of concentration, sleeplessness, and moodiness. The items were coded on a four-level Likert scale ([@CIT0001]--[@CIT0004]). The Likert scores were averaged across the 12 items and multiplied by 12 in order to construct a summative index while avoiding bias due to item non-response. The final index thus had a theoretical range of 36, and sample Cronbach\'s alpha was 0.89, which is similar to the Swedish validation study ([@CIT0031]). ### Exposure: intersectional positions by gender and income Gender and income were based on information from population registers of Statistics Sweden. Gender comprised the categories woman and man. Income was measured as annual disposable income for the respondent, which was divided into tertiles (cut-offs at 144,718 and 232,065 SEK) in order to form three equally sized categories of economic affluence: low, medium, and high. In accordance with the procedure by Sen and Iyer ([@CIT0015]), gender and affluence were combined to form six mutually exclusive intersectional positions or groups, including the *extreme groups* of low-income women and high-income men; the *dominant middle groups* of high-income women and medium-income men; and the *subordinate middle groups* of medium-income women and low-income men. ### Explanatory factors: processes of privilege and oppression Explanatory variables were selected to capture processes of privilege, oppression, or marginalization relevant for economic affluence and gender, and which potentially could be used as leverage points to gain mental health advantages. An overview of all variables is shown in [Table 1](#T0001){ref-type="table"}. ###### Descriptive statistics of all variables in the total sample and by intersectional positions of gender and affluence. Women Men ------------------------------- ------------ ------------ ------------ ------------ ------------ ------------ ------------ Depressive symptoms, *M* (SD) 21.4 (4.7) 22.2 (5.1) 21.6 (4.9) 21.2 (4.6) 21.6 (4.9) 21.2 (4.3) 20.4 (3.8) Demographic factors  Age   Young adult: 16--35 years 22.9 37.1 20.7 10.3 41.5 15.0 12.5   Middle age: 36--64 years 44.3 17.4 51.4 76.1 20.5 33.0 66.9   Old age: 66--85 years 32.9 45.5 27.9 13.6 38.0 52.0 20.6  Born outside Sweden 6.3 8.9 6.2 5.0 9.7 4.8 3.6  Education   Low 50.1 66.6 43.8 25.2 68.4 55.8 42.0   Medium 33.3 25.8 35.9 35.4 26.6 34.6 39.6   High 16.6 7.6 20.3 39.4 5.0 9.6 18.4 Material conditions  Diff. to make ends meet   Sometimes 4.9 6.0 5.7 3.7 6.9 4.9 2.8   Often 5.9 8.8 6.4 3.9 9.8 4.8 2.2  Low cash margin 16.4 28.5 16.2 8.6 29.2 11.8 4.9  Residential ownership   Resident-owned 74.8 60.6 78.3 86.4 57.2 76.1 88.1   Rental 18.2 26.3 18.5 11.9 23.2 19.4 10.1   Other arrangements 7.0 13.2 3.3 1.8 19.6 4.5 1.8 Job relations  Job dissatisfaction 7.2 5.5 8.3 6.9 7.4 7.7 7.8  Job insecurity 7.7 7.5 9.5 7.6 7.3 6.1 7.4 Violence  Fear of violence 13.2 25.8 18.6 18.1 5.6 3.9 2.5  Threat/violence experience 4.2 4.5 4.3 4.7 6.2 3.5 2.9  Degrading treatment 16.4 20.5 19.9 19.3 16.8 11.6 9.8 Domestic burden  Elderly care 9.7 7.6 10.8 13.7 7.7 8.5 10.1  Child illness 3.7 1.9 5.0 7.4 1.3 1.8 4.6 Healthcare contacts  Unmet medical needs   Inaccessibility 5.0 6.9 4.7 3.3 6.0 4.8 4.1   Negative experiences 5.6 7.4 6.3 4.3 4.9 5.9 4.1   Other reasons 8.7 11.7 8.7 7.5 10.2 8.3 5.9  Unmet dental care needs   Economic reasons 7.3 8.9 7.7 4.7 11.1 8.2 4.2   Other reasons 9.1 11.0 8.1 7.7 10.3 10.1 7.8 Numbers are column percentages within each variable, unless otherwise noted. Material conditions were measured by three variables: *Low cash margin* (whether the respondent would be able to get hold of 15,000 SEK \[approx. 1,600 EUR\] in 1 week) was coded as 'no' (1) and 'yes' (0); *Difficulties to make ends meet* (whether the respondent has had difficulties paying running costs during the past 12 months) was coded as 'no' (0), 'yes, once' (1), and 'yes, multiple times' (2); and *Residential ownership* was coded as owned house or apartment (0), rental apartment (1), and other living arrangements (2). Job relations were measured by two variables: *Job dissatisfaction* ('How well do you enjoy your work tasks?') was coded as satisfied (0) and dissatisfied (1); and *Job insecurity* ('Are you worried about losing your job within the coming year?') was coded as 'no' (0) and 'yes' (1). Violence was measured by three variables: *Fear of violence* ('Do you ever refrain from walking out alone in fear of being assaulted, robbed, or harassed in any other way?') was coded as 'no' (0) or 'yes' (1). *Threat/violence* was based on two items on whether the respondent during the past 12 months had been exposed to physical violence or threat of violence, which were combined into one variable coded as 'no' (0) and 'yes' (1). *Degrading treatment* was based on whether the respondent during the past 12 months had been treated in a way that was perceived as degrading or humiliating, which was coded as 'no' (0) or 'yes' (1). Domestic burden was measured through *Child with illness* (having a child with chronic disease or functional limitation), coded as 'no' (0) and 'yes' (1); and *Elderly care* (whether the respondent takes care of everyday tasks for someone close who is ill or old), coded as 'no' (0) and 'yes' (1). *Healthcare contacts* were measured by five variables concerning unmet medical and dental care needs in the past 3 months, and the reasons for refraining from seeking care. Three variables concerned unmet medical care needs: *Inaccessibility* (including 'too long waiting times', 'difficulties to reach the provider by phone', 'too late appointment', and 'not knowing where to turn'), *Negative experiences*, and *Other reasons*. Two variables concerned unmet dental care needs, due to *Financial reasons* and *Other reasons*. All variables were coded as 'no' (0), including people without perceived healthcare needs and those who sought care when in need, and 'yes' (1), including people who refrained from seeking care when in need, that is, with unmet care needs. ### Demographic background variables Demographic background was measured by three variables: *Age* was coded as 16--35 years (0), 36--65 years (1), and 66--85 years (2); *Country of birth* was coded as Sweden (0) and outside Sweden (1); and *Education* was coded as low (0), medium (1), and high (2). Data analysis {#S0002-S20003} ------------- ### Aim 1 analyses Following the principal approach by Sen et al. ([@CIT0015], [@CIT0016]) but adapted for continuous outcomes, the first aim of illustrating mental health inequalities between intersectional positions was addressed by reporting GHQ-12 means and confidence intervals for all six groups. Absolute mean differences in GHQ-12 score between the groups were subsequently tested through a one-way analysis of variance (ANOVA), with post-hoc tests for all pairwise comparisons using the Games-Howell correction method. Analyses were done using SPSS v23. ### Aim 2 analyses The second aim, seeking to explain any health gaps between dominant and subordinate middle groups displayed in the Aim 1 analyses, was addressed by Blinder-Oaxaca decomposition analysis ([@CIT0032]) using the *oaxaca* command ([@CIT0033]) in Stata v13. The basic idea of Blinder-Oaxaca decomposition is to explain the distribution of the outcome, in this case group difference in mean GHQ-12 score between dominant and subordinate middle groups, by a set of explanatory factors that vary systematically across the groups. The method is based on two linear regression models that are fit separately for each of the groups, and the technique then partitions the health gap between the groups into a fraction attributable to differences in the explanatory factors (the explained part) and to differences in coefficients (the unexplained part; 32). Blinder-Oaxaca decomposition was applied to decompose absolute difference in mean GHQ-12 score for the following three comparisons between dominant versus subordinate middle groups: 1) middle-income men (dominant) versus middle-income women (subordinate); 2) middle-income men (dominant) versus low-income men (subordinate); and 3) high-income women (dominant) versus middle-income women (subordinate). All explanatory factors described above were included to explain the health gap. The total explained and unexplained parts, as well as the independent contribution of each of the explanatory factor, are reported as absolute contributions (i.e. on the same scale as the outcome) and as relative contributions (percentages). Relative contributions are calculated with respect to the absolute health gap for the total explained and unexplained parts, and relative to the absolute explained part for the individual contributions of each explanatory factor. Results {#S0003} ======= Descriptive statistics of explanatory factors across intersectional groups {#S0003-S20001} -------------------------------------------------------------------------- To illustrate the variations in explanatory variables between the different intersectional positions, [Table 1](#T0001){ref-type="table"} shows descriptive statistics for the total sample as well as for the six intersectional positions. As expected, the doubly disadvantaged group of low-income women generally reported the least favorable life conditions, whereas the doubly advantaged group of high-income men reported the most favorable circumstances. The distributions were more mixed for the four middle groups. The most striking finding was that the economic gradient in fear of violence, present in both women and men, was on a level 5--7 times more frequent in women than in men. Similarly, whereas there was an economic gradient in degrading treatment in men, the gradient was much weaker in women, but at a considerably higher level. Mental health inequalities between intersectional groups (Aim 1) {#S0003-S20002} ---------------------------------------------------------------- See [Fig. 1](#F0001){ref-type="fig"} for a display of mean mental health for the six intersectional positions, corresponding to Aim 1. The bar graph shows the intersections ordered by mental health from worst to best, with statistical inference of absolute difference in means from a one-way ANOVA \[F(5, 25563) =88.81, *p*\<0.001\] and *p*-values for all pairwise comparisons in the lower part of the figure. At the extremes are the doubly disadvantaged and advantaged groups of low-income women and high-income men, who reported worst and best health, respectively (mean difference (95% CI)=1.81 (1.57--2.06); *p*\<0.001). The health inequalities between the middle groups were as expected of smaller magnitude but were still significant. The subordinate mid-income women reported worse mental health than the dominant mid-income men (0.42 (0.14--0.71); *p*\<0.001) and high-income women (0.44 (0.13--0.74); *p*\<0.001), and the subordinate low-income men reported worse mental health than the dominant mid-income men (0.46 (0.13--0.78); *p*\<0.001). In contrast, very similar mean health was reported by the two subordinate (0.03 (−0.35−0.28); *p*=1.000) and the two dominant (0.01 (−0.32−0.29); *p*=1.000) middle groups. As such, mental health inequalities were observed across the range of extreme, subordinate, and dominant middle groups of gender and affluence. ![Mental health (mean GHQ-12 score) in intersections by affluence (low, middle, and high income) and gender (woman, man): illustration of means with extreme (white), dominant (dark grey), and subordinate (light grey) groups; and tests of significance between groups (middle groups' comparisons with border). *p*-values are derived from one-way ANOVA \[F(5, 25563)=88.81, *p*\<0.001\] using Games--Howell post-hoc tests for multiple comparisons.](GHA-9-32819-g001){#F0001} Decomposition of mental health gaps between intersectional middle groups (Aim 2) {#S0003-S20003} -------------------------------------------------------------------------------- [Table 2](#T0002){ref-type="table"} displays a summary of Blinder-Oaxaca decompositions of the mental health gap between structurally adjacent dominant and subordinate middle groups, corresponding to Aim 2. The models explained 68.2% of the gap between mid-income women and mid-income men; 75.3% of the low-income men versus mid-income men gap; and merely 33.5% of the mid-income women versus high-income women gap. ###### Summary of Blinder-Oaxaca decomposition analyses of mental health (GHQ-12 score) between intersectional middle groups by affluence and gender ----------------------------------------------------------------------------------------------------------------------------------------------------------------------- Mid-income women (group 1)\ Low-income men (group 1)\ Mid-income women (group 1)\ Mid-income men (group 2) Mid-income men (group 2) High-income women (group 2) ----------------------------- ----------------------------- --------------------------- ----------------------------- -------- ------- ------- -------- ------- -------  GHQ-12 mean (group 1) 21.57 0.000 21.64 0.000 21.57 0.000  GHQ-12 mean (group 2) 21.11 0.000 21.11 0.000 21.16 0.000  Health gap 0.469 0.000 0.534 0.000 0.412 0.000  Explained fraction 0.320 68.2 0.000 0.402 75.3 0.000 0.138 33.5 0.023  Unexplained fraction 0.149 31.8 0.161 0.132 24.7 0.296 0.274 66.5 0.010 Factor contributions  Country of birth −0.002 −0.7 0.488 0.008 1.9 0.384 −0.004 −2.9 0.212  Age −0.074 −23.1 0.004 −0.160 −39.7 0.000 −0.026 −18.7 0.342  Education 0.010 3.0 0.568 0.038 9.4 0.022 −0.073 −52.8 0.003  Diff. make ends meet 0.041 12.7 0.001 0.071 17.6 0.002 0.059 42.6 0.000  Low cash margin 0.023 7.2 0.017 0.148 36.8 0.000 0.037 26.5 0.021  Residential ownership −0.015 −4.7 0.037 0.095 23.6 0.008 0.015 10.9 0.226  Job dissatisfaction 0.011 3.5 0.384 −0.008 −2.0 0.491 0.040 29.0 0.024  Job insecurity 0.027 8.4 0.003 0.017 4.3 0.038 0.018 12.9 0.012  Fear of violence 0.097 30.3 0.000 0.025 6.2 0.023 0.001 0.8 0.778  Threat/violence experience 0.011 3.4 0.098 0.017 4.3 0.058 −0.001 −0.9 0.773  Degrading treatment 0.163 51.0 0.000 0.081 20.2 0.000 0.014 10.5 0.429  Elderly care 0.002 0.6 0.636 0.000 0.0 0.900 −0.011 −8.1 0.045  Child illness 0.022 6.9 0.038 0.002 0.5 0.525 −0.019 −13.6 0.010  Medical: inaccessibility −0.002 −0.7 0.756 0.015 3.9 0.076 0.023 16.6 0.016  Medical: neg. experiences 0.008 2.4 0.276 −0.010 −2.5 0.306 0.022 15.7 0.006  Medical: other 0.008 2.5 0.443 0.039 9.6 0.006 0.020 14.5 0.111  Dental: other −0.004 −1.3 0.294 0.002 0.4 0.499 0.001 0.4 0.653  Dental: economic reasons −0.004 −1.2 0.416 0.023 5.6 0.028 0.023 16.8 0.012 ----------------------------------------------------------------------------------------------------------------------------------------------------------------------- Estimates are absolute (Abs.) and relative (Rel.) contributions and *p*-values. The better mental health among mid-income men compared with mid-income women was mostly explained by experiences of degrading treatment (alone standing for half of the explained portion of the health gap) and fear of violence in public space (30% of the explained gap). These two factors were also much more frequent in women than in men, across the economic spectrum ([Table 1](#T0001){ref-type="table"}). Difficulties to make ends meet, low cash margin, and job insecurity made smaller but still significant (*p*\<0.05) and sizable (7--13%) contributions to the health gap. In contrast, material conditions were the most important in explaining the mental health advantage of middle-income relative to low-income men. Here, low cash margin, residential ownership, and difficulties to make ends meet jointly explained almost 80% of the explained health gap. Again, this is mirrored by the description in [Table 1](#T0001){ref-type="table"}, where low cash margin and frequent difficulties to make ends meet were 2--3 times more common in low-income than middle-income men. Moreover, experience of degrading treatment made a considerable contribution (20%), with fear of violence and unmet medical and dental needs also making smaller contributions (5--10% each). The explained portion of the mental health gap between mid- and high-income women was, as noted above, smaller than for the other comparisons, which was due to education statistically offsetting the health gap in the model, as indicated by a large negative contribution (−53%). This offsetting contribution of education appeared due to high-income women having higher education than middle-income women ([Table 1](#T0001){ref-type="table"}), in combination with high education being associated to worse, not better, mental health among mid- to high-income women (data not shown). The most important factors explaining the gap were more spread across different spheres than in the two comparisons mentioned above. Material conditions, including difficulties to make ends meet and low cash margin together explained 69%; job relations, including dissatisfaction and insecurity explaining 42%; and unmet medical needs due to inaccessibility and previous negative experiences jointly explained 32% of the total explained part of the health gap. However, degrading treatment was of less importance, mirroring the similar frequencies of this experience in high- and middle-income women ([Table 1](#T0001){ref-type="table"}). Discussion {#S0004} ========== The present study from northern Sweden first found patterns of mental health across intersections of gender and affluence marked by both *inequalities* -- with dominant middle groups reporting better mental health than subordinate middle groups -- and *equalities* -- with the two dominant middle groups reporting similar mental health, as did the two subordinate middle groups. Second, the observed mental health inequalities between dominant and subordinate middle groups were explained partly by processes of specific importance for each comparison and partly by processes of more universal importance. There is considerable evidence on mental health differentials between genders and socioeconomic groups ([@CIT0034], [@CIT0035]), including in Sweden ([@CIT0019], [@CIT0036], [@CIT0037]). In this study, the overall population pattern of mental health across dominant and subordinate middle groups highlights the value of considering middle groups in intersectionality research. For example, despite the structural distinctiveness of the positions of high-income women versus middle-income men and low-income men versus mid-income women, these pairs reported close to identical mental health. These similarities in mental health and the overall small mental health inequalities found in this study should be interpreted in the context of the welfare systems of Sweden, which are expected to reduce health inequalities and also the possibilities for leveraging between groups. Previous research has attributed gender inequalities in health to, for example, working conditions ([@CIT0021]), domestic violence ([@CIT0023]), and also to the unequal distribution of sociodemographic factors ([@CIT0038]). Economic inequalities in health have similarly been attributed to, for example, material factors such as financial strain and employment conditions ([@CIT0036], [@CIT0039], [@CIT0040]). This study illustrates how social inequalities underlie the health inequalities between intersectional middle groups. Although the design and analysis do not allow for causal inference, these findings may reflect processes of leveraging, and how gender and economic intersections become embodied, and health inequalities are upheld. As such, the findings could be viewed as suggesting that middle-income men are able to successfully leverage their structural advantages into a mental health advantage relative to both middle-income women and low-income men, but by different processes. The gender advantage indicated by feeling safe in the public space and infrequent experiences of degrading treatment stood out as valuable resources underlying the mental health advantage relative to similarly affluent women. In contrast, economic advantages as indicated by being able to pay for running costs, financial security, and privileged residential conditions seemed to be valuable points of leverage to gain a mental health advantage relative to low-income men. In addition, degrading treatment and making ends meet emerged as important factors to explain the health inequalities among all middle group comparisons. These factors have been emphasized as critical social determinants of health in Swedish women and men ([@CIT0019], [@CIT0024]), which can reflect their usefulness for both gender and economic leveraging. For the inequality between mid- and high-income women, the health gap was slightly smaller than the economic comparison between middle- and low-income men, and was not explained as well to the same degree as the other comparisons. These findings bear some resemblance with the findings by Griffin ([@CIT0041]) from the United Kingdom, where GHQ scores were not different across social grades in women, but in men. Similar to middle-income men, the economic advantage of high-income women was reflected in material advantages, as well as in better work conditions and access to health care. However, high-income women experienced fear of violence and degrading treatment as frequently as did their middle-income counterparts, and these factors also made less contributions to the mental health gap than for the comparison of middle- and low-income men. This suggests that in contrast to men, economic advantage may not be a very effective leverage point to gain a secure life for women, due to the pervasive power of gender disadvantage. Methodological considerations {#S0004-S20001} ----------------------------- The methodological strengths of the study include a large population-based sample, with a well-validated outcome measure, and utilization of novel statistical approaches. However, the cross-sectional nature of the data precludes any causal inferences. Since only half of those invited participated in the survey, selection bias might have been introduced into the sample. For example, people with severe mental disorders such as major depression could be expected to be under-represented in the sample. This would lead to biased estimates of, for example, mean population mental health, but unless the non-participants simultaneously differ from the sample with respect to, for example, the gender or income, the point estimates of the main analyses would not be expected to be severely biased. Nevertheless, the impact of selection bias is ultimately unknown. The mental health outcome, GHQ-12, has performed well in the Swedish validation study ([@CIT0031]), but is still a screening instrument which potentially can involve inaccurate responses. The use of self-reports can also introduce common-method bias, although in this study at least the main attributes of exposure (gender and income) were measured through independent registers and not self-reported. Moreover, social categories are an issue of contention within intersectionality research, and even pragmatic provisional use of conventional social categories ([@CIT0008]) has received criticism ([@CIT0042]). Moreover, the sole focus on the affluence--gender intersection naturally disguises within category heterogeneity along, for example, ethnicity or sexuality. Additional factors not covered by the questionnaire could also be of interest, for example, gender equality at home ([@CIT0020], [@CIT0043]), job strain ([@CIT0037]), and discrimination ([@CIT0044]). Labor market position was an additional factor which we preliminarily included in the analysis but was highly collinear with age due to the age-diverse sample. As such, the estimate of age can be viewed as capturing both biological age, as well as the age-related labor market circumstances. Conclusions {#S0005} =========== This study gives some indications as to how dominant middle groups in the intersectional space of economic affluence and gender can leverage strategic resources tied to their structural position, in order to gain a mental health advantage relative to subordinate middle groups. From a population health perspective, this highlights how complex pathways of embodiment of entangled gender and economic inequalities may contribute to population patterns of mental health. Future intersectional research should pay attention to intersectional middle groups, and the complex processes of leveraging underlying health inequalities between them. Offering a safe public space, combatting discrimination, and improving financial security are three areas which stand out as promising for policy and prevention seeking to improve gender and economic equity in mental health. This work was supported by Forte -- Swedish Research Council for Health, Working Life and Welfare \[grant numbers 2014-2725, 2014-0451\]. Authors\' contributions {#S0006} ======================= PEG analyzed and wrote the first draft, with guidance from PM. PM and MSS revised the manuscript, and all have agreed to the final version. Conflict of interest and funding ================================ The authors report no conflict of interest. The research was supported by Forte -- Swedish Research Council for Health, Working Life and Welfare. Paper context {#S0007} ============= Intersectional population health research has focused mostly on multiple advantaged or disadvantaged groups, but less on the middle groups of simultaneous advantage and disadvantage. This northern Swedish study of mental health inequalities between middle groups at intersections of gender and income indicates how the inequalities are explained by partly general and partly unique circumstances. Policies aiming toward gender and economic equity in mental health ought to consider the complex processes between intersectional middle groups.

Introduction ============ Among many applications of biodegradable polymers, one of the important applications is in the area of sustained and controlled drug delivery, due to low cytotoxicity and higher tissue compatibility ([@B1], [@B2]). Proteins, being a versatile class of biopolymers, are widely studied for their application in nanoparticulate drug delivery systems. Casein, a major milk protein, due to a number of interesting properties, is regarded as a good candidate for drug delivery system. The efficiency of casein as a nanocarrier for drug delivery has been reviewed by Elzoghby *et al*. ([@B3]). In the recent past, casein based nano delivery systems are being developed for the delivery of drug and nutraceuticals varying from nano sized micelles to nanoparticles prepared by polyelectrolyte, copolymerization, or ionic complexion ([@B3]). Also, casein has a tendency to associate with other bioactives, which is applicable to the process of nanoencapsulation ([@B4]). Green tea polyphenols (GTP) are known for its medicinal properties. Various forms of catechins such as epicatechin (EC), epigallocaetchin (EGC), epicatechin-3-gallate (ECG), and epigallocatechin-3-gallate (EGCG) constitute major part of polyphenols. Among all these forms, epigallocatechin-3-gallate (EGCG) constitutes for 50-80% of total catechins ([@B5]). There are several reports on the applications of GTP in the prevention of cancer, cardiovascular diseases, neurodegeneration, and diabetes ([@B6]-[@B10]). GTP is also a strong antioxidant, and has psychotropic effect ([@B11], [@B12]). In spite of these health benefits, GTP has several limitations with regard to bioavailability, stability, and biotransformation ([@B13]-[@B15]). To overcome these limitations, nanoparticulate formulations have been prepared to maximize the health benefits of GTP. The time required for the new delivery device could be reduced if the mechanism of drug release is known ([@B16]). Mathematical modelling of the *in-vitro* release data plays an important role in providing tools to analyze the experimental data in which the formulation and design factors influence the release data ([@B17]). To the best of our knowledge, no reports have been published on encapsulation of GTP into casein nanoparticles. In this study casein nanoparticles have been used to encapsulate GTP. *In-vitro* studies have been conducted to achieve a sustained release and the mechanism of drug release was determined through mathematical modelling. Experimental ============ Green tea polyphenols (GTP) and glutaraldehyde were purchased from Sigma Chemical Co., USA. Casein and other chemicals were obtained from HiMedia, India. *Preparation of casein nanoparticles* Casein nanoparticles were prepared by cross linking the particles with glutaraldehyde using desolvation step. About 200 mg of casein was dissolved in 2 mL of 1 N NaOH. The solution was stirred continuously and 8 mL of ethanol was added dropwise at a constant rate of 1 mL/min which resulted in the formation of casein nanoparticles. The particles formed were stabilized by the addition of 8% glutaraldehyde (1.175 μL/mg casein). The cross linking was performed for about 24 h at room temperature under constant stirring through a magnetic stirrer. For the preparation of GTP loaded casein nanoparticles, 1 to 10 mg/mL GTP was added to the reaction mixture prior to cross linking, followed by the addition of glutaraldehyde. The mixture was centrifuged at 8000 rpm for 20 min at 4 ºC. The pellet contained nanoparticles while the supernatant was used for quantification of unloaded GTP. *Characterization of casein nanoparticles* Mean diameter and surface charge of the nanoparticles were determined through dynamic light scattering and ζ potential analysis. Analysis was carried out using Nanopartica, Nanoparticle analyzer SZ-100. Morphology and polydispersity of the particles were analyzed through AFM (Nano Surf Easy Scan2, Switzerland) and HR SEM (FEI Quanta FEG 200 -- High Resolution Scanning Electron Microscope). Major constituents of tea polyphenols (EGCG, ECG, EGC, EC and (+) Catechin) were docked with the target protein, casein to confirm its potential of binding. The docking analysis was carried out using Auto dock tools (ADT) v1.5.4 and AutoDock v4.2 programs ([@B18]). The search was extended over the whole receptor protein used as blind docking. *Determination of encapsulation efficiency* *(EE)* Estimation of GTP in the supernatant was carried out by Folin-Ciocalteu assay as described by Swain and Hillis ([@B19]). Briefly, 0.5 mL of sample was mixed with 0.5 mL Follin - Ciocalteu reagent (1 M) and 0.5 mL sodium carbonate (35%) and incubated in dark for 30 min. Absorbance was recorded at 700 nm in a spectrophotometer. The concentration of GTP was calculated from standard curve of GTP. All experiments were performed in triplicates. Encapsulation efficiency (EE) was calculated using the formula: EE (%) = {(Amount of GTP - Amount of free GTP)/Total amount of GTP} × 100 *Drug release profile of GTP loaded casein nanoparticles* About 1 mL of GTP loaded nanoparticles were placed in a tube containing 10 mL phosphate buffer saline (PBS), pH 7.4 at 37 °C. At regular intervals (1, 2, 3, 4, 5, 6, 7, 8, 12, 24 and 48 h), one mL of sample from the tube was withdrawn and equal volume of fresh PBS was added. The samples were centrifuged and the supernatant was used to determine the amount of GTP released. The estimation of GTP was carried out by the method of Swain and Hillis ([@B19]). The drug release studies were carried out for different concentrations of GTP (1, 2.5, 5 and 10 mg/mL), at different pH conditions (pH 3, 7.4 and 9) and at different temperatures (25, 37 and 45 ºC). The drug release kinetics and the mechanism of release were determined for each profile. *Mathematical modelling and release kinetics* To study the mechanism of drug release, the *in-vitro* release data was fitted in various mathematical models like Zero order, First order, Higuchi, Hixson--Crowell and Korsemeyer- Peppas models ([@B20]). The correlation coefficient (R^2^) was used as an indicator for the best fit for each of the models and release exponent (n) to determine the mechanism of release ([Table 1](#T1){ref-type="table"}). *Antioxidant assay* Radical scavenging activity (antioxidant activity) of released GTP was determined *in-vitro* by 1, 1-diphenyl-2-picrylhydrazyl (DPPH) assay following the method reported by Williams *et al*. ([@B21]). DPPH was prepared in absolute ethanol at a concentration of 0.1 mM. About 100 μL sample was added to 550 μL PBS (pH 7.4) and 0.5 mL of DPPH. The reaction mixture was incubated in dark at room temperature for about 30 min and the absorbance was recorded at 517 nm in UV-visible spectrophotometer against control solution. The experiment was repeated three times. *Statistical analysis* All data analysed are presented as mean ± SD. The means were separated by Duncan's multiple range test (DMRT) at *p ≤*0.05. Results ======= Casein nanoparticles were prepared through desolvation technique. The casein nanoparticles were incubated in aqueous solution of GTP ranging from 1 to 10 mg/mL to determine the optimum loading concentration of GTP. AFM and HR SEM images showed spherical nanoparticles with sizes of 63 and 97 nm respectively ([Figure 1](#F1){ref-type="fig"}). Average size of the synthesised casein nanoparticles was observed to be 275.4 nm through particle size analyser. Surface charge of the particles was determined for both unloaded and GTP loaded casein nanoparticles. Unloaded nanoparticles were negatively charged with an electric potential of -11.7 mV. Loading of GTP resulted significant increase in the charge of the particles and was observed to be -29.7, -24.5, -24.2 and -22.9 mV when GTP was loaded at a concentration of 1, 2.5, 5 and 10 mg/mL respectively ([Table 2](#T2){ref-type="table"}). The interactions between the protein and the ligands were determined through molecular docking studies. The docked orientation of ligands with the target protein is shown in [Figure 2](#F2){ref-type="fig"} and the hydrophobic interactions and the docked amino acids are presented in [Table 3](#T3){ref-type="table"}. Highest interactions were observed with casein and ECG and the least interactions between casein and EGCG with no hydrogen bonds formed and a much weaker bond was observed with a bonding energy of -1.67 kcal/mol. ###### Equations for various kinetic models and their parameters -------------------------------------------------------------------------------------------------------------------------------------- **Model** **Equation** **Parameters definition** ------------------------ --------------------------- --------------------------------------------------------------------------------- Zero order C = kot C: Concentration of drug ko: Zero order constant\ t: Time First order logCo - logCt = k1t/2.303 Co: Initial concentration Ct: Concentration at time t k1: First order constant\ t: Time Higuchi Model Q = k t1/2\ Q: Amount of drug released in time t: Time\ H kH: Higuchi constant t: Time Hixson Crowell Model Q 1/3 - Q 1/3 = k t\ Qo: Initial amount of drug\ o t HC Qt: Remaining amount of rug at time t kHC: Hixson Crowell constant\ t: Time Korsmeyer-Peppas Model M - M = ktn\ Mt - Mα: Fraction of drug released at time t k: Korsmeyer-Peppas constant\ t α t: Time\ n: Release exponent -------------------------------------------------------------------------------------------------------------------------------------- ###### Hydrodynamic diameter, polydispersity index (PDI), ζ potential and encapsulation efficiency (%) (EE), of GTP loaded casein nanoparticles **Conc. of GTP (mg/mL)** **Hydrodynamic diameter (nm)** **PDI** **ζ potential (mV)** **EE (%)** -------------------------- -------------------------------- --------------- ---------------------- ----------------- 1.0 77.3 ± 15.65d 0.532 ± 0.15b -29.7 ± 1.23a 65.409 ± 2.151c 2.5 313.4 ± 9.87b 0.178 ± 0.17d -24.5 ± 0.27b 71.101 ± 0.133b 5.0 275.4 ± 9.15c 0.276 ± 0.20c -24.2 ± 0.52b 76.945 ± 2.935a 10.0 428.1 ± 13.79a 2.591 ± 0.35a -22.9 ± 1.02b 70.261 ± 0.953b Data represented as mean ± SD (n = 3). Different letters (a, b, c, and d) indicate significant difference within the group according to DMRT, *p*≤ 0.05. ###### Hydrophobic interactions and docked amino acid residues of albumin with EC, ECG, EGC, EGCG and (+)- catechin ------------------------------------------------------------------------------------------------------------------------------------------------------------------- **Ligand** **Protein** **Binding Energy** **Ligand efficiency** **Intermole energy** **Ligand atoms** **Docked amino**\ **Acid residue(bond length)** ------------------ ------------- -------------------- ----------------------- ---------------------- ------------------ ------------------------------------------- EC Casein -3.25 -0.1 -4.79 C-7' OH A-Met 115/HN (2.9 A⁰) ECG Casein -3.44 -0.11 -5.15 D-6' OH\ A-Pro 116/O (2.6, 2.2 A⁰) D-7' OH EGC Casein -3.39 -0.15 -4.25 A-3'OH\ A-Met 115/O (1.8 A⁰) A-Pro 116/O (1.7 A⁰) B-4' OH EGCG Casein -1.67 -0.05 -3.88 \- A-Ser 120/HN **(+)-**Catechin Casein -3.3 -0.16 -4.38 B-2' OH\ A-Pro 116/O (2.0, 2.1 A⁰) B-3' OH ------------------------------------------------------------------------------------------------------------------------------------------------------------------- ###### GTP release profiles with various models at different loading concentrations of GTP and at different conditions **Condition** **Zero order** **First order** Higuchi model Hixson- Crowell model **Korsmeyer- Peppas model** **Drug transport mechanism** --------------- ---------------- ----------------- --------------- ----------------------- ----------------------------- ------------------------------ -------- ------- -------- ------- ----------- 1 mg/mL GTP 0.9901 0.501 0.7792 0.019 0.8106 5.49 0.898 0.002 0.859 0.641 Anomalous 2.5 mg/mL GTP 0.9807 0.892 0.9879 0.023 0.7529 9.78 0.9922 0.002 0.8395 0.438 Fickian 5 mg/mL GTP 0.9964 1.321 0.9829 0.015 0.7699 17.7 0.9936 0.053 0.8548 0.548 Anomalous 10 mg/mL GTP 0.9970 1.717 0.9577 0.011 0.7868 26.1 0.9842 0.001 0.8613 0.573 Anomalous pH 3 0.9794 0.730 0.9537 0.018 0.8432 9.80 0.9649 0.059 0.8483 0.674 Anomalous pH 7.4 0.9964 1.321 0.9829 0.015 0.7699 17.7 0.9936 0.053 0.8548 0.548 Anomalous pH 9 0.9956 1.52 0.9534 0.014 0.7912 20.3 0.9779 0.051 0.8616 0.636 Anomalous RT 0.9912 1.21 0.9407 0.010 0.8385 18.8 0.9634 0.038 0.8844 0.647 Anomalous 37 °C 0.9964 1.321 0.9829 0.015 0.7699 17.7 0.9936 0.053 0.8548 0.548 Anomalous 45 °C 0.9947 1.286 0.8589 0.010 0.8208 19.9 0.9320 0.037 0.8695 0.711 Anomalous {#F1} {#F2} {#F3} {#F4} Encapsulation efficiency (EE) of GTP loaded casein nanoparticles was determined for GTP concentrations ranging from 1 to 10 mg/mL and is presented in [Table 2](#T2){ref-type="table"}. Highest EE (76.9%) was observed when GTP was used at a concentration of 5 mg/mL. Therefore, all other experiments were carried out using 5 mg/mL GTP. *In-vitro* release of GTP from casein nanoparticles was studied on loading different concentrations of GTP (1, 2.5, 5 and 10 mg/mL) to casein nanoparticles, at different pH (pH 3, 7.4 and 9) and temperatures (25, 37 and 45 ºC) ([Figure 3](#F3){ref-type="fig"}). When 1 mg/mL GTP was used for loading, about 63% of the drug was released in 8 h and a maximum of 85% was released in 48 h. On increasing the GTP concentration to 2.5 and 5 mg/mL, cumulative percentage of drug release was observed to be 75 and 72% respectively in 48 h. On further increasing the concentration of GTP to 10 mg/mL, about 82% release in 48 h was observed. Release was also influenced by pH of release medium. When pH was shifted from acidic to basic, there was an increase in the cumulative percentage of drug release. At pH3 about 57% of the drug was released in 48 h while at pH 7.4, about 72% and at pH 9 about 82% of the drug was released in 48 h. No significant difference was observed when release was carried out at 25 and 37 ºC, where about 68 and 73% was released in 48 h respectively but at 45 ºC about 90% of GTP was released in 48 h. *In-vitro* release data obtained for all the parameters were fitted into various kinetic models to evaluate the mechanism of drug release. The regression parameters *i.*e., regression coefficients (R^2^) and release kinetic constants (k) are presented in [Table 4](#T4){ref-type="table"} for different parameters. It was observed that all the release data showed good correlation with Zero order kinetics, except for that when GTP was used at a concentration of 2.5 mg/mL, which followed Hixson Crowell model of kinetics. Predominance in anomalous mode of release mechanism was observed for all the batches, except when GTP was used at a concentration of 2.5 mg/mL. The release exponent was observed to be 0.438, indicating a Fickian mode of drug release. Radical scavenging activity of the released GTP samples were carried out through DPPH assay and the percentage inhibition was calculated. Free radical scavenging activity of the released GTP samples was observed to be about 88% in 4 h when GTP was used at a concentration of 5 mg/mL for encapsulation in casein nanoparticles ([Figure 4](#F4){ref-type="fig"}). Discussion ========== Several studies have been published regarding the applications of casein based nano delivery systems. Livney reviewed the applications of milk proteins for their use as nanovehicles for the delivery of bioactives and the achievements of using casein nanovehicles for drug delivery systems in the recent past ([@B22]). All the studies showed promising results for using casein as a nanovehicle for the delivery of bioactive compounds. In our study, casein nanoparticles were prepared through desolvation technique using glutaraldehyde as a cross linker. GTP was used as a bioactive compound due to its several health benefits. The synthesised particles were characterized for their size and the surface charge ([Table 2](#T2){ref-type="table"}). Size of the particles plays a major role in delivering the drugs to the target site. In our study, larger size of the particles in DLS could be due to the aggregation of the particles in the dispersion medium, which was much higher for polyphenol loaded chitosan nanoparticles (average particle diameter being \> 400 nm) and GTP loaded PLA-PEG nanoparticles (average particle diameter being \> 300 nm) ([@B24], [@B25]). As observed in [Table 2](#T2){ref-type="table"}, the polydispersity index was shown to be the maximum at a GTP concentration of 10 mg/mL, which also might be due to aggregation of the particles as the concentration of GTP increased. ζ potential measurements were carried out for casein nanoparticles, before and after loading GTP and a difference in the surface charge was observed. As the concentration of GTP in the nanoparticles increased, a decrease in the ζ potential value was observed. This difference in the electric potential could be due to change in the protein conformation to protect the encapsulated GTP ([@B23]). All the GTP loaded nanopaticles showed ζ potential of -22.9 mV and above which would provide good stability, which showed a better result compared to the work carried out by Dube *et al*. (2010) on tea polyphenol loaded chitosan nanoparticles and for GTP loaded PLA-PEG nanoparticles ([@B24], [@B25]). To further confirm the entrapment of GTP in casein nanoparticles, molecular docking studies were carried out to identify the interactions between casein and the individual constituents of GTP. All the ligands showed good interactions with the target molecule and ECG showed highest binding energy ([Figure 2](#F2){ref-type="fig"}, [Table 3](#T3){ref-type="table"}). Maximum EE was observed with 5 mg/mL GTP and on further increasing the concentration to 10 mg/mL, decrease in the EE was observed ([Table 2](#T2){ref-type="table"}). On comparing the EE of GTP to PLA-PEG nanoparticle, GTP loaded casein nanoparticles in our study showed good results with a maximum EE at a comparatively lower concentration of GTP ([@B25]). *In-vitro* release data for different concentrations of GTP showed no particular trend in the release of GTP and maximum amount of drug release was observed when 1 mg/mL GTP was encapsulated to casein nanoparticles ([Figure 3](#F3){ref-type="fig"}). Thus, it could be said that the concentration of GTP entrapped and the cumulative percentage of drug release were independent. Drug release was further monitored at different pH conditions. When a drug is orally administered, it first passes through the gastro intestinal tract, where most of the drug is metabolized due to the harsh pH of the gastric juice. In our study, release of GTP was carried out in acidic conditions to mimic the conditions of the gastric juice. Although 57% of drug was released in acidic pH in 48 h, by the end of 4 h (equal to normal retention time of food in stomach) cumulative percentage of drug release was only 21%. Thus maximum amount of drug is retained within the nanoparticles and could be available for delivery to the site of action. Release studies were further carried out by altering the temperature. As the temperature increased, an increase in the cumulative percentage of drug release was observed which indicates the possible role of temperature in the release of GTP from casein nanoparticles. It has been reported that casein particles have no fixed structures and due to the lack of a rigid three dimensional structure, change in temperature may alter its structure ([@B26]). Thus, the observed increase in the cumulative percentage of drug release from casein nanoparticles at 45 ºC would be due to the alterations in the casein structure at a higher temperature. Mathematical modelling of the released data showed that all release profiles have good correlation with Zero order reaction kinetics, where the concentration of drug release and time are independent ([Table 4](#T4){ref-type="table"}). Ideal model of drug release for nanoparticulate dosage forms or sustained release formulations is the Zero order kinetics ([@B20]). Few of the advantages of Zero order kinetics include a constant release rate as long as the drug is present in the core, no inhibitory build up of drug takes place in the dissolution medium, and a prolonged release is achieved, which is an ultimate goal for a controlled/ sustained release system ([@B17]). Therefore, as our data showed good correlation with Zero order reaction kinetics, casein nanoparticles could be used as an ideal carrier for the delivery of GTP, achieving a prolonged or a sustained release system. According to Korsmeyer-Peppas, a pure Fickian release has a release exponent n value limiting to 0.5, 0.45, and 0.43 for release from slabs, cylinders, and spheres ([@B27]). In our study n values ≤ 0.43 are considered as pure Ficikan release where the release is through diffusion and that ranging from 0.43-0.89 are considered as anomalous transport which involves a coupling of diffusion and erosion mechanism and that greater than 0.89 are referred to as super case II transport. The transport is characterized by polymer relaxation due to polymer erosion when enzymatic degradation occurs ([@B28]). When the relative relaxation time of the polymer is much shorter than the diffusion time, Fickian diffusion occurs. Once solvated, polymers assume an equilibrium state immediately, leading to polymer relaxation, known as case II transport. In many cases, transport mechanism leads to behaviour intermediate to Fickian and case II transport, referred to as anomalous transport ([@B29]). In our study, anomalous mode of drug transport was predominantly observed for all the batches and the mechanism of drug transport/release was observed to be a coupling of diffusion and erosion mechanism, where the rate of drug release is constant over time. *In-vitro*antioxidant studies showed no particular trend in the activity for the released samples ([Figure 4](#F4){ref-type="fig"}). Since GTP is known to be a good antioxidant source, there was a direct correlation between the concentration of GTP released and its antioxidant activity. As the concentration of drug release was maximum at about 4-5 h, proportionally there was an increase in the radical scavenging activity. Since all the released samples showed the scavenging activity, GTP loaded casein nanoparticles would serve as a potent antioxidant source. Conclusion ========== GTP encapsulated casein nanoparticles were prepared and characterized for the first time. *In-vitro*release showed that GTP loaded casein nanoparticles would serve as a good source for oral administration, as the release in acidic pH (simulated gastric juice/ stomach conditions) was much lower than the physiological and basic pH. All the release profiles showed good correlation with Zero order kinetics, which is known as an ideal model of release for nanoparticulate systems, showing anomalous mode of drug transport. Potent antioxidant activity was observed for all the released samples. All these observations indicate that casein nanoparticles could serve as a versatile vehicle for delivery of GTP. The authors would like to thank the management of Vellore Institute of Technology for providing the required facilities to carry out the work.

Jacques C, Jonas J, Maillard L, Colnat‐Coulbois S, Rossion B, Koessler L. Fast periodic visual stimulation to highlight the relationship between human intracerebral recordings and scalp electroencephalography. Hum Brain Mapp. 2020;41:2373--2388. 10.1002/hbm.24952 32237021 **Funding information** Fédération Wallonie‐Bruxelles, Grant/Award Number: ARC 13/18‐053; Fondation Louvain; Fonds National de la Recherche ‐ FNRS, Grant/Award Number: PDR T.0207.16 1. INTRODUCTION {#hbm24952-sec-0001} =============== Since its first report and validation in humans (Adrian & Matthews, [1934](#hbm24952-bib-0001){ref-type="ref"}) scalp electroencephalography (EEG) has been widely used to study dynamic neurofunctional processes and their pathology in large‐scale brain networks (Lopes da Silva, [2013](#hbm24952-bib-0037){ref-type="ref"}; Nunez & Srinivasan, [2005](#hbm24952-bib-0050){ref-type="ref"}; Regan, [1989](#hbm24952-bib-0057){ref-type="ref"}). Given that EEG noninvasively provides information about the unfolding of brain processes at the millisecond time resolution, understanding the relationship between scalp EEG signals and their source(s) at the cortical level is important for fundamental research. It is also of primary importance for clinical studies, in particular for the neurological study of epileptic patients, in order to define and localize brain sources of epileptic seizures (Coito et al., [2019](#hbm24952-bib-0009){ref-type="ref"}; Gavaret, Badier, Marquis, Bartolomei, & Chauvel, [2004](#hbm24952-bib-0017){ref-type="ref"}; Koessler et al., [2010](#hbm24952-bib-0030){ref-type="ref"}). Unfortunately, knowledge about the relationship between scalp EEG signals and their cortical source(s) remains severely limited, for several reasons. First, scalp EEG signals are attenuated by the electrical resistance of head tissues, which remain unknown in human in vivo (especially the skull resistivity) and very difficult to estimate noninvasively and during in vivo measurements (Goncalves et al., [2003](#hbm24952-bib-0020){ref-type="ref"}; Koessler et al., [2017](#hbm24952-bib-0032){ref-type="ref"}; Malmivuo & Suihko, [2004](#hbm24952-bib-0040){ref-type="ref"}). Second, the distance from brain sources to scalp sensors reduce the amplitude of EEG signal, making it difficult to capture and estimate brain sources at the deepest portions of sulci or the medial brain structures (Koessler et al., [2015](#hbm24952-bib-0031){ref-type="ref"}; Seeber, Cantonas, Sesia, Visser‐vandewalle, & Michel, [2019](#hbm24952-bib-0064){ref-type="ref"}; see also Pizzo et al., [2019](#hbm24952-bib-0053){ref-type="ref"} in MEG). Third and perhaps most importantly, many brain sources (often co‐activated in interlocked time‐courses) contribute to EEG recording. Electrical signals generated by these co‐activated sources are mixed when measured on the scalp with EEG sensors, making it difficult to assign a specific source to a specific EEG signal characteristic (nonlinear relationship; Kovach, Oya, & Kawasaki, [2018](#hbm24952-bib-0033){ref-type="ref"}) and requiring to solve an undetermined inverse problem (Grech et al., [2008](#hbm24952-bib-0021){ref-type="ref"}; Kaiboriboon, Luders, Hamaneh, Turnbull, & Lhatoo, [2012](#hbm24952-bib-0028){ref-type="ref"}; Michel et al., [2004](#hbm24952-bib-0044){ref-type="ref"}). In humans, the relationship between scalp EEG signals and their cortical sources can be potentially addressed by comparing simultaneous scalp and intracranial EEG recordings. These rare studies derive from invasive investigations performed mainly with foramen ovale or subdural electrodes (e.g., Alarcon et al., [1994](#hbm24952-bib-0002){ref-type="ref"}; Lantz, Holub, Ryding, & Rosen, [1996](#hbm24952-bib-0034){ref-type="ref"}; Van Der Loo, Congedo, Plazier, Van De Heyning, & De Ridder, [2007](#hbm24952-bib-0069){ref-type="ref"}; Wennberg & Cheyne, [2014](#hbm24952-bib-0071){ref-type="ref"}) and more rarely with intracerebral electrodes (Alarcon et al., [1994](#hbm24952-bib-0002){ref-type="ref"}; Gavaret, Dubarry, Carron, & Bartolomei, [2016](#hbm24952-bib-0018){ref-type="ref"}; Koessler et al., [2015](#hbm24952-bib-0031){ref-type="ref"}). In the latter case, intracerebral EEG (also named stereoelectroencephalography, SEEG) records electrical activity directly from specific brain areas with high anatomical accuracy by means of implanted multicontact electrodes. This technique is used in the presurgical evaluation of drug‐resistant epilepsies (Chauvel, Gonzalez‐Martinez, & Bulacio, [2019](#hbm24952-bib-0007){ref-type="ref"}; Talairach & Bancaud, [1973](#hbm24952-bib-0067){ref-type="ref"}). Beyond its clinical significance, the recorded signal can be analyzed in parallel for research purposes (e.g., Allison, Puce, Spencer, McCarthy, & Belger, [1999](#hbm24952-bib-0003){ref-type="ref"}; Barbeau et al., [2008](#hbm24952-bib-0005){ref-type="ref"}; Halgren et al., [1994](#hbm24952-bib-0022){ref-type="ref"}; Jacques et al., [2016](#hbm24952-bib-0024){ref-type="ref"}; Jonas et al., [2016](#hbm24952-bib-0026){ref-type="ref"}). The few studies which investigated the relationship between simultaneously recorded scalp EEG and SEEG signals relied on event‐related potential (ERP) approaches, in which brain activity is recorded to the sudden occurrence of an event (external or internal such as epileptic spikes) and then averaged in the time domain (Dubarry et al., [2014](#hbm24952-bib-0013){ref-type="ref"}; Jacques et al., [2019](#hbm24952-bib-0023){ref-type="ref"}; Koessler et al., [2015](#hbm24952-bib-0031){ref-type="ref"}; Merlet et al., [1998](#hbm24952-bib-0043){ref-type="ref"}; Rosburg et al., [2010](#hbm24952-bib-0058){ref-type="ref"}). At least two main factors make it difficult to perform these studies and therefore seriously limit their availability. First, relating scalp to intracerebral EEG requires a high signal‐to‐noise ratio (SNR) on the scalp, which typically results in long‐duration experiments to collect data from a large number of events (exogenous or endogenous; Luck, [2014](#hbm24952-bib-0039){ref-type="ref"}). SNR is particularly an issue in these clinical settings where recordings can take place over several days without the possibility to fix or replace noisy scalp electrodes. Second, the amplitude and shape of evoked responses in the time domain are difficult to relate across scalp and intracerebral EEG because of the diversity and the unknown spatiotemporal dynamic of activated sources within the brain (Lopes da Silva, [2019](#hbm24952-bib-0038){ref-type="ref"}). One approach to overcome these difficulties would be to use a stimulus presentation technique that provides high SNR and allows to more objectively characterize the signal to be compared across recordings. A potentially powerful approach that fits these criteria is the frequency tagging approach where a stimulus is presented at a (relatively fast) fixed frequency rate, for instance, a flickering light, eliciting a neural response exactly at this frequency rate which can be therefore objectively tracked and quantified in simultaneously recorded scalp EEG and SEEG signals. This approach was discovered shortly after the first descriptions of EEG recordings in humans (Adrian & Matthews, [1934](#hbm24952-bib-0001){ref-type="ref"}), that is, well before the first ERP recordings (Dawson, [1951](#hbm24952-bib-0012){ref-type="ref"}; see Regan, [1972](#hbm24952-bib-0056){ref-type="ref"}). It was already considered at the time as offering a powerful mean to understand the nature and the source(s) of EEG recordings, as stated by Adrian ([1944](#hbm24952-bib-0500){ref-type="ref"}):""All the messages which reach the cortex will produce their own electrical accompaniment, and this can be recorded well enough if electrodes can be placed on the surface of the brain. But if we can get no nearer than the scalp, the potential changes generated in any group of nerve cells will usually be obscured by those of other groups nearby, and the record will then show us nothing... Fortunately this difficulty can be overcome, in part at least, by making all the cells work in unison. This can be done, as far as vision is concerned, by making the field more or less uniform and lighting it with a flickering light. The nerve cells are then forced to work in unison at the frequency of the flicker, and we can record their electrical activity through the skull up to frequencies of about 30 a second. This gives us a method of tracing the visual messages in the brain, for by means of the flicker rhythm they can be made easy to recognize" (Adrian, 1944, p. 361)." Thanks to the subsequent application of Fourier analysis to EEG recordings (Regan, [1966](#hbm24952-bib-0055){ref-type="ref"}), these "frequency‐tagged" neural responses can be investigated in the frequency domain in various sensory modalities in neurotypical adults, but also in developmental and clinical populations (Regan, [1989](#hbm24952-bib-0057){ref-type="ref"}; see Norcia, Appelbaum, Ales, Cottereau, & Rossion, [2015](#hbm24952-bib-0048){ref-type="ref"} for a recent review in vision research). The main advantages of this approach are its objectivity (i.e., responses are identified at a frequency known by the experimenter) and high sensitivity (i.e., high SNR) (Norcia et al., [2015](#hbm24952-bib-0048){ref-type="ref"}; Regan, [1989](#hbm24952-bib-0057){ref-type="ref"}; Rossion, [2014](#hbm24952-bib-0059){ref-type="ref"}). Moreover, by carefully manipulating the nature of the stimulus property that is periodically modulated, sensory processes, but also higher‐level brain processes such as face or word categorization for instance (e.g., Lochy, Van Belle, & Rossion, [2015](#hbm24952-bib-0036){ref-type="ref"}; Rossion & Boremanse, [2011](#hbm24952-bib-0060){ref-type="ref"}), can be selectively tracked in all modalities. Despite these advantages, to our knowledge, the frequency‐tagging approach has never been applied to simultaneous EEG and intracerebral recordings in order to shed light on the relationship between the two types of signals.[1](#hbm24952-note-0002){ref-type="fn"} Here we report simultaneous frequency‐tagged scalp and intracerebral EEG responses in a unique epileptic patient implanted with three intracerebral electrodes (27 recording contacts) in the right occipitotemporal (OT) cortex and equipped simultaneously with 27 scalp electrodes on the scalp surface. Thanks to a fast periodic (6 Hz) visual stimulation with highly salient stimuli (faces), we objectively relate the quantified face‐evoked responses observed inside and outside the brain. Specifically, we address the following questions: (a) How do objectively related signals recorded simultaneously inside the brain and on the scalp differ in terms of amplitude and SNR? and (b) Can frequency‐tagging significantly improve the precise identification of the sources of activity recorded on the scalp? 2. MATERIALS AND METHODS {#hbm24952-sec-0002} ======================== The epileptic patient, as well as the recording settings, are identical to those reported in Jacques et al. ([2019](#hbm24952-bib-0023){ref-type="ref"}). The patient has also been described in Jonas et al. ([2014](#hbm24952-bib-0027){ref-type="ref"}). Therefore, a shortened version of the methods is reported here. 2.1. Case description {#hbm24952-sec-0003} --------------------- KV is a right‐handed female suffering from refractory occipital epilepsy related to a focal cortical dysplasia involving the right lingual gyrus and posterior collateral sulcus. The patient was 32 year‐old at the time of testing. Her case was also reported as evidence of strong face identity repetition suppression effects in the lateral cortex of right IOG using fast periodic visual stimulation (FPVS) with unfamiliar faces (Jonas et al., [2014](#hbm24952-bib-0027){ref-type="ref"}). 2.2. Simultaneous intracerebral---Scalp EEG recordings {#hbm24952-sec-0004} ------------------------------------------------------ The patient underwent simultaneous intracerebral and scalp EEG recordings. The co‐locations of these electrodes are shown in Figure [1](#hbm24952-fig-0001){ref-type="fig"}. The originality of electrode placements is a relatively dense spatial coverage of the occipital‐temporal cortex with both intracerebral (27 intracerebral contacts) and surface electrodes (including sO1, sOz, sO2, sPO7, sPO8, sP9, sP10, sP5, sP6). {#hbm24952-fig-0001} ### 2.2.1. Intracerebral electrodes {#hbm24952-sec-0005} The patient was stereotactically implanted with three intracerebral multicontact electrodes targeting the right ventral OT cortex, according to a well‐defined and previously described procedure (Jonas et al., [2016](#hbm24952-bib-0026){ref-type="ref"}; Salado et al., [2018](#hbm24952-bib-0062){ref-type="ref"}). Each intracerebral electrode consists of a cylinder of 0.8 mm diameter and contains 8--11 independent recording contacts of 2 mm in length separated by 1.5 mm from edge to edge and by 3.5 mm center to center (DIXI Medical, Besançon, France). Electrodes D and L (eight recording contacts each: D1--D8 and L1--L8) sampled the right inferior occipital gyrus and posterior collateral sulcus. Electrode F (11 contacts, F1--F11) was more anterior and went from the right inferior temporal gyrus to the lingual gyrus. All intracerebral contacts except D8 were in direct contact with the gray matter. The recording surface of contact D8 was located \~2 mm from the cortical surface, likely within the meninges (Figure [1](#hbm24952-fig-0001){ref-type="fig"}c,d). ### 2.2.2. Scalp electrodes {#hbm24952-sec-0006} Simultaneous scalp EEG recordings were acquired with 28 Ag/AgCl electrodes of 10 mm diameter placed according to the 10--20 system (Figure [1](#hbm24952-fig-0001){ref-type="fig"}, Seeck et al., [2017](#hbm24952-bib-0065){ref-type="ref"}) using sterile procedures, with a particular spatial coverage of bilateral OT regions. Some of the posterior electrodes were slightly displaced relative to the 10--20 positions due to the presence of depth electrodes. Scalp electrode positions were determined using a 3D digitizer system (3 space Fastrak, Polhemus, Colchester, VT). ### 2.2.3. Recordings {#hbm24952-sec-0007} Simultaneous SEEG---scalp EEG signals were recorded at a 1,024 Hz sampling rate with a 128‐channel amplifier (SD LTM 128 Headbox; Micromed, Italy). The reference electrode was a prefrontal midline scalp electrode (sFPz). The recording of the periodic visual stimulation experiment reported here was performed 2 days after the scalp electrode placement. Due to the low diameter of the skull defect at the penetration points of intracerebral electrodes (1.2 mm) and the low electrical conductivity of the guidance screw (titanium), no leakage of current was observed in scalp EEG recordings. 2.3. FPVS {#hbm24952-sec-0008} --------- ### 2.3.1. Rationale {#hbm24952-sec-0009} The main aspects of the procedure for this experiment have been previously described in three different studies comparing the presentation of trains of different faces to identical faces at a fixed frequency rate (Jonas et al., [2014](#hbm24952-bib-0027){ref-type="ref"}; Rossion & Boremanse, [2011](#hbm24952-bib-0060){ref-type="ref"}; Rossion, Prieto, Boremanse, Kuefner, & Van Belle, [2012](#hbm24952-bib-0061){ref-type="ref"}). From a methodological perspective, this FPVS approach---which leads to so‐called steady‐state visual evoked potentials (SSVEPs, Regan, [1989](#hbm24952-bib-0057){ref-type="ref"}, Regan, [1966](#hbm24952-bib-0055){ref-type="ref"})---has multiple advantages: objectivity of definition and quantification of the response of interest, high SNR, short duration of the experiment, and recording of the response of interest during a simple incidental task (Regan, [1989](#hbm24952-bib-0057){ref-type="ref"}; Rossion, [2014](#hbm24952-bib-0059){ref-type="ref"}), making it a tool of choice for the study of patients implanted with intracerebral electrodes. Here, faces were presented at a 6 Hz rate because this frequency rate provides the largest repetition suppression effect on the scalp over the right OT cortex (Alonso‐Prieto, Van Belle, Liu‐Shuang, Norcia, & Rossion, [2013](#hbm24952-bib-0004){ref-type="ref"}), as well as in face‐selective areas of the right inferior occipital gyrus and middle section of the lateral fusiform gyrus (Gentile & Rossion, [2014](#hbm24952-bib-0019){ref-type="ref"}). ### 2.3.2. Stimuli {#hbm24952-sec-0010} Full‐front color face pictures of 18 unfamiliar individuals (7° × 10° of visual angle for the base face size) equalized for global luminance were used. These face stimuli were the same as used in previous studies (Alonso‐Prieto et al., [2013](#hbm24952-bib-0004){ref-type="ref"}; Rossion & Boremanse, [2011](#hbm24952-bib-0060){ref-type="ref"}) and taken from a well‐known set of laser‐scanned faces from the Tubingen Max Planck Institute (MPI) database of laser‐scanned (Cyberware TM) human heads. They were cropped to remove external features (hair and ears) but their overall shape was preserved. ### 2.3.3. Procedure {#hbm24952-sec-0011} In each condition, a face stimulus appeared and disappeared (sinusoidal contrast modulation) on the screen, at a stimulation rate of six faces per second (one face every 166.66 ms; Figure [2](#hbm24952-fig-0002){ref-type="fig"}). A trigger was sent to the parallel port of the EEG recording computer at each minimal level of visual stimulation (gray background), using a photodiode placed on the left upper corner of a laptop monitor. In the *same face* condition, a randomly selected face picture was presented repeatedly during the whole stimulation duration (70 s). In the *different faces* condition, the sequence started with the repeated presentation of a randomly selected face picture for the first 15 s, after which the face identity changed at every cycle for the remainder of the sequence (i.e., from 16 to 70 s, see Rossion et al., [2012](#hbm24952-bib-0061){ref-type="ref"}). In these *different faces* condition, 18 individual faces of the same sex were used and presented in random order. The same face identity never appeared twice in a row, so that the face identity change rate was always 6 Hz. To minimize repetition suppression effects due to low‐level visual cues, the face stimulus changed substantially in size with each presentation, that is, at a rate of 6 Hz, in all conditions (random face size between 82 and 118% of base face size). The experiment consisted of four sequences of 70 s: each condition (same face or different faces) was repeated two times (face gender: male or female). The order of conditions was randomized. During each 70 s run, the patient was instructed to fixate on a small black cross located centrally on the face, slightly below the bridge of the nose. The fixation cross changed color (black to red) briefly (200 ms) 6 to 8 times during each run and the patient was instructed to report the color changes by pressing a response key. {#hbm24952-fig-0002} 2.4. Data processing and analyses {#hbm24952-sec-0012} --------------------------------- ### 2.4.1. Frequency domain analyses {#hbm24952-sec-0013} All analyses were performed using Letswave 5 (Mouraux & Iannetti, [2008](#hbm24952-bib-0047){ref-type="ref"}) and MATLAB v7.8 (The Mathworks, Inc.). Segments of 50 s of recording during visual stimulation (i.e., 300 face stimulation cycles at 6.0 Hz) from 17 to 67 seconds were considered for analysis. These segments were cropped to contain an exact integer number of 6 Hz cycles. Segments were averaged in the time domain separately for each condition and a Fast Fourier Transform (FFT) was applied to these averaged segments to compute the amplitude and phase spectra at a high spectral resolution of 1/50 = 0.02 Hz. SNR was computed from the amplitude spectra as the ratio between the amplitude at each frequency bin and the average amplitude of the corresponding 20 neighboring bins (up to 11 bins on each side, i.e., 22 bins, but excluding the 2 bins directly adjacent to the bin of interest, i.e., 20 bins, e.g., Rossion et al., [2012](#hbm24952-bib-0061){ref-type="ref"}). Significant responses above noise level at the stimulation frequency at each channel were defined by computing Z‐scores on the amplitude spectra, using the mean and *SD* of the 20 neighboring bins around the frequency of interest (e.g., Liu‐Shuang, Norcia, & Rossion, [2014](#hbm24952-bib-0035){ref-type="ref"}). Statistical comparisons between conditions were similarly made by computing Z‐scores on amplitude spectra obtained by subtracting the spectra measured in the *same face* condition from the spectra measured in the *different faces* condition. ### 2.4.2. Time domain analyses {#hbm24952-sec-0014} For this and further analyses, we only used data from the *different faces* condition which generated the largest responses both in scalp and intracerebral recordings. Each recording sequence from 17 to 67 s relative to sequence onset was divided into epochs of 1 s duration centered on the appearance of a face. Epochs containing blinks were rejected and remaining epochs were averaged and the mean amplitude was centered on zero (dc correction). ### 2.4.3. Correlation between intracerebral and scalp EEG signals {#hbm24952-sec-0015} We examined the relationship between visually‐driven signal recorded at intracerebral contacts relative to scalp electrodes by correlating the variations of the 6 Hz response amplitude over time across all scalp electrodes and intracerebral contacts (Figure [3](#hbm24952-fig-0003){ref-type="fig"}). {#hbm24952-fig-0003} We first applied a Morlet wavelet transform on the raw signal to compute the time‐varying amplitude envelope of the electrophysiological signal around 6 Hz. The parameters of the mother wavelet (central frequency: 6 Hz, full width at half maximum \[FWHM\] in the frequency‐domain = 0.8 Hz; FWHM in the time domain = 0.47 s) were chosen to provide a relatively high frequency resolution while preserving the dynamic in the amplitude variation over time. We kept the amplitude envelope from 15 to 69 s from each recording sequence and divided the envelope from each sequence in 9 segments of 6 s, resulting in 18 segments in total (2 sequences with 9 segments). Then, for each intracerebral contact, the signal in each segment was correlated (Pearson\'s coefficient) with the corresponding segment in each scalp electrode and the correlations were averaged across the 18 segments, resulting in a 27 (intracerebral) × 27 (scalp) correlation matrix. We also computed the across‐segments *SD* of the correlation coefficients. We determined whether correlations were significantly different from zero using a randomization procedure in which, for each electrode, we randomly shuffled (5,000 times) the order of the 6 s segments prior to computing correlations and averaging the correlations across segments. For each intracerebral contact correlated with all scalp electrodes, we determined significance using a cluster‐based correction (cluster‐mass) for multiple comparisons (Maris & Oostenveld, [2007](#hbm24952-bib-0041){ref-type="ref"}; Pernet, Latinus, Nichols, & Rousselet, [2015](#hbm24952-bib-0052){ref-type="ref"}) with a cluster‐forming threshold of *p* \< .05. Note that the wavelet analysis used here removes the phase information so that the correlations across channels are only determined by the variation of amplitude across time at each channel. Disregarding the phase information for this inter‐channel correlation analysis is crucial to avoid correlations being driven by simple phase coherence across channels triggered by a common stimulation. Even if we observe significant scalp‐intracerebral correlations, they could still occur because scalp and intracerebral channels are sensitive to the same electrophysiological responses (i.e., generated by a common cortical territory) unrelated to the FPVS stimulation. To determine whether the pattern of scalp‐intracerebral correlations obtained using the original signal at the stimulation frequency during FPVS (*Ori\@6 Hz*) is specific to the signal with visual stimulation, we compared this correlation pattern with patterns obtained in several control situations: (a) *NotchOri\@6 Hz*: Pattern of scalp‐intracerebral correlations with the visually‐driven signal filtered‐out from the original signal using a narrow notch filter (butterworth notch filter order 4: \[5.9--6.1\] Hz); (b) *Ori\@4 Hz*: Pattern of correlations on the original signal using the amplitude envelope at another frequency than the stimulation frequency (i.e., 4 Hz); (c) *Rest\@6 Hz*: Pattern of correlations obtained in (S)EEG recording where no periodic visual stimulation was presented and the patient was resting with eyes open. In this third control situation (rest), sections of the amplitude envelope corresponding to eye blinks were removed prior to computing the correlations. Correlations during rest were computed using 44 segments of 6 s. We also used further benchmark tests for the control situations to evaluate the effect of notch filtering and of using a different analysis frequency on the correlation patterns. We reasoned that if the difference in the correlation patterns obtained in the *Ori\@6 Hz* versus the control situations is due to the notch filtering or to the use of a different analysis frequency rather than to FPVS, then we should observe similar differences when applying notch filtering or using a different frequency on signal which does not contain a visually‐driven periodic response. The benchmark situations to evaluate the effect of notch‐filtering on the patterns of correlations were the following: (a) *NotchOri\@4 Hz*: correlations using amplitude envelope at 4 Hz when the original signal has been notched filtered at 4 Hz (\[3.9--4.1\] Hz); (b) *NotchRest\@6 Hz*: correlations using amplitude envelope at 6 Hz when the signal during rest has been notched filtered at 6 Hz (\[5.9--6.1\] Hz). The benchmark situation to evaluate the effect of using the amplitude envelope at a different frequency was the following: (c) *Rest\@4 Hz*: correlations using the amplitude envelope at 4 Hz from the signal recorded during rest. The pattern of correlations in the original, control and benchmark situations were summarized and statistically compared by computing an index of right‐hemispheric lateralization: we subtracted the correlations averaged over left‐hemispheric OT scalp electrodes from the correlations averaged over corresponding electrodes in the right hemisphere (sO2, sPO8, sP4, and sP10). These indices were compared against zero and against each other using a permutation test (10,000 permutations). 3. RESULTS {#hbm24952-sec-0016} ========== 3.1. Patient KV shows a significant and typical synchronization to 6 Hz face identity stimulation on the scalp {#hbm24952-sec-0017} -------------------------------------------------------------------------------------------------------------- Fast periodic face stimulation generated robust and significant (z‐score \> 2.33, *p* \< .01, one‐tailed) visual responses over patient KV\'s scalp as evidenced by the distinct peak in the EEG spectrum at the 6 Hz stimulation frequency (Figure [4](#hbm24952-fig-0004){ref-type="fig"}a). As in typical subjects (Alonso‐Prieto et al., [2013](#hbm24952-bib-0004){ref-type="ref"}; Rossion et al., [2012](#hbm24952-bib-0061){ref-type="ref"}; Rossion & Boremanse, [2011](#hbm24952-bib-0060){ref-type="ref"}), the EEG response in the *different faces* condition was clearly strongest over right OT electrodes (largest at sO2: 0.67 μV and sPO8: 0.63 μV). In addition, scalp topographies of the identity adaptation effect (i.e., larger amplitudes in the *different faces* relative to the *same face* conditions) revealed an even more focused positive difference at right OT electrodes (Figure [4](#hbm24952-fig-0004){ref-type="fig"}b) with a maximum at sPO8 (0.23 μV). The adaptation effect was statistically significant at two scalp electrodes, sPO8 and sP10 (z‐scores: 3.870 and 3.02, respectively; *ps* \< .01, two‐tailed). These observations largely replicate findings from healthy subjects using a similar experimental procedure (Alonso‐Prieto et al., [2013](#hbm24952-bib-0004){ref-type="ref"}; Rossion et al., [2012](#hbm24952-bib-0061){ref-type="ref"}; Rossion & Boremanse, [2011](#hbm24952-bib-0060){ref-type="ref"}; see also Liu‐Shuang et al., [2014](#hbm24952-bib-0035){ref-type="ref"}; Figure [4](#hbm24952-fig-0004){ref-type="fig"}c), and suggest that patient KV\'s electrophysiological responses related to (unfamiliar) face individuation are typical. Further, this demonstrates that significant functional responses can be obtained from the scalp with the FPVS approach in single epileptic patient tested for a few minutes only. {#hbm24952-fig-0004} 3.2. Corresponding spatial location of maximal response amplitude to faces between scalp and intracerebral EEG recordings {#hbm24952-sec-0018} ------------------------------------------------------------------------------------------------------------------------- A detailed report of periodic intracerebral responses recorded in the *same face* and *different faces* conditions in patient KV\'s right OT cortex is described in Jonas et al. ([2014](#hbm24952-bib-0027){ref-type="ref"}). Here, for comparison with scalp recordings, we will summarize responses measured in the *different faces* condition and the effect of identity adaptation. In intracerebral recordings, we found robust responses to the visual presentation of different faces at the 6 Hz stimulation frequency (Figure [5](#hbm24952-fig-0005){ref-type="fig"}a,b) in ventral and lateral sections of the occipital and posterior temporal cortex. SEEG Responses were significantly above noise (all zs \> 3.45, *ps* \< .001) for all contacts except three adjacent contacts of electrode F (F9 to F11). The largest responses were measured in contacts located in the posterior section of the collateral sulcus (D1 to D5 and L1 to L5, amplitude range: 0.7 to 6.9 μV, mean amplitude: 2.8 μV, Figure [5](#hbm24952-fig-0005){ref-type="fig"}b) and the lateral section of the inferior occipital gyrus (D6 to D8 and L6 to L8, amplitude range: 4.4 to 7.6 μV, mean amplitude: 5.8 μV). Responses in electrode F, going through the CoS (F1 to F5), occipitotemporal sulcus (F6 to F10) and posterior inferior temporal gyrus (F11) were overall much weaker (amplitude range: 0.4 to 2.5 μV, mean amplitude: 1.3 μV). SEEG responses also displayed a significant face identity adaptation effect (larger response in the *different faces* condition; z \> 2.33, *p* \< .01 Figure [5](#hbm24952-fig-0005){ref-type="fig"}c) at contacts in or near the right lateral IOG (D5, D7, D8, L7, and L8: mean amplitude for different faces: 5.8 μV, mean amplitude for same face: 3.2 μV) or more anterior contacts in the CoS (F3) or in the OTS above the lateral fusiform gyrus (F7, F8, \~8 mm from the cortical surface of the lateral fusiform gyrus). {#hbm24952-fig-0005} Importantly, there is a correspondence in the spatial location of intracerebral contacts and scalp electrodes showing the strongest response in the *different faces* condition, or the strongest effect of adaptation. First, in the *different faces* condition (Figure [5](#hbm24952-fig-0005){ref-type="fig"}b), the scalp electrodes displaying the strongest scalp EEG responses were the closest to the intracerebral contacts in lateral IOG (sPO8, sO2, sP6, sP4 scalp electrodes; Euclidean distance to D8 = 27, 26, 32, and 43 mm, respectively; Euclidean distance to L8 = 28, 35, 33, and 46 mm, respectively). Second, the adaptation effect in intracerebral recording was maximally measured at two separate cortical locations: around the lateral IOG (D5, D7, D8, L7, and L8 contacts) and more anteriorly in the OTS above the lateral fusiform gyrus (F7, F8 contacts). Interestingly, this pattern of intracerebral response was associated with a maximal adaptation effect on the scalp over slightly more anterior electrodes (e.g., sP10) compared to the response in the *different faces* condition: scalp electrode sPO8 was closest to intracerebral contacts in the lateral IOG and scalp electrode sP10 was closest to intracerebral contacts in the lateral fusiform gyrus compared to contacts in lateral IOG (Euclidean distances from sP10 = 35 mm to F8, 45 mm to L8, and 51 mm to D8). 3.3. Strong amplitude and SNR attenuation in scalp compared to intracerebral EEG {#hbm24952-sec-0019} -------------------------------------------------------------------------------- While we observed a spatial correspondence of the largest responses in scalp and intracerebral recordings, the amplitude measured on the scalp was very much attenuated relative to the intracerebral signal (Figure [5](#hbm24952-fig-0005){ref-type="fig"}a, compare top and bottom plots). Indeed, relative to intracerebral contacts D8 and L8 where the amplitude at 6 Hz in the *different faces* condition was 6.3 and 5.2 μV respectively, the amplitudes at the closest scalp electrodes sO2 and sPO8 were 0.67 and 0.63 μV respectively, which is between 7.7 and 9.9 times smaller than the corresponding intracerebral signal. Interestingly, the SNR at the stimulation frequency (i.e., 6 Hz) was less attenuated than the absolute signal amplitude when comparing scalp to intracerebral recordings. SNR was between 2 and 3.5 times lower in scalp (SNR = 8.3 and 8.1 for sO2 and sPO8, respectively) compared to intracerebral (SNR = 29 and 16.7 for D8 and L8, respectively) recordings. This is due to the signal amplitude being comparatively more attenuated from intracerebral to scalp recordings than the mean noise amplitude in frequency bins around 6 Hz (mean noise at D8/L8 = 0.26 μV; mean noise at sO2/sPO8 = 0.08 μV; ratio of intracerebral to scalp noise = 3.4). 3.4. Focal correspondence between intracerebral and scalp EEG signals over the right occipitotemporal cortex {#hbm24952-sec-0020} ------------------------------------------------------------------------------------------------------------ ### 3.4.1. Phase investigation {#hbm24952-sec-0021} For the remainder of the analyses, we will focus on the signal measured in the *different faces* condition, as it generated the strongest responses over the right OT cortex, in contrast to the *same face* condition which generated low responses over OT regions and a maximal response over medial occipital regions (Supplementary Figure [S1](#hbm24952-supitem-0001){ref-type="supplementary-material"}) as in previous publications (Alonso‐Prieto et al., [2013](#hbm24952-bib-0004){ref-type="ref"}; Rossion et al., [2012](#hbm24952-bib-0061){ref-type="ref"}; Rossion & Boremanse, [2011](#hbm24952-bib-0060){ref-type="ref"}). This latter condition is therefore sub‐optimal to investigate the relationship between intracerebral and scalp responses in the OT region. Moreover, this avoids relying on a post hoc subtraction of conditions (i.e., adaptation effect), which provided an optimal opportunity to investigate the relationship between simultaneously recorded intracerebral and scalp EEG. To further characterize the relationship between intracerebral and scalp recordings, we computed the phase of the signal at the stimulation frequency for intracerebral and scalp channels and visualized the signal in the time domain (Figure [6](#hbm24952-fig-0006){ref-type="fig"}). Phase provides additional information about response timing and allows to further characterize the relationship between neighboring recording sites. For instance, perfect phase alignment or phase‐reversal (i.e., 180° difference) at two separate recording sites is indicative of a common neural source generating the signal measured at the two sites. This revealed that although lateral contacts of the D and L electrodes (D5--D8 and L5--L8) all exhibited strong visual responses at the stimulation frequency (Figures [5](#hbm24952-fig-0005){ref-type="fig"} and [6](#hbm24952-fig-0006){ref-type="fig"}), only the four most lateral contacts (D7--D8, L7--L8) were in phase with the signal measured at nearby scalp electrodes (Figure [6](#hbm24952-fig-0006){ref-type="fig"}). The signal of adjacent---more medial---contacts (D5--D6, L5--L6) was out of phase relative to the more lateral contacts. This is clear when visualizing data in the time domain (Figure [6](#hbm24952-fig-0006){ref-type="fig"}a), where the crests and troughs of the responses measured at the most lateral intracerebral contacts (D7--D8, L7--L8) are temporally aligned with responses measured at the scalp, but misaligned with signal measured at more medial contacts (D5--D6, L5--L6). This is also reflected in the phase of the frequency spectra (Figure [6](#hbm24952-fig-0006){ref-type="fig"}b), where the phase values are similar between D8/L8 and right OT scalp electrodes (sO2, sPO8, sP10, mean phase difference: 11°), but very dissimilar between contiguous contacts D5--D6 and D7--D8 (mean phase difference: 158°) and between L5--L6 and L7--L8 (mean phase difference: 116°). The observation of only a partial phase opposition between these two groups of intracerebral contacts (i.e., rather than a phase difference close to 180°) suggests that the signal measured at these two groups arises from partly distinct generators. Only one of these neural generators, the generator contributing to the signal at lateral IOG contacts (D7--D8, L7--L8), also contributes to the EEG signal measured at OT scalp electrodes. {#hbm24952-fig-0006} ### 3.4.2. Correlation investigation {#hbm24952-sec-0022} To directly and formally quantify the relationship between scalp and intracerebral signals, we correlated the variations of the 6 Hz response amplitude over time across all scalp and intracerebral contacts (Figure [3](#hbm24952-fig-0003){ref-type="fig"}, methods). As a reminder, this analysis disregards phase information so that only coordinated temporal variation of amplitude across channels could result in meaningful correlation coefficients. This was done to avoid correlations (positive or negative) driven simply by phase coherence across channels triggered by a common visual stimulation. While the temporal variations of the amplitude of the 6 Hz response were strongly correlated between contacts D5--D6/L5--L6 and between contacts D7--D8/L7--L8, the 6 Hz signal was not correlated between these two groups of adjacent contacts. Specifically, correlations were high between contacts D5, D6, L5, and L6 (both within electrode: r = 0.84 ± 0.07 and 0.82 ± 0.13 for D5--D6 and L5--L6, respectively, and across the D and L electrodes: r = 0.5 ± 0.26 and 0.42 ± 0.22 for D5--L5 and D6--L6, respectively) and between contacts D7, D8, L7, and L8 (within electrode: r = 0.97 ± 0.03 and 0.76 ± 0.16 for D7--D8 and L7--L8, respectively, and across the D and L electrodes: r = 0.5 ± 0.23 and 0.68 ± 0.16 for D7--L7 and D8--L8, respectively). In contrast, there was no meaningful correlation between signals at immediately adjacent contacts D6--D7 (r = 0.08 ± 0.28) and L6--L7 (r = 0.02 ± 0.27). Thus, these analyses support our view that despite the partial phase opposition between contacts D5--D6/L5--L6 and contacts D7--D8/L7--L8 (Figure [6](#hbm24952-fig-0006){ref-type="fig"}), the signal measured at these two groups of contacts arises from different cortical generators. In addition, these analyses confirmed and refined our observation above that signal measured at right OT scalp electrodes arises mainly from regions in the lateral IOG around contacts D7--D8 and L7--L8 rather than from more medial cortical regions around contacts D5--D6, L5--L6. Indeed, we observed a very focal pattern of positive correlations between contacts D/L 7--8 and right OT scalp electrodes closest to the location of the corresponding intracerebral contacts (Figure [7](#hbm24952-fig-0007){ref-type="fig"}a). Statistical analyses indicated that the 6 Hz signal at each of these four intracerebral contacts was significantly correlated with a cluster of 4 to 5 scalp electrodes. D7, D8, L7, and L8 were each significantly correlated with sPO8, sO2, sP10, and sP4, and L7 was also significantly correlated with sP6. Among these channels, Pearson\'s coefficients ranged from 0.32 (*SD* across segments: ± 0.25) to 0.37 ± 0.23 for sPO8, from 0.28 ± 0.28 to 0.36 ± 0.22 for sO2, from 0.19 ± 0.22 to 0.25 ± 0.19 for sP4 and from 0.17 ± 0.29 to 0.21 ± 0.31 for sP10. In contrast, correlations of the same scalp electrodes with intracerebral contacts D5--D6, L5--L6 were around zero (range of correlations: −0.11 ± 0.25 to −0.02 ± 0.22). Moreover, these positive correlations were restricted to the right hemisphere: the correlations for contacts D7--D8, L7--L8 with electrodes in the OT left hemisphere (sO1, sPO7, sP9, sP7) ranged from −0.08 to 0.11 (mean r = 0.01 ± 0.24). {ref-type="fig"} for anatomical location of the contacts). (b) Topographical maps showing only significant correlation coefficients (*p* \< .05, cluster‐based correction for multiple comparisons). White indicates no significant correlation](HBM-41-2373-g007){#hbm24952-fig-0007} Correlations were slightly but significantly higher (t\[8\] = 7.6, *p* \< .001) for the most external contact (D8, L8, mean r across channels: 0.28) relative to the immediately adjacent more internal contact (D7, L7, mean r across channels: 0.25). Over the more anterior F electrode, only the signal from F11 tended to be correlated with surrounding scalp electrodes (Figure [7](#hbm24952-fig-0007){ref-type="fig"}a, sP10: r = 0.2 ± 0.25; sP6: r = 0.2 ± 0.3; sPO8: r = 0.18 ± 0.33; sT8: r = 0.14 ± 0.3). However, these correlations did not reach significance at the cluster level. 3.5. Focal and right‐lateralized intracerebral‐scalp correlations are specific to the FPVS signal {#hbm24952-sec-0023} ------------------------------------------------------------------------------------------------- The observed correlations suggest that these scalp and intracerebral channels pick up electrophysiological responses coming from the same cortical region responding to the periodic visual stimulation. However, these correlations may also be driven by a general electrophysiological activity (i.e., unrelated to the stimulation) coming from cortical territories to which both intracerebral contacts and scalp electrodes are sensitive to, given their spatial proximity. If this is the case, we should observe the same correlation pattern with or without the presence of a visually‐driven response in the recorded electrophysiological signal. We, therefore, compared the patterns of scalp‐intracerebral correlations obtained at the stimulation frequency during periodic visual stimulation (*Ori\@6 Hz*) with a series of "control" situations: (a) *NotchOri\@6 Hz*: Pattern of correlations when the visually‐driven signal has been selectively filtered‐out; (b) *Ori\@4 Hz*: Pattern of correlations at another frequency than the stimulation frequency (4 Hz); (c) *Rest\@6 Hz*: Pattern of correlations obtained using 6 Hz signal recorded during a rest period. These analyses revealed that intracerebral‐scalp correlations were more focal when a visually‐driven periodic response was present in the electrophysiological signal. Specifically, in the three control conditions, while the patterns of scalp correlations measured for intracerebral contacts D7--D8 and L7--L8 were overall similar to the one observed at 6 Hz using the original signal, these patterns were all more widespread and included significant correlations on the scalp both in the right and the left hemisphere (Figure [8](#hbm24952-fig-0008){ref-type="fig"}). This observation was reflected in the larger number of scalp electrodes significantly correlated with each intracerebral contacts---D7, D8, L7, and L8---in the three control situations (mean number of significant scalp electrodes across D7, D8, L7, and L8 = 9.25 to 14.5, Figure [9](#hbm24952-fig-0009){ref-type="fig"}a) compared to the original FPVS condition (*Ori\@6 Hz*: 4.25 significant scalp electrodes). In addition, the right hemispheric dominance of the correlations pattern when using the original signal at 6 Hz was assessed and compared to control conditions by subtracting the correlations averaged over left‐hemispheric OT scalp channels from the correlations averaged over corresponding electrodes in the right hemisphere (sO2, sPO8, sP4, sP10). This revealed that, while scalp correlations were significantly stronger in the right than in the left hemisphere in all conditions (all *p*\'s \< .02, one‐tailed permutation test, Figure [9](#hbm24952-fig-0009){ref-type="fig"}b), the right hemispheric dominance was significantly stronger when using the signal at 6 Hz in the original signal where periodic visual stimulation is present (*Ori\@6 Hz*: right hemispheric dominance = 0.24 ± 0.16) compared to the three control situations (0.09 ± 0.14, 0.1 ± 0.1, 0.1 ± 0.13, all *p*\'s \< .005, one‐tailed permutation test). In contrast, there was no significant difference among any of the control and additional benchmark situations (*p*'s \> 0.5, Figure [9](#hbm24952-fig-0009){ref-type="fig"}b, see methods for benchmark situations). {#hbm24952-fig-0008} {#hbm24952-fig-0009} 4. DISCUSSION {#hbm24952-sec-0024} ============= This study shows that a few minutes of periodic visual stimulation suffice to generate a robust signal, objectively identifiable at the exact frequency of stimulation (and harmonics) in the frequency spectrum of both SEEG and EEG signals recorded simultaneously. Here we find in the single epileptic patient tested that the response peaks over the right occipitotemporal scalp region, as in neurotypical participants tested with this paradigm (Figure [4](#hbm24952-fig-0004){ref-type="fig"}), suggesting the generalizability of the present observations to the normal population. By combining FPVS with correlation analyses we thus provide an original approach to investigate the relationship between functional brain electrophysiological activity measured simultaneously inside the brain and on the scalp. Quantification of the 6 Hz response in the frequency domain is straightforward and reveals a tenfold decrease of amplitude at 6 Hz between the most external intracerebral contacts and the nearest scalp EEG electrodes (i.e., 25--30 mm), providing unique information about skull attenuation of electrophysiological activity (Oostendorp, Delbeke, & Stegeman, [2000](#hbm24952-bib-0051){ref-type="ref"}; Wendel, Vaisanen, Seemann, Hyttinen, & Malmivuo, [2010](#hbm24952-bib-0070){ref-type="ref"}). The choice of the 6 Hz stimulation frequency was dictated by previous studies showing robust responses at this frequency for face stimulation, in particular when different face identities are presented at every stimulation cycle (Alonso‐Prieto et al., [2013](#hbm24952-bib-0004){ref-type="ref"}). Note that this attenuation might even be underestimated, given that the intracranial sampling was limited and that larger responses might have been found at other nearby locations inside the brain. Nevertheless, the attenuation in SNR between intracerebral and surface EEG responses was of "only" 2--3.5. This reduced ratio indicates that the electrophysiological noise, as computed as in the present study, is significantly larger inside than outside the brain. A major factor contributing to this reduction of the ratio between amplitudes and SNR is that, rather than being computed over a prestimulus baseline as in standard ERP studies, electrophysiological noise is computed here within a small theta range frequency around the signal of interest that is, 6 Hz (Meigen & Bach, [1999](#hbm24952-bib-0042){ref-type="ref"}; Rossion et al., [2012](#hbm24952-bib-0061){ref-type="ref"}; Srinivasan, Russell, Edelman, & Tononi, [1999](#hbm24952-bib-0066){ref-type="ref"}). Hence, EEG noise is "free" of alpha activity, environmental noise, eye and muscle artifacts, and so forth which typically greatly contaminate scalp EEG signals (Luck, [2014](#hbm24952-bib-0039){ref-type="ref"}). Additionally, the cortical surface to which an EEG scalp electrode is sensitive to is likely larger than that of an SEEG electrode. Noise in scalp EEG might thus be smaller compared to SEEG by virtue of averaging uncorrelated electrophysiological "noise" over a larger surface. The narrow band of the 6 Hz signal allows objective and finer‐grained tracking of the phase and identification of the intracerebral electrodes generating that signal on the scalp. Using FPVS in relation with our correlation approach, we found that activity recorded over right OT scalp regions directly relates to the activity measured at intracerebral contacts located in the inferior occipital gyrus either in the cortex (D7, L7, L8) or in the meninges (D8). This latter intracerebral contact likely receives electrical field propagation from the nearby cortex (i.e., about 2 mm distance) (Zaveri, Duckrow, & Spencer, [2009](#hbm24952-bib-0074){ref-type="ref"}). This strongly suggests that one major neural source for the 6 Hz FPVS signal measured on the scalp is located in the inferior occipital gyrus at or near contacts D7--D8 and L7--L8. Other sources in neighboring location likely also contribute to the measured scalp activity, but they could not be captured here due to the limited intracerebral sampling in the clinical case. Nevertheless, the main contribution of lateral brain sources to scalp EEG by comparison to medial sources is in line with previous studies in epilepsy and especially in mesial temporal lobe epilepsy (Koessler et al., [2015](#hbm24952-bib-0031){ref-type="ref"}; Merlet et al., [1998](#hbm24952-bib-0043){ref-type="ref"}). The identification of the right OT scalp region is similar to our findings in a previous study measuring the relationship between the face‐evoked N170 potential on the scalp and in the cortex in the same epileptic patient (Jacques et al., [2019](#hbm24952-bib-0023){ref-type="ref"}). However, this relationship was relatively more widespread on the scalp in the previous ERP study (see Figures [5](#hbm24952-fig-0005){ref-type="fig"} and [6](#hbm24952-fig-0006){ref-type="fig"} in Jacques et al., [2019](#hbm24952-bib-0023){ref-type="ref"}) compared to the current study. Specifically, in the previous study, significant correlations involved up to six scalp electrodes (right OT and medial occipital) when correlating N170 latency and up to 11 scalp electrodes (bilateral OT and medial occipital) when correlating N170 amplitude, while the relationship found here with the frequency‐tagging approach was restricted to four scalp electrodes (Figure [7](#hbm24952-fig-0007){ref-type="fig"}). Notwithstanding this comparison across studies of the spatial spread of the correlations on the scalp, it should be noted that differences in the experimental design, in the type of signal used to compute the correlations and in the analyses schemes prevent a more formal comparison across the current and the previous study. For instance, in the current study, the use of FPVS allows taking into account the amplitude variations in a relatively narrow frequency range (i.e., around 6 Hz), which is less affected by ongoing broadband EEG noise (mostly in the theta and alpha range) compared to when using the ERP signal as in the previous study. Such broadband noise in the ERP signal may be correlated across a large cortical surface (e.g., Buzsáki, [2002](#hbm24952-bib-0006){ref-type="ref"}; Klimesch, [1999](#hbm24952-bib-0029){ref-type="ref"}; Miller, Foster, & Honey, [2012](#hbm24952-bib-0045){ref-type="ref"}), therefore being picked up simultaneously by intracerebral contacts in the IOG and by many posterior scalp electrodes, resulting in a broad correlation pattern on the scalp. Therefore, while the cortical regions involved in generating the N170 and the 6 Hz FPVS face signals are likely similar given the overall similarity in the scalp topographies of the correlations (irrespective of their spatial spread) across the two studies, the FPVS approach allows revealing a more focused relationship between scalp and intracerebral recordings. In addition, we demonstrate that the relationship between scalp and intracerebral channels at 6 Hz is significantly more focal and right‐lateralized during FPVS as compared to rest EEG or during stimulation but considering a frequency range outside the stimulation frequency (i.e., around 4 Hz). Hence, this relationship is specific to the stimulation frequency and cannot be attributed to general factors such as an increase of arousal during visual stimulation for instance. Moreover, other benchmark control situations (Figure [9](#hbm24952-fig-0009){ref-type="fig"}) further indicate that the specificity of this relationship during FPVS is not driven by analyses or testing parameters. Note that since we did not stimulate with other frequencies, whether 6 Hz provides the tightest correlation between scalp and intracerebral activity cannot be determined in the present study---although this is likely given the particularly large response that it generates in this paradigm (Alonso‐Prieto et al., [2013](#hbm24952-bib-0004){ref-type="ref"}). A peculiar observation is the broader pattern of correlations for control situations, such as when using a different frequency of analysis (i.e., Ori\@4 Hz), compared to when using FPVS. One might have expected no or reduced correlations in control situations when no visual response is present in the signal at the frequency of analyses. The observation of a correlation for the control situations likely stems from the close spatial proximity between scalp right OT electrodes and the most lateral intracerebral contacts, which makes it likely that these recording channels are sensitive to a similar cortical territory. This would result in correlations across channels inside the brain and on the scalp stemming from the fact that they are sensitive to the same ongoing cortical EEG activity unrelated to a visual response (i.e., "noise"). As indicated above, this EEG "noise" in the theta and alpha range (e.g., Buzsáki, [2002](#hbm24952-bib-0006){ref-type="ref"}; Klimesch, [1999](#hbm24952-bib-0029){ref-type="ref"}; Miller et al., [2012](#hbm24952-bib-0045){ref-type="ref"}) is more likely to correlate across a larger brain surface than neural signal related to the visual stimulation, therefore being measured over a broader scalp surface. This would yield patterns of correlations that are more broadly distributed on the scalp over both hemispheres. This phenomenon is also likely to take place in the control situation where the visually‐driven signal at 6 Hz was removed using a very narrow notch filter (i.e., NotchOri\@6 Hz). In this particular situation, one possibility is that when the visually‐driven 6 Hz signal is filtered out, the variations of the wavelet amplitude envelope across time (which is used to obtain the correlations) are related to other frequencies around 6 Hz (i.e., theta range) that fall within the frequency bandwidth of the wavelet at 6 Hz (full width at half maximum in the frequency‐domain at 6 Hz is 0.8 Hz), resulting in broad scalp correlation patterns. In contrast, when visually‐driven signal at 6 Hz is present (Ori\@6 Hz), it completely dominates over surrounding frequencies in the amplitude envelope, allowing to reveal the more local neural activity specifically related to face periodic stimulation. Several studies (Cosandier‐Rimele, Merlet, Badier, Chauvel, & Wendling, [2008](#hbm24952-bib-0011){ref-type="ref"}; Ebersole, [1997](#hbm24952-bib-0014){ref-type="ref"}; Ramantani et al., [2014](#hbm24952-bib-0054){ref-type="ref"}; Tao, Ray, Hawes‐Ebersole, & Ebersole, [2005](#hbm24952-bib-0068){ref-type="ref"}) have shown that a large cortical surface from 6 to 30 cm^2^ is required to generate detectable scalp EEG signals. According to a recent computational study (Cosandier‐Rimele et al., [2008](#hbm24952-bib-0011){ref-type="ref"}), a cortical surface of 26cm^2^ is required to obtain a SNR equal to eight in scalp EEG, corresponding roughly to the SNR observed in our study. However, the FPVS approach is known to generate extremely high SNR responses on the scalp, so that more focal sources in the lateral section of the inferior occipital gyrus, corresponding to smaller cortical patches, may have been sufficient to generate this response. While fMRI studies indicate that the (right) IOG plays a key role in (unfamiliar) face individuation responses in the human brain (e.g., Gauthier, Tarr, Moylan, Skudlarski, & Gore, [2000](#hbm24952-bib-0016){ref-type="ref"}; Schiltz et al., [2006](#hbm24952-bib-0063){ref-type="ref"}), these responses are not confined to this region but extend to more anterior regions of the lateral fusiform gyrus. Although these regions may also contribute to the signal recorded on the scalp, the cortical orientation of the lateral fusiform gyrus makes it unlikely for this region to majorly contributes to the EEG measured over lateral OT cortex scalp regions (Jacques et al., [2019](#hbm24952-bib-0023){ref-type="ref"}). While simultaneous scalp and intracerebral studies very often rely on relatively complex signal processing methods (e.g., blind source separation) to extract biomarkers from scalp background activity (Koessler et al., [2015](#hbm24952-bib-0031){ref-type="ref"}; Pizzo et al., [2019](#hbm24952-bib-0053){ref-type="ref"}), the spontaneous visibility (i.e., even without averaging method or baseline correction method) of the scalp responses during FPVS highlights the interest of this approach to clarify the scalp EEG correlates of focal cortical generators. Therefore, compared to signal analysis performed in time domain and on spontaneous activity (like epileptic spikes or seizures), FPVS in combination with our correlation approach may represent a major tool in order to understand and characterize the link between cortical source activity and scalp EEG signals. Moreover, our observation of robust intracerebral vs. scalp EEG correlations with just two sequences of stimulation (2 × 50 s of data) is a clear advantage over more conventional stimulation approach in clinical settings where these recordings take place. Given that we report data from a single patient implanted with three intracerebral electrode arrays, this approach should be further validated to ensure its applicability in different experimental research domains and settings. Nevertheless, the frequency‐tagging approach is readily used to measure a wide range of responses in the visual domain (e.g., to luminance or contrast changes, but also visual words, quantities, or objects, Norcia et al., [2015](#hbm24952-bib-0048){ref-type="ref"}) and in other modalities (e.g., low‐level and high‐level auditory responses: Fujiki, Jousmaki, & Hari, [2002](#hbm24952-bib-0015){ref-type="ref"}; Nozaradan, Peretz, Missal, & Mouraux, [2011](#hbm24952-bib-0049){ref-type="ref"}; somatosensory responses: Colon, Legrain, Huang, & Mouraux, [2015](#hbm24952-bib-0010){ref-type="ref"}), as well as their modulation by attentional factors (Chen, Seth, Gally, & Edelman, [2003](#hbm24952-bib-0008){ref-type="ref"}; Colon et al., [2015](#hbm24952-bib-0010){ref-type="ref"}; Morgan, Hansen, & Hillyard, [1996](#hbm24952-bib-0046){ref-type="ref"}; Yan, Liu‐Shuang, & Rossion, [2019](#hbm24952-bib-0073){ref-type="ref"}). This suggests that the original approach introduced here could be extended to larger samples of individual brains and other brain functions, pending appropriate adaptation in the stimulation and analyses parameters. Moreover, this approach could be extended to understand functional connectivity between intracerebral contacts in close or remote regions with a wider sampling across the brain. Supporting information ====================== ###### **Supplementary Figure 1 FPVS responses and correlation between EEG and SEEG in the "same face" condition.** A. Patient KV\'s scalp topographical distribution showing the 6 Hz SNR response to faces in the "different face" condition (top; same as in Figure [4](#hbm24952-fig-0004){ref-type="fig"}b of the main manuscript) and "same face" condition (bottom). While the largest response is focused over the right occipitotemporal region in the "different face" condition, the response is maximal over the medial occipital scalp region in the "same face" condition as in previous publications (Rossion and Boremanse, [2011](#hbm24952-bib-0060){ref-type="ref"}; Rossion et al., [2012](#hbm24952-bib-0061){ref-type="ref"}). B Correlations between intracerebral and scalp signals during FPVS in the "same face" condition. Top: Scalp topographical maps of the unthresholded Pearson correlation coefficients between the signal around 6 Hz measured at each intracerebral contact and each scalp electrode. Each map represents the correlations of one intracerebral contact with all scalp electrodes. Different intracerebral electrodes (D, L, F) are show in rows and adjacent recording contacts are shown in columns (see Figure [1](#hbm24952-fig-0001){ref-type="fig"} for anatomical location of the contacts). Bottom. Topographical maps showing only significant correlation coefficients (*p* \< 0.05, cluster‐based correction for multiple comparisons). White indicates no significant correlation. The pattern of significant correlations appears broader on the scalp compared to the "different face" condition (see Figure [7](#hbm24952-fig-0007){ref-type="fig"} in the main manuscript), likely because the 6 Hz response in the "same face" condition is relatively weak in right OT cortical regions (Figure [5](#hbm24952-fig-0005){ref-type="fig"}) and significant correlations are driven by EEG noise (see related discussion in the manuscript). This is consistent with the observation of consistent correlations between scalp and contact F11 (located close to the scalp) where no significant 6 Hz response was measured. **Supplementary Table 1.** Phase of the 6 Hz FPVS signal for intracerebral and scalp channels where a significant 6 Hz response was measured (Z \< 2.33). ###### Click here for additional data file. This work was supported by the Belgian Fonds National de la Recherche Scientifique (FNRS --PDR T.0207.16), Fédération Wallonie‐Bruxelles under Grant No. ARC 13/18--053, and the Fondation Louvain. A recent study recorded 48‐channel ECoG and 27‐channel scalp‐EEG data simultaneously during light flickering in a single patient (Wittevrongel et al., [2018](#hbm24952-bib-0072){ref-type="ref"}). However, due to insufficient quality of the simultaneously recorded scalp‐EEG (dried conductive gel, as well as the influence of scarred and swollen tissue with the particularly invasive ECoG procedure), the patient\'s scalp‐EEG was excluded from further analysis. DATA AVAILABILITY STATEMENT {#hbm24952-sec-0027} =========================== The data that support the findings of this study are available from the corresponding author upon reasonable request. [^1]: Bruno Rossion and Laurent Koessler contributed equally to this work.

{#sp1 .378}

Complete anterior capsulotomy is one of the critical steps in cataract surgery. Even in cases where the surgeon performs a perfect capsulotomy, changes in the anterior capsule opening may occur, usually several weeks after surgery \[[@B1]\]. Reduction of the anterior capsule opening area leads to a reduced optic zone area and severe contraction may cause a reduction in visual outcomes. Recently, several studies have focused on how to prevent anterior capsule contraction. Some pathologic conditions, like uveitis, pseudoexfoliation syndrome, and retinitis pigmentosa, are correlated with anterior capsule contraction \[[@B2][@B3][@B4]\]. In normal eyes, reduction of the anterior capsule opening is related to the initial anterior capsulotomy size, roundness, circularity, and intraocular lens (IOL) design and material \[[@B5][@B6][@B7]\]. Some studies have reported that silicone IOLs with a plate-haptic design are more susceptible to anterior capsule shrinkage due to the limitation of capsular dilation \[[@B8]\]. Tsinopoulos et al. \[[@B9]\] reported that anterior capsule contraction syndrome occurs more frequently with hydrophilic lenses than with hydrophobic lenses. Due to advances in minimally invasive surgery, foldable acrylic IOLs are preferred over various other materials. We hypothesized that IOL haptic design may influence anterior capsule stability in the very early post-surgery period, even when IOLs are made of the same material. A previous study compared IOLs made with different designs and different materials. There are few studies that have evaluated differences in anterior capsule stability after single piece acrylic IOL implantation with modified L-haptic design or plate-haptic design. The purpose of our study was to compare anterior capsule contraction and IOL decentration following implantation of single piece acrylic IOLs with different haptic designs. Materials and Methods ===================== This was a prospective, observer-blinded study of patients who had cataract surgery between March 2012 and November 2013 at Seoul St. Mary\'s Hospital, Seoul, Korea. In all, 260 eyes from 150 patients were included in this study. Study protocol followed the guidelines of the Declaration of Helsinki and was approved by the hospital\'s institutional review board. Written informed consent was obtained from patients before the study and potential complications were explained in detail. Inclusion criteria were the presence of a senile cataract in adults older than 55 years of age, normal axial length between 22 mm and 25.5 mm, and a dilated pupil larger than 8.0 mm. No patient had a history of ocular surgery, pupil abnormality, or ocular disease other than cataract. Each patient underwent a complete ophthalmologic evaluation. Patients were randomly implanted with one of three types of IOL during cataract surgery: Matrix Aurium 404 IOL (Medennium, Irvine, CA, USA), EC-1 PAL IOL (Aaren Scientific, Ontario, CA, USA), and CT Asphina 509M IOL (Carl Zeiss, Jena, Germany). The three IOLs have different haptic designs but are made of the same acrylic material and consist of a single piece ([Table 1](#T1){ref-type="table"}). One surgeon (CKJ) performed all of the cataract surgeries using a standard procedure. Continuous curvilinear capsulorhexis of 5.0 mm to 5.5 mm diameter was carefully performed. After successful phacoemulsification, the IOL was implanted. Digital retroillumination photospots through a dilated pupil were obtained to evaluate the area of the anterior capsule opening immediately after continuous curvilinear capsulorhexis before viscoelastic removal, 1 day and 2 months postoperatively ([Fig. 1A-1C](#F1){ref-type="fig"}). IOL decentration was measured using vector analysis between the pupil center and IOL center. Maximal and minimal distances between the edge of capsulotomy and the edge of the IOL optic were calculated to determine capsulotomy overlap (overla*p* = distance minimum / distance maximum). All measurements were repeated three times and performed by a single experienced technician. Photospots were imported into Image J (National Institutes of Health, Bethesda, MD, USA) for measurement of capsulotomy diameter and area. IOL optic size was used as a reference scale. We used the same amount and type of mydriatic eyedrops to eliminate the effect of mydriatic drops on changes in the shape and size of the pupil before taking the photospot. Statistical analyses -------------------- Statistical analysis was performed using PASW ver. 18.0 (SPSS Inc., Chicago, IL, USA). Paired *t*-tests and Mann-Whitney tests were used for pair-wise comparisons. One-way ANOVA was used to compare all three IOLs. Pearson correlations were used to evaluate linear relationships. All data are expressed as the mean ± standard deviation. A *p*-value less than 0.05 was considered statistically significant. Results ======= Seventy-four patients received a modified L-haptic IOL, 97 patients received a modified C-haptic IOL, and 89 patients received a plate-haptic IOL implant. The average patient age was 66.25 ± 15.22 years in the modified L-haptic IOL group, 68.00 ± 10.12 years in the modified C-haptic IOL group, and 68.80 ± 8.71 years in the plate-haptic IOL group (*p* = 0.705) ([Table 2](#T2){ref-type="table"}). The initial area of the anterior capsule opening was not significantly different between the three IOL groups (*p* = 0.129). [Fig. 2](#F2){ref-type="fig"} shows the area of the anterior capsule opening on day 1 and 2 months postoperatively according to the IOLs. The differences in area between the time point after continuous curvilinear capsulorhexis and postoperative day 1 was 0.38 ± 1.26 mm^2^ for the modified L-haptic IOL, −0.54 ± 1.88 mm^2^ for the modified C-haptic IOL, and 0.03 ± 2.13 mm^2^ for the plate-haptic IOL. All IOLs showed significant reductions in anterior capsule opening at 2 months compared to on day 1, and a greater reduction was shown in the modified C-haptic IOL (modified L-haptic IOL, −1.81 ± 2.76 mm^2^; modified C-haptic IOL, −3.91 ± 2.91 mm^2^; and plate-haptic IOL, −2.50 ± 1.95 mm^2^) groups. The mean area reduction was significantly different between the three IOLs (*p* = 0.003). Two months postoperative, IOL decentration was significantly different between the three IOLs (modified L-haptic IOL, 0.24 ± 0.12; modified C-haptic IOL, 0.31 ± 0.12; and plate-haptic IOL, 0.26 ± 0.11, respectively; *p* = 0.002). No significant differences were founded in capsule overlap between the three IOLs (modified L-haptic IOL, 0.51 ± 0.16; modified C-haptic IOL, 0.48 ± 0.17; and plate-haptic IOL, 0.51 ± 0.16, respectively; *p* = 0.233). In all three IOLs, the degree of IOL decentration was significantly correlated with mean area reduction between the initial capsulotomy and postoperative month 2, with the strongest correlations observed for the modified C-haptic IOL ([Fig. 3](#F3){ref-type="fig"}). Discussion ========== Various reports have shown that IOL design and material are correlated with anterior capsule contraction \[[@B8][@B10]\]. With respect to IOL material, contraction occurs more frequently with silicone IOLs than with polymethylmethacrylate or acrylic IOLs \[[@B8]\]. Hayashi et al. \[[@B10]\] reported that hydrophilic acrylic IOLs are associated with a more pronounced reduction of the anterior capsule opening than hydrophobic acrylic IOLs. However, other authors have found no correlation between hydrophilicity of acrylic IOLs and anterior capsular shrinkage \[[@B11][@B12][@B13]\]. Gonvers et al. \[[@B6]\] reported that anterior capsule contraction in eyes with plate-haptic silicone IOLs was more extensive than that in eyes with three piece silicone/polymethylmethacrylate IOLs. Some studies have reported no significant differences between single piece and three piece acrylic IOLs or between two haptic and four haptic IOLs \[[@B11][@B14]\]. We evaluated changes in the anterior capsule opening at 1 day and 2 months after uncomplicated cataract surgery using single piece acrylic IOLs with different haptic designs. Reduction of the anterior capsule opening results from an imbalance between the centripetal force of anterior capsular fibrosis and centrifugal force of the zonule \[[@B15][@B16]\]. Anterior capsular fibrosis is induced by cytokines produced by lens epithelial cells (LEC) \[[@B17]\]. Blocking LEC proliferation, migration, and differentiation prevents anterior capsule contracture. Acrylic IOLs strongly adhere to the anterior capsule, which inhibits LEC migration \[[@B18][@B19]\]. There was a significant reduction in anterior capsule opening at 2 months compared to 1 day after surgery for all three IOLs. This reduction was largest in the modified C-haptic IOL, followed by the plate-haptic IOL and the modified L-haptic IOL. Despite being composed of the same material and having the same hydrophilicity, in our study we measured less reduction of the anterior capsule opening with the modified L-haptic IOL than the modified C-haptic IOL. This suggests the haptic of the modified C-haptic IOL may have had poor tensile strength and did not overcome centripetal forces resulting from capsular fibrosis \[[@B20][@B21]\]. The greater stability of the modified L-haptic IOL seemed to be related to proper tension on the bag along its equator which resulted in less change in the anterior capsulotomy area. A previous study reported that a single piece IOL with modified L-haptic showed less capsulotomy ovaling and maintained the capsular bag robustly \[[@B22]\]. Our result also supports the finding that the modified L-haptic design is associated with less anterior capsule contraction. Plate-haptic designs are known to cause less stretching of the capsular bag and more pronounced anterior capsule contraction \[[@B5][@B6]\]. Previous studies have mostly evaluated cases of first generation silicone plate-haptic IOL implantation. Compared with past silicone IOLs with plate-haptic design, the acrylic IOLs with plate-haptic design used in this study have relatively larger overall diameters and interfacial haptic planes. These IOLs showed less contraction compared to modified C-haptic IOLs and were comparable to the modified L-haptic IOLs with regard to a reduction in anterior capsule contraction. Single piece acrylic IOLs are known to produce less haptic force than three piece IOLs \[[@B23]\]. Hayashi and Hayashi \[[@B24]\] reported that single piece acrylic IOLs with modified-L haptic design resulted in no significant change in IOL decentration compared with three-piece acrylic IOLs. We found similar results for the degree of IOL decentration with the modified L-haptic design. In our study, decentration of the modified L-haptic design was significantly less than that of the modified C haptic IOL, although the difference was only less than 0.1 mm. This suggests that haptic design is associated with better IOL centration. In a correlation analysis, IOL decentration was positively correlated with the mean area reduction between the initial value and the value 2 months after surgery. Additionally, the strongest correlation was shown for the modified-C haptic IOL. Ohmi and Uenoyama \[[@B25]\] found that IOL decentration after surgery is influenced by asymmetric capsular contraction. When we performed a well-centered circular capsulotomy, the degree of IOL decentration was significantly different according to IOL type. It is possible that different haptic designs may result in different forces on the capsular bag and affect anterior capsulotomy shape after IOL implantation. The strengths of our study include the relatively large population compared with previous studies and the prospective randomized study design. Although our study had a relatively short follow-up period, our results are instructive. It is known that anterior capsule contraction after cataract surgery progresses rapidly during the first month and then slows after 3 months \[[@B10]\]. The proliferation of LEC and capsular contracture are active for 3 months after surgery and the initial results correlate with later results. In our study, none of the patients presented with visually significant contracture during the first 2 months. We were able to observe uncomplicated changes to their anterior capsule openings. In summary, our study demonstrated that contraction of the anterior capsule opening area was greater with the modified C-haptic IOL than modified L-haptic and palate haptic IOLs. The degree of IOL decentration was significantly different by IOL type, and showed a positive correlation with the mean area reduction over 2 months. Recently, acrylic has been a popular IOL material and IOL design is an important factor influencing outcomes. Anterior capsule contraction and IOL decentration vary according to IOL haptic design and, consequently, haptic design should be taken into consideration when determining the IOL to be used in selected cases. **Conflict of Interest:** No potential conflict of interest relevant to this article was reported. {#F1} {#F2} {#F3} ###### Characteristics of the three types of acrylic intraocular lenses  ###### Patient characteristics  OD = right eye; OS = left eye.

Introduction {#Sec1} ============ Multiple sclerosis (MS) is an immune-mediated disease that affects the entire central nervous system (CNS) \[[@CR1]--[@CR3]\]. Magnetic resonance imaging (MRI) lesions are well-scattered at white matter (WM) and grey matter (GM) \[[@CR4]\], while normal-appearing brain tissue in MRI also seems to be affected in pathological studies \[[@CR4]\]. Brain atrophy, the gradual loss of brain volume, is quite extensive in MS, nearly 0.5--1.35% per year, far off the limits of normal aging \[[@CR5], [@CR6]\]. It arises early in the course of the disease, accelerates with disease progression \[[@CR7]--[@CR12]\] but is attenuated by disease-modifying drugs \[[@CR13]\]. There has been increasing interest in measuring tissue loss in CNS, as it represents the net effect of all destructive pathogenic processes during the disease course \[[@CR14]--[@CR17]\]. It is worth recalling that neurons occupy almost half (46%) of the tissue volume, myelin is 24%, and glial and other cells almost 30% \[[@CR5]\]. GM \[[@CR4]\] holds much less myelin than WM (about one tenth), while neurons comprise its most abundant component \[[@CR18]\]. Relative to glial cells, oligodendrocytes outweigh the number of astrocytes, microglia and oligodendrocyte progenitor cells, although the exact percentage is still unknown \[[@CR19], [@CR20]\]. Atrophy in MS is often considered to be the result of extensive axonal transection and demyelination \[[@CR21]--[@CR23]\]. The contribution of neuroglia may be less clear; reactive gliosis has the potential to mask considerable tissue loss in WM lesions \[[@CR24], [@CR25]\]. Measurement of brain atrophy is also considerably influenced by the amount of tissue fluids \[[@CR26]\], which is increased by active inflammation and vasogenic edema in WM plaques, and decreased during treatment with agents with strong anti-inflammatory properties (pseudoatrophy effect) \[[@CR14], [@CR26]\]. Transient volume changes could also be attributed to idiosyncrasic and technical factors \[[@CR14]\]. Dehydration may affect functional integrity of neuroglial cells, while decreased protein levels mainly affect synaptic densities \[[@CR26]\]. Unlike demyelination, water volume fluctuations and transient biological factors, neuroaxonal damage is irreversible in CNS, and atrophy is primarily considered to reflect this neurodegenerative component in MS \[[@CR27]--[@CR30]\]. Finally, the atrophy rates may also be influenced to some extent by the genetic makeup of a person; Human leukocyte antigen (HLA) genotypes considered as 'high risk for MS', namely DRB1 and DQB1, have been associated with significantly lower WM and GM volumes, alongside with higher mean annualized percentage of brain volume change (PBVC) compared with medium and low risk HLA genotypes independent from patients clinical features (age, gender, disease course) or the DMTs used \[[@CR31]\]. Pathogenesis of brain atrophy {#Sec2} ============================= The time trajectory of brain atrophy {#Sec3} ------------------------------------ Focal tissue loss in WM plaques is undoubtedly a major contributor to brain atrophy. However, the correlation between demyelination foci and whole brain atrophy is still a matter of debate \[[@CR16]\]. Some studies have found a strong association \[[@CR32], [@CR33]\], while others have not \[[@CR25], [@CR34]--[@CR36]\], suggesting that separate pathologic processes may also contribute to tissue destruction. Chard et al. \[[@CR37]\] in a longitudinal 14-year study found that atrophy is more related to early rather than late focal lesion volumes. Inflammation may be an important contributor to global tissue loss in early disease stages (i.e. in clinically isolated syndrome). As the disease progresses, additional mechanisms emerge that are, at least partly, independent from WM injury, such as microglia activation, meningeal inflammation, iron deposition, oxidative stress and diffuse axonal damage in normal appearing white matter (NAWM). The lack of a significant relationship between white matter fraction (WMF) and T2 lesion load \[[@CR34], [@CR38]\] further support this hypothesis. Biopsy studies also confirm that the atrophy may proceed even in the absence of inflammation \[[@CR39], [@CR40]\]. Regional atrophy studies may also be helpful. Indeed, the volume loss of deep GM structures may be present in the early stages of the disease and it is strongly correlated with the disease course \[[@CR41]\]. In MS, brain atrophy may develop in different CNS structures and varies depending on the clinical disease phenotypes; ventricular enlargement is more prominent in relapsing--remitting MS \[RRMS\], whereas cortical atrophy seems to be more important in the progressive forms of the disease \[[@CR42]\]. All things considered, it has been suggested that the pathogenic trajectory of brain atrophy changes with disease progression; from primarily inflammatory to less inflammatory and primarily neurodegenerative in the late stages of the disease \[[@CR43], [@CR44]\]. Pathogenesis of acute demyelination and axonal injury {#Sec4} ----------------------------------------------------- In the initial stages of MS, many different components of the adaptive and the innate immunity induce demyelination and neuronal loss \[[@CR43]\]. The activation of auto-reactive CD4+ T lymphocytes in the peripheral immune system is necessary for their migration across the blood--brain-barrier (BBB) and into the CNS. After myelin destruction, T cells are in situ reactivated by antigens within myelin debris and their clonal expansion results in multifocal demyelinating plaques \[[@CR45]\]. Peripheral B lymphocytes are involved in the antigen presentation and initial stimulation of CD4 T cells. Also, they are an essential source of pro- and anti-inflammatory cytokines (IL-6 among others) promoting every autoimmunity response (driven by Th1, Th2, Th 17 cells) driving MS. In addition, the presence of chemokines (CXCL13) and survival factors (BAFF and APRIL) in the CSF of patients with MS, promotes the formation of meningeal follicle like structures, in progressive phases but also in early RRMS \[[@CR46]\]. T cells and B cells may, therefore, play an equally important role in the immunopathology of MS \[[@CR47]\]. Axonal destruction is quite extensive (up to 60--80%) in all active WM lesions \[[@CR9], [@CR12], [@CR48]\] and the extend of axonal loss is related to the number of immune cells within the plaques \[[@CR49]\]. Activated immune cells (T and B cells) and microglia/macrophages release a number of pro-inflammatory cytokines (e.g. TNFa, INFγ), proteolyticenzymes (e.g. perforin, granzymes) and free radicals (e.g. nitric oxide, glutamate) that can directly damage axons \[[@CR50]\]. Additionally, axons may die secondarily, due to the loss of pre- and post-synaptic signals (i.e. dying-back and Wallerian axonal degeneration) in regions far from the lesion site \[[@CR43]\]. Active MS lesions are characterized by profound heterogeneity regarding their demyelination pattern \[[@CR51]\], which is persistent over time \[[@CR52]\]. The most commonly observed patterns are pattern II, which is a complement- and antibody-mediated demyelination, and pattern III, in which the initial event in lesion formation is a brief yet exorbitant oligodendrocyte injury \[[@CR53]\]. In other patients with RRMS, new lesions are associated with T cells, and activated microglia only. Pathologic heterogeneity across individuals in demyelination may imply different stimuli in the initial inflammation or different vulnerability to tissue loss across individuals \[[@CR54]\]. In WM lesions, inflammation and brain edema, demyelination, axonal loss, gliosis, and remyelination, all happen simultaneously \[[@CR35], [@CR55]\]. Brain edema which increases brain volume might bias atrophy measurements, but it resolves in the first few weeks after lesion formation. Notably, CNS has the capacity to use a great number of compensatory mechanisms (i.e. remyelination, redistribution of sodium channels, expression of neurotrophic factors etc.) to re-establish lost functioning to demyelinated foci \[[@CR48]\]. To conclude, tissue loss due to inflammation and demyelination maybe partly reversible in RRMS \[[@CR56], [@CR57]\], while tissue loss and axonal damage due to mechanisms other than inflammation is irreversible, and remains the major component of brain atrophy especially in the progressive disease stages. Mechanisms of late axonal loss (Fig. [1](#Fig1){ref-type="fig"}) {#Sec5} ---------------------------------------------------------------- While the destruction of CNS myelin is associated with clinical relapses, acute or late axonal loss is considered to be the main cause of permanent clinical disability in MS \[[@CR49]\]. Axons are more vulnerable to acute injury by inflammatory mediators, due to their shape and structure, compared to cell bodies or dendrites \[[@CR43]\], while thin axons (\< 2.5 μm in diameter) are mainly affected \[[@CR24], [@CR58]\]. Neurofilament light chain (NfL) protein is only expressed in neurons. It is an essential component of the axonal cytoskeleton, and reflects the axonal integrity and the stability of neurons. Under conditions of acute axonal transection, NfL are released and can be found as a result, in the cerebrospinal fluid (CSF) and blood of patients with MS. Of note, ultra high versus low blood NfL levels have been associated with MRI related (increased number of gadolinium enhancing or T2 lesion load, whole brain atrophy) and clinical measures (number of relapses, disability worsening) of disease activity and evolution and may, therefore, have prognostic value for patients and clinicians \[[@CR59]\].Fig. 1Mechanisms of late axonal loss. Molecular and cellular mechanisms driving neurodegeneration and atrophy. Key elements are considered to be: (1) Mitochondria Dysfunction: Inflammation in acute demyelinating lesions lead to respiratory protein complexes inhibition, mitochondrial injury and dysfunction, release of apoptosis-inducing factors and mitochondrial DNA deletions. In chronic inactive plaques, ionic imbalance, high energy demands and clonal expansion of defective mitochondria further impair oxidative damage. These mitochondrial alterations of functional impairment and structural damage lead to histotoxic hypoxia and energy failure and consequently to neurodegeneration. \[[@CR146]\] Upregulation of sodium channels, acid sensing ion channels and expression of maladaptive isoforms (Nav1.6 channels), paranodal (Caspr) and juxtparanodal (Kv1.2) protein lead to high energy demands, intra-axonal calcium accumulation, and subsequent axonal degeneration. (3) Glutamate Excitotoxicity: Increased glutamate production by activated microglial cells and lymphocytes, and impaired clearance by resident cells such as astrocytes lead to higher lever of glutamate. High levels of glutamate lead to over-activation of *N*-methyl-[d]{.smallcaps}-aspartate (NMDA) and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors (which are permeable for calcium and sodium ions) and subsequent calcium overload and oligodendocyte and neuron cell death. (4) Iron release: In MS lesions free iron \[Fe2+\] is released in the extracellular space leading to production of highly reactive hydroxyl molecules (OH^−^) by the Fenton reaction. Further, iron is released by activated glial cells, which become dystrophic and disintegrate, leading to a second wave of Fe^2+^ release Transected axons and ovoids are abundant in MS lesions \[[@CR9], [@CR27]\] but, abnormalities have also been reported in chronic inactive plaques, in normal appearing white matter (NAWM), and cortical areas, in which inflammation is less prominent \[[@CR48], [@CR57]\]. Therefore, additional mechanisms of axonal loss coexist with disease progression. It should be noted that these mechanisms have been postulated for both acute and late axonal loss (i.e. "late" signifying the absence of apparent inflammation): ### Ion overload {#Sec6} Several ion channels show compensatory changes a few weeks after demyelination \[[@CR60]\] a process that eventually promotes energy deficiency, and neurodegeneration. Aberrant expression of sodium channels, acid sensing ion channels, increased expression of maladaptive isoforms (i.e. Na~v~1.6 channels) \[[@CR61]\], paranodal (Caspr) and juxtparanodal (K~v~1.2) protein alterations \[[@CR62]\] have also been detected in WM lesions, in NAWM, and GM. Alternation in the expression of these ion channels lead to intra-axonal calcium accumulation, and subsequent axonal degeneration and atrophy, particularly in secondary progressive MS \[[@CR49]\]. ### Mitochondria dysfunction {#Sec7} There has been increasing interest in the role of mitochondrial injury in MS demyelination and axonal destruction. In acute inflammatory lesions mitochondrial nicotinamide adenine dinucleotide-hydrogen (NADH) oxidase \[[@CR63]\] and complex IV defects (COX I) have been described, in axons, oligodendrocytes, and astrocytes \[[@CR58]\]. In chronic inactive plaques, ionic imbalance and high energy demands result to swollen and dysfunctional mitochondria \[[@CR64], [@CR65]\], a phenomenon in which is partially reversed in remyelinating axons \[[@CR66]\]. There are also additional mtDNA deletions in GM structures of patients with SPMS \[[@CR67]\]. Furthermore, the respiration deficient neurons were diffusely distributed in the subcortical WM resulting in axonal loss in the absence of demyelination or inflammation. In oligodendrocytes, mitochondrial damage results in cell death and demyelination. Progenitor cells are also impaired, regarding their capacity to differentiate and produce myelin \[[@CR48]\]. Plus, genetic defects in mitochondrial genes potentiate MS lesions \[[@CR68]\]. From what can be deducted, mitochondrial dysfunction, in neurons and glia, is recognized as an important cause of atrophy and degeneration in MS and in other primarily neurodegenerative deceases such as Alzheimer's disease and Parkinson's disease \[[@CR65], [@CR69]\]. ### Iron dysregulation {#Sec8} Iron \[Fe\] loading accumulates with age and in patients with MS, it can further increase oxidative tissue loss. In the CNS, iron is mainly stored in oligodendrocytes, binding with ferritin. Under conditions of oxidative stress, such as MS lesions, when oligodendrocytes are destroyed, free iron \[Fe^2+^\] is released in the extracellular space and becomes an additional source of reactive oxygen species (Fenton reaction: Fe^2+^ + H~2~O~2~ = Fe^3+^ + OH. + OH−) \[[@CR48]\]. Further, iron is released by activated glial cells, which become dystrophic and disintegrate, leading to a second wave of Fe^2+^release. Diffuse T2hypointenselesions, which represent increased iron deposition \[[@CR70]\] are commonly found in patients with MS in cortical and deep GM areas (i.e. thalamus, basal ganglia, dentate nucleus \[[@CR71]--[@CR73]\] and WM plaques \[[@CR74]\]. Notably, T2 hypointensity has been associated with brain atrophy and early axonal loss \[[@CR73]\]. Furthermore, in progressive MS, there is a significant decrease in iron levels in NAWM \[[@CR75]\]. Iron is important for myelin synthesis and neurogenesis, and iron depletion in normal appearing tissue, may further promote diffuse axonal loss and CNS atrophy. ### Glutamate excitotoxicity {#Sec9} Several lines of evidence suggest that glutamate could also mediate injury to myelin, oligodendrocytes and neurons in the autoimmune experimental encephalomyelitis (EAE) model and in MS \[[@CR76]\]. Glutamate levels are elevated in CSF \[[@CR77]\], in the centre of active plaques, on the borders of chronic active lesions \[[@CR78]\], and in NAWM \[[@CR79]\]. There are two factors intertwining for glutamate accumulation: increased glutamate production by activated microglial cells and lymphocytes, and impaired clearance by resident cells such as astrocytes. High levels of glutamate lead to the over-activation of *N*-methyl-[d]{.smallcaps}-aspartate (NMDA) and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) \[[@CR80]\] receptors (permeable for calcium and sodium ions) and subsequent calcium overload and oligodendocyte and neuron cell death. Clinical correlates of brain atrophy {#Sec10} ==================================== Clinical symptoms and signs do not usually correlate with changes seen on conventional MRI measures (the "clinical-MRI paradox") \[[@CR81], [@CR82]\]. Whole brain atrophy, on the other hand, has a significant imaging association with physical disability as measured by Expanded Disability Status Scale (EDSS) score \[[@CR83]--[@CR88]\]. In a longitudinal study, whole brain (WB) and cortical atrophy as well as other MRI related metrics such as the enlargement of ventricular CSF spaces have been associated with disability progression over a 10 year follow up \[[@CR89]\]. Furthermore, brain volume changes during the first year after disease onset, estimated by PBVC, were the best predictor of future neurologic impairment \[[@CR90]\] regardless of the intermediate relapse rate \[[@CR91]\]. Increased brain volume loss (BVL) has been correlated disability progression, independent from the number of previous relapses or the T2 lesion load in RRMS \[[@CR92]\]. In a similar vein, when patients with clinically definite MS were compared to patients with clinically isolated syndrome (CIS), at baseline, all brain volume metrics, except for cortical GM, were significantly lower in the MS cohort. Over a mean follow-up period of about 3 years, the annual PBVC values were significantly lower in CIS patients when compared to the MS cohort \[[@CR93]\]. Neuropsychological impairment, affecting mental speed processing, episodic memory, executive functions and attention, may be present in up to 50% of patients with MS \[[@CR94]\] and has been found to occur early in the disease course \[[@CR95]\]. Changes of brain parenchymal fraction (BPF) have been shown to predict cognitive impairment over 2 years in patients with early MS \[[@CR96]\]. Cortical atrophy was the best predictor of poor cognitive functioning, even when mild impairment was detected. Poorcognitive functioning has been associated with significant cortical thinning \[[@CR97]\], especially in the fronto-parietal cortical and subcortical regions \[[@CR98]\]. Pravatà et al. \[[@CR98]\] specifically reported that the thinning of the right precuneus and high T2 lesion load were the best predictors of cognitive impairment. Strong correlations have also been reported between cognitive impairment and thalamic atrophy \[[@CR80], [@CR98], [@CR99]\]. Not surprisingly, patients with brain atrophy and higher education or high "cognitive reserve" are relatively protected against cognitive decline \[[@CR100]\]. Other clinical aspects of CNS atrophy include mood and personality disorders (i.e. euphoria, disinhibition, aggression, major depressive disorder) \[[@CR101]\] autonomic dysfunction and sexual disorders \[[@CR85]\]. Fatigue has been reported to be associated with GM atrophy in frontal regions \[[@CR102]\] and depressed patients were found to present selective cortical thinning in the fronto-temporal regions, while the frontal thinning was found to be the best predictor for depression in MS patients \[[@CR98]\]. Taken together, this growing body of evidence suggests that brain atrophy is a valid and sensitive measure of disease burden and progression in MS patients and may effectively be used in routine clinical practice and treatment trials. Effect of disease modifying treatments (Tables [1](#Tab1){ref-type="table"}, [2](#Tab2){ref-type="table"} and [3](#Tab3){ref-type="table"}) {#Sec11} =========================================================================================================================================== Approved DMTs and brain volume outcomes {#Sec12} --------------------------------------- The need of agents to control the inflammatory process in multiple sclerosis pathology is obvious, but the need for medications to halt brain atrophy progression and neurodegeneration is also evident. Currently approved treatments for MS differ in their effects on brain atrophy \[[@CR103]\] (Table [1](#Tab1){ref-type="table"} for the first line therapies, Table [2](#Tab2){ref-type="table"} for the second line therapies and Table [3](#Tab3){ref-type="table"} for the emerging therapies).Table 1First line therapies and their effect on brain volume loss (BVL)ReferencesDMT and trial designClinical trialBaseline/MRI cohortsType of MSMain effect on brain atrophyRudick et al. \[[@CR83], [@CR87]\]IFN β-1a i.m. 30 mcg weekly vs placeboPhase III MS.C.RG 2 years140 IFN β-1a n = 68, placebo n = 72RRMSPercent change inbrain parenchymal fraction was lower in IFN β-1a treated patients compared to placebo, during the second year of treatment (p = 0.03) but not the first (p = 0.71)Fisher et al. \[[@CR104]\]IFN β-1a i.m. 30 mcg weekly vs placeboRetrospective analysis of Phase III MS.C.RG 2 years131 IFN β-1a n = 62, placebo n = 69RRMSIFN β-1a significantly preserved GM \[[@CR4]\] atrophy (p = 0.03) and whole brain atrophy (p = 0.04) during the second year of treatment, but not WM atrophy (at any point)Fillipi et al. \[[@CR106]\]IFN β-1a s.c. 22 μg weekly vs placeboPhase III ETOMS 2 years262 IFN β-1a n = 131, placebo n = 132CISSignificant reductions in PBVC the IFN β-1a treated arm (p = 0.0031) compared to placebo, from baseline to second yearDe Stefano et al. \[[@CR173]\]IFN β-1a s.c. 44mcg TIW vs placebodouble-blind and rater-blind phase IMROVE 40 weeks180 (double-blind phase) IFN β-1a n = 120, placebo n = 60RRMSNon-significant differences in mean \[[@CR102]\] PBVC between treatment groups (placebo: − 0.24% \[0.48%\]; IFN β-1a: − 0.22% \[0.54%\]; p = 0.76) at week 16 (end of double-blind phase)De Stefano et al. \[[@CR105]\]IFN β-1a s.c. 44 mg TIW vs once a week vs placeboPhase III REFLEX 2 years517 IFN β-1a TIW n = 171, IFN β-1a, once a week n = 175, placebo n = 171CISNo differences in BVL (from baseline to 2 years) in patients receiving once or three times a week IFNβ-1a vs placebo. The greatest loss was in the TIW IFN β-1a group compared with the once a week IFN β-1a and placebo groupsKappos et al. \[[@CR108]\]IFN β-1a s.c. 44 or 22 mcg TIW vs placebo (2 years); then open label (4 years); long term follow up (2 years)Phase III PRISMS \~ 8 years382 44 mcg sc TIW n = 136, 22 mcg sc TIW n = 123, placebo n = 123RRMSNon-significant differences in median BPV (from baseline to long term follow-up and each study period therein) for all treatment armsHardmeier et al. \[[@CR174]\]IFN β-1a i.m. 30 μg or 60 μgRetrospective of The European IFNb-1a Dose- Comparison Study 3 yearsannual MRI cohort n = 386, frequent MRI cohort n = 138RRMSThe greatest BVL took place during the first 4 months of therapy in frequent MRI cohort (from baseline to month 4, p \< 0.05). Non-significant reduction in the brain atrophy in the 2nd and 3rd year of treatmentMolyneux et al. \[[@CR175]\]INF β-1b s.c. 8 MIU every other day vs placeboPhase III 3 years92 INF β-1b n = 48, placebo n = 44SPMSNot significant effect of treatment with INF β-1b on cerebral volume loss (p = 0.343, from baseline to 3 years) compared with placebo.Kappos et al. \[[@CR176], [@CR177]\]INF β-1βs.c. 250 μg every other day (early vs delayed treatment)Extension study of the Phase III BENEFIT trial (3 and 5 years follow up)Follow-up phase n = 418 early treatment n = 261, delayed treatment n = 157 5-year completers n = 358 early treatment n = 235, delayed treatment n = 123CISMarginal, non-significant differences between early and delayed treatment (p = 0.15, from baseline to 3 years, p = 0.121 from baseline to 5 years)Calabresi et al. \[[@CR111]\]Peginterferon b-1a s.c. 125 μg Q2 W vs Q4 W vs placeboPhase III ADVANCE 1 year, then open label1512 PEG-IFN β-1a 125 μg Q2 W n = 512, PEG-IFN β-1a 125 μg Q4 W n = 500, placebo n = 500RRMSCore study: During the first 6 months of treatment there was a significant "pseudoatrophy" effect (PEG-IFN β-1a 125 μg Q2 W vs placebo, p = 0.0170)Baseline to year 1: Νo significant differences on whole brain volume between groups (Q2Wvs placebo p = 0.0841; Q4 W vs placebo p = 0.3747)Arnold et al. (F2069, 1rst EAN Congress 2015) \[[@CR112]\]Peginterferon b-1a s.c. 125 μg Q2 W vs Q4 WExtension study of Phase III ADVANCE 2 yearsAt week 96/569 PEG-IFN β-1a 125 μg Q2 W n = 384, PEG-IFN β-1a 125 μg Q4 W n = 185 (delayed treatment)RRMSFrom week 24 to week 96, the delayed treatment PEG-IFN β-1a 125 μg Q4 W n = 185 demonstrated a significantly greater decrease in whole brain volume compared with the Q2 W group (p = 0.0034)Sorensen et al. \[[@CR110]\]INF β-1a s.c. 44 μg plus Methylprednisolone orally 200 mg or placeboorallyPhase III NORMI.M.S 2 years110 INF β-1a and oral methylprednisolone n = 54, IFN β-1a and placebo n = 56RRMSMean changes in normalized brain parenchymal volume favored pulsed treatment with oral methylprednisolone combined with INF β-1a vs INFβ-1a monotherapy, but the benefit was not significant (p = 0.25) between baseline and week 96Ravnborg et al. \[[@CR109]\]INF β-1a i.m. 30 μg once weekly plus Methylprednisoloneorally 500 mg daily (3 consecutive days per month for 3--4 years) or placeboPhase III MECOMBIN 3 years338 INF β-1a plus placebo n = 167, INF β-1a plus methylprednisolone n = 171RRMSThe study showed no effect on brain parenchymal volume (p = 0.58) or change in normalized brain volume (p = 0.52)Comi et al. \[[@CR114]\]GAs.c. 20 mg daily vs PlaceboPhase III PreCISe 2 years481 GA n = 243, Placebo n = 238CISNo significant difference in percentage change from baseline to last observed value in brain volume between the treatment groups (− 0.33% in GA vs − 0.38% in placebo)Comi et al. \[[@CR115]\]GAs.c. 20 mgdailyvs placeboopen-label, extension phase of Phase III PreCISe 2 years409 early-treatment group n = 198, placebo (delayed-treatment) n = 211CISSignificant reduction of BVL in early versus delayed treatment onset (28% reduction, p = 0.0209)Ge et al. \[[@CR116]\]GAs.c. 20 mg daily vs placeboPhase III The US GA study 2 years27 GA treated n = 14, placebo n = 13RRMSGA significantly reduced the rate of BVL (77% reduction) in the 2-year treatment period (p = 0.007) compared with placeboRovaris et al. \[[@CR119]\]GA s.c. 20 mg daily vs placebo for 9 months, then GA open-labelPhase III European/ Canadian GA trial 18 months227 GA n = 113, placebo n = 114RRMSDuring the double-blind, placebo-controlled phase of the study, GA treatment did not have any measurable impact on the absolute or percentage change of BV (from baseline to 9 months, p = 0.88)In the subsequent open-label phase, early GA treatment showed a 40% reduction in the rate of brain atrophy (from 9th to 18th month)Rovaris et al. \[[@CR118]\]GA s.c. 20 mg dailyExtension of the Phase III European/ Canadian GA trial 5 years142 Early treatment n = 73 Delayed treatment n = 69RRMSBaseline to 5 years: Non-significant differences in median PBVC in early vs delayed treatment groups.Comi et al. \[[@CR113]\]GA s.c. 20 mg vs 40 mg (dose comparison)Phase III FORTE 1 year1155 GA 20 mg n = 586, GA 40 mg n = 569RRMSPBVCs were similar in both groups (p = 0.423). Higher dose of GA did not have a clear-cut impact on brain volume loss. Slower atrophy rates, compared with the Eur/Canadian GA trialKhan et al. \[[@CR178]\]GA s.c. 40 mg TIW vs placeboPhase III GALA 1 year1263 GA 40 mg TIW n = 840, placebo n = 423RRMSThe percentage change in brain volume (from baseline to 1 year) was not statistically different between treatment arms (− 0.706 with GA vs − 0.645 with placebo; p = 0.2058)Lublin et al. \[[@CR179]\]INF β-1a i.m. 30 mg weekly, GA s.c. 20 mg dailyPhase III CombiRx 3 years790 IFN + GA n = 388, IFN n = 187, GA n = 215RRMSCombination treatment was not superior to either INF β-1a or GA agents alone (CSF volume change from baseline to month 36; IFN β-1a + GA vs IFN, p = 0.008, INF β-1a vs GA p = 0.48). Whole brain tissue loss was reflected by the change in normalized CSF from baselineO'Connor et al. \[[@CR180]\]INF β- 1b s.c. 250 μg or 500 μg, every other day or GA s.c. 20 mg dailyPhase III BEYOND 2 years2244 IFN β-1b 500 μg n = 899, IFN β-1b 250 μg n = 897, GA n = 448RRMSNon-significant differences between treatment groupsHigh dose INF β-1b was not superior to the standard dose (500 μg IFN β-1b vs 250 μg IFN β-1b p = 0.74). Both doses of IFN β-1b had similar measurable brain volume (BV) effect as compared with GA (500 μg IFN β-1b vs GA p = 0.33; 250 μg IFN β-1b vs GA p = 0.46). During year 1, patients under IFN β-1b had a significantly greater reduction in mean brain volume than did patients treated with GA (250 μg IFN β-1b vs GA p = 0.02; 500 μg IFN β-1b vs GA p = 0.007)Arnold et al. \[[@CR136]\]DMF orally 240 mg BID vs TID vs placeboPhase III DEFINE 2 years540 DMF BID n = 176, DMF TID n = 184, Placebo n = 180RRMSSignificant results for the DMF BID versus placebo on brain atrophy, from either baseline or 6 months to second year (baseline to 2 years p = 0.0449, 6 months to 2 years p = 0.0214). Non-statistically results for the DMF TID dose groupMiller et al. \[[@CR137]\]DMF orally 240 mg BID vs TID vs GA 20 mg once daily vs placeboPhase III CONFIRM 2 years681 DMF BID n = 169, DMF TID n = 170, GA n = 175, placebo n = 167RRMSAt 2 years, PBVC favored DMF BID, but not TID or GA, compared to placebo (BID vs placebo; p = 0.0645; TID vs placebo; p = 0.2636; GA vs placebo p = 0.8802)Kappos et al. (P7.243, AAN) \[[@CR181]\]DMF orally 240 mg BID vs TID vs placebo8 year follow-up study of Phase III ENDORSE Ongoingyear 1/464 DMF BID n = 197, GA n = 88, placebo n = 179RRMSThere was no significant effect in brain volume loss for the placebo/DMF and the GA/DMF groups relative to the group treated continuously with DMF BID (BID/BID group) (median PVC, from baseline to 5 years, BID/BID vs placebo/DMF p = 0.1165, BID/BID vs GA/DMF p = 0.3436)Miller et al. \[[@CR129]\]Teriflunomide orally 7 or 14 mg once-daily vs placeboPhase III TOPIC 4 years614 Teriflunomide 14 mg n = 214, Teriflunomide 7 mg n = 203, placebo n = 197CISNo significant differences were recorded for brain atrophy (SIENA). (Mean change from baseline at week 108 vs placebo, 14 mg p = 0.4495; 7 mg p = 0.4462)O'Connor et al. \[[@CR130]\]Teriflunomide orally 7 or 14 mg once-daily vs placeboPhase III TEMSO 2 years1086 Teriflunomide 14 mg n = 358, Teriflunomide 7 mg n = 365, placebo n = 363RRMSNo effect on relative BPF change among the study groups (from baseline to 2 years: Teriflunomide 7 mg vs. placebo p = 0.19; Teriflunomide 14 mg vs. placebo p = 0.35)Wolinsky et al. \[[@CR131]\]Teriflunomide orally 7 or 14 mg once-daily vs placeboPost hoc analysis of Phase III TEMSO 108 weeks1088 Teriflunomide 14 mg n = 359, Teriflunomide 7 mg n = 366, placebo n = 363RRMSThere was a significant decrease in WM fraction (from baseline to108 weeks) for both doses of Teriflunomide (WMF change 14 mg vs placebo p = 0.0002; 7 mg vs placebo p = 0.0609)Radue et al. (P3-089 AAN 2016) \[[@CR134]\]Teriflunomide orally 7 or 14 mg once-daily vs placeboPost hoc analysis of Phase III TEMSO and TOWER 9 years969 808 baseline and week 48, 709 baseline and week 108RRMSSignificant gain in brain volume loss, by using an alternative method (SIENA). Median PVC, from baseline to first year, Teriflunomide 14 mg vs placebo p = 0.0001; Teriflunomide 7 mg vs placebo p = 0.0011; from baseline to second year: Teriflunomide 14 mg vs placebo p = 0.0001; Teriflunomide 7 mg vs placebo p = 0.0019)Sprenger et al. P3.047 \[[@CR132]\]Teriflunomide orally 14 mg once-daily vs placeboPost hoc analysis of Phase III TEMSO and TOWER 2 years969 808 first year, 709 s yearRRMSTeriflunomide resulted in lower atrophy rate in patients with and without disability progression vs placebo. Without disability progression: Median PBVC, from baseline to first year, Teriflunomide 14 mg vs placebo (22%) p = 0.0128 and from baseline to second year (23%) p = 0.0129 With disability progression: Median PBVC, from baseline to first year, Teriflunomide 14 mg vs placebo (69%) p = 0.0037 and from baseline to second year (44%) p = 0.0043Wuerfel et al. P3.052 \[[@CR135]\]Teriflunomide orally 14 mg once-daily vs placeboPost hoc analysis of Phase III of TEMSOYear one cohort. 0--2 in previous 2 year: Teriflunomide 14 mg n = 191, placebo n = 197 2--3 in previous 2 year Teriflunomide 14 mg n = 195, placebo n = 198RRMSSignificant impact on median PBVC regardless of the level of disease activity (prior relapse rate)Patients with few prior relapses (0--2 in previous 2 years): Baseline to year 1: Teriflunomide 14 mg vs placebo, relative change in percentage brain volume 40% p = 0.0001. Year 1to year 2: relative change 36%, p = 0.0001. This finding was confirmed in patients with a greater number of relapses (2--3 in previous 2 years): p = 0.0018 at year 1 and p = 0.0067 at year 2Freedman et al. (P734 ETCRIMS 2016) \[[@CR133]\]Teriflunomide orally 7 or 14 mg once-daily vs placeboSubgroup analysis of Phase III TEMSO971 treatment-naïve n = 704, 1 Prior DMT n = 208, ≥2 Prior DMTs n = 57RMSPositive results on median PBVC regardless of treatment history. PVC change from baseline to year 1, Teriflunomide 14 mg No prior DMT vs placebo p = 0.0025; baseline to year 2: p = 0.0109; Teriflunomide 14 mg prior DMT vs placebo p = 0.0119, baseline to year 2: p = 0.0109. PVC change from baseline to year 1: Terilunomide 7 mg No prior DMT vs placebo p = 0.0002; baseline to year e: p = 0.0089. Teriflunomide 7 mg prior DMT vs placebo p = 0.0119, baseline to year 2: p = 0.0109mg: milligrams; mcg: micrograms; μg: micrograms; vs: versus; PBVC: percentage of brain volume change; BPF: brain parenchymal fraction; BPV: brain parenchymal volume; SIENA: structural imaging evaluation using normalization of atrophy; i.m.: intramuscular; s.c: subcutaneous; i.v.: intravenous; ΤΙW: three times weekly; SD: standard deviation;Q2W: once every 2 weeks; Q4W: once every 4 weeks; BID: twice daily; TID: thrice daily; DMT: disease modifying therapies; CSF: cerebrospinal fluid; PVC: percentage volume change Table 2Second line therapies and their effect on brain volume loss (BVL)ReferencesDMT and trial designClinical trialBaseline/MRI cohortsType of MSMain effect on brain atrophyKappos et al. \[[@CR122]\]Fingolimod orally 0.5 mg or 1.25 mg once daily vs placebo for 2 years, then FTY open-labelPhase III FREEDOMS 2 years1272 Fingolimod 1.25 mg n = 429, Fingolimod 0.5 mg n = 425, placebo n = 418RRMSSignificant reductions in the rate of brain volume loss were detected as early as 6 months for the 12 mg Fingolimod treatment group (PBVC values from baseline to 6 months, 1.25 mg Fingolimodvs placebo p = 0.003; 0.5 mg Fingolimodvs placebo p = 0.006) and remained significant at 24 months (P \< 0.001 in all comparisons)Kappos et al. \[[@CR124]\]Fingolimod orally 0.5 mg or 1.25 mg once daily (FTY open label)Extension of Phase III FREEDOMS 2 years920RRMSSignificantly lower atrophy rates in the continuous Fingolimod groups relative to the combined switch group, over 4 years (Continuous Fingolimod 0.5 mg p = 0.0013; Continuous Fingolimod 1.25 mg p = 0.001)Patients who switched to Fingolimod 0.5 mg during the extension study experienced significant improvements in rates of brain volume decline (Placebo---Fingolimod 0.5 mg p = 0.0084, months 24--48 vs months 0--24)Cohen et al. \[[@CR121]\]Fingolimod orally 1.25 or 0.5 mgonce daily vs INF β-1a i.m. 30 μg (1 year, then open-label)Phase III TRANSFORMS 1 year1280 Fingolimod 1.25 mg n = 420, Fingolimod 0.5 mg n = 429, INF β-1a N = 431RRMSCompared to i.m. INF β-1a, patients treated with Fingolimod presented less brain volume loss, over 1 year (all p \< 0.001)Khatri et al. \[[@CR125]\]Fingolimod orally 1.25 or 0.5 mg once dailyExtension of Phase III TRANSFORMS 2 years799 INF β-1a to 0.5 mg Fingolimod n = 124, INF β-1a to 1.25 mg Fingolimod n = 130. Continuous 0.5 mg Fingolimod n = 290, Continuous 1.25 mg Fingolimod n = 255RRMSPatients switching from INF β-1a to Fingolimod (either 1.25 or 0.5 mg) reduced their brain volume decrease (PBVC: months 13--24 vs months 0--12, p = 0.006 for the INF β-1a to 0.5 mg Fingolimod switch group p = 0·007 for the INF β-1a to 1.25 mg FTY720 switch group. No further gain in BVL for patients on continuous Fingolimod treatment (p values non-significant at months 13--24 vs months 0--12)Calabresi et al. \[[@CR120]\]Fingolimod orally 1.25 or 0.5 mg once daily vs placeboPhase III FREEDOMS II 1 year1083 Fingolimod 1.25 mg n = 370, Fingolimod 0.5 mg n = 358, placebo n = 355RRMSPatients given Fingolimod had decreased brain volume loss compared with those given placebo from baseline to months 6 (Fingolimod 1.25 mg vs placebo, p = \<0.0001; Fingolimod 0.50 mg vs placebo, p = 0.012) 12 (Fingolimod 1.25 mg vs placebo, p = \<0.0001; Fingolimod 0.50 mg vs placebo, p = 0.004) and 24 (Fingolimod 1.25 mg vs placebo, p = \<0.0001; Fingolimod 0.50 mg vs placebo, p = 0.013). (Total treatment effect on PBVC vs placebo p \< 0·0001 and p = 0·0002 respectively)Cohen et al. \[[@CR123]\]Fingolimod orally 1.25 or 0.5 mg once daily IFN β-1a i.m. 30 μg once a weekExtension of Phase III TRANSFORMS 4.5 yearsFingolimod 0.5 mg n = 356, IFN β-1a-switch Fingolimod 0.5 mg n = 167, Fingolimod 1.25 mg n = 330, IFN β-1a switch fingolimod 1.25 mg n = 174RRMSNon-significant long term benefit on mean PBVC (from baseline to 4.5 years): continuous-fingolimodvs IFN β-1a-switch group −1.01% (−0.8) vs −0.96% (−0.8); p = 0.937. The PBVC from baseline to month 12 was reduced significantly by fingolimod compare to IFN β-1a (p \< 0.0001) and the low rate was maintained through the study completionLublin et al. \[[@CR126]\]Fingolimod orally 0.5 mg once daily vs placeboPhase III INFORMS 3 years714 Fingolimod 0.5 mg n = 293, placebo n = 421PPMSIn patients with primary progressive MS, percentage change in brain volume did not differ between Fingolimod and placebo groups (p = 0.673)Miller et al. \[[@CR139]\]Natalizumab i.v. 300 mg every 4 weeks vs placeboPhase III AFFIRM 2 years942 Natalizumab n = 627, placebo n = 315RRMSOverall, not significant effect of treatment with Natalizumabvs placebo (mean percentage change in BPF, 0.80% vs 0.82%, p = 0.822, from baseline to 2 years). During the first year, natalizumab-treated patients presented greater BVL compared to placebo (0.56% vs 0.40%, p = 0.002) but the rate of brain atrophy was significantly less in natalizumab-treated patients over the second year of treatment (0.24% vs 0.43% p = 0.004)Radue et al. \[[@CR140]\](IFN β-1a i.m.30 μg + Natalizumab i.v.300 mg every 4 weeks) vs IFN β-1a i.m. 30 μg + placebo once weeklyPhase III SENTINEL 2 years1171 IFN β-1a + natalizumab n = 589, IFN β-1a + placebo n = 582RRMSFrom baseline to second year, no significant differences were reported between the 2 treatment arms regarding change in BPF (p = 0.926). During the first year, there was a significant reduction in BPF in the Natalizumab treated arm (p = 0.058), but lower rates during the 2nd year of treatment (-- 0.31% versus -- 0.40%; p = 0.020)mg: milligrams; mcg: micrograms; μg: micrograms; vs: versus; PBVC: percentage of brain volume change; BPF: brain parenchymal fraction; BPV: brain parenchymal volume; SIENA: structural imaging evaluation using normalization of atrophy; i.m.: intramuscular; s.c: subcutaneous; i.v.: intravenous; ΤΙW: three times weekly; SD: standard deviation;Q2W: once every 2 weeks; Q4W: once every 4 weeks; BID: twice daily; TID: thrice daily; DMT: disease modifying therapies; CSF: cerebrospinal fluid; PVC: percentage volume change Table 3New or emerging therapies and their effect on brain volume loss (BVL)ReferencesDMT and trial designClinical trialBaseline/MRI cohortsType of MSMain effect on brain atrophyComi et al. \[[@CR157]\]Laquinimod orally 0.6 mg once daily vs placeboPhase III ALLEGRO 2 years1106 Laquinimod n = 550, placebo n = 556RRMSLaquinimod had a significant effect on reducing brain volume loss vs placebo (p \< 0.001, from baseline to 2 years)Vollmer et al. \[[@CR158]\]Laquinimod orally 0.6 mg once daily vs IFN β-1a i.m. 30 μg once weekly vs oral placeboPhase III BRAVO 1 year1331 Laquinimod n = 434, IFN β-1a i.m. n = 4 47, placebo n = 450RRMSRobust effects on reducing brain atrophy are replicated for Laquinimod (p \< 0.001, from baseline to year 1), whereas IFN β-1a showed no benefit at all (non-significant. increased BVL 11% vs placebo, p = 0.14)Cohen et al. \[[@CR147]\]Alemtuzumab i.v. 12 mg (once per day for 5 days at baseline and once per day for 3 days at 12 months) vs INF β-1a s.c. 44 μg TIWPhase III CARE-MS I 2 years563 Alemtuzumab n = 376, INF β-1a n = 187RRMSMedian change in brain parenchymal fraction was less in Alemtuzumab (− 0.867%) was compared with INF β-1a (1.488%), p \< 0.001)Coles et al. \[[@CR148]\]Alemtuzumab i.v. 12 mg once per day vs 24 mg once per day (once per day for 5 days at baseline and for 3 days at 12 months) vs INF β-1a s.c. 44 μg TIWPhase III CARE-MS II 2 years628 Alemtuzumab 12 mg n = 426, INF β-1a n = 202RRMSCompared to 44 μgsc IFN β-1a (− 0.810%), alemtuzumab-treated (− 0.615%) patients showed less reduction in median parenchymal brain fraction during the first year of the trial (p = 0.01)Traboulsee et al. P1181 ECTRI.M.S \[[@CR6]\]Alemtuzumab i.v. 12 mg once daily received 2 annual courses (on 5 consecutive days at baseline and on 3 consecutive days 12 months later). Patients could receive additional treatment with alemtuzumab (12 mg on 3 consecutive days ≥ 1 year after the most recent course) during the extension studyExtension of Phase III CARE-MS I, CARE-MS II 4 years93% of CARE-MS I n = 325, 88% of CARE-MS II n = 393RRMSDurable MRI positive outcomes (i.e. sustained low brain atrophy rates, in the absence of continuous treatment with Alemtuzumab or other DMTs during the follow up period)Coles et al. \[[@CR150]\]Alemtuzumab i.v.(12 mg on 3 consecutive days) Alemtuzumab-treated patients who completed CARE-MS II could enroll in the extension and receive, at the investigator's discretion, additional alemtuzumab courses (12 mg on 3 consecutive days) ≥ weeks after the most recent course, if they had evidence of MS disease activity. Patients who received s.c. IFN-b-1a for 2 years in the core study could also enroll in the extension and switch to alemtuzumab treatment; results for these patients will be reported separatelyCARE-MS II 5 years follow-upMost alemtuzumab-treated patients (92.9%) who completed CARE-MS II entered the extension; 59.8% received no alemtuzumab retreatmentRRMSMedian yearly BVL remained low in the extension (years 1--5: − 0.48%, − 0.22%, − 0.10%, − 0.19%, − 0.07%). Yearly BVL rate continued to decrease in year 3 compared with the core study, remaining low in years 4 and 5. Median BPF change from baseline to year 5 was − 0.855%Arnold et al. (P558, ECTRI.M.S 2015) \[[@CR151]\]Daclizumab s.c 150 mg every 4 weeks vs INF β-1ai.m. 30mcg once weeklyPost hoc of Phase III DECIDE 2 years1806 Daclizumab n = 899, INF β-1a n = 907RRMSDaclizumab showed a significant effect in limiting the rate of brain atrophy vs IFN β-1a, between baseline and week 96 (p \< 0.0001), week 0 and week 24 (p = 0.0325) and between week 24 and week 96 (p \< 0.0001)Montalban et al. \[[@CR155]\]Ocrelizumab i.v. 600 mg (two 300 mg infusions 14 days apart) vs placeboPhase III ORATORIO732 Ocrelizumab 600 mg, n = 488, placebo n = 244PPMSOcrelizumab reduced the rate of whole brain volume loss from week 24 to week 120 by 17.5%120 (p = 0.0206) compared with placeboArnold et al. \[[@CR154]\]Ocrelizumab i.v 600 mg.every 24 weeks vs INF β-1as.c. 44 mcg TIWPhase III OPERA I 96 weeks821 Ocrelizumab n = 410, IFN β-1a n = 411RMSOcrelizumab reduced brain volume loss compared with INF β-1a. (p \< 0.001 from baseline to 96th week and p = 0.0042 from 24th to 96th week)Arnold et al. \[[@CR154]\]Ocrelizumab i.v 600 mg.every 24 weeks vs INF β-1as.c. 44 mcg TIWPhase III OPERA II835 Ocrelizumab n = 417, IFN β-1a n = 418RMSOcrelizumab reduced brain volume loss compared with INF β-1a. \[p = 0.001 from baseline to 96th week and p = 0.09 (non-significant) from 24th to 96th week\]De Stefano et al. \[[@CR162]\]Cladribine3.5 mg/kg or Cladribine5.25 mg/kgvs placeboPhase III CLARITY1025 Cladribine 3.5 mg/kg n = 336, Cladribine 5.25 mg/kg n = 351, placebo n = 338RMSPatients treated with cladribine had significantly less annualized brain atrophy over 2 years compared with patients receiving placebo. At 18 months, patients treated with cladribine had 20% reduction in brain atrophy compared with patients receiving placeboIn patients under cladribine tablets 3.5 mg/kg (− 0.56% ± 0.68, p = 0.010) and 5.25 mg/kg (− 0.57% ± 0.72, p = 0.019), the annualized PBVC was reduced compared with placebo (− 0.70% ± 0.79)mg: milligrams; mcg: micrograms; μg: micrograms; vs: versus; PBVC: percentage of brain volume change; BPF: brain parenchymal fraction; BPV: brain parenchymal volume; SIENA: structural imaging evaluation using normalization of atrophy; i.m.: intramuscular; s.c: subcutaneous; i.v.: intravenous; ΤΙW: three times weekly; SD: standard deviation;Q2W: once every 2 weeks; Q4W: once every 4 weeks; BID: twice daily; TID: thrice daily; DMT: disease modifying therapies; CSF: cerebrospinal fluid; PVC: percentage volume change In general, studies of traditional injectable treatments have not exerted robust beneficial effects in the rate of brain atrophy. Intramuscular IFN-β-1a produced lower rates of brain volume loss (BVL) when compared to placebo during the second year of treatment in relapsing--remitting MS patients (− 0.23% vs − 0.51%; p = 0.03) \[[@CR83], [@CR104]\]. However, the subcutaneous (sc) IFN-β-1a produced inconsistent results in both CIS and RRMS patients \[[@CR105]--[@CR108]\]. BV data for intramuscular INF-β-1a in CIS patients and for subcutaneous INF-β-1b in relapsing MS patients has not been made available to date. The addition of monthly oral methylprednisolone pulses to subcutaneous interferon beta-1a treatment provided no further gain in normalized BV change in two published trials against placebo \[[@CR109], [@CR110]\]. The approved long-acting pegylated interferon beta-1a has only shown limited and inconclusive evidence for a beneficial effect on BV change in RRMS \[[@CR111], [@CR112]\]. A possible delayed effect in reducing brain atrophy has been reported for Glatiramer acetate \[GA\] \[[@CR113]--[@CR119]\]. In the PReCISe clinical trial, GA failed to show an immediate effect on brain volume outcomes versus placebo (− 0.38% vs 0.33%), but the subsequent open label phase of the trial showed a clear--cut benefit on PBCV for the early treatment group, when compared to patients with delayed treatment onset (40% reduction, p = 0.0209) \[[@CR114], [@CR115]\]. In relapsing--remitting MS, data from the extension phase of the European/Canadian GA trial come back as negative \[[@CR118]\]. Available oral therapies (Fingolimod, Teriflunomide, Dimethyl fumarate) have shown various effects on BV decline. Fingolimod has been reported consistentin reducing median PBVC by approximately 30 to 45% versus placebo or IFNβ-1a, in its three phase III clinical trials \[[@CR120]--[@CR122]\] and their extensions \[[@CR123]--[@CR125]\]. Of note, this reduction was observed as early as 6 months after treatment onset \[[@CR120], [@CR122]\]. In the extension phase of the TRANFORMS trial, patients switching from intramuscular (i.m.) INF β1a to FTY720 slowed their median PBVC, and patients continuing on FTY720 sustained low atrophy rates, over the following 4.5 years of therapy \[[@CR123]\]. However, no similar effects were reproduced in patients with the primary progressive form of MS, a finding that that could have otherwise strengthened the evidence for a direct action of fingolimod on brain cellular components \[[@CR126]\]. Finally, further condoning the aforementioned observations, in a study by Yousuf et al. \[[@CR127]\], cortical GM, alongside T2 lesion volume, remained stable in the cohort treated with fingolimod, as compared to the untreated group, where it decreased and increased respectively, in the first 2 years of treatment. Regarding Teriflunomide \[[@CR128]\], brain volume outcomes have been reported for clinically isolated syndrome and relapsing- remitting MS in the TOPIC and TEMSO clinical trials respectively. Both doses of 7 mg or 14 mg failed to show a clear effect on slowing BVL when compared to placebo \[[@CR129], [@CR130]\]. However, when tissue specific volume changes were examined a significant reduction in the rate of WM loss was detected for the 14 mg teriflunomide treatment arm versus placebo \[[@CR131]\]. Similar results have recently been reported in 4 retrospective analyses of TOWER and TEMSO trials when an alternative method of brain loss evaluation was implemented \[[@CR132]--[@CR135]\]. Dimethyl fumarate (DMF/BG12) showed a 21% reduction in BVL compared to placebo in the DEFINE study (the 240 mg twice daily regimen only) \[[@CR136]\] and produced only marginal but beneficial effects in BVL reduction in the CONFIRM study \[[@CR137]\]. A recent pilot study of 20 patients with RRMS showed a protective effect of DMF treatment in whole brain atrophy (PBVC: − 0.37 ± 0.49% vs. − 1.04 ± 0.67%, p = 0.005) and putamen atrophy (− 0.06 ± 0.22 vs. − 0.32 ± 0.28 ml, p = 0.02), but no effect on other subcortical volumes or total GM atrophy \[[@CR138]\]. Natalizumab, a monoclonal antibody against the cell adhesion molecule a 4-integrin, in two pivotal clinical trials was found to increase the rate of BVL in the first year of treatment and then significantly reduced it when compared to the placebo in the second year \[[@CR139], [@CR140]\]. Post--marketing observational studies confirmed that most of the BVL occurring while on Natalizumabtherapy takes place during the first months of therapy, and that it primarily involves WM volume changes \[[@CR141], [@CR142]\]. One trial has shown superiority of Natalizumab over conventional MS therapies (IFN-β and GA) and placebo regarding cortical atrophy \[[@CR143]\]. Recently, treatment with Natalizumabdid not affect the loss of brain volume compared to placebo in secondary progressive MS patients (ASCENT) \[[@CR144]\]. Τhe study by Arpín et al. \[[@CR145]\] also suggests a neuroprotective effect of Natalizumab, after the measurement and comparison of the corpus calosum index, and the absence of brain atrophy in several patients under treatment during the follow up. Alemtuzumab, a monoclonal antibody against cells that express the CD52, has demonstrated greater MRI and clinical improvement in comparison to IFNb-1a in its three pivotal studies in active relapsing MS patients \[[@CR146]--[@CR148]\]. Additionally, most patients remained free of disability and MRI progression, for the following 6 years of the initial treatment \[[@CR6]\]. Brain atrophy measures showed that brain parenchymal fraction was smaller in Alemtuzumab compared to the INF β-1a treatment arm either in treatment naïve patients \[[@CR149]\] or in participants who had relapsed on prior therapy \[[@CR147]--[@CR149]\]. Extension studies showed sustained low brain atrophy rates, in the absence of continuous treatment with Alemtuzumab or other DMTs during the follow up period \[[@CR149]\]. The CARE-MS II 5-year follow-up study (2017) provided class III evidence that Alemtuzumab slows brain atrophy; the annual BVL rate continued to drop during the third year and remained low through the fourth and fifth year as well \[[@CR150]\]. The immune-modulatory agent Daclizumab in a 3-year post hoc analysis of 899 RRMS patients was compared to IFN beta-1a on brain volume change. Median annualised PBVC was significantly reduced in the DAC treatment group during both the first and the second year of treatment (baseline---24th weeks: − 0.674 vs − 0.745; 24th--96th weeks: − 0.511 vs − 0.549; all p \< 0.0001) in comparison to INF β treatment \[[@CR151]\], a finding which was consistent with previous longitudinal data \[[@CR151]--[@CR153]\]. Ocrelizumab is a humanized mAb designed to target CD20+ B cells. MRI outcomes hint towards a positive effect on BVL and clinical disability progression. Treatment with Ocrelizumab has significantly slowed brain atrophy rates in comparison to INF-β1a (baseline to 96 weeks: 23.5% p \< 0.0001 in OPERA 1 and 23.8% p \< 0.0001 in OPERA 2) along with clinical disability \[[@CR154]\]. Ocrelizumab reduced the rate of whole BVL in PPMS from week 24 to week 120 by 17.5%120 (p = 0.0206) compared with placebo (ORATORIO) \[[@CR155]\]. Emerging DMTs and their effect on PBVC {#Sec13} -------------------------------------- Several new agents are currently undergoing clinical development, including immuno-modulatory, neuroprotective or remyelinating compounds. Laquinimod, a linomide derivative, has also shown promising results on PVC rates in RRMS, most probably as a result of reduced astrocytic activation within the CNS \[[@CR156]\]. In the ALLEGRO clinical trial, after adjusting for baseline active inflammation, laquinimod markedly reduced BVL as compared to the placebo \[[@CR157]\]. Positive effects on PBVC are replicated in one active comparator trial \[BRAVO\] versus im IFN-β-1a \[[@CR158], [@CR159]\]. At present, the agent is further investigated in RRMS \[CONCERTO\] and PPMS patients \[ARPEGGIO\]. Cladribine, an antiproliferative agent that takes effect by interfering with DNA synthesis, has shown significant effects in terms of relapse rate and disability progression \[[@CR160], [@CR161]\]. Data from CLARITY study suggested that at 18 months, patients treated with cladribine had 20%reduction in brain atrophy compared with patients receiving placebo \[[@CR162]\]. However, further studies are needed, in order to cladribine's effect on brain atrophy rates, be fully elucidated \[[@CR161], [@CR163], [@CR164]\]. Conclusions {#Sec14} =========== MS is an evolving disease, now considered of both inflammatory and neurodegenerative nature \[[@CR165]--[@CR168]\]. Axonal injury and loss accounting for brain atrophy may be either acute (i.e. due to inflammation) or chronic/late due to pathogenic mechanisms primed by the preceding inflammation and later perpetuating with disease progression \[[@CR169]--[@CR171]\]. Brain atrophy occurs as early as CIS, progresses faster than it does in healthy adults, and is the best predictor of future disability, physical and cognitive \[[@CR166], [@CR172]\]. It is widely accepted to be a valid, sensitive and reproducible measure of neuroprotection in MS research studies and therapeutic trials. There is now a variety of approved DMDs, with secondary neuroprotective properties, and an even greater number of novel compounds, in various stages of development and investigation. A firm belief remains that for a therapy to be effective in delaying the disease progression, its impact on axon and neuronal survival needs to be monitored. Conventional MRI findings (T1-hypotensive or T2 hypertensive lesion load) have already shown their limits for monitoring the disease burden and progression in MS patients. Newly introduced sophisticated imaging methods hold promise for the future of the clinical surveillance of the disease. Trials incorporating brain atrophy in their endpoints are providing accumulating evidence that rises substantial hopes for treating neurodegeneration in the near future. **Publisher\'s Note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Athina Andravizou and Efthimios Dardiotis shared first authorship Not applicable AA, ED, AA, MS, VS, GMH were involved in the conception of the study. AA, ED, AA, MS, VS, GMH, ZT, AMA, IN, CB, GT, GD, NG, DPB were involved in the acquisition of the data and study design. AA, ED, AA, MS, VS, GMH, ZT, AMA, IN, CB, GT, GD, NG, DPB were involved in the writing of the article. AA, ED, AA, MS, VS, GMH, ZT, AMA, IN, CB, GT, GD, NG, DPB critically revised the manuscript. All authors read and approved the final manuscript. The author(s) received no specific funding for this work. Compliance with ethical standards. Human and animal rights: This article does not contain any studies with animals performed by any of the authors. This article does not require informed consent due to the lack of human participants. This article does not require informed consent due to the lack of human participants. The authors declare that they have no competing interests.

1. Introduction {#sec1-sensors-19-05449} =============== Wearable exoskeletons have recently been extensively researched \[[@B1-sensors-19-05449],[@B2-sensors-19-05449]\]. Regarding medical applications, the lower limb exoskeleton has widely been researched, because it can assist elderly people or those suffering from muscle weakness to walk easily and also help people suffering from strokes or paraplegics to walk again. In past decades, numerous exoskeletons were designed such as the hybrid assistive limb (HAL) \[[@B3-sensors-19-05449]\], EKSO \[[@B4-sensors-19-05449]\], ReWalk \[[@B5-sensors-19-05449]\], Indego \[[@B6-sensors-19-05449]\] and brain-machine-interface (BMI)-controlled rehabilitation exoskeletons \[[@B7-sensors-19-05449],[@B8-sensors-19-05449],[@B9-sensors-19-05449]\]. The exoskeleton robots work in different modes under different gait phases. Gait phase classifiers could be used to control walking-assistive devices for people \[[@B10-sensors-19-05449]\]. The method of providing the specific joint torque in a specific gait phase is widely used for controlling exoskeleton robots. The applications of gait phase classifier can be found in exoskeletons \[[@B11-sensors-19-05449]\], smart prostheses \[[@B12-sensors-19-05449]\] and functional electronic simulation (FES) \[[@B13-sensors-19-05449],[@B14-sensors-19-05449]\]. Based on the existing research results, the gait phase classifiers often use single-type sensors or a combination of multiple types of sensors, such as the angular velocity, attitude, force, electromyograph (EMG), IMU, camera and so on. However, the fewer sensors we use the easier we control the exoskeleton. This paper mainly focuses on detecting the gait phase of a human-exoskeleton coupling system using only joint angular sensors considering the gait differences between the wearers. Different people have different gait features while wearing exoskeleton robot due to the gap between the exoskeleton and the wearer and the operation habits, such as the correspondence between the joint angle and the moment at which the foot contacts the ground, the amplitude of the joint angle and others. These differences are analyzed in detail in [Section 4.1](#sec4dot1-sensors-19-05449){ref-type="sec"}. We define these user-dependent features as the unique feature of one wearer. At the same time, the gaits have some general features, such as the angle of hip joints which have one peak and one valley in one gait cycle, nearby the peak angle of the hip and knee joints correspond to the switch of different gait phases. The gait phase can be obtained by the optimized joint angle threshold based on the general features. However, the present gait phase classifiers concentrate on utilizing the general features of the gait and omitting the unique feature when wearing the exoskeleton robot. Hence, we will investigate whether the unique gait feature is relevant to the decrease in the gait phase classification accuracy or to the performance of predicting the assist torque of the exoskeleton robot. In this work, we adopt the KRLS algorithm to construct a gait phase classifier. There are two main advantages of using the KRLS algorithm to design the classifier. On the one hand, the computational complexity of the KRLS algorithm is low and it can be deployed in real-time control loops. On the other hand, the algorithm could update the classifier model with online changing data. In other words, the KRLS-based classifier model could adaptively obtain features of different data sets, which means that the KRLS-based classifier model has the ability to learn specific gait characteristics from the particular wearer. To validate the performance of the KRLS based gait phase classifier, two classical gait phase recognition algorithms were used for comparison---SVM and MLPNN. To obtain the optimal parameter of the SVM algorithm, the particle swarm optimization (PSO) algorithms was applied. To validate the adaptation abilities of the KRLS algorithm, we designed the assist torque prediction experiment. There are two main contributions of this work. First, to the best of our knowledge, we were the first to apply KRLS to gait phase classification and assist torque prediction to control the exoskeleton robot. Second, we found that the unique gait feature has an effect on the generalization and robustness of the gait phase classifier. Therefore, the online adaptation ability is needed for the gait phase classification or assist torque prediction. The remainder of this paper is organized as follows. We introduce and discuss the related works in [Section 2](#sec2-sensors-19-05449){ref-type="sec"}. In [Section 3](#sec3-sensors-19-05449){ref-type="sec"}, we introduce the data acquisition platform, data source and data processing at first, then KRLS, MLPNN, and SVM algorithms are described for constructing gait phase classifiers and assist torque predictors. Experimental processes and results are comprehensively described in [Section 4](#sec4-sensors-19-05449){ref-type="sec"}. We discuss the results in [Section 5](#sec5-sensors-19-05449){ref-type="sec"}. [Section 6](#sec6-sensors-19-05449){ref-type="sec"} concludes this work and the future work. 2. Related Work {#sec2-sensors-19-05449} =============== Many researchers used the gait phase classification information to control the lower limb exoskeleton. Kazerooni \[[@B11-sensors-19-05449]\] presented a hybrid control scheme to control the Berkeley lower extremity exoskeleton (BLEEX). In their work, the walking gait cycle was divided into two phases, swing phase and stance phase. Meanwhile, two controllers with different control schemes were adopted. A sensitivity amplification controller was used for the swing leg, while a position control was used for the stance leg. The position controller can be implemented in many ways \[[@B15-sensors-19-05449],[@B16-sensors-19-05449]\]. Li \[[@B17-sensors-19-05449]\] utilized a finite state machine to estimate the current gait period of the wearer among the six major periods. After that, an impedance-based controller was adopted to provide functional gait assistance in each gait period. Chen \[[@B18-sensors-19-05449]\] presented a novel gait phase-based control strategy for a portable knee-ankle-foot robot in which the assistive torque was only provided for selected gait phases. Thanks to recent advances in wearable sensor technologies such as goniometers, accelerometers, gyroscopes, inertial measurement units \[[@B19-sensors-19-05449]\], force sensing registers (FSRs) \[[@B20-sensors-19-05449],[@B21-sensors-19-05449],[@B22-sensors-19-05449],[@B23-sensors-19-05449],[@B24-sensors-19-05449]\] and sensor fusion technology \[[@B25-sensors-19-05449]\], various sensor-based gait phase classification methods were developed in exoskeleton robots in which different sensors were used to distinguish the gait phase. Jung \[[@B10-sensors-19-05449]\] utilized two types of artificial neural networks to classify the gait phase using signals from the angular sensors in the lower limb exoskeleton. The classifiers use the angle of each lower limb and the angular velocities of joints as input to obtain the gait phase. Guo \[[@B26-sensors-19-05449]\] proposed a C4.5 decision tree to identify the gait phase. Joint angles, its angular velocity and the contact force of the foot were measured. Next, five phases of the gait were identified by the C4.5 decision tree algorithm. Heo \[[@B27-sensors-19-05449]\] used recurrent neural networks (RNN) to categorize a gait cycle into a swing phase and a stance phase. The input data of the RNN was composed of joint angles alongside its velocities and a back propagation algorithm was used to train the network. Kong \[[@B28-sensors-19-05449]\] noted that, although ground contact forces (GCF) provide detailed information to detect gait phase, the idea of simply using a threshold to predict the gait phase could not achieve satisfactory results. Therefore, a fuzzy logic method was applied to obtain a smooth and continuous gait phases detection. Mannini \[[@B29-sensors-19-05449],[@B30-sensors-19-05449]\] used hidden Markov models (HMMs) to estimate four gait phases using the signals from the gyroscope mounted on the foot instep. Another work \[[@B31-sensors-19-05449]\] adopted linear discriminant analysis (LDA) to separate eight different phases of a gait cycle using the electromyographic (EMG) signals of the lower limb. Zheng \[[@B32-sensors-19-05449]\] presented a walking-speed-adaptive estimation method of the gait phase through novel noncontact capacitive sensors and adaptive oscillators. These were aimed at real-time gait phase estimation tasks. In Reference \[[@B33-sensors-19-05449]\], the MLPNN algorithm was used to achieve gait phase recognition for lower limb exoskeletons using only joint angular sensors. Nevertheless, all of these studies neglected the unique feature of different peoples' gaits. The KRLS algorithm was used in the load forecasting of power systems for online, real-time property \[[@B34-sensors-19-05449]\]. In Reference \[[@B35-sensors-19-05449],[@B36-sensors-19-05449]\], Pang employed KRLS in an FPGA to implement online non-linear regression. Inspired by the real-time and nonlinear properties of KRLS, this work employs KRLS to address the gait phase classification problem in the exoskeleton robot. 3. Materials and Methods {#sec3-sensors-19-05449} ======================== In this section, the data acquisition platform is described, then the data source is described. After that, we show how we divided the gait phase and how to define the assist torque. Finally, the gait phase classification and prediction model is introduced. 3.1. Data Acquisition Platform {#sec3dot1-sensors-19-05449} ------------------------------ The data acquisition platform is shown in [Figure 1](#sensors-19-05449-f001){ref-type="fig"}. There are two kinds of sensors in this exoskeleton robot, the goniometers and the force sensing resistors. The lower limb exoskeleton designed by the Shenzhen Institutes of Advanced Technology (SIAT), Chinese Academy of Science called the SIAT exoskeleton was used in our experiments \[[@B37-sensors-19-05449]\]. This exoskeleton was designed with the aim of helping the paraplegic patients to walk again normally or in the rehabilitation training for people suffering from lower limb injuries, such as accidents or strokes. Since the first generation exoskeleton designed in 2012, a third version of the SIAT exoskeleton has been redesigned. After several improvement proposals in recent years, the newest SIAT exoskeleton has a more reasonable mechanism structure with lighter weight (27 kg), which means that users would feel more comfortable when wearing the exoskeleton. Currently, our exoskeleton has three joints in each leg, including ankle joint, knee joint and hip joint. Overall, there are three degrees of freedom (DoFs) in each leg of the exoskeleton and a total of six joint DoFs for the entire exoskeleton. The three joints in each leg can all produce the flexion/extension (F/E) motion to produce movement in fore-and-aft directions. The distributions of DoFs and actuator types of joint are shown in [Table 1](#sensors-19-05449-t001){ref-type="table"}. It should be mentioned that the knee and hip joints are active and driven by the DC motors, while the ankle joint is passive with compliant springs, so as to reduce the impact and shock when contacting the ground. To guarantee that the joints can provide enough torque for the entire exoskeleton and the wearer, we chose the 200 watt RE 50 maxon DC motor. Additionally, the length of the thigh and shank in each leg of the exoskeleton can be adjusted to fit the different height of wearers. The main characteristics of SIAT Exoskeleton robot are shown in [Table 2](#sensors-19-05449-t002){ref-type="table"}. In order to obtain the state information of the SIAT exoskeleton and use it for the controller to decide the following movement, goniometers are installed in the hip and knee joints to collect their angle information. The used goniometers are WDG-AM23-360, which can record the absolute angle range from 0 to 360 with 12-bit resolution. This means that angular measurement precision can be as high as $0.087{^\circ}$. Meanwhile, three force sense resistors (FSRs) are contained in an insole place in the shoes to detect the relationship between the foot and the ground. As shown in [Figure 1](#sensors-19-05449-f001){ref-type="fig"}, three FSRs are placed on the heel, middle and toe. The FSR 402 sensors equipped in our exoskeleton are produced by Interlink Electronics. FSR is a polymer thick film (PTF), shown in the bottom of [Figure 1](#sensors-19-05449-f001){ref-type="fig"}, whose resistance decreases when the force that is applied to the active surface increases. As the wearer walks in a gait cycle, the output voltage changes corresponding to the variation resistance of the FSR. The SIAT exoskeleton has two working modes---power-assist mode and zero-torque mode---which are used for different condition. The clutches are designed on each joint to switch between the two modes \[[@B33-sensors-19-05449]\]. The power-assist mode offers enough torque to let the wearers walk in the predefined trajectory with the smart crutches to maintain balance. In the zero-torque mode, the wearers can walk following their own will and the exoskeleton offers torque as needed to follow the pilot. The zero-torque mode is used for collecting gait data with angle sensors described above. 3.2. Data Source {#sec3dot2-sensors-19-05449} ---------------- The gait data we collect comes from 10 healthy volunteers who never suffered from gait dysfunctions. The heights of the volunteers varied between 161 cm to 177 cm and their weights ranged from 53 kg to 70 kg. Before data collection, all the participants understood the purpose and the procedure of the data collection. The detailed information of the volunteers is described in [Table 3](#sensors-19-05449-t003){ref-type="table"}. Before gait data collection, the shank and thigh of the exoskeleton were adjusted to ensure that the joint axes of the exoskeleton and the wearer are aligned. Each volunteer was then asked to wear the SIAT exoskeleton robot set in zero-torque mode and to walk in any comfortable way for a while. After the above adjustment, we started to collect gait data. In this procedure, volunteers were asked to walk 5 m on a flat surface with their normal walking speed while a host computer recorded the data from the goniometers and FSR. Each volunteer repeated the above procedure 5 trials. Then start to record data from another subject. [Figure 2](#sensors-19-05449-f002){ref-type="fig"} shows how to define the joint angles: the angle of the hip is the deviation from the standing baseline to thigh baseline, while the angle of the knee is the deviation from the thigh baseline to shank baseline. The direction of the hip extension is defined as negative, while the direction of the knee flexion is defined as positive. Finally, we obtained angle information from the knee and hip joints and FSRs data from the feet with 50 Hz sampling frequency. One segmentation of angle information of the hip and knee joints and FSRs of the foot are shown in [Figure 3](#sensors-19-05449-f003){ref-type="fig"}. The data collected while the subjects stopped walking and were standing still are removed to ensure all the data used in our experiments is walking gait data. Fifty series of gait data have been collected with each subject having different walking phase, step length, step height and walking cadence to ensure the volunteer operate the exoskeleton robot as their will. Finally, we obtained a total of 24,408 samples from the collected data that are to be used in our later experiments. Every sample is composed of five dimensions, that are the hip joint angle, knee joint angle, heel pressure, paw pressure, and toe pressure. The FSR data are to be used to confirm the current gait phase, which is described in detail later. 3.3. Gait Phase Division and Assist Torque Definition {#sec3dot3-sensors-19-05449} ----------------------------------------------------- The gait phase division and the assist torque definition are used to generate the label of the gait data in the classification experiment and the prediction experiment, respectively. ### 3.3.1. Gait Phase Division {#sec3dot3dot1-sensors-19-05449} Rancho Los Amigos (RLA) terminology is often used as the standard way to divide gait phase into 8 phases, this has become the mainstream since the late 1980s \[[@B38-sensors-19-05449]\]. RLA terminology divides gait phases into initial contact, loading response, mid-stance, terminal stance, pre-swing, initial swing, mid-swing and terminal swing. In this work, our aim is to enhance the rehabilitation of hemiplegic patients who have only one dysfunctional leg. This could be done by controlling the exoskeleton robot to assist the disabled leg using the coupling relationship between the two legs. The gait phase information from the healthy leg is used to estimate the walking phase of the disabled leg and the different manually adjusted torques are used to assist the disabled leg in different walking phases. Hence, all we need to do is to correctly identify the gait phase of the healthy leg. Motion information of the healthy leg is described by the RLA stance phase states. When we decouple the phase relationship of the left and right legs in RLA gait terms, it is easy to find that the states of one leg phase can be classified into 4 phases based on the relationship between the foot and the ground which can be defined as---heel strike (HS), foot flat (FF), heel off (HO) and swing (SW) \[[@B39-sensors-19-05449]\] as shown in [Figure 4](#sensors-19-05449-f004){ref-type="fig"}. This relationship in stance foot (STF) and swing foot (SWF) can represent one-leg phase information. In this paper, we used HS, FF, HO, and SW to divide the gait phase. The one leg gait phase could be divided based on the FSR data and used as the switching value. Therefore, the label of HS, FF, HO and SW should be calculated before gait classification. Each FSR and series of walking has a different threshold, the thresholds could be separately calculated using the following equation $$\begin{array}{l} \begin{array}{r} {D_{threshold} = \frac{y_{max} + 2\left( 1 - \sigma \right)y_{min}}{1 + 2\left( 1 - \sigma \right)},} \\ \end{array} \\ \end{array}$$ where $D_{threshold}$ is an FSR threshold of one series of the walking gait, $y_{max}$ is the maximum value of an FSR series, $y_{min}$ is the minimum value of the FSR series and $\sigma$ is the coefficient that represents the percentage of FSR threshold. In these experiments, $\sigma$ is set to 50%. Therefore, we can obtain three thresholds for heel FSR, sole FSR and toe FSR. If the voltage of the FSR data is larger than the corresponding threshold, then the state of this FSR is high, otherwise it is low. After obtaining the threshold of each FSR in each series, a truth table is created to divide the gait phase, as shown in [Table 4](#sensors-19-05449-t004){ref-type="table"} \[[@B37-sensors-19-05449]\]. ### 3.3.2. Assist Torque Definition {#sec3dot3dot2-sensors-19-05449} The hip joint assist torque is a set of continuous smooth curves and it is defined based on the gait phase. According to the clinical gait analysis (CGA), from the HS to HO of one leg, the corresponding hip joint performs a flexion torque, while from the HO to SW of one leg, the corresponding hip joint performs an extension torque. Hence, the contralateral leg is the opposite. We define the extension torque to be positive and the flexion torque to be negative. At the middle of HS and FF of the healthy leg, the assist torque is defined as 5 Nm. At the rising edge of FF to HO of the healthy leg, the assist torque is defined as 0 Nm. At the middle of HO and SW of the healthy leg, the assist torque is defined as −5 Nm. Finally, at the falling edge of SW to HS of the healthy leg, the assist torque is defined as 0 Nm. The spline interpolation method in MATLAB is employed to get the continuous assist torque. 3.4. Gait Phase Classification and Prediction Model {#sec3dot4-sensors-19-05449} --------------------------------------------------- In this paper, the KRLS, MLPNN and SVM algorithms are employed to classify the gait phase. The particle swarm optimization (PSO) is used to optimize the parameters of SVM. KRLS and MLPNN are also employed in the regression of the assist joint torque of the exoskeleton robot. In the gait phase classification experiment, a sliding window for five sample points is employed. The input of classifier is defined as $$\begin{matrix} \left\lbrack x_{k}\left( t \right)~x_{k}\left( t - 1 \right)~x_{k}\left( t - 2 \right)~x_{k}\left( t - 3 \right)~x_{k}\left( t - 4 \right) \right. \\ {x_{h}\left( t \right)~x_{h}\left( t - 1 \right)~x_{h}\left( t - 2 \right)~x_{h}\left( t - 3 \right)~x_{h}\left( t - 4 \right)\rbrack,} \\ \end{matrix}$$ where $x_{k}\left( t \right)$ and $x_{h}\left( t \right)$ are the angles of the knee and hip joints at time *t*, respectively. The output of these classifiers is a discrete gait phase. In the assist torque prediction experiment, a sliding window for five sample points is also employed. The input of the predictor is defined as $$\begin{array}{r} \left\lbrack x_{k}\left( t \right)~x_{k}\left( t - 1 \right)~x_{k}\left( t - 2 \right)~x_{k}\left( t - 3 \right)~x_{k}\left( t - 4 \right) \right. \\ {x_{h}\left( t \right)~x_{h}\left( t - 1 \right)~x_{h}\left( t - 2 \right)~x_{h}\left( t - 3 \right)~x_{h}\left( t - 4 \right)} \\ {y_{T}\left( t - 1 \right)~y_{T}\left( t - 2 \right)~y_{T}\left( t - 3 \right)~y_{T}\left( t - 4 \right)~y_{T}\left( t - 5 \right)\rbrack,} \\ \end{array}$$ where $x_{k}\left( t \right)$ and $x_{h}\left( t \right)$ are the angles of the knee and hip joints at time *t* and $y_{T}\left( t \right)$ is the target assist torque at time *t*. The target assist torque at times $t - 1$, $t - 2$, $t - 3$, $t - 4$, $t - 5$ is employed as the input in the prediction task. The output of these predictors is a continuous assist torque. ### 3.4.1. The Kernel Recursive Least-Squares Algorithm {#sec3dot4dot1-sensors-19-05449} The KRLS algorithm \[[@B40-sensors-19-05449]\] is the nonlinear version of the recursive least-squares (RLS) algorithm, which could change the structure and parameter of the classification model adaptively when using varying data. Using a sequential sparse process, the KRLS algorithm was allowed to run in real time. Here, we introduce the algorithm. Given a recorded set of data $$\begin{array}{l} \begin{array}{r} {Z^{t} = {\left( x_{1},y_{1} \right),\left( x_{2},y_{2} \right),\ldots,\left( x_{t},y_{t} \right)},} \\ \end{array} \\ \end{array}$$where *Z* is the data set of the sampling point of the walking data, *t* is the time step, $x_{i}$ is the joint angle at time step *i*, and $y_{i}$ is the corresponding gait phase. The simple form of RLS minimizes the sum of squared errors at each time step *t* and could be calculated as follows $$\begin{array}{l} {\mathcal{L}\left( w \right) = \sum\limits_{i = 1}^{t}\left( {f\left( x_{i} \right) - y_{i}} \right)^{2} = {\parallel {\mathsf{\Phi}_{t}^{T}w - y_{t}} \parallel}^{2},} \\ \end{array}$$ $$\begin{array}{l} {w_{t} = \sum\limits_{i = 1}^{t}\mathbf{\alpha}_{i}\phi\left( x_{i} \right) = \mathsf{\Phi}_{t}^{T}\mathbf{\alpha},} \\ \end{array}$$ $$\begin{array}{l} {\mathcal{L}\left( \mathbf{\alpha} \right) = {\parallel {K_{t}\mathbf{\alpha} - y_{t}} \parallel}^{2},} \\ \end{array}$$ where $\phi$ is a mapping: $\left. \chi\longrightarrow\mathcal{H} \right.$ from the original space to the Hilbert space, $\mathbf{\alpha} = \left( \mathbf{\alpha}_{1},\ldots,\mathbf{\alpha}_{t} \right)^{T}$, $\mathsf{\Phi}_{t} = \left\lbrack \phi\left( x_{1} \right),\phi\left( x_{2} \right),\ldots,\phi\left( x_{t} \right) \right\rbrack$, substituting into Equation ([3](#FD5-sensors-19-05449){ref-type="disp-formula"}) we have Equation ([5](#FD7-sensors-19-05449){ref-type="disp-formula"}). Where $K_{t} = \left\lbrack \mathsf{\Phi}_{t}^{T}\mathsf{\Phi}_{t} \right\rbrack$ and $K_{ij} = \left\langle {\phi\left( x_{1} \right),\phi\left( x_{2} \right)} \right\rangle$, $\left\langle {\cdot , \cdot} \right\rangle$ is the inner product in the feature space can be expressed in terms of the kernel function. In this paper, the Gaussian kernel is used $\mathbf{\alpha}_{t} = K_{t}^{+}y_{t}$. Usually, we minimize Equation ([3](#FD5-sensors-19-05449){ref-type="disp-formula"}) w.r.t. *w*, to obtain $$w_{t} = \min\limits_{w}{\parallel {\mathsf{\Phi}_{t}^{T}w - y_{t}} \parallel}^{2} = \left( \mathsf{\Phi}_{t}^{T} \right)^{+}y_{t},$$ where $\left( \cdot \right)^{+}$ denotes the pseudo-inverse. Fortunately, it can be easily verified that the optimal weight vector can be expressed as in Equation ([4](#FD6-sensors-19-05449){ref-type="disp-formula"}) by simply adding some vector $\overline{w}$ that is orthogonal to $\phi\left( x_{i} \right)$ to $w_{t}$ and substituting into Equation ([3](#FD5-sensors-19-05449){ref-type="disp-formula"}). To reduce the space and time required for the calculation and increase adaptive ability, we use the sparse method and have $$w_{t} = \mathsf{\Phi}_{t}\mathbf{\alpha}_{t} \approx \widetilde{\mathsf{\Phi}_{t}}A_{t}^{T}\mathbf{\alpha}_{t} = \widetilde{\mathsf{\Phi}_{t}}\widetilde{\mathbf{\alpha}_{t}},$$ where $\widetilde{\mathbf{\alpha}_{t}}$ is a vector of *m* "reduced" coefficients that can be used in the online scenario. Here we use the one-versus-all strategy to distinguish one gait phase from another. Therefore, for *n* classification tasks, we have *n* classifications and the predicated class is expressed as $$y = \max\limits_{c}f_{c}\left( x_{i} \right),$$ where $c \in {1,\ldots,n}$ represents the corresponding gait phase class, $x_{i}$ is the input vector, and $f_{c}$ represents the *c*-th classifier. ### 3.4.2. Multi-Layer Perceptron Neural Network Method {#sec3dot4dot2-sensors-19-05449} Neural networks are often used in gait phase estimation or regression as a normal method. The multi-layer perceptron neural network (MLPNN) consists of an input layer, one or more hidden layers and an output layer. It is a commonly used type of neural networks and has many applications in various disciplines. Prior research proved that 3-layer forward neural networks could approach any multivariate nonlinear functions when the used activation function is sigmoid. The relationship between the input layer and hidden layer can be formulated as $$\begin{array}{l} {h = f_{1}\left( W_{1}x + b_{1} \right),} \\ \end{array}$$ $$\begin{array}{l} {y = f_{2}\left( W_{2}h + b_{2} \right),} \\ \end{array}$$ where *x* is the input feature vector, $W_{1}$ and $W_{2}$ are weight matrices, $b_{1}$ and $b_{2}$ are bias vectors, and $f_{1}$ and $f_{2}$ are sigmoid functions. The weight matrices and bias vectors are optimized using back propagation algorithms. Two hidden layers are employed in the MLPNN. The activation function is sigmoid and the maximum number of iterations is set to 1000. The number of neurons is set to 30 in data set analysis experiment. ### 3.4.3. Support Vector Machine {#sec3dot4dot3-sensors-19-05449} Support vector machines (SVM) have the advantages of low cost and low generalization error, and are widely used in practical problems. We use the spatial features from walking data to train an SVM classifier to distinguish different gait phases. The basic SVM \[[@B41-sensors-19-05449],[@B42-sensors-19-05449]\] attempts to find a linear boundary that separates data from two classes and tries to orient the boundary so as to maximize the distance between the boundary and the nearest data points for each class. Given a dataset consisting of *m* pairs of data. $D = \left\{ \left( x_{1},y_{1} \right),\left( x_{2},y_{2} \right),\ldots,\left( x_{m},y_{m} \right) \right\}$ where the indicator $y_{i} \in \left\{ ‒1, + 1 \right\}$, the SVM will construct a boundary that can be expressed as $$\begin{array}{l} \begin{array}{r} {f\left( x \right) = w^{T}x + b = 0,} \\ \end{array} \\ \end{array}$$ optimization problem $$\begin{array}{lr} \min\limits_{w,b} & {\left( {\frac{1}{2}w^{T}w + C\sum\limits_{i = 1}^{m}\xi_{i}} \right) = 0,} \\ {s.t.} & {\xi_{i} \geq 0,} \\ & {y_{i}\left( w^{T}x_{i} + b \right) \geq 1 - \xi_{i},i = 1,2,\ldots,m,} \\ \end{array}$$ where *C* controls the trade-off between the model complexity and empirical risk. Generally, we obtain linear boundaries that achieve a better training-class separation in this enlarged space and that are equivalent to nonlinear boundaries in the original space. Here, we use *G* to represent the parameter of the Gaussian kernel function. For 4-classes classification, the one-versus-all strategy is used in the experiment as same as the method used in KRLS. ### 3.4.4. Optimization Using Particle Swarm Optimization Algorithm {#sec3dot4dot4-sensors-19-05449} To acquire satisfactory results, we must fine tune the above-mentioned parameters *C* and *G*. Thus, we employ the PSO algorithm \[[@B43-sensors-19-05449]\] to optimize the parameters. The PSO is an effective method in multi-objective optimization. The PSO algorithm randomly initializes a group of particles, then optimizes the parameters *C* and *G* by iterations. There are two fitness values in the update iteration: the individual fitness representing the current local optimal solution, and the global fitness indicating the global optimal solution. Here, we use the prediction success rate (PSR) produced by SVM through cross validation as the fitness values of each particle. The particle velocity is updated as follows $$\begin{array}{l} \begin{array}{r} {v_{i} = wv_{i} + c_{1}r_{1}\left( p_{i} - x_{i} \right) + c_{2}r_{2}\left( g_{i} - x_{i} \right),} \\ \end{array} \\ \end{array}$$ where $v_{i}$ is the velocity of *i*-th particle and $v_{i} \in \left\lbrack - v_{max},v_{max} \right\rbrack$, *w* represents the inertia weight factor, $c_{1} \geq 0$ and $c_{2} \geq 0$ are learning factors and $r_{1} \in \left\lbrack 0,1 \right\rbrack$ and $r_{2} \in \left\lbrack 0,1 \right\rbrack$ are random numbers. The position of the particle is updated by $$\begin{array}{l} \begin{array}{r} {x_{i} = x_{i} + v_{i},} \\ \end{array} \\ \end{array}$$ where $x_{i}$ is the position of *i*-th particle in the search space. ### 3.4.5. Evaluation Criterion {#sec3dot4dot5-sensors-19-05449} In order to compare the performances of different classifiers, we define the correct rate (CR) of the dataset as C R = c o r r e c t l y e s t i m a t e d s a m p l e p o i n t s t o t a l s a m p l e p o i n t s × 100 \% . In order to compare the effects of different predictors, we define the evaluation criteria as $$\begin{array}{l} \begin{array}{r} {{MSE} = 10\lg\left( {\sum\limits_{i = 1}^{N}\left( \left( y\left( i \right) - y_{p}\left( i \right) \right)^{2} \right)} \right)} \\ \end{array} \\ \end{array}$$ where *N* is the size of the test set, $y\left( i \right)$ and $y_{p}\left( i \right)$ are the target value and prediction value, respectively. The smaller the mean squared error (MSE), the smaller the error between the predicted value and the true value, and vice versa. 4. Results {#sec4-sensors-19-05449} ========== In this section, we first analyze the collected dataset, then we analyze the effect of using the unique feature of gaits on the performance of the MLPNN classifier. After that, we compare the classification results of KRLS, MLPNN and SVM without considering the effects of unique feature. Finally, considering the effects of unique feature, the assistive joint torque curves are adaptively predicted by KRLS. Simultaneously, the predicted results are compared to that of MLPNN that do not adaptively adjust the weight matrix in the test set. 4.1. The Results of Data Set Analysis {#sec4dot1-sensors-19-05449} ------------------------------------- The results of joint angle data and gait phase label of the five trials done by volunteers No. 1, No. 2 and No. 4 are shown in [Figure 5](#sensors-19-05449-f005){ref-type="fig"}. Each walking gait phase is divided as discussed above. By comparing the results of each volunteer, we can see that the walking gaits in different trials are similar, and the gait phasing results are consistent. However, comparing the results of different volunteers, the gait phasing results differ from one to another. The following remarks can be seen---No. 1 has a longer FF phase shape than No. 2 and No. 4, No. 4 has the longest SW phase shape overall and the shape of HS and FF phase in No. 1 and No. 2 are similar. Hence, in this dataset, the unique gait features of different volunteers are included. By comparing the joint angles of different volunteers and the corresponding gait phase segmentation results, it can be found that different volunteers obviously have different correspondence between the joint angle and the moment at which the foot contacts the ground, as well as the amplitude of joint angle. In [Figure 6](#sensors-19-05449-f006){ref-type="fig"}, the dynamic time warping (DTW) method is employed to mathematically compare the similarity of joint angle curves. It is shown that the joint curve of different volunteers has different formations. Hence, it is difficult to accurately classify the gait phase when different people are using the exoskeleton robot as the data results shown in [Figure 5](#sensors-19-05449-f005){ref-type="fig"} suggest. In order to analyze the common feature of gaits in the collected data set and the unique gait feature of different volunteers, we first analyze the data consistency of the same volunteer. That is, to classify the user-dependent gait phase. The MLPNN, the commonly used method, is employed to classify the gait phase using Equations ([9](#FD11-sensors-19-05449){ref-type="disp-formula"}) and ([10](#FD12-sensors-19-05449){ref-type="disp-formula"}). The gait data of each volunteer adopts 5-fold cross-validation. The classification results of each volunteer are shown in [Figure 7](#sensors-19-05449-f007){ref-type="fig"}. In this figure, the volunteers No. 3, 4, 6, 7, 8, 9 and 10 have an CR over $85\%$ and the distribution of 5-fold cross-validation results is concentrated over the average line. This suggests that the data of these volunteers have less variation and have more consistency. The results of No. 1 and 2 are more variation than the other volunteers. However, the maximum CR is near $90\%$, which means that the gait feature of some trial could be accurately obtained from other trials but the data of some trials do not contain enough gait features. To avoid the disturbance of non-consistency, we use the data of volunteers No 3, 4, 6, 7, 8, 9 and 10 to validate the effects of their unique gait features and common gait features. Two groups of experiments are designed---Trial 1 (T1) and Trial 2 (T2). In T1, we randomly select one or two sets of gaits as the test set for each volunteer's gait, that is using 70% of the gait data for the training set and 30% for the test set. In T2, we randomly select two volunteers' gait data as the test set, using 25 trials (from 5 volunteers) of the gait data for the training set and 10 trials (from 2 volunteers) for the test set. The datasets of T1 and T2 used as an input to the MLPNN classifier that has the same configuration parameters as the cross-validation experiments. Both T1 and T2 were repeated 35 times. The CR results of T1 and T2 are shown in [Figure 8](#sensors-19-05449-f008){ref-type="fig"}. The CR results of T1 are concentrated between $86\%$ and $88\%$, while the CR results of T2 have more deviation ranging from $81\%$ to $85\%$. From a qualitative view, the CR of T1 is obviously better than T2. For the sake of analysis, we separately sort the CR results of T1 and T2 and focus on the 15 results near the median value shown in [Table 5](#sensors-19-05449-t005){ref-type="table"}. The average CR of T1 and T2 in [Table 5](#sensors-19-05449-t005){ref-type="table"} are $86.49\%$ and $82.67\%$. The average CR of T1 is nearly $4\%$ higher than T2. The results show that the gait features between different people are different. If the untrained gait is classified by the classifier with fixed parameters, the gait phase cannot be correctly classified. 4.2. Comparison and Analysis of KRLS, MLPNN and SVM {#sec4dot2-sensors-19-05449} --------------------------------------------------- In this section, the data of 7 volunteers (No. 3, No. 4 No. 6, No. 7 No. 8, No. 9 and No. 10) are used and we show the accuracy of different parameters of all three classifiers. The training set and the test set are chosen similar to T1, but the combination of data are fixed in this part. All three classifiers use the same training and test sets. The purpose of this experiment is to verify the performance of three classification models when the data set don't have unique gait features. The KRLS algorithm uses the kernel method, so the memory size and the width of the kernel will affect the results. In order to determine the reasonable memory size and kernel width, we try different values. Memory size is set within the range of \[200 800\] with 100 strides, while the kernel width is set within the range of \[0.5 80.5\] with 0.5 stride. Since the accuracy will stop increasing when the memory size is higher than 500 with fixed kernel width, the memory size is set to 600. The results of different kernel width values are shown in [Figure 9](#sensors-19-05449-f009){ref-type="fig"}. [Figure 9](#sensors-19-05449-f009){ref-type="fig"} shows how the average accuracy increases when the kernel width is in the range of \[0.5 46\] and achieves the highest CR ($88.71\%$) when the kernel width is 46. The accuracy decreases in the range of \[46 80.5\]. Regarding MLPNN networks, we use two hidden layers. In order to obtain reliable results, we train each MLPNN model for 10 times and calculate the average accuracy to represent the performance of that model. The number of the hidden layer nodes ranges from 5 to 40 in this model with 5 strides. The activation function is sigmoid and the maximum number of iterations is set to 1000. The results are shown in [Figure 10](#sensors-19-05449-f010){ref-type="fig"} that shows different combinations of the number of hidden layers' nodes. When the first hidden layer has 5 nodes and the number of nodes in the second one ranges from 5 to 40, the CR is always lower than 88%. When the number of nodes ranges from 10 to 25 in the first hidden layer and from 5 to 35 in the second one, the accuracy increases as the number of neurons increases and is almost always higher than 88% and achieves the best performance (88.45%) when the first hidden layer has 25 nodes and the second one has 35. When it comes to the SVM model, we used the PSO algorithm to determine the optimal parameters *C* and *G*. We set the population to 20, the maximum iterations to 100, learning factor $c_{1} = 1.6$, $c_{2} = 1.9$; inertia weight factor $\omega = 0.6$, the search range of parameter *C* is \[0.001 20\], the search range of parameter *G* is \[0.001 30\]. After optimization, The best results of *C* and *G* are 1.2 and 4.6, respectively. The maximum CR is 88.69%. Finally, we report the best parameters of the three models based on our training and testing datasets. Using these parameters, the best CR results of the models built by KRLS, MLPNN, and SVM are $88.71\%$, $88.45\%$, and $84.22\%$, respectively. The best results of these classifiers using the fixed dataset are similar by selecting their optimal parameters. To test the model generalization ability of the three algorithms, five trials of data from volunteers No. 2, 3, 4, 6, 7, 8, 9, and 10 are mixed together, and a K-fold cross-validation method is used to select the training set and the test set. During cross-validation, the combined dataset is divided into K sections. Every section could be the testing set, while others are used as the training set in the cross-validation loop. Every trial is repeated 5 times and the average results are employed. [Table 6](#sensors-19-05449-t006){ref-type="table"}, [Table 7](#sensors-19-05449-t007){ref-type="table"} and [Table 8](#sensors-19-05449-t008){ref-type="table"} provide the 3-Fold, 5-Fold, and 10-Fold CR results of cross validation, respectively. The CR result of the best performing classifier under the specific fold index is marked in bold. The CR results show that KRLS has almost the best performance under various cross-validation conditions. [Table 9](#sensors-19-05449-t009){ref-type="table"}, shows that the average CR of KRLS is nearly $2.33\%$ higher than MLPNN and $2.49\%$ higher than SVM under 3-Fold condition, about $3.62\%$ higher than MLPNN and $3.35\%$ higher than SVM under 5-Fold condition and finally about $3.04\%$ higher than MLPNN and $2.98\%$ higher than SVM under 10-Fold condition. This shows that the KRLS method has a robust and stable performance under different data conditions. From the above comparison and analysis, we can find that the KRLS method performs better than the other two classification algorithms in gait phase classification tasks and shows a robust and stable performance in cross-validation experiments. 4.3. Assist Torque Prediction {#sec4dot3-sensors-19-05449} ----------------------------- Since the features of different wearers' gaits are not similar, that is unique gait features, it is difficult for the gait phase classifiers trained using a certain dataset to accurately classify the gait phase of a wearer whose gait data are not included in the database (shown in [Figure 8](#sensors-19-05449-f008){ref-type="fig"} and [Table 5](#sensors-19-05449-t005){ref-type="table"}). Hence, the gait phase classifier should adaptively learn the gait features. KRLS has the ability to adaptively update its model. To validate that the KRLS method could be used in exoskeleton robot with good adaptive performance, we designed an experiment to compare the hip joint assist torque prediction task with MLPNN. The reason why we do not construct the adaptive gait phase classifier is that if the ground truth of the gait phase is known, we don't need to build the adaptive classifier. In order to verify the ability of adaptation, we used the data set including the gait data from all volunteers. We randomly selected the gait of 3 volunteers as the test set and used the rest as the training set. This way, it is guaranteed that the training set will not have the special gait features of the volunteers included in the test set. This data set is chosen similar to T2. KRLS and MLPNN are employed in this section to predict the assist torque. The parameters of KRLS and MLPNN are the same as in the classification part. Since the KRLS algorithm does not require iterative optimization, the KRLS based predictor is updated adaptively in the training set and the test set. The MLPNN requires iterative optimization and the MLPNN based predictor is only updated in the training set. The results of assist torque prediction are shown in [Table 10](#sensors-19-05449-t010){ref-type="table"}. The MSE of KRLS is two times smaller than MLPNN. These results show that the KRLS algorithm which has adaptive ability could predict the results quite well, even if there are unlearned unique gait features in the test set. [Figure 11](#sensors-19-05449-f011){ref-type="fig"} shows a picked period of prediction results, in which we can find that the prediction results of MLPNN have an obvious lag and the results from KRLS basically coincide with the target assist torque trajectory. 5. Discussion {#sec5-sensors-19-05449} ============= This paper employs three classification methods---KRLS, SVM, and MLPNN---to build recognition models of one leg gait phase and employ KRLS and MLPNN to predict assist torque of the hip joint. Angle data from goniometers on the hip and knee of a single leg were used from 10 healthy participants to train and test these models. T1 and T2 experiments are compared to show the effectiveness of unique gait features. Comparing results of KRLS, MLPNN and SVM showed that the KRLS network has the best robustness in cross-validation experiments. The KRLS network could adaptively obtain the unique gait feature of each volunteer and the results of the KRLS predictor show a great performance in the assist torque estimation task. Using the goniometer only, the gait phase can be well identified and the sensor and control systems of the exoskeleton robot can be simplified. Our experiments show that KRLS is more robust compared with the other two algorithms and easier to obtain optimal results in the overall experiment. Additionally, the unique gait feature is useful in the gait phase-based controller to improve the robustness. Although the gait phase recognizers obtain satisfactory accuracy, there still remain unstable predictions in the case of two continuous gait phase. Therefore, some gait information is still not used to increase the accuracy. 6. Conclusions {#sec6-sensors-19-05449} ============== Considering the unique gait features of each person, the KRLS algorithm can adaptively distinguish the gait phase based on the unique features. KRLS also has a robust performance in gait phase classification tasks compared with classical SVM and MLPNN. Compared with the gait phase classification method by plantar sensor, the proposed method uses only the joint angle information and has better adaptability to uneven surfaces. It also reduces the number of sensors and the complexity of the exoskeleton system. Compared with gait classification methods using surface EMG sensor, the proposed method is insensitive to fatigue, sweating and installation. Currently, the exoskeleton is not able to perform activities of daily living (ADL) standards and is still expensive for consumers. Furthermore, the applications of the exoskeleton robot in the medical institutes would be investigated. We expect that the exoskeleton robot will be widely used as shared equipment \[[@B44-sensors-19-05449]\] in the architectural industry, logistics industry and many more domains. Therefore, the technologies of adaptively learning the unique gait feature would be widely applied. We aim to increase the accuracy of gait classification in future work by using a larger data set. Research on how the exoskeleton affects normal gait of each participant is valuable to be conducted in the future to increase the accuracy of gait classification. At the same time, the online control architecture with gait classifier considering the unique gait feature for shared applications will be researched in our future work. Y.M. and C.W. conceived of the research project. X.W. and C.W. developed the SIAT exoskeleton. Y.M. preformed the experiments. X.W., C.W., Z.Y. and G.L. provided support and discussions. Y.M. wrote the paper. C.W. and Z.Y. performed the English corrections. The work described in this paper is partially supported by the National Key Research and Development Program of China (2017YFB1302303), the National Natural Science Foundation-Shenzhen Joint Research Program (U1613219) and Shenzhen Institute of Artificial Intelligence and Robotics for Society. The authors will thank Yuhao Luo for his great help in data organization. The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. {#sensors-19-05449-f001} {#sensors-19-05449-f002} {#sensors-19-05449-f003} {#sensors-19-05449-f004} {#sensors-19-05449-f005} {#sensors-19-05449-f006} {#sensors-19-05449-f007} {#sensors-19-05449-f008} {#sensors-19-05449-f009} {#sensors-19-05449-f010} {#sensors-19-05449-f011} sensors-19-05449-t001_Table 1 ###### The joint actuator types. Joint DoFs Actuator Type ----------- ------ ---------------- Hip F/E 2 Motor-actuated Knee F/E 2 Motor-actuated Ankle F/E 2 Passive sensors-19-05449-t002_Table 2 ###### The main characteristics of SIAT exoskeleton. Item Value --------------------------- ------------------------------------- Weight 27 kg Hip joint angle range $- 10^{{^\circ}}$--$100^{{^\circ}}$ Knee joint angle range $0^{{^\circ}}$--$95^{{^\circ}}$ Hip joint maximum torque 179.8 Nm Knee joint maximum torque 192 Nm sensors-19-05449-t003_Table 3 ###### Volunteers information. No. Volunteers Weight (kg) Height (cm) ---------------- ------------- ------------- 1 65 165 2 67 175 3 53 161 4 58 175 5 61 175 6 62 170 7 55 177 8 54 170 9 58 172 10 59 176 sensors-19-05449-t004_Table 4 ###### Truth table of gait phase. Gait Phase HS FF HO SW ------------ ------ ------ ---------- ----- Heel FSR High High Low Low Sole FSR Low High Low/High Low Toe FSR Low High High Low sensors-19-05449-t005_Table 5 ###### 15 times CR results near the median value of T1 and T2. T Number T1 T2 Diff ---------- ----------- ----------- ---------- Average **86.49** **82.67** **3.82** Max 87.46 84.24 3.22 Min 85.87 81.30 4.57 sensors-19-05449-t006_Table 6 ###### 3-fold cross-validation results. Fold Index 1 2 3 ------------ ----------- ----------- ----------- KRLS **85.56** **87.41** 83.5 MLPNN 80.58 84.18 **84.74** SVM 82.74 86.27 79.98 sensors-19-05449-t007_Table 7 ###### 5-fold cross-validation results. Fold Index 1 2 3 4 5 ------------ ----------- ----------- ----------- ----------- ----------- KRLS **87.85** **86.38** **85.17** **86.23** 84.55 MLPNN 87.17 81.70 75.13 83.20 **84.89** SVM 86.33 85.83 82.68 82.99 75.60 sensors-19-05449-t008_Table 8 ###### 10-fold cross-validation results. Fold Index 1 2 3 4 5 6 7 8 9 10 ------------ ----------- ----------- ----------- ----------- ----------- ----------- ----------- ----------- ----------- ----------- KRLS 83.52 **84.77** **82.96** **86.58** **87.20** **88.57** **87.42** 85.85 **86.66** **89.10** MLPNN **86.94** 83.83 77.90 72.31 81.23 88.00 86.64 84.70 83.72 87.01 SVM 69.06 83.93 81.76 84.69 81.83 85.86 87.03 **86.38** 84.96 87.38 sensors-19-05449-t009_Table 9 ###### Average accuracy and differential between KRLS and other two classifiers. K-Fold 3-Fold 5-Fold 10-Fold ----------------- ----------- ----------- ----------- KRLS 85.49% 86.04% 86.26% MLPNN 83.17% 82.42% 83.23% SVM 83.00% 82.69% 83.29% Diff with MLPNN **2.33**% **3.62**% **3.04**% Diff with SVM **2.49**% **3.35**% **2.98**% sensors-19-05449-t010_Table 10 ###### MSE of KRLS and MLPNN in assist torque prediction task. Trials No. 1 2 3 4 5 -- -------------- ------------ ------------ ------------ ------------ ------------ MSE of KRLS −28.05 −29.91 −28.10 −32.77 −31.61 MSE of MLPNN −12.26 −11.18 −13.41 −12.20 −12.26 Diff **−15.79** **−18.73** **−14.69** **−20.57** **−19.35**

1. Introduction {#sec1} =============== *M. genitalium* was first identified in 1980 from the urethral specimens of two men with nongonococcal urethritis (NGU) \[[@B1], [@B2]\]. Its prevalence in men presenting with urethritis is between 30 and 40% \[[@B1], [@B3]\]. Further, the presence of*M. genitalium* in men is associated with a 5.5-fold increased risk of NGU \[[@B4], [@B5]\]. In women,*M. genitalium* has been linked to cervicitis, endometritis, pelvic inflammatory disease (PID), infertility, human immunodeficiency virus (HIV), and adverse birth outcomes \[[@B1]\]. In the United States, the prevalence of*M. genitalium* among women is thought to be around 1%, slightly higher than*Neisseria gonorrhoeae* prevalence (0.4%) and less than*Chlamydia trachomatis* prevalence (3%) based on a nationally representative sample of young adults \[[@B3]\]. Despite initial contradictory findings regarding the association between*M. genitalium* and female genital tract pathology \[[@B6], [@B7]\], several studies have since confirmed this association \[[@B1], [@B3], [@B8]\], which has recently been reviewed in a meta-analysis \[[@B9]\]. The development of nucleic acid amplification in the 1990s has facilitated several epidemiologic studies that have examined the association of*M. genitalium* with PID \[[@B1], [@B10], [@B11]\]. In the 2015 sexually transmitted infection (STI) treatment guidelines, the Centers for Disease Control (CDC) calls attention to*M. genitalium* as an emerging sexually transmitted pathogen in women \[[@B12]\]. PID is a polymicrobial disease commonly diagnosed in women of reproductive age \[[@B13]\]. The prevalence among women aged 15--44 in the United States declined since 1995 from \~8.6% down to \~5.7% in 2002 and leveled off between 2006 and 2010 at 5% \[[@B14]\]. Although a third to half of PID cases have been associated with*N. gonorrhoeae* and/or*C. trachomatis*, many cases have an unknown etiology \[[@B12]\]. It is well known that PID can lead to serious reproductive problems including infertility, chronic pelvic pain, ectopic pregnancy, and recurrent infections. The independent association between*M. genitalium* and PID confirmed by several studies \[[@B15]\] raises concern that*M. genitalium* may play a pathogenic role, particularly in cases where other STIs are not identified. Thus, there is a need for PID treatment regimens to cover*M. genitalium*. Further complicating management, several studies have now identified*M. genitalium* treatment resistance among infected women \[[@B15]--[@B18]\]. Most studies on*M. genitalium* are observational studies of variable sample sizes, some very small, with very few randomized trials. Consequently, some of the study findings may not be transferable to most populations. The prevalence of*M. genitalium* from cohort studies done in high-risk populations is considerably higher than the prevalence found in the general population \[[@B1], [@B2]\].*M. genitalium* has also been described in sexually abstinent women, putting into question the criteria for screening for this pathogen. Our objective is to review the evidence in the literature regarding the association of*M. genitalium* with genital tract pathology in women and to identify needed areas of research regarding the pathophysiology, clinical manifestations, screening, and treatment of this pathogen. 2. Materials and Methods {#sec2} ======================== An initial PubMed search was conducted using the terms "*Mycoplasma genitalium,*" "*Mycoplasma genitalium* women," and "prevalence of*Mycoplasma genitalium* in asymptomatic women," which identified 1064 articles. Articles were excluded for lack of relevance by reviewing titles, abstracts, and content. English articles presenting relevant data stratified by sex and conducted exclusively in women were included. As a result, 66 articles were included in this review. 3. Results and Discussion {#sec3} ========================= 3.1. Clinical Updates on*M. genitalium* as a Sexually Transmitted Infection in Women {#sec3.1} ------------------------------------------------------------------------------------ ### 3.1.1. Epidemiology {#sec3.1.1} *M. genitalium* is one of the most common microorganisms associated with genital tract infections and is increasingly recognized as a STI \[[@B1], [@B19]--[@B21]\].*M. genitalium* infection has several clinical features consistent with sexual transmission, including higher detection among sexually active individuals compared to sexually naïve adolescents, detection in partners of infected individuals, and predominance in younger individuals with multiple sexual partners and men who have sex with men (particularly those infected with HIV) \[[@B1], [@B3], [@B8], [@B19], [@B24], [@B25]\]. Several studies that had identified*Mycoplasma* as a STI have showed statistically significant increased rates of infection among sexually active women, with rate/risk of infection increasing with 2 or more sexual partners. One study reported that the prevalence of*M. genitalium* increases by 10% with each additional sexual partner \[[@B3]\]. It has also been shown that women with infected partners are also at increased risk \[[@B3]\] so sexual activity in itself appears to be a major risk factor but is not the only determinant factor for infection. Further, several studies have shown an independent association between*M. genitalium* and genital infection. It however appears that not all carriers are symptomatic as evidenced by general population studies \[[@B26]\]. Several clinical associations with*M. genitalium* infection have been identified. In one prospective study,*M. genitalium* was found most frequently among women aged ≤24years, those with a history of abortion, and those with first intercourse after 20 years \[[@B32]\]. This last association seems counterintuitive but may be related to a higher chance of clearing the infection when women are first exposed to the organism at a younger age, although there are no studies to date to support this argument. Overall, most evidence suggests a low prevalence of*M. genitalium* among asymptomatic women \[[@B22]\], which may make screening efforts low-yield \[[@B21], [@B22], [@B33], [@B32], [@B23]\]. Most of the epidemiological studies on*Mycoplasma* infection have been conducted in high-risk populations, such as symptomatic and asymptomatic patients attending STI clinics. This introduces a sampling bias and limits the conclusions regarding*M. genitalium* as an independent STI in the general population \[[@B22], [@B23]\]. [Table 1](#tab1){ref-type="table"} summarizes the studies regarding the prevalence of*M. genitalium*. The prevalence in the general population is not known since routine screening is not done but some studies have estimated the global prevalence of*M. genitalium* among women to range between 1 and 6.4% \[[@B27]--[@B29]\]. Prevalence studies have usually included women attending STI clinics or those infected with HIV. Clarivet et al. found a low rate of 0.1% in asymptomatic women \[[@B22]\] whereas Gaydos et al. found a rate close to 20% among women attending STI clinic in Baltimore (\~70% of these women were symptomatic) \[[@B2]\]. Studies from adolescent clinics, STI clinics, and emergency departments in the United States have identified*M. genitalium* as a genital tract microorganism in 15--20% of young women reporting genitourinary symptoms or at risk for STIs based on clinical history \[[@B13]\]. *M. genitalium* coinfection with*C. trachomatis* has also been recognized. In a cross-sectional case-control study, 4.5% of asymptomatic patients were found to be positive for*M. genitalium* \[[@B31]\], and \~5% of individuals infected with*C. trachomatis* were coinfected with*M. genitalium* \[[@B31]\]. Asymptomatic study participants, usually recruited from a convenience group of STI clinic attendees reported no genital tract symptoms \[[@B21]\]. The prevalence was higher among younger women 18--24 years of age compared to older women (7.9% for*C. trachomatis* and 2.4% for*M. genitalium*, resp.) \[[@B32]\]. ### 3.1.2. Clinical Manifestations {#sec3.1.2} *M. genitalium* has been associated with typical PID symptoms such as pelvic pain, abnormal vaginal discharge, fever, nausea, and vomiting. Symptomatic women who are positive for*M. genitalium* are more likely to report postcoital bleeding, which could be due to cervicitis, compared to women negative for the organism (AOR 5.8; 95% CI 1.4--23.3, after adjusting for age and coinfections) \[[@B23]\]. Most*M. genitalium* infections are asymptomatic in women \[[@B26]\] and roughly half of women (56.2%) who test positive for the organism are asymptomatic \[[@B27]\]. Like*C. trachomatis, M. genitalium* can lead to "silent" PID infections with mild symptoms relative to*N. gonorrhoeae* associated PID symptoms \[[@B13]\]. Bjartling et al. found comparable rates of abnormal vaginal wet smear, cervical friability or tenderness, fever, and level of serum C-reactive protein (CRP) between*M. genitalium*-positive women and negative controls \[[@B31]\]. However,*M. genitalium*-positive women were more likely to report postcoital bleeding than the negative controls \[AOR 2.00 (1.10--3.61)\]. Further, women with*M. genitalium* were more likely to have combined cervical tenderness, postcoital bleeding, and abnormal vaginal discharge \[AOR 2.71 (1.50--4.90)\] compared to women not infected with*M. genitalium* \[[@B31]\]. ### 3.1.3. *M*.*genitalium*: An Emerging Cause of Pelvic Inflammatory Disease (PID) {#sec3.1.3} PID is an inflammatory disease that can include one or more of the following conditions: endometritis, salpingitis, tuboovarian abscess, and pelvic peritonitis. PID is described as a polymicrobial syndrome, mainly caused by anaerobic bacterial species \[[@B1]\].*N. gonorrhoeae* and*C. trachomatis* are the most commonly diagnosed organisms in PID, yet up to 70% of cases are of indeterminate etiology \[[@B1], [@B34]\]. Organisms of the vaginal flora such as*Mycoplasma*,*Ureaplasma, Gardnerella vaginalis, Escherichia coli,* and anaerobes have also been associated with PID. With the development of nucleic acid amplification tests (NAATs), the incidence of biopsy-proven endometritis or clinical PID associated with*M. genitalium* has increased \[[@B1], [@B31]\]. Women positive for*M. genitalium* were found to be twice as likely to have histology-proven endometritis than women testing negative after adjusting for age, race,*N. gonorrhoeae,* and*C. trachomatis* \[AOR 2.0 (1.0 to 4.2)\] \[[@B15], [@B35]\]. Despite being less studied, postabortal PID has been shown to be strongly associated with*M. genitalium* \[AOR 6.3 (1.6--25.3)\] \[[@B1], [@B13]\]. Several cross-sectional studies have investigated the independent association between*M. genitalium* and PID. For example, one prospective study reported a thirteenfold increased incidence of endometritis in the presence of*M. genitalium* at 30-day follow-up visits among an urban population of women in the United States with clinical PID without concurrent*N. gonorrhoeae* and*C. trachomatis* infection \[[@B15]\]. A recent meta-analysis shows pooled odds ratios of 1.66 \[95% CI, 1.35--2.04\] for cervicitis and 2.43 for infertility \[95% CI, .93--6.34\] among*M. genitalium* infected women \[[@B9]\]. ### 3.1.4. *Mycoplasma genitalium* and Its Association with Other STIs and Malignancies {#sec3.1.4} *M. genitalium* has been associated with increased susceptibility to HIV infection \[[@B36], [@B37]\]. Unlike most other*Mycoplasma* species,*M. genitalium* can attach to the surface of epithelial cells and invade the cells with a specialized tip structure \[[@B4]\]. In an*in vitro* model, Das et al. showed that*M. genitalium* increased the risk of HIV infection by infecting the epithelial layer, reducing its integrity, and activating HIV cell targets beyond the epithelial layer, thereby promoting transmission and reproduction within the host and increasing viral shedding through mucosal surfaces \[[@B36]\]. Vandepitte et al. in their nested case-control study found evidence of a temporal relationship between*M. genitalium* and HIV acquisition \[[@B37]\]. The association was only found among the subgroup that was tested for*M. genitalium* three months prior to first HIV-positive results compared to the group with earlier HIV testing (aOR = 7.19; 95% CI 1.68 to 30.77) \[[@B37]\]. Further studies have shown a positive association between*M. genitalium* and high-risk human papilloma virus (HR-HPV) infection. For example, one study of female sex workers showed that 39.6% were positive for*M. genitalium* and HR-HPV \[[@B38]\]. In addition, Zarei et al. have demonstrated an association between chronic*M. genitalium* infection and ovarian cancer and lymphoma \[[@B39]\]. However, these studies did not control for the sexual behavior of women and their partners, limiting the generalizability of the results. ### 3.1.5. Mycoplasma in Pregnancy {#sec3.1.5} All*Mycoplasma* species have been associated with perinatal morbidity and mortality \[[@B40]\]. A US-based cohort study demonstrated a 2.5-fold increase in preterm birth in women with*M. genitalium* infection who presented with contractions between 23 and 32 weeks of gestation compared to noninfected women (AOR 2.5; 95% CI 1.2--6.0) \[[@B1]\]. In a meta-analysis of six studies,*M. genitalium* was associated with preterm birth with a pooled OR of 1.89 (95% CI, 1.25--2.85) and also associated with spontaneous abortion with a pooled OR of 1.82 (95% CI, 1.10--3.03) \[[@B9]\]. Of the*Mycoplasmas*,*M. hominis* and*Ureaplasma* have been most associated with chorioamnionitis and are thought to contribute to these adverse effects \[[@B40]\]. While*M. hominis* has not been associated with PID, it has been associated with upper respiratory infections, nervous system infections, neonatal bacteremia, and meningoencephalitis, unlike*M. genitalium* \[[@B41]\]. ### 3.1.6. Diagnosis and Screening {#sec3.1.6} *Mycoplasma Diagnosis*.*M. genitalium* is a small bacterium of the Mollicutes class with no cell wall and a genome of only 580 kilobases in size \[[@B1], [@B42]\]. Consequently, it cannot be detected by gram stain and is extremely difficult to culture requiring up to 6 months for growth \[[@B12]\]. Its genome is most similar to*Mycoplasma pneumonia* \[[@B43]\], which causes atypical bacterial pneumonia. Currently there is no FDA-approved diagnostic test for*M. genitalium* \[[@B1]\]. Given the difficulty with culturing the organism and the lack of standardized serological tests for*M. genitalium,* NAATs in the form of polymerase chain reaction (PCR) assays are almost exclusively carried out for the diagnosis of*M. genitalium* in the research setting. Some PCR assays have demonstrated \>95% specificity and sensitivity \[[@B44]\]. A recent study reported loop-mediated isothermal amplification (LAMP) as a novel NAAT, which has similar sensitivity to a PCR assay \[[@B45]\]. To date four types of specimens can be collected for the detection of*M. genitalium*: vaginal swab, first void urine, and endocervical and rectal swabs. Some studies in the United States have shown that NAATs with vaginal swab specimens have the highest relative sensitivity compared to urine and endocervical specimens \[[@B10], [@B46]\]. Further, self-obtained vaginal swabs have been found to yield similar test sensitivities to clinician-obtained specimens \[[@B10]\]. In an earlier study conducted in Seattle, WA, among symptomatic women attending a STI clinic, the specimen with the highest sensitivities was the vaginal specimen PCR: reported sensitivities were 91%, 53%, and 65% for vaginal, cervical, and urine specimens, respectively \[[@B46]\]. In a subsequent cross-sectional study among women attending a STI clinic in New Orleans, the relative sensitivity of PCR was 85.7% for the vaginal swab specimen, 74.3% for the endocervical swab specimen, 61.4% for the urine specimen, and 24.3% for the rectal swab specimen for the detection of*M. genitalium* in women \[[@B10]\]. Consequently, vaginal swabs are currently the most commonly used specimens for detecting*M. genitalium* through PCR. *To Screen or Not to Screen*. The 2015 CDC sexually transmitted disease treatment guidelines recommend that all women diagnosed with PID should also be tested for HIV, gonorrhea, and chlamydia \[[@B12]\]. There are no recommendations regarding*M. genitalium* screening given the lack of data around the utility of screening and the lack of a FDA-approved testing modality for commercial use \[[@B12]\]. Given the higher prevalence of*M. genitalium* in high-risk women \[[@B1]\] and its reported association with PID, infertility, and adverse pregnancy outcomes, it would be reasonable to test symptomatic women for*M. genitalium* if NAAT is available. Further, in patients whose symptoms are refractory to appropriate antibiotic therapy for PID, cervicitis, and endometritis, testing for*M. genitalium* may be clinically beneficial and indicated based on current data. There is ongoing debate regarding possible cost, benefits, and harm of universal screening for*M. genitalium* among asymptomatic patients given that most carriers are likely asymptomatic. Given limited data, this decision should be based on a discussion between providers and patients in the context of personal risk factors, as official screening recommendations will not be made until better quality data on cost, harm, and benefits are available. 3.2. Treating*M. genitalium* Infection {#sec3.2} -------------------------------------- Azithromycin and doxycycline are the current first-line treatment for cervicitis and NGU \[[@B5]\]. One of the initial randomized controlled trials on*Mycoplasma genitalium* treatment reported more effective treatment with a single 1 g of azithromycin compared to doxycycline 100 mg BID for 7 days in the USA \[[@B47]\]. Cure rates with azithromycin ranged from 67 to 87% \[[@B5]\]. However, higher treatment failures with single 1 g of azithromycin were reported with a decline in efficacy down to 60% \[[@B48], [@B49]\] and to 39% in the most recent study \[[@B50]\]. Treatment failure with azithromycin is due to an isolated point mutation on 23 rRNA gene in numerous*M. genitalium* populations \[[@B17]\], with up to 50% of cases reported \[[@B12]\]. Due to these poor efficacy rates, alternative azithromycin regimens have been investigated \[[@B5]\]. Several studies have examined an extended 1.5 g azithromycin (500 mg on day 1, followed by 250 mg daily for 4 days) and a single higher dose of 2 g azithromycin once with the rationale that an extended azithromycin-containing regimen decreases the risk of acquired macrolide resistance when initiated first-line among patients without preexisting macrolide resistance. A similar trend has also been noted with doxycycline \[[@B47]\]. Unfortunately, a recent randomized controlled trial found declining microbiological cure rate for the extended regimen and single 2 g regimen to 25--81% (wide range based on different studies) and 73%, respectively \[[@B5]\]. In light of the rising azithromycin resistance, moxifloxacin had been introduced as a second-line treatment option. Moxifloxacin, a fluoroquinolone, was thought to be a reliable alternative with a reported 100% cure rate initially \[[@B52], [@B51]\]. According to the 2015 CDC guidelines, women with PID who do not respond to the first-line treatment within 7--10 days should be considered as possibly infected with*M. genitalium* and treated with moxifloxacin 400 mg/day for 14 days \[[@B12], [@B53]\]. It is not used as first-line due to more significant adverse effects associated with moxifloxacin relative to azithromycin, such as tendon rupture, although these significant adverse effects remain rare. However, as of 2013, increasing treatment failures have also been noted due to bacterial resistance to moxifloxacin with failure rates ranging between 10% and 15% \[[@B17], [@B18], [@B54]\]. Given increasing moxifloxacin resistance, monotherapy has the potential to increase the risk of multidrug-resistant strains. Other fluoroquinolones that have been investigated and proven to remain effective include gatifloxacin and sitafloxacin \[[@B55]\]. Other fluoroquinolones such as gemifloxacin, sparfloxacin, grepafloxacin, trovafloxacin, and garenoxacin have been shown to be effective against*M. genitalium in vitro* but lack human studies \[[@B55]\]. Ciprofloxacin, ofloxacin, and levofloxacin reportedly have poor activity against the microbe relative to moxifloxacin \[[@B55]\]. Pristinamycin is a streptogramin that is used to treat vancomycin-resistant*Enterococcus faecium* bacteremia and complicated skin infections due to MRSA \[[@B56]\]. Treatment of*M. genitalium* with pristinamycin (1 g 6 hourly for 10 days) led to negative PCR results 28 days after treatment \[[@B56]\]. This regimen appears promising for the treatment of multidrug-resistant*M. genitalium* but has not been well studied to inform optimal dosing and is reportedly expensive with limited availability \[[@B56]\]. Given the organism\'s propensity for drug resistance, follow-up testing to document treatment response is reasonable. Some authors advocate for a test of cure (TOC) in 3-4 weeks after treatment with resistance profiling in those with persistent infection despite treatment \[[@B16]\]. Most studies on TOC have been conducted in men, with fewer studies done in women. A retrospective cohort study performed TOC at 1 month from the initiation of therapy with azithromycin-containing regimens to identify resistant infections \[[@B57]\]. However, a later prospective cohort study investigated the optimal time for TOC and reported negative TOC within an average of 14 days (12--15 days) for infected patients that were susceptible to a single 1 g of azithromycin, which was used as the first-line treatment \[[@B50]\]. Those that appeared resistant were further treated with moxifloxacin 400 mg daily for 10 days with a negative TOC at 28 days for responders \[[@B50]\]. Furthermore, Falk et al. showed that individuals treated with azithromycin had a negative PCR within 8 days and those treated with moxifloxacin had a negative PCR within 1 week \[[@B58]\]. However, it was further discussed that early negative PCR may be related to low DNA levels for detection soon after treatment initiation with resistance detected at 10 days after treatment initiation with azithromycin and eventually recolonization requiring further treatment \[[@B58]\]. Hence it was concluded that optimal timing for the most reliable TOC should take place 3-4 weeks after treatment \[[@B58]\], which correlates with an earlier Japanese study performed among men \[[@B59]\]. Testing and/or empirical treatment of partners within the preceding 60 days of diagnosis are also strongly recommended for women with confirmed positive*M. genitalium* to prevent reinfection \[[@B12]\]. Partners are recommended to abstain from sexual intercourse until adequate treatment is completed and symptoms resolve if initially present \[[@B12]\]. There is no specific evidence regarding the utility of condom use in these circumstances. 3.3. Long-Term Sequelae of*M. genitalium* Infection {#sec3.3} --------------------------------------------------- The long-term reproductive consequences of*M. genitalium* infection have not been clearly determined. However, the association with PID indicates that infertility, chronic pelvic pain, and risk of ectopic pregnancy may be potential sequelae of infection with this pathogen like for*C. trachomatis* and*N. gonorrhoeae* infection \[[@B60]\]. This may be another argument for screening in certain populations. [Table 2](#tab2){ref-type="table"} summarizes the studies that investigated the association between*M. genitalium* and infertility.*M. genitalium* can persist for months or years in infected individuals \[[@B61]\]. In a recent meta-analysis, it had been reported that women carrying*M. genitalium* infection are usually asymptomatic with reported estimated clearance rate of 15 months based on a large London study \[[@B26]\]. Despite spontaneous clearance, chronic infection may lead to tissue damage prior to clearance causing long-term health problems. Reinfection due to the partner\'s carrier state may also lead to reinfection leading to more chronic infection. The PID Evaluation and Clinical Health (PEACH) Study is a multicenter, randomized prospective clinical trial, the largest treatment trial of mild to moderate acute PID in the United States, involving 586 women in several centers in North America who presented with signs and symptoms of PID \[[@B35]\]. This study showed higher rates of infertility (22%), chronic pelvic pain (42%), and recurrent PID (31%) among women in whom*M. genitalium* had been detected on endometrial samples by PCR compared to women testing negative, but these findings were not statistically significant \[[@B15], [@B35]\]. It is unclear whether the increased risk of other infections such as chlamydia or gonorrhea lead to infertility or if*M. genitalium* itself primarily leads to infertility. Given that untreated PID can lead to long-term adverse reproductive outcomes,*M. genitalium* may contribute to adverse effects on the reproductive tract. One prospective study identified strong*M. genitalium* antibody responses among women with a diagnosis of infertility that were asymptomatic, suggesting an adverse effect of*M. genitalium* on fertility \[[@B62]\]. Another prospective study showed that fertile women were less likely to have PCR-proven*M. genitalium* infection compared to women with idiopathic infertility (4.4% versus 29.2%, *P* = 0.0479) \[[@B63]\]. Consequently, some authors would recommend screening for*M. genitalium* as part of the STI work-up given possible adverse effects such as infertility, chronic pelvic disease, risk of ectopic pregnancy, and preterm labor as well as any other health consequence associated with PID. However, evidence regarding other reproductive sequelae is even more limited, and the few studies that have evaluated reproductive sequelae have not shown any statistically significant difference between women with and without*M. genitalium* infection \[[@B35]\]. A single case-control study on risk of ectopic pregnancy did not find any significant association either (OR 1.0, 95% CI 0.5--2.0) \[[@B65]\]. Well-powered prospective studies that control for other genital tract infections and compare*M. genitalium* cases to asymptomatic noninfected women are needed to establish the long-term reproductive consequences of chronic*M. genitalium* infection. 4. Conclusions and Areas for Future Research {#sec4} ============================================ *M. genitalium* is now increasingly recognized as a STI and has been associated with PID, endometritis, cervicitis, and HIV in women though clinical manifestations and risk factors overlap with other STIs. The availability of NAAT for PCR detection of this organism will allow further investigation into the effects of*M. genitalium* infection on long-term reproductive health outcomes such as infertility, chronic pelvic pain, ectopic pregnancy, and obstetric outcomes such as preterm deliveries. Due to antibiotic resistance patterns, alternatives to azithromycin and moxifloxacin must be investigated. In the interim, clinicians should consider testing for and treating*M. genitalium* on a case-by-case basis, particularly in women diagnosed with PID or cervicitis without clinical improvement using standard regimens. Competing Interests =================== The authors declare that there is no conflict of interests regarding the publication of this paper. ###### Summary of *M. genitalium* prevalence according to various studies in women. Source Study design Study population Overall *M. genitalium* prevalence (%) --------------------------------- ------------------------------------------------------------------ -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------- Gaydos et al. \[[@B2]\] Cross-sectional study 324 women attending STI clinics in Baltimore. Detected by transcription mediated amplification from vaginal, endocervical, and urine swabs 19.2 Oakeshott et al. \[[@B8]\] Prospective study 2378 sexually active female students (mean age of 21) followed up between 2004 and 2008 in London. Tested vaginal swabs by PCR 3.3 Haggerty et al. \[[@B15]\] Multicenter randomized controlled prospective study, PEACH study Stored cervical and endometrial specimens of 682 women treated with cefoxitin and doxycycline for clinically suspected PID tested by PCR 15 Clarivet et al. \[[@B22]\] Cross-sectional study 743 asymptomatic women attending free and anonymous STI clinics from April to August 2009. Detected by PCR in first void urine (FVU) sample 0.1 Falk et al. \[[@B25]\] Cross-sectional study 465 female STI clinic attendees (mean age of 24) in Orebro, Sweden. Tested FVU and endocervical samples by PCR 6 Hancock et al. \[[@B30]\] Cross-sectional study 1090 women attending the Public Health-Seattle & Kig County STI Clinic in Seattle, WA. *M. genitalium* detected by TMA from self-obtained vaginal swabs 7.7 Bjartling et al. \[[@B31]\] Cross-sectional case-control study 679 women attending a gynecological outpatient clinic from 2003 through 2008. Tested urine and vaginal swabs by PCR 2.1 Uno et al. \[[@B33]\] Cross-sectional study 200 women visiting the Obstetrics and Gynecology Department in Kizawa Memorial Hospital and Jaysaki Women\'s Clinic in Japan. Tested cervical swabs using PCR. 6.8 Gomih-Alakija et al. \[[@B38]\] Cross-sectional study 350 female sex workers aged 18--50 years in Nairobi, Kenya. Tested cervical samples by TMA 12.9 Bradshaw et al. \[[@B52]\] Prospective study 313 women attending Melbourne Sexual Health Center, Australia, between March 2005 and November 2007 with cervicitis/pelvic inflammatory disease and sexual contacts of proven *M. genitalium,*infected partners. Cervical, vaginal swabs, or FVU samples analyzed by PCR 10 Andersen et al. \[[@B64]\] Cross-sectional study 921 women aged 21--23 provided self-collected vaginal samples by PCR 2.3 ###### Summary of studies regarding *M. genitalium* and female infertility. Source Study design Study population Findings ----------------------------- ---------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Clausen et al. \[[@B66]\] Cross-sectional study 308 women undergoing IVF treatment in Aarhus, Denmark *M. genitalium* was detected in 22% of women with tubal factor infertility (TFI) versus 6.3% in women without TFI Tosh et al. \[[@B19]\] Multicenter (North America) randomized controlled prospective study, PEACH study Stored cervical and endometrial specimens of 682 women treated with cefoxitin and doxycycline for clinically suspected PID *M. genitalium* was associated with baseline endometritis (AOR 3.0, 95% CI 1.5 to 6.1). Nonsignificant trend towards increased infertility, chronic pelvic pain and recurrent PID, decreased pregnancy, and live birth were found in this study. Svenstrup et al. \[[@B62]\] Prospective study 212 couples attending a fertility clinic in Horsens-Brædstrup or the Holstebro fertility clinic in Denmark *M. genitalium* was found to be independently associated with TFI (AOR 4.5, 95% CI 1.2--15.6) Grześko et al. \[[@B63]\] Prospective study 51 patients with primary infertility (24 women with idiopathic infertility) and 23 women with proven fertility *M. genitalium* was found in 19.6% of all infertile women and 4.4% of fertile women (*P* = 0.156); 29.2% among women with idiopathic infertility versus 4.4% in fertile women (*P* = 0.0479) [^1]: Academic Editor: Louise Hafner

Introduction {#Sec1} ============ Cereals and cereal-based products are the most important commodities in human nutrition. The health benefits of whole grain cereal products are currently widely recognized and understood because of the presence of a broad range of bioactive compounds \[[@CR1]\]. Whole grain cereals are a rich source of carbohydrates, oils, proteins, vitamins, and minerals. The increasing demand for cereals and products thereof is reflected also by the fact that the world cereal production has reached its maximum level, exceeding 2.609 × 10^9^ tonnes in 2016, having increased steadily in the last 8 years \[[@CR2]\]. Although it is estimated that it is possible to maintain present food consumption levels by increasing overall food supplies in quantitative terms, providing quality food that is nutritious and free from contaminants is becoming a very challenging task. Among food contaminants, mycotoxins can have serious consequences in terms of both human and animal health as well as huge economic impacts \[[@CR3]\]. Mycotoxins are toxic products of secondary metabolism of microscopic filamentous fungi. These ubiquitous microorganisms are able to colonize various agricultural commodities either before harvest or under postharvest conditions, thus causing, in addition to mycotoxin contamination, a serious loss of harvest yield and quality of the infested commodity. According to the European Commission \[[@CR4]\], it has been estimated that 5--10% of global production is lost annually because of mycotoxin contamination. However, in a recent survey on mycotoxin contamination, more than 80% of samples were contaminated with at least one mycotoxin and 45% contained more than one secondary metabolite of fungi \[[@CR5]\]. Currently, there are approximately 100,000 described fungal species. The number of the most commonly occurring species in foods/feeds and indoor environments is estimated to be around 175 \[[@CR6]\]. The most toxigenic species belong mainly to three fungi genera: *Fusarium*, *Aspergillus*, and *Penicillium* \[[@CR7]--[@CR10]\]. These fungi can produce a wide range of mycotoxins differing not only in their chemical structures but also in the mode of toxicological actions. However, the health risk to humans and animals is currently attributed to only a limited number of them. With regard to high incidences of contamination (and thus possible dietary exposures) and their toxicity, aflatoxins (AFs) (from *Aspergillus*), ochratoxins (from *Aspergillus* and *Penicillium*), trichothecenes, fumonisins B (FBs), and zearalenone (ZEN) (from *Fusarium*) are of greatest concern. Their toxic effects range from various effects on the liver, kidney, hematopoietic system, immune system, and fetal and reproductive systems to significantly contributing to carcinogenetic and mutagenic developments \[[@CR11], [@CR12]\]. The effect of mycotoxin exposure differs greatly. The susceptibility of animals and humans to the toxicological effects of mycotoxins differs with species, age, nutrition, length of exposure, and other factors. Evaluation of adverse health effects is complicated by exposure to various co-occurring mycotoxins, which may lead to additive, synergic, or antagonist toxic effects \[[@CR13]\]. With regard to the health hazards attributed to mycotoxins and their impact on consumers (and farm animals), many countries have set up regulations for their control in the food and feed chain. On the world scale, the Joint FAO/WHO Expert Committee on Food Additives has assessed the toxicity of various mycotoxins and related health risks. In the European Union (EU), scientific opinions on the risk assessment of mycotoxins have been issued by the European Food Safety Authority (EFSA), which advises the European Commission Directorate-General for Health and Food Safety. Currently, only maximum levels for AFs, deoxynivalenol (DON), ZEN, ochratoxin A (OTA), FBs, and patulin in various foodstuffs have been set in EU countries \[[@CR14], [@CR15]\]. The chemical structures of some EU-regulated mycotoxins and other toxicologically important mycotoxins are depicted in Fig. [1](#Fig1){ref-type="fig"}. The establishment and acceptance of new regulatory limits is a long-term and complex process consisting of the evaluation of occurrence data that are important for overall risk assessment, availability, and understanding of toxicological data. To obtain such information, advanced analytical methods are required in both research and official control laboratories for reliable and accurate determination of mycotoxins. The determination of mycotoxins in cereals and cereal-based products is a challenging task because of their low occurrence levels and the complexity of the matrices \[[@CR13]\].Fig. 1Chemical structures of the most important mycotoxins belonging to the group of *Aspergillus* toxins (aflatoxin B~1~, alfatoxin G~1~, ochratoxin A, patulin), *Fusarium* toxins (deoxynivalenol, T-2 toxin, zearalenone, fumonisin B~1~, beauvericin, enniatin A), *Alternaria* toxins (alternariol), ergot alkaloids (ergotamine), and *Penicillium* toxins (patulin, ochratoxin A) In the past, the analytical methods were more focused on routine analysis (i.e., mainly methods for a single analyte/matrix or groups of structurally related mycotoxins in a respective matrix had been developed). These methods were usually based on specific sample preparation protocols followed by traditional chromatographic separation. Primarily liquid chromatography (LC) coupled with ultraviolet/diode array detection and fluorescence detection, was used, but detection by mass spectrometry (MS) was rarely used. Gas chromatography with either electron capture detection or MS was used in routine determination of mycotoxins (e.g., trichothecenes) after time-consuming and laborious derivatization \[[@CR16]\]. However, ongoing developments in the field of LC--MS technology have led to the availability of high-throughput instrumentation meeting the current demands of scientists and regulatory authorities for mycotoxin detection. The use of LC--MS in the determination of low molecular weight contaminants and residues at trace levels has significantly increased during the past two decades. Because of the unique features of this technique, LC--MS became a tool of choice to deal with a number of analytical challenges related to chemical food and feed safety testing in both research and routine commercial laboratories. In the field of mycotoxin determination there is a clear trend toward the use of multiple-analyte methods based on ultrahigh-performance LC (UHPLC) coupled with MS with various mass analyzers. LC--MS-based workflows provide significantly higher selectivity and sensitivity, increased confidence in the identification of analytes, and wider analyte/matrix scope as compared with traditional methods using conventional detectors \[[@CR17]\]. LC--MS also facilitates the use of streamlined sample preparation procedures that save time and labor and reduce the overall costs associated with mycotoxin testing. State-of-the-art LC--MS methods for the determination of mycotoxins may largely differ in terms of analyte scope depending on the intended use of the resulting data. Although some of the available methods focus exclusively on mycotoxins with regulatory limits in place, others may allow (semi)quantitative or qualitative analysis of hundreds of mycotoxins, metabolites, or degradation products in a single analytical run. Having a large analyte scope usually results in the need for compromises in terms of the method performance characteristics achieved for the respective compounds. Besides targeted applications, LC--MS also plays an invaluable role in metabolomics-based studies dealing with the discovery and elucidation of new toxins or metabolites, as discussed in detail later in this review. Currently, there are a wide range of LC-compatible MS instruments available on the market. A detailed overview and discussion of the advantages and disadvantages of current MS systems allowing measurements in low, medium and (ultra)high mass resolving power modes was provided in several recent comprehensive reviews \[[@CR18], [@CR19]\]. From the available studies, the MS detectors used in mycotoxin analysis include triple quadrupole (QqQ), ion trap, time-of-flight (TOF), and orbital ion trap mass analyzers, as well as hybrid systems that combine two types of analyzers. The latter group includes quadrupole--linear ion trap (QLIT), double quadrupole--TOF (QqTOF), quadrupole--orbital ion trap quadrupole--Orbitrap; Q--Orbitrap, and linear ion trap--orbital ion trap systems. The distribution of the LC--MS techniques applied in mycotoxin determination from 2012 to 2016 is depicted in Fig. [1](#Fig1){ref-type="fig"}. In the past 10 years, several review articles on the occurrence and determination of mycotoxins have been published \[[@CR11], [@CR13], [@CR16], [@CR20]--[@CR24]\]. Moreover, new developments and updates in this field are covered on a yearly basis in *World Mycotoxin Journal* \[[@CR25]--[@CR28]\]. This review thus provides insight into LC--MS-based methods for the determination of co-occurring mycotoxins. The aim is not to give a comprehensive overview of all published methods, but rather to focus on LC--tandem MS (MS/MS) targeted approaches and on untargeted analysis using LC--high-resolution MS (HRMS), including application of these approaches in the analysis of cereals and cereal-based food products. The advantages and limitations of both methods are critically assessed. Requirements and guidance for quantification and proper validation {#Sec2} ================================================================== Basic validation of an analytical method is a crucial part of the overall process of implementation of a new method. It must demonstrate that the analytical method complies with the criteria applicable for the relevant performance characteristics (i.e., it confirms that the method is suitable for the intended applications and provides reliable results). In the EU, only general guidelines on the performance of analytical methods and the interpretation of results are laid down in Commission Decision 2002/657/EC \[[@CR29]\]. This document gives the specification of individual performance characteristics that have to be evaluated during method validation. According to the in-house validation approach, specificity, trueness, recovery, repeatability, reproducibility, decision limit (CCα), detection capability (CCβ), calibration curve (linearity), and ruggedness should be evaluated (Table [1](#Tab1){ref-type="table"}). Strict guidelines on how to perform the experiments for the evaluation of the individual performance characteristics are also given here. Additionally, the term "confirmatory method" has been established. Any confirmatory method has to provide full information on the chemical structure of an analyte. Furthermore, an internal standard, preferably isotope labeled, should be used. Therefore, LC--MS/MS and LC--HRMS are recommended as the techniques of first choice.Table 1Overview of performance characteristics of an analytical method defined in Commission Decision 2002/657/ECPerformance characteristicDefinitionAccuracyThe closeness of agreement between a test result and the accepted reference value. It is determined by determining trueness and precisionDetection capability (CCβ)The smallest content of the substance that may be detected, identified, and/or quantified in a sample with an error probability of *β*. For mycotoxins with no legislation limit, the detection capability is the lowest concentration at which a method is able to detect truly contaminated samples with a statistical certainty of 1-*β*. For mycotoxins with a legislation limit, the detection capability is the concentration at which the method is able to detect the legislation limit concentration with a statistical certainty of 1-*β*. *β* error means the probability that the tested sample is truly noncompliant, even though a compliant measurement has been obtained (false compliant decision)Decision limit (CCα)The limit at and above which it can be concluded with an error probability of *α* that a sample is noncompliant. *α* error means the probability that the tested sample is compliant, even though a noncompliant measurement has been obtained (false noncompliant decision)PrecisionThe closeness of agreement between independent test results obtained under stipulated (predetermined) conditions. The measure of precision is usually expressed in terms of imprecision and computed as the standard deviation of the test results. Less precision is determined by larger standard deviationRecoveryThe percentage of the true concentration of a substance recovered during the analytical procedure. It is determined during validation if no certified reference material is availableRepeatabilityThe precision under repeatability conditions. Repeatability conditions means conditions where independent test results are obtained with the same method on identical test items in the same laboratory by the same operator using the same equipmentReproducibilityThe precision under reproducibility conditions. Reproducibility conditions means conditions where test results are obtained with the same method on identical test items in different laboratories with different operators using different equipment. Participation in ring trials is neededRuggednessThe susceptibility of an analytical method to changes in experimental conditions that can be expressed as a list of the sample materials, analytes, storage conditions, and environmental and/or sample preparation conditions under which the method can be applied as presented or with specified minor modifications. For all experimental conditions that could in practice be subject to fluctuation, any variations that could affect the analytical result should be indicated.SpecificityThe ability of a method to distinguish between the analyte being measured and other substances. This characteristic is predominantly a function of the measuring technique described, but can differ according to the class of the compound and the matrixTruenessThe closeness of agreement between the average value obtained from a large series of test results and an accepted reference value. Trueness is usually expressed as bias LC separation should be done with the appropriate LC column. The minimum acceptable retention time of the analyte of interest has to be at least twice the retention time corresponding to the dead volume of the column and has to match that of the calibration standard. The width of the retention time window should correspond to the resolving power of the chromatographic system. Moreover, the relative retention time of the analyte should match that of the calibration standard with a tolerance of ±2.5%. MS detection should be done by use of MS techniques such as recording of full mass spectra or selected-ion monitoring (SIM), as well as MS/MS techniques such as selected-reaction monitoring (SRM). In HRMS the resolution should typically be greater than 10,000 for the entire mass range (according to the 10% valley definition \[[@CR30]\]). In the full scan, the presence of all diagnostic ions (protonated and deprotonated molecules, characteristic fragment ions, and isotope ions) with a relative intensity of more than 10% in the reference spectrum of the calibration standard is obligatory. For SIM and SRM, a protonated or deprotonated molecule should preferably be one of the diagnostic ions selected. The diagnostic ions selected should not exclusively originate from the same part of the molecule. The signal-to-noise ratio for each diagnostic ion should be higher than 3:1. For the interpretation of data, a system of identification points should be used. In the case of mycotoxins, a minimum of three identification points are required per analyte. That means that two precursor ion to product ion transitions or one precursor ion and one product ion are required per analyte when low-resolution MS/MS or accurate mass HRMS is used, respectively \[[@CR29]\]. The trueness of a quantitative confirmatory method has to be verified either by the repeated analysis of a certified reference material or, if this is not available, through the recovery of additions of a known amount of the analyte to a blank matrix. The analytical standards and certified reference materials for mycotoxins are supplied by two companies (Romer Labs and Sigma-Aldrich) \[[@CR31], [@CR32]\]. The guideline ranges for the deviation of the experimentally determined recovery corrected mean mass fraction from the certified value are -50% to +20% for mass fraction below 1 μg/kg, -30% to +10% for mass fraction between 1 and 10 μg/kg, and -20% to +10% for mass fraction above 10 μg/kg \[[@CR29]\]. The precision of a quantitative confirmatory method expressed as the interlaboratory coefficient of variation (CV) for the repeated analysis of a reference or fortified material under reproducibility conditions (definition in Table [1](#Tab1){ref-type="table"}) should not exceed the level calculated by the Horwitz equation: CV = 2^(1-0.5log*C*)^, where *C* is the mass fraction. For instance, the CV for a mass fraction of 100 μg/kg should not exceed 23% \[[@CR29]\]. The trueness and precision of the analytical methods intended for the official control of mycotoxins are laid down in Commission Regulation (EC) No 401/2006 \[[@CR33]\]. Participation in interlaboratory testing is an efficient tool to demonstrate a sufficient level of method trueness. A drawback of both aforementioned European Commission decision and regulation is that the LC--MS methods counting more than 100 mycotoxins are not considered. Therefore, the performance of some validation experiments for such a high number of analytes is not feasible. For instance, to find a blank material free of a broad spectrum of mycotoxins is almost impossible, the definition of matrix effects and their evaluation is missing, the term "recovery" is not exactly specified, and the determination of the limit of detection (LOD) and limit of quantification (LOQ) by the spiking of 20 replicates at one level for various matrices is not feasible for hundreds of analytes because of the cost of analytical standards. Moreover, the availability of isotopically labeled standards (as it is recommended they be used) and certified reference materials on the market is also limited. Therefore, in practice, guidance document SANTE 11945/2015 \[[@CR34]\] designed for multiresidue determination of pesticides was very useful also for validation of LC--MS methods for multiple determination of mycotoxins. Briefly, matrices are grouped on the basis of water/sugar/fat content, and for each group one representative matrix should be validated. Sensitivity, mean recovery (extraction efficiency), precision, and LOQ have to be evaluated in the method validation. The spiking experiments have to be performed on a minimum of five replicates at two different levels (a low level to check the sensitivity and a higher level). The method LOQ is defined as the lowest validated spiking level meeting the method performance acceptability criteria \[mean recovery of 70--120% with a relative standard deviation (RSD) of 20% or less\]. The criteria for LC--MS in this document do not differ from those specified in Commission Decision 2002/657/EC. To ensure the accuracy of the results generated, a quality control is recommended to be introduced in the laboratory. In practice, most of laboratories regularly participate in proficiency ring trials. Proficiency testing is an effective procedure for quality assurance and performance verification in chemical analysis laboratories, ensuring that laboratory validation and within-laboratory procedures are working satisfactorily. The individual laboratory performance is expressed in terms of the *z* score in accordance with ISO 13525:2015 \[[@CR35]\] and is calculated as *z* = (*χ*~lab~ - *χ*~assigned~)/*σ*~p~, where *χ*~lab~ is the mean of the two measurement results reported by a participant, *χ*~assigned~ is the assigned value (robust mean), and *σ*~p~ is the standard deviation for proficiency assessment derived from the truncated Horwitz equation. The *z* scores obtained are interpreted as follows: \|*z*\| ≤ 2 is an acceptable result, 2 \< \|*z*\| ≤ 3 is considered a questionable result, and \|*z*\| \> 3 is an unacceptable result. An example of the *z* scores obtained by the "dilute and shoot" multimycotoxin LC--MS/MS method \[[@CR36]\] in routine proficiency testing organized by the Bureau Interprofessionnel des Études Analytique (BIPEA) is displayed in Fig. [2](#Fig2){ref-type="fig"}.Fig. 2An example of the *z* score compilation obtained by the multimycotoxin liquid chromatography--tandem mass spectrometry method in proficiency testing organized by the Bureau Interprofessionnel des Études Analytique (BIPEA). Green lines borders of acceptable range of *z* scores, red lines borders of questionable range of *z* scores, area outside red lines unacceptable values Currently, several proficiency testing schemes for mycotoxins are available in Europe, such as those from FAPAS (UK), BIPEA (France), Dienstleistung Lebensmittel Analytik (Germany), DUCARES (Netherlands), LGC Standards Proficiency Testing (UK), and Test Veritas (Italy). A detailed list of these schemes is available from \[[@CR37]\]. However, most proficiency testing schemes organized by the aforementioned providers are focused on only a single mycotoxin or mycotoxins belonging to the same group. The first multimycotoxin proficiency testing scheme was organized by the Institute of Sciences of Food Production of the National Research Council of Italy in 2011. Since then, several other multimycotoxin proficiency testing schemes has been organized. The results of these trials have been summarized by De Girolamo et al. \[[@CR38]\]. LC--MS/MS-based approaches intended for the targeted determination of mycotoxins {#Sec3} ================================================================================ Almost 80% of all published LC--MS studies on mycotoxins since 2012 used methods based on LC--MS/MS (Fig. [3](#Fig3){ref-type="fig"}). The term "targeted" analysis implies that only "known" mycotoxins can be determined. In targeted mycotoxin determination, the complexity of the analyzed matrix and the range of "target mycotoxins" are the most important factors in determining a suitable instrument to be used for that particular application. A detailed description of the analytical methods discussed in the following text is given in Table [2](#Tab2){ref-type="table"}.Fig. 3An overview of use of liquid chromatography (LC)--mass spectrometry (MS) instruments in studies focused on mycotoxin analysis published between 2012 and 2016. LC--MS/MS covers studies using instruments equipped with triple quadrupole, quadrupole--linear ion trap, and ion trap mass analyzers; LC--high resolution MS (HRMS)/MS covers studies using instruments equipped with quadrupole--time of flight or quadrupole--orbital ion trap mass analyzersTable 2Detailed description of the setup of some liquid chromatography (LC)--mass spectrometry (MS) methods for mycotoxin determinationReferenceExtractionCleanupAnalytesMatrixLC--MS instrumentLC conditionsMS conditions\[[@CR39]\]CH~3~CN--H~2~O (84:16, v/v)MycoSep 226 AflaZON+, MycoSep 227 (both Romer Labs)NIV, DON, FUS-X, 3ADON, 15ADON, DAS, HT2, T2, ZENMaizeQTRAP MS/MS instrument (Sciex) coupled to 1100 series LC system (Agilent Technologies)Aquasil RP-18 column (100 mm × 4.6 mm, 3 μm) + C~18~ guard column; 25 °C, flow rate 1000 μL/min, injection volume 25 μL; eluent A H~2~O--CH~3~OH (80:20, v/v), eluent B H~2~O--CH~3~OH (10:90, v/v), both containing 5 mM NH~4~CH~3~COO^-^; gradient 0.5 min 0% eluent B, linear gradient to 100% eluent B to 4.5 min, 100% eluent B to 7 min, 7.1 min 0% eluent B, reequilibration 3 min, total run 10 minAPCI± MRM, monitoring of 2 transitions (1 quantifier and 1 qualifier), dwell time 100 ms, polarity switching (2 periods)\[[@CR40]\]CH~3~CN--H~2~O (84:16, v/v)MycoSep 226 AflaZON+AFs, NIV, DON, 3ADON, 15ADON, FUS-X, HT2, T2, ZEN, OTA, STER, CIT, verruculogenVarious foods and feedQuattro Ultima QqQ instrument (Micromass) coupled to Acquity UHPLC system (Waters)UPLC BEH C~18~ (100 mm × 2.1 mm, 1.7 μm); 35 °C, flow rate 300 μL/min, injection volume 5 μL; eluent A ESI+ 10 mM NH~4~CH~3~COO^-^, ESI− 0.1% 0.1% (v/v) aqueous NH~3~, eluent B CH~3~OH; gradient initially 20% eluent B, linear increase to 5.5 to 85% eluent B, 100% eluent B within 0.3 min, reequilibration for 2 min at 20% eluent B, total run 10 minESI+, ESI−, MRM, 2 chromatographic runs, monitoring of 2 transitions (1 quantifier and 1 qualifier)\[[@CR41]\]2-step extraction: (1) PBS; (2) 70% CH~3~OHAOZFDT2 (VICAM)AFs, OTA, FBs, DON, ZEA, T2, HT2MaizeQTRAP MS/MS (Sciex) instrument coupled to 1100 micro LC system (Agilent Technologies)Gemini C~18~ column (150 mm × 2 mm, 5 μm) + Gemini C~18~ guard column (4 mm × 2 mm, 5 μm); 40 °C, flow rate 200 μL/min, injection volume 20 μL; eluent A H~2~O, eluent B CH~3~OH, both containing 0.5% CH~3~COOH and 1 mM NH~4~CH~3~COO^-^; gradient 3 min at 20% eluent B, jump to 40% eluent B, linear increase to 63% eluent B within 35 min, 63% eluent B for 11 min, reequilibration at 20% eluent B for 10 min, total run 59 minESI+, ESI−, dMRM, monitoring of 2 transitions (1 quantifier and 1 qualifier), time window of 1 MRM 0.8 min, cycle time 0.55 s\[[@CR42]\]CH~3~CN--H~2~O--CH~3~COOH (79.5:20:0.5, v/v/v). Evaporation and redissolution in PBS before IACMyco6in1+ (VICAM)AFs, OTA, FBs, DON, ZEN, T2, HT2Barley, maize breakfast cereals, peanutsQTRAP 4500 instrument (Sciex) coupled to a Prominence UFLC XR chromatography system (Shimadzu)Acquity UPLC HSS T3 end-capped C~18~ column (100 mm × 2.1 mm, 1.7 μm), 40 °C, flow rate 400 μL/min, injection volume 10 uL; eluents A H~2~O, eluent B CH~3~OH, both containing 5 mM NH~4~CH~3~COO^-^; gradient 5% eluent B increased to 50% eluent B in 1 min, linear increase to 100% eluent B within 6 min, 100% eluent B to 8 min, at 8.1 min initial conditions 5% eluent B, reequilibration at 5% eluent B for 2 min, total run 10 minESI± MRM, 2 periods, monitoring of 2 transitions (1 quantifier and 1 qualifier), dwell time 50 ms, polarity switching (2 periods)\[[@CR43]\]2-step extraction: (1) H~2~O; (2) CH~3~OH. Evaporation and redissolution in PBS before IACMyco6in1+ (VICAM)AFs, OTA, FBs, DON, ZEN, T2, HT2, NIVMaize, durum wheat, corn flakes, maize crackersQTRAP MS/MS instrument (Sciex) coupled to 1100 micro LC system (Agilent technologies)Gemini C~18~ column (150 mm × 2 mm, 5 μm) + Gemini C18 guard column (4 mm × 2 mm, 5 μm); 40 °C, flow rate 200 μL/min, injection volume 20 μL; eluent A H~2~O, eluent B CH~3~OH, both containing 0.5% CH~3~COOH and 1 mM NH~4~CH~3~COO^-^; gradient 3 min at 20% eluent B, jump to 40% eluent B, linear increase to 63% eluent B within 35 min, 63% eluent B for 11 min, reequilibration at 20% eluent B for 10 min, total run 59 minESI± MRM, monitoring of 2 transitions (1 quantifier and 1 qualifier), dwell time 100 ms, polarity switching (2 periods)\[[@CR44]\]NaCl + H~2~O--CH~3~OH (30:70, v/v). Dilution with PBS before IACOCHRAPREP + DZT MS-PREP, AOF MS-PREP + DZT MS-PREP, AFLAOCHRA PREP + DZT MS-PREP (R-Biopharm)OTA + DON, ZEN, T2, HT2; AFs, FBs, OTA + DON, ZEN, T2, HT2; AFs, OTA + DON, ZEN, T2, HT2Wholemeal bread, maize and maize-based products including infant foods, oat-based muesliAcquity TQD tandem QqQ MS instrument (Waters)Gemini C~18~ column (150 mm × 2 mm, 5 μm), 40 °C, flow rate 300 μL/min, injection volume 20 μL; eluent A H~2~O--CH~3~OH (95:5, v/v), eluent B H~2~O--CH~3~OH (98:3, v/v), both containing 0.5% HCOOH and 1 mM NH~4~HCOO^-^; gradient 20% eluent B for 0.1 min, to 10 min linear increase to 90% eluent B, 90% eluent B to 15 min, reequilibration at 20% eluent B, total run 20 minESI+ MRM, monitoring of 2 transitions (1 quantifier and 1 qualifier), 6 acquisition periods, dwell times from 0.1 to 0.27 s\[[@CR45]\]QuEChERSAFs, FBs, NIV, DON, 3ADON, 15ADON, FUS-X, HT2, T2, ZEN, OTA, DAS, NEORice, corn, wheat, rye, oat, barley, infant cereals, soya, corn glutenQTrap 4000 instrument (Sciex) coupled to 1100 series LC system (Agilent Technologies)Zorbax Bonus-RP column (150 mm × 2.1 mm, 3.5 μm) + Zorbax RB C~8~ guard column (12.5 mm x 2.1 mm, 3.5 μm), flow rate 250 μL/min, injection volume 40 μL; eluent A 0.15% (v/v) HCOOH + 10 mM NH~4~HCOO^-^, eluent B 0.05% HCOOH (v/v) in CH~3~OH; gradient: 0% eluent B at 1 min, linear increase to 100% eluent B until 15 min, 100% eluent B for 5 min, reequilibration at 0% eluent B for 5 min, total run 25 minESI± MRM, monitoring of 2 transitions (1 quantifier and 1 qualifier), 3 acquisition periods\[[@CR46]\]QuEChERSAFs, FBs, DON, HT2, T2, ZEN, OTAWheat, maize, riceMicromass Quattro Micro QqQ coupled to Alliance 2695 system (Waters)Atlantis RP C~18~ column (150 mm × 2.1 mm, 5 μm), 30 0°C, flow rate 300 μL/min, injection volume 20 μL; eluent A H~2~O--CH~3~OH (90:10, v/v), eluent B H~2~O--CH~3~OH (10:90, v/v), both containing 5 mM NH~4~CH~3~COO^-^; gradient 20% eluent B for 0.1 min, until 10 min linear increase to 90% eluent B, 90% eluent B to 15 min, reequilibration at 20% eluent B, total run 20 minESI± MRM, monitoring of 2 transitions (1 quantifier and 1 qualifier), 3 acquisition periods\[[@CR47]\]QuEChERS, dilute and shoot38 mycotoxins and 288 pesticidesApple baby food, wheat flour, paprika, black pepper, sunflower seedQTRAP 5500 instrument (Sciex) coupled to Acquity UHPLC system (Waters)Acquity UPLC HSS T3 end-capped C~18~ column (100 mm × 2.1 mm, 1.8 μm), 40 °C, flow rate 350--700 μL/min, injection volume 3 μL; ESI+ eluent A: H~2~O, eluent B: CH~3~OH, both containing 0.2% HCOOH + 5 mM NH~4~HCOO^-^; ESI− eluent A H~2~O, eluent B CH~3~OH, both containing 0.2% HCOOH + 5 mM NH~4~HCOO^-^; gradient: 10% eluent B with flow rate 350 μL/min increased to 50% eluent B in 1 min, linear increase to 100% eluent B within 10 min and simultaneous increase of flow rate to 550 μL/min, flow rate 0.7 μL/min at 100% eluent B, reequilibration for 2.5 min at 10% eluent B at 450 μL/min, total run 15.5 minESI+, ESI−, dMRM, time window for 1 MRM 0.8 min, cycle time 0.55 s\[[@CR48]\]Dilute and shootNo39 mycotoxinsWheat, maizeQTRAP 4000 instrument (Sciex) coupled to 1100 series LC system (Agilent Technologies)Gemini C~18~ column (150 mm × 2 mm, 5 μm) + Gemini C~18~ guard column (4 mm × 2 mm, 5 μm); 40 °C, flow rate 1000 μL/min, injection volume 5 μL; eluent A CH~3~OH--H~2~O--CH~3~COOH (10:89:1, v/v/v), eluent B CH~3~OH--H~2~O--CH~3~COOH (97:2:1, v/v/v), both containing 5 mM NH~4~CH~3~COO^-^; gradient 2 min at 100% eluent A, linear increase to 100% eluent B within 12 min, held at 100% eluent B for 3 min, reequilibration at 100% eluent A for 4 min, total run 19 minESI+, ESI−, dMRM, dwell time 100 ms, pause time 5 ms\[[@CR36]\]Dilute and shootNo295 analytesApple puree, hazelnut, maize, green pepperQTRAP 5500 instrument (Sciex) coupled to 1290 series LC system (Agilent Technologies)Gemini C~18~ column (150 mm × 2 mm, 5 μm) + Gemini C~18~ guard column (4 mm × 2 mm, 5 μm); 40 °C, flow rate 1000 μL/min, injection volume 5 μL; eluent A CH~3~OH--H~2~O--CH~3~COOH (10:89:1, v/v/v), eluent B CH~3~OH--H~2~O--CH~3~COOH (97:2:1, v/v/v), both containing 5 mM NH~4~CH~3~COO^-^; gradient: 2 min at 100% eluent A, linear increase to 50% eluent B within 3 min, linear increase zo 100% eluent B within 9 min, hold at 100% eluent B for 4 min, reequilibration at 100% eluent A for 2.5 min, total run 20.5 minESI+, ESI−, dMRM, MRM window ±27 s for positive mode, ±42 s for negative mode, scan time 1 s\[[@CR49]\]Raw extract, SIDANoAFs, FBs, DON, HT2, T2, OTA, ZENMaize, cereal-based products6490 triple-quadrupole instrument coupled to 1290 series UHPLC system (both Agilent Technologies)Zorbax RRHL Eclipse Plus C~18~ column (100 mm × 2.1 mm, 1.8 μm); 30 °C, flow rate 350 μL/min, injection volume 3 μL; eluent A H~2~O--HCOOH (99.9:0.1, v/v), eluent B CH~3~OH--HCOOH (99.9:0.1, v/v) both containing 5 mM NH~4~HCOO^-^; gradient: 0.5 min at 30% eluent B, linear increase to 100% eluent B in 7.5 min, hold at 100% eluent B for 1.5 min, at 9.6 min back to 30% eluent B, reequilibration at 30% eluent B for 2 min, total run 11.5 minESI±, dMRM, monitoring of 2 transitions (1 quantifier and 1 qualifier)\[[@CR50]\]CH~3~CN--H~2~O (84:16, v/v), SIDABond Elut Mycotoxin SPE cartridges (Agilent Technologies)NIV, DON, FUS-X, DON-3-Glc, 3ADON, 15ADON, HT2, T2, ENNs, BEA, ZENBarley, malt, oat, wheat, maizeQTRAP 4000 instrument (Sciex) coupled to LC-20A Prominence system series LC system (Shimadzu)Hydrosphere RP-C18 column (100 mm × 3 mm, 3 μm) + C~18~ guard column; 40 °C, flow rate 200 μL/min, injection volume 10 μL; eluent A H~2~O--HCOOH (99.9:0.1, v/v), eluent B CH~3~OH--HCOOH (99.9:0.1, v/v); gradient ESI− 2 min at 10% eluent B, linear increase to 99% eluent B in 6 min, hold at 99% eluent B for 7.5 min, for 2 min back to 10% eluent B, reequilibration at 10% eluent B for 9.5 min, total run 25 min; ESI+ 2 min at 10% eluent B, linear increase to 87% eluent B in 6 min, hold at 87% eluent B for 7 min, increase to 100% eluent B in 5 min, hold at 100% eluent B for 3.5 min, for 2 min back to 10% eluent B, reequilibration at 10% eluent B for 9.5 min, total run 34.5 minESI−, ESI+, dMRM, 2 single chromatographic runs, monitoring of 2 transitions (1 quantifier and 1 qualifier)\[[@CR51]\]CH~3~CN--H~2~O (84:16, v/v), evaporation, reconstitution in CH~3~OH and H~2~OSPE (Oasis HLB columns)AFs, OTA, DON, ZEN, T2, HT2Wheat flour, barley flour, crisp breadAccela HPLC system, Exactive HRMS instrument (Thermo Fisher Scientific); 1100 micro-LC system (Agilent Technologies), QTRAP instrument (Applied Biosystems)Kinetex C~18~ column (100 mm × 2.1 mm, 2.6 μm); 40 °C, flow rate 200 μL/min, injection volume 20 μL; eluent A H~2~O, eluent B CH~3~OH, both containing 0.5% CH~3~COOH and 1 mM NH~4~CH~3~COO^-^; gradient 10% eluent B start, until 4 min linear increase to 40% eluent B, 60% eluent B in 27 min, keep for 5 min, reequilibration at 10% eluent B for 7 min, total run 20 minHESI-II (heated-electrospray, ESI+, HCD fragmentation (in-source fragmentation)\[[@CR52]\]QuEChERS (2 g sample, 10 mL 0.1% HCOOH in H~2~O, 3 min shaking, 10 mL CH~3~CN, 3 min shaking, 4 g MgSO~4~, 1 g NaCl, shaking)No additional cleanup3ADON, 15ADON, DON, DON-3-Glc, FUS-X, NIV, HT2, T2, DAS, NEO, AFs, OTA, FBs, STER, ZEN, penitrem A, BEA, *Alternaria* toxins, ergot alkaloidsbarleyAccela HPLC system, Exactive HRMS instrument (Thermo Fisher Scientific)Acquity UPLC HSS T3 column (100 mm × 2.1 mm, 1.8 μm); 40 °C, flow rate 300 μL/min, injection volume 5 μL; eluent A H~2~O with 5 mM NH~4~HCOO^-^ and 0.1% HCOOH, eluent B CH~3~OH; gradient: start with 5% eluent B, increase to 50% eluent B in 6 min, increase to 95% eluent B within 4 min, keep until 15 min of the run, reequilibration at 5% eluent B for 3 minHESI-II, ESI+/ESI−\[[@CR53]\]QuEChERS (2 g sample, 10 mL 0.1% HCOOH in H~2~O, 3 min shaking, 10 mL CH~3~CN, 3 min shaking, 4 g MgSO~4~, 1 g NaCl, 0.5 trisodium citrate dihydrate, shaking)No additional cleanup3ADON,15ADON, DON, DON-3-Glc, FUS-X, NIV, HT2, T2, DAS, NEO, AFs, OTA, FBs, STER, ZEN, mycophenolic acid, MON, BEA, *Alternaria* toxins, ergot alkaloids, culmorinsmalting barleyAcquity UHPLC system (Waters), Q-Exactive system (Thermo Fisher Scientific)Atlantis T3 column (100 mm × 2.1 mm, 3 μm); 30^o^C, flow rate 300 μL/min, injection volume 2 μL; eluent A CH~3~CN--H~2~O--CH~3~COOH (95:4.9:0.1, v/v/v), eluent B H~2~O--CH~3~COOH (99.9:0.1, v/v), both containing 5 mM NH~4~CH~3~COO^-^; gradient 5% eluent A start for 1 min, increase to 15% eluent A in 14 min, increase to 100% eluent A in next 15 min, kept at 100% eluent A for 3 min, reequilibration for 4.4 minHESI-II (positive, negative)*15ADON* 15-acetyldeoxynivalenol, *3ADON* 3-acetyldeoxynivalenol, *AFs* aflatoxins, *APCI* atmospheric pressure chemical ionization, *BEA* beauvericin, *CIT* citrinin, *DON-3-Glc* deoxynivalenol 3-glucoside, *DAS* diacetoxyscirpenol, *dMRM* dynamic multiple-reaction monitoring, *DON* deoxynivalenol, *ENNs* enniatins, *ESI* electrospray ionization, *FBs* fumonisins B, *FUS-X* fusarenon X, *HCD* high-energy collisional dissociation, *HPLC* high-performance liquid chromatography, *HRMS* high-resolution mass specrometry, *HT2* HT-2 toxin, *IAC* immunoaffinity chromatography, *MON* moniliformin, *MRM* multiple-reaction monitoring, *NEO* neosolaniol, *NIV* nivalenol, *OTA* typo in ocratoxin A, *PBS* phosphate-buffered saline, *QqQ* triple quadrupol, *QuEChERS* quick, easy, cheap, effective, rugged, and safe, *SIDA* stable isotope dilution assay, *SPE* solid-phase extraction, *STER* sterigmatocystin, *T2* T-2 toxin, *UFLC* ultrafast liquid chromatography, *UHPLC* ultrahigh-performance liquid chromatography, *UPLC* ultraperformance liquid chromatography, *ZEA* zearalanol, *ZEN* zearalenone In MS/MS, the generation of analyte ions is followed by the selection of suitable precursor ions, and collision-induced dissociation (fragmentation) to form product ions that are further detected. LC--MS/MS techniques using a QqQ (and QLIT) mass spectrometer are the most frequently used approach in targeted mycotoxin determination. QqQ systems are typically operated in multiple-reaction monitoring (MRM) mode based on recording two or more precursor ion to product ion transitions. In contrast to the poor performance in full-scan mode, the use of MRM data acquisition provides a significant gain in both sensitivity and selectivity. The excellent sensitivity of QqQ analyzers when operating in SRM mode makes it possible to achieve low microgram per kilogram detection levels \[[@CR54]\]. The average number of mycotoxins cited in the most recently published studies was about 30. Such methods cover a range of well-established mycotoxins, for which analytical standards are available (e.g., trichothecenes, enniatins, AFs B, *Alternaria* toxins, FBs, and ergot alkaloids). The matrices of interest are typically cereals, nuts, seeds, or baby food, but also more complex and problematic matrices such as spices, fruits, herbs, or food supplements are tested. Although MS/MS is generally considered to be a very selective technique especially for problematic matrices, in the case of challenging matrices the MS/MS signal might be overestimated and lost because of the complexity of some samples, which can finally result in false positive findings. In LC--MS methods, positive-mode electrospray ionization (ESI) is almost exclusively used to couple high-performance LC (HPLC) or UHPLC and MS detection. In theory, QqQ mass spectrometers can simultaneously detect a large number of targeted analytes. However, to achieve low LOQs compliant with the regulatory level and have an acceptable number of points per chromatographic peak to ensure accurate and reproducible quantification, the number of simultaneously recorded MRM transitions is limited. This is because of the need for sufficient dwell times for the recording of the respective MRM transitions. Faster instrument electronics and improved design of the collision cell significantly shorten the minimum dwell times that need to be used for each precursor ion--product ion pair monitored. The rapid multimethods that fully exploit the potential of the state-of-the-art QqQ/QLIT instruments are capable of the simultaneous analysis of up to 300 mycotoxins, their metabolites, or other related food contaminants depending on the length of the chromatographic run \[[@CR36], [@CR49], [@CR55]\]. Such multimycotoxin methods are considered to be semiquantitative methods, since no pure analytical standards are available on the market and only in-house purified standards are used for quantification. This brings up a significant disadvantage associated with the use of QqQ MS detectors that do not allow nontargeted analysis (e.g., mycotoxin metabolites), as the detection conditions need to be typically optimized for each analyte with the use of a standard. The main difficulty in LC--MS analysis of complex matrices is matrix effects. Components of the sample matrix can cause suppression (in most cases) or enhancement of the analyte signal during the ionization process and thus affect accurate quantification of analytes, leading to incorrect results, when pure solvent standards are used. A review dealing with matrix effects in LC--MS/MS methods was published in 2010 \[[@CR56]\]. Matrix effects are a complex phenomenon. Their extent is dependent on many factors. First, the chemical structure and polarity of the analyte of interest play an important role. Furthermore, the matrix type and the relative concentrations of the components competing for the charges in the MS interface are also significant. Therefore, the entire sample preparation procedure and the chromatographic and MS conditions have to be optimized to decrease the matrix effects to a minimum \[[@CR56]\]. Ion suppression might be caused by the presence of matrix compounds that are co-eluted with the target analytes and that reduce the ionization efficiency as well as affect the reproducibility and accuracy of the method \[[@CR57]\]. Huge differences in the extent of matrix effects are not only seen for different matrices; high deviations between individual samples of one matrix type are also observed \[[@CR58], [@CR59]\]. Matrix effects were present also after highly extensive immunoaffinity cleanup \[[@CR41], [@CR42]\]. Standard addition is often used in routine analysis. It is applied in single-analyte methods rather than in multiple-analyte determination as it is laborious and costly because of the high consumption of analytical standards and double the number of LC runs \[[@CR60]\]. The last and currently most frequently used approach is isotopically labeled internal calibration. Stable isotopically labeled standards share the same chemical and physical properties as the target analytes, but are still distinct by their different molecular masses. Additionally, they are not present in naturally contaminated samples. The principle of the use of isotopically labeled standards is that of the dilution of naturally abundant isotopic distribution. Therefore, this procedure is called "stable isotope dilution assay" (SIDA). The basic SIDA principle is to transfer the concentration of the analyte into an isotopologue ratio, which has to be stable during all analytical steps. Therefore, the first prerequisite for an internal standard is stable labeling. As carbon--carbon and nitrogen--carbon bonds are very unlikely to be cleaved, mainly labels consisting ^13^C and ^15^N are used for preparation of the isotopically labeled standards for mycotoxins. In contrast, losses of ^18^O and ^2^H can occur if these labels are at labile positions. For instance, ^18^O in carboxyl moieties can be exchanged in acidic and basic solutions. As a concern, deuterium (^2^H) is susceptible to "protium--deuterium exchange" if it is activated by an adjacent carbonyl group or aromatic systems. Moreover, chromatographic shifts caused by the isotope effect are more common in the case of deuterated standards than in the case of ^13^C- and ^15^N-labeled ones. Therefore, ^13^O and ^2^H labeling are less common in mycotoxin quantification using SIDA \[[@CR61]\]. The following paragraphs highlight examples of currently used approaches for LC--MS/MS determination of mycotoxins in cereals and products thereof. The methods range from those intended for official control to multiclass methods used in research for both screening and quantitative purposes. In addition, how to cope with determination of conjugated mycotoxins will also be part of this section. Most of the methods discussed have been developed for the determination of EU-regulated mycotoxins in various matrices to fulfill the strict requirements of EU legislation (e.g., low LODs for AFB~1~ and OTA in processed cereal-based food and baby food for infants and young children) \[[@CR42], [@CR44], [@CR62], [@CR63]\]. Therefore, to minimize the amount of undesirable matrix co-extracts (further discussed later), purification steps are usually required during sample preparation, especially in the case of complex matrices such as cereal-based products. Typical cleanup strategies involve commercially available solid-phase extraction (SPE) cartridges or immunoaffinity columns (IACs). However, despite the use of highly sensitive LC--MS instrumentation, achieving trace detection levels of some analytes in more than 200 multidetection methods is impossible when compromises have to be made with regard to both sample preparation and LC--MS/MS conditions. These methods rely on the injection of raw extracts \[[@CR43], [@CR64]\], "dilute and shoot" approaches \[[@CR36], [@CR47]\], or different modifications of the "quick, easy, cheap, effective, rugged, and safe" (QuEChERS) approach \[[@CR48], [@CR62], [@CR64], [@CR65]\]. Strategies to eliminate matrix effects in (multi)mycotoxin determination {#Sec4} ------------------------------------------------------------------------ ### Solid-phase extraction {#Sec5} Cleanup techniques in mycotoxin analysis were last critically reviewed in 2008 \[[@CR16]\]. For multitoxin analysis, mainly SPE cartridges are used. Nowadays, there are plenty of these from different manufacturers available on the market. For example, multifunctional columns, MycoSep^®^, containing a mixture of charcoal, ion-exchange resins, and other exchange materials are available for various combinations of mycotoxins or single mycotoxins \[OTA, moniliformin, nivalenol (NIV)\] \[[@CR66]\]. Use of MycoSep^®^ columns has become a routine purification step especially for trichothecenes since the 1990s \[[@CR66], [@CR67]\]. The procedure involves extraction with aqueous acetonitrile and passing the extract through the SPE column, and preconcentration of the analytes of interest. Matrix co-extracts are retained on the column packing, and the analytes of interest pass to the supernatant. No washing step is required. One of the first pioneering LC--MS/MS methods using MycoSep^®^ 226 for purification of maize samples before simultaneous detection of *Fusarium* mycotoxins \[NIV, DON, fusarenon X (FUS-X), 3-acetyldeoxynivalenol (3ADON), 15-acetyldeoxynivalenol (15ADON), diacetoxyscirpenol, HT-2 toxin (HT2), T-2 toxin (T2), and ZEN\] was published in 2005 \[[@CR39]\]. Although MycoSep^®^ 227 was also tested and satisfactory recoveries were obtained for most of the trichothecenes, 50% retention of NIV and complete retention of ZEN on the column packing was observed. MycoSep^®^ 226, designed for a broader range of polarities, increased the recovery rate for ZEN. However, more matrix components that also passed through the column caused high signal suppression, which resulted in a recovery of 30% for ZEN. Therefore, the addition of an internal standard (zearalanol) was used for ZEN quantification. The low recovery (50%) obtained for NIV was within the manufacturer's specification. The method performance characteristics summarized in the [electronic supplementary material](#Sec13){ref-type="sec"} show that because of MycoSep-based purification, low LODs can be achieved without highly sensitive state-of-the-art LC--MS/MS systems having to be used. An in-house validation of another method including MycoSep^®^ 226 was done for 17 mycotoxins, including AFs and trichothecenes, in various foods and feed \[[@CR40]\]. Special attention was paid to the selection of proper eluents, the chromatographic column, and MS/MS conditions to achieve the highest possible sensitivity. The LOQs below 0.01 μg/kg for AFs, low intraday and interday precision (RSD below 10%), and high recoveries (70--110%) (see the [electronic supplementary material](#Sec13){ref-type="sec"}) achieved passed the criteria of the official methods for mycotoxin determination in baby food \[[@CR40]\]. Recently, MycoSpin^TM^ 400 for multimycotoxin LC--MS/MS analysis (optimized for most of the EU-regulated mycotoxins) in a spin format became available. So far, it has been applied to the purification of maize-based silages \[[@CR68]\]. The manufacturer claims that highly accurate results are achieved by use of ^13^C-labeled standards \[[@CR66]\]. Dispersive magnetic SPE might be an attractive technique as an alternative to classic SPE in the future \[[@CR96]\]. However, compared with a classic on-column SPE, dispersive magnetic SPE requires the same time for a single extraction. The mechanism occurring in magnetic SPE is analogous to that in classic on-column SPE. The dispersion of the magnetic nanoparticles into the solution containing mycotoxins ensures a continuous and dynamic contact with the adsorbent surface, leading to more efficient analyte retention. The separation of the magnetic material with the adsorbed analytes from the solution is then realized by application of a magnet outside the vessel, avoiding centrifugation or filtration steps. Finally, after washing, analytes are eluted from the magnetic material by a proper solvent mixture. So far, only one study on the use of dispersive magnetic SPE in mycotoxin determination (AFs, ZEN, and OTA) in cereals has been published \[[@CR96]\]. ### Immunoaffinity columns {#Sec6} IACs provide an even more specific cleanup compared with SPE cartridges. The principle is based on antibodies that entrap an analyte of interest. The matrix co-extracts are removed by washing, and "pure analyte" is released from the antibody with use of an organic solvent. IAC cleanup should provide a purified sample completely free of the matrix, which allows the use of a solvent calibration curve for quantification. However, as discussed later \[[@CR41], [@CR42]\], ion suppression/enhancement (caused by the IAC sorbent) has been observed. In the past, IACs were mainly designed for one or two toxins. However, legislative requirements led to the development of IACs for multiple regulated toxins; for example, AFLAOCHRA PREP^®^ (AFs and OTA), AO ZON PREP^®^ (AFs, OTA, and ZEN), and DZT MS-PREP^®^ (DON, ZEN, HT2, and T2). The only IAC that covers most of the EU-regulated mycotoxins is Myco6in1^+^ (VICAM) containing antibodies for AFs, OTA, FBs, DON, ZEN, NIV, T2, and HT2. Although, the application of IACs is a preferred and selective tool for sample cleanup, the sample extraction and handling procedure recommended by the manufacturer is very laborious and time-consuming, and produces large volumes of solvent waste. Special attention has to be paid to the extraction solvent, elution rate, and column capacity to achieve optimal recoveries of targeted analytes \[[@CR41], [@CR42]\]. In addition, column capacity may be hindered as a result of antibody cross-reactivity (affinity for other structurally related toxins). This, on the other hand, is a big advantage in the determination of conjugated mycotoxins (discussed later). An LC--MS/MS method for simultaneous determination of AFs, OTA, and *Fusarium* mycotoxins in maize using multifunctional IACs (AOZFDT2^TM^) was developed and validated in-house \[[@CR41]\]. Despite the laborious cleanup and long chromatographic gradient (59 min), matrix effects were observed (i.e., significant suppression for AFB~1~ and AFG~1~ and a slight suppression for OTA). Although no matrix effects were observed for DON, T2, HT2, FBs, and ZEN, the study authors decided to use matrix-matched standards for quantification. When validated for maize, at EU maximum permitted levels, the method showed satisfactory performance in terms of recovery and repeatability (see the [electronic supplementary material](#Sec13){ref-type="sec"}). The application of these IACs in analysis of maize-based cereals, barley, and peanuts was documented in another study \[[@CR42]\]. The sample preparation procedure was markedly simplified, and the compatibility of several extraction solvents with the IACs was tested. Assessment of matrix effects confirmed the ion suppression observed for AFs and OTA \[[@CR41]\] and showed significant ion enhancement for FBs. The follow-up evaluation on the "IAC solvent" calibration curve (when standards were prepared in solvent obtained from passing through the IAC) revealed that the ion suppression/enhancement was mainly caused by compounds flushed from the IAC packing. Therefore, "IAC solvent" standards can be used for quantification instead of matrix-matched calibration, and so this approach allows the use of one calibration for various matrices \[[@CR42]\]. Other measures were taken to enhance the method; for example, changing from HPLC to UHPLC decreased the chromatographic run time from 59 to 10 min. Similarly, the sample preparation procedure \[[@CR41]\] was simplified, and the modified method was validated for maize, durum wheat, corn flakes, and maize crackers \[[@CR43]\]. Full in-house validation was performed for 12 mycotoxins at three levels (see the [electronic supplementary material](#Sec13){ref-type="sec"}). Another option in determining multiple mycotoxins using IAC cleanup more efficiently is to use different IACs in tandem (i.e., the first column is connected below the glass vessel and the second column is connected below the first one). Thus sample loading, washing, and elution is achieved for two columns simultaneously. The use of columns in tandem offers a range of desired combinations of mycotoxins in accordance with EU regulations, so analytical cleanup can focus on target mycotoxins known to co-occur in specific matrices. OCHRAPREP^®^ and DZT MS-PREP^®^ columns have been used in tandem for analysis of wholemeal bread, AOF MS-PREP^®^ and DZT MS-PREP^®^ IACs have been combined for analysis of a range of maize and maize-based products, including infant foods, and AFLAOCHRA PREP^®^ and DZT MS-PREP^®^ columns have been combined for the analysis of oat-based muesli containing dried fruit and nuts \[[@CR44]\]. In summary, immunoaffinity chromatography is still the most powerful cleanup method for up to six mycotoxins especially when used before LC-based (no MS) detection \[[@CR62], [@CR63]\]. Because of recent trends leading to the use of LC--MS/MS, it is no longer necessary in laboratories with sophisticated instrumentation. However, IACs are often used in LC--MS/MS when low LODs are required (e.g., OTA and AFB~1~ in baby food) \[[@CR64]\]. ### LC--MS/MS with limited cleanup or without cleanup {#Sec7} The availability of sensitive LC--MS instruments that are less prone to matrix effects has led to the development and application of so-called multiple-analyte approaches. Hence, a clear trend toward the determination of a range of different mycotoxin classes can be observed, both in the literature and in research and routine laboratories. Because of the broad physicochemical properties of mycotoxins, a compromise, using nonspecific, minimal cleanup (if needed), has to be used to avoid a discrimination of some analytes during sample preparation processing. #### QuEChERS approach {#FPar1} As a minimal cleanup, the QuEChERS approach is being used in multiclass analysis \[[@CR69]\]. This approach was developed for multipesticide analysis, for very fast extraction and purification. The key principle is the partitioning of an acetonitrile--water mixture induced by addition of inorganic salts. While the analytes are largely transferred into an organic phase, more polar matrix impurities are left in an aqueous layer. Nevertheless, use of such a basic cleanup for multiresidue analysis in complex matrices leads inevitably to matrix effects, and thus affects sensitivity adversely. For this reason, LC--MS/MS methods including QuEChERS-based protocols are generally inefficient for the detection of AFs and OTA in baby foods at the EU limits. Hence, for these specific metabolites in baby foods, a multiresidue approach is often abandoned in favor of dedicated methods making use of specific cleanup with IACs \[[@CR64]\] or a combination with another cleanup technique is used \[[@CR41], [@CR42], [@CR45], [@CR46], [@CR64]\]. The original QuEChERS method for determination of pesticide residues consists of several steps: (1) extraction with acetonitrile, (2) partitioning step with magnesium sulfate (MgSO~4~) and sodium chloride (NaCl), (3) addition of an internal standard, (4) dispersive SPE---aliquot purified with MgSO~4~ and SPE sorbents \[e.g., primary--secondary amine (PSA) salts, C~18~, C~8~\], and (5) addition of an analyte protectant and adjustment of the pH. The implementation of the QuEChERS method in mycotoxin determination required some modifications depending on the spectrum of the analytes and the character of the matrix. For instance, PSA used for the removal of polar matrix components caused significant loss of FBs \[[@CR45]\], and/or addition of water before acetonitrile extraction was needed for low water content matrices to increase the recovery yields \[[@CR45], [@CR47]\]. The QuEChERS-like approach was used for the development of an LC--MS/MS method for 17 mycotoxins in cereals for human consumption and infant cereals \[[@CR45]\]. Besides the aforementioned modifications, the extraction was done with acidified acetonitrile to increase recoveries for FBs. Direct analysis of the extract after the partitioning step resulted in significant matrix effects, and thus insufficient sensitivity (especially for AFs). Several dispersive solid-phase extraction (SPE) sorbents were tested (SPE, Oasis HLB, Carbograph 4, C~18~, or dispersive SPE with both PSA and C~18~ modified silica gel). Finally, a simple defatting step with *n*-hexane followed by a two-step sequential reconstitution in aqueous methanol was shown to be the best adaptation for all analyte--matrix combinations. Although the performance characteristics of the method fulfilled the EU legislation criteria \[[@CR33]\], it is not appropriate for official control of infant cereals. The maximum EU legislation level for AFB~1~ is 0.1 μg/kg but the LOQ was 1 μg/kg. The current method was applied in the analysis of more matrices (cereals, cocoa, oil, spices, infant formula, coffee, and nuts) and validated. Matrix effects were successfully corrected by ^13^C-labeled standards. To achieve lower LOQs for AFs and OTA in baby food, an additional cleanup step (immunoaffinity chromatography) was applied \[[@CR64]\]. Positive identification of mycotoxins in the matrix was conducted according to the confirmation criteria defined in Commission Decision 2002/657/EC \[[@CR29]\], while quantification was performed by isotopic dilution using ^13^C-labeled mycotoxins as internal standards. The LOQs were at or below the maximum levels set by Commission Regulation (EC) No 1881/2006 \[[@CR14]\] for all regulated mycotoxin--matrix combinations. In particular, the inclusion of an immunoaffinity chromatography step allowed LOQs as low as 0.05 and 0.25 μg/kg to be achieved in cereals for AFs and OTA, respectively \[[@CR64]\]. Another QuEChERS modification followed by LC--ESI-MS/MS was introduced for the determination of EU-regulated mycotoxins in wheat, maize, and rice \[[@CR46]\]. Water soaking and acidified acetonitrile extraction with a mixture of magnesium sulfate, sodium chloride, sodium citrate tribasic dihydrate, and sodium citrate dibasic sesquihydrate (4:1:1:0.5) was followed by dispersive SPE with a mixture of magnesium sulfate and C~18~ sorbent. Purified extract was evaporated and reconstituted in aqueous methanol. The validation data obtained are summarized in the [electronic supplementary material](#Sec13){ref-type="sec"}. It is worth noting that the greatest matrix effects were observed for maize. #### "Dilute and shoot" approach {#FPar2} "Dilute and shoot" and the injection of raw extract, without any cleanup, is now commonly performed in multiresidue LC--MS analysis \[[@CR47], [@CR48], [@CR65]\]. The dilute and shoot LC--MS/MS-based method is considered a pioneering method in the field of multimycotoxin determination \[[@CR48]\]. Simple extraction with a mixture of acetonitrile--water--acetic acid (79:20:1, v/v/v) further diluted 1:1 with acetonitrile--water--acetic acid (20:79:1, v/v/v) was used in the determination of 39 mycotoxins, including conjugated metabolites, in cereals. Although the MS/MS parameters for most of the analytes could have been optimized in both polarities (i.e., analytes give the MS/MS signal in negative ionization mode as well as in positive ionization mode), some analytes (moniliformin, NIV, ZEN 14-glucoside) gave no or very weak signals in the positive mode. Because of the high number of analytes, the determination was performed in two chromatographic runs (positive and negative) to avoid losses in sensitivity caused by polarity switching. The method was validated for wheat and maize. Ion suppression effects caused by co-eluted matrix components were negligible in the case of wheat, whereas significant signal suppression for 12 analytes was observed in maize. The apparent recoveries were within the range of (100 ± 10)% for half of the analytes; in extreme cases the apparent recovery dropped to 20%. As an example of an analyte with low apparent recovery, FB~1~ (34%) can be given. Apparent recovery of 113% was observed for a common contaminant of maize, ZEN. Nevertheless, this was caused by matrix effects rather than low extraction recovery, and can thus be compensated by the use of matrix-matched standards. The method performance characteristics for some analytes are given in the [electronic supplementary material](#Sec13){ref-type="sec"}. The method has been continuously extended to include 331 bacterial, plant, and fungal metabolites, and has been fully validated for 295 analytes in maize and three other matrices \[[@CR36]\]. To successfully acquire as many MRM transitions with acceptable sensitivity and repeatability in a reasonable time, the method was transferred onto the next generation of a QTRAP instrument (QTRAP 5500). As no guidelines are available for multidetection of mycotoxins, the validation procedure was performed according to SANTE 11945/2015 \[[@CR34]\], and the trueness of the method was demonstrated with use of samples from organized ring trials. With regard to the apparent recovery in maize, 62% of 295 analytes matched the acceptable recovery range of 70--120% laid down in SANTE 11945/2015 at the highest spiking level. At the levels close to the LOQ, 57% of the analytes fulfilled this criterion. The extent of matrix effects was strongly dependent on the analyte--matrix combinations. No matrix effects were observed for 45% analytes at the highest spiking level and 35% of analytes at the lowest spiking level. The repeatability of the method was acceptable (RSD ≤ 20%) for 95% of the analytes. The trueness of the method was proved by participation in ring trials. The calculated *z* scores were satisfactory for all maize samples analyzed (i.e., between −2 and 2) and also for a broad variety of different matrices, which proves that the method provides accurate results also for other "nonvalidated" matrices. A critical assessment of extraction methods in the simultaneous analysis of 288 pesticides and 38 mycotoxins was performed in another study \[[@CR47]\]. Three extraction procedures were performed for wheat and other matrices: aqueous acetonitrile extraction followed by a modified QuEChERS approach, aqueous acetonitrile extraction, and pure acetonitrile extraction. Different eluent modifiers were used for positive-mode ESI and negative-mode ESI measurements to obtain high sensitivity and sharper peak shape. For positive-mode ESI, two to four times higher responses were observed in the presence of ammonium formate for most of the analytes. Extraction with pure acetonitrile was not efficient in terms of recoveries, whereas the QuEChERS approach and extraction with aqueous methanol showed satisfactory recoveries within the range of 70--120% with RSD less than 20% \[[@CR34]\] for most of the analyte--matrix combinations. Although the QuEChERS-like method resulted in lower LOQ and more consistent results, the recoveries were lower in particular for polar analytes \[DON 3-glucoside (DON-3-Glc), NIV, T2 tetraol\] because of the partitioning step. Finally, QuEChERS-like extraction was chosen as the most suitable approach for the analytes tested \[[@CR47]\]. #### SIDA methods {#FPar3} The correction of matrix effects in LC--MS/MS methods using limited sample cleanup can be done with SIDAs; however, this approach also has its limitations. Although the spectrum of commercially available isotopically labeled standards is getting broader, it is still mainly limited to EU-regulated mycotoxins or to those considered by EFSA. Moreover, the cost of these internal standards is far greater than that of those of natural origin, which significantly increases the cost of the analysis \[[@CR56]\]. A review of the application of SIDA in mycotoxin analysis was published in 2008 by Rychlik and Asam \[[@CR61]\]. Therefore, the following discussion is focused on a few examples of SIDA applications in multimycotoxin LC--MS/MS methods. A UHPLC--MS/MS method for the determination of all EU-regulated mycotoxins in maize and cereal-based products was developed \[[@CR70]\]. The accuracy was enhanced by the application of ^13^C-labeled compounds for each of the target analytes before UHPLC--MS/MS analysis. The simple raw-extract-injection technique was validated as a confirmatory method according to Commission Decision 2002/657/EC \[[@CR29]\]. The trueness of the method was verified by the measurement of 12 test materials from different providers with well-defined analyte concentrations. Method performance parameters were evaluated for maize (see the [electronic supplementary material](#Sec13){ref-type="sec"}). The sample preparation consisted of two extraction steps: (1) extraction with acetonitrile--water--formic acid (80:19:0.1, v/v/v); (2) extraction of the residue with acetonitrile--water--formic acid (20:79.9:0.1, v/v/v). Both extracts were combined and centrifuged, and the raw extract was fortified with a mixture of labeled standards in a ratio of 4:1 (v/v) before injection. A drawback of this procedure is that more matrix compounds are extracted because of the high water content of the second extraction solvent. However, the use of internal standards efficiently compensated for all matrix effects for all target analytes. An 11.5-min UHPLC gradient elution using methanol and water containing 5 mM ammonium formate and 0.1% formic acid provided a capacity factor *k*′ of the first analyte eluted (DON) of more than 1 (actually 1.5), and acceptable resolution and peak shape for all analytes, which fulfills the Commission Decision 2002/657/EC criteria \[[@CR29]\]. The MS detection was performed using the dynamic MRM mode and fast polarity switching. Dynamic MRM is a technique that monitors the analytes only around the expected retention time, and thus decreases the number of co-occurring MRM transitions, allowing both the cycle time and the dwell time to be optimized for the highest sensitivity, accuracy, and reproducibility. Fast polarity switching needs less than 0.5 s for switching between positive and negative modes, which reduces the losses in sensitivity. As the method fulfilled all the European Commission criteria, it is suitable for routine analysis of maize for official control \[[@CR33]\]. A combination of SIDA and SPE cleanup was used in the determination of EU-regulated and EFSA-recommended mycotoxins in cereals. Moreover, conjugated forms of DON (DON-3-Glc, 15ADON, 3ADON) were included \[[@CR50]\]. Isotopically labeled standards of 3ADON, T2, enniatins, and beauvericin were prepared in-house. For the ^13^C-labeled equivalents of 15ADON and DON-3-Glc, ^13^C-labeled 3ADON and ^13^C-labeled DON, respectively, were used for correction. Similarly to the previous study \[[@CR70]\], internal standards were added to the raw extract, which then underwent SPE cleanup using Bond Elut Mycotoxin cartridges (Agilent Technologies). Special attention had to be paid to the chromatographic separation of DON and DON-3-Glc because of in-source fragmentation of DON-3-Glc. To avoid decreases in sensitivity, the analysis was performed in two single chromatographic runs (positive-mode ESI and negative-mode ESI). Detailed information about the method performance characteristics obtained with use of SIDA is given in the [electronic supplementary material](#Sec13){ref-type="sec"}. The mycotoxins for which ^13^C-labeled standards were not available were evaluated by matrix-matched calibration of potato starch. The accuracy was confirmed by analysis of commercially available reference materials and samples from interlaboratory testing. This method was later applied to beer \[[@CR71]\]. Although satisfactory recoveries were achieved, the LOQ of 20 μg/L for DON-3-Glc is too high for beer control, where common levels of DON-3-Glc are usually lower than 20 μg/L. To achieve a lower LOD and LOQ, the method needs to be optimized further (i.e., use of any cleanup, change of the chromatographic gradient, or use of a more sensitive LC--MS/MS instrument). A dilute and shoot method for the LC--MS/MS determination of multiple mycotoxins (AFs, OTA, FBs, ZEN, DON, T2, and HT2) in wines and beers has been developed and validated. Separation was accomplished by UHPLC with an analysis time of less than 10 min. Mycotoxins were detected by dynamic MRM in positive-mode ESI. To reduce matrix effects, ^13^C-labeled mycotoxin standards were added to the sample extracts before LC--MS/MS analysis. With external calibration, the recoveries were 18−148% for white wines, 15−118% for red wines, and 20−125% for beers, at three spiking levels. The ^13^C-labeled internal standards compensated for matrix effects effectively, with overall recoveries of 94−112% for white wines, 80−137% for red wines, and 61−131% for beers, with greater recoveries for FBs, at three spiking levels. The RSD was less than 20% for all analytes in the wines and beers. This method was applied in a survey of domestic and imported wines and beers for the determination of OTA, and was extended to include other mycotoxins \[[@CR72]\]. LC--MS for the determination of masked mycotoxins {#Sec8} ================================================= Masked mycotoxins (also conjugated by plants), which are plant metabolites of mycotoxins, are a well-known and significant subgroup of these natural contaminants. Research on these derivatives has expanded tremendously, especially within the last 10 years, during which time their forms, occurrence, and principles of origin and metabolization in plants have been continuously elucidated. As a result, the development of analytical methods to measure these metabolites has progressed. The topic of masked mycotoxins was comprehensively reviewed by Berthiller et al. \[[@CR73]\]. First, the terminology used in the field of masked mycotoxins should be clarified. The term "masked mycotoxins" refers to a specific group of mycotoxin metabolites that are created by plants as their defense against various xenobiotics. The masked mycotoxins can be further subdivided into the two categories of conjugated and bound mycotoxins. Conjugated mycotoxins can be extracted from samples (plants, cereals, food) and detected by further described analytical strategies. In contrast, bound mycotoxins cannot be extracted directly from samples of interest since they are covalently or noncovalently attached to polymeric carbohydrate or protein structures and have to be released from the matrix by chemical or enzymatic treatment first \[[@CR74]\]. Regarding analytical methods, basically there are two possible strategies for testing, direct and indirect, depending on the particular analytes and the available laboratory equipment. In the field of direct methods, which cover the soluble forms of these metabolites, the LC--MS approaches are the techniques of choice for accurate, fast, and specific analysis of masked mycotoxins. Unfortunately, for most of the masked mycotoxins analytical standards are not commercially available, and thus in-house purified standards have to be prepared first. This production is laborious and time-consuming. Since masked mycotoxins tend to be more polar than their parent toxins (glycosylated, sulfated, acetylated forms), their determination can be done easily by an analytical procedure developed for the determination of native forms of mycotoxins. This means, as also previously demonstrated by several authors, that the best recovery of masked mycotoxins can be obtained through extraction with acidified acetonitrile--methanol and water mixtures, which is the procedure widely applied in multimycotoxin analysis. From the available published methods and protocols, it is clear that acetonitrile--water mixtures ranging from 80% to 84% acetonitrile, simple or acidified with 1% acetic acid, are the most versatile extraction solvents, yielding sufficient recoveries (mostly above 70% regardless of the type of matrix) of various masked trichothecenes (DON-3-Glc, T2 glucoside, HT2 glucoside, NIV glucoside) or ZENs (ZEN 14-glucoside, ZEN 14-sulfate). Although several studies dealt with various cleanup strategies, it was uniquely determined that neither SPE cartridges nor IACs are suitable for their analysis, since only very low recoveries (less than 50%) are commonly obtained \[[@CR73]\]. Generally, the dilute and shoot strategy has been proven to provide the best recoveries and data repeatability. In contrast, the QuEChERS approach provides significantly lower recoveries of masked forms compared with the parent compounds. In the case of the QuEChERS approach, analytes are typically not transferred into the acetonitrile phase because of their polar character, which also causes retention problems in SPE cartridges. Chromatographic separation can be easily achieved by use of traditional C~18~ columns, although methods based on hydrophobic interaction LC could also be advantageous. Problems as a result of the lack of analytical standards can be avoided by the application of indirect methods when primary hydrolysis (enzymatic, acetic, and basic) of conjugated mycotoxins is required before analysis \[[@CR73]\]. The significant disadvantage of these methods is the impossibility of complete hydrolysis of conjugates and its confirmation, which results in problematic quantification and underestimation of results. High-resolution (LC--HRMS) approaches for the targeted determination of mycotoxins {#Sec9} ================================================================================== Currently, the use of HRMS is gaining increasing interest in both research and routine laboratories. Advanced HRMS instruments combine crucial features such as increased selectivity and mass resolution, lower cost, and relatively easy maintenance. HRMS is provided by two types of analyzers, TOF and Orbitrap analyzers, with resolving power of 10,000--100,000 and 140,000--240,000 (full width at half maximum defined at *m*/*z*), respectively. In contrast to MS/MS, HRMS techniques can overcome all the limiting factors of SIM/MRM analyte detection, thus representing an appropriate alternative to the use of QqQ instruments for targeted as well as nontargeted compound detection. Generally, HRMS measurements have enhanced performance in terms of confirmatory capabilities compared with MS/MS measurements. Especially when one is dealing with complex sample matrices, adequate mass resolution is essential, and the absence of noise can cause a significant decrease of the signal-to-noise ratio \[[@CR47]\]. The use of accurate mass measurement permits full spectral data acquisition for all ions, does not rely on fragmentation of analytes, which means that it can overcome the problems caused by production of transitions (nonspecific transitions and stable adducts), and allows retrospective data mining of the chromatogram to look for additional compounds of interest, predominantly metabolites. Overall, HRMS analyzers operate in full-scan mode, and thus allow target, posttarget, and nontarget analysis in a single run without time-consuming optimization of MRM conditions for each compound. The analysis of accurate mass reduces matrix interferences within one mass unit of the target analytes that are normally detected in QqQ systems; however, sole measurement of accurate mass for analyte identification can lead to false positive results or misidentifications. A new generation of hybrid instruments, quadrupole--TOF (QTOF) and Q--orbital ion trap (Q Exactive) instruments, combine the benefits of high-performance quadrupole selection of precursor ions with those of high-resolution mass detection. Furthermore, the use of these systems also allows higher selectivity in comparison with targeted QqQ systems and simultaneously allows retroactive processing of samples for other untargeted compounds of interest. To collect MS/MS data for nontargeted compounds either a data-dependent acquisition (DDA) strategy or a data-independent acquisition (DIA) strategy can be used. In DDA, a limited number of ions with highest abundance detected in the full MS scan are isolated and fragmented in a product ion scan experiment. The approach based on DIA involves sequential isolation of windows across a mass range for MS/MS. The cycle is repeated throughout the LC run ensuring that product ion spectra of all ions are recorded. Moreover, the DIA approach allows the use of fragment ions for quantification \[[@CR75]\]. Application of DIA and DDA in analysis of mycotoxins was demonstrated in several recent studies \[[@CR76], [@CR77]\]. Finally with respect to Commission Decision 2002/657/EC \[[@CR29]\] for HRMS measurements, each ion earns two identification points, so quantification with the molecular ion and confirmation with one product ion gives enough points to confirm any substance. The advantages and disadvantages of LC--HRMS approaches over targeted LC--MS/MS are summarized in Table [3](#Tab3){ref-type="table"}.Table 3Comparison, advantages and drawbacks of mass spectrometry (MS) instrumentationMS techniqueAnalyzerProsConsLRMS(/MS)QqQ, IT, QLITHigh sensitivity and selectivity in MRM mode\ Wide linear dynamic range\ Robust gold standard instrumentation\ Lower purchase costsNumber of simultaneously detected analytes in MRM mode is limited\ Poor/moderate sensitivity in full MS mode\ Need to optimize detection conditions for each analyte\ Screening without reference standard not usually possibleHRMS(/MS)TOF, QTOFHigh sensitivity and selectivity in full MS mode\ Postacquisition data interrogation for nontargeted analytes\ Identification of unknowns is possibleNarrower dynamic range compared with LC--MS(MS/MS) instruments\ High purchase costLower mass resolving power and mass accuracy compared with Orbitrap systemsOrbitrap, Q--OrbitrapAcquisition speed limited at high mass resolving power settings*HRMS* high resolution mass spectrometry, *LC* liquid chromatography, *LRMS* low-resolution mass spectrometry, *MRM* multiple-reaction monitoring, *QLIT* quadrupole--linear ion trap, *Q--Orbitrap* quadrupole--Orbitrap, *QqQ* triple quadrupole, *TOF* time of flight Despite the ability of HRMS instruments to simultaneously detect various analytes that can be confirmed by MS/MS, this technique is not extensively used for multimycotoxin analysis, and only a couple of methods have been published so far (Fig. [3](#Fig3){ref-type="fig"}). Tanaka et al. \[[@CR78]\] successfully applied the combination of LC--atmospheric pressure chemical ionization TOF MS for the determination of trichothecenes, ZEN, and AFs in cereal-based food. However, additional SPE cleanup of samples was needed to obtain sufficient sensitivity for all analytes. This pioneering study was followed by several others, and a comprehensive study by Mol et al. \[[@CR65]\]. In that study, four different extraction approaches and UHPLC--TOF MS and MS/MS for the determination of mycotoxins, pesticides, plant toxins, and veterinary drugs were compared. It was shown that a procedure using water--acetonitrile--1% formic acid is applicable for the extraction of multiple food contaminants from different matrices. UHPLC--TOF and UHPLC--Orbitrap mass analyzers were applied to examine major *Fusarium* mycotoxins (FBs, DON, 3ADON, 15ADON, NIV, HT2, T2, ZEN, DON-3-Glc, FUS-X) in cereals. QuEChERS and aqueous acetonitrile extractions were applied before instrumental determination of the analytes. From published results, it can be concluded that both techniques are fit for purpose for the determination of the mycotoxins tested, but the approach using TOF MS requires an additional cleanup strategy to achieve sufficient sensitivity for all targeted analytes \[[@CR79]\]. A hybrid QTOF system for the determination of trichothecenes in wheat, corn, rice, and noodles was used in another study \[[@CR80]\]. The most comprehensive studies describing the use of HRMS techniques in multimycotoxin analysis were published by Herebian et al. \[[@CR81]\], Rubert et al. \[[@CR52]\], De Dominics et al. \[[@CR82]\], and Beccari et al. \[[@CR53]\]. These articles cover the main mycotoxin representatives of *Fusarium*, *Claviceps*, *Aspergillus*, *Penicillium,* and *Alternaria* fungi and the applicability of TOF and Orbitrap MS systems for both screening and quantitative analysis of the respective toxins in cereal-based matrices. In one study \[[@CR81]\] the determination of undiluted acetonitrile--water extracts of cereals was with HPLC--MS/MS and microcapillary HPLC--LTQ Orbitrap MS instruments. It was also concluded that the HRMS technique is fully usable and a time-saving method for rapid and accurate analysis of mycotoxins. LC--Orbitrap MS instrumentation was used for the determination of all legislatively regulated mycotoxins in wheat and barley flours and crisp bread \[[@CR52]\]. With the support of the SPE cleanup of extracts, it was possible to quantify all desired mycotoxins, and all validation data met the current European regulatory requirements for LC--MS confirmatory analysis \[[@CR29]\]. Critical evaluation of four different extraction procedures (modified QuEChERS extraction, matrix solid-phase dispersion, solid--liquid extraction, and SPE cleanup) for simultaneous determination of 32 different mycotoxins was conducted \[[@CR52]\]. Separation and detection of the analytes was performed by the UHPLC--Orbitrap MS method. The only extraction procedure that was mutually developed with the HRMS method and capable of determining all mycotoxins tested was the QuEChERS procedure (see Table [2](#Tab2){ref-type="table"}). LC--HRMS-based approaches for the qualitative screening of mycotoxins {#Sec10} ===================================================================== LC--HRMS-based techniques have clearly been shown to be applicable as reliable detection tools for the screening of conjugated mycotoxins and various mycotoxin metabolites for which no analytical standards are available. The increasing interest in the use of HRMS is evident from the number of articles that have been published \[[@CR51], [@CR83], [@CR84]\]. The presence of glycosylated metabolites of HT2 and T2 in wheat, oat, and barley was confirmed by Veprikova et al. \[[@CR83]\]. Cereal extracts were purified through IACs, which allowed the purification and preconcentration of the glycosylated metabolites. The use of this specific cleanup in combination with a UHPLC--QqTOF system operating in MS and MS/MS modes allowed the identification of HT2 glucoside and T2 glucoside. Moreover, HT2 diglucoside and T2 diglucoside in barley were found for the first time \[[@CR83]\]. The occurrence of HT2, T2, and their glycosides in cereals and their fate during malting were studied \[[@CR51]\]. Because of the use of an HPLC--Orbitrap system (HRMS), a retrospective analysis of the full-scan HRMS chromatograms was possible, and presence of neosolaniol, DAS, and their monoglucosides was also successfully evaluated. The study confirmed a common co-occurrence of HT2, T2, HT2 glucoside, and T2 glucoside in wheat, oat, and barley raw grains. Furthermore, also traces of neosolaniol glucoside and DAS glucoside were found. The preliminary investigation on the fate of HT2 and T2 and their glucosylated forms during malting revealed a general mycotoxin reduction from cleaned barley to malt \[[@CR51]\]. In another study, novel masked mycotoxins, FUS-X glucoside and NIV glucoside, were identified in wheat with use of an HPLC--Orbitrap system operating in full-scan and in-source MS/MS modes. The extract was purified by an SPE column to achieve higher signal for these analytes and thus facilitate their reliable identification. Both glucosides were detected in negative mode with a mass deviation of less than -1.1 ppm for \[M − H\]^-^ and -0.83 ppm for \[M − CH~3~COO\]^-^ \[[@CR84]\]. Metabolomics approaches to study mycotoxin metabolism in cereals {#Sec11} ================================================================ Recently, metabolomics-based approaches have found their place in the mycotoxin research field to (1) study the metabolism of mycotoxins in grains, especially in resistant plant varieties such as *Fusarium*-resistant wheat \[[@CR85]\], (2) explore the co-occurrence of these secondary fungal metabolites with plant metabolites, and (3) reveal the formation of so far unknown metabolites that can also be present in cereal-based food. In line with this, two different strategies can be distinguished: targeted and untargeted metabolomics. In targeted approaches, the abundances of metabolites of a set of predefined known substances are determined. Such an approach allows absolute quantification but it is usually limited to metabolites for which reference standards are available. Untargeted approaches aim to record MS features of all detectable compounds, including those unknown at the time of sample analysis. This approach has therefore the advantage of probing the entire metabolic space and can obtain relative abundances of several hundred known and unknown metabolites. Both targeted and untargeted LC--HRMS-based metabolomic studies follow a general workflow consisting of several steps: (1) experimental design, (2) sample preparation, (3) chromatographic separation and MS detection, (4) data acquisition, (5) data processing, and (6) data analysis and interpretation. These workflows were recently summarized in several comprehensive reviews \[[@CR86]--[@CR88]\]. Several studies dealing with the biotransformation of mycotoxins in plants using a recently developed untargeted stable isotope labeling (SIL)-assisted approach using ^13^C-labeled standards in combination with LC--HRMS have been published \[[@CR89]--[@CR91]\]. All the studies share a common experimental setup. Briefly, the plant ears were spiked with a 1:1 mixture of ^12^C toxin and ^13^C-labeled toxin in triplicate at different time points depending on the experiment. As blank control samples, "mock" (i.e., water) treated plant ears were prepared. The harvested ears frozen at -80 °C were milled, and then extracted with a common extraction mixture used for mycotoxins. A UHPLC--LTQ Orbitrap XL system equipped with an ESI source was used. The chromatography columns (commonly C~18~) and gradient as well as the HRMS conditions were chosen and set on the basis of the parent toxin studied. Qualitative analysis of LC--HRMS data was performed with Xcalibur 2.1.0 QualBrowser, and for relative and absolute quantification of known analytes Xcalibur 2.1.0 QuanBrowser was used. Data were further evaluated with MetExtract \[[@CR92]\], which was programmed to automatically extract the corresponding MS peak pairs in mass spectra of a 1:1 mixture of the sample containing^12^C toxin and ^13^C-labeled toxin. The metabolites were putatively identified on the basis of the specific criteria \[[@CR92]\]. In a DON-metabolism study in wheat, MetExtract data processing revealed a total of 57 ion pairs in the full-scan chromatogram of DON-treated samples that passed the aforementioned criteria. DON and its nine biotransformation products were detected in wheat samples treated with a 1:1 mixture of ^12^C DON and ^13^C-labeled DON (Fig. [4](#Fig4){ref-type="fig"}). All DON metabolites identified contained the intact carbon skeleton of DON in their molecular structure. Moreover, no DON biotransformation products containing fewer than 15 carbon atoms were detected. DON-3-Glc was found as a main metabolite, which was confirmed by measurement of an analytical standard. Beside glycosylation, the glutathione pathway, where DON *S*-cysteine and DON *S*-cysteinylglycine were identified, was described in wheat for the first time in this study \[[@CR89]\]. Moreover, five unknown DON conjugates were found.Fig. 4Metabolism of deoxynivalenol (DON) in wheat. Sample treated with 1:1 native DON and ^13^C-labeled DON (tracer). Deoxynivalenol 3-glucoside (DON-Glc) was found as a major metabolite. Mass spectrometry (MS) scan of DON and ^13^C-labeled DON (A). Extracted ion chromatogram of DON (B). Extracted ion chromatogram of DON-Glc (C). MS scan of DON-3-Glc and ^13^C-labeled DON-Glc (D). Extracted ion chromatogram of ^13^C-labeled DON (E). Extracted ion chromatogram of ^13^C-labeled DON-Glc (F) Combination of SIL and LC--Orbitrap MS in fast polarity switching mode followed by MetExtract data processing was applied in a study of HT2 and T2 metabolism in barley. Additionally, a QTOF instrument was used for acquisition of MS/MS spectra, which were needed for further structure annotation \[[@CR90]\]. In total, nine HT2 and 13 T2 metabolites were annotated and partly identified. The metabolism routes in barley covered hydrolysis of acetyl and isovaleryl groups, and hydroxylation as well as covalent binding of glucose, malonic acid, acetic acid, and ferulic acid. A major metabolite of HT2 and T2 metabolism, HT2 3-*O*-β-glucoside, formed at the maximum level as soon as the first day after toxin application and was further metabolized. Other putatively identified metabolites included 15-acetyl-T2 tetraol malonylglucoside, hydroxy-HT2 glucoside, hydroxy-HT2 malonylglucoside, HT2 diglucoside, HT2 malonylglucoside, and feruloyl-T2 \[[@CR90]\]. Similarly, a QTOF system in MS and MS/MS mode was used for the identification of HT2 and T2 metabolites in wheat \[[@CR91]\]. Together with the use of SIL, it allowed putative annotation of 11 HT2 and 12 T2 metabolites. It was confirmed that the metabolism route did not differ from that in barley (i.e., T2 was rapidly converted to HT2, which was further metabolized to HT2 3-*O*-β-glucoside). In contrast to DON metabolism in wheat, no glutathione metabolite was found in case of HT2 and T2. Untargeted SIL profiling using sensitive HRMS instrumentation is undoubtedly a highly efficient tool for studying mycotoxin metabolism in plants. Currently, this technique is being used during baking on an industrial scale to reveal the formation of degradation products of mycotoxins. Moreover, such tracer-fade studies could also be used for elucidation of elevated levels of masked mycotoxins found during malting and brewing \[[@CR93], [@CR94]\]. Conclusions and outlook {#Sec12} ======================= One of the biggest challenges in mycotoxin analysis is still the sampling issue, for which guidance has become available \[[@CR33]\]. However, it remains a difficult and tedious task to obtain a representative sample. Subsequent extraction with appropriate solvents that match the range of multiple co-occurring mycotoxins to be determined is another crucial step, followed by proper cleanup. The latter is dependent on the final determination step, ideally involving chromatography and MS. Chromatography will continue to play a crucial role in the determination of mycotoxins unless a radically different approach to separate complex mixtures is developed. With the advent of small particles in relation to UHPLC, smaller amounts of samples can be processed faster than ever. To quantify the wealth of potential mycotoxins and other potentially toxic substances in our food and feed chain in highly complex matrices, separation remains as important as ever. The great increases in sensitivity and selectivity of LC--MS instruments have made a significant contribution in qualitative and quantitative determination of mycotoxins in cereal-based food and other commodities. However, matrix effects, isobaric interference, and maintaining confidence in the assignment of identity are still the major limitations for LC--MS methods used for the quantification and identification of food contaminants, including mycotoxins, in complex matrices. The increasing use of HRMS instruments has reduced the problems associated with low selectivity and errors in identification because of the capability of accurate mass measurement. However, how to fully eliminate matrix effects has not been fully technically solved yet. Despite these remaining challenges, dilute and shoot approaches before LC--MS/MS, which do not require any cleanup, are increasingly being used for the quantification of several hundred mycotoxins and other secondary metabolites of fungi and plants in food and feed matrices. Moreover, for all regulated toxins, matrix effects and especially signal suppression can be compensated for through the use of fully ^13^C-labeled internal standards, which have become commercially available. Ensuring comparability of measurement results is another challenge especially for mycotoxin--commodity combinations for which no certified reference materials exist. In the past few years, mycotoxin analysis has been moving from the targeted analysis of individual mycotoxins to untargeted metabolite profiling and metabolomics of (ideally) all secondary metabolites that are involved in plant--fungi interactions. The major method used in this area is based on in vivo stable isotope ^13^C labeling and subsequent measurement of biological samples by full-scan high-resolution LC--MS. We anticipate SIL will become a major technique to study the fate of mycotoxins during food processing. SIL can also be used to detect deviations of secondary metabolites of fungi, plants, and bacteria from normal patterns, flagging suspicious food and cereal samples for further analysis and confirmation, and for more accurate quantification and identification of compounds. Electronic supplementary material ================================= {#Sec13} ESM 1(PDF 4.86 kb) ESM 2(XLSX 12.8 kb) Published in the topical collection celebrating *ABCs 16th Anniversary.* **Electronic supplementary material** The online version of this article (10.1007/s00216-017-0750-7) contains supplementary material, which is available to authorized users. Open access funding provided by University of Natural Resources and Life Sciences Vienna (BOKU). This work received funding from the European Union's Horizon 2020 research and innovation program under grant agreement no. 692195 (MultiCoop). The authors thank Lukas Vaclavik, Franz Berthiller, and Ondrej Lacina for scientific consultations on the sections on high-resolution mass spectrometry and Christoph Büschl for providing Fig. [4](#Fig4){ref-type="fig"}. Conflict of interest {#FPar4} ==================== The authors declare that they have no competing interests.

Introduction {#Sec1} ============ Attention-deficit/hyperactivity disorder (ADHD) is a disorder of inattention, impulsivity, and hyperactivity that affects 8% to 12% of children worldwide^[@CR1],\ [@CR2]^. It has been suggested that individuals with ADHD may have an elevated risk for obesity, which comes from studies of adults and youth in both clinical and population-based settings^[@CR3]--[@CR6]^. However, when looking at a younger age in children, the comorbidity between ADHD and obesity has not been examined sufficiently, especially with the lack of data based on population. A study based on a small sample of clinical data confirmed that ADHD would increase the risk of obesity in boy ADHD patients^[@CR7]^. Another study based on a national survey with 62887 children and adolescents aged 5 to 17 years suggested that children with ADHD had 1.5 times the odds of being overweight, after adjustment for age, gender, and other demographic information^[@CR8]^. A recent meta-analysis study revealed that adolescents with ADHD only have 1.1 times of odds of being overweight^[@CR9]^. In that national survey, a doctor-diagnosed ADHD was used to determine ADHD status, therefore, the finding is based on ADHD patients as well. Few studies have been done in children based on general population. Furthermore, the potential underlying mechanisms between ADHD, eating disorder, and obesity has not been examined sufficiently^[@CR10]^. A recent meta-analysis study demonstrated that there are significant overlapping neurobehavioral circuits in ADHD, eating disorder, and obesity in pediatric populations^[@CR11]^. Research has mainly examined ADHD as a risk factor for obesity^[@CR12],\ [@CR13]^. A prospective study indicated that inattention-hyperactivity symptoms at 8 years old were associated with indices of obesity at 16 years old, after adjustment for gender and baseline BMI, rather than the opposite^[@CR14]^. With better insight into the potential mechanism of comorbidity, abnormal eating such as Bulimia Nervosa and especially impulsivity-induced eating, is considered to be an important mediator for the association between ADHD and obesity^[@CR15]^. An updated meta-analysis study indicated that people with ADHD are 3.8 times more likely to present with any eating disorder as a comorbid diagnosis, and the risk increased to 5.7 times for Bulimia Nervosa^[@CR16]^. Bulimia Nervosa is characterized by recurrent episodes of binge eating, and recurrent inappropriate compensatory behavior in order to prevent weight gain^[@CR17]^. Community-based studies suggested that Bulimia Nervosa symptoms partially explained the associations between ADHD and obeisty^[@CR18],\ [@CR19]^. In addition, emotional eating has been defined as eating in response to a range of negative emotions such as anxiety, depression, and loneliness to cope with negative affect emotional eating^[@CR20],\ [@CR21]^. Emotional eating is highly correlated with other eating disorders, like Bulimia Nervosa symptoms, since both of them are related to emotion-focused coping, maladaptive coping strategies, and a strong aversion to negative feelings and stimuli^[@CR22]--[@CR24]^. It has been suggested that ADHD symptoms and impulsivity predicted both emotional eating and binge eating in adult women, and such abnormal eating behaviors were, in turn, positively associated with BMI^[@CR25]^. Similar results have been replicated in males^[@CR26]^. In addition, adult patients with Bulimia Nervosa and childhood ADHD were more impulsive than patients with Bulimia Nervosa alone, and these patients also displayed more severely disordered eating patterns^[@CR27]^. ADHD inattention symptoms, depressive symptoms and trait-impulsivity in obese women patients highly predicted their binge eating behaviors^[@CR28]^. Considering the sample of children and adolescents, ADHD comorbidity of Bulimia Nervosa symptoms has been discussed in few studies, but the findings are confounding. A study suggested that childhood ADHD diagnosis predicted eating pathology in female adolescents^[@CR29]^. In another study, both boys and girls with ADHD displayed more Bulimia Nervosa symptoms than controls^[@CR30]^. Conversely, an inconsistent outcome indicated that no significant differences in rates of Bulimia Nervosa symptoms were identified in children with ADHD when compared to the sex-matched control group^[@CR31]^. This might be explained by the fact that full Bulimia Nervosa syndrome is not that common in the pediatric group. It was said that children/adolescents with oppositional defiant disorder (ODD) symptoms showed increased eating in response to external cues, even though binge eating and ADHD symptoms were not associated with disordered eating behaviors in overweight and obese children/adolescents^[@CR32]^. However, to our knowledge, no study has examined the relationship between ADHD and emotional eating in children. To date, the evidence regarding the relationships among ADHD, abnormal eating, and overweight is extremely limited in a sample of children. To our best knowledge, only 1 recent study showed that adolescents diagnosed with ADHD showed more binge eating problems, and after adjusting for binge eating, the relationship between ADHD and BMI z-scores was attenuated^[@CR33]^. The study was carried out based on community mental health clinics, and the effect of binge eating was excluded from the model, so little is known about the mediating effect of abnormal eating on the association between ADHD and obesity. Therefore, the first purpose of this study is to clarify the relationships among ADHD, abnormal eating (including Bulimia Nervosa symptoms and emotional eating), and BMI in children at the population level. Except for abnormal eating, ADHD is associated with a broader range of psychiatric comorbidity, e.g., depression is a common internalizing disorder^[@CR34],\ [@CR35]^. A review of studies in community samples has reported that the rate of depression in youth with ADHD is 5.5 times higher than in youths without ADHD, with rates ranging from 12% to 50%^[@CR36]^. Children/adolescents with symptoms of depression showed emotional and binge eating^[@CR37]^. Those with ADHD and comorbid depression had significantly increased risk for substance abuse and suicide compared to children with ADHD or depression alone^[@CR35],\ [@CR37]^. Obese women with both of ADHD and eating disorders have higher rates of substance abuse; moreover, inattention symptoms combined with depression predict binge eating^[@CR28]^. However, little is known about the role of depression between ADHD and eating problems, even though emotional eating and Bulimia Nervosa symptoms are highly correlated with mood. A preliminary study had explored the relationship among them, but the effect of depression was often excluded from the analysis models. For instance, in a clinical sample of severely obese adolescents aged 12 to 17 years, ADHD symptoms were significantly associated with Bulimia Nervosa symptoms after controlling for depressive symptoms^[@CR38]^. Thus, the third purpose of this study is to identify the mediating effect of depression between ADHD and abnormal eating in children. Results {#Sec2} ======= This study mixed a sample with different socioeconomic strata from China. Table [1](#Tab1){ref-type="table"} showed the statistical difference between the selected three schools in parents' education level, age and household income. Generally to say, parents in School A have highest socioeconomic level, e.g. high education level, high household income and young parents. On the contrary, the parents in School C have lowest socioeconomic level in the current samples. As displayed in Table [2](#Tab2){ref-type="table"}, boys (mean = 15.3) showed more ADHD symptoms than girls (mean = 11.6; F = 36.33, *P* \< 0.0001). The behaviors of emotional undereating in boys (mean = 9.2) was also slightly higher than girls (mean = 8.6; F = 4.54, *P* \< 0.05). There were no gender differences in Bulimia Nervosa symptoms, emotional overeating behaviors, and depression. The results indicated that 98 children (12.9%) were considered obese and 150 children (19.7%) were considered overweight in the sample. There were more boys than girls in both the overweight and obesity classifications (χ^2^ = 13.449, *P* \< 0.01). Children in obesity/overweight group have a slightly higher ADHD score (mean = 14.3) than children in normal weight group (mean = 13.0; t = −1.85, *P* = 0.064). Children in obesity/overweight group have more Bulimia Nervosa symptoms (mean = 1.3) than normal weight group (mean = 1.1), but the difference is not statistically significant. Similarly, there are no significant differences in emotional overeating/undereating and depression symptoms within two groups.Table 1The variance in the demographic information of participated families of three elementary schools.School A n (%)School B n (%)School C n (%)χ^2^ ~MH~Father's education level Illiteracy/Elementary school6(1.8)9(3.4)13(7.2)102.2\*\*\*\* Middle school98(30.1)82(31.4)106(59.2) Senior school71(21.8)87(33.3)47(26.3) Undergraduate college135(41.5)78(29.9)13(7.3) Graduated college15(4.6)5(1.9)0(0)Mother's education level Illiteracy/Elementary school17(5.3)21(7.9)53(29.8)146.6\*\*\*\*\* Middle school94(29.4)86(32.5)93(52.3) Senior school80(25.0)78(29.4)25(14.0) Undergraduate college127(39.7)77(29.1)7(3.9) Graduated college2(0.6)3(1.1)0(0)Annual household income (US\$) ≤30009(2.9)16(6.3)25(15.1)71.3\*\*\*\* 3000--600030(9.7)34(13.4)22(13.2) 6000--900052(16.9)51(20.2)39(23.5) 9000--1200038(12.3)29(11.5)38(22.9) 12000--1500063(20.5)38(15.0)26(15.7) ≥15000116(37.7)85(33.6)16(9.6)Father's age \<305(1.6)2(0.7)8(4.5)46.4\*\*\*\* 30--40256(80.5)154(58.8)112(62.9) 40--5054(16.9)97(37.0)54(30.3) 50--603(1.0)9(3.5)4(2.2)Mother's age \<306(1.9)6(2.3)13(7.4)43.5\*\*\*\* 30--40288(90.6)210(79.5)120(67.8) 40--5023(7.2)46(17.4)42(23.7) 50--601(0.3)2(0.8)2(1.1)Single child Yes239(73.5)160(59.3)55(30.4) No86(26.5)110(40.7)126(69.6) ADHD symptoms1.0Yes247(90.8)294(89.4)162(88.0) No25(9.2)35(10.6)22(12.0) Total329(41.9)272(34.6)184(23.4)\*\*\*\**p* \< 0.0001. Table 2The partial correlation among variables controlling for gender.TotalBoysGirlsFmean (SD)mean (SD)mean (SD)ADHD^a^13.4 (8.9)15.3 (9.5)11.6 (7.9)36.33\*\*\*\*Bulimia Nervosa symptoms^b^1.2 (2.4)1.4 (2.5)1.2 (2.2)2.34Emotional overeating^c^6.1 (2.4)6.2 (2.5)6.0 (2.3)1.34Emotional undereating^c^8.9 (3.7)9.2 (3.6)8.6 (3.7)4.54\*Depression^d^3.1 (3.3)3.1 (3.4)3.0 (3.1)0.34BMI**n (%)n (%)n (%)χ** ^**2**^Obesity98 (12.9)63 (8.3)35 (4.6)Overweight150 (19.7)90 (11.8)60 (7.9)13.449\*\*Normal or thinness513 (56.4)245 (32.2)268 (35.2)^a^ADHD Rating Scale-IV; ^b^Children's Eating Attitude Test; ^c^Child Eating Behaviour Questionnaire; ^d^Child Behavior Checklist. \* *P* \< 0.05, \*\* *P* \< 0.01, \*\*\*\**P* \< 0.0001. The results in Table [3](#Tab3){ref-type="table"} indicated the correlations between ADHD, abnormal eating, and depression after adjustment for gender. We found that ADHD was positively correlated with Bulimia Nervosa symptoms (r = 0.18916, *P* \< 0.0001), emotional overeating (r = 0.31435, *P* \< 0.0001), emotional undereating (r = 0.27953, *P* \< 0.0001), and depression (r = 0.48574, *P* \< 0.0001). Bulimia Nervosa symptoms was positively correlated with emotional overeating as well (r = 0.15883, *P* \< 0.0001). Emotional undereating was positively correlated with depression (r = 0.31332, *P* \< 0.0001). Emotional overeating is highly correlated with emotional undereating as well (r = 0.47550, *P* \< 0.0001). No significant correlation was found between BMI and ADHD or abnormal eating.Table 3The partial correlation among variables controlling for gender.ADHDBulimia Nervosa symptomsEmotional overeatingEmotional undereatingDepressionBulimia Nervosa symptoms0.18916\*\*\*\*Emotional overeating0.31435\*\*\*\*0.15883\*\*\*\*Emotional undereating0.27953\*\*\*\*0.064900.47550\*\*\*\*Depression0.48574\*\*\*\*0.061220.27988\*\*\*\*0.31332\*\*\*\*BMI0.006670.04496−0.00340−0.03110−0.04129\*\*\*\**P* \< 0.0001. In Fig. [1](#Fig1){ref-type="fig"}, SEM model 1 explored the effect pathway among ADHD, emotional eating and BMI. We found that ADHD positively contributed to emotional eating (*β* = 0.41, *P* \< 0.001) and Bulimia Nervosa symptoms (*β* = 0.20, *P* \< 0.001) in children. However, neither emotional eating (*β* = 0.01, *P* \> 0.05) nor Bulimia Nervosa symptoms (*β* = 0.05, *P* \> 0.05) was related to BMI. The proposed Model 1 represents an excellent fit to the data (GFI = 0.994). The details are provided in Table [4](#Tab4){ref-type="table"}. The loading factors for the 2 variables of emotional eating were 0.73 to 0.66.Figure 1SEM model 1 testing correlations among ADHD symptoms, emotional eating, bulimia and BMI in children. \*\|t\| \< 1.96, \*\*\|t\| \< 2.58, \*\*\*\|t\| \< 3.28. Table 4Comparing model 1 with model 2 in model fitness indices.Chi-SquareP value of Chi-SquareGoodness of Fit Index (GFI)Adjusted GFI (AGFI)RMSEA EstimateRMSEA Lower 90% Confidence LimitRMSEA Upper 90% Confidence LimitModel 110.3997\<0.050.9940.97030.05980.02290.1013Model 270.6114\<0.00010.9630.88960.11770.09370.1434 In Fig. [2](#Fig2){ref-type="fig"}, compared with SEM model 1, the mediating effect of depression was added between ADHD and abnormal eating in SEM model 2. We found that depression would directly predict emotional eating (*β* = 0.43, *P* \< 0.001), rather than bulimia (*β* = 0.06, *P* \> 0.05). ADHD significantly contributed to depression (*β* = 0.48, *P* \< 0.001). It suggested that ADHD not only contributed to abnormal eating directly, but also had an effect on emotional eating through depression when two symptoms were comorbid. Similar to model 1, neither emotional eating (*β* = 0.03, *P* \> 0.05) nor Bulimia Nervosa symptoms (*β* = 0.06, *P* \> 0.05) was related to BMI. SEM model 2 also demonstrated a good fit to the data (GFI = 0.963). The loading factors for the 2 variables of emotional eating were 0.73 to 0.65.Figure 2SEM model 2 testing correlations among ADHD symptoms, depression, emotional eating, bulimia and BMI in children. \*\|t\| \< 1.96, \*\*\|t\| \< 2.58, \*\*\*\|t\| \< 3.28. Discussion {#Sec3} ========== Even though an eating disorder has been demonstrated to be an important mediating factor between ADHD and obesity in adults, the evidence is very limited in children^[@CR15]^. The current study indicated that ADHD symptoms in children contributed to their Bulimia Nervosa symptoms. Impulsivity might be the cause of ADHD comorbidity with Bulimia Nervosa symptoms, since impulsivity is considered to be a central symptom for both disorders^[@CR30]^. Impulsive symptoms such as interrupting, intruding, and difficulty waiting one's turn are part of the diagnostic criteria for the Combined Type of ADHD, which has been conceptualized as a problem of inhibition and impulse control. Impulsive personality traits are thought to be central to Bulimia Nervosa, and up to 40% of individuals with Bulimia Nervosa show multi-impulsivity symptoms in other life domains beyond eating, such as risky sexual behavior, stealing, and drug abuse^[@CR39]^. Even though a study with a small sample of college students indicate that impulsivity hasn't mediated the relationship between ADHD and binge eating disorder^[@CR40]^, impulsivity has been shown to predict the onset of Bulimia Nervosa symptoms 9 months later, among adolescent girls in a community sample^[@CR41]^. Therefore, impulsivity and lack of inhibition in children with ADHD may play a role in triggering Bulimia Nervosa^[@CR42]^. Additionally, some genic feature might be the cause of the relationship. It had been suggested that the dopaminergic genes, especially the dopamine receptor D4 (DRD4), have been associated independently with both binge eating and ADHD^[@CR43]^. Reward Deficiency Syndrome caused by abnormalities of dopaminergic genes could be a transdiagnostic feature between ADHD, eating disorder, and obesity^[@CR44],\ [@CR45]^. In addition, a unique finding in the present study was that ADHD can predict emotional eating behaviors in children, including both emotional overeating and undereating. Some people eat less in the face of strong emotions, but some may turn to impulsive or binge eating to struggle with emotional distress. To our best knowledge, only 1 study examined the relationship between ADHD and emotional overeating, but no specific interpretation was discussed. A nationwide twin study suggests that children with ADHD show significant higher prevalence of restrictive eating behaviors than their counterparts, especially for girls^[@CR46]^. A speculated mechanism is that the stress and negative emotion faced by children with ADHD increased the risk of emotional eating behaviors, because individuals engage in emotional eating only when they are experiencing negative emotions^[@CR47]^. Emotional eating is associated with a range of emotional problems, such as low self-esteem, social anxiety, and feelings of inadequacy^[@CR48]^. Children with ADHD often suffer from low social recognition, poor school achievement, and high accompaniment with mood disorders, which may contribute to emotional eating^[@CR49]^. Meanwhile, those children probably perceive subjectively greater stress that may result in anxiolytic quality of eating^[@CR50]^. Given emotional eating is closely related with emotional problems, ADHD is highly comorbid with emotional disorders, especially depression^[@CR51]^. Our present study clarified the mediating effect of depression between ADHD and eating behaviors. As expected, we found that depression mediated the relationship between ADHD and emotional eating in children, rather than ADHD and bulimia. Depression has long been documented as a significant comorbid condition of eating disorders among female adolescents^[@CR52]^. It has been suggested that the presence of depressive symptoms and ADHD inattention symptoms can predict the present of binge eating in obese women^[@CR28]^. Depressive symptoms would predict eating disorders in 13 to 18-year-old girls over a 4-year period, but not the reverse^[@CR53]^. One finding in our study indicated that children with ADHD who are comorbid with depression would have an increased risk of emotional eating. To summarize shortly, the complex relationships between ADHD and eating disorder not just rely on single factors (e.g., impulsivity, genetic abnormalities, depression), but the interactions among these factors might interpret the associations. It has been suggested that not only ADHD neuropsychological functioning could alter information processing for food-related decision making, but also changes in nutritional or emotional state promoted by eating disorder could make ADHD individuals more prone to disinhibition^[@CR16]^. Taking BMI into consideration, although the relationships between ADHD and eating disorders have been confirmed in this study, no significant relationship among ADHD, eating disorder and BMI was established in children. This finding is consistent with a previous study in which the diagnosis of ADHD was not associated with any of the obesity trajectories (no obesity, childhood obesity, adolescent obesity, and chronic obesity) in the general pediatric population (age ranged from 9 to 16 years)^[@CR54]^. To date, only a few studies based on ADHD patients demonstrated that ADHD could increase the risk of obesity in children^[@CR7],\ [@CR8]^ and none of these were carried out at the population level. The close relationship between ADHD and eating disorders in children potentially increase the risk of obesity when they are grown. Obesity has been found to be a long-term consequence of binge eating disorder in epidemiological studies^[@CR55]^. Another interpretation is that the median age for the onset of an eating disorder is between 12 and 13 years of age^[@CR56]^; this means children who are comorbid with ADHD and eating disorders might take a long time to become obese. Limitations {#Sec4} =========== The current study is a cross-section design, so a longitudinal study is needed to explore the causality among ADHD, abnormal eating, and BMI, even though the method of SEM is suggested to be a good method to explore the causal relationship for cross-sectional data. ADHD was assessed with the parent-reported questionnaire, rather than an interview by a professional. In addition, except for abnormal eating and depression, multi-factors related to both ADHD and BMI should be taken into consideration to construct a SEM model, such as sleep problems and screen time. Method {#Sec5} ====== Participants {#Sec6} ------------ This study was approved by the medical ethics committee of Fudan University (Shanghai, China). The corresponding author confirms that this study was performed in accordance with the approved social experiment guidelines and regulations. Informed consent was obtained from legal parents of all participating students. A total of 785 primary students, aged 9 to 13 years (Mean = 10.6, SD = 1.1), and their parents were recruited by stratified random sampling (from third grade to fifth grade) of 3 primary schools in Shanghai, China. Among the sample, 409 (52.1%) were boys (mean age = 10.6, SD = 1.1) and 376 (47.9%) were girls (mean age = 10.6, SD = 1.1). Measures {#Sec7} ======== ADHD {#Sec8} ---- Both the student self-reported questionnaire and parent-filled questionnaire were used in this study. The 2 questionnaires were matched by student identification (ID). The specific assessment instruments covered in the 2 questionnaires were reported as follows: The parent-report version of the ADHD Rating Scale-IV (ADHDRS-IV) was used to assess ADHD symptoms^[@CR57]^. The ADHDRS-IV is an 18-item ADHD assessment scale, and it consisted of 2 subscales, inattention and hyperactivity-impulsivity, each containing 9 items. Each item was mapped onto 1 of the 18 Diagnostic and Statistical Manual of Mental Disorders, 4^th^ Edition (DSM-IV) symptoms of ADHD. Parents were required to rate the frequency of each of the ADHD symptoms, which occurred over the previous 6 months, on a 5-point Likert scale that ranged from 0 to 4 (never \[0\], rarely \[1\], sometimes \[2\], often \[3\], and very often \[4\]). The sum of all the scores on the18 items resulted in a total score that ranged from 0 to 54. The reliability and validity of the home version of ADHDRS-IV had been verified in a sample of Chinese children aged 6 to 17 years^[@CR58]^. Cronbach's alpha was 0.92 for the ADHDRS-IV in this sample. ### Abnormal eating {#Sec9} Two separate scales were used to assess abnormal eating behaviors:Emotional eatingThe Child Eating Behaviour Questionnaire (CEBQ) is a parent-report questionnaire designed to assess variation in eating styles among children^[@CR59]^. It has been shown to have a robust factor structure and good internal reliability^[@CR60]^. Two subscales involving emotional overeating and emotional undereating were used in our present study. Parents were asked to rate their child's eating behavior on a 5-point Likert scale that ranged from 1 to 5 (never \[1\], rarely \[2\], sometimes \[3\], often \[4\], and always \[5\]). Sample scale items included, for example, "My child eats less when s/he is angry; my child eats more when s/he is happy; my child eats more when worried; and my child eats more when annoyed." Cronbach's alpha was 0.86 for both subscales of emotional overeating and emotional undereating.Bulimia Nervosa symptoms The Children's Eating Attitude Test (ChEAT), a modified version of the Eating Attitudes Test (EAT) is a self-reported questionnaire filled out by children themselves^[@CR61]^. The psychometric properties and the application of ChEAT in different cultures have been certified in previous studies^[@CR62]--[@CR65]^. The ChEAT contained 26 items which were rated on a 6-point Likert scale that ranged from 0 to 3: always (3), very often (2), often (1), sometimes (0), rarely (0), and never (0). A subscale of Bulimia Nervosa was also used in this study. Sample scale items included, for example, "I have gone on eating binges where I feel that I might not be able to stop" and "I give too much time and thought to food." The Cronbach's alpha for the bulimia nervosa subscale was 0.67. ### Depression {#Sec10} The parent of each participant completed the Chinese version of the Child Behavior Checklist for ages 4 to 18 years (CBCL/4--18). The CBCL is a very common tool because its psychometric properties and Chinese version have been certified in previous studies^[@CR66],\ [@CR67]^. The subscale of depression contained 17 items. Sample scale items included, for example, "Deliberately harms self or attempts suicide; Cries a lot; and Worries." Each item was rated on a 3-point Likert scale that ranged from 0 to 2 (not true \[0\], somewhat or sometimes true \[1\], and very true or often true \[2\]). The Cronbach's alpha for the depression subscale was 0.75. ### Body Mass Index {#Sec11} Body Mass Index (BMI) was calculated from height and weight (weight \[kg\]/height^2^ \[m^2^\]) measured with the participant wearing indoor clothing and standing in stocking feet. The height and weight were measured by a healthcare provider in the school hospital. Specific age and gender BMI cut-off points were used to define overweight and obesity according to the International Obesity Task Force (IOTF) Body Mass Index Cut-Offs^[@CR68]^. Statistical analysis {#Sec12} -------------------- The data was analyzed by Statistical Analysis System (SAS) 9.3 (Institute Inc., Cary, NC, USA). Partial correlation analysis was used to examine the correlation between ADHD, abnormal eating, and depression. The relationship between ADHD symptoms, emotional eating, Bulimia Nervosa symptoms, and BMI was analyzed by using Structural Equation Modeling (SEM). **Publisher\'s note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. The study was funded by the National Natural Science Foundation of China (Grant No. 81402693), Shanghai Pujiang Program (Grant No.14PJC012) and Award from Shanghai Municipal Health Bureau (Grant No. 15GWZK0402). T.L. designed the study, T.L., S.H.J. and L.X.R. collected the data, T.L. analyzed the data and wrote the paper. Competing Interests {#FPar1} =================== The authors declare that they have no competing interests.

Introduction and background {#Sec1} =========================== At any one time, there are one billion people worldwide who are in the second decade of their life \[[@CR1]\] and 1.8 billion in the 10--24 age range \[[@CR2]\]. Broadly speaking, adolescence occurs in and around the second decade and is the transitional developmental period between childhood and adulthood. The term adolescence, derived from the latin meaning for 'growing up', has no universally accepted definition. For example, WHO \[[@CR2]\] describes adolescence as a period in human growth and development that occurs nominally between the ages of 10 and 19. Others define it as incorporating young adulthood, spanning a variable period of time between the ages of 10 and 24 \[[@CR3]\]. There is widespread agreement that adolescence starts at puberty and ends with the uptake of mature social roles, such as employment and child rearing. Due to the later transition into these roles in many developed countries, the longer and later definition may therefore be more appropriate. There are, however, recognised wide variations in the start and end points of adolescence, including differences between cultures and individuals \[[@CR4]\]. Whilst a great deal of focus has been placed on the importance of healthy early childhood development \[[@CR5]--[@CR7]\], the adolescent years are a unique period in their own right. For example, during this period, there is increased exposure to health-influencing behaviours such as alcohol consumption and cigarette smoking. These behaviours, in conjunction with peer group pressure, may contribute to the establishment of patterns that have significant health consequences in later life \[[@CR8]\]. However, whilst there is often an emphasis on risk and detrimental health behaviours during adolescence \[[@CR9]\], these years also provide significant opportunities for behaviour to be shaped in positive ways that may improve longer term health outcomes \[[@CR1]\]. The capacity of adolescents to make decisions and take control of their own future life course should not be underestimated, especially if they are appropriately informed. Intervening to support positive health choices during adolescence therefore has the potential to lessen the likelihood of developing a health condition in later life, with possible positive implications for future health service provision and resources. An understanding of the complex physiological changes that are taking place within the human body during this period and the relationship with health-related behaviours is the first step towards recognising adolescence as a transition period that warrants a specific focus as a unique period of opportunity. Such knowledge can help to inform health policy and intervention development. The aim of this systematic review is to gain an understanding of the relationship between physiological development and health-related behaviours in adolescence. The objectives are to:Identify and describe the range of evidence that explores the relationship between physiological development and health behaviours in adolescenceDetermine the strength of such evidence This review forms the first phase of a wider adolescent and young adult research programme (see Additional file [1](#MOESM1){ref-type="media"}). It is hoped that the findings of the review will help to inform policy, practice, and future research priorities. Methods {#Sec2} ======= Previous associated reviews \[[@CR9], [@CR10]\], which have both focussed on adolescent health in more general terms, indicate that data may be found from a broad range of sources. In order to encompass such a wide variety of literature, this systematic review will be based on the principles of an integrative review. This approach can be used to capture an extensive range of studies with a common focus on the chosen topic of interest \[[@CR11]\]. Integrative reviews are the most comprehensive of all review approaches, facilitating the inclusion of different methodological approaches, and both published or unpublished literature \[[@CR12]\]. Such reviews can combine existing knowledge to produce a more extensive understanding \[[@CR13]\]. Within a systematic integrative review, similar studies can be grouped together (e.g. clinical trials and comparative studies or qualitative and descriptive studies). Quality criteria instruments and analytical methods can then be selected to be relevant to each type. The stages of the review are summarised in Table [1](#Tab1){ref-type="table"}, with key elements of the review being further detailed below.Table 1Five stages of an integrative literature review (summarised from Whittemore and Knafl \[[@CR11]\])Stages of reviewAim/purposeDetails(1) Problem formulationTo clearly state topic of interest and purpose of review• List variables of interest\ • Set focus and boundaries(2) Literature searchTo make explicit and justify search strategy and sampling criteria• Specify databases and other search methods\ • State type of literature to be included (e.g. published/unpublished)\ • Detail key words\ • Acknowledge publication bias(3) Data evaluationTo assess type, scope, diversity, and quality of accessed literature• Specify different types of study found and classify into sub-groups\ • Decide on quality criteria instruments for each type of study(4) Data synthesisTo specify systematic analytical method\ To create an innovative synthesis\ To formulate a unified and integrated conclusion• Data reduction: simplify sub-groups into a manageable framework according to the type (e.g. qualitative, comparative, experimental); create short summaries of each primary source\ • Data display: create charts or visual network displays to show connection within each sub-group type\ • Data comparison: identify patterns, themes, relationships, and major variables within and between sub-groups\ • Conclusion drawing and verification: creative and critical analysis of data, acknowledging commonalities and differences, and including any justifiable generalisations\ • Production of integrative summation(5) PresentationTo capture the depth and breadth of the topic and produce a comprehensive understanding• Summary should contribute to a new understanding\ • Specify implications for practice, research, and policy\ • Note limitations of the review as a whole Variables of interest {#Sec3} --------------------- The purpose of the review is to gain a comprehensive understanding of the complex and potentially multifaceted relationship between physiological development and health-related behaviours in adolescence. The main variables of interest are theories and hypotheses related to physiological development in adolescence and relationships with health-related behaviour. In particular, the review will aim to assess the strength of such evidence. Focus and boundaries {#Sec4} -------------------- To assist with the classification of the review studies, all identified material will be screened for relevance using the broad inclusion criteria of adolescent physiological development and health behaviours. Physiological development will encompass a broad range of biological systems (e.g. musculo-skeletal, nervous, endocrine, integumentary, cardiovascular, respiratory, digestive, reproductive) and associated biochemical and hormonal processes. Health-related behaviours will include areas such as diet and nutrition, physical activity, substance use (including smoking), sexual behaviours, and sleep. Outcome measures will potentially refer to any of these systems or behaviours. Where there is insufficient evidence in the title and abstract to make a decision, full-text papers/reports will be retrieved and narrow screening criteria used on all full text papers. Narrow screening will focus on the study design, population, intervention, and outcome (SPIO) criteria: ### Study design {#FPar1} This review will consider all research designs and types of publication. Randomised controlled trials, non-randomised controlled trials, comparative studies, case control studies, and cohort studies will all be considered for inclusion. Qualitative data will be included, if relevant, as will published reports and grey literature such as unpublished dissertations/theses (where accessible) to reduce the likelihood of publication bias. ### Population {#FPar2} Studies that include adolescents aged nominally between ≥10 and ≤24 years will be included, although no definitive age barriers will be used to avoid excluding potentially relevant research, with the proviso that the topic focus relates to the adolescent life stage. Papers reporting exclusively on younger children and/or adults will be excluded. Studies of non-human subjects will be included if they supplement the knowledge base (e.g. add further information regarding mechanisms). Studies with an exclusive focus on participants with specific clinical health conditions will be excluded, since these groups may have distinct needs in terms of development and health behaviour. These groups include individuals diagnosed with eating disorders, addictions, or major mental health disorders. Studies that focus exclusively on individuals with morbid (or severe) obesity will be excluded, due to the clinical impact on health and subsequent health-related behaviours. ### Interventions {#FPar3} All types of intervention will be included, if relating to adolescent physiological development and including explicit or implicit links to health behaviour. ### Outcomes {#FPar4} Included outcome measures will not be restrictive, with the proviso that there is a relationship between physiological development and health behaviour. Physiological development will include such elements as relate to healthy and normal functioning and the physical and chemical phenomena involved therein \[[@CR14]\]. Health behaviours will relate to behaviours that have the potential to impact on health, such as dietary intake, physical exercise, sexual behaviour, sleep, and substance use (including alcohol, drugs, smoking, and solvents). Papers that solely focus on situations where substance use had evolved into an addiction will be excluded, as will studies involving gambling. We will include studies that examine pre-adolescent biological development if these studies also relate this to adolescent physiology and health-related behaviour. We will also include studies that assess the impact of adolescent physiological development on health-related behaviour in adulthood. Theoretical or discussion papers will be included if they contribute to our understanding of mechanisms or effects. We will exclude protocols that do not contain results. Studies dating from 1980 will be included. Health promotion was formally acknowledged and defined as an important part of healthcare professional's work by WHO in 1986, as part of the Ottawa Charter \[[@CR15]\]. It was therefore largely from this decade onwards that more formal initiatives started to be used and became the focus of research work. However, important earlier theoretical papers which contribute to our understanding will be considered for inclusion. No initial language limitations will be applied. If necessary, we will seek translation of papers that appears to meet the inclusion criteria. Similarly, no geographical boundaries will be set. Details regarding inclusion and exclusion criteria are summarised in Table [2](#Tab2){ref-type="table"}.Table 2Inclusion/exclusion criteriaInclusion criteriaExclusion criteriaStudies/reports involving:Studies/reports involving:• Adolescents (nominal age range 10--24 years^a^)\ and\ • Physiological development, including developmental influencers (e.g. biochemical influences, hormonal activity, enzyme activity, and neurotransmitters), with related outcomes\ and\ • Health behaviour (explicit or implicit), including lifestyle or behavioural factors/outcomes, and decision making• Children (e.g. under 10 years^a^) as sole focus and/or\ • Adults (e.g. \>24 years of age^a^) as sole focus and/or\ • Atypical or pathological physical/psychological/physiological developmental focus (e.g. eating disorders, addictions, gambling, or major mental health disorders) and their treatment interventions\ • Sport performance studies\ • Adolescent pregnancy (as sole focus)\ • Incidence, prevalence, or trend papers (as sole focus)\ • Protocols for studies yet to be undertaken^a^No definitive age barriers will be used to avoid excluding potentially relevant research The project group for this review will monitor all stages of the review process and is comprised of researchers with a background in public health, neuroscience, adolescence, and systematic reviewing, as well as policy/decision makers with a remit to improve health for young people across Scotland. An advisory group, including lay adolescent members, will provide additional guidance as necessary. Literature search {#Sec5} ----------------- ### Search strategy {#Sec6} A comprehensive search strategy that aims to be both sensitive and specific will be used in this review. An initial search strategy will be devised using the MEDLINE thesaurus and indexing system to identify appropriate MeSH headings and key/text words associated with the terms 'adolescence', 'physiology', and 'health-related behaviour'. This will be adapted for use across all included databases as necessary. ### Stages in the literature search {#Sec7} Stages in the literature search are summarised in Fig. [1](#Fig1){ref-type="fig"}. Although not limited to empirical evidence, we will use the Preferred Reporting of Systematic Reviews and Meta-Analysis (PRISMA) flow chart format to visually display the processes and findings of the review.Fig. 1Stages in the literature search process (summarised from Aveyard \[[@CR18]\]) ### Databases {#Sec8} Details of databases to be used in the review are given in Table [3](#Tab3){ref-type="table"}.Table 3Databases to be used in the systematic reviewDatabaseDetailsAMEDAllied and Complimentary Medicines Database, with a focus on complimentary medicine, palliative care, and professions allied to medicineASSIAApplied Social Sciences Index and Abstracts; includes literature from psychology, sociology, medicine, anthropology, politics, and lawCENTRALCochrane Central Register of Controlled Trials; summary details of published and unpublished trialsCINAHLCumulative Index to Nursing and Allied Health Literature. Includes over 2.6 million records dating back to 1981; books and dissertations also includedEMBASEExcerpta Medica Database covers biomedical and pharmacological literatureERICEducation Resources Information Center (ERIC); described as the world's largest database of educational literatureHMICThe Health Management Information Consortium; brings together the bibliographic databases of two UK health and social care management systems: the Dept of Health's library and information services and King's Fund information and library services. Includes grey literatureMEDLINECovers health-related journals worldwide, focusing on evidence/research based work; includes in-process and non-indexed itemsPsycINFOAbstract database providing systematic coverage of psychological literature as far back as the 1800sPubMedIncludes Medline, plus a comprehensive and broad-ranging selection of health-related journals and booksDiscoveryUniversity of Edinburgh accumulated databases and resources ### Grey literature {#Sec9} The term 'grey literature' came into common use in the 1970s \[[@CR16]\] and is generally considered to refer to unpublished research. In this review, Electronic theses On-line Service (EthOS) 1980--2015 and Zetoc conference proceedings (1980--2015) will be searched electronically. Google and Google Scholar will also be searched using key phrases. In addition, key professionals and voluntary/third sector organisations will also be contacted for resources and reports that may not have been identified through routine searches of databases. Reference lists of all eligible research or reports will also be scanned to identify potentially relevant references not retrieved by the database searches. ### Screening {#Sec10} Two review authors will independently screen potential studies for eligibility at both broad and narrow screening stages. Members of the project or advisory teams will act as third reviewers in the case of eligibility uncertainty or discrepancy. ### Bibliographic management {#Sec11} A bibliographic data management system (RefWorks™) will be used to store and manage the results of the electronic database searches. This will also provide an audit trail of identified resources and give indicators of how they are subsequently classified and sub-divided during the broad and narrow screening processes. Data evaluation {#Sec12} --------------- ### Classification of result resources {#Sec13} Data extraction sheets and tables will be developed, tailored to the resources found. These will give an overview summary to assist in the production of groupings and classifications of the types of resources identified by the literature searches. The different types of study or report will be classified into major groups and then divided into sub-groups, as necessary and relevant. ### Critical appraisal {#Sec14} To assess the quality of papers selected for inclusion in the review, the UK Critical Appraisal Skills Programme \[[@CR17]\] provides tools suitable for the evaluation of individual study quality. Quality assessment will address issues such as whether study aims are clearly stated and whether findings and conclusions are valid and/or credible. Two review authors will independently quality appraise included studies and agree the final assessment; a third reviewer will assist in the event of a discrepancy. The critical appraisal of studies will assist with assessing the strength of evidence from individual and grouped studies. Data synthesis {#Sec15} -------------- Once the final number of studies has been identified and appraised, the task of data synthesis will commence. It is anticipated that this will involve analysis within and across groupings, which could potentially include and compare:Population/age groupingsBehaviours and riskDevelopmental stageSocioeconomic factorsGenderIndividual or multiple lifestyle factorsOutcome measuresTheory or hypotheses bases Depending on the findings, the type and scope of further analyses (e.g. meta-analysis, meta-synthesis, meta-summary) will be decided upon, and sub-set analyses carried out as necessary and/or feasible. Relevant statistical software will be used, if appropriate. Charts and other visual displays will be used to show the patterns, connections, and variables within and across the data sources. These will assist in the production of an integrative summation, giving a critical analysis of the results as a whole and providing any conclusions, generalisations, and recommendations from the review. Presentation {#Sec16} ------------ The integrative summation will form the basis of a report on the findings of the review. Presentation and dissemination of results will be through local, national, and international publications and conferences, via a variety of media (e.g. slide presentations, podcasts, written summaries) and for a variety of audiences (e.g. academic/health professionals, lay audiences, including adolescents and pictorial versions for those with reading difficulties). Discussion {#Sec17} ========== Strengths and limitations of the review {#Sec18} --------------------------------------- This review seeks to provide a comprehensive understanding of the complexities surrounding adolescent physiological development and the relationship and links to health behaviour. As such, the review aims are ambitious in their scope and depth. Although the review will use a clear and systematic approach to searching, screening, and reviewing studies, the search strategies and engines may not capture all relevant material. The findings may also be open to selection bias. The presence of the wider project research team and advisory/reference groups will serve to minimise such bias. Relevance of the review {#Sec19} ----------------------- It is anticipated that the results of the review will be of interest to a wide variety of policy makers, organisations, and individuals working with adolescents across health, social care, and education. We will therefore seek to make the findings from the review widely available. The findings will also be used to inform future research strategies, by identifying areas where further evidence would be beneficial and providing the necessary background information to justify research funding allocation and priority. Additional file {#Sec20} =============== Additional file 1:**Adolescent project overview.** This provides details of where this current review is placed within the broader context of adolescent study by the project team. (DOCX 28 kb) **Competing interests** The authors declare that they have no competing interests. **Authors' contributions** JMcA, RJ, EH, NA, and SJB were involved in the conceptualisation of the study; JP, JMcA, KM, and RJ contributed to the protocol construction; and JP, JMcA, and KM carried out the manuscript draft. All authors read and approved the final manuscript. We would like to thank our advisory group members for supporting this project and funders: MRC (MR/KO 023209/1), CSO, and NHS Health Scotland.

We thank Professor Home for his interest ([@B1]) in our publication on a novel basal insulin analog, LY2605541 ([@B2]). Total, LDL, and HDL cholesterol levels along with triglyceride levels (or lipid profiles) were measured as noted in the last paragraph on page 2144 ([@B2]). Whereas the serum triglyceride measurements for the LY2605541-treated patients did not differ from baseline but were higher compared with insulin glargine--treated patients, the total, HDL, and LDL cholesterol did not differ from baseline for the LY2605541-treated patients and did not differ from those treated with insulin glargine. In contrast, in the recently published type 1 diabetes study ([@B3]), LY2605541-treated patients demonstrated a modest increase in triglycerides and LDL cholesterol compared with baseline and to insulin glargine-treated patients, whereas the HDL cholesterol with LY2605541 decreased from baseline and compared with insulin glargine. As these findings were first noted in these exploratory phase 2 studies, the phase 3 studies will measure routine lipids and hepatic transaminase levels at frequent intervals to characterize the chronology of these changes. Additionally, in a subset of patients, these studies will measure hepatic fat content by magnetic resonance imaging, plasma lipoprotein subclass particle concentration (nmol/L), and particle size measured by nuclear magnetic resonance, total cholesterol efflux capacity, free fatty acids, cholesterylester transfer protein activity and mass, adiponectin, and apolipoproteins (Apo-A1, Apo-A2, Apo-B100, and Apo-CIII). Although high-sensitivity C-reactive protein would provide some information regarding inflammation, the determination of hepatic fat content was considered to be a fundamental investigation. The subsequent effects on insulin sensitivity along with determining inflammation were considered to be secondary. Lastly, because these patients were treated with exogenous insulin prestudy, and circulating levels of LY2605541 (similar to insulin detemir) are very different than that of human insulin, we question the validity of the homeostasis model assessment calculations under these circumstances. R.M.B. has served on scientific advisory panels for Abbott Diabetes Care, Amylin Pharmaceuticals, Bayer HealthCare, Eli Lilly and Company, Hygieia, Johnson & Johnson, Roche Pharmaceuticals, Sanofi, and Valeritas; has served as a consultant to Abbott Diabetes Care, Amylin Pharmaceuticals, Bayer HealthCare, Boehringer Ingelheim Pharmaceuticals, Calibra Medical, Eli Lilly and Company, Halozyme Therapeutics, Helmsley Trust, Hygieia, Johnson & Johnson, Medtronic, ResMed, Roche Pharmaceuticals, Sanofi, Takeda Pharmaceutical Company, Valeritas, and Becton, Dickinson and Company; has received research support from Abbott Diabetes Care, Amylin Pharmaceuticals, Bayer HealthCare, Boehringer Ingelheim Pharmaceuticals, Calibra Medical, Dexcom, Eli Lilly and Company, Halozyme Therapeutics, Helmsley Trust, Hygieia, Intarcia Therapeutics, Johnson & Johnson, MannKind Corporation, Medtronic, National Institutes of Health, ResMed, Roche Pharmaceuticals, Sanofi, Takeda Pharmaceutical Company, and Becton, Dickinson and Company; and holds stock in Merck & Co. J.R. has served on scientific advisory panels for Roche Pharmaceuticals, Sanofi, Novo Nordisk, Eli Lilly and Company, MannKind Corporation, GlaxoSmithKline, Takeda Pharmaceutical Company, Daiichi-Sankyo, Johnson & Johnson, Novartis Pharmaceuticals Corporation, Boehringer Ingelheim Pharmaceuticals, Lexicon Pharmaceuticals; has served as a consultant to Roche Pharmaceuticals, Sanofi, Novo Nordisk, Eli Lilly and Company, MannKind Corporation, GlaxoSmithKline, Takeda Pharmaceutical Company, Daiichi-Sankyo, Johnson & Johnson, Novartis Pharmaceuticals Corporation, Boehringer Ingelheim Pharmaceuticals, and Lexicon Pharmaceuticals; and has received research support from Pfizer, Sanofi, Novo Nordisk, Roche Pharmaceuticals, Bristol-Myers Squibb Company, Eli Lilly and Company, Forest Laboratories, GlaxoSmithKline, Takeda Pharmaceutical Company, Novartis Pharmaceuticals Corporation, AstraZeneca LP, Amylin Pharmaceuticals, Johnson & Johnson, Daiichi-Sankyo, MannKind Corporation, Lexicon Pharmaceuticals, and Boehringer Ingelheim Pharmaceuticals. R.F.A. has been a consultant to Novo Nordisk and Sanofi; has received research support from Eli Lilly and Company, Sanofi, Novo Nordisk, Novartis Pharmaceuticals Corporation, AstraZeneca LP, and Reata Pharmaceuticals; and has participated in Speaker's Bureaus for Amylin Pharmaceuticals and Eli Lilly and Company. M.J.P., Y.Q., and S.J.J. are employees of and hold stock in Eli Lilly and Company. D.C.H. is a retired employee of Eli Lilly and Company and holds stock in Eli Lilly and Company. V.P.S. is a past employee and is currently a stock holder of Eli Lilly and Company. No other potential conflicts of interest relevant to this article were reported. [^1]: V.P.S. is currently affiliated with the U.S. Food and Drug Administration, Silver Spring, Maryland. The opinions expressed in this article are those of the authors and do not necessarily reflect the views of the U.S. Food and Drug Adminstration.

Since the first child was born after *In Vitro* Fertilisation (IVF) in 1978, several studies have been conducted on the possible consequences of Assisted Reproductive Technologies (ART), which include standard IVF, Gamete Intra-Fallopian Transfer (GIFT) and Intra-Cytoplasmatic Sperm Injection (ICSI). Most papers on the consequences of ART concentrated on short-term outcomes, such as perinatal mortality, multiple pregnancies, weight at birth and malformations (reviewed in [Schieve *et al*, 2004](#bib30){ref-type="other"} and [Hansen *et al*, 2005](#bib16){ref-type="other"}), while few studies have considered the long-term effects of these techniques. Now that several children born after ART have reached adolescence, it is useful to study the possible long-term consequences of this procedure, such as cancer incidence. Cases of cancer in children born after ART have been reported ([White *et al*, 1990](#bib35){ref-type="other"}; [Toren *et al*, 1995](#bib34){ref-type="other"}; [Rizk *et al*, 2000](#bib28){ref-type="other"}; [Cruysberg *et al*, 2002](#bib9){ref-type="other"}; [Lee *et al*, 2004](#bib18){ref-type="other"}); the hypotheses behind a possible association between ART and cancer could be the repeated hormonal exposure and/or the epigenetic modification of gene expression that may be activated by the manipulation of the gametes in the laboratory; however, the studies on this topic are scarce ([De Rycke *et al*, 2002](#bib10){ref-type="other"}; [Thompson *et al*, 2002](#bib33){ref-type="other"}; [Ayhan *et al*, 2004](#bib1){ref-type="other"}; [Brinton *et al*, 2004](#bib7){ref-type="other"}). We review here the cohort studies that have considered the association between ART and cancer in children, and performed a meta-analysis of the available data. MATERIALS AND METHODS ===================== Published guidelines for meta-analysis of observational studies ([Stroup *et al*, 2000](#bib32){ref-type="other"}) were followed to perform the literature search and the analysis, and to report the results. Studies included in the present analysis had to meet the following inclusion criteria: they had to be cohort studies involving children born after ART (which represents the exposure of interest), and cancer (all types) had to be the end point. A search on Medline and Embase was performed for articles reported up to February 2005, using combinations of the keywords 'IVF\', 'ART\', 'children\', 'cohort\' and 'cancer\', and restricting the search to articles published in English. A broad search yielded more than 2500 potentially relevant titles. The titles and the abstracts of the papers were screened independently by two experts, and 161 articles which contained information on both short- and long-term health outcomes of children born after ART were selected. Citation indices, bibliographies of the articles and review papers ([Brinton *et al*, 2004](#bib7){ref-type="other"}; [Schieve *et al*, 2004](#bib30){ref-type="other"}; [Lightfoot *et al*, 2005](#bib21){ref-type="other"}) were also checked to complete the search. We selected a total of 14 studies that met the inclusion criteria for the meta-analysis. A description of the studies is reported in [Table 1](#tbl1){ref-type="table"}. Out of the 14 studies, five ([Bergh *et al*, 1999](#bib3){ref-type="other"}; [Ericson *et al*, 2002](#bib14){ref-type="other"}; [Pinborg *et al*, 2003](#bib27){ref-type="other"}, [2004a](#bib25){ref-type="other"}, [2004b](#bib26){ref-type="other"}) were partially overlapping, therefore only the two most recent publications, with the larger cohort were included in the meta-analysis ([Pinborg *et al*, 2004b](#bib26){ref-type="other"} for the Danish data set, [Ericson *et al*, 2002](#bib14){ref-type="other"} for the Swedish data set). The meta-analysis was therefore performed on 11 of the 14 data sets ([Table 2](#tbl2){ref-type="table"}), which reported pertinent, nonoverlapping, and comparable data. Three studies have different designs ([White *et al*, 1990](#bib35){ref-type="other"}; [Odone-Filho *et al*, 2002](#bib24){ref-type="other"}; [Moll *et al*, 2003](#bib22){ref-type="other"}), but were included since it was possible to calculate cancer incidence ratios from the available published data; however, a sensitive analysis was conducted by including and excluding these studies. Statistical analysis -------------------- Two studies ([Rufat *et al*, 1994](#bib29){ref-type="other"}; [Pinborg *et al*, 2004b](#bib26){ref-type="other"}) did not provide cancer incidence rates in a reference population. For the French study ([Rufat *et al*, 1994](#bib29){ref-type="other"}), the expected cases were extracted from the literature ([Bernard *et al*, 1993](#bib4){ref-type="other"}; [Gembara *et al*, 1995](#bib15){ref-type="other"}). Since the average period of follow-up of children born after ART was not reported in the original paper by Rufat, we estimated an average follow-up of 2.2 years, which is the mean between the minimum and the maximum period of follow-up. For the Danish data sets ([Pinborg *et al*, 2003](#bib27){ref-type="other"}, [2004a](#bib25){ref-type="other"}, [2004b](#bib26){ref-type="other"}) the expected number of cases was calculated by applying the cancer incidence rate provided by the Danish Cancer Registry for children 0--6 years and for the period 1995--1999. The average period of follow-up used to calculate the number of expected cases was 4.1 years, as reported in a subsequent publication on the same cohort of children ([Lidegaard *et al*, 2005](#bib20){ref-type="other"}). For the meta-analysis, the observed and expected cases from each study were added and the overall Standardised Incidence Ratio (SIR) was calculated as the ratio between the number of observed and the number of expected cases. The details for the calculation of the expected number of cases are reported for each study in [Table 2](#tbl2){ref-type="table"}. The exact confidence interval for SIR was obtained by using the Poisson\'s distribution. The heterogeneity across studies was analysed with the Cochran\'s test. A SIR adjusted for study was then calculated, using either a fixed or a random-effects model, according to the results of the Cochran\'s test ([Normand, 1999](#bib23){ref-type="other"}). The potential for publication bias was examined by drawing a 'funnel plot\' in which study-specific log effect estimates were plotted against their s.e. ([Sterne *et al*, 2000](#bib31){ref-type="other"}). Egger\'s test was performed to assess the symmetry of the funnel plot ([Begg and Mazumdar, 1994](#bib2){ref-type="other"}; [Egger *et al*, 1997](#bib12){ref-type="other"}). A significant asymmetry indicates the presence of bias, which was set in this analysis at a *P*-value \<0.05. The statistical analysis was performed using STATA package, version 8. RESULTS ======= Out of the 11 studies included in this meta-analysis, two were conducted in Australia, six in Europe, one in Israel, one in USA and one in Brasil. Eight were cohort studies, three ([White *et al*, 1990](#bib35){ref-type="other"}; [Odone-Filho *et al*, 2002](#bib24){ref-type="other"}; [Moll *et al*, 2003](#bib22){ref-type="other"}) the last two based on a hypothetical cohort and one based on national statistical data without a systematic follow-up of all the children in the cohort ([White *et al*, 1990](#bib35){ref-type="other"}).The follow-up varied from about 1 to 13 years ([Table 1](#tbl1){ref-type="table"}). None of the cohort studies reported a significant association between ART and childhood cancer. The three studies with a different design reported a significant increase of neuroectodermal cancer ([White *et al*, 1990](#bib35){ref-type="other"}), retinoblastoma ([Moll *et al*, 2003](#bib22){ref-type="other"}) or cancer in general ([Odone-Filho *et al*, 2002](#bib24){ref-type="other"}) in children conceived after IVF. The SIR for each study are presented in [Figure 1](#fig1){ref-type="fig"}. The overall assumption of homogeneity between study-specific SIRs was rejected (*P*-value for Cochran\'s test \<0.001), even when restricted to the eight studies with similar design (*P*-value for Cochran\'s test: 0.003). The lack of homogeneity seemed to be due to two studies ([Lerner-Geva *et al*, 2000](#bib19){ref-type="other"}; [Bradbury and Jick, 2004](#bib6){ref-type="other"}), since the SIR=0 could strongly influence the result even though the studies included a small number of subjects (332 and 176, respectively). When these two studies were excluded from the analysis, the hypothesis of homogeneity could be accepted (*P*-value for Cochran\'s test: 0.76). The final cohort included 38 815 subjects, with 38.21 cases of cancer expected *vs* 47 observed, giving a SIR of 1.23 (95% CI 0.93--1.37). The analysis restricted to eight studies (excluding the studies by White, Odone-Filho and Moll, with different designs) indicates 36.22 expected cases of childhood cancer and 35 observed, giving a SIR of 0.97 (95% CI 0.69--1.10). The study-adjusted SIR was 1.33 (95% CI 0.62--2.85) when all the 11 studies were included, while it was 0.77 (95% CI 0.41--1.42) when the analysis was restricted to eight studies. The overall analysis did not show publication bias (Egger\'s test *P*-value: 0.70), while there was evidence of publication bias for the restricted analysis (eight studies, Egger\'s test *P*-value: 0.02). When the two studies with SIR=0 were excluded, no evidence of publication bias was observed (*P*-value for the Egger\'s test: 0.40). DISCUSSION ========== Studies on childhood cancer in children born after ART have been conducted only recently, and are analysed in this paper. Overall, no increased risk of childhood cancer was found in the present analysis. A previous review on four of the studies included in this meta-analysis ([Lightfoot *et al*, 2005](#bib21){ref-type="other"}) provided a meta-SIR of 1.03 (95% CI 0.61--1.63). Meta-analysis is a useful approach when studying rare diseases, such as childhood cancer, because the pooled data set has greater power than each individual study ([Egger and Smith, 1997](#bib13){ref-type="other"}; [Blettner *et al*, 1999](#bib5){ref-type="other"}). However, pooling data can also have certain limitations, ([Blettner *et al*, 1999](#bib5){ref-type="other"}; [Stroup *et al*, 2000](#bib32){ref-type="other"}), such as publication bias. Although our analysis suggests overall a lack of such bias, the number of studies included was small, and therefore our results are not conclusive. Another issue is that the studies in a meta-analysis may differ considerably in quality, design, methods of data collection, definition of the exposure and type of confounding variables. To promote homogeneity, we included only cohort studies of children followed up for several years. Three studies ([White *et al*, 1990](#bib35){ref-type="other"}; [Odone-Filho *et al*, 2002](#bib24){ref-type="other"}; [Moll *et al*, 2003](#bib22){ref-type="other"}) presented a slightly modified design, so we performed a sensitivity analysis by including and excluding these studies from the metaestimates. Our meta-analysis could not take into account the length of follow-up as a covariate, since not all the studies included specified it, and when they did, it appeared obvious that the follow-up period was different from study to study. In this analysis, we concentrated on ART and did not consider studies of the possible negative consequences of hormones administered to the mothers for infertility problems, some of which have included suggestions of (nonsignificant) increases of childhood cancer. Similarly, studies of congenital malformation in relation to ART have not fallen within the scope of our meta-analysis. In conclusion, this meta-analysis does not suggest an association between ART and childhood cancer, even though the limited number of studies prevent a firm conclusion and a pooled analysis would be useful. This work was partially supported by a Grant from the Ministry of Health (no. 521C/7-I). We thank Jeanette Falck Winther, Hans Storm and Milan Fajber for providing cancer incidence rates from the Danish Cancer Registry. We are also grateful to Raffaella Masucci and Luca Brenga for technical assistance. {#fig1} ###### Cohort studies on *in vitro* fertilization and childhood cancer **Authors (year)** **Country** **Number of exposed children** **Follow**-**up** **Cancer results** -------------------------------------------------------- ------------- ------------------------------------- -------------------------------------- ----------------------------------------------------------------------------------------- [White *et al* (1990)](#bib35){ref-type="other"} Australia 2285^a^ Absent Three neuroectodermal tumours [Rufat *et al* (1994)](#bib29){ref-type="other"} France 1637 Minimum follow-up: 1 year One leukaemia [Doyle *et al* (1998)](#bib11){ref-type="other"} Britain 2507 Average follow-up: 8.6 years Two unspecified cancers [Bergh *et al* (1999)](#bib3){ref-type="other"} Sweden 5586 Maximum follow-up: 13 years One ALL, one reticulosis, one upper extremities, one peripheral nerves cancer [Bruinsma *et al* (2000)](#bib8){ref-type="other"} Australia 5249 Average follow-up: 3 years, 9 months One brain, one connective tissue, three leukaemia, one salivary gland cancer [Lerner-Geva *et al* (2000)](#bib19){ref-type="other"} Israel 332 710 person years 0 cancers [Klip *et al* (2001)](#bib17){ref-type="other"} Netherlands 9479 (429 after hormonal treatment) Average follow-up: 4.6 years Three leukemia, four unspecified cancers [Ericson *et al* (2002)](#bib14){ref-type="other"} Sweden 9056 Maximum follow-up: 13 years Three ALL, two histiocytosis, two sarcomas, two CNS, one retinal, one hepatic carcinoma [Odone-Filho *et al* (2002)](#bib24){ref-type="other"} Brazil  ^b^ Maximum follow-up: 5 years One AML, one neuroblastoma, two rhabdo myosarcoma [Moll *et al* (2003)](#bib22){ref-type="other"} Netherlands  ^c^ Maximum follow-up: 1 year, 2 months Five retinoblastoma [Pinborg *et al* (2003)](#bib27){ref-type="other"} Denmark 1080 (454 twins) Maximum follow-up: 4 years One 1 ALL, one germinal cell tumour [Pinborg *et al* (2004a)](#bib25){ref-type="other"} Denmark 3393 (twins) Minimum follow-up: 1 year 0 cancers [Pinborg *et al* (2004b)](#bib26){ref-type="other"} Denmark 8523 (3393 twins) Minimum follow-up: 1 year ALL, hepatoblastoma, unspecific tumours of thorax, heart, cerebrum [Bradbury and Jick (2004)](#bib6){ref-type="other"} USA 176 Maximum follow-up: 13 years 0 cancers Cohort without a systematic follow-up. Hypothetical cohort, assuming that approximately 2000 children were conceived after IVF during the period 1996--2000. Hypothetical cohort, assuming that 1--1.5% of children were conceived after IVF. ALL=acute lymphoblastic leukaemia; CNS=central nervous system; AML=acute myelocytic leukaemia. ###### Cohort studies included in the meta-analysis **Author (year)** **Number of exposed children** **Calculation of expected cases (EC)** **Observed/expected cases (*N*)** **Standardized incidence ratio (95%CI)** ---------------------------------------------------------------------------------------- -------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ----------------------------------- ------------------------------------------ [White *et al* (1990)](#bib35){ref-type="other"} 2285 No calculation in the original study. EC calculated by applying cancer incidence rates from the original study to the IVF cohort 3/0.13 23.08 (8.38--67.31) [Rufat *et al* (1994)](#bib29){ref-type="other"} 1637 No calculation in the original study. EC calculated by applying published cancer incidence rates ([Bernard *et al*, 1993](#bib4){ref-type="other"}; [Gembara *et al*, 1995](#bib15){ref-type="other"}) to the IVF cohort 1/0.7 1.43 (0.31--3.79) [Doyle *et al* (1998)](#bib11){ref-type="other"} 2507 Application of cumulative national cancer rates, taking into account year of birth and length of follow-up 2/3.5 0.57 (0.07--2.06) [Bruinsma *et al* (2000)](#bib8){ref-type="other"} 5249 Application of the Victorian cancer incidence rates, taking into account age and length of follow-up 6/4.33 1.39 (0.62--3.09) [Lerner-Geva *et al* (2000)](#bib19){ref-type="other"} 332 Application of specific national cancer incidence rates, taking into account age, gender and year of diagnosis 0/1.7 0 (0--2.18)^a^ [Klip *et al* (2001)](#bib17){ref-type="other"} 9050^b^ Application of cancer incidence rates from the Eindhoven and the Netherlands Cancer Registries, taking into account age, gender and calendar period 6/6.78 0.88 (0.41--1.98) [Ericson *et al* (2002)](#bib14){ref-type="other"} 9056 Application of the Swedish Cancer Registry cancer incidence rates, taking into account year of birth, maternal age, parity and length of involuntary childlessness 11/12.5 0.88 (0.5--1.13) [Odone-Filho *et al* (2002)](#bib24){ref-type="other"} Around 2000 Application of annual incidence rate of cancer for children aged 0--4 years 4/1.17 3.42 (1.42--8.76) [Moll *et al* (2003)](#bib22){ref-type="other"} Not known Application of the 1-year age-specific mortality rates from statistics in the Netherlands 5/0.69 7.25 (3.19--17.03) [Pinborg *et al* (2004a](#bib25){ref-type="other"}, [2004b)](#bib26){ref-type="other"} 8523 No calculation in the original study. EC calculated by applying the 0--6 years cancer incidence rates from the Danish Cancer Registry to the IVF cohort 9/6.7 1.34 (0.71--1.78) [Bradbury and Jick (2004)](#bib6){ref-type="other"} 176 No calculation in the original study. EC were calculated by applying retinoblastoma incidence rates from the original study to the IVF cohort 0/0.01 0 (0--7.40)^a^ The upper confidence limit was calculated using 0.5 as observed number of cases. In all, 429 children conceived after hormonal treatment, but not after ART, were excluded.

Introduction ============ The pyran structural motif can be found in numerous bioactive natural products. Possible strategies towards their efficient preparation include Prins cyclizations, intramolecular substitutions, ring closure metathesis, and hetero Diels--Alder reactions \[[@R1]--[@R2]\]. Our group recently reported the synthesis of enantiopure aminopyrans employing as the key step a Lewis acid induced rearrangement of 1,2-oxazines to bicyclic ketones. This strategy allowed a simple access to unusual carbohydrates with C2-branched 4-amino sugar units and related carbohydrate mimetics \[[@R3]--[@R7]\]. Moreover, pursuing a similar strategy, we described the stereocontrolled preparation of 3,6-dihydro-*2H*-pyrans **D** as precursors for the synthesis of enantiopure aminopyrans **E** ([Scheme 1](#C1){ref-type="fig"}). The sequence of nucleophilic addition of lithiated enol ethers **A** to nitrones **B** and Lewis acid promoted cyclization of the resulting 1,3-dioxolanyl-substituted hydroxylamine derivatives **C**, delivered the enantiopure dihydropyrans **D** in a highly stereodivergent fashion \[[@R8]\]. In this report we disclose full details about this new method, including expansion of the reaction scope and further evaluation of the synthetic potential of the easily accessible functionalized dihydropyrans **D**. {#C1} Results and Discussion ====================== The addition of organometallic species to nitrones has been extensively studied in the past. Highly stereoselective reactions of enantiopure nitrones allow rapid access to functionalized intermediates that have been elegantly used for the preparation of a wide range of natural products \[[@R9]\]. It is known that pre-complexation of carbohydrate-derived nitrones by Lewis acids can reverse the stereochemical outcome of these reactions \[[@R10]\] allowing, for example, a stereodivergent access to enantiopure 1,2-oxazines \[[@R11]\]. When lithiated enol ethers **1** were used as the organometallic species in the addition to nitrones **2** \[[@R12]--[@R14]\], either *syn*- or *anti*-configured hydroxylamine derivatives **3** were obtained, depending strongly on the reaction conditions ([Table 1](#T1){ref-type="table"}) \[[@R15]--[@R16]\]. Subsequent treatment under basic conditions with *tert*-butyldimethylsilyl triflate afforded TBS-protected compounds *syn*- and *anti*-**4** in good overall yields \[[@R17]\]. The stereochemical outcome of the reaction is in line with the addition of other nucleophiles to nitrone **2a** \[[@R9]--[@R11]\]. The configurations of compounds **4** could be unambiguously assigned by X-ray crystallographic analysis of an aminopyran derivative derived from *anti*-**4b** (compound **29**, [Scheme 12](#C12){ref-type="fig"}) \[[@R8]\]. The enol ethers used, i.e., ethyl vinyl ether, dihydropyran, and dihydrofuran, all added smoothly to glyceraldehyde-derived nitrones **2a** or **2b** after lithiation with *tert*-butyllithium. Whereas in the reaction with uncomplexed nitrones, 1.5 equivalents of the respective lithiated enol ether were sufficient to obtain the desired *syn*-products with good yields and stereoselectivities, 5 equivalents of the nucleophile were necessary to achieve reasonable yields in the addition to the nitrone/chlorodiethylaluminium complexes. Despite this, diastereomeric ratios in favour of the *anti*-products were generally excellent. ###### Stereodivergent additions of lithiated enol ethers **1** to glyceraldehyde-derived nitrones **2**. ---------------------------------------- ---------------------------------------- --------------- ---------------------------------------- --------------------------------  Lithiated enol ether Nitrone Conditions^a^ Product Yield and diastereomeric ratio   a\)  *syn*: 61%\ *anti*: 9%\ (d.r. = 88:12) **1a** **2a** b\)  *syn*: 4%\ *anti*: 48%\ (d.r. = 93:7)  **2a** a\)  *syn*: 44%\ *anti*: 18%\ (d.r. = 72:28) **1b** **2a** b\)  *syn*: 8%\ *anti*: 56%\ (d.r. = 88:12)  **2a** a\)  *syn*: 58%\ *anti*: 10%\ (d.r. = 86:14) **1c** **2a** b\)  *syn*: 9%\ *anti*: 54%\ (d.r. = 86:14)   a\)  *syn*: 70%\ (d.r. \> 95:5) ---------------------------------------- ---------------------------------------- --------------- ---------------------------------------- -------------------------------- ^a^Condition a): 1. THF, −78 °C, 1 h (1.5 equiv of enol ether); 2. TBSOTf, 2,6-lutidine, 30 min; Condition b): 1. **2** + Et~2~AlCl, THF, 5 min, then addition to **1**, THF, −78 °C, 15 min (5 equiv of enol ether); 2. TBSOTf, 2,6-lutidine, 30 min. Lithiated ethyl vinyl ether was also added to pre-complexed D-arabinose-derived nitrone **2c** ([Scheme 2](#C2){ref-type="fig"}). After a subsequent TBS-protection step, the desired hydroxylamine derivative *anti*-**4e** was obtained in low yield, but with excellent diastereoselectivity. {#C2} Next, the protected hydroxylamine derivatives *syn*-**4** and *anti*-**4** were treated with Lewis acid to provide enantiopure 3,6-dihydro-*2H*-pyrans *cis*- and *trans*-**5** ([Table 2](#T2){ref-type="table"}). The cyclization proceeded in moderate to excellent yields when two equivalents of trimethylsilyl triflate were used as Lewis acid. Catalytic amounts of Lewis acid were not sufficient to obtain the dihydropyrans, even after prolonged reaction times. The reactions with stoichiometric amounts of TMSOTf gave monocyclic, bicyclic, and spirocyclic dihydropyrans depending on the 1,3-dioxolanyl-substituted enol ether employed ([Table 2](#T2){ref-type="table"}). Unfortunately, attempts to generate an eight-membered (or six-membered) ring by cyclization of *anti*-**4e** were unsuccessful, only decomposition was observed. ###### Lewis acid promoted cyclization of 1,3-dioxolanyl-substituted enol ethers *syn*- and *anti*-**4** to 3,6-dihydro-*2H*-pyrans *cis*- and *trans*-**5**. ---------------------------------------- ---------------------------------------- -------  Cyclization precursor Product Yield   79%   84%   74%   82%   85%   55%   90%   -- ---------------------------------------- ---------------------------------------- ------- As a mechanistic rationale, shown below for the transformation of *syn*-**4a** into *cis*-**5a**, we propose that the Lewis acid coordinates to the distal oxygen atom of the dioxolane moiety in **4** leading to a ring opening and formation of a stabilized carbenium ion **6** ([Scheme 3](#C3){ref-type="fig"}). Subsequent attack of the oxocarbenium ion on the enol ether moiety leads to ring closure affording the cationic pyran intermediate **7**. Subsequent proton transfer to the (moderately basic) hydroxylamine nitrogen re-establishes the enol ether moiety. During aqueous work-up the OTMS-group is apparently removed due to the fairly acidic conditions to produce *cis*-**5a**. Overall, the described cyclization can be classified as an aldol-type reaction of an enol ether to an acetal, or as a Prins-type reaction. {#C3} Encouraged by these results, we decided to try a related Friedel--Crafts-type cyclization. Thus, 2-lithiofuran was added to nitrone **2a** \[[@R18]\] and then protected with TBSOTf to give hydroxylamine derivative *syn*-**4f** in 55% ([Scheme 4](#C4){ref-type="fig"}). Remarkably, subsequent treatment with 2 equiv of TMSOTf did not lead to the desired bicyclic product. Since no conversion was observed, the lower reactivity of *syn*-**4f**, caused by the considerably weaker nucleophilicity of the furan compared to an enol ether \[[@R19]\], might in the future be overcome with more forcing reaction conditions. {#C4} The enol ether moiety in dihydropyrans **5** represents a very versatile handle for further synthetic modifications ([Scheme 5](#C5){ref-type="fig"} and [Scheme 6](#C6){ref-type="fig"}). Starting from *cis*-**5a**, acidic hydrolysis with concurrent loss of the TBS-group by treatment with saturated methanolic HCl followed by the addition of a saturated solution of NaHCO~3~ in water gave ketone **8** in good yield ([Scheme 5](#C5){ref-type="fig"} and [Scheme 6](#C6){ref-type="fig"}). Dihydroxylation using the K~2~OsO~4~/*N*-methylmorpholinoxide-system afforded α-hydroxyketone **9** via in situ hydrolysis of an initially formed hemiacetal. Bromination with NBS in MeCN/H~2~O led to α-bromoketone **10** and its desilylated derivative **11** as single diastereomers. We presume that the NBS attacks the enol ether moiety, as is also the case with the dihydroxylation reagent OsO~4~, from the sterically less hindered side, leading to incorporation of the bromo substituent *trans* to the hydroxymethyl and the bulky hydroxylamine moiety. The partial removal of the TBS-group is undoubtedly caused by HBr formed in the course of the reaction. {#C5} When these reaction conditions were applied to the diastereomeric dihydropyran *trans*-**5a**, different behaviour was noted ([Scheme 6](#C6){ref-type="fig"}). A smooth hydrolysis of the enol ether moiety leading to ketone **12** was neither feasible with methanolic HCl nor with aqueous HCl. After dihydroxylation, a partial hydrolysis of the hemiacetal initially formed in the dihydroxylation was observed leading to an inseparable mixture of compounds **13** and **14**. The TBS-protection group was completely removed during bromination of the enol ether moiety with the formation of compound **15** in good yield \[[@R20]\]. {#C6} The internal enol ether moiety of the bicyclic compound *trans*-**5b** could also be converted into the corresponding α-bromoketone with the bromination reagent NBS to give the ring opened product **16** ([Scheme 7](#C7){ref-type="fig"}). The two possible diastereomers were formed in a ratio of 92:8, as indicated by ^1^H NMR spectroscopy and have not been separated. {#C7} Bromination of the cyclohexylidene-substituted dihydropyran *cis*-**5d** afforded α-bromo ketone **17** which, upon treatment with NaBH~4~, produced epoxypyran **18** ([Scheme 8](#C8){ref-type="fig"}). Epoxide **18** was obtained as a single diastereomer and should be a promising substrate for subsequent nucleophilic addition reactions leading to another series of enantiopure aminopyrans. {#C8} An additional option for the use of the enol ether moiety is oxidative cleavage \[[@R21]\]. Thus, benzyl protection of compound *cis*-**5b** followed by treatment of the resulting **19** with RuCl~3~ and NaIO~4~ gave the ten-membered ring lactone **20** by selective cleavage of the internal double bond ([Scheme 9](#C9){ref-type="fig"}). {#C9} With the objective of preparing a diaminopyran, we attempted to substitute the bromide atom in compound **15** with the azido group. Surprisingly, instead of the desired α-azido ketone, the formation of nitrone **21** was observed, which apparently involves an unusual internal redox reaction ([Scheme 10](#C10){ref-type="fig"}). Sodium azide presumably acts as base to initiate the reaction by deprotonation of the hydroxylamine moiety. A hydride shift from the benzylic position to the 3-position of the pyran ring produces the nitrone moiety of **21** and simultaneously displaces the axially positioned bromo substituent by an intramolecular substitution. This mechanistic hypothesis was supported by the fact that heating **15** in DMF in the absence of NaN~3~ did not lead to conversion of the substrate. {#C10} When the diastereomeric α-bromoketone **11** was reacted with sodium azide, the identical nitrone **21** was obtained in moderate yield ([Scheme 11](#C11){ref-type="fig"}). We presume that an initially formed nitrone **22** probably undergoes base induced epimerization to generate the thermodynamically more stable *trans*-substituted nitrone **21**. As shown below, the nitrone moiety of **21** can be regarded as a masked hydroxylamine, however, it should also be very useful for further diversification of our pyran derivatives, e.g., by 1,3-dipolar cycloaddition leading to isoxazole derivatives \[[@R22]\] or, by addition of nucleophiles to this electrophilic unit \[[@R9]\]. {#C11} Compounds **8** and **21** proved to be excellent precursors for the stereoselective preparation of enantiopure aminopyrans with high structural similarities to carbohydrates. Reduction of the carbonyl group of **8** by NaBH~4~ followed by hydrogenolysis of the *N*,*O*- and *N*-benzyl-bonds furnished aminopyran **24** ([Scheme 12](#C12){ref-type="fig"}). Similarly, nitrone **21** was smoothly converted to the diastereomeric aminopyran **26** by analogous reductive transformations. {#C12} α-Hydroxyketones **9** and **13** were used as precursors for the preparation of compounds with an additional hydroxyl group, such as aminopyrans **28** and **30** ([Scheme 13](#C13){ref-type="fig"}). Reduction of **9** with NaBH~4~ in the presence of CeCl~3~ and subsequent hydrogenolysis furnished compound **28** in excellent yield. The presence of CeCl~3~ during the reduction of the carbonyl group proved to be essential to achieve excellent diastereoselectivity. The mixture of compounds **13** and **14**, which was formed during dihydroxylation of dihydropyran *cis*-**5a** ([Scheme 6](#C6){ref-type="fig"}), was also treated with NaBH~4~ to afford the diastereomeric compound **29** \[[@R23]\]. A final hydrogenation step gave aminopyran **30** in high overall yield. Aminopyrans **24**, **26**, **28** and **30** can be regarded as mimetics of differently configured (2-deoxy) 4-amino sugars \[[@R24]--[@R26]\]. {#C13} Aminopyrans **24** and **28** have already been linked via amide bonds to the thiol shell of gold nanoparticles (particle diameter 6 nm). After sulfation of the hydroxyl groups the multivalent conjugates obtained displayed strong binding to P- and L-selectins, thus demonstrating that compounds such as **24** and **28** are of interest for the development of new anti-inflammatory agents \[[@R27]--[@R28]\]. Inspired by these first findings, we decided to prepare also smaller conjugates for evaluation as selectin ligands. Thus, the free hydroxyl groups of aminopyrans **24** and **28** were temporarily protected as TMS ethers in order to achieve smooth reaction with 1,3,5-benzenetricarboxylic acid chloride ([Scheme 14](#C14){ref-type="fig"}) \[[@R29]\]. After successful conversion to the triamide, the TMS-groups were removed by treatment with trifluoroacetic acid to give the desired compounds **31** and **32**. The low yield of **32** was due to difficulties encountered in the purification of the compound which was carried out by crystallization. {#C14} Conclusion ========== We have presented an approach to novel 3,6-dihydro-2*H*-pyrans that can be transformed into a series of new carbohydrate mimetics and into other enantiopure heterocycles. The stereodivergent addition of lithiated enol ethers to carbohydrate-derived nitrones produced 1,3-dioxolanyl-substituted hydroxylamine derivatives as suitable substrates for Lewis-acid induced cyclizations furnishing mono-, di-, or spirocyclic dihydropyrans. Subsequent reactions such as bromination, acidic hydrolysis or dihydroxylation provided differently substituted and configured pyranone derivatives as precursors for the stereoselective synthesis of aminopyran derivatives that can be regarded as carbohydrate mimetics. Trimeric versions of these carbohydrate mimetics were constructed via their attachment to a tricarboxylic acid core to produce novel polyhydroxylated compounds that will be evaluated as potential selectin binders at a future date. Experimental ============ Addition of lithiated enol ethers to glyceraldehyde-derived nitrones without precomplexation of the nitrone ----------------------------------------------------------------------------------------------------------- The respective enol ether (1.5 equiv) was dissolved in THF (2 mL/mmol of enol ether) and cooled to −78 °C. *t*BuLi (1.6 M in pentane, 1.5 equiv) was added and the reaction mixture stirred for 1 h during which time it was allowed to warm to 0 °C. After further stirring for 1 h at this temperature, it was cooled once more to −78 °C. A solution of the respective nitrone (1 equiv) in THF (0.7 mL/mmol nitrone) was added dropwise over a 15 min period. The mixture was stirred at this temperature for 1 h and the reaction quenched by the addition of H~2~O. After the mixture reached room temperature it was extracted three times with Et~2~O. The combined organic phases were dried (Na~2~SO~4~) and the solvent was removed in vacuo. The crude product (1 equiv) was dissolved in CH~2~Cl~2~ (2.5 mL/mmol), and 2,6-lutidine (2 equiv) and TBSOTf (1.5 equiv) were added slowly at 0 °C. The mixture was stirred at room temperature for 30 min and then the reaction was quenched by the addition of a sat. NH~4~Cl solution. The layers were separated and the aqueous phase was extracted three times with CH~2~Cl~2~. The combined organic phases were dried (Na~2~SO~4~) and the solvent was removed in vacuo. Purification by column chromatography (silica gel, hexane/EtOAc = 20:1) yielded the products as colourless oils. Addition of lithiated enol ethers to glyceraldehyde-derived nitrones precomplexated by Et~2~AlCl ------------------------------------------------------------------------------------------------ The respective enol ether (5 equiv) was dissolved in THF (2 mL/mmol of enol ether) and cooled to −78 °C. *t*BuLi (1.6 M in pentane, 5 equiv) was added and the reaction mixture stirred for 1 h during which time it was allowed to warm to 0 °C. After further stirring for 3 h at this temperature, it was cooled once more to −78 °C. In a separate flask, a solution of nitrone **2** (1 equiv) in THF (2.5 mL) was treated with Et~2~AlCl (1 M in hexane, 1 equiv) for 5 min. The prepared solution was added dropwise to the solution of the lithiated enol ether over a 15 min period. The mixture was stirred at this temperature for a further 15 min and the reaction quenched by addition of 2 M NaOH solution. After the mixture reached room temperature it was extracted three times with Et~2~O. The combined organic phases were dried (Na~2~SO~4~) and the solvent was removed in vacuo. The crude product (1 equiv) was dissolved in CH~2~Cl~2~ (2.5 mL/mmol), and 2,6-lutidine (2 equiv) and TBSOTf (1.5 equiv) were added slowly at 0 °C. The mixture was stirred at room temperature for 30 min and then the reaction was quenched by the addition of a sat. NH~4~Cl solution. The layers were separated and the aqueous phase was extracted three times with CH~2~Cl~2~. The combined organic phases were dried (Na~2~SO~4~) and the solvent was removed in vacuo. Purification by column chromatography (silica gel, hexane/EtOAc = 20:1) yielded the products as colourless oils. Lewis acid-induced cyclization of 1,3-dioxolanyl-substituted enol ethers ------------------------------------------------------------------------ To a solution of the respective 1,3-dioxolanyl-substituted enol ether (1 equiv) in CH~2~Cl~2~ (6 mL/mmol) at −30 °C, was added TMSOTf (2 equiv) and the resulting solution stirred until it slowly reached room temperature (6 h). The reaction was quenched by the addition of water. After separation of the layers, the aqueous phase was extracted three times with CH~2~Cl~2~. The combined organic phases were dried (Na~2~SO~4~) and the solvent was removed in vacuo. Purification by column chromatography (silica gel, hexane/EtOAc = 6:1) yielded the products as colourless oils. Supporting Information ====================== ###### Experimental procedures, characterization data, ^1^H NMR and ^13^C NMR spectra of synthesized compounds. Generous support of this work by the Deutsche Forschungsgemeinschaft (SFB 765), the Fonds der Chemischen Industrie (PhD fellowship to FP) and the Bayer-Schering Pharma AG is most gratefully acknowledged. We thank O. Reimann for his experimental contributions and Dr. R. Zimmer for help during the preparation of the manuscript.

Introduction {#Sec1} ============ Microalgae have been proposed as a commercially important crop in support of a variety of products, ranging from biofuels to pharmaceuticals and feed-stocks for aquaculture (Greenwell et al. [@CR23]; Milledge [@CR36]; Borowitzka [@CR4]). As with all crops, there is potential for production to be decreased or spoilt by the activity or presence of pests. In closed bioreactors, under laboratory conditions, there is the opportunity to exclude or control pests by using clean techniques, but in open pond systems and in large commercial bioreactor operations, there is enhanced scope for entry of pests. For microalgal crops, there are three potential pest types: contamination by other microalgae (Smith et al. [@CR50]), infections caused by viruses which can destroy algal growth very rapidly (as noted sometimes in nature; Schroeder et al. [@CR48]) and fungal attack such as those by chytrid fungi (Gutman et al. [@CR25]; Strittmatter et al. [@CR55]) and the presence of predators (Day et al. [@CR10]). Predators may be present at very low levels in stock cultures and, depending on the growth conditions, may not normally be apparent at all (especially likely with protistan pests). In other instances, and notably in open ponds, predators of various types (protists, rotifers and crustacea such as *Daphnia* and copepods), may be introduced from the wider environment. To date, there is no single established method successfully used to maximise microalgal production with simultaneous minimisation of crop loss through zooplanktonic predation. Cultivators of microalgae have resorted to a wide variety of strategies to control contamination. Most approaches rely on the culturing of extremophiles under highly selective growth conditions (Borowitzka [@CR3]) with regards to pH or, in the case of marine species, salinity, with the latter having the potential double benefit of stimulating productivity while suppressing increases in invader populations (Bartley et al. [@CR2]). Other methods can include filtration and the use of chemical pesticides (Bacellar Mendes and Vermelho [@CR1]; Wang et al. [@CR58]; McBride et al. [@CR34]), although the former can only work when predators are relatively large, while imprudent use of pesticides can destroy the microalgae along with the predator (Méndez and Uribe [@CR35]). Pulsed electric fields, intended to cause structural and functional damage to predators while leaving the microalgal cells intact, have also been suggested (Rego et al. [@CR46]). Other workers have proposed a more "top-down" approach to the problem by turning the hunter into the hunted with the introduction of zooplanktivorous fish into the system (Smith et al. [@CR51]). The rationale behind such suggestions stems from the belief that monoculture states are naturally unstable; so, it is better to manage the inevitable increase in diversity by the creation of a "synthetic community" (Kazamia et al. [@CR30]; Smith and Crews [@CR49]). It has also been suggested that such a top-down bio-manipulation of trophic cascades may pay dividends through increased lipid production (Sturm et al. [@CR56]). However, to preserve biochemical consistency within the crop at the point of harvest (which is usually a commercial imperative), uni-algal cultivation will most likely remain the favoured approach except perhaps for the formulation of aquaculture feeds. In view of the difficulty in applying effective predation mitigation strategies in an industrial setting, it is unsurprising that progress in this area remains slow (Chisti [@CR8]); this was the motivator for the current work. The growth rate of microalgae, and the form of their biomass in biochemical terms (most basically, as indicated by their C/N/P elemental stoichiometry), is of paramount importance for commercial viability, crop production and also for the growth of grazing pests. Traditionally, a biomass C/N/P stoichiometry in accordance with the work of Redfield ([@CR45]), termed the Redfield ratio, is deemed to be optimal for microalgal growth and health (Geider and LaRoche [@CR21]). Often, light limitation developing through self-shading affects the scope for nutrient limitation within dense microalgal populations, and in consequence, the so-called optimal N/P nutrient supply ratio does not simply, nor necessarily at all, drive balanced growth (Flynn [@CR14]). There is a broadly linear relationship between cellular N/C and N-limiting growth rate and a strongly curvilinear relationship for cellular P/C under P-limiting growth (Elrifi and Turpin [@CR11]; Flynn [@CR13]); in consequence, some level of P-limitation can be incurred by microalgae without a significant impact on growth rate, nor significantly affect biochemical quality (Mayers et al. [@CR33]). Research on the relationship between microalgal C/N/P and growth of its natural predator, the zooplankton, indicates that prey C/N/P aligning closely with Redfield ratios best supports predator growth (Sterner and Elser [@CR52]). Furthermore, this interaction is self-reinforcing; a predator feeding on poor quality algae (i.e. low N/C and/or low P/C) releases (regenerates) less nutrients to be re-assimilated by the remaining algae, and hence, the nutrient status of the microalgal population as prey for the grazer can deteriorate further (Mitra and Flynn [@CR39]). For commercial exploitation, the relationship between the microalgal C/N/P and its value as a crop is most obviously divisible on whether the crop is intended for use as feedstocks that are either protein-rich (high N/C) or alternatively C-rich (high C/N, having accumulated extra C as carbohydrate and/or fatty acids). The options for optimising the rate of production for bulk biomass either for high-value chemicals, such as carotenoids and phycobilins (Borowitzka [@CR4]), or for PUFAs and biofuel feedstocks (Fon Sing et al. [@CR19]) are thus largely mutually exclusive. As a high microalgal C/N is associated with low growth rates and hence low biomass productivity, maximising production of C-rich products requires careful selection of microalgal physiology (Flynn et al. [@CR17], [@CR18]) and also careful management of bioreactors with respect to their design, lighting and nutrient loading and harvesting or bioreactor dilution rates (Kenny and Flynn [@CR31]). For commercial growth of microalgae, the organisms are grown to high densities (with dry weight biomass concentration typically between 0.1 and 0.5 kg m^−3^; Chisti [@CR6]; Ozkan et al. [@CR43]) and supplied with high nutrient loads. Under these conditions, light limitation of growth is likely (Richmond [@CR47]) so that growth reactors contain dense suspensions of potentially high-quality prey for zooplankton. Further, in contrast to the situation in nature, prey availability is not limiting for zooplankton growth. If the reactor is operated in a continuous dilution (chemostat-like) mode, then conditions may further conspire to favour the growth of the zooplankton over the microalgae, as biomass-specific grazing rates typically exceed biomass-specific growth rates of the crop (Hansen et al. [@CR26]). However, if the crop is grown for a high C-content (high C/N), then there is scope to minimise losses through the formation of a crop that is intrinsically of low nutrient quality for predators (Sterner and Elser [@CR52]; Mitra and Flynn [@CR38]). This work uses computational stoichiometric ecology to identify approaches (forgoing genetic modification, GM, of the crop) for the control of predation in commercial microalgal cultivation systems through manipulation of factors such as nutrient regimes and culture harvesting. Methods {#Sec2} ======= System configuration {#Sec3} -------------------- The investigation described here exploited the use of dynamic variable stoichiometric models (i.e. for algae, variable C/N/P/Chl). This is in recognition of the importance of simulating variable stoichiometry not only for describing biomass and carbohydrate/lipid production by microalgae (Kenny and Flynn [@CR31]) but also for predator-prey interactions (Sterner and Elser [@CR52]; Mitra and Flynn [@CR38]). A schematic of the model is given in Fig. [1](#Fig1){ref-type="fig"}; a full description, with equations and examples of prior usage to establish the models providence, is given in the Supplementary material ([Appendix_A\_Model_Info.xls](#MOESM1){ref-type="media"}).Fig. 1Schematic of the model. Items within *dark blue boxes* are state variables defined in the model. Nutrients for algal consumption include dissolved inorganic C (DIC), ammonium (DINa) and nitrate (DINn) and phosphate (DIP). The increase in microalgal C-biomass (TC) is a function of the nutrient status (N/C and P/C) of the cells, the status of the photosystems (ChlC, which is itself a function of nutrient status and photoacclimation) and light availability. Light reaching the microalgae depends on surface irradiance, light absorbance by water and the pigmented microalgae and operation depth (OD; pond depth or diameter of a bioreactor tube). A proportion of algal biomass (TC) is C-rich storage products (carbohydrate + lipid; CexC). Zooplankton predator C-biomass (ZC) increases through grazing, with part of the ingestion biomass being voided as organic C/N/P, and part regenerated as nutrients for re-assimilation by the microalgae. The efficiency of grazing (conversion of TC to ZC) depends on algal N/C and P/C, according to stoichiometric rules. Harvesting contributes to harvested biomass (hC), including harvested material for biofuels (hexC). Indicated by *diamonds* are the key parameters explored in the simulations: *Dil* dilution rate and harvesting frequency, *Inoc* inoculation of the system with zooplankton, *Nut* nutrient concentration, *OD* operational depth, *Time* day and hour, *μ* ~*max*~ maximum growth rate of the algae and of the zooplankton predator. Light available for microalgal photosynthesis is a function of that surface irradiance over the day-night cycle, absorbance by the algal suspension with reference to OD, TC and ChlC (the latter a function of NC and of light availability via photoacclimation). The value of hexC depends on TC and CexC, which in turn relates to Dil and Nut, such that growth rate is optimised while N/C is low (noting that N/C is linearly related to N-limited growth potential) The microalgal sub-model was similar to that used in our studies on optimisation of microalgal production (Flynn et al. [@CR18]; Kenny and Flynn [@CR31], [@CR32]). Thus, microalgal growth was described using a mechanistic, acclimative, variable stoichiometric model of microalgal physiology. Changes in microalgal C/N/P/Chl occur with acclimation in response to changes in nutrient and light availability. With nutrient exhaustion, especially of nitrogen (N), excess carbon (C) is accumulated (i.e. cellular C/N increases; N/C declines). Whether that excess C accumulates in reality as carbohydrate and/or fatty acids depends on the taxonomic characteristics of the organism. This particular microalgal model has been used to describe the growth dynamics of many different species under different situations ([Appendix_A\_Model_Info.xls online](#MOESM1){ref-type="media"}). A demonstration of the model operating against a published data set for a real reactor system (Quinn et al. [@CR44]) is shown in Kenny and Flynn ([@CR32]). To this original microalgal-bioreactor model, we added a model describing the growth dynamics of zooplankton (Mitra [@CR37]). That zooplankton model has been previously configured to simulate the growth of both microzooplankton (protists) and mesozooplankton (crustacea) and has been used under various scenarios (Mitra and Flynn [@CR39]; Mitra et al. [@CR40], [@CR41] and references therein; see [Appendix_A\_Model_Info.xls](#MOESM1){ref-type="media"}). The physiological descriptions of the microalgae and zooplankton were as used in Flynn et al. ([@CR18]). The maximum growth rate of the microalgae enabled a division per day (i.e. μ = 0.693 day^−1^) under the 12:12-h light/dark cycle employed. This growth rate is consistent with the enzyme characteristics and cellular activity of RuBisCO (the enzyme that fixes CO~2~), which may enable a maximum growth rate in continuous illumination approaching two divisions per day (Flynn and Raven [@CR15]). The zooplankton were simulated with either a growth rate equivalent to a doubling per day, or two doublings per day. The maximum zooplankton assimilation efficiency was set at 0.75, but this decreased with deterioration in prey quality with reference to the prey and predator N/C and P/C ratios, using which ever was the lower of preyNC/predNC or preyPC/predPC (Mitra [@CR37]). The result of this linkage is an increasingly poor trophic transfer from the prey to the predator as the prey quality declines (i.e. as microalgal N/C and/or P/C falls); this description accords with empirical evidence (e.g. Sterner et al. [@CR53]; Flynn et al. [@CR16]; Young et al. [@CR59]). The whole model, describing microalgae and zooplankton growth within a description of the physico-chemical environment of a bioreactor or pond, was run with an integration step size of 11.25 min. The growth environment included descriptions of the supply of inorganic dissolved N and P nutrients (and their recycling with any zooplankton activity), of light (changing in its availability to the microalgae as a function of the self-shading activity of the plankton biomass, as well as over a 12:12-h light/dark cycle), of bioreactor depth (assuming the typical homogenous mixing of the contents) and of harvesting rates (see below). The default primary limiting nutrient concentration was 440 μM inorganic N (which is half the concentration of the classic f/2 marine algal growth medium; Guillard [@CR24]). We have previously shown that this nutrient provision allows potential nutrient exhaustion under non-light limiting conditions over a range of the shallower optical paths ("depths") typically used in photobioreactors (Kenny and Flynn [@CR31]). In addition to this, P was supplied at either the Redfield ratio with N (mole ratio for N/P of 16) or at twice that ratio (N/P of 32) to enable a potentially moderate level of co-P stress in the microalgae. Growth at a nutrient N/P of 32 appears not to be deleterious to phytoplankton physiology or fatty acid production (Mayers et al. [@CR33]), consistent with the distinct curvi-linear (quota) relationship between cellular P/C and growth rate (Flynn [@CR13]). Harvesting techniques {#Sec4} --------------------- We explored different crop harvesting approaches. Harvesting can be continuous (akin to a chemostat operation, i.e. described by a dilution rate) or discontinuous (as in a batch growth system) with some fraction of the culture harvested every few days. However, in reality, in preparation for developing systems for either approach, a culture must first be grown up in what amounts to a batch growth system. Accordingly, here, growth was simulated from a low inoculum into a low-dilution continuous dilution system; this gave a period of nutrient-replete exponential growth (batch-culture like) prior to steady-state (chemostat-like) growth but without risk of an abrupt nutrient stress than may damage the crop. Introducing the zooplankton {#Sec5} --------------------------- For the simulations presented here, the introduction of the zooplankton predator was considered to occur either concurrent with the microalgae at the start of the growth cycle (i.e. as a contamination in the original inoculum), or after establishment of steady-state conditions (here contamination occurred at day 20 of the simulation, into a well-developed algal system). In both instances, it was assumed that the initial zooplankton biomass was only 0.05% of that of the microalgae. If one considers a predator with a size (volume) of at least 10 times that of its prey (as would typically be the case---Hansen et al. [@CR26]), then numerically, this level of zooplankton contamination would equate to only 0.005% (1 in 20,000). Such contamination levels would likely go unnoticed in routine microscope sample examination of cell counts of the microalgae (Day et al. [@CR10]). Presentation of results {#Sec6} ----------------------- Results are given in terms of the rates of areal biomass production (AP; gC m^−2^ day^−1^) and areal production of C-rich components (fatty acids and/or carbohydrate; AXP; gC m^−2^ day^−1^), as well as volumetric production (VP; gC m^−3^ day^−1^). Areal productivity takes into account VP and bioreactor depth (m). Table [1](#Tab1){ref-type="table"} gives a summary of the simulated test conditions and their associated figures. Here, we highlight specific examples of simulations; additional results (as indicated in Table [1](#Tab1){ref-type="table"}) are provided in the [Supplementary Material Appendix B](#MOESM2){ref-type="media"} and referenced in the format Fig. S*x*. It should be noted that maximum (standing stock) biomass levels reflect a balance of growth rate set against dilution rate. Emphasis here is placed on production rate (the commercial imperative) and not upon terminal biomass concentration.Table 1Summary of simulation conditions and plot locationsNutrient mole N/PDilution rate (day^−1^)Areal biomass: algal and zooplanktonVolumetric biomass: algal and zooplanktonAlgal N/C and N/PAlgal AP and AXPZooplankton growth and ingestion ratesZoo-plankton AE and GGE160.1[Fig. S1](#MOESM2){ref-type="media"}320.1[Fig. S2](#MOESM2){ref-type="media"}160.3Fig. [2](#Fig2){ref-type="fig"}Fig. S5Fig. [4](#Fig4){ref-type="fig"}[Fig. S7](#MOESM2){ref-type="media"}[Fig. S9](#MOESM2){ref-type="media"}[Fig. S11](#MOESM2){ref-type="media"}320.3Fig. [3](#Fig3){ref-type="fig"}Fig. S6Fig. [5](#Fig5){ref-type="fig"}[Fig. S8](#MOESM2){ref-type="media"}[Fig. S10](#MOESM2){ref-type="media"}[Fig. S12](#MOESM2){ref-type="media"}160.5[Fig. S3](#MOESM2){ref-type="media"}320.5[Fig. S4](#MOESM2){ref-type="media"}Plots given in the Supplementary material ([Appendix B online](#MOESM2){ref-type="media"}) are indicated by Fig. S*x*. In all instances, the simulations were performed over a range of reactor depths from 0.025 to 0.5 m*AP* areal biomass production, *AXP* areal production of C-rich components (fatty acids and/or carbohydrate), *AE* assimilation efficiency (proportion of material ingested by zooplankton that is not voided), *GGE* zooplankton gross growth efficiency (ratio of growth rate to ingestion rate) Results {#Sec7} ======= Control systems, with no pest introduction {#Sec8} ------------------------------------------ Figures [2a](#Fig2){ref-type="fig"} and [3a](#Fig3){ref-type="fig"} show microalgal growth in systems of different depths (0.025 to 0.5 m), with different N/P nutrient supply ratios and different dilution rates (panels (a) in [Figs. S1--S4](#MOESM2){ref-type="media"}). There are no substantial differences in the microalgal growth rates between systems supplied with nutrients at the higher N/P (N/P = 32) versus the lower, "optimal", N/P of 16. Peak areal biomass (i.e. as gC m^−2^) decreases with increasing dilution rate (Figs. [2a](#Fig2){ref-type="fig"} and [3a](#Fig3){ref-type="fig"} versus [Figs. S1--S4](#MOESM2){ref-type="media"}). The volumetric biomass density of microalgae exceeded 100 g C m^−3^ in shallow systems operating at a dilution rate of 0.3 day^−1^ ([Figs. S5a and S6a](#MOESM2){ref-type="media"}).Fig. 2Areal biomass of algae and of the zooplankton contaminant when grown at 6 different reactor operation depths (OD; 0.025--0.5 m), with the supply nutrient mole ratio N/P at 16 and a dilution rate of 0.3 day^−1^. **a** No contamination. **b** Contamination at 0 day. **c** Contamination at 20 days. **d** Contamination at 0 day with fast growing zooplankton. **e** Contamination at 20 days with fast growing zooplankton Fig. 3As Fig. [2](#Fig2){ref-type="fig"} but with a supply nutrient mole ratio N/P of 32. **a** No contamination. **b** Contamination at 0 day. **c** Contamination at 20 days. **d** Contamination at 0 day with fast growing zooplankton. **e** Contamination at 20 days with fast growing zooplankton Microalgal N/C was high (i.e. indicative of N-replete and/or light-limiting conditions) in all reactor systems except those with depths \<0.1 m (Figs. [4](#Fig4){ref-type="fig"} and [5](#Fig5){ref-type="fig"}). Accordingly, significant areal production of high-C metabolites, such as could be used for lipid nutrition or biofuels, only occurs in these shallow systems (AXP; [Fig. S7a](#MOESM2){ref-type="media"}) where significant nutrient limitation can develop. Nutrient supply N/P does not greatly affect cellular N/C (Figs. [4](#Fig4){ref-type="fig"} and [5](#Fig5){ref-type="fig"}) nor AXP ([Figs. S7 and S8](#MOESM2){ref-type="media"}). Microalgal P/C in the N/P = 16 systems remains high at depths \>0.2 m (Fig. [4](#Fig4){ref-type="fig"}), but in N/P = 32 systems (Fig. [5](#Fig5){ref-type="fig"}), only the deepest system (depth 0.5 m) enables an elevated microalgal P/C with shallower systems enabling the development of P-stress.Fig. 4Algal mass ratios of N/C and P/C when grown at 6 different reactor operational depths (OD; 0.025--0.5 m), with the supply nutrient mole ratio N/P at 16 and a dilution rate of 0.3 day^−1^. Cf. Fig. [2](#Fig2){ref-type="fig"} for biomass and legend to line types. **a** No contamination. **b** Contamination at 0 day. **c** Contamination at 20 days. **d** Contamination at 0 day with fast growing zooplankton. **e** Contamination at 20 days with fast growing zooplankton Fig. 5As Fig. [4](#Fig4){ref-type="fig"}, but with a supply nutrient mole N/P of 32. Cf. Fig. [3](#Fig3){ref-type="fig"} for biomass and legend to line types. **a** No contamination. **b** Contamination at 0 day. **c** Contamination at 20 days. **d** Contamination at 0 day with fast growing zooplankton. **e** Contamination at 20 days with fast growing zooplankton Systems contaminated with zooplanktonic pests {#Sec9} --------------------------------------------- In the simulations, we considered two scenarios, one with contamination by zooplankton at the point of inoculation, and the other into an established crop. When contaminated with zooplankton with a maximum growth rate similar to that of the microalgae at the start of the system operation (Figs. [2b](#Fig2){ref-type="fig"} and [3b](#Fig3){ref-type="fig"}; [Figs. S1--S4](#MOESM2){ref-type="media"}), there was no effective grazing loss in the shallow systems (0.025 and 0.05 m deep) when supplied with a low nutrient N/P. This is because these systems contained microalgae with a C/N/P stoichiometry that makes them a poor food source for zooplankton growth. In the high N/P systems (Figs. [3](#Fig3){ref-type="fig"}, [S2 and S4](#MOESM2){ref-type="media"}), which produce microalgae with low P/C (Fig. [5](#Fig5){ref-type="fig"}), microalgal losses in the 0.1-m-deep system were also slower, with low grazing pressure ([Figs. S9 and S10](#MOESM2){ref-type="media"}) and poor zooplankton assimilation and growth efficiencies ([Figs. S11 and S12](#MOESM2){ref-type="media"}). Contamination of established systems is less likely to lead to crop losses than an initial contamination event (panels c versus panels(b in Figs. [2](#Fig2){ref-type="fig"} and [3](#Fig3){ref-type="fig"}; [Figs. S1--S4](#MOESM2){ref-type="media"}), provided that the system is shallow enough for nutrient limitation of microalgal growth to develop, and/or where the resultant zooplankton growth rates are slower than the system dilution rate (panels c in [Figs. S3 and S4](#MOESM2){ref-type="media"}). The likelihood of some level of grazing resistance is enhanced in shallow depths by the P-depletion that developed in high N/P systems, which resulted in a depressed algal P/C (Fig. [3](#Fig3){ref-type="fig"} [, S1, S2, S4 and S6](#MOESM2){ref-type="media"}); because of the resultant poor food quality, zooplankton in such systems have poor assimilation and growth efficiencies ([Figs. S11 and S12](#MOESM2){ref-type="media"}). Contamination with fast growing zooplankton leads to a more rapid demise of the microalgal crop; this is catastrophic when it occurs at the start of the culture process, when neither light nor nutrients are limiting and hence good food quality is assured (Figs. [2d](#Fig2){ref-type="fig"} and [3d](#Fig3){ref-type="fig"}; [Figs. S1--S4](#MOESM2){ref-type="media"}). However, when contamination occurs into established shallow systems with a moderate system dilution rate (Figs. [2e](#Fig2){ref-type="fig"} and [3e](#Fig3){ref-type="fig"}; [Figs. S1--S4](#MOESM2){ref-type="media"}), then microalgal losses are low. This is especially obvious with a high N/P medium (Fig. [3](#Fig3){ref-type="fig"}), where the poor P/C of the microalgae (i.e. poor quality food) restricts zooplankton growth to rates similar to, or below, the dilution rates of the systems ([Figs. S9 and S10](#MOESM2){ref-type="media"}). In our simulations, a high microalgal N/C ratio is maintained for depths \>0.05 m and a high microalgal P/C for depths \>0.2 m (Figs. [4](#Fig4){ref-type="fig"} and [5](#Fig5){ref-type="fig"}), regardless of the timing of pest entry. These systems are more likely to support effective feeding by zooplankton, with higher assimilation and growth efficiencies ([Figs. S11 and S12](#MOESM2){ref-type="media"}), and thence higher pest growth rates ([Figs. S9 and S10](#MOESM2){ref-type="media"}). Discussion {#Sec10} ========== Control systems, with no pest introduction {#Sec11} ------------------------------------------ Our simulated microalgal biomass values are comparable to peak concentrations measured in commercial open ponds (Ozkan et al. [@CR43]) assuming a conversion of C-biomass to dry weight, C/dw, between 0.3 and 0.5 (Heymans [@CR28]; Geider and LaRoche [@CR21]). Areal productivity levels, peaking at 2 gC m^−2^ day^−1^ and equating to 7.3 tC ha^−1^ year^−1^, fall in the upper half of rates of the rate of 10--30 t dry weight ha^−1^ year^−1^ reported for real systems (Garcia-Gonzalez et al. [@CR20]; Jiménez et al. [@CR29]; Crowe et al. [@CR9]), again assuming a value for C/dw between 0.3 and 0.5. The simulations of microalgal growth thus give results consistent with expectations (see also Kenny and Flynn [@CR32]). Systems contaminated with zooplanktonic pests {#Sec12} --------------------------------------------- The outcome of zooplankton-microalgal predator-prey interactions is dictated by the balance of growth and loss rates of both parties. Growth of the microalgal prey requires adequate nutrients and light, but as the formation of algal biomass increases at higher nutrient supply concentrations so does the likelihood of self-shading which leads to light limitation of photosynthesis. Loss of the microalgal crop is intrinsically related to the growth of the zooplanktonic predator which in turn depends on the nutritional quality and quantity of the available prey. In nature, both quantity and quality of prey affect zooplankton grazing. However, in artificial systems where microalgae are grown for biomass production, prey quality is the overriding issue. It is important to note that prey quality in this context need not equate to commercial crop quality; a microalgal crop grown for high lipid content will have a high C/N and thus constitute poor food for a predator (Sterner and Elser [@CR52]; Mitra and Flynn [@CR38]). In theory, genetic modification (GM) approaches could be used to configure microalgae that are unpalatable to zooplanktonic pest. Indeed, an analysis on the optimal configuration of GM-microalgae for biofuels and lipid production describes an organism that is coincidentally very poor feed for zooplankton (Flynn et al. [@CR18]). While outwardly a win-win situation, the (inevitable) escape of such a GM organism from large-scale open ponds into the wild would carry a very real risk of generating a harmful algal species *par excellence*, being able to grow rapidly using little nutrient and be ungrazable (Flynn et al. [@CR18]). For this reason, here we consider exploiting stoichiometric ecology to control pest growth. The greatest risk from zooplankton pests is at the initial phases of crop growth, when the microalgal biomass density is relatively low (though far above levels likely to limit zooplankton grazing rates---Hansen et al. [@CR26]) and hence neither light nor nutrients are limiting for algal growth. Microalgae in such situations are typically of good feed value for predators. Contamination of established systems is less likely to lead to crop losses if the cultivation system is shallow enough and the nutrient loading low enough to enable nutrient limitation of microalgal growth to develop. Minimising risks is improved further if a high N/P nutrient regime is operated allowing some level of P-depletion to develop, depressing algal P/C (Figs. [3](#Fig3){ref-type="fig"}, [S1, S2, S4 and S6](#MOESM2){ref-type="media"}). High dilution systems populated by microalgae with growth rates that are higher relative than that of their predators are also less susceptible to zooplankton attack (Moheimani and Borowitzka [@CR42]). However, contamination with fast growing zooplankton lessens the likelihood of crop survival whenever that contamination should occur. To optimise production of biofuels or lipids, shallow culture systems populated with microalgae grown into nutrient limitation at a low fraction of their maximum possible growth rate (μ~max~) are required (Kenny and Flynn [@CR31]). The higher the microalgal μ~max~, the better, as the system can also be operated at a higher absolute dilution rate and this then also washes out any zooplankton. Hence, to minimise crop losses through predation, a balance needs to be struck with bioreactor dilution rates. A slow diluted, light limiting system will favour predators. Conversely, a high dilution rate system configured for high production rates of a C-rich crop appears of lesser susceptibility to predators because the crop is of poor value as food and the elevated dilution rate then exceeds the zooplankton growth rate, washing out the pest. To illustrate this, compare the high peaks of slow-growing zooplankton contamination in [Fig. S1b and S1c](#MOESM2){ref-type="media"} (where dilution = 0.1 day^−1^) to the fast dilution (0.5 day^−1^) situation in [Fig. S3b and S3c](#MOESM2){ref-type="media"}; in the latter, zooplankton biomass only starts to rise very late in the simulation period. Manipulating the systems to commercial advantage {#Sec13} ------------------------------------------------ Setting the best culture conditions to attain optimal chemical composition and elemental C/N/P for commercial growth of microalgae is challenging. For instance, any naturally lit system is susceptible to diel and weather-induced changes in irradiance levels that pose a serious risk to production. This is not solely due to a decrease in growth rate, but because during periods of lowered irradiance, there is an attendant improvement in food quality resulting from changes in microalgal C/N/P. Hence, decreased illumination over a significant period of a day (due to cloud cover, for example) may be expected to enhance the likelihood of grazer control due to a combination of a decreased growth rate of the phototroph, coupled with a concomitant improvement in prey nutritional quality as microalgal N/C and P/C increases. Thus, in studies by Sterner et al. ([@CR54]) and Urabe et al. ([@CR57]), the balance of light and nutrient limitation affected whether the stoichiometric quality of the microalgae remained low under high light, suppressing zooplankton growth, or was improved under low light conditions, where the zooplankton quickly dominated the microalga, *Scenedesmus*. If prey are of good nutritional quality (i.e. microalgal N/C and P/C broadly similar to that of their predator), then grazing and assimilation efficiency are expected to be optimal (Sterner and Elser [@CR52]). Through a positive feedback loop, the nutrients regenerated by the zooplankton enhance the nutrient status of their prey and the predator population grows rapidly; the microalgal crop is thence removed rapidly as the well-fed zooplankton proliferate. However, if microalgal N/C and/or P/C become decreased due to nutrient exhaustion, then trophic transfer to the grazer is less effective; nutrient regeneration by the zooplankton is decreased or even stalled and the crop remains of poor nutritional status. Such effects on grazer activity were explored by Hessen et al. ([@CR27]) under a variety of nutrient/light conditions. When zooplankton were introduced to a stable algae culture, an increase in P-availability combined with low light levels increased the algal P/C and the zooplankton soon dominated. Conversely, zooplankton growth was slowest under high light/low P conditions, and hence, algal loss was low due to poor grazing (Hessen et al. [@CR27]). The effects of such positive feedback processes have been demonstrated using models (Mitra and Flynn [@CR39]). The approach needed in commercial microalgal production is to exploit these ecological features to advantage. That is relatively easy if the crop is grown for high-C products, as such microalgae are naturally poor prey. But what if the crop is grown for high protein content? Recent evidence indicates that biomass and biochemical production by microalgal crops may be maintained when the microalgae are grown to N/P ratios higher than the Redfield N/P mole ratio of 16 (Mayers et al. [@CR33]). This is of immediate benefit because it decreases the demand for phosphorus fertiliser---an expensive and dwindling resource (Elser and Bennett [@CR12]; Chisti [@CR7]). However, there is an additional benefit because growth of zooplankton pests appears to be affected more by a lack of P (Hessen et al. [@CR27]) than are the microalgae, providing a more grazer-resistant crop (Fig. [2](#Fig2){ref-type="fig"} vs [3](#Fig3){ref-type="fig"}). In the simulations run here, the stoichiometric interactions between predator and prey were, in accordance with simple stoichiometric ecology (Sterner and Elser [@CR52]), those directly and simply (linearly) related to differences in elemental C/N/P. In reality, deviations in C/N/P in phototrophs are often associated with other biochemical events such as the accumulation of secondary metabolites which are distasteful to grazers, if not potent toxins. Being stressed by P, rather than N, is most closely allied to accumulation of secondary metabolites that are noxious to zooplankton (Granéli and Flynn [@CR22]). N-stress may also produce microalgae that are distasteful to zooplankton (Flynn et al. [@CR16]) even though they may not be classed as noxious in a human nutrition context. Given that zooplankton grazing is most damaging when the prey (the microalgal crop) has a well-balanced C/N/P, it is fortuitous that many products of value from microalgae are derived from high-C metabolites (Greenwell et al. [@CR23]) which are products of cells with elevated C/N and that minor changes in algal C/P may not be damaging to crop production (Mayers et al. [@CR33]). Interestingly, P-limitation also limits chytrid parasitic infections of microalgae (Bruning [@CR5]). Coupled with the projected increase in the cost of P-fertilisers, there appear sound reasons for minimising the addition of P in all microalgal culture systems. Electronic supplementary material ================================= {#Sec14} ESM 1(XLSX 158 kb) ESM 2(PDF 475 kb) **Electronic supplementary material** The online version of this article (doi:10.1007/s10811-017-1112-8) contains supplementary material, which is available to authorized users. This work was supported through the EnAlgae project (ref. 215G), which received ERDF funding through the INTERREG IVB North West Europe programme with co-financing from the Welsh Government, and Natural Environment Research Council (UK) grants NE/K001345/1 (via subcontract to A.Mitra) and NE/F003455/1.

###### What is already known? - Abdominal and thoracic cancers cause debilitating illness, and surgery is associated with significant decline in physical function. - Exercise initiated after completion of active cancer treatment has a beneficial effect on health-related quality of life. ###### What are the new findings? - There is insufficient evidence that preoperative or postoperative resistance muscle-strengthening exercise improves or negatively affects functional outcomes in patients undergoing abdominal surgery for cancer. - Large-scale, well-designed clinical trials are required to determine whether resistance muscle-strengthening exercise is beneficial for patients undergoing abdominal surgery for cancer. Introduction {#s1} ============ Background {#s1a} ---------- Abdominal and thoracic cancers affect about 12 000 people annually in the UK. Many of these patients will undergo surgery, after which there is a high risk of postoperative complications and significant decline in physical function. A systematic review of exercise for people with cancer by Stevinson *et al*[@R1] found some evidence that those who exercised had better physical function compared with those who did not exercise, but there was insufficient evidence to demonstrate improvement in quality of life. In addition, they were not able to determine which type of exercise intervention was best or if any had long-term benefit. A more recent Cochrane review of exercise for people with cancer by Mishra *et al*[@R2] found that exercise initiated after completion of active cancer treatment (ie, surgery, chemotherapy, radiation therapy or hormone therapy) has a beneficial effect on health-related quality of life, although no parallel improvement in self-reported physical function was found. The exercise interventions included in this review varied greatly and included strength training, yoga, walking, cycling, tai chi and qi gong. However, due to the small number of studies available, these authors were not able to evaluate the effect of different modes and intensities of exercise. Furthermore, studies of exercise in the preoperative and early postoperative stages were not included in the review. Therefore, it is not known whether exercise, when commenced before the end of active cancer treatment, would have additional benefit on physical function for those undergoing surgery. While there is growing evidence on the beneficial effects of aerobic exercise, resistance exercise training has received much less attention.[@R3] It is thought that resistance exercise training could act to aid recovery of muscle function.[@R7] It has long been established that resistance exercise training is effective in stimulating muscle anabolic processes and increasing muscle strength.[@R8] It may even counteract some of the metabolic pathophysiology associated with cachexia.[@R9] Furthermore, it can be performed with very little equipment and space and while patients are bed-bound in hospital or at home. Although there have been previous systematic reviews of the effects of exercise training, there have not been any that have specifically focused on resistance training. Previous reviews, relating to exercise training for patients with cancer, have mostly focused on specific outcomes such as fatigue and quality of life,[@R4] and most have centred on specific types of cancer.[@R10] Galvão and Newton[@R18] published a review of exercise intervention studies for all cancers and a meta-analysis of exercise training interventions. However, their review included a heterogeneous group of studies including some that were not randomised or had no control group. Quality systematic reviews require critical appraisal of the quality of the reviewed studies and share accurate descriptions of the design, delivery and interpretation of what was done in the study. In some instances detailed description of these aspects is not available.[@R19] One of the main challenges in studying the effects of a resistance exercise programme on physical function in cancer surgery patients is in identifying an appropriate outcome measure. The review by Mishra and colleagues found no significant improvement in physical function as evaluated using self-report questionnaires, but they did not measure any index of physical performance.[@R2] Therefore, our aim was to undertake a systematic review of the literature on interventional studies investigating the effects of preoperative and postoperative resistance exercise training on recovery of physical function in patients undergoing abdominal surgery for cancer. The findings will provide clinicians and investigators a basis to choose exercise interventions for use in clinical practice or for future research. Methods {#s2} ======= The Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines on systematic reviews were followed for this review.[@R20] [Figure 1](#F1){ref-type="fig"} summarises the review process. {#F1} Search strategy {#s2a} --------------- The Cochrane Library, EBSCO (SPORTDiscus and Cumulative Index to Nursing and Allied Health Literature (CINAHL)), PLOS, PubMed (Medline) and Elsevier (Scopus) electronic databases were searched up to and including December 2014. The search strategy used was exercise OR training OR isometric OR static OR isotonic OR concentric OR eccentric OR resistance OR strengthening exercise OR exercise therapy OR circuit training OR rehabilitation OR physiotherapy; AND neoplasm OR abdominal cancer OR stomach cancer OR gastric cancer OR bowel cancer OR pancreatic cancer OR colorectal cancer OR colon cancer OR rectal cancer OR gastrointestinal cancer OR ovarian cancer OR endometrial cancer OR cervical cancer OR renal cancer OR kidney cancer OR bladder cancer OR uterine cancer OR gynaecological cancer OR urological cancer; AND abdominal surgery OR laparotomy OR laparoscopy OR laparoscopic OR anterior resection OR colectomy OR hemicolectomy; AND clinical trial OR random controlled trial OR quasi-randomised controlled trial OR controlled trial OR comparative trial. All titles and abstracts generated by the search were independently screened for inclusion by three authors (DS, FH and KC). Disagreement between authors was discussed and consensus was reached. The search was restricted to English language and were included if the following criteria were met: (1) randomised, quasi-randomised or controlled trial study design comparing a muscle-strengthening exercise intervention (ie, exercise using resistance to induce muscular contraction) ± other therapy with a comparative group; (2) included adult participants (≥18 years) who underwent abdominal surgery (ie, surgery pertaining to the contents of the abdominal cavity, its walls and orifices) for cancer; and (3) included muscle strength, physical function, self-reported functional ability, range of motion and/or performance-based test as an outcome measure. Data extraction {#s2b} --------------- Participants' age, gender, diagnosis, surgical procedure and sample size were extracted from the included studies, along with a description of the exercise intervention, including muscle group or groups exercised, contraction effort, number of repetitions and frequency, length of programme, length of follow-up, group or individual exercise programme, home or supervised exercise programme, and timing of programme (presurgery and/or postsurgery). Data synthesis and analysis {#s2c} --------------------------- The aim of this review was to evaluate the effect of resistance muscle strengthening on physical function in people undergoing abdominal surgery for cancer. For each study, means and SD of outcomes focused on physical function were extracted. Outcomes relating directly to surgery, length of stay, infection and other postsurgical complications were not considered in this review. Assessment was made of the outcome measures for physical function that were used in different studies, before progression to pooling of data for analysis of the most common outcome measure. Treatment effect of individual studies is reported as mean difference and 95% CIs, and the data summarised. Risk of bias was assessed with the Physiotherapy Evidence Database (PEDro) scale.[@R21] Items assessed included exclusion criteria, procedures for group allocation and missing data, participant, therapist and assessor blinding, and reporting of results. Studies were then graded using the Cochrane Reviews Grading of Recommendations Assessment, Development and Evaluation criteria.[@R21] Results {#s3} ======= Search strategy and selection of articles {#s3a} ----------------------------------------- The initial search strategy resulted in 588 publications. Following screening of titles and abstracts, 24 studies met the inclusion criteria and were accessed for review of the full text, of which 2 eligible studies[@R23] were included in the review (see [table 1](#T1){ref-type="table"} and [figure 1](#F1){ref-type="fig"}). Full-text studies were excluded for a number of reasons: (1) the study lacked a well-defined muscle-strengthening intervention (n=18); (2) the study did not include patients undergoing abdominal surgery for cancer (n=4); and (3) the study did not use a physical function outcome measure (muscle strength, self-report questionnaires or physical performance measures). ###### Characteristics of included studies Methods Participants Intervention Relevant outcomes Risk of bias ------------------------ ---------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------ ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------- Dronkers *et al*[@R23] Randomised study investigating the preoperative effect of an exercise programme in participants with colon cancer. Exercise group, n=22Age: 71.1±6.3Gender: 15 male, 7 femaleControl group, n=20Age: 68.8±6.4Gender: 16 male, 4 female Supervised programme 2×week for 2--4 weeks (mean 5.1±1.9) and home-based programme of walking or cycling for a minimum of 30 min per day (perceived exertion of 11--13 Borg Scale). Programme:Warm up.Resistance training of the lower limb extensors---equipment and method not stated (maximum of 1 set of 8--15 repetitions at 60%--80% of the one repetition maximum).Inspiratory muscle training (10%--60% max inspiratory pressure for 240 breathing cycles.Aerobic training---method and equipment not stated (55%--75% max HR or perceived exertion of 11--13 Borg Scale for 20--30 min).Functional activities according to patients' capabilities and interests (Vreede *et al*,[@R28] regimen---no other information provided). Timed Up and GoChair rise timePhysical Activity QuestionnaireAbbreviated Fatigue QuestionnaireEORTC QLQ-C30 Global Health/Functional Scale/Symptom Scale PEDro score 8/11GRADE criteria---moderate Ahn *et al*[@R24] Randomised study investigating the effect of a postsurgical, inpatient exercise programme in patients with stages I--III colon cancer. Exercise group, n=17Age: 55.61±7.11Gender: 12 male, 5 femaleControl group, n=14Age: 57.43±6.12Gender: 5 male, 9 female Supervised exercise programme 2×day, 15 min/sessionSubdivided into three phases:Implemented while subjects were still unable to get out of bed: stretching (neck, shoulder, wrist, ankle and pelvis), pelvic tilt---isometric, resistance exercise (ankle dorsiflexion and plantar flexion against the hand of the therapist), unsupervised sitting or walking in the ward.Performed once subjects were able to get out of the bed, but had limited ambulation: stretching (whole body, leg and shoulder), pelvic tilt and thrust, one leg raise, crunch, resistance exercise (1 set, 10 repetitions) with 1 lb weight (chest, shoulder, arm, thigh and calf), unsupervised walking.Performed when subjects were able to ambulate without any discomfort; in addition to phase 2 exercises, resistance strengthening increased to 12 repetition×3 sets, supervised balance exercises---one leg standing, one leg calf raise, hip adduction, hip abduction, hip flexion with knee bent, hip extension, unsupervised walking. Timed one-leg standSit-to-stand in 30 sTecumseh step test PEDro score 8/11GRADE criteria---moderate EORTC QLQ-C30, European Organization for Research and Treatment of Cancer Quality of Life Questionnaire; GRADE, Grading for Recommendations Assessment, Development and Evaluation; HR, heart rate; PEDro, Physiotherapy Evidence Database. Description of included studies {#s3b} ------------------------------- Characteristics of the participants and intervention of the two included studies are presented in [table 1](#T1){ref-type="table"}. Both were small (n=42 and 31) single-centre studies investigating participants undergoing abdominal surgery for excision of cancer of the colon. Dronkers *et al*[@R23] investigated the effect of a preoperative exercise programme on preoperative outcomes, and Ahn *et al*[@R24] investigated the effect of a postoperative exercise programme on short-term outcomes at discharge from hospital. The participants in the preoperative study were aged 10--15 years older than those in the postoperative study. In terms of gender, a higher proportion of men participated in both studies. The preoperative intervention of Dronkers *et al*[@R23] included a twice-weekly supervised exercise programme and a home-based programme of walking or cycling for a minimum of 30 min per day for 2--4 weeks before admission for surgery. In addition to a single set of resistance strengthening exercises of the leg (8--15 repetitions at 60%--80% of the one repetition maximum), the programme included inspiratory muscle training, aerobic training at 55%--75% max heart rate (HR) or perceived exertion of 11--13 Borg Scale for 20--30 min, and functional activities. A full description of the resistance exercise was not published. Three of the intervention groups (13.6%) did not complete the study with their data analysed as intention to treat. The postoperative intervention of Ahn *et al*[@R24] comprised a twice-daily 15 min supervised exercise programme performed by the participant until discharge from hospital (mean 8.87±2.28 days). In addition to resistance strengthening exercises of the chest, shoulder, arm, thigh and calf leg, the programme included stretching exercises for the neck, shoulder, wrist, ankle and pelvis, core trunk exercises and ambulation. In terms of the strengthening exercises, resistance was applied manually by the therapist initially and then using 1 lb free weights. During phase 2, one set of 10 repetitions was performed, and in phase 3, three sets of 12 repetitions were performed. Because these studies used different outcome measures, it was not possible to pool the data in order to analyse mean changes in physical function outcomes. Risk of bias of included studies {#s3c} -------------------------------- The methodological quality of the two included studies was rated as moderate according to the GRADE criteria, that is, randomised studies with unclear bias or well-designed observational studies with large, consistent and precise estimates of the magnitude of an intervention effect. Difficulty in blinding trial participants and therapists to the intervention meant studies were not rated as high. Both studies scored 8 out of 11 on the PEDro scale. Block randomisation using prepared envelopes, stratified by age (60--70 and \>70) by someone independent of the study, was used in the preoperative study. Randomisation, at a 1-to-1 ratio, into study groups via minimisation to balance prognostic factors between groups (age and gender) was used in the postoperative study. In the preoperative study the gender distribution was similar in the control and intervention groups; however, in the postoperative study, twice as many men were randomised to the exercise group than the control group despite the minimisation procedures to balance gender between groups. In relation to the description of the intervention, some information was lacking in terms of equipment and methodology with regard to the aerobic and functional activity components of the preoperative intervention. Effect of strengthening exercise {#s3d} -------------------------------- ### Preoperative muscle strengthening {#s3d1} The mean difference and upper and lower 95% CI between the control and intervention group in the study by Dronkers *et al*[@R23] are shown in [table 2](#T2){ref-type="table"}. The five-session preoperative exercise programme had no significant effect on preoperative Timed Up and Go, chair rise time test, self-reported physical activity, quality of life and fatigue. Statistical power for six out of the seven measures was unacceptably low. Effect on postsurgery outcomes was not evaluated. ###### Summary of effect of exercise intervention Mean between-group difference Lower 95% CI Upper 95% CI Statistical power\* ---------------------------------------------------- ------------------------------- -------------- -------------- --------------------- Dronkers *et al*,[@R23] preoperative intervention  Timed Up and Go (s) −1.20 −2.78 0.38 31.2  Chair rise (s) −5.40 −9.24 −1.56 77.3  Physical activity (min/day) 44.00 −141.82 229.82 7.3  Abbreviated Fatigue Questionnaire −3.90 −7.41 −0.39 57.6  EORTC QLQ-C30 (Global Health) −4.00 −15.57 7.57 10.2  EORTC QLQ-C30 (Functional Scale) 12.00 −28.26 52.86 87.4  EORTC QLQ-C30 (Symptom Scale) 36.00 −31.09 103.09 17.7 Ahn *et al*,[@R24]  postoperative intervention  Timed one-leg stand (s) −7.28 −16.25 1.69 40.0  Sit-to-stand (repetitions) −2.00 −5.78 1.78 17.7  Tecumseh step test (heart rate, beats/min) 10.29 1.63 18.95 64.8 \*Probability of rejecting a false null hypothesis (where α=0.05), for a between-group comparison of means at study endpoint. ### Postoperative muscle strengthening {#s3d2} The mean difference and upper and lower 95% CI between the control and intervention groups in the study by Ahn *et al*[@R24] are also shown in [table 2](#T2){ref-type="table"}. The inpatient postoperative exercise programme had no significant effect at time of discharge from hospital on ability to balance on one leg, number of sit-to-stands in 30 s or aerobic capacity (estimated from performance of the Tecumseh step test). Statistical power was not sufficient to allow any conclusion for or against the preferential use of any of the outcome measures that were used in this trial. Effect on functional recovery postdischarge from hospital was not evaluated. Discussion {#s4} ========== Our aim was to systematically review the evidence on the effectiveness of preoperative and postoperative strengthening exercises on short-term and long-term recovery of physical function in patients undergoing abdominal surgery for cancer. Two studies were included, which represented 73 patients (48 men and 25 women) undergoing abdominal surgery for cancer. One exercise programme was undertaken preoperatively and the other postoperatively until discharge from hospital. This represents insufficient evidence to determine whether this type of preoperative or postoperative resistance muscle-strengthening exercise programme improves or negatively affects functional outcomes in patients undergoing abdominal surgery for cancer. The study by Dronkers *et al*,[@R23] which investigated a preoperative exercise programme, was statistically underpowered with the exception of the functional measure derived from the quality of life scale. The programme included resistance strengthening of the lower limb muscle extensors and was performed for a mean of five sessions. This may not be sufficient to provide an adequate training stimulus to significantly increase muscle strength. Indeed, guidelines published by the American College of Sports Medicine recommend resistance exercise 2--3 times per week with 2--4 sets of 10--15 repetitions to improve strength in middle-aged and older persons.[@R25] In contrast, the study by Ahn *et al*[@R24] investigated a postoperative exercise programme, but this was also statistically underpowered and provides inconclusive evidence in support of the intervention and the use of particular outcome measures. The intervention was different from that of Dronkers *et al*[@R23] in that it used a progressive resistance programme involving the upper and lower limbs, together with stretching, functional balance strengthening and walking. Also, isometric strengthening exercises were commenced early postoperatively while the patient was still in bed and then progressed to 'resistance-through-range' strengthening as well as balance strengthening exercises, until discharge from hospital. The mean hospital length of stay, for the study of Dronkers *et al*[@R23] was 7 days for the control group, and in the exercise group it was 8 days. Similarly, for the study by Ahn *et al*,[@R24] it was 8 days of exercise, and it is likely that this will not provide an adequate training stimulus to significantly increase muscle strength and function. There are some limitations to our review. We limited our inclusion by study design, only including randomised or quasi-randomised studies where there was a clear resistance muscle strengthening component as part of an exercise programme. It is possible that other studies have included muscle-strengthening exercises or functional exercises that will have an effect on muscle strength that have not been included in this review due to our inclusion criteria, and we advocate the Consensus on Exercise Reporting Template guidelines for reporting exercise intervention studies.[@R26] The two studies included in the review recruited almost twice as many men as women, and the results may not reflect the general population. Future studies should focus on detailed descriptions of the exercise intervention, consistent outcome measures and longer intervention and follow-up times.[@R27] Our systematic review suggests that the use of resistance exercise interventions for recovery of physical function in patients undergoing abdominal surgery for cancer must be considered with caution. The small number of included underpowered studies and the inability to pool the results due to the heterogeneity of outcome measures mean that there is a lack of evidence for or against the use of this type of resistance muscle-strengthening exercise programmes to improve functional outcomes in these patients. While the studies give encouraging preliminary evidence that muscle-strengthening programmes may be feasible for abdominal cancer surgery patients, further large-scale, well-designed clinical trials are required to determine whether this type of exercise intervention is beneficial for this group of patients. **Contributors:** DS, FH, KC performed the systematic review. DS, FH, KC, AB, MG, RPH, MH, IH, DPL, TPH, CS, IS, LT, WH and HA contributed to study design, data analysis and interpretation and preparation of the manuscript. **Funding:** The study was funded by an NIHR Research for Patient Benefit grant (PB-PG-0613--31107). **Competing interests:** None declared. **Patient consent:** Not required. **Provenance and peer review:** Not commissioned; externally peer reviewed.

This series on immunotherapies in acute myeloid leukemia (AML) aims to give readers new insights on established but also emerging immunotherapeutic approaches for AML patients. The therapeutic landscape in AML is rapidly changing, and several drugs have been developed and approved such as first and second generation FLT3 inhibitors \[[@B1-jcm-08-02054],[@B2-jcm-08-02054],[@B3-jcm-08-02054]\], IDH1 and 2-inhibitors \[[@B4-jcm-08-02054],[@B5-jcm-08-02054]\], demethylating agents, liposomal cytarabine and daunorubicin (CPX-351) \[[@B6-jcm-08-02054]\], venetoclax \[[@B7-jcm-08-02054],[@B8-jcm-08-02054]\] and the hedgehog pathway inhibitor glasdegib. However, relapse after intensive chemotherapy or allogeneic hematopoietic stem cell transplantation is one of the major obstacles impeding the complete elimination of all AML cells \[[@B9-jcm-08-02054]\]. Thus, although the median overall survival for AML patients has increased, it still remains relatively low \[[@B10-jcm-08-02054]\]. Therefore, immunotherapeutic approaches might be an option to prevent disease relapse and to eliminate leukemic cells or leukemic stem cells (LSC) that survive intensive treatment approaches. The efficacy of immunotherapeutic approaches has become ever more evident in solid tumors, especially immune-checkpoint inhibitors that are routinely used in several solid tumor entities, but also lymphoma \[[@B11-jcm-08-02054],[@B12-jcm-08-02054]\]. Our focus in this special issue is different strategies of immunotherapeutic approaches in AML. Some of the immunotherapies in the treatment of AML, such as allogeneic hematopoietic stem cell transplantation (HSCT) and donor lymphocyte infusion (DLI), have been part of routine clinical practice in the treatment of AML for a long time, whereas other immunotherapeutic approaches have only recently entered clinical practice or need to be further developed. A key aspect is the mechanisms underlying the cure of AML patients, which are based on the graft-versus-leukemia (GvL) effect, in which allogeneic T cells recognize target antigens on malignant cells by T cell approaches including DLI. An effective and well-tolerated regimen for HSCT in patients with AML and MDS is the FLAMSA-RIC regimen, and therefore novel data of this approach are presented in this issue \[[@B13-jcm-08-02054]\]. It is very appropriate to utilize DLI after allogeneic HSCT to prevent relapse, to prolong progression-free survival, to establish full donor chimerism, and to restore the GvL effect in patients with hematological malignancies. There are different strategies to use DLI in a therapeutic setting for the treatment of morphological relapse, and also for prophylactic use in AML/MDS and DLI administered preemptively. There is also the approach of antigen-directed immunogenic and specifically stimulated and modified DLI as well as virus-specific donor T cells and third-party DLI \[[@B14-jcm-08-02054]\]. DC-based immunotherapies also have the potential to bring about demonstrable clinical responses in AML patients, although there has not been a complete breakthrough for this type of therapy until today. Van Acker et al. have highlighted different DC strategies in AML \[[@B15-jcm-08-02054]\]. Leukemia-associated antigens (LAAs) represent immunogenic structures to target LSC \[[@B16-jcm-08-02054],[@B17-jcm-08-02054]\], and LAA might be relevant for the elimination of malignant cells by cytotoxic T lymphocytes. Therefore, LAAs might be a good target for specific immunotherapeutic approaches. Several LAAs have been identified in the context of malignant hematological diseases \[[@B16-jcm-08-02054],[@B18-jcm-08-02054],[@B19-jcm-08-02054]\], and in clinical phase I/II peptide vaccination trials, some LAAs showed immunological as well as clinical responses \[[@B20-jcm-08-02054],[@B21-jcm-08-02054],[@B22-jcm-08-02054],[@B23-jcm-08-02054]\]. In this special issue, we also elucidate antibody-based therapies in AML, such as T cell activating antibodies including immune-checkpoint inhibitors and diverse monoclonal antibodies \[[@B11-jcm-08-02054],[@B12-jcm-08-02054],[@B24-jcm-08-02054]\]. Immune-checkpoint inhibitors have changed clinical treatment algorithms of malignant diseases such as malignant melanoma, lung cancer, as well as lymphoma. Today, immune-checkpoint inhibitors are not yet established in the routine treatment of AML but should be considered as further immunotherapeutic options in the future, especially in the context of allogeneic stem cell transplantation \[[@B24-jcm-08-02054]\]. Further antibody-directed approaches such as unconjugated, toxin-conjugated, radio-conjugated, and multivalent formats of antibody-based therapy, are demonstrating the potential of a diverse leukemia-derived antibody strategy which is already established in acute lymphoblastic leukemia and are summarized in one section of this issue \[[@B25-jcm-08-02054]\]. Chimeric antigen receptor T cells (CARs) are highly effective in the treatment of refractory and relapsed acute lymphoblastic leukemia, to some lower extent in aggressive lymphoma, but also in multiple myeloma \[[@B26-jcm-08-02054]\]. However, early CAR-T cell approaches are also being tested in AML with interesting target structures, and these strategies are described in this issue \[[@B27-jcm-08-02054]\]. Immune responses are complex and are also influenced by T cell cross-talk and communication by cytokines and the communication of leukemic cells with their microenvironment, as presented by Reikvam et al. \[[@B28-jcm-08-02054]\] in this issue. All of these aspects emphasize the high potential of immunotherapeutic approaches to improve the survival of AML patients in the future, where combination therapies utilizing immunotherapeutic drugs could represent further innovation strategies to further improve the treatment of AML. The author declares no conflict of interest.

1. Introduction {#sec0005} =============== In comparison to human and mice, germ-line immunoglobulin variable region gene diversity in cattle is highly limited. The antibody repertoire is derived from a single polymorphic VH gene family ([@bib0015]) and is dominated by one of two VL gene families (Vλ1) ([@bib0030]). Similar to sheep and horses, 90% of the light chains that are expressed in cattle are the λ-isotype. It is generally agreed that V region diversification in cattle is generated by somatic hyper mutation following VDJ segment rearrangement ([@bib0145]). The limited sequence diversity in the expressed Vλ repertoire in cattle has led to the suggestion that it contributes relatively little to antigen recognition and that most of the immune response resides in the VH ([@bib0115]). A unique feature of the antibody response in cattle is the generation of a subset (up to 10%) of heavy chains that have a highly extended H3 of over sixty residues. This compares to an average of 20 residues for most bovine heavy chains, which in itself is longer than in other species such as human and mouse, typically 8--16 residues. The ultra-long H3s contain a large number of cysteine residues that cross-link to stabilise the structure ([@bib0110]; [@bib0155]). We have investigated the role of the light chain in the antigen-binding activity of two bovine monoclonal antibodies (mAbs) by analysing the functional and structural consequences of exchanging their light chains. The two antibodies, designated B4 and B13, previously produced from bovine x mouse heterohybridomas ([@bib0125]; [@bib0115]) are specific for the fusion (F) protein of respiratory syncytial virus (RSV). The F protein epitopes recognised by mAbs B4 and B13 are conserved in both human (h) and bovine (b) RSV, and are a major cause of lower respiratory tract infections in young children and calves, respectively. Both B4 and B13 are potent neutralizing, fusion-inhibiting antibodies and protect mice against hRSV infection and calves against bRSV infection ([@bib0075]; [@bib0135]). Competitive binding assays, recognition of antibody-escape mutants, and binding to synthetic peptides have shown that mAb B4 and B13 recognise different antigenic sites on the F protein. B4 recognises antigenic site II (residues 255--275) ([@bib0005]; [@bib0125]), whereas B13 binds to an epitope contained within antigenic site IV, V, VI (residues 417--438) ([@bib0165]). We have produced recombinant Fab fragments of B4 and B13 and also molecules where the heavy and light chains have been exchanged between the two molecules. Analysis of their structures and antigen binding properties suggests a dominant role for the bovine H3 in antigen-binding but shows that the VL also plays a key role in both the folding and positioning of H3 where the functional consequences of this depend on the structure of H3. 2. Materials and methods {#sec0010} ======================== 2.1. Protein production {#sec0015} ----------------------- Two vectors were constructed containing resident bovine Cλ and IgG1 CH1 sequences and a signal sequence. The light chain constant region is *Bos taurus* allotypic variant IGLC3a (Genbank DQ537487: and [HQ456934](ncbi-n:HQ456934){#intr0005}; ([@bib0035])). The CH1 heavy chain is *Bos taurus* IgG1 (Genbank S82409: ([@bib0070])). Synthetic genes encoding the constant regions were inserted by Infusion^®^ cloning into PmeI-HindIII cut pOPING-ET ([@bib0090], [@bib0100]) The vectors have been engineered so that VH and VL sequences can be inserted into the KpnI-PstI (pOPINBOVL) and KpnI-SfoI (pOPINBOVH) restriction sites by Infusion^®^ cloning. Synthetic genes encoding the candidate variable regions (B4 and B13 VH and VL) ([@bib0010]) were purchased from IDT Technology as "Infusion-ready" gBlocks and inserted into the pOPIN expression vectors. All vectors were sequenced to confirm clones were correct. Recombinant Fabs were produced by co-transfection of VH and VL vectors into Expi293™ cells according to the manufacturer's protocol (Invitrogen). Proteins were purified from culture supernatants by a combination of immobilised metal affinity and size-exclusion chromatography using an automated protocol implemented on an ÄKTAxpress (GE Healthcare) ([@bib0095]). Purified proteins were analysed by size-exclusion multi-angle light scattering (SEC-MALS) using a Superdex 200 10/300 Increase column (GE Healthcare) and an AktaPure 25 System (GE Healthcare). The protein sample (100 μL), at a concentration of 1.0 mg/mL, was loaded onto the gel filtration column and eluted with one column volume (24 mL) of buffer, at a flow rate of 0.7 mL/min. The eluting protein was monitored using a DAWN HELEOS-II 18-angle light scattering detector (Wyatt Technologies) equipped with a WyattQELS dynamic light scattering module, a U9-M UV/Vis detector (GE Healthcare), and an Optilab T-rEX refractive index monitor (Wyatt Technologies). Data were analysed by using Astra (Wyatt Technologies) using a refractive increment value of 0.185 mL/g. Purified proteins were also analysed under reducing and non-reducing conditions by electrospray mass spectrometry using a Waters Q-Tof Micro mass spectrometer ([@bib0085]). Recombinant bRSV F protein was a generous gift from Peter Kwong (National Institutes of Health, Bethesda, MD) ([@bib0175]). 2.2. Thermal shift assay {#sec0020} ------------------------ A thermal shift assay ([@bib0085]), using the fluorescent dye SYPRO Orange (Invitrogen), was employed to determine protein stability. Solutions of Fabs at 4 μM and 1000x SYPRO orange were made in 20 mM Tris pH 7.5, 200 mM NaCl and 40 μl was added to a 96-well thin-wall PCR plate (Thermo Scientific). The plates were sealed with adhesive PCR seal (4titude) and heated in an Mx3005p qPCR machine (Stratagene) from 25 to 95 °C at a rate of 1 °C/min. Fluorescence changes were monitored with excitation and emission wavelengths of 492 and 610 nm respectively. 2.3. Antigen-binding assays {#sec0025} --------------------------- An antigen-binding ELISA assay was performed as previously described ([@bib0130]). Briefly, 50 μl of virus antigen (lysate from BRSV infected Vero cells) was coated onto immunoassay plates alongside control antigen (lysate from mock-infected Vero cells) and left to dry overnight at 37 °C. The wells were then blocked for 1 h at room temperature with 200 μl PBS containing 5% pig serum and 0.05% Tween 20. After washing the wells four times with PBS containing 0.05% Tween 20, serial dilutions of Fabs were added to the wells (final volume of 50 μl in each well) followed by incubation at room temperature for 1 h. The wells were washed five times with PBS containing 0.05% Tween 20 and 50 μl 1:4000 Rabbit anti-Bovine IgG F(ab′)2HRP (Sigma SAB3700010) conjugate added to each well. The plate was incubated at room temperature for 1 h. After washing five times in PBS containing 0.05% Tween 20, 100 μl freshly prepared TMB substrate was added to each well. The reaction was stopped with 50 μl 1 M H~2~SO~4~ when the colour change had reached the optimum. Absorbance was read at 450 nm and 690 nm. 2.4. Surface plasmon resonance {#sec0030} ------------------------------ The surface plasmon resonance experiments were performed using a BiaCore T200 (GE Healthcare) equipped with a Series S CM5 sensor chip. The F protein was immobilized using amine-coupling chemistry. The surfaces of the sample and reference flow cells were activated for 7 min with a 1:1 mixture of 0.1 M NHS (*N*-hydroxysuccinimide) and 0.1 M EDC (3-(*N,N-*dimethylamino) propyl*-N-*ethylcarbodiimide) at a flow rate of 5 μL/min. The F protein, at a concentration of 6.7--20 μg/mL in PBS pH 6.5, was immobilized at a density of 20--100 RU on the sample flow cell. The reference flow cell was left blank. All the surfaces were blocked with a 7 min injection of 1 M ethanolamine, pH 8.0. To collect steady-state affinity data, the Fabs were injected over the two flow cells at a range of concentrations prepared by serial four-fold dilution of the Fab stock from 5 μM (including five duplicates and zero concentration samples), at a flow rate of 30 μL/min and at a temperature of 20 °C. For B4 and B4\* the assay was performed in PBS pH 7.4, the complex was allowed to associate and dissociate for 180 and 600 s, respectively, and surfaces were regenerated with two 120 s injections of 0.1 M Glycine pH 3.0, 0.5 M NaCl. For B13 and B13\* the assay was performed in PBS pH 7.4 with 0.5% (v/v) surfactant P20 to reduce non-specific binding, the complex was allowed to associate and dissociate for 300 and 300 s, respectively, and surfaces were regenerated with one 120 s injection of 0.1 M Glycine pH 1.5. The data were fitted to a 1:1 interaction steady-state binding model using the BiaEvaluation 1.0 software. 2.5. Crystallization, data collection and structure determination {#sec0035} ----------------------------------------------------------------- Protein crystallizations were carried out using standard OPPF protocols ([@bib0150]). Crystals were grown by the sitting drop vapor diffusion method at room temperature. Each Fab crystallized in different buffer conditions containing PEG 3350 or PEG ME 2000 (see Table S1 for details). All crystals were flash-frozen in liquid nitrogen and then kept at −173 °C under a stream of nitrogen gas during data collection. X-ray data were collected on Diamond beamlines I03 for B4 and B13\*, I04-1 for B13 and I24 for B4\* Fabs (Diamond Light Source, Harwell, UK), see [Table 1](#tbl0005){ref-type="table"}.Table 1Data collection and refinement statistics.Table 1B4B4[a](#tblfn0005){ref-type="table-fn"}B13B13[a](#tblfn0005){ref-type="table-fn"}**Data collection**Space groupP21212P21P21P1Cell dimensions *a*, *b*, *c* (Å)78.4, 135.1, 42.142.4, 86.0, 165.087.9, 132.5, 118.243.4, 76.3, 130.0 α, β, γ (°)90.0, 90.0, 90.090.0, 98.6, 90.090.0, 102.9, 90.089.7, 87.4, 88.6Resolution (Å)67.8--1.90 (1.95--1.90)[a](#tblfn0005){ref-type="table-fn"}76.1--2.15 (2.29--2.15)85.7--2.12 (2.16--2.12)65.9--2.62 (2.67--2.62)*R*~merge~0.116 (0.922)0.094 (--)0.128 (--)0.044 (0.782)*I* / σ*I*14.4 (2.2)10.3 (1.0)8.6 (1.8)13.8 (1.4)Completeness (%)99.2(98.0)95.7 (72.4)99.2 (97.6)98.2 (98.1)Redundancy23.5 (10.9)6.1 (3.7)6.8 (6.0)3.5 (3.6)CC~1/2~--/--[b](#tblfn0010){ref-type="table-fn"}1.0 (0.55)1.0 (0.58)1.0 (0.52)**Refinement**Resolution (Å)67.7--1.9276.1--2.1585.7--2.1265.9--2.62No. reflections33138 (1723)58184 (2912)140485 (7327)46980 (2230)*R*~work~/*R*~free~0.229/0.2770.213/0.2350.256/0.2760.237/0.274No. atoms: Protein294364521956513125 Water59321139653*B*-factors: Protein75592396 Water53513960R.m.s. deviations: Bond lengths (Å)0.0160.0050.0100.012 Bond angles (°)1.80.81.41.6[^1][^2] The diffraction data were indexed, integrated and merged with the automatic data processing program Xia2 using the 3dii or Dials protocols ([@bib0170]; [@bib0160]). The statistics of the data are shown in [Table 1](#tbl0005){ref-type="table"}. The structure of the B4 Fab was solved first by molecular replacement using the coordinates of a human Fab, FabOX108 (PDB ID, 3DGG) as the search model. The orientation and position of the model in the crystal were determined using the program MOLREP ([@bib0140]) and refined with rigid-body refinement followed by restrained maximum-likelihood positional and B-factor refinements, using either REFMAC ([@bib0080]) or PHENIX ([@bib0060]). The refined structure of the B4 Fab was then used as a starting model for the structure determinations of the three remaining Fabs. Model rebuilding was carried out with COOT ([@bib0055]). 3. Results {#sec0040} ========== 3.1. Expression and purification of recombinant Fabs {#sec0045} ---------------------------------------------------- Cloning of the heavy and light chain variable regions of the anti-bRSV bovine monoclonal antibodies (B4 and B13) has been previously reported ([@bib0010]) and this information was used to construct recombinant Fabs of the two antibodies. Synthetic genes encoding the variable regions were inserted into plasmids containing resident signal sequences and light and heavy chain constant regions to produce B4 and B13 Cλ LC and IgG1Fd HC expression vectors respectively. A hexahistidine tag was added to the C-terminus of the HC for detection and separation of assembled Fabs from LC dimers. To investigate the role of the LCs in the assembly of the Fabs, Expi293™ cells were co-transfected with all combinations of light and heavy chains to produce four Fabs B4, B4\*, B13 and B13\*. Transfection of HEK cells with either B4 and B13 HC alone did not produce any heavy chain only antibody fragments confirming that assembly with the LC was essential for production of secreted protein. The secreted products were purified by IMAC followed by gel filtration and purity confirmed by mass spectrometry. Assembly of the recombinant B4, B13 Fd HC fragments with either B4 or B13 LCs into Fab fragments was demonstrated by SEC-MALS run under non-reducing conditions with the molecular weights of all four proteins corresponding to the expected size for a Fd-LC dimer ([Fig. 1](#fig0005){ref-type="fig"}). The thermal stability of the assembled Fabs was probed using a Thermal Shift Assay ([@bib0085]) and all, including the LC swapped Fabs, were found to be very stable ([Fig. 2](#fig0010){ref-type="fig"}). The transition melting temperatures (*T~m~*) for the B4 and B13 Fabs and the hybrid B4\* were within 1 °C of each other whereas the B13\* showed a 5 °C +/- 1 °C increase in *T~m~* indicating a stronger association of heavy and light chains.Fig. 1SEC-MALS profiles for the anti-RSV Fabs showing expected molecular weights for the intact Fabs with the inter-molecular disulphide bond intact. The results showed the molecular weight of B4 to be 46 kDa (calculated 47.7 kDa); B13 to be 48 kDa (calculated 48.4 kDa); B4\* to be 46 kDa (calculated 47.7 kDa) and B13\* to be 48 kDa (calculated 48.4 kDa)(For colour coding of SEC-MALS profiles in this figure , the reader is referred to the web version of this article).Fig. 1Fig. 2Melt curves from the thermal shift assay on all four Fabs showing high stability for both the matched and the chain-swapped versions. The *T*~m~ values for each Fab are: 71 ± 1 °C for B4; 70 ± 1 °C for B13; 68 ± 1 °C for B4\* and 75 ± 1 °C for B13\*.Fig. 2 3.2. Antigen-binding activity {#sec0050} ----------------------------- The location of the epitopes recognized by B4 and B13 were mapped onto the structure of the bRSV F protein to illustrate that their binding sites are spatially separate ([Fig. 3](#fig0015){ref-type="fig"}a). The recombinant Fabs were assessed for binding to bovine RSV (bRSV) in an antigen-binding ELISA using lysed Vero cells infected with bRSV as the source of antigen and the lysate of mock-infected cells as the control. [Fig. 3](#fig0015){ref-type="fig"}b shows that both the matched Fabs (B4 and B13) and one of the hybrid Fabs, B13\*, showed similar binding, whereas the B4\* light chain-swapped Fab showed a considerably reduced binding activity that was not significantly different from the mock control.Fig. 3(A) Surface representation of bovine RSV F trimer (PDB code 5TDG)^27^. 3 monomers are coloured in light blue, pale green and salmon, respectively. The antigenic site for B4 is in red and B13 in green. (B) ELISA data showing binding of Fabs to bRSV antigen against a negative control antigen. (C) SPR data fitted to a 1:1 steady-state affinity model. The vertical black lines indicate the *K~dapp~* obtained from the fit, with the exception of B4\* where this is shown by a dashed red line. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Fig. 3 Detailed analysis of antigen-binding to purified bRSV F protein was carried out by Surface Plasmon Resonance (SPR). The F protein was immobilised on the surface of the sensor chip and the Fabs passed over the surface. The assays were initially conducted in PBS pH 7.4 but it was found that B13 and B13\* displayed a high proportion of non-specific binding. Therefore, the assay conditions were extensively optimised for these two Fabs to reduce the effects of non-specific binding. The sensogram data were fitted to a 1:1 steady-state binding model under the assumptions that each subunit of the trimeric F protein binds one Fab, and that each subunit binds the Fab with the same affinity i.e. that there is no cooperativity between the F protein subunits upon Fab binding. These assumptions, and the presence of an element of non-specific binding, mean that the *K~d~*s obtained from this analysis should only be treated as apparent *K~d~*s. The *K~dapp~* obtained for the F protein/B4, B13 and B13\* were 84 nM, 125 nM and 234 nM respectively ([Fig. 3](#fig0015){ref-type="fig"}c). For B4\* an accurate *K~dapp~* could not be calculated as the assay had not reached saturation at a Fab concentration of 5 μM which was the maximum B4\* concentration used in the assay, but suggests that the affinity would be in the micromolar range ([Fig. 3](#fig0015){ref-type="fig"}b). Overall, the results of binding to purified F protein are consistent with the results of the cell lysate ELISA and demonstrate that B4\* shows much weaker binding to the F protein than the other Fabs. 3.3. Overall structure of Fabs {#sec0055} ------------------------------ To investigate the structural basis of the different functional effects of exchanging the light chains between B4 and B13, all four Fabs were crystallized, diffraction data collected and their structures solved by molecular replacement using an initial model of human FabOX108 (PDB ID: 3DGG) ([@bib0090]) ([Table 1](#tbl0005){ref-type="table"}). There were 1, 6, 2 and 4 Fab molecules in the crystallographic asymmetric units (AU) of the B4, B13, B4\* and B13\* crystals, respectively. For B4 a large portion of the CH1 domain of the heavy chain was disordered, including residues 149--153, 169--179, 198--215 and 222−233. For the Fabs with more than one molecule in the AU superimposition showed that the 6 B13 Fabs differed by rmsds of 0.7--1.4 Å, the 2 B4\* Fabs by only 0.3 Å and the 4 B13\* Fabs by 0.5--2.3 Å based on Cα atoms. The large structural differences in B13 and B13\* Fabs are mainly due to variations in the elbow angles between the variable and constant domains. Thus, superimposition of the variable domains of the 6 B13 and 4 B13\* Fabs showed smaller conformational differences between the variable domains with rmsds of 0.4−0.6 Å and 0.8−0.9 Å respectively ([Fig. 4](#fig0020){ref-type="fig"}).Fig. 4Overall structures of 4 Fabs. (A) B4 with HC in red and LC in blue. (B)-(D) Overlaps of 6 B13 (B), 2 B4\* (C) and 4 B13\* (D) Fabs in the crystallographic asymmetric units. Superimpositions are based on the variable domains only. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Fig. 4 3.4. Comparison of the variable domains of B4 and B13 {#sec0060} ----------------------------------------------------- The H3s of B4 and B13 are of different lengths, comprising 21 and 25 residues respectively. Both contain a pair of cysteines, at positions cys97-cys100G (B4), and cys95-cys100E (B13) (Kabat numbering) ([Fig. 5](#fig0025){ref-type="fig"}) and structures of the Fabs revealed that they form disulphide bridges irrespective of the paired LC. Both H3s adopt a double loop structure, though with different configurations. The two loops of B4 H3 are of a similar length with the N-terminal loop contacting the CDR3 of the LC (L3) and the C-terminal loop interacting with CDR1 (L1) and CDR2 (L2). The N-terminal loop of the longer B13 H3 adopts an extended β-hairpin structure contacting L1 and L2, and a shorter C-terminal loop folded between the H1 and L2 loops ([Fig. 5](#fig0025){ref-type="fig"}). Therefore, in contrast to the H3 of B4, that of B13 protrudes from the surface of the variable regions. Viewed from the top, the antigen binding regions of the two Fabs show differences in both shape and electrostatic potential ([Fig. 5](#fig0025){ref-type="fig"}). Both Fabs have a negatively charged cavity surrounded by positively charged areas, but the B13 cavity is much bigger and the negatively charged area is broader, consistent with the fact that the two Fabs bind to different epitopes of the bRSV F protein ([Fig. 3](#fig0015){ref-type="fig"}a).Fig. 5Conformations of CDR H3s. (A) B4 Fab. (B) B13 Fab. In both (A) and (B) HC is coloured in red and LC in blue. The disulphides are shown as orange sticks (note C100E is Kabat numbering). Surface charge profiles of the variable domains of B4 Fab (C) and B13 Fab (D) viewed from the top. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Fig. 5 3.5. Comparison of light chain exchanged Fabs {#sec0065} --------------------------------------------- In order to investigate how light chain pairing may affect the structure of the heavy chain variable domain and hence antigen binding, the structure of B4 was compared with B4\*, and B13 with B13\*. Interface analysis of the paired variable regions of the heavy and light chains showed little difference for the B13 HC with either LC. Table S1 shows a similar number of residues involved in the interface and similar interface surface area. A detailed analysis shows LC residues 47--52 in the B13\* structure are less buried than those in the B13 structure. For example, three residues in the B13\* Fab structure are solvent exposed (Ile47, Glu50 and Ser52) whereas they form part of the interface in the B13 structure (Ile47 is 100% buried, Asn50 is 50% buried and Lys 52 is 20% buried) (Table S1). Overlapping the H3 of B13\* and B13 shows that the overall β-hairpin structure of the H3 remains unchanged when the B13 LC is replaced by the B4 LC, despite being pushed slightly away by 2.5 Å from the light chain due to the presence of Tyr49 in the B4 LC instead of a glycine ([Fig. 6](#fig0030){ref-type="fig"}b). The observation that only small structural differences in the B13 HC occur on exchanging the B13 LC for the B4 LC is consistent with the very modest reduction in antigen-binding affinity measured by SPR.Fig. 6The effects of LC on the structure of HC. (A) Structural differences between B4 H3 and B4\* H3. B4 HC is in red and B4 LC in blue, B4\* is in grey. The running directions of H3s are indicated by arrows. (B) Structural differences between the H3s of B13 (red and blue) and B13\* (grey). Side chains and disulphides are shown as sticks. Residues are shown in Kabat numbering. For the light chains the final letter is in brackets and represents the one letter code for the sequence of the hybrid Fab (in all cases the letter preceding the residue represents the amino acid of the native Fab). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Fig. 6 By contrast, swapping the B4 LC for B13 LC leads to large changes in both orientation of the B4 VH domain and conformation of the H3 loop. The VH domain of B4\* differs by approximately 10° in orientation relative to the VL domain compared to B4 VH ([Fig. 6](#fig0030){ref-type="fig"}a). On exchanging the light chain from B4 to B13, the H3 flips around to include a π-interaction between Tyr100C (Kabat numbering) of B4 HC and Phe31 of the CDR1 of B13 LC. In the B4 Fab structure, this space is occupied by Trp31, which does not allow space for the Tyr100C interaction. In addition, the orientation of the H3 loop changes from the sequence running clockwise when viewed from the top in the B4 structure to anticlockwise in the B4\* structure ([Fig. 6](#fig0030){ref-type="fig"}a). The area of the interface between the variable domains of the B4\* Fab is lower than that for the B4 (Table S1). This analysis shows that amino acids 100E to 100 L of the HC in the B4\* structure have very little involvement in the interface except for Gly100 G, which is completely buried. In the B4 Fab structure, all of these amino acids are involved in the interface with ≥20% of the amino acid buried (Table S1). This string of amino acids at the interface in the B4 structure contributes four hydrogen bonds, involving Cys100E, Thr100 F, Gly100 G and Glu100 L, which are not present at the B4\* interface. The large conformation change that occurs on replacing the B4LC with B13LC radically alters the conformation of H3 and this disruption of the antigen-binding surface is consistent with the loss of activity. 4. Discussion {#sec0070} ============= In this study, we have examined the role of light and heavy chain pairing in the antigen-binding activity of two bovine monoclonal antibodies to the F protein of bRSV, B4 and B13. We have analysed the functions and structures of the Fab fragments from both parent antibodies expressed as recombinant proteins and versions of the Fabs in which the light chains had been exchanged. Overall, these results demonstrate that for B4 and B13, binding to the cognate LC is not essential for the formation of a stable Fab fragment. However, although all combinations of light and heavy chain assembled, there were different outcomes in respect of antigen-binding. In the case of B13 there was only a small reduction in binding affinity on swapping with the B4 light chain whereas exchanging B4LC for B13LC significantly compromised antigen-binding affinity by Fab B4. By examining the structures of all four Fabs, these observations could be explained by differences in the role of the LCs in positioning the H3 loops of the B4 and B13. In the case of B13, the conformation of H3 was only marginally affected by pairing with B4 VL, whereas for B4, the conformation of H3 was radically altered. The CDR loops in the B13 and B4 VL domains showed little change in conformation between the parent and chain swapped Fabs pointing to the key role of H3 in antigen-binding. It has been shown that all CDR loops of antibody variable regions, with the exception of H3, adopt one of a limited number of canonical structures ([@bib0040]) including bovine antibodies ([@bib0120]), though the number of structures available for analysis is very small compared to human and mouse. Of the six CDRs, H3 shows the greatest structural variability and in comparison with loops matched in length from non-antibody proteins, adopts a high proportion of unique conformations ([@bib0105]) and is generally considered to play the major role in antigen-binding interactions in most antibodies. The pairing of immunoglobulin light and heavy chains contributes to the diversification of antibody binding sites and the stability of the antibody structure ([@bib0025]). The preference in pairing of particular heavy and light chains was presumed to occur at random ([@bib0015]; [@bib0020], [@bib0045]). In humans, LCs from multiple sub-groups can pair with a single IGHV and in several cases use of an IGLV or IGKV combines with identical VDJ combinations without loss of antigen specificity ([@bib0050]). However, pairing preferences have been shown to exist in the human and mouse germline for a small proportion of gene segments, whilst others are promiscuous and show no pairing preferences ([@bib0065]). Specific pairings in cattle have been shown for the ultra-long H3 in cattle; these ultra-long sequences have unique structural properties and restricted LC pairings ([@bib0115]) that may have evolved specifically to provide a structural framework for supporting the ultra-long H3 ([@bib0155]). The limited diversity of the cattle VL germ line sequences leads to the hypothesis that the LC repertoire performs a greater structural role than in other species. In conclusion, the analysis of two bovine antibodies with H3 of more typical length (21--25 residues) presented here supports this view, and shows that the pairing of LC/HC can be crucial for the structure of the antibody paratope. Thus the LC variable domain (VL) is important for antigen recognition, contrary to what might have been expected from the limited VL repertoire in cattle but in line with the co-evolution of these protein domains. Further work on the structure of the B13 and B4 in complex with the RSV F protein will be required to determine whether this involves any direct interactions between the LC and the antigen. Accession numbers {#sec0075} ================= The atomic coordinates and structure factors of the Fabs have been deposited in the Protein Data Bank under the accession codes 6QN9, 4QN8, 6QN7 and 6QNA for B4, B13, B4\* and B13\* respectively. Appendix A. Supplementary data {#sec0085} ============================== The following is Supplementary data to this article: J.R. is supported by the Wellcome Trust (101122/Z/13/Z), and D.I.S. and E.E.F. by the UK Medical Research Council (MR/N00065X/1). The Wellcome Centre for Human Genetics is supported by the Wellcome Trust (grant 090532/Z/09/Z). R.J.O., N.R and J.E.N. acknowledge support from the UK Medical Research Council (MR/K018779/1). We thank the staff of beamlines I03, I04-1 and I24 at Diamond Light Source for help with data collection (BAG nos. mx10627 & mx14744). Supplementary material related to this article can be found, in the online version, at doi:<https://doi.org/10.1016/j.molimm.2019.04.026>. [^1]: Values in parentheses are for highest-resolution shell. [^2]: CC50 was not incorporated in the data processing programs used for this data set.

1. Introduction {#sec1-marinedrugs-14-00017} =============== 1.1. Marine Natural Products {#sec1dot1-marinedrugs-14-00017} ---------------------------- Dating back to ancient civilizations, natural products have played a vital role in drug discovery. Modern day drugs such as penicillin, morphine, and paclitaxel (Taxol™), which are used for the treatment of bacterial infections, pain, and cancer, respectively, are examples of natural products that have successfully progressed through the drug development pipeline. Natural products are abundant in several sources and are routinely extracted from plants, microorganisms, and animals. While terrestrial plants and terrestrial microbes continue to be the primary contributors to the natural product derived drug market, marine sources are an emerging resource that has been relatively unexplored until recent times \[[@B1-marinedrugs-14-00017]\]. Marine organisms are considered to yield superior natural products compared to terrestrial organisms in terms of novelty of structures and in producing potent bioactivities, due to the distinctive environmental conditions in which marine organisms live \[[@B2-marinedrugs-14-00017],[@B3-marinedrugs-14-00017],[@B4-marinedrugs-14-00017]\]. Factors such as predation, competition for space on highly populated coral reefs, and biochemical warfare between organisms have greatly contributed to the evolution of compounds with unique structures and potent biological effects \[[@B5-marinedrugs-14-00017]\]. 1.2. Cyanobacteria {#sec1dot2-marinedrugs-14-00017} ------------------ Microbial organisms in the marine environment have increasingly become one of the major focal points for investigations seeking to identify new chemical entities with novel structural backbones and diverse biological activities. In the 1970s, Professor Richard E. Moore at the University of Hawaii began exploring the chemistry of marine cyanobacteria \[[@B6-marinedrugs-14-00017]\]. Cyanobacteria are one of the oldest organisms on Earth, with an existence record of at least 2.7 billion years \[[@B7-marinedrugs-14-00017]\]. These prokaryote organisms are also known as blue green algae, cyanoprokaryotes, and cyanophytes due to their blue-green pigment, c-PC (c-phycocyanin) \[[@B8-marinedrugs-14-00017]\]. This pigment is used for photosynthesis and is considered to have played a crucial role in releasing oxygen into the primitive atmosphere \[[@B7-marinedrugs-14-00017],[@B9-marinedrugs-14-00017]\]. Cyanobacteria possess a broad geographical distribution, ranging from limnic and marine environments to terrestrial habitats \[[@B10-marinedrugs-14-00017],[@B11-marinedrugs-14-00017]\]. A subclass of marine natural compounds is produced by cyanobacteria. The most studied species of marine cyanobacteria include *Nostoc*, *Calothrix*, *Lyngbya*, and *Symploca* \[[@B12-marinedrugs-14-00017]\]. Cyanobacteria are known to produce potent toxins. Several studies have investigated the pharmaceutical and biotechnological potential of the secondary metabolites isolated from cyanobacteria. These studies revealed a wide range of potent pharmacological effects that include anti-inflammatory, antimalarial, antiprotozoal, antimicrobial, immunosuppressant, anticancer, anti-HIV, antibacterial, anticoagulant, antifungal, anti-tuberculosis, antiviral, and antitumor activities \[[@B12-marinedrugs-14-00017],[@B13-marinedrugs-14-00017],[@B14-marinedrugs-14-00017]\]. A number of compounds that were isolated from aquatic cyanobacteria have showed promise as anticancer leads. Examples of such promising compounds include apratoxin A, cryptophycin, dolastatin 10, and largazole ([Figure 1](#marinedrugs-14-00017-f001){ref-type="fig"}). {#marinedrugs-14-00017-f001} Apratoxin A is a cyclodepsipeptide derivative of the apratoxin family of cytotoxins isolated from a *Lyngbya* sp. collected from Guam \[[@B15-marinedrugs-14-00017],[@B16-marinedrugs-14-00017]\]. It induces G~1~ phase cell cycle arrest and apoptosis, and has exhibited IC~50~ values ranging from 0.36 to 0.52 nM *in vitro* against 60 human tumor cell lines. However, *in vivo* studies revealed only marginal activity against early stage adenocarcinoma \[[@B17-marinedrugs-14-00017],[@B18-marinedrugs-14-00017],[@B19-marinedrugs-14-00017]\]. Another class of highly potent anticancer agents produced by cyanobacteria is cryptophycins. For example, cryptophycin 1, which was isolated from *Nostoc* sp. GSV224, has an IC~50~ of 5 pg/mL against KB human nasopharyngeal cancer cells, and 3 pg/mL against LoVo human colorectal cancer cells \[[@B20-marinedrugs-14-00017]\]. Its mechanism of action is through the suppression of microtubule dynamics, thereby inhibiting cells in G~2~/M phase. Cryptophycin 52, a chemical analog of cryptophycin 1, entered clinical trials but produced only marginal activity \[[@B21-marinedrugs-14-00017]\]. Dolastatin 10 is cyanobacterial metabolite that was synthesized by Pettit *et al.* \[[@B22-marinedrugs-14-00017],[@B23-marinedrugs-14-00017]\] in the 1980s and then its origin was later confirmed when its direct isolation occurred from *Symploca* sp. Dolastatin 10 binds to tubulin on the rhizoxin-binding site and is an established antiproliferative agent that affects microtubule assembly, which leads to cell death during the G~2~/M phase \[[@B24-marinedrugs-14-00017]\]. In 2005, the efficacy and toxicity of dolastatin 10 were investigated in Phase II clinical trials in patients with advanced prostate cancer but it was discontinued due to the development of peripheral neuropathy in 40% of the patients \[[@B25-marinedrugs-14-00017]\]. Many research groups have sought after SAR studies of the synthetic analogs of this compound, which are now classified as auristatins. This group of compounds showed the most amount of promise as antibody drug conjugates. For example, brentuximab vedotin (adcetris, Seattle Genetics ) is now a FDA-approved drug against lymphoma. This compound is currently undergoing phase 3 studies to study its effect on cutaneous T-cell lymphoma, B-cell lymphomas, and mature T-cell lymphomas \[[@B26-marinedrugs-14-00017],[@B27-marinedrugs-14-00017]\]. The cyclic depsipeptide largazole, of the genus *Symploca*, is a marine natural product that contains a methylthiazoline linked to a thiazole, as well as a 3-hydroxy-7-mercaptohept-4-enoic acid unit, and a thioester, which had previously not been observed in marine cyanobacterial natural products. This compound has been shown to selectively target transformed over non-transformed cells \[[@B28-marinedrugs-14-00017],[@B29-marinedrugs-14-00017]\], through the inhibition of class I histone deacetylases (HDACs) \[[@B30-marinedrugs-14-00017],[@B31-marinedrugs-14-00017]\]. Largazole's interesting structure and biological activity have attracted strong interest from the synthetic chemistry community, which seeks to establish synthetic routes to largazole and to investigate its potential as a cancer therapeutic \[[@B32-marinedrugs-14-00017],[@B33-marinedrugs-14-00017]\]. Thus, cyanobacterial metabolites have the potential for expanded utilization in drug discovery. Despite their potent biological activities, very few cyanobacterial compounds have entered clinical trials; one of the reasons is the complexity of synthesis of the natural products. The focus of this review is the natural products called calothrixins, with an emphasis on their synthesis and bioactivities. 1.3. Calothrixins {#sec1dot3-marinedrugs-14-00017} ----------------- Calothrixins A and B ([Figure 2](#marinedrugs-14-00017-f002){ref-type="fig"}) are two cyanobacterial metabolites that were first isolated from *Calothrix* in 1999 by Rickards *et al.* \[[@B34-marinedrugs-14-00017]\]. Briefly, lyophilized cells of *Calothrix* strains were extracted with dimethyl sulfoxide (DMSO) and then with ethyl acetate (AcOEt) using Soxhlet extraction conditions. These extracts were fractionalized using a combination of bioassays, differential solubility, and chromatography, which yielded the relatively insoluble calothrixin A (1a) and its more soluble co-metabolite, calothrixin B (1b). Both structures were elucidated by Electron Impact Mass Spectra (EIMS), ^1^H, ^13^C and ^1^H--^1^H Correlation Spectroscopy (COSY) NMRs. Calothrixins possess an unique indolo\[3,2-*j*\]phenanthridine framework with an assembly of quinoline, quinone, and indole pharmacophores. Calothrixin B is generally known to be the neutral analog while calothrixin A is known to be its *N*-oxide analog. {#marinedrugs-14-00017-f002} These scaffolds have attracted the interests of organic chemists and biologists alike due to their unique structures and the promise they hold as lead compounds for cancer drug discovery. Herein, we summarize the recent advances in synthetic work on calothrixins and their analogs. 2. Synthesis of Calothrixins {#sec2-marinedrugs-14-00017} ============================ Due to the structural complexity and drug discovery potential, organic and medicinal chemists around the world have embarked on developing efficient synthetic methodologies to produce these natural products. Earlier synthesis of calothrixins was reviewed in 2009 by Satoshi Hibino *et al.* \[[@B35-marinedrugs-14-00017]\]. There has been a lot of progress made in the synthesis of calothrixins and their analogs since then. Herein, the discussion will focus on the progress achieved within the past six years in developing new synthetic routes to achieve the large-scale production of calothrixins. The calothrixin scaffold consists of five rings (A--E), as shown in [Figure 2](#marinedrugs-14-00017-f002){ref-type="fig"}. Multiple synthetic strategies have been used to synthesize calothrixins starting from suitably substituted indole, quinoline, or carbazole derivatives. The synthetic routes for calothrixins are categorized into three main groups based on the strategy of the last ring closure step in the construction of the calothrixin scaffold (1) ring B closure; (2) ring C closure; and (3) ring D closure ([Table 1](#marinedrugs-14-00017-t001){ref-type="table"}). No syntheses have been reported in which either ring A or ring E are constructed as the last ring closure step. Recent reports on the synthesis of calothrixins and the key reactions involved in these syntheses are summarized in [Table 1](#marinedrugs-14-00017-t001){ref-type="table"}. marinedrugs-14-00017-t001_Table 1 ###### Summary of reported syntheses of calothrixin B since 2009. Research Group Year Ring Closure Key Step Reference ------------------------- ------------- -------------- --------------------------------------------------------------------------- ------------------------------------------------------------- Velu *et al.* 2014 B Mn(OAc)~3~ mediated oxidative free radical reaction \[[@B36-marinedrugs-14-00017]\] Ishikura *et al.* 2011 & 2012 C Palladium catalyzed tandem cyclization/cross-coupling \[[@B37-marinedrugs-14-00017],[@B38-marinedrugs-14-00017]\] Dethe *et al.* 2014 C LTA mediated rearrangement of *o*-hydroxy aryl hydrazone \[[@B39-marinedrugs-14-00017]\] Nagarajan *et al.* 2014 C Friedel-Crafts hydroxyalkylation and directed *o*-lithiation \[[@B40-marinedrugs-14-00017]\] Mal *et al.* 2014 C The anionic annulation of MOM-protected furoindolone \[[@B41-marinedrugs-14-00017]\] Kusurkar *et al.* 2012 D Two one pot reaction sequences: a nucleophilic substitution and reduction \[[@B42-marinedrugs-14-00017]\] Nagarajan *et al.* 2013 D Pd catalyzed intramolecular cross-coupling reaction via C--H activation \[[@B43-marinedrugs-14-00017]\] Mohanakrishnan *et al.* 2014 D Electrocyclization of 2-nitroarylvinyl-3-phenylsulfonylvinylindole \[[@B44-marinedrugs-14-00017]\] Kumar *et al.* 2014 D Pd mediated intramolecular C-X/C-H cross coupling reactions \[[@B45-marinedrugs-14-00017]\] Hibino *et al.* 2012 D Allene mediated electrocyclic reaction of the 6π-electron system \[[@B46-marinedrugs-14-00017]\] Mohanakrishnan *et al.* 2013 C & D FeCl~3~ mediated domino reaction of enamines \[[@B47-marinedrugs-14-00017]\] 2.1. Formation of Indole (Rings A and B) as the Last Step in the Construction of the Indolo\[3,2-j\]Phenanthridine Framework {#sec2dot1-marinedrugs-14-00017} ---------------------------------------------------------------------------------------------------------------------------- ### Velu's Synthesis In 2014, Velu *et al.* \[[@B36-marinedrugs-14-00017]\] reported a synthesis of calothrixins in which the indole ring is formed last to construct the five-ring scaffold. This approach relies on the construction of the indole ring (rings A and B) on a phenanthridine dione (ring C, D, and E) via a novel oxidative free radical reaction mediated by manganese triacetate Mn(OAc)~3~, as outlined in [Scheme 1](#marinedrugs-14-00017-f003){ref-type="scheme"}. The method of Mn(OAc)~3~ mediated oxidative reaction of 2-cyclohexenone with quinones was originally developed by Chuang *et al.* \[[@B48-marinedrugs-14-00017],[@B49-marinedrugs-14-00017],[@B50-marinedrugs-14-00017]\]. None of the existing reports of calothrixin synthesis utilizes the late-stage indole construction strategy. Through this methodology, calothrixin B was synthesized in seven steps, with an overall yield of 19% starting from 2,4,5-trimethoxyindole. {#marinedrugs-14-00017-f003} 2.2. Ring C Closure as the Last Step in the Construction of the Indolo\[3,2-j\]Phenanthridine Framework {#sec2dot2-marinedrugs-14-00017} ------------------------------------------------------------------------------------------------------- A number of calothrixin syntheses have been reported where the ring C closure was used as the last step in the construction of calothrixin scaffold. This includes the two syntheses reported by Ishikura *et al.* in 2011 and 2012 \[[@B37-marinedrugs-14-00017],[@B38-marinedrugs-14-00017]\], and the three contributions made by Dethe *et al.* \[[@B39-marinedrugs-14-00017]\], Nagarajan *et al.* \[[@B40-marinedrugs-14-00017]\], and Mal *et al.* \[[@B41-marinedrugs-14-00017]\] in 2014. ### 2.2.1. Ishikura's Synthesis {#sec2dot2dot1-marinedrugs-14-00017} Ishikura *et al.* \[[@B37-marinedrugs-14-00017],[@B38-marinedrugs-14-00017]\] was one of the first groups that designed a synthetic route to calothrixins in which ring C was cyclized last as shown in [Scheme 2](#marinedrugs-14-00017-f004){ref-type="scheme"}. In their ongoing studies of trialkyl(indol-2-yl)borates, they previously found that indolylborates show high reactivity in palladium-catalyzed cross-coupling reactions, such as carbonylative cross-coupling and tandem cyclization/cross-coupling reactions. This new approach for the synthesis of calothrixins A and B was demonstrated through a palladium-catalyzed cross-coupling reaction of 1-methoxyindolylborate (generated *in situ* from 1-methoxyindole by treatment with n-BuLi and BEt~3~) with the intermediate compound **13** to form the compound **15**. Additional novelty of this synthetic route is the strategic use of Cu(OTf)~2~.toluene complex for the 6π-elecrocyclization of the hexatriene intermediate (**16**) to cyclize the ring C to form the compound **17**. Ishikura's synthesis yielded calothrixin B in nine steps with an overall yield of 9% starting from 2-iodoaniline. {#marinedrugs-14-00017-f004} ### 2.2.2. Dethe's Synthesis {#sec2dot2dot2-marinedrugs-14-00017} Dethe *et al.* \[[@B39-marinedrugs-14-00017]\] reported a concise five-step total synthesis of calothrixin B using LTA-mediated rearrangement of a suitable *o*-hydroxy aryl hydrazone into the corresponding quinone as the key step, as outlined in [Scheme 3](#marinedrugs-14-00017-f005){ref-type="scheme"}. The synthesis began with the coupling of *N*-PMB (p-methoxy benzyl) protected indolyl hyrazide (**21**) and the quinolinone derivative (**22**) to form the intermediate hydrazone **23**. The hydrazone (**23**) was treated with Pb(OAc)~4~ to undergo the LTA-mediated oxidative rearrangement followed by a BF~3~.OEt~2~-mediated cyclization to form the PMB-protected calothrixin B (**25**). This synthesis is short and high yielding. The overall yield of the calothrixin B synthesis from ethyl indole-2-carboxylate (**20**) is 39% for five steps. {#marinedrugs-14-00017-f005} ### 2.2.3. Nagarajan's Synthesis {#sec2dot2dot3-marinedrugs-14-00017} Nagarajan *et al.* \[[@B40-marinedrugs-14-00017]\] outlined a strategy for the synthesis of calothrixin B featuring a directed *o*-metalation reaction as a key step, as outlined in [Scheme 4](#marinedrugs-14-00017-f006){ref-type="scheme"}. The synthesis began with the coupling of the commercially available reagents, ethyl indole-2-carboxylate (**20**) and quinoline-3-carboxaldehyde (**26**) in the presence of TMG in MeOH followed by the oxidation of the intermediate product with Dess-Martin periodinane (DMP) in CH~2~Cl~2~/AcOH. This two-step process yielded compound **27**, which was then cyclized by an intramolecular directed *o*-lithiation reaction using lithium tetramethylpiperidide (LiTMP) to form calothrixin B in 48% yield. This synthetic methodology offers a three-step route to calothrixin B from commercially available reagents in an overall yield of 38%. {#marinedrugs-14-00017-f006} ### 2.2.4. Mal's Synthesis {#sec2dot2dot4-marinedrugs-14-00017} Mal *et al.* \[[@B41-marinedrugs-14-00017]\] reported a two-step synthesis of calothrixin B from the *N*-protected indololactone (**28**) as outlined in [Scheme 5](#marinedrugs-14-00017-f007){ref-type="scheme"}. This reagent was prepared in three simple steps from commercially available α-acetobutyrolactone \[[@B51-marinedrugs-14-00017]\]. Treatment of compound **28** with 4-bromoquinoline (**29**) in the presence of lithium diisopropylamide (LDA) in tetrahydrofuran (THF) at −78 °C afforded an inseparable mixture of compounds **30a** and **30b**. However, after the removal of methoxymethylene (MOM) group using HCl, it gave a mixture of calothrixin B (**1b**) and its regioisomer (**31b**) that could be easily separated in the ratio of 40% and 6%, respectively. This synthetic methodology afforded calothrixin B with an overall yield of 47% for two steps starting from compound **28**. {#marinedrugs-14-00017-f007} 2.3. Ring D Closure as the Last Step in the Construction of the Indolo\[3,2-j\]Phenanthridine Framework {#sec2dot3-marinedrugs-14-00017} ------------------------------------------------------------------------------------------------------- Another strategy for the calothrixin synthesis involves cyclizing ring D as the last step in the construction of calothrixin scaffold. A number of contributions have been made to this category of syntheses. These include four independent publications by Kusurkar *et al.* in 2012 \[[@B42-marinedrugs-14-00017]\], Nagarajan *et al.* in 2013 \[[@B43-marinedrugs-14-00017]\], Mohanakrishnan *et al.* in 2014 \[[@B44-marinedrugs-14-00017]\], Kumar *et al.* in 2014 \[[@B45-marinedrugs-14-00017]\], and Hibino *et al.* in 2012 \[[@B46-marinedrugs-14-00017]\]. ### 2.3.1. Kusurkar's Synthesis {#sec2dot3dot1-marinedrugs-14-00017} Kusurkar *et al.* \[[@B42-marinedrugs-14-00017]\] have published a synthesis of calothrixins starting from 4-hydroxy carbazole as outlined in [Scheme 6](#marinedrugs-14-00017-f008){ref-type="scheme"}. The novelty of their synthetic route is the unprecedented use of DMF-NaOMe as a reagent for the reduction of the aldehyde group and the use of a high-yielding Pd-catalyzed coupling reaction to construct the ring D of the pentacyclic system. For example, the treatment of compound **35** with NaOMe in dry dimethyl formamide (DMF) and CuI at 120 °C resulted in the substitution of both bromine and benzyloxy groups with methoxy groups with concomitant reduction of the aldehyde resulting in the formation of the compound **36**. Pd-catalyzed cyclization of ring D of the scaffold to form compound **39** is another novel key step in this synthesis. The overall yield for Kusurkar's synthesis of calothrixin B was 25% for nine steps. {#marinedrugs-14-00017-f008} ### 2.3.2. Nagarajan's Synthesis {#sec2dot3dot2-marinedrugs-14-00017} In addition to developing a synthetic route for calothrixins that involved the closure of ring C in the last step, Nagarajan *et al.* \[[@B43-marinedrugs-14-00017]\] has also accomplished a synthesis in which ring D was closed last in the preparation of the five-membered ring system, followed by the installation of the necessary substitutions to produce calothrixin B as shown in [Scheme 7](#marinedrugs-14-00017-f009){ref-type="scheme"}. This synthesis featured a prominent Pd-catalyzed intramolecular cross-coupling reaction via C-H activation using an appropriately substituted carbazole derivative to construct the indolophenanthridine core ring system of calothrixins. This is the first reported synthesis of calothrixin pentacycles without any protection on the indole N atom. The overall yield for calothrixin B in this Nagarajan's synthesis is 35% for five steps starting from 4-methoxycarbazole. {#marinedrugs-14-00017-f009} ### 2.3.3. Mohanakrishnan's Synthesis {#sec2dot3dot3-marinedrugs-14-00017} Mohanakrishnan *et al.* have reported a linear synthesis of calothrixin B using the thermal electro cyclization of 2-nitroarylvinyl-3-phenylsulfonylvinylindole as the key step as outlined in [Scheme 8](#marinedrugs-14-00017-f010){ref-type="scheme"} \[[@B44-marinedrugs-14-00017]\]. In this key step, the 2,3-divinylidole intermediate generated from compound **52** by treatment with Me~2~SO~4~ was subjected to thermal cyclization in refluxing xylenes to afford the carbazole derivative **53**. Mohanakrishnan's total synthesis thus afforded calothrixin B from 2-methylindole in 14 steps with an overall yield of about 8%. {#marinedrugs-14-00017-f010} ### 2.3.4. Kumar's Synthesis {#sec2dot3dot4-marinedrugs-14-00017} Kumar *et al.* \[[@B45-marinedrugs-14-00017]\] reported the total synthesis of calothrixin B via multiple palladium-mediated C-X/C-H cross intramolecular coupling reactions as outlined in [Scheme 9](#marinedrugs-14-00017-f011){ref-type="scheme"}. The salient feature of this synthesis is the two intramolecular C-X/C-H cross coupling reactions of the intermediates **60** to form compound **61** and **63** to form the compound **64**. The intermediate **60** was cyclized under intramolecular cross coupling reaction condition using Pd(OAc)~2~, PCy~3~, and JohnPhos to afford the carbazole derivative **61**. The cyclization of **63** was carried out using Pd(OAc)~2~, PCy~3~, and K~2~CO~3~ to afford the compound **64.** Kumar's methodology thus accomplished the synthesis of calothrixin B in 7 steps starting from 2,5-dimethoxybenzaldehyde in an overall yield of 50%. {#marinedrugs-14-00017-f011} Compound **60** in Kumar's synthesis has also been converted to the key intermediate **64** in three steps as shown in [Scheme 10](#marinedrugs-14-00017-f012){ref-type="scheme"}. In this synthesis, rings B and D of the calothrixin scaffold were cyclized in one step by intramolecular C-X/C-H cross coupling reaction using Pd(OAc)~2~, PCy~3~, and JohnPhos in the presence of K~2~CO~3~ in DMF to afford the key intermediate **64**. {#marinedrugs-14-00017-f012} ### 2.3.5. Hibino's Synthesis {#sec2dot3dot5-marinedrugs-14-00017} Hibino *et al.* \[[@B46-marinedrugs-14-00017]\] synthesized calothrixin B and other *N*-alkyl-calothrixins using a biomimetic approach featuring a key allene-mediated electrocyclic reaction of the 6π-electron system, as outlined in [Scheme 11](#marinedrugs-14-00017-f013){ref-type="scheme"}. The synthesis of the key intermediate, indolo\[2,3-*a*\]carbazole (**72**), was carried out by an allene-mediated electrocyclic reaction of MOM-protected compound **71** in the presence of tBuOK in tBuOH-THF. This electrocyclization involved the two \[*b*\]-bonds of indole units and the allene double bond generated *in situ*. The overall yield for Hibino's synthesis of calothrixin B from 2-bromoindole-3-carboxaldehyde is 22% for nine steps. {#marinedrugs-14-00017-f013} 2.4. Simultaneous Closure of Rings C and D {#sec2dot4-marinedrugs-14-00017} ------------------------------------------ Only one methodology of calothrixin synthesis has been developed in which multiple rings were cyclized in a one-pot synthesis. This report was published by Mohanakrisnan *et al.* \[[@B47-marinedrugs-14-00017]\] in 2013, prior to their report in 2014 \[[@B44-marinedrugs-14-00017]\] in which ring D was the last ring closure step. ### Mohanakrishnan's Synthesis Mohanakrishnan *et al.* \[[@B47-marinedrugs-14-00017]\] reported another novel synthesis of calothrixin B involving a key FeCl~3~ mediated domino reaction of an enamine derivative, as outlined in [Scheme 12](#marinedrugs-14-00017-f014){ref-type="scheme"}. In this key step, FeCl~3~ was used to cyclize the key intermediate **79** to afford calothrixin B. Both rings C and D were created in a sigle step in this synthesis. The overall yield of calothrixin B from the compound **75** is 45% for six steps. {#marinedrugs-14-00017-f014} 3. Bioactivities of Calothrixins {#sec3-marinedrugs-14-00017} ================================ The most therapeutically relevant biological activity reported for marine cyanobacterial metabolites is their cytotoxicity. While some antimicrobial studies were also conducted, calothrixins were majorly pursued to serve as novel anticancer molecules. Herein, we summarize all bioactivities conducted on calothrixins and their analogs. Following their isolation, calothrixins quickly gained recognition due to their anticancer activity as preliminary studies revealed a high potency against the human cervical cancer cell line, HeLa cells. Since then, calothrixins have been evaluated against several different cell lines and have been studied for their mode of action. Calothrixins were also evaluated for their antiparasitic activity. A summary of their bioactivities is found below. 3.1. Antiparasitic Activity {#sec3dot1-marinedrugs-14-00017} --------------------------- Along with their discovery in 1999, Smith *et al.* \[[@B34-marinedrugs-14-00017]\] also reported that the cell extracts from cyanobacteria *Calothrix* were found to inhibit the growth of malarial strains. The two calothrixins, calothrixin A and B, were evaluated against the chloroquine (QC)-resistant malarial strain, *P. falciparum* FCR-3 strain (QC-sensitive), and were found to be effective in a dose-dependent manner. The IC~50~ values were observed to be 58 nM and 180 nM, respectively, compared with 83 nM IC~50~ value for chloroquine in the same assay \[[@B34-marinedrugs-14-00017]\]. In an attempt to find a better lead against malaria, Hibino *et al.* \[[@B46-marinedrugs-14-00017]\] synthesized and evaluated several *N*-alkyl calothrixin analogs for their activity against the chloroquine-resistant strain ([Table 2](#marinedrugs-14-00017-t002){ref-type="table"}). The results for antimalarial activity were compared against chloroquine. Calothrixin B was observed to be the most active (IC~50~ = 120 nM), concurring with the findings of Rickards *et al.* \[[@B34-marinedrugs-14-00017]\]. Calothrixin A was found to be almost equally active (IC~50~ = 185 nM ) as calothrixin B. Although all compounds showed antimalarial activity, substitution of the indole nitrogen atom with various alkyl groups led to a decrease in activity. Among the analogs, 2-hydroxyethyl group showed slightly better activity in comparison to the other alkyl substitutions. In addition to its antimalarial activities, Smith *et al.* \[[@B52-marinedrugs-14-00017],[@B53-marinedrugs-14-00017]\] reported the antibacterial activity of calothrixin A against *Bacillus subtilis* 168, where it was found to be inhibiting the bacterial growth in a dose-dependent manner without any cell lysis. A complete growth inhibition was achieved at 16 µM concentration. marinedrugs-14-00017-t002_Table 2 ###### Antimalarial activity calothrixin analogs against FCR-3 strain. Compound IC~50~ against FCR-3 Strain (nM) Compound IC~50~ against FCR-3 Strain (nM) ------------------------------------ ---------------------------------- ------------------------------------ ----------------------------------  185  120  380  490  380  220  320  640  250  330  180 3.2. Anticancer Activity {#sec3dot2-marinedrugs-14-00017} ------------------------ Along with their antimalarial activity, Rickards *et al.* \[[@B34-marinedrugs-14-00017]\] reported calothrixins to be potent against HeLa cells. The corresponding IC~50~ values for calothrixin A and B against the human cervical cancer cell line, HeLa cells, were observed to be 40 nM and 350 nM, respectively. Calothrixins showing similar inhibitory effects against cancer cell lines and malarial strains indicated that they may share a common mode of action. Following this report, in 2003, Waring *et al.* \[[@B54-marinedrugs-14-00017]\] studied the effect of calothrixin A on apoptosis in human Jurkat cancer cells. A well-known inducer of apoptosis, menadione, was used as a positive control to compare the effects of calothrixin A \[[@B55-marinedrugs-14-00017]\]. Calothrixin A was observed to induce cell death via apoptosis in a time- and concentration-dependent manner, and was found to be more potent than menadione in anti-proliferative activity, with IC~50~ values of 1.6 µM and 4.7 µM, respectively ([Table 3](#marinedrugs-14-00017-t003){ref-type="table"}). IC~50~ values for calothrixin A- and menadione-induced apoptosis were 0.6 µM and 12 µM respectively. Both calothrixin A and menadione were observed to induce cell cycle arrest in the G~2~/M phase, but the concentration required for calothrixin A was much lower than for menadione. The cell cycle arrest in the G~2~/M phase indicated intracellular DNA damage, which was further supported by direct DNA damage observed in cell-free experiments. Both calothrixin A and menadione were found to be redox active through the observation of additional oxygen intake when added to reductant dithiothreitol (DTT, 2 mM at pH 8.0). Quinones can redox cycle by one or two electron transfers, generating reactive oxygen species (ROS), by which they can damage DNA and induce apoptosis \[[@B56-marinedrugs-14-00017]\]. It was postulated that the ring structure of calothrixin might act as a DNA intercalator that might be responsible for its anticancer activity. In addition, calothrixin A was observed to be cleaving DNA, though less effectively than menadione. Menadione caused significant DNA damage at 10 µM after 60 min incubation, whereas calothrixin A exhibited the same effect at 500 µM. marinedrugs-14-00017-t003_Table 3 ###### Cell cytotoxicities and apoptotic activities of calothrixin A and menadione. Compound Structure IC~50~ for Cytotoxicity IC~50~ for Apoptosis --------------- ------------------------------------ ------------------------- ---------------------- Calothrixin A  1.6 µM 0.6 µM Menadione  4.7 µM 12 µM One of the known modes of action for anticancer activities of quinones is by the generation of ROS through redox cycling \[[@B56-marinedrugs-14-00017]\]. In 2004, Wilkes *et al.* \[[@B57-marinedrugs-14-00017]\] carried out a comparative study of bioactivities of calothrixins and some structurally related quinones. This was the first structure--activity relationship study done for calothrixins. The cytotoxic effect of various quinones was evaluated against human cervical cancer cells, HeLa cells, using 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay ([Table 4](#marinedrugs-14-00017-t004){ref-type="table"}). Calothrixin A and calothrixin B showed cytotoxicity against HeLa cells with EC~50~ values of 0.12 µM and 0.24 µM, respectively. Calothrixin B and ellipticine quinone, which differ by the ring E, showed comparable activities, indicating that there might be only a limited role for the ring E in dictating the bioactivity of calothrixins. Introducing the MOM group on the indole nitrogen seemed to decrease the potency of both calothrixin B and ellipticine quinone with EC~50~ values of 0.42 µM and 0.37 µM, respectively. Further, replacement of the heteroaromatic ring-D with a carbocyclic aromatic ring in ellipticine and its *N*-methoxymethyl analog also led to a decrease in potency, benzocarbazoledione (0.43 µM), and *N*-MOM-benzocarbazoledione (1.6 µM). Menadione showed activity at 3.7 µM. Absence of activity for uncyclized precursors of benzocarbazoledione and *N*-MOM-benzocarbazoledione indicated the importance of a tetracyclic quinone ring structure. In case of all the quinones with EC~50~ \< 1.6 µM, the quinones were observed to be reduced to their respective semiquinones, though no direct correlation was observed between the reduction potentials and bioactivity. marinedrugs-14-00017-t004_Table 4 ###### Cytotoxicity of quinones against HeLa cells. Compound Structure EC~50~ (µM) HeLa Cells ----------------------------------------------------- ------------------------------------ ------------------------ Calothrixin A  0.12 ± 0.01 Calothrixin B  0.24 ± 0.04 *N*-MOM-calothrixin B  0.42 ± 0.02 Ellipticine quinone  0.15 ± 0.09 *N*-MOM-ellipticine quinone  0.37 ± 0.08 Benzocarbazoledione  0.43 ± 0.01 *N*-MOM-Benzocarbazoledione  1.6 ± 1.0 Menadione  3.7 ± 0.3 Uncyclized precursor to benzocarbazoledione  \>50 Uncyclized precursor to *N*-MOM-benzocarbazoledione  \>50 To study the involvement of the ring structure and the mechanism of action in the inhibitory effects, the same group \[[@B58-marinedrugs-14-00017]\] synthesized and analyzed the activity of various simple quinone analogs of calothrixin B ([Table 5](#marinedrugs-14-00017-t005){ref-type="table"}) in 2007. The compounds were evaluated for their selectivity and antiproliferative activity, using an MTT assay, against three different cell lines: human cervical cancer (HeLa) cells, murine P388 macrophage cancer cells, and simian non-cancerous CV-1 cells. In this study, calothrixin B was found to be the most potent against HeLa cells (0.25 µM) followed by indolophenanthrene-7,13-dione analog (1.5 µM), indicating the importance of nitrogen in ring D. The activity was further decreased with the deletion of ring E of this analog to form benzoncarbzoledione (1.8 µM). The rest of the analogs showed moderate activity ranging from 7 to 13 µM, while the carbazole-1,4-dione analog was found to be inactive. The results were not as consistent for the P388 cell line, where murrayaquinone (2.3 µM) and 2-methylcarbazoledione (1 µM) were found to be more active compared to calothrixin B (9 µM) or other analogs, whereas calothrixin B and 2-methylcarbazoledione showed similar activity with 2.4 µM and 1.7 µM, respectively, against non-cancerous cell line CV-1. The rest of the analogs had either very low activity or no measurable activity. These studies indicated the importance of rings A--D in the activity of calothrixins. Also, a selective usage of calothrixin B against human cervical cancer cells was indicated by the 10-fold higher effectiveness against HeLa cells (0.25 µM) as compared to CV-1 (2.4 µM) and 38 times more compared to P388 (9 µM). The quinones containing tetra- and penta cyclic systems (calothrixin-B, indolophenanthrene-7,13-dione and benzocarbazole-1,4-dione) were found to be more active against HeLa cells as compared to p388 and CV-1 cells. However, the trend was reversed in the quinones containing bi- and tricylic systems (murrayaquinone, 2-methylcarbazoledione, and isoquinoline-5,8-dione). In 2009, another group, Hecht *et al.* \[[@B59-marinedrugs-14-00017]\], published a study to determine the mechanism of action and the stage of the cell cycle that is targeted by the calothrixins. Calothrixin A and B and the *N*-methylated derivative were synthesized and tested against CEM leukemia cells to measure their cytotoxicity using an MTT assay. The activity of calothrixin analogs were compared to camptothecin, which is known to cause irreversible DNA damage during the S phase. Delay in S phase due to DNA damage is associated with a block in replication. The IC~50~ value for calothrixin A was found to be five-fold higher than campothecin and the other two analogs were observed to be much less potent, with *N*-methylcalothrixin B being the least at 5 µM compared to calothrixin B at 1 µM ([Table 6](#marinedrugs-14-00017-t006){ref-type="table"}). marinedrugs-14-00017-t005_Table 5 ###### Antiproliferative activity of calothrixin B and analogs in MTT assay against HeLa cells, P388 cells, and CV-1 cells. Compound Structure EC~50~ (µM) HeLa Cells EC~50~ (µM) P388 Cells EC~50~ (µM) CV-1 Cells ------------------------------- ------------------------------------ ------------------------ ------------------------ ------------------------ Calothrixin B  0.25 ± 0.05 9 ± 2 2.4 ± 0.7 Indolophenanthrene-7,13-dione  1.5 ± 0.3 \>50 \>50 Benzocarbzoledione  1.8 ± 0.1 \>50 \>50 Carbazole-1,4-dione  \>50 \>50 \>50 Murrayaquinone  13 ± 1 2.3 ± 0.3 10 ± 2 2-Methylcarbazoledione  7 ± 1 1.0 ± 0.1 1.7 ± 0.4 Isoquinoline-5,8-dione  12 ± 1 9 ± 2 \>50 The effect on the cell cycle was studied for these compounds using a mitotic inhibitor nocodazole, which blocks re-entry of cells into the G~1~ phase. Calothrixin B was observed to arrest the G~1~ phase at 0.1 μM concentrations, whereas calothrixin A and *N*-methylcalothrixn B showed no effects at the same concentration. At higher concentrations, calothrixin A and *N*-methylcalothrixn B led to cell accumulation in S and G~2~/M phase. Compared to camptothecin, these effects were found to be readily reversible. Calothrixins were also evaluated for their activity against topoisomerase I. The effects of calothrixins were studied keeping camptothecin as a positive control, which is a known topoisomerase I poison \[[@B60-marinedrugs-14-00017]\]. It was observed that calothrixins A, B, and *N*-methylcalothrixin B were capable of stabilizing the covalent topoisomerase I/DNA complex at 18%, 13%, and 11%, respectively, compared to 100% at 5 µM concentration of camptothecin. Calothrixins and their analogs were observed to be affecting different stages of cell cycle in a reversible manner, leading to low micromolar range cytotoxicities. Thus, calothrixins have shown a wide array of activity including antimalarial, anticancer, and antibacterial. The modes of action of calothrixins' bioactivities have not been fully deciphered yet. marinedrugs-14-00017-t006_Table 6 ###### Cytotoxicity effects against CEM leukemia cell line and topoisomerase I inhibition. Compound Structure IC~50~ (µM) in CEM Leukemia Cells \% Topo I DNA Cleavage at 5 µM \% NaCl Induced Reversibility at 5 µM ------------------------ ------------------------------------ ----------------------------------- -------------------------------- --------------------------------------- Camptothecin  0.04 ± 0.01 100 100 Calothrixin A  0.20 ± 0.02 18 17 Calothrixin B  1.05 ± 0.30 13 16 *N*-methycalothrixin B  5.13 ± 0.72 11 13 4. Conclusions {#sec4-marinedrugs-14-00017} ============== While the medicinal and biosynthetic potential of terrestrial plants and microbes is fairly well studied, comparatively little is known regarding the chemistry and biological activity of organisms in the marine environment. A unique group of oxygenic photosynthetic bacteria known as cyanobacteria populate diverse habitats throughout the world. Their potential as a good source of new therapeutic lead compounds has been realized during the last few decades. Calothrixins, which are cyanobacterial metabolites, have demonstrated a diverse range of bioactivities that include antimalarial, anticancer, and antibacterial properties. They have been observed to target various aspects of RNA synthesis in bacteria. Further investigation of the exact mechanism for their bioactivity is still required for many analogs, which will be beneficial for the ongoing development and lead optimization. Several research groups have developed synthetic routes to obtain these natural products. This review emphasized the synthetic progress accomplished within the past six years in developing new synthetic routes. Since 2009, 11 novel syntheses have been published to improve upon the efficiency in the production of calothrixins. The number of steps in the various synthetic strategies ranges from 2 to 14, while the overall yield to construct calothrixin B ranges from 8% to 50%. Thus, significant progress has been made in the last decade in attaining more efficient synthesis of calothrixins, which may aid to establish calothrixins as a potential anti-cancer candidate in the future. We thank the current and former members of Velu lab for their contributions to the research projects that were cited in this review. Sadanandan E. Velu was supported by a Breast Spore pilot grant and a Collaborative Programmatic Development grant from University of Alabama at Birmingham Comprehensive Cancer Center. He was also supported by grant number 1UL1RR025777 from the NIH National Center for Research Resources. Sadanandan E. Velu conceived the idea, Su Xu, Bhavitavya Nijampatnam and Shilpa Dutta wrote the manuscript. The authors declare no conflicts of interest. AcCl : Acetyl chloride Ac~2~O : Acetic anhydride AcOEt : Ethyl acetate AcOH : Acetic acid AIBN : Azobisisobutyronitrile AlCl~3~ : Aluminum chloride BEt~3~ : Triethylborane or triethylboron BF~3~.OEt~2~ : Boron trifluoride diethyl etherate CAN : Ceric ammonium nitrate CCl~4~ : Carbon tetrachloride CH~2~Cl~2~ : Dichloromethane CH~3~CN : Acetonitrile COSY : Correlation spectroscopy c-PC : c-Phycocyanin CSA : Camphorsulfonic acid CuI : Copper(I) iodide (CuOTf)~2~.toluene : Trifluoromethanesulfonate toluene complex DMAP : 4-Dimethylaminopyridine DMF : Dimethylformamide DMP : Dess-Martin periodinane DMSO : Dimethyl sulfoxide DoM : Directed *o*-metalation DTT : Dithiothreitol Et~2~NH : Diethylamine Et~3~SiH : Triethylsilane FeCl~3~ : Iron(III) chloride HCO~2~H : Formic acid HDACs : Histone deacetylases HIV : Human immunodeficiency virus H~2~O~2~ : Hydrogen peroxide K~2~CO~3~ : Potassium carbonate LDA : Lithium diisopropylamide LiTMP : Lithium tetramethylpiperidide LTA : Lead tetracetate MeOH : Methanol Mn(OAc)~3~ : Manganese triacetate MOM : Methoxymethyl protecting group MTT : 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide NaCNBH~3~ : Sodium cyanoborohydride NaH : Sodium hydride NaOH : Sodium hydroxide NaOMe : Sodium methoxide NBS : *N*-bromosuccinimide *n*-BuLi : *n*-Butyllithium NH~2~NH~2~.H~2~O : Hydrazine hydrate solution Pb(OAc)~4~ : Lead(IV) acetate PCC : Pyridinium chlorochromate Pd/C : Palladium on carbon PdCl~2~(PPh~3~)~2~ : Bis(triphenylphosphine)palladium(II) dichloride Pd~2~(dba)~3~ : Tris(dibenzylideneacetone)dipalladium(0) Pd(dppf)Cl~2~ : \[1,1′-Bis(diphenylphosphino)ferrocene\] dichloropalladium(II) Pd(OAc)~2~ : Palladium acetate Pd(TFA)~2~ : palladium(II) trifluoroacetate PhCN : Benzonitrile (PhSe)~2~ : Diphenyl diselenide PMB : *p*-Methoxybenzyl PMBCl : *p*-Methoxybenzyl chloride P(*o*-tol)~3~ : Tri(*o*-tolyl)phosphine POCl~3~ : Phosphoryl chloride PPh~3~ : Triphenylphosphine *p*-TSA : *p*-Toluenesulfonic acid PTT : Polytrimethylene terephthalate rac-BINAP : Racemic 2,2′-bis(diphenylphosphino)-1,1′-binaphthalene ROS : Reactive oxygen species TBAD : Bis(tetrabutylammonium) dichromate TBHP : *Tert*-butyl hydroperoxide Tf~2~O : Trifluoromethanesulfonic anhydride THF : Tetrahydrofuran THP : Tetrahydropyran TMG : 1,1,3,3-Tetramethylguanidine TMP : 2,2,6,6-Tetramethylpiperidine TMSI : Trimethylsilyl iodide

1. Introduction {#sec1-jof-03-00035} =============== The encapsulated, yeast-like fungi of the genus *Cryptococcus* are prevalent throughout the environment worldwide. The most common species that cause disease in humans are *Cryptococcus neoformans* and *Cryptococcus gattii*. These pathogens can cause a life-threatening meningoencephalitis after acquisition through the respiratory tract and subsequent dissemination to the central nervous system (CNS). While *C. gattii* can infect apparently immunocompetent hosts, *C. neoformans* is more often an opportunistic pathogen, affecting immunocompromised patients including those with HIV/AIDS, cancer and solid organ transplantation \[[@B1-jof-03-00035]\]. Cryptococcal meningitis has been estimated to affect up to 1 million people worldwide each year \[[@B2-jof-03-00035],[@B3-jof-03-00035]\]. Despite modern-day combination antifungal therapy, the mortality rate for cryptococcal meningitis is estimated at 15--25% \[[@B4-jof-03-00035],[@B5-jof-03-00035]\], and the at-risk population is expanding with the development of new immunosuppressive regimens for autoimmunity and cancer \[[@B6-jof-03-00035]\]. More effective approaches to treating cryptococcosis may necessitate the incorporation of immunomodulatory therapies. Therefore, it is essential to understand the cellular and molecular mechanisms of immunity to *Cryptococcus* in mammalian hosts. While the adaptive immune response to *Cryptococcus* is an important arm of anti-cryptococcal immunity (reviewed in \[[@B1-jof-03-00035],[@B7-jof-03-00035],[@B8-jof-03-00035]\]), this review will focus on our current knowledge of innate immune responses to the species *C. neoformans* and identify significant questions that remain to be investigated. 2. Animal Models of Cryptococcosis {#sec2-jof-03-00035} ================================== Different vertebrate and invertebrate animal models have been utilized in the study of cryptococcosis (for more comprehensive reviews see \[[@B1-jof-03-00035],[@B9-jof-03-00035],[@B10-jof-03-00035]\]). Predominantly, murine models have been used to study innate immune responses to *C. neoformans* due to the relative ease of genetic modification, manipulation and maintenance of this mammalian host. Therefore, results from mouse studies will comprise the majority of this review. The use of mouse models of cryptococcosis does have its challenges. Different mouse strains develop different T helper cell (Th) responses to *C. neoformans*; mice that develop Th type 2 (Th2) responses are more susceptible to cryptococcosis, while those that develop Th type 1 (Th1) responses are more resistant \[[@B11-jof-03-00035],[@B12-jof-03-00035],[@B13-jof-03-00035],[@B14-jof-03-00035],[@B15-jof-03-00035]\]. Mouse susceptibility can further vary depending on the virulence of the *C. neoformans* strain, the type and amount of infectious propagule (i.e., spore versus yeast form), and the route of administration \[[@B11-jof-03-00035],[@B12-jof-03-00035],[@B13-jof-03-00035],[@B14-jof-03-00035],[@B16-jof-03-00035],[@B17-jof-03-00035],[@B18-jof-03-00035]\]. *C. neoformans* has two main variants: var. *grubii* (Serotype A), which is the most common clinical isolate, and var. *neoformans* (Serotype D) \[[@B1-jof-03-00035]\]. The most physiologic route of infection is through the respiratory tract, either intranasal or intratracheal. However, respiratory infection in mice can result in variable dissemination to the CNS, so systemic infection (intravenous or intraperitoneal) and direct inoculation into the cerebrospinal fluid have been used to study the pathology of *C. neoformans* in the CNS \[[@B13-jof-03-00035],[@B19-jof-03-00035],[@B20-jof-03-00035]\]. As an example of the differences between mouse models of cryptococcosis, respiratory infection of C57BL/6 mice with the highly virulent serotype A strain H99 leads to a Th2-skewed immune response that results in an acute and uniformly fatal infection \[[@B21-jof-03-00035],[@B22-jof-03-00035],[@B23-jof-03-00035]\]. On the other hand, respiratory infection of BALB/c mice with a less virulent serotype D strain like 52D leads to a Th1-skewed immune response that results in a more chronic infection that can eventually be cleared in a CD4^+^ T-cell-dependent manner \[[@B11-jof-03-00035],[@B12-jof-03-00035],[@B15-jof-03-00035],[@B24-jof-03-00035]\]. A protective model of pulmonary cryptococcosis has also been established in which mice are infected with a *C. neoformans* strain H99-γ, that has been modified to express murine interferon gamma (IFNγ) \[[@B25-jof-03-00035]\]. 3. Host Recognition of *Cryptococcus* {#sec3-jof-03-00035} ===================================== Fungal pathogens are typically sensed through the detection of fungal antigens, or pathogen-associated molecular patterns (PAMPs), by pattern recognition receptors (PRRs) on host immune cells. Engagement of PRRs induces signal transduction that coordinates innate immune processes like phagocytosis and cytokine production. Common fungal PAMPs include components of the cell wall, such as β-glucans, mannans, and chitin. However, *C. neoformans* provides an interesting challenge due to its polysaccharide capsule that can mask these potential PAMPs. Correspondingly, many PRRs that are known to detect other fungal pathogens, including members of the C-type lectin receptor (CLR) and Toll-like receptor (TLR) families, do not have similar roles in the recognition of *C. neoformans.* Therefore, the mechanisms by which *C. neoformans* is sensed by the host are still not fully defined. 3.1. C-Type Lectin Receptors {#sec3dot1-jof-03-00035} ---------------------------- The CLRs are a large family of receptors that can recognize fungal carbohydrate ligands like β-glucans or mannans. An engaged CLR typically initiates downstream signaling pathways either through its own intracellular signaling domain, if present, or else through signaling adapters that contain an immunoreceptor tyrosine-based activation motif (ITAM), such as Fc receptor γ-chain (FcRγ or FcεRIγ chain) or DNAX activation protein of 12 kDa (DAP12). While CLRs have established roles in host innate immune responses to other pathogenic fungi (reviewed in \[[@B26-jof-03-00035]\]), their ability to mediate immunity to *C. neoformans* is less robust. There is evidence that β-glucans can be accessible on encapsulated yeast \[[@B27-jof-03-00035]\] and spore \[[@B28-jof-03-00035]\] forms of *C. neoformans* and that the fungal cell wall can be exposed at daughter bud sites prior to capsule assembly \[[@B29-jof-03-00035]\], but it is likely that the capsule is interfering with many of these potential interactions in vivo \[[@B30-jof-03-00035],[@B31-jof-03-00035]\]. Mannose receptor (MR/CD206) binds to fucose and terminal mannose moieties and is known to have roles in phagocytosis as well as antigen processing and presentation as a receptor of the endocytic pathway (reviewed in \[[@B32-jof-03-00035]\]). MR does not have any known intracellular signaling motifs and can also exist in a soluble form \[[@B32-jof-03-00035],[@B33-jof-03-00035]\], suggesting it may work in concert with other receptors for signal transduction, such as TLR2 \[[@B34-jof-03-00035]\]. Human MR has been shown to bind cryptococcal mannoproteins in vitro \[[@B35-jof-03-00035]\]. It is unclear if MR binds whole cryptococcal cells since MR-deficient murine phagocytes had no changes in binding and uptake of spores or yeast cells compared to wild-type (WT) phagocytes by microscopy \[[@B36-jof-03-00035]\]. Nevertheless, MR^−/−^ mice challenged with *C. neoformans* in an acute respiratory infection model appear to have a moderate increase in fungal burden and susceptibility to infection \[[@B37-jof-03-00035]\]. Several studies have investigated the ability of MR to facilitate the priming of adaptive T cell responses by dendritic cells (DCs). MR-deficient bone marrow-derived DCs (BMDCs) from mice had no changes in uptake of cryptococcal mannoproteins and no differences in expression of maturation markers like MHCII, CD40 and CD86 \[[@B37-jof-03-00035]\]. In contrast, studies with human cells indicate that blocking MR can inhibit maturation marker expression by DCs in response to cryptococcal mannoproteins \[[@B38-jof-03-00035]\] and can inhibit fungal uptake by DCs and subsequent lymphocyte proliferation \[[@B39-jof-03-00035]\]. Therefore, the mechanisms by which MR mediates innate immune responses to *C. neoformans* warrants continued investigation. Dendritic cell-specific intercellular adhesion molecule-3 grabbing non-integrin (DC-SIGN/CD209) binds fucose and mannose residues and is involved in antigen uptake as well as cellular adhesion (reviewed in \[[@B40-jof-03-00035]\]). Complicating its study, DC-SIGN has eight homologs in mice, designated DC-SIGN-related proteins (SIGNR) 1-8. SIGNR3 (CD209d) is considered the closest homolog to human DC-SIGN \[[@B41-jof-03-00035]\], and SIGNR3 and SIGNR1 (CD209b) are the only homologs shown to bind fungal ligands \[[@B42-jof-03-00035]\]. Human DC-SIGN binds cryptococcal mannoproteins in vitro \[[@B43-jof-03-00035]\], but murine SIGNR1 does not influence the ability of splenic macrophages to internalize the cryptococcal capsular polysaccharide glucuronoxylomannan (GXM) \[[@B44-jof-03-00035]\]. Additional studies on the potential role of DC-SIGN or its murine homologs during cryptococcal infection are currently lacking. Collectins are secreted carbohydrate-binding proteins and include the lung surfactant proteins (SPs) SP-A and SP-D and serum mannose binding lectin (MBL), also referred to as mannose binding protein (MBP). Collectins have been shown to engage with various fungal pathogens \[[@B45-jof-03-00035],[@B46-jof-03-00035],[@B47-jof-03-00035],[@B48-jof-03-00035]\] and regulate cytokine responses by binding to cell surface receptors like CD14, TLR2 and TLR4 \[[@B49-jof-03-00035],[@B50-jof-03-00035]\]. Interestingly, SP-D appears to be detrimental to the host, as SP-D^−/−^ mice have improved survival after infection with *C. neoformans* \[[@B51-jof-03-00035]\]. SP-D binds to and protects *C. neoformans* from macrophage killing, and its activity has been correlated with increased IL-5 production and pulmonary eosinophilia \[[@B29-jof-03-00035],[@B52-jof-03-00035]\]. The cryptococcal PAMP recognized by SP-D in vivo is unclear. In vitro, SP-D can bind to capsular GXM and mannoprotein 1 (MP1), but has higher affinity to pustulan, an analog of β-1,6-glucan found in the cryptococcal cell wall \[[@B29-jof-03-00035]\]. This higher affinity for a cell wall component correlates with the observation that acapsular *C. neoformans* mutants are more susceptible to agglutination and phagocytosis in the presence of SP-D compared to encapsulated strains \[[@B29-jof-03-00035],[@B53-jof-03-00035],[@B54-jof-03-00035]\]. Further studies are needed to determine which interactions and signaling mechanisms are essential for the harmful effects of SP-D on the host response. In contrast, SP-A can bind to *C. neoformans* but does not affect phagocytosis \[[@B55-jof-03-00035]\] and does not regulate murine susceptibility to infection \[[@B56-jof-03-00035]\]. MBL is known to bind mannose and *N*-acetylglucosamine (GlcNAc) and has been shown to act as an opsonin for complement activation \[[@B57-jof-03-00035]\]. However, soluble human MBL can only bind acapsular *C. neoformans* and minimally improves phagocytosis of these fungal cells by human polymorphonuclear cells in vitro \[[@B30-jof-03-00035],[@B54-jof-03-00035],[@B58-jof-03-00035],[@B59-jof-03-00035]\]. Thus, the overall role of collectins in anti-cryptococcal responses appears to be minimal or else harmful to the host. Other CLRs have been investigated but do not appear to have links to anti-cryptococcal immunity. Dectin-1 (CLEC7A) does not mediate immune responses in vitro or in vivo to either yeast or spore forms of *C. neoformans* \[[@B36-jof-03-00035],[@B60-jof-03-00035]\]. Co-expression of Dectin-1 and TLR2 in vitro also does not facilitate signal transduction in response to the fungus \[[@B61-jof-03-00035]\]. Dectin-2 (CLEC6A/CLEC4N) is not essential in host defense against *C. neoformans* yeast or spore forms despite molecular evidence of increased Th2 and decreased Th1 responses in Dectin-2^−/−^ mice \[[@B36-jof-03-00035],[@B62-jof-03-00035]\]. Dectin-3 or macrophage C-type lectin (MCL/CLEC4D/CLECSF8) does not regulate murine outcomes after *C. neoformans* infection or phagocytosis of fungal cells \[[@B63-jof-03-00035],[@B64-jof-03-00035]\] and cannot initiate signal transduction in response to *C. neoformans* spores \[[@B36-jof-03-00035]\]. Macrophage inducible C-type lectin (Mincle) does not bind *C. neoformans* or induce signal transduction in response to the fungus in vitro \[[@B36-jof-03-00035]\]. Langerin (CD207) does not bind to either encapsulated or acapsular *C. neoformans* \[[@B65-jof-03-00035]\]. Work remains to determine whether other CLRs, including novel receptors like CD23/FcεRII \[[@B66-jof-03-00035]\], may play a role in host recognition of *C. neoformans*. 3.2. Toll-Like Receptors {#sec3dot2-jof-03-00035} ------------------------ The potential role of TLRs as cryptococcal PRRs has been supported by evidence that myeloid differentiation primary response gene 88 (MyD88), a signaling molecule downstream of most TLRs, plays a role in murine anti-cryptococcal responses \[[@B67-jof-03-00035],[@B68-jof-03-00035],[@B69-jof-03-00035]\]. However, direct experimental evidence supporting a role for many of the TLRs in cryptococcosis is limited. Whether TLR signaling is relevant to human disease is unclear, as people with Mendelian defects in MyD88 do not have increased susceptibility to cryptococcosis \[[@B70-jof-03-00035],[@B71-jof-03-00035]\]. Studies on TLR2 have had conflicting results regarding the ability of this receptor to influence infectious outcomes and to initiate signal transduction in response to *C. neoformans*, perhaps related to differences in experimental design. Biondo et al. demonstrated that TLR2^−/−^ mice have increased susceptibility to systemic (intraperitoneal) infection with *C. neoformans,* as measured by survival, organ fungal burden and cytokine production \[[@B67-jof-03-00035]\]. Yauch et al. found that TLR2^−/−^ mice have increased susceptibility to respiratory infection but not systemic (intravenous) infection; however, there were no differences in lung fungal burden or cytokine production in the TLR2^−/−^ mice compared to WT mice \[[@B68-jof-03-00035]\]. Nakamura et al. also found no differences in fungal burden or cytokine production in TLR2^−/−^ mice infected through the respiratory tract, and *C. neoformans* did not induce nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) activation through TLR2 in an in vitro cell reporter assay, even with co-expression of Dectin-1 \[[@B61-jof-03-00035]\]. TLR4, in conjunction with its co-receptor CD14, can respond to cryptococcal GXM in vitro by inducing NF-κB but not mitogen activated protein (MAP) kinase pathways or tumor necrosis factor alpha (TNFα) secretion, suggesting incomplete activation \[[@B72-jof-03-00035]\]. Monoclonal antibodies against TLR4 can inhibit Fas ligand expression \[[@B73-jof-03-00035]\] and partially block GXM uptake by human peripheral blood mononuclear cell (PBMC)-derived macrophages \[[@B74-jof-03-00035]\]. However, TLR4 has not been shown to regulate murine susceptibility to infection \[[@B67-jof-03-00035],[@B68-jof-03-00035]\]. The strongest evidence for direct TLR involvement in anti-cryptococcal responses is for TLR9, an intracellular receptor of the endocytic pathway that typically recognizes unmethylated cytosine-phosphate-guanine (CpG) motifs common in the DNA of bacteria and viruses (reviewed in \[[@B75-jof-03-00035],[@B76-jof-03-00035]\]). More recently, the fungus *Aspergillus fumigatus* was found to contain unmethylated CpG motifs that can stimulate cytokine responses by DCs in vitro in a TLR9-dependent manner \[[@B77-jof-03-00035]\]. Several groups have also used synthetic CpG-oligodeoxynucleotides to boost the immune response against *C. neoformans* \[[@B78-jof-03-00035],[@B79-jof-03-00035],[@B80-jof-03-00035],[@B81-jof-03-00035]\]. TLR9^−/−^ mice are more susceptible to cryptococcosis, potentially due to decreased recruitment and maturation of DCs and the development of Th2 immune responses, including alternative activation of macrophages \[[@B69-jof-03-00035],[@B82-jof-03-00035],[@B83-jof-03-00035],[@B84-jof-03-00035]\]. Cryptococcal DNA can stimulate in vitro cytokine responses by DCs, which can be partially inhibited by deletion of TLR9 or MyD88 \[[@B84-jof-03-00035]\]. Subsequently, it has been shown that polymerase chain reaction (PCR) products amplified from cryptococcal genes involved in virulence including *URA5*, *CNLAC1*, and *CAP59* can induce the same cytokine responses by DCs \[[@B85-jof-03-00035]\]. Interestingly, these genes do not contain canonical CpG motifs. Thus, cryptococcal DNA can function as a PAMP for TLR9, but the specific nucleic acid motifs involved in its recognition have not been elucidated. 3.3. Nucleotide-Binding Oligomerization Domain (NOD)-Like Receptors {#sec3dot3-jof-03-00035} ------------------------------------------------------------------- The NOD-like receptors, or nucleotide-binding domain leucine-rich repeat-containing receptors (NLRs), are a family of cytoplasmic receptors that, upon activation, can form an inflammasome complex that cleaves and activates pro-IL-1β and pro-IL-18 generated after initial microbial detection induces NF-κB (reviewed in \[[@B86-jof-03-00035]\]). NLR family, pyrin domain-containing 3 (NLRP3) is an NLR that has been shown to play a role in immunity against *A. fumigatus* and the yeast *Candida albicans* \[[@B87-jof-03-00035],[@B88-jof-03-00035],[@B89-jof-03-00035]\], although the ligand for NLRP3 remains unidentified. Biofilms of encapsulated *C. neoformans*, opsonized and encapsulated *C. neoformans*, and acapsular yeast forms of *C. neoformans* stimulate formation of the NLRP3 inflammasome, and mice deficient in components of the NLRP3 inflammasome are more susceptible to infection \[[@B90-jof-03-00035],[@B91-jof-03-00035],[@B92-jof-03-00035]\]. However, additional studies will be needed to further clarify the role of NLRs and inflammasome formation in antifungal responses to *C. neoformans*. 3.4. Scavenger Receptors {#sec3dot4-jof-03-00035} ------------------------ Scavenger receptors are classically known to bind and internalize oxidized low-density lipoproteins. In more recent years, they have been found to have very diverse ligands and can serve as PRRs that detect microbial PAMPs and complex with other receptors like TLRs (reviewed in \[[@B93-jof-03-00035]\]). In vitro studies indicate that the scavenger receptors CD36 and scavenger receptor class F member 1 (SCARF1), also known as scavenger receptor expressed by endothelial cells 1 (SREC1), can bind to and internalize encapsulated *C. neoformans*, thereby inducing cytokine responses that can further be enhanced by synergy with TLR2; competition assays suggest that CD36 and SCARF1 may bind to β-glucans and, to a lesser extent, mannan, although they do not contain classic lectin-binding domains \[[@B94-jof-03-00035]\]. In the same study, neutralizing anti-SCARF1 antibody inhibited binding of *C. neoformans* to alveolar macrophages in vivo, CD36^−/−^ mice were found to be more susceptible to systemic infection with *C. neoformans*, and deletion of CD36 and SCARF1 orthologues in the nematode *Caenorhabditis elegans* resulted in increased susceptibility to fungal challenge. Macrophage receptor with collagenous structure (MARCO) has been shown to enhance early lung recruitment of monocyte-derived immune cells and protective cytokine responses after murine respiratory infection with *C. neoformans* that correlate with a transient improvement in fungal clearance \[[@B95-jof-03-00035]\]. Interestingly, MARCO-deficient macrophages and DCs exhibit no defect in fungicidal activity though they do have decreased interactions with fungal cells \[[@B95-jof-03-00035]\]. Scavenger receptor A (SRA/SR-AI/II/CD204/SCARA1) has been reported to have detrimental effects on host immunity to *C. neoformans*. SRA^−/−^ mice have decreases in lung fungal burden likely related to regulation of cytokine responses that influence innate immune cell recruitment and activation \[[@B96-jof-03-00035]\]. The potential cryptococcal ligands for SRA and MARCO and additional mechanistic details for how all these scavenger receptors influence anti-cryptococcal responses have not yet been determined. 3.5. Natural Antibodies {#sec3dot5-jof-03-00035} ----------------------- Natural antibodies, that are predominantly of the immunoglobulin M (IgM) isotype, are constitutively produced in mammalian hosts by an innate subset of B lymphocytes called B-1 cells; opsonization of microbial antigens with natural IgM can result in complement activation, phagocytosis by macrophages, and priming of adaptive immune responses (reviewed in \[[@B97-jof-03-00035],[@B98-jof-03-00035]\]). It has been shown that IgM produced by murine B-1 cells in vitro can bind to cell wall laminarin, capsular GXM, and acapsular and heat-killed encapsulated *C. neoformans* \[[@B99-jof-03-00035]\]. Secretory IgM-deficient (sIgM^−/−^) mice have increased susceptibility to respiratory infection with *C. neoformans* compared to control mice and exhibit defects in Th1 polarization and phagocytosis of fungi by alveolar macrophages; the defect in phagocytosis can be ameliorated by administration of IgM into the lungs \[[@B100-jof-03-00035]\]. Additionally, depletion of B-1 cells in pulmonary infected mice increases fungal burden and decreases phagocytosis of fungal cells by alveolar macrophages compared to non-depleted controls; adoptive transfer of B-1 cells into depleted mice can restore the phenotype to that of control mice \[[@B99-jof-03-00035]\]. On the other hand, sIgM^−/−^ mice infected systemically with *C. neoformans* have improved survival compared to control mice \[[@B101-jof-03-00035]\]. It was found that these sIgM^−/−^ mice have an increased baseline number of B-1 cells \[[@B101-jof-03-00035]\], and, interestingly, B-1 cell derivatives may have direct fungicidal effects against *C. neoformans* \[[@B102-jof-03-00035]\]. Thus, IgM and B-1 cells may play different roles in the anti-cryptococcal response depending on the tissue compartment. X-linked immunodeficient (XID) mice, that have a defect in B cell development and IgM production due to a mutation in Bruton's tyrosine kinase (Btk), exhibit increased susceptibility to both respiratory and systemic infection with *C. neoformans* \[[@B6-jof-03-00035],[@B103-jof-03-00035]\]. However, adoptive transfer of B-1 cells into pulmonary infected XID mice could neither reverse this susceptibility to *C. neoformans* nor fully restore serum IgM levels, suggesting that B-1 cells may not be the only source of protective IgM or that additional immune mechanisms are contributing to the phenotype in this particular model \[[@B6-jof-03-00035]\]. Human studies support a role for IgM in protective immune responses against *C. neoformans.* The percentage of IgM-expressing memory B cells inversely correlates with the risk for developing cryptococcosis among HIV-positive patients \[[@B104-jof-03-00035]\]. In solid organ transplant recipients, pre-transplantation levels of GXM-reactive IgM inversely correlate with the development of post-transplant cryptococcosis \[[@B105-jof-03-00035]\]. The ability to identify B-1 cells in humans has recently been reported (reviewed in \[[@B106-jof-03-00035]\]), which may facilitate future studies on the role of these innate immune cells and natural antibodies in human cryptococcosis. 3.6. Complement and Other Soluble Mediators {#sec3dot6-jof-03-00035} ------------------------------------------- The complement system is an important mediator for the phagocytosis of *C. neoformans* by innate immune cells. Opsonization by complement has been shown to improve uptake and killing of *C. neoformans* by phagocytes \[[@B107-jof-03-00035],[@B108-jof-03-00035]\] and to mediate DC responses to *C. neoformans* \[[@B109-jof-03-00035]\]. Activation of complement can occur through three pathways: alternative, classical, and lectin (reviewed in \[[@B110-jof-03-00035]\]). Disruption of the alternative, but not the classical, pathway of complement reduces phagocytosis of *C. neoformans* in vitro and increases the mortality of guinea pigs after infection \[[@B111-jof-03-00035],[@B112-jof-03-00035]\]. The lectin pathway likely does not play a significant role given minimal interactions between MBL and *C. neoformans*, as discussed earlier in this review. It has been shown that complement component 3 (C3) binds to the capsule of *C. neoformans* and then is degraded to inactivated C3b (iC3b) \[[@B113-jof-03-00035],[@B114-jof-03-00035],[@B115-jof-03-00035]\]. Phagocytosis can then proceed via the action of complement receptors (CR). Blocking CR1, CR3 and CR4 decreases the interaction between *C. neoformans* and human macrophages in vitro \[[@B108-jof-03-00035]\]. CR3 has been shown to facilitate complement-mediated phagocytosis of *C. neoformans* by murine macrophages \[[@B116-jof-03-00035]\], but CR3 and CR4 can also mediate phagocytosis independent of complement \[[@B117-jof-03-00035]\]. Additionally, signaling by C5a through its receptor C5aR appears to be important for neutrophil uptake and killing of *C. neoformans* in mice \[[@B118-jof-03-00035]\]. Other potential soluble mediators of anti-cryptococcal immunity have been studied. Pentraxin 3 (PTX3) expression is induced in the brains of mice infected intracerebrally with *C. neoformans* \[[@B119-jof-03-00035]\], but it is not yet known what function PTX3 may play in anti-cryptococcal responses. Recombinant rat ficolin-A can bind and facilitate uptake of acapsular mutants of *C. neoformans* by lung epithelial cells in vitro but does not bind encapsulated *C. neoformans* \[[@B120-jof-03-00035]\], so it is unclear if ficolins play any significant role in anti-cryptococcal immunity. Finally, production of antimicrobial peptides is increased in a protective model of cryptococcosis \[[@B121-jof-03-00035]\], but their specific functions in the response to *C. neoformans* are not understood. 3.7. Other Recognition Pathways {#sec3dot7-jof-03-00035} ------------------------------- Additional potential cryptococcal PAMPs have been identified, but their receptors remain unclear. Chitin is a long chain polymer of GlcNAc that can also be deacetylated to chitosan; both forms are components of the cryptococcal cell wall \[[@B1-jof-03-00035]\] and appear to have detrimental effects on the host immune response upon recognition. The chitin content of cryptococcal cells has been shown to correlate with Th2 cell accumulation and increased mortality in the murine host \[[@B122-jof-03-00035]\], and a chitosan-deficient strain of *C. neoformans* promotes protective Th1 host responses and is avirulent in mice \[[@B123-jof-03-00035]\]. Cryptococcal chitin has been shown to induce IL-10 secretion from human and murine macrophages \[[@B124-jof-03-00035]\] and induce Th2 responses through CD11b^+^ conventional DCs, although this process does not seem to occur through direct sensing of chitin by DCs \[[@B122-jof-03-00035]\]. The PRRs for chitin and chitosan are still unknown (reviewed in \[[@B125-jof-03-00035]\]). Studies using *C. albicans*-derived chitin suggest that chitin recognition is dependent on MR, NOD2, and TLR9 \[[@B124-jof-03-00035]\], and purified chitosan can induce inflammasome activation \[[@B126-jof-03-00035],[@B127-jof-03-00035]\]. The hypervirulent *rim101*Δ *C. neoformans* mutant, that has increased chitosan content and exposure of chito-oligomers on its cell surface, induces TNFα secretion by murine bone marrow-derived macrophages but not IL-1β, suggesting cryptococcal chitosan does not induce the inflammasome; however, the induction of TNFα appears to be dependent on the caspase recruitment domain-containing 9 (CARD9) and MyD88 signaling molecules, indicating a potential role for CLRs and TLRs \[[@B128-jof-03-00035],[@B129-jof-03-00035],[@B130-jof-03-00035]\]. Another possible source of cryptococcal PAMPs are extracellular vesicles (EVs), also referred to as exosomes, which are bilayer vesicles released by *C. neoformans* \[[@B131-jof-03-00035]\] and can contain an array of cellular components including polysaccharides, nucleic acids, and proteins (reviewed in \[[@B132-jof-03-00035],[@B133-jof-03-00035]\]). Although some EVs may be able to promote the virulence of *C. neoformans* \[[@B134-jof-03-00035],[@B135-jof-03-00035]\], cryptococcal EVs have also been shown to be internalized by macrophages and stimulate cytokine secretion, NO production, and uptake and killing of the fungus in vitro \[[@B136-jof-03-00035]\]. As we improve our technical capability to isolate extracellular vesicles, it will be interesting to perform further analysis of their contents under different host conditions and determine if there are specific EV-borne PAMP interactions with host PRRs. 4. Intracellular Signaling Molecules {#sec4-jof-03-00035} ==================================== Another approach to defining innate immune responses to *C. neoformans* has been to study the role of molecules that commonly integrate signals from PRRs after fungal recognition. These include CARD9, MyD88, and the signaling adapters DAP12 and FcRγ (reviewed in \[[@B137-jof-03-00035],[@B138-jof-03-00035]\]). CARD9 is best known as a downstream mediator of signaling through CLRs like Dectin-1, but it can also transmit signals from TLRs and NOD2 and can facilitate activation of NF-κB or MAP kinase pathways (reviewed in \[[@B139-jof-03-00035],[@B140-jof-03-00035]\]). Respiratory infection of CARD9^−/−^ mice with *C. neoformans* results in increased lung fungal burden and neutrophilia along with defective early IFNγ production by NK cells and memory T cells \[[@B141-jof-03-00035]\]. CARD9 may not play a direct role in phagocytic pathways, as CARD9-deficient murine phagocytes have no defect in binding or uptake of *C. neoformans* spores or yeast forms as evaluated by microscopy \[[@B36-jof-03-00035]\], but it may regulate cytokine responses. For example, TNFα production is reduced in CARD9-deficient murine macrophages in response to the chitosan-enriched, acapsular *rim101*Δ *cap59*Δ *C. neoformans* mutant \[[@B128-jof-03-00035]\]. Together, these studies suggest that CARD9 may play a role in the host response to *Cryptococcus*, but the signaling pathway requires further definition. MyD88 has well-established roles in signal transduction for most TLRs but can also function downstream of the cytokine receptors IL-1R and IL-18R \[[@B142-jof-03-00035],[@B143-jof-03-00035]\]. MyD88^−/−^ mice have increased susceptibility to both systemic and respiratory infection with *C. neoformans* \[[@B67-jof-03-00035],[@B68-jof-03-00035],[@B69-jof-03-00035]\]. Since the increased susceptibility of TLR2^−/−^ mice to cryptococcosis is not as pronounced as that of MyD88^−/−^ mice \[[@B67-jof-03-00035],[@B68-jof-03-00035]\], MyD88 may mediate non-TLR signaling in response to *C. neoformans* as well. Indeed, IL-18R^−/−^ mice, but not IL-1R^−/−^ mice, have increased susceptibility to respiratory infection with *C. neoformans*, and knockout of either receptor causes significant changes in lung cytokine production compared to WT mice \[[@B69-jof-03-00035]\]. Additionally, mice deficient in IL-18 have increased susceptibility to cryptococcosis \[[@B144-jof-03-00035],[@B145-jof-03-00035]\]. Thus, MyD88 may integrate signals from multiple cryptococcal recognition pathways during the host innate immune response. DAP12 is an ITAM-containing signaling adapter that pairs to a variety of carbohydrate- and protein-binding immunoreceptors on myeloid and NK cells, including CLRs and other tyrosine kinase-signaling receptors (reviewed in \[[@B138-jof-03-00035],[@B146-jof-03-00035],[@B147-jof-03-00035],[@B148-jof-03-00035]\]). DAP12 has been shown to have roles in the regulation of macrophage activation and survival \[[@B149-jof-03-00035],[@B150-jof-03-00035]\]. Interestingly, DAP12-deficient macrophages have enhanced fungal uptake and killing and TNFα production in response to *C. neoformans*, and DAP12^−/−^ mice are more resistant to respiratory infection with *C. neoformans* than WT mice \[[@B21-jof-03-00035]\]. Thus, DAP12 appears to inhibit beneficial fungicidal macrophage responses to *C. neoformans*. Further research will be needed to identify the DAP12-associated PRRs that trigger these immunosuppressive effects and could be potential immunomodulatory targets for the treatment of cryptococcosis. FcRγ is also an ITAM-containing signaling adapter utilized by receptors on myeloid and NK cells (reviewed in \[[@B138-jof-03-00035]\]). In contrast to DAP12, there is no current evidence that supports a role for FcRγ in innate immune responses to *C. neoformans*. Murine phagocytes from FcRγ^−/−^ mice demonstrate no changes in binding or uptake of spores or yeast \[[@B36-jof-03-00035]\]. Any other potential roles of FcRγ during cryptococcosis are still unknown. Additional important signaling molecules in fungal sensing pathways, including spleen tyrosine kinase (Syk), have not yet been investigated for their roles in cryptococcosis. As these gaps in our knowledge are filled, we may gain further insight into the signaling network that enables coordination of the innate immune response by effector cells. 5. Effector Functions of Innate Immune Cells {#sec5-jof-03-00035} ============================================ After a fungal pathogen is recognized by the innate immune system, signal transduction coordinates the effector functions of innate immune cells, which may include phagocytosis and the generation of inflammatory response mediators such as cytokines, fungicidal compounds and acute phase reactants. These processes can regulate clearance of the fungus or initiate the development of adaptive immune responses. In the case of *C. neoformans*, these pathways can also be subverted by the pathogen to suppress the host innate immune response and allow the fungus to proliferate instead. 5.1. Inflammatory Monocytes {#sec5dot1-jof-03-00035} --------------------------- Inflammatory monocytes are innate immune cells that are recruited from the bone marrow to sites of infection or inflammation, whereupon they can differentiate into macrophages or DCs \[[@B151-jof-03-00035],[@B152-jof-03-00035],[@B153-jof-03-00035]\]. Although monocytes from HIV-positive patients have been reported to have impaired chemotaxis and cytotoxicity \[[@B154-jof-03-00035],[@B155-jof-03-00035]\], studies using human monocytes and macrophages have had conflicting results about the role of these cells during cryptococcosis. Some researchers have found that human PBMCs can kill *C. neoformans* in vitro \[[@B156-jof-03-00035],[@B157-jof-03-00035],[@B158-jof-03-00035]\], and blood monocyte deactivation was associated with early mortality in HIV-associated cryptococcal meningitis \[[@B159-jof-03-00035]\]. In other studies, human PBMCs and monocyte-derived macrophages were merely fungistatic \[[@B160-jof-03-00035]\] or even permissive for intracellular cryptococcal proliferation and dissemination \[[@B161-jof-03-00035],[@B162-jof-03-00035],[@B163-jof-03-00035]\], and there was no difference in antifungal activity of monocyte-derived macrophages from cryptococcosis patients compared to normal controls \[[@B161-jof-03-00035]\]. In mice, inflammatory monocytes are defined as cells expressing lymphocyte antigen 6 complex, locus C1 (Ly6C) and C-C chemokine receptor type 2 (CCR2) that can migrate in response to the chemokines monocyte chemoattractant protein (MCP1), also known as C-C chemokine ligand 2 (CCL2), and CCL7 (reviewed in \[[@B152-jof-03-00035]\]). In chronic models of respiratory cryptococcosis, inflammatory monocytes appear to be beneficial to the host because CCR2^−/−^ mice, that have a defect in monocyte recruitment, develop Th2 responses and have increased fungal burden and decreased lung macrophages, CD11b^+^ DCs and CD8^+^ T cells \[[@B164-jof-03-00035],[@B165-jof-03-00035],[@B166-jof-03-00035]\]. Further, in response to infection with *C. neoformans*, Ly6C^hi^ CCR2^+^ monocytes differentiate into fungicidal exudative macrophages and CD11b^+^ DCs that promote fungal clearance and Th1 adaptive immune responses, respectively \[[@B166-jof-03-00035],[@B167-jof-03-00035]\]. However, it is interesting to note that in an acute model of respiratory cryptococcosis, enhancing Th2 responses worsens survival and correlates with increased recruitment of monocytes to the lungs \[[@B122-jof-03-00035]\], suggesting that monocytes and their derivatives could play different roles depending on the host environment. This theory could potentially account for the differences observed in studies on human monocyte responses to *C. neoformans*. 5.2. Macrophages {#sec5dot2-jof-03-00035} ---------------- Macrophages are phagocytic cells that include tissue-resident, embryonic-derived cells like lung alveolar macrophages as well as monocyte-derived macrophages that are of hematopoietic cell origin \[[@B168-jof-03-00035]\]. Since macrophages, in the guise of alveolar macrophages, are present in the lung at the time that *C. neoformans* is inhaled into the lungs, they have long been considered to be the first line innate immune cell in host defense against the fungus. Indeed, fungi are seen within lung macrophages in patients with cryptococcosis \[[@B169-jof-03-00035]\], and in murine models, alveolar macrophages have been visualized to quickly take up cryptococcal cells after respiratory infection \[[@B170-jof-03-00035],[@B171-jof-03-00035]\]. However, there have been differing results regarding the ability of macrophages to clear *C. neoformans* from the host. While some groups have observed that murine macrophages can kill *C. neoformans* in vitro \[[@B172-jof-03-00035],[@B173-jof-03-00035]\], others have found that the fungus can actually replicate within these cells, which may lead to dissemination by way of a Trojan Horse mechanism \[[@B170-jof-03-00035],[@B174-jof-03-00035],[@B175-jof-03-00035],[@B176-jof-03-00035]\]. Interestingly, clinical *C. neoformans* isolates that exhibit higher rates of uptake by macrophages in vitro predict poor patient outcomes \[[@B177-jof-03-00035]\]. In murine respiratory infection models, depletion of macrophages using liposomal clodronate reduces fungal burden \[[@B176-jof-03-00035],[@B178-jof-03-00035]\]. In contrast, ablation of macrophages, along with DCs, using transgenic CD11c-diphtheria toxin receptor (DTR) mice was found to worsen survival without any differences in lung fungal burden \[[@B179-jof-03-00035]\]. It is important to note that the ablation protocol for CD11c-DTR mice can induce fatal toxicity, even in the absence of any infection (reviewed in \[[@B180-jof-03-00035]\]). Thus, it will be necessary to confirm this result using alternative strategies. It has become apparent that macrophage polarization may be a key determinant of whether macrophages are beneficial or detrimental during cryptococcosis. M1 (classically activated) macrophages produce nitric oxide (NO) through inducible NO synthase (iNOS) expression, secrete TNFα, and are fungicidal against *C. neoformans*, while M2 (alternatively activated) macrophages typically express the markers arginase 1 (Arg1), chitinase-like 4 (Chil4 or Ym2), resistin like alpha (Retnla or Fizz1), and MR (CD206) and are permissive for fungal growth (reviewed in \[[@B181-jof-03-00035]\]). M2 polarization has been associated with severe cryptococcal disease in non-HIV patients \[[@B182-jof-03-00035]\], though not in HIV-positive patients \[[@B159-jof-03-00035]\]. In mice, alternative activation of macrophages worsens cryptococcosis in the brain \[[@B183-jof-03-00035]\]. In a chronic respiratory infection model in mice, lung macrophages cycle from a resting state to an M2 phenotype, that corresponds with initial proliferation of *C. neoformans* in the lungs, followed by an M1 phenotype, that correlates to a period of fungal clearance, and then back to a resting state; this cycling could be simulated in vitro by modifying the cytokine environment with either IFNγ (M1) or IL-4 (M2) \[[@B184-jof-03-00035],[@B185-jof-03-00035]\]. IFNγ^−/−^ mice have increased lung fungal burden and demonstrate alternative activation of macrophages after pulmonary challenge with *C. neoformans* \[[@B185-jof-03-00035],[@B186-jof-03-00035]\]. IL4^−/−^ mice have improved fungal clearance and demonstrate classical activation of macrophages \[[@B185-jof-03-00035],[@B187-jof-03-00035]\]. *C. neoformans* cells weakly stimulate expression of iNOS and Arg1 in murine macrophages in vitro, suggesting that direct interaction between fungus and phagocyte is not the only determinant of macrophage polarization \[[@B184-jof-03-00035]\]. From a therapeutic perspective, it will be helpful to further dissect the signaling mechanisms that can influence the polarization of macrophages during cryptococcosis. Various signaling components have been identified, including DAP12 \[[@B21-jof-03-00035]\], heat shock protein 70 (Hsp70) \[[@B188-jof-03-00035]\], and signal transducer and activator of transcription 1 (STAT1) \[[@B189-jof-03-00035],[@B190-jof-03-00035]\]. Studies on other intracellular pathogens suggest that TLR signaling can induce Arg1 in macrophages \[[@B191-jof-03-00035]\]. Understanding these processes will allow testing of the idea that macrophage polarization drives infectious outcomes in mammalian hosts and could lay the foundation for potential new immunomodulatory strategies for the treatment of cryptococcosis. 5.3. Dendritic Cells {#sec5dot3-jof-03-00035} -------------------- The primary function of DCs in antifungal responses is to take up, process and present antigens to prime T cells and trigger adaptive immunity (reviewed in \[[@B192-jof-03-00035],[@B193-jof-03-00035],[@B194-jof-03-00035]\]). DCs are a heterogeneous group of cells whose classification continues to evolve. Generally, it is recognized that the main subsets of DCs include classical or conventional DCs (cDCs), monocyte-derived DCs (moDCs), plasmacytoid DCs (pDCs), and Langerhans cells (reviewed in \[[@B151-jof-03-00035],[@B195-jof-03-00035]\]). DCs appear to have roles in protective immunity against *C. neoformans*. Ablation of DCs, along with macrophages, using CD11c-DTR mice increases murine mortality after infection \[[@B179-jof-03-00035]\], although there are limitations to this mouse model as mentioned previously in this review. DCs have been shown to take up and present cryptococcal glycoantigens \[[@B43-jof-03-00035]\]. Researchers have found that protective adaptive immune responses to cryptococcal antigen can be mediated by CD11b^+^ DCs and Langerhans cells \[[@B196-jof-03-00035]\], and moDCs have been shown to enhance Th1 responses after respiratory infection with *C. neoformans* \[[@B166-jof-03-00035]\]. Cryptococcal cells and cryptococcal antigen have been shown to stimulate IL-12 and IL-23p40 secretion and expression of activation markers by DCs in vitro \[[@B38-jof-03-00035],[@B197-jof-03-00035]\]. DCs upregulate the CD80 activation marker in response to pulmonary *C. neoformans* challenge in vivo and can stimulate T cell activation ex vivo \[[@B198-jof-03-00035]\]. In addition, DCs can phagocytose and kill *C. neoformans* \[[@B39-jof-03-00035],[@B109-jof-03-00035],[@B198-jof-03-00035]\]. However, CD11b^+^ cDCs can also mediate harmful Th2 immune responses stimulated by chitin, as discussed earlier in this review \[[@B122-jof-03-00035]\]. The potential role of pDCs during cryptococcosis has not been as closely examined as that of cDCs and moDCs. *C. neoformans* does not appear to activate pDCs in vitro \[[@B197-jof-03-00035]\]. Other reports suggest that pDCs phagocytose *C. neoformans* and limit fungal growth through a Dectin-3 and ROS-dependent mechanism \[[@B63-jof-03-00035]\]. However, infectious outcomes are not altered in Dectin-3^−/−^ mice \[[@B63-jof-03-00035],[@B64-jof-03-00035]\]. 5.4. Neutrophils {#sec5dot4-jof-03-00035} ---------------- Neutrophils are granulocytes that can phagocytose microorganisms, release antimicrobial enzymes, and produce neutrophil extracellular traps (NETs) (reviewed in \[[@B137-jof-03-00035]\]). Neutrophils have established roles in the innate immune response to fungal pathogens like *A. fumigatus* \[[@B199-jof-03-00035]\], but their role in anti-cryptococcal immunity remains poorly defined. Human neutrophils can kill *C. neoformans* in vitro \[[@B157-jof-03-00035],[@B200-jof-03-00035]\], and treatment of mice with human recombinant granulocyte-colony stimulating factor (G-CSF) in combination with fluconazole improves survival from intracerebral infection \[[@B201-jof-03-00035]\]. At the same time, *C. neoformans* can inhibit human neutrophil migration \[[@B202-jof-03-00035]\], and its capsule blocks neutrophil binding of fungal cells \[[@B203-jof-03-00035]\]. Human neutrophils release NETs in response to acapsular *C. neoformans* mutants and the capsular polysaccharide glucoronoxylomannogalactan (GXMGal) but not in response to encapsulated *C. neoformans* or capsular GXM \[[@B204-jof-03-00035]\]. However, if already formed, NETs can kill encapsulated *C. neoformans* \[[@B204-jof-03-00035]\]. In a systemic model of murine cryptococcosis, anti-Ly6G (1A8) antibody depletion of neutrophils suggests that these cells are needed for fungal clearance in the brain and lungs \[[@B205-jof-03-00035]\], and neutrophils have been visualized to swarm the fungus for removal from the brain microvasculature \[[@B206-jof-03-00035],[@B207-jof-03-00035]\]. In a protective model of cryptococcosis, neutrophils are the primary source of IL-17A that enhances protective immune responses, although they are not essential as γδ T cells can produce IL-17A in their absence \[[@B208-jof-03-00035]\]. On the other hand, after pulmonary challenge with *C. neoformans*, depletion of neutrophils and inflammatory monocytes with anti-Gr-1 (RB6-8C5) antibody improves murine survival and causes an overall reduction in inflammatory lung damage, suggesting a detrimental role for neutrophils \[[@B209-jof-03-00035]\]. In the same study, treatment with anti-Gr-1 had no effect on murine survival after systemic infection. Further supporting a harmful role for neutrophils, mice with genetically-induced neutrophilia appear to have increased susceptibility to cryptococcal disease \[[@B210-jof-03-00035]\]. Therefore, the role of neutrophils in anti-cryptococcal responses is still not clear and may depend on the specific host and/or tissue environment. 5.5. Natural Killer Cells {#sec5dot5-jof-03-00035} ------------------------- NK cells are cytotoxic lymphocytes of the innate immune system. Studies in murine models of systemic cryptococcosis suggest that NK cells may participate in early anti-cryptococcal immune responses through direct fungal interactions \[[@B211-jof-03-00035],[@B212-jof-03-00035],[@B213-jof-03-00035],[@B214-jof-03-00035],[@B215-jof-03-00035],[@B216-jof-03-00035]\]. Other groups find that instead of direct cytotoxic effects against *C. neoformans*, NK cells may enhance the fungicidal activity of macrophages in mice by producing IFNγ \[[@B217-jof-03-00035],[@B218-jof-03-00035]\]. Mice lacking NK cells have increased fungal burden, but they do not have increased susceptibility to infection \[[@B211-jof-03-00035],[@B213-jof-03-00035]\]. The role of NK cells in anti-cryptococcal responses has been more closely examined in human cells. NK cells from HIV-positive patients are impaired in their growth inhibition of *C. neoformans* \[[@B219-jof-03-00035]\]. Human lymphocytes and NK cells have been shown to inhibit cryptococcal growth through direct interaction \[[@B220-jof-03-00035],[@B221-jof-03-00035]\]. In studies using human primary NK cells or cell lines, Mody and colleagues have demonstrated that binding of *C. neoformans* by NK cells leads to signaling through the PI3K-ERK1/2 pathway \[[@B222-jof-03-00035]\] and triggers perforin degranulation to facilitate cryptococcal killing \[[@B223-jof-03-00035]\]. The natural cytotoxicity receptor NKp30, an immunoglobulin-like protein, has been identified as a human NK cell PRR for *C. neoformans* \[[@B224-jof-03-00035]\]. In the same study, blocking NKp30 impaired PI3K-ERK1/2 signaling, perforin release and ultimately fungal killing in response to *C. neoformans*. Additionally, it was shown that NK cells from HIV patients have decreased expression of NKp30 and decreased toxicity against *C. neoformans*, both of which can be reversed by IL-12 treatment in vitro. Work remains to identify any additional cryptococcal PRRs on NK cells as well as the cryptococcal ligand for NKp30. Studies on the detection of *Candida glabrata* by the related receptor NKp46 suggest that fungal adhesins could be potential ligands for this class of receptors \[[@B225-jof-03-00035]\]. 5.6. Eosinophils {#sec5dot6-jof-03-00035} ---------------- Eosinophils are granulocytes that are best known for their roles in allergic responses and parasitic infections (reviewed in \[[@B226-jof-03-00035]\]). Eosinophilia has been associated with cryptococcal disease in humans and mice \[[@B11-jof-03-00035],[@B227-jof-03-00035],[@B228-jof-03-00035],[@B229-jof-03-00035],[@B230-jof-03-00035],[@B231-jof-03-00035],[@B232-jof-03-00035],[@B233-jof-03-00035],[@B234-jof-03-00035],[@B235-jof-03-00035]\] and positively correlated to murine susceptibility to cryptococcosis \[[@B11-jof-03-00035],[@B52-jof-03-00035]\], but it is not clear if eosinophils have an essential role in the innate immune response to *C. neoformans* or if their recruitment is the byproduct of an ineffectual Th2 response. After infection with *C. neoformans*, eosinophil-deficient ΔdblGATA mice have enhanced Th1 and Th17 responses and decreased lung recruitment of other inflammatory cells, although fungal burden in the lung and brain are not significantly different from WT mice \[[@B236-jof-03-00035]\]. It is interesting to note that in rats, eosinophils can phagocytose *C. neoformans* and prime T and B cells in order to generate Th1 responses that are protective for the host \[[@B233-jof-03-00035],[@B237-jof-03-00035],[@B238-jof-03-00035]\]. Therefore, the role of eosinophils during cryptococcosis may depend on the particular host setting. 5.7. Other Innate Immune Cells {#sec5dot7-jof-03-00035} ------------------------------ Innate lymphoid cells (ILCs), other than NK cells, have not been extensively studied in cryptococcosis, but type 2 ILCs may be detrimental to host anti-cryptococcal responses \[[@B239-jof-03-00035]\]. Derivatives of B-1 cells may have direct antifungal effects against *C. neoformans* \[[@B102-jof-03-00035]\], as discussed earlier in this review. Epithelial and endothelial cells not only serve as a physical barrier to microbial invasion, but can also participate as effector innate immune cells (reviewed in \[[@B240-jof-03-00035],[@B241-jof-03-00035]\]). Lung epithelial cells can bind *C. neoformans* and produce cytokines in response to the fungus \[[@B242-jof-03-00035],[@B243-jof-03-00035]\], and endothelial cells may enhance anti-cryptococcal activity of neutrophils \[[@B244-jof-03-00035]\]. The potential role of γδ T cells is still unclear. Mice deficient in γδ T cells have improved infectious outcomes after *C. neoformans* challenge \[[@B245-jof-03-00035]\], but studies in a protective model of cryptococcosis suggest that γδ T cells are a source of beneficial IL-17A in the setting of neutropenia \[[@B208-jof-03-00035]\]. 6. Conclusions {#sec6-jof-03-00035} ============== By methodically investigating common mammalian antifungal mechanisms, researchers have established important roles for cellular PRRs, in particular MR, TLR9, and NKp30, and for signal transduction through CARD9 and MyD88 in protective immune responses against *C. neoformans*. Other promising PRR candidates include NLRs like NLRP3 and certain scavenger receptors. Furthermore, soluble mediators including natural IgM and complement have key functions in facilitating host recognition and immunity to *C. neoformans*. Many additional signaling pathways have been studied, but they either require further evaluation as to their specific anti-cryptococcal functions or appear to have limited or even detrimental roles in host responses to *C. neoformans*. Whether the limited findings are due to redundancies in the immune system remains to be determined \[[@B246-jof-03-00035]\]. Several innate immune cell types appear to have effector functions that facilitate *C. neoformans* clearance and prime adaptive immune responses under certain conditions, but the mechanisms that coordinate these processes require further definition. Much of the work on anti-cryptococcal immunity has been performed in vitro, so it will be important to confirm these pathways in vivo and in human hosts, when possible. Since *C. neoformans* is equipped with unique virulence factors, like its polysaccharide capsule, that enable it to evade or subvert the host immune response \[[@B1-jof-03-00035]\], it is not unexpected that the fungus would stimulate distinct innate immune responses compared to other fungal pathogens. Thus, while it is important to study the potential roles of established antifungal pathways in the response to cryptococcosis, it is also critical to work towards identifying immune mechanisms that may be specific to *C. neoformans*. Identification of additional patient populations susceptible to cryptococcosis, such as those with anti-granulocyte macrophage colony-stimulating factor (GM-CSF) autoantibodies \[[@B247-jof-03-00035],[@B248-jof-03-00035]\], may reveal previously unknown immune processes important for the host response to *C. neoformans*. Additionally, the rise of new bioinformatics approaches like next-generation sequencing \[[@B249-jof-03-00035]\] and tools like CRISPR-Cas gene editing \[[@B250-jof-03-00035]\] and fluorescent probes \[[@B251-jof-03-00035]\] may enable the discovery of novel pathways in anti-cryptococcal immunity. The author thanks Tobias Hohl and Bing Zhai for helpful comments and regrets that the valuable contributions of many other researchers in the field could not be included due to space constraints. Lena J. Heung is funded by a National Institutes of Health (NIH) grant K08 AI130366, a Stony Wold-Herbert Fund Fellowship, and the Dana Foundation as a Memorial Sloan Kettering Cancer Center Clinical Scholar in Biomedical Research. This work was funded in part through the NIH/National Cancer Institute (NCI) Cancer Center Support Grant P30 CA008748. The author declares no conflict of interest.

CASE REPORT {#s1} =========== A previously healthy 35-year-old Brazilian woman presented with persistent redness, swelling, and multiple wounds on the right hand 2 weeks after a cat bite in her home country. While in Minas Gerais, Brazil, the patient had reached out to touch a wounded cat (possibly wild) and was bitten on the right hand. Soon after, she developed erythema, oedema, and pain. She was evaluated at a local hospital where she was prescribed 10 days of amoxicillin-clavulanate. She also received rabies vaccine and rabies immunoglobulin. She completed the 10-day course of antibiotics, after which she experienced an initial improvement in the hand swelling and erythema. Her symptoms returned 2 weeks after the attack. She then traveled to Boston, where she was seen in an emergency department. She was diagnosed with cellulitis and discharged with another 7 days of amoxicillin-clavulanate. She completed the antibiotics, but her hand swelling and pain progressed. The patient then presented to our institution. Her vital signs were normal. She appeared generally well. Puncture wounds were noted on the dorsal aspect of the right hand, with wounds on the third and fifth digits ([Image 1](#I1){ref-type="fig"}). Erythema and fluctuance were present and extended into the forearm. Range of motion was preserved in the hand and wrist joints. There was no palpable lymphadenopathy in the right upper extremity or axilla. Notable laboratory tests included the following: C-reactive protein 35.0 (reference range, 0--5 mg/L), erythrocyte sedimentation rate 72 (reference range: 0--20 mm/hour), and white blood cells 12.6 (reference range, 4.0--11.0 K/uL). An x-ray of the right hand showed soft tissue swelling at the third and fourth proximal phalanx. Osteoarticular infection was ruled out with x-ray as well as orthopedic surgery evaluation. {#I1} Incision and drainage were performed, and a wound culture was obtained. The patient was then admitted to the hospital for intravenous antibiotics. She received treatment with levofloxacin and clindamycin. A second incision and drainage were performed on the right hand on hospital day 4, after which her symptoms improved. She was discharged with a prescription for metronidazole and doxycycline. Four days after discharge, the wound culture resulted positive for 1+ (Rare) *Sporothrix (Sporotrichum) schenckii*. Identification was performed by matrix-assisted laser desorption ionization time-of-flight mass spectrometry. The patient was seen in the infectious disease clinic on day 6 after discharge. On that day, No improvement was identified in the hand wounds despite antibiotic therapy. Given the new diagnosis of sporotrichosis, she was prescribed itraconazole 200 mg once daily. Over the subsequent months, her hand oedema and wounds improved dramatically ([Image 2](#I2){ref-type="fig"}). {#I2} Discussion {#s2} ========== Sporotrichosis is a subcutaneous mycosis that infects both humans and animals. Sporotrichosis is endemic in many South American countries including Brazil, and it is considered the main subcutaneous mycosis in that region \[[@CIT0001]\]. The first recognition of sporotrichosis in Brazil was in 1907 \[[@CIT0001], [@CIT0002]\]. In the state of Rio de Janeiro, there were approximately 2200 human cases reported from 1998 to 2009. Cats play a major role in transmitting this disease to humans by bites and scratches because they carry the fungus in their claws and oral cavity \[[@CIT0003]\]. A case series conducted in Rio de Janeiro reported that 90.7% of 172 people with sporotrichosis had contact with cats \[[@CIT0003]\]. The most common clinical presentation is the lymphocutaneous form of disease (55.6%) \[[@CIT0004]\]. Although *Sporothrix brasiliensis* was the primary involved species in the outbreak in Rio de Janeiro, *S schenckii* was also thought to be transmitted in Brazil \[[@CIT0001]\]. Considering the ongoing sporotrichosis epidemic in South America, patients who are coming from endemic areas presenting after cat bite with cellulitis should have this illness included on the differential diagnosis. Conclusions {#s3} =========== Sporotrichosis is yet another example of an emerging infectious disease due to increased contact between humans and animals. Implementing a "One Health" approach is essential to achieving better public health outcomes \[[@CIT0005]\]. This can be achieved through considering the sporotrichosis epidemic as a priority for clinicians, public health practitioners, and the general public. The sporotrichosis epidemic can be viewed as a "canary in the coal mine," a concern due to the anthropogenic forces of urbanization, globalization, and industrialization \[[@CIT0006]\]. Medical education should increase its emphasis on emerging infectious diseases. ***Potential conflicts of interest.*** All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.

{#sp1 .400}

Introduction ============ Background ---------- Globally, health care systems have to deal with the exponential growth of the aging population, adding strain to health care service provision while still having to commit to achieving the sustainable development goal of the right to good health for all ages \[[@ref1]\]. There is a general shift of the population toward older age being referred to as population aging \[[@ref2]\], who not only live longer but also deal with complex interrelated factors related to their quality of life (QoL) \[[@ref1]-[@ref3]\]. At the same time, fast technology advances add pressure to the development of relevant technology-enabled health care services and the potential mediating role that technology could play toward the QoL of the older person \[[@ref1]\]. The exponential growth of both the aging population with the complexity of health states and fast technology advances results in health care services struggling to deliver quality health services to maintain the human dignity of the aging population \[[@ref4]-[@ref6]\]. The needs and preferences of older people to enhance their intrinsic capacity to negotiate their own changing world toward new ways of functioning as part of healthy aging are complex \[[@ref2],[@ref7]\]. Design thinking offers an approach that deals with complexity with a focus on empathy, context, ideation, and iteration as part of designing human-centered services \[[@ref3]\]. The danger of fast technology advances is that they could potentially widen the gap between younger and older population groups, especially considering the diversity and multitude of cognitive and physical abilities and health status associated with aging groups, which may make adoption of technology more difficult \[[@ref3]\]. Efforts should be made to include the older users in the service design and implementation processes, owing to their emergence as important consumers of services rather than merely regarding them as passive recipients of technology \[[@ref1],[@ref8]\]. The design of fit for purpose service should focus on integrated care that is person-centered and empowers the care recipient for the older user to continue to contribute to society \[[@ref7]\]. As a recipient of care services, it is important to understand the older users' experience in their own situated context \[[@ref9]\]. *Gerontechnology* is a concept for considering the impact of technology on the QoL of the aging population, where technology has the ability to enable services \[[@ref6]\]. When considering the relationship between technology and aging, it is important to regard older persons as active consumers of technology-enabled services and active cocreators of technology during the design process \[[@ref8]\]. Many gerontechnology research studies deal with the design processes and methodologies \[[@ref8]\]. Both participatory design and codesign approaches speak to the need to include participants in the exploration of a design challenge and the subsequent design process. One of the challenges older individuals face is that technology has not always been designed with their specific wants or needs in mind \[[@ref10]\]. A participatory design approach can bridge this challenge by including older persons in the design process of health and well-being technologies and services \[[@ref11]-[@ref13]\]. Studies on the implication of Internet of Things for future health care services and devices for older individuals position a participatory approach as critical \[[@ref14],[@ref15]\]. As technology develops, there remains a need to explore the complexities of technology use in health care services intersecting with the older user \[[@ref3],[@ref16]\], as well as the role of the older user in the design process \[[@ref6]\]. It is also necessary to unpack the inherent tension among design research that allows for iterations, ambiguity, rapid prototyping, and health research that is hypothesis-driven and where evidence-based research is the norm \[[@ref3]\]. This paper attempts to contribute toward the understanding of the design of technology-enabled services for the older user. The aim of this study was to propose a theory that can be used when designing health and well-being services with the aging population as a target group. Findings from the study contribute a substantive theory to the service design body of knowledge, which explains the engagement of older individuals with Web-based services (including services that support health and well-being). The study adds value to Web-based health care service design practice by developing a deeper understanding of user perceptions and experiences within a sociotechnical context. Designing for an Aging Individual --------------------------------- Decreased physical mobility, eyesight, and cognitive processing may impact the QoL of older individuals. These individuals require support from family, friends, or the community to complete basic daily tasks and activities relating to their health and well-being. Technology and devices that support health care could offer alternative solutions to these challenges. These technologies can range from services that enable increased socialization (social media and communication applications) to health monitoring devices and emergency notification services. The Web-based market for an aging population has been an area of research interest for a long period, but it has not yielded many insights into user-driven design in collaboration with older users \[[@ref17]\]. This may be influenced by how aging users are viewed by both the service and goods providers, as well as the developers of health care and well-being technologies and services. A number of recent software solutions, apps and devices reflect the spectrum of interaction with aging individuals during the conceptualization and development of interventions and supportive solutions. These range from including aging individuals as end-user testers instead of cocreators \[[@ref18]\] to collaborating with those close to the aging person (family and caregivers) \[[@ref19]\]; finally, these also span to projects that include aging individuals in the process as cocreators \[[@ref20]\]. The influence of age on the likelihood of engaging with technology is less extreme than once imagined \[[@ref6],[@ref21]\], but the nature of, and influences on, the engagement of an older individual has been noted. Before the potential impact of health care technologies and services on the lives of aging individuals can be understood, the usage and perceptions of the aging population must be explored to identify possible barriers to participation. From an economical perspective, it is crucial to consider the global growing aging community within our technological society \[[@ref1],[@ref22]\]. If not considered during the conceptualization and design of products and services, it can be hypothesized that aging users who do not feel confident on use of the Web would cease to use the services that could possibly improve the support for their everyday activities and offer them greater independence \[[@ref8],[@ref23]\]. Technology to Support Aging in Place ------------------------------------ Gerontechnologies offer technologies and innovations specifically designed for an aging community \[[@ref6],[@ref24],[@ref25]\]. Another definition offered by Iffländer \[[@ref26]\] is *age-based innovations*, which are defined as products and services specifically designed to acknowledge the needs of older users. Irrespective of the assigned title, these products and services focus on enabling an aging community to remain autonomous and contribute to a greater sense of well-being (including increased social engagement). A possible contribution of supportive technology and access to Web-based services is facilitating aging in place (AiP) of aging individuals. A lack of understanding specific user needs can have a major impact on the innovation of various technologies that can support AiP \[[@ref27]\]. This understanding must include a review of human factors that consider an individual's limitations, his or her capabilities, as well as personal and cultural contexts \[[@ref28]\]: > Only when the real needs of the elderly are correctly understood by innovators, fully specified in AiP digitalization, together with stakeholders' inclusion in the innovation process and proper consideration of human factors and other contextual factors...can then ensure the success of AiP implementation \[[@ref26]\]. To encourage continued engagement among aging users who can benefit from technology that facilitates AiP, the nature of their Web-based interaction, as well as the process of learning how to engage, must be considered. Aging individuals and those who care for them will embrace technological products and services that support AiP \[[@ref20]\]. Methods ======= Grounded Theory Method ---------------------- Grounded theory (GT) emerged from the research and practice of Glaser and Strauss in 1967 \[[@ref29]\]. Since then, 3 main streams of GT have developed. The first represents the original ideals of Glaser (often referred to as Glaserian GT), the second variation of the method was conceptualized by Strauss and Corbin (in response to the earlier Glaserian variation), and finally, Charmaz's constructivist GT \[[@ref30]\]. At the heart of each GT stream is the exploration of real-world situations through rigid analysis and documentation to gain insights, and it is not based on preconceived ideas or assumptions \[[@ref31]\]. The types of data collection tools vary, but qualitative methods, such as in-depth interviews, are prominent. The information gained is analyzed through coding processes, followed by making sense of the complex data and finally coming to a cohesive theory grounded in the data. The process aims to conceptually explain how participants respond to a certain concept, phenomenon, or challenge \[[@ref32]\]. The development of the theory is based on 3 foundations: constant data analysis (where data collection can happen simultaneously), theoretical saturation (data are collected and analyzed until nothing new is discovered), and theoretical sampling, which facilitates the emergence of theory \[[@ref33]\]. Recruitment and Participants ---------------------------- Participants were recruited through a gatekeeper organization, which has a broad reach throughout the Western Cape and South Africa and attracts individuals from varying ethnic and socioeconomic backgrounds. Presentations were made to members to introduce the project and highlight the parameters of participation. The parameters were that the participants had to be over 65 years and have access to Web-based services through a personal or shared device. The study did not explore problems with connectivity or internet penetration within South Africa. Following the presentations, 23 individuals requested to collaborate on the study, out of which 13 were interviewed ([Table 1](#table1){ref-type="table"}). The 2 workshops were hosted with 15 participants, out of which 5 participated in the interviews before the workshop. Not all of the interviewed participants joined the workshops. This study thus featured a convenience sampling with regard to the open call for participation; however, all participants met the defined project parameters. For this reason, the sampling method aligns itself with the traditional theoretical sampling one would expect in a GT study. Ethical considerations were a primary focus in the study, and recruitment was dependent on the participants' ability to give informed consent. Furthermore, 3 aspects contributed to the concept of informed consent practiced in the project: (1) participation was voluntary, (2) the nature of the project (including all benefits and risks) was explained before commencement of the research activities, and (3) participants' consent was *valid*. The validity of consent was defined through the work of Ratzan \[[@ref34]\], which proposes that, "...the elderly research subject\'s actual understanding of the experiment be accurate and complete..." Ethical considerations in the project aimed to establish an empathetic grounding for interaction and placing the participant at the center of any consideration or project decision. The research process was initiated with an open discussion with participants, during which the goals of the project were introduced along with research activities. Participants had the opportunity to discuss their concerns or excitement, from which comments were noted down on the *participant coding and information form*. The form was part of a research process map and toolkit document, which supported the gathering of informed consent and first stage research observations ([Multimedia Appendix 1](#app1){ref-type="supplementary-material"}). To quote interviewed participants directly, in an anonymous manner, they were asked to select their own pseudonym. This allowed participants to select the name they wanted to be known by and shifted participant identification protocols associated with anonymity and research ethics into a more human-centered realm. ###### Participant group and research activity participation. Participant code (PC) Pseudonym selected by participant (only for interviewed participants) Age (years) Gender Device type Participation in research ----------------------- ----------------------------------------------------------------------- ------------- -------- --------------------------------------- --------------------------- PC1 Victoria 76 Female Mobile phone, tablet, desktop Interview and workshop PC2 Marie 72 Female Mobile phone, desktop Interview PC3 Convict1 85 Male Mobile phone, tablet, laptop Interview PC4 Cordelia 82 Female Mobile phone, desktop Interview PC5 Susan 74 Female Mobile phone, desktop Interview and workshop PC6 Ina 82 Female Mobile phone, laptop Interview PC7 Doll 65 Female Mobile phone, tablet, desktop Interview PC8 Hellet 78 Female Mobile phone Interview and workshop PC9 Cody 76 Female Mobile phone, desktop Interview PC10 Diana 69 Female Mobile phone, desktop Interview PC11 Leonie 71 Female Mobile phone, tablet, desktop, laptop Interview PC12 Dick 69 Male Mobile phone, desktop Interview and workshop PC13 Taffy 74 Female Mobile phone, desktop Interview and workshop PC14 N/A^a^ 68 Female Mobile phone, tablet Workshop PC15 N/A 66 Female Mobile phone, desktop Workshop PC16 N/A 83 Female Mobile phone Workshop PC17 N/A 65 Female Mobile phone, laptop Workshop PC18 N/A 67 Female Mobile phone Workshop PC19 N/A 71 Female Mobile phone, desktop Workshop PC20 N/A 77 Female Mobile phone, laptop Workshop PC21 N/A 70 Female Mobile phone, desktop Workshop PC22 N/A 77 Male Mobile phone, tablet, desktop Workshop PC23 N/A 79 Female Mobile phone, desktop Workshop ^a^N/A: not applicable. Data Collection and Initial Analysis and Coding ----------------------------------------------- The project followed a systematic process with regard to the collection and analysis of data in-line with GT. Data collection and analysis happen simultaneously in GT. The process of data analysis progressed through 3 main analysis cycles: initial coding, focused coding, and finally, theoretical coding. The initial coding cycle is rooted in iterative data collection and analysis. Once no new codes emerge as new data are analyzed, the focused coding and theoretical coding processes are completed. Data analysis was completed through line-by-line review and coding. ATLAS.ti, a Computer Assisted Qualitative Data Analysis software, was used to collate and analyze the data gathered. A key aspect of the software that assisted in the analysis of the data was its ability to code both abstractly, as well as in vivo. Using in vivo codes helps preserve participants' reactions and the meanings of their views and comments. The data collection and analysis process mirrored the *double diamond* \[[@ref35]\] design process ([Figure 1](#figure1){ref-type="fig"}). In the first diamond, 2 phases were completed: (1) the gathering and analysis of interview data and observational notes, followed by (2) the reflection and code review phase. Interviews were a critical method in this study as they allowed the researcher the opportunity to listen and reflect on experiences of older individuals. Participants could decide on where it was most convenient for them to meet, resulting in a number of interviews being conducted in participants' homes (10 in total); however, the remaining 3 interviews were conducted in a café. Participants were interviewed by a single researcher, who recorded the whole interview for transcription. To encourage a more open approach to interviews, participants were given contact details (email and phone number) to allow them to stay in contact with the researchers. If they wanted to share more information at a later stage, or ask a question, they could engage directly with the researcher. {#figure1} At the interviews, participants were asked to respond to the following questions: - How do you use Web-based services in your everyday life? - How do you feel about technology and Web-based services? - What, if any, are the challenges you have noted when engaging with Web-based services? - Would you want to engage with technology and Web-based services? If yes, what would you like to see and how would you want to engage with devices and Web-based services? The driver of each interview was to develop a narrative around the participants' perceptions and experiences, and thus the questions were not considered as fixed. The interviewer remained open to conversational shifts as participants shared relevant stories. The process of *initial* coding during the analysis of 13 interview transcripts and observational notes was iterative, in that new data were compared with reviewed and previously coded data. Interviews ranged from 65 min to 93 min in length. The emerging codes from the interview data analysis were seen as *provisional*, as they changed and evolved as more data were coded and compared. In the initial coding phase, the emphasis was on extracting data from interviews, comments, and observations captured during interactions with participants. The open process of initial coding yielded 155 individual codes. Following the code review, the number of codes decreased to 130 codes. This was the result of code-to-code analysis and the merging of conceptually similar codes. The second diamond also comprised 2 phases: (1) the gathering and analysis of workshop data (transcripts, observational notes, and workshop materials), followed by (2) the reflection and code review phase. In total, 2 workshops were completed, one with 7 and another with 8 participants. The workshops included 3 activities. The first activity was an introduction to a range of Web-based services, demonstrated on different devices and operating systems (Apple, Windows, and Android). At the interactive demonstration session, participants shared their own experiences. The second set of activities focused on the creation of a collaborative persona. The collaborative persona template ([Figure 2](#figure2){ref-type="fig"}) served as a discussion catalyst, probing the groups of participants to respond collaboratively. Several key factors were highlighted through the activity. The discussions were recorded for transcription and analysis. The template and group discussion recording allowed for the gathering of experiences and perceptions from 3 distinct perspectives: the person, the person and technology, as well as the person and Web-based services ([Figure 3](#figure3){ref-type="fig"}). Each workshop group was facilitated by a design researcher who observed interactions and posed probing questions during group discussions and cocreation of personas. The paper-based personas were digitized following the workshops ([Multimedia Appendix 2](#app2){ref-type="supplementary-material"}) and shared with workshop participants. The final workshop activity focused purely on the participants' needs. All facilitators and technical support worked with individual participants to answer questions regarding Web-based experiences, set up access where needed, or offered device support where possible. Following the workshops, *initial* coding continued. Workshop materials (in the form of 4 cocreated personas), discussion transcripts for each of the 4 working groups, and observational notes were analyzed and coded. During the analysis and coding of data from the second workshop, no new codes emerged. The analysis and coding of workshop data were followed by a review of the captured initial codes. As part of the review, each code was assigned an introductory word or phrase, which identified the larger focus area and the theme of which the code was a part ([Table 2](#table2){ref-type="table"}). {#figure2} {#figure3} ###### Code groups resulting from code review. Introductory word or phrase Code group description ---------------------------------- ------------------------------------------------------------------------------------------------------------------------------------ Cognitive impact Codes relating to the positive and negative impact of using technology and Web-based services on cognitive functions. Design Codes relating to the impact of design and interface on the user experience of technologies and services. Emotive response Codes relating to both positive and negative emotions experienced by participants when engaging with technologies and services. Infrastructure and affordability Codes relating to the affordability of Web-based access and the infrastructure that supports it in South Africa. Learning facilitation Codes relating to the method, nature, and process of learning as experienced by the users. Nature of the internet Codes relating to the scope and nature of the internet. Perception Codes relating to how users perceive technologies and services: the benefit, value, challenges, and expectations. Physical and digital Codes relating to the intersection of the physical and digital space and when participants identified comparisons between the two. Privacy and security Codes relating to concepts of safety on the Web, Web-based threats, and privacy. Reason for adoption Codes relating to reasons for adoption of technologies and services. Using services Codes relating to actual technologies and services being used and the experience of specific service elements. Interaction Codes relating to all aspects of user interaction with technologies and services from both a positive and negative perspective. Subsequent to the code review, a final list of 135 codes was defined. The code review process informed the next data analysis phase within the grounded study. The focused coding process critically reviewed each code and the emerging code groups, in relation to the theme identified throughout the data gathering process. The focused coding process resulted in a set of core categories. Focused and Theoretical Analysis and Coding ------------------------------------------- Through a systematic review and analysis of codes (in relation to original data, initial codes, and identified themes), the focused coding process resulted in a set of core categories. All initial 135 codes were absorbed within the emerging core categories, and they reflected a more conceptual connection and relationship of themes noted during the code review. The core categories descriptions are grounded in the original comments and perceptions shared by participants. Through the coding process, the data gathered were deconstructed into essential elements, reviewed for patterns and relationships, and then constructed into more complex concepts. The core categories are *digital context*, *cognition and learning*, *emotive response to Web-based interaction*, *user context*, *perceived benefits*, *nature of user interaction,* and *design to support use*. ### Digital Context Aging users viewed technology as products and services that allow access to both convenience and information. Participants felt that products, apps, and services were not designed specifically with aging users in mind. Factors noted in this study that impact user engagement and willingness to interact relate to personal perception, emotional responses, and how each individual views his or her own ability to learn and master new technology. ### Cognition and Learning Diverse learning networks were noted in the study. When new skills, ways of working, and ways of learning are introduced, users must adapt their cognitive understanding to engage with the process. Cognitive ability plays a role not only in an aging user's willingness to engage but also in the long-term usage. Olphert and Damodaran \[[@ref23]\] hypothesize that the difficulties that aging users face on the Web may be because of the complexity of technology combined with the cognitive load required to engage actively. In this study, participants mentioned 3 other main learning scenarios: learning from peers, learning from family members, and exchanges with external individuals (such as technical assistants at shops). The engagement with others resulted in mutual learning, during which the aging individuals expanded their knowledge and skill base, whereas those they interacted with developed a keener understanding of the challenges and opportunities that aging presents. ### Emotive Response to Web-Based Interaction Both positive and negative emotions were observed when participants discussed engaging with new technologies. One of the key emotional triggers noted by participants was that they often felt *forced* to engage with new technologies and services. The sense of frustration that they felt was heightened if their immediate network either could not or chose not to offer support and tutelage. Hellet (a pseudonym selected by the participant) noted the following: > I really think I need to go for lessons. My daughter always says "what lessons? Just sit and try" like I'm not trying in the first place. These early reactions to the use of a Web-based service can influence the way that older users experience the technology as a whole. ### User Context The requirements of aging users are often overlooked by developers and vendors as they do not play an active role in the conceptualization or design of Web-based services and other technologies \[[@ref36]\]. Participants in this study acknowledged that they should play a greater role in becoming computer and device literate, to enable them to engage with new technologies and services with more confidence. Cordelia (a pseudonym selected by the participant) noted the following: > Programs are often designed by people who are too clever. Sometimes, I don't know what's going on, then what do you do? Explain to people. As we work, we learn\...I need to learn by doing it myself. Acknowledging the unique personal context of an aging user is a key factor in facilitating Web-based engagement. The impetus to engage, the perceived value of Web-based engagement, and the design characteristics of an enabling environment must all be considered. ### Perceived Benefits The rise of Web-based activity among aging users is linked to their perception of services, technologies, and apps as tools to support everyday activities. One participant (with the pseudonym Victoria) noted the following: > So the value is when you have limited mobility, when you\'re housebound...You could have your Skype, which you can use to check in with your neighbors. If you are housebound and you can contact your friends or family. You can phone your neighbor. So the access is absolutely brilliant. You have unlimited access, you can access anybody, your children, anybody and let them know your situation. In the context of an everyday tool, supportive devices and services can offer users clear value propositions. Digital technologies and Web-based services play a major role in addressing the so-called *burden of care* often associated with an aging population \[[@ref23]\]. Participants stated that Web-based access could benefit an individual by allowing access to emergency services, lessen feelings of loneliness and isolation, and benefit individuals with limited mobility. Access and support provided through technology could extend aging individuals' ability to age in place, allowing them to age in their own home even though various physical, cognitive, or social capacities may decline with age \[[@ref26]\]. ### Nature of User Interaction Once the perceived value of the interaction is high enough for an aging user to engage, even though he or she may have negative perceptions, the nature of the interaction space and platform may still impact continued engagement. In many cases the emotions noted included frustration at one's own inability to understand and successfully navigate a Web-based service and supportive devices, the fear of the unknown, and feelings of frustration when forced by family or community members to engage with digital technologies. These emotional reactions form part of the human experience of a technological interaction. Digital interaction and the design of this interaction may influence the user experience, as well as the process through which the user learns to navigate the interaction. If not considered during the conceptualization and design of Web-based services, it can be hypothesized that aging users who do not feel confident would cease to use the services that could possibly improve support for their everyday activities \[[@ref23]\]. This result has led researchers to define a *fourth digital divide*, which is not characterized by a lack of access or skill but rather by a lack of clear motivation or interest \[[@ref23]\]. ### Design to Support Use Using devices and services does pose a number of challenges to aging users, including limited experience with devices, limited experience with Web-based services, or a lack of interest \[[@ref37]\]. It is not only the design of the system that needs to support interaction and learning but also the design of interfaces and interactions. The applied practice of designing for aging users and creating more accessible design have received much attention from various projects and researchers over the past decades \[[@ref37]-[@ref42]\]. In addition to individual researchers, various projects have investigated accessibility from an aging perspective, including the *Web Accessibility Initiative: Ageing Education and Harmonisation* \[[@ref43]\]---a European Commission Specific Support Action project (Information Society Technology 035015). Overwhelmingly, participants stressed the need for design elements that favor simplicity; however, this may not be an easily generalizable standard. One of the primary reasons for ending one's use of Web-based services has been described as *excessive complexity of the technology* \[[@ref23]\]. Participants said that they preferred clear action buttons and a simple style of communication. During the focused coding process, memos allowed the researcher to keep track of considerations and decisions. The writing up of memos also provides for a structured moment to pause and reflect on the process, encouraging a holistic reflective practice. Finally, the process of theoretical coding responded to the 2 coding aims stated by Flick \[[@ref44]\]. First, the process aims to clearly understand and explore the research context or question. Second, it identifies the relationships among categories or components extracted from what was found. During the theoretical coding process, the emergent core categories were reviewed from a structural perspective to understand the relationships between categories and the nuances at play when considering core categories as part of a whole and not individual sectors. The result of the theoretical coding process was the substantive theory, which explored the process and factors that impacted an aging individual's willingness to engage with technology and services. Saturation was explored at data level (when no new codes emerged through constant comparative data analysis) and again at theoretical level (when the emerging theoretical constructs, literature, and memos were compared and yielded no new variations). Through the process of theoretical coding, these categories formed the foundation for the emerging theory of *Ageing User Decision-Driven Engagement* (AUDDE). Results ======= The result of the study was a grounded, substantive theory that explores the factors, and process, of aging individuals engaging with services on the Web. The term *cyber-seniors* refers to 2 emerging groups of aging users, the first is that of the *technology lovers* (who engage willingly and are fascinated by technology) and the second is that of the *technology users* \[[@ref45]\]. Technology users see technology as a tool to achieve a specific goal. Participants who took part in this study were overwhelmingly *technology users*. Only 1 participant in the study could be defined as a technology lover; all other participants commented on the *tool* nature of Web-based services or their ability to *help you do something*. AUDDE ([Figure 4](#figure4){ref-type="fig"}) highlights the decision-determined engagement that characterized the aging technology users within this study. The theory proposes an iterative process in the decision-making cycle. When aging users decide to engage with a supportive device, app or service, or decide not to engage, 2 main factors form the basis for their ultimate decision. These factors are defined as *perceived benefits* and *Web-based user context*. For a user to engage, the perceived benefits must outweigh any hesitation that forms part of the Web-based user context. The perceived benefits must be made clear through a value statement. This value statement could be found in the form of an advertisement, but among the aging, it will more likely be word of mouth or the suggestion of a trusted medical professional or a similar individual. Family members and peers are the most likely candidates to share a value statement with an aging individual. Once the individual is aware of the perceived benefit and value that an engagement may have, the decision to interact is dependent on the level of resistance within the user context. The Web-based user context is shaped by 2 spheres of influence, the *social context* and the *use context.* The first sphere of influence that impacts the decision to engage is the user's *social context*. Here, the social context refers to the perceptions of the aging individual's social group, communities of interest, and communities of practice with regard to the technology. Within these social constructs, the aging individual may feel pressured to share the communal peer point of view of his or her social groups. The view of individual family members or friends has a similar ability to shape the perspective of an aging individual. The emotional and social influences on the willingness of an individual to interact with a supportive technology or service are crucial to potential engagement. Equally crucial is the emerging and constantly evolving *context of use*. The use context of aging individuals is informed by every interaction they have had with the device or service in question, and it may even include a broader range of technological interactions. Every interaction contributes to an individual's perception of ease of use, cognitive demand, convenience, and overall advantage. In this way, the use context is constantly evolving. The context of use is formed through a process that occurs when the user has decided to interact and revolves around 2 process points, the *outcome of the interaction* and *reflection on interaction*. {#figure4} The outcome of the interaction is a process of evaluation that the user completes at the end of a task. The task does not have to be completed successfully, or even completed at all, for the outcome of the interaction to be evaluated. The design of technologies, devices, and services plays a pivotal role in user interaction. Services that take into account different levels of physical ability and focus on enabling simple and specific tasks are of greater value to aging users. Design choices that create an enabling experience elicit a positive emotional response from users. These emotions include pride in one's ability to complete the task unaided, a heightened sense of accomplishment, and joy. Given the nature of the activity, the completion of the task may also elicit a sense of relief. A design that restricts task completion elicits feelings of frustration and confusion. The emotional reaction to a task can often permeate the entire interaction. If aging users, for example, struggle to switch on the device or use it as intended, their initial feeling of minor annoyance can be aggravated by subsequent challenges. The emotional reactions experienced throughout the interaction affirm or challenge perceptions which the user had before. A positive experience may challenge a personal resistance to engage or strengthen previous beliefs that engagement has value. A negative experience may call into question previous perceptions relating to the value of engaging or reinforce previous resistance to engagement. Linked to the emotional experience of aging users, while interacting with a Web-based service, is the cognitive experience and the potential for learning. Both physical and cognitive decline are key markers of the aging process; however, both manifest in varying degrees and escalate differently among the aging community. Products and services, which offer guidance in the case of possible lapses in user memory or provide a step-by-step task guide, support the unpredictable nature of cognitive ability among this user group. The learning process for aging users is mediated by either an external stakeholder (a friend, a family member, or a professional encountered when seeking help) or self-exploration. These learning experiences may happen before engagement or may take the form of coached support while using a device or service. How an aging user interacts, the cognitive experience of the interaction, and the emotional consequences and resulting learning process are all interwoven elements of an iterative cycle. A user may go through multiple cycles within this sociotechnical experience before reaching the end of the interaction. The outcome of the interaction could be a successful task completion, an unsuccessful task completion, or an interrupted task (ended before either a successful or unsuccessful task completion). Irrespective of the nature of the outcome, the experience of the interaction will result in the user's reflection on the perceived value expectation in relation to the interaction experience. This process will inform the future *use context*. Discussion ========== Involving Users in the Design Process ------------------------------------- When reflecting on the emerging AUDDE theory, the complex nature of the networked society becomes clear. Knowledge and experience within the Web-based and digital service realm transcend single disciplines to create service systems: > Actual service systems can be described as complex sociotechnical systems, being approached in an interdisciplinary vision that integrates business functions, technology, and human resources, with the final aim of creating value and benefit through the generated services. > > 46 Acknowledging this complexity is crucial when conceptualizing and designing devices, technologies, app, and Web-based services, which aim to support the health and wellness of aging individuals. Researchers and designers must systematically review both the technical and social systems that are at play during an interaction \[[@ref47]\]. The technical systems contribute engineered interaction spaces that are designed to be anticipatable and reliable. The social systems are in many ways dependent on the technical systems and evolve throughout interaction encounters. The fast pace of technological development today requires that we endeavor to gain a greater understanding of social systems, to navigate new sociotechnical interactions as they emerge \[[@ref48]\]. End users are social beings, who evolve and grow. As such, it is impossible to define social value as a constant, and the social impact on interactions must be considered within the *more institutionalized traditions or regulations inside various user communities* \[[@ref49]\]. To understand sociotechnical interactions, it is important to understand that the technology and social context develop reciprocally \[[@ref50]\], as well as know the value system that these relationships represent. In AUDDE, the perceived value of the interaction is a crucial catalyst for engagement. Aging users continuously make meanings of their experiences, which affect their current and future actions. Limitations ----------- It must be emphasized that AUDDE is firmly grounded in the perceptions and shared stories of a specific group of aging users living in Cape Town, South Africa. The relevance of this theory to aging users' experiences outside this geographic area must be verified. The regularity of participants' use of supportive technologies was not a consideration within the delimitations of the study. AUDDE is a substantive theoretical contribution to the body of knowledge of design, but it has the potential to be explored within the context of other studies. To ensure that the emerging theory represented the views and shared experiences of participants, an interactive discussion session was conducted with a participant who took part in the interview and workshop. The session was discussion driven and presented the research process and emerging theory before requesting feedback and comments from the participant. The feedback supported the emerging theory, but feedback from a larger audience will be beneficial. Conclusions ----------- As the global health care systems become strained with growing aging populations \[[@ref1]\], technology may facilitate alternative ways of engaging with and supporting older individuals. Health and well-being products and services draw on the potential of Web-based interaction and technology to support independence and AiP. Product and service design projects and initiatives encourage contributions from aging users to varying degrees. Some projects position older individuals as final validators of a product or service. In this case, input and feedback from the older individuals are gathered subsequent to the development of the artifact or service. Projects that focus on a human-centered approach position older individuals as central to the design and conceptualization process \[[@ref3]\]. This approach allows for continuous feedback and interaction between the design team and the envisaged end users. AUDDE contributes to the theoretical body of knowledge in design and aims to explain how and why aging users engage with technology. The theory proposes an iterative cycle of evaluation during which users decide to engage when the perceived value of the interaction is greater than the perceived challenge of interacting. If health care products and services are to be contextually relevant and thus viable, aging individuals must be included in the design process. Authors\' Contributions: VDP recruited the participants, collected the data, conducted the analysis, and formulated the theory. RDLH supervised the study. VDP and RDLH drafted the manuscript and approved the final version. Conflicts of Interest: None declared. Research process map and toolkit. Digitized collaborative personas. AiP : aging in place AUDDE : Ageing User Decision-Driven Engagement GT : grounded theory QoL : quality of life

Introduction {#S0001} ============ The fourth Millenium Development Goal (MDG 4) calls for efforts to reduce child mortality by two-thirds by 2015. Although there has been a decline in child mortality in recent decades in many regions of the world, about 7 million child deaths occur annually in the world \[[@CIT0001]\]. The highest rates of child mortality occur in sub-Saharan Africa - where 1 in 9 children die before age five. The rate of decline in the region is also slower and most countries are unlikely to meet MDG 4 \[[@CIT0001]\]. Ethiopia, one of the countries with high child mortality, has shown a remarkable decline in child mortality in the last decade. The under-five mortality rate declined from 166 deaths per 1000 live births in 2000 to 88 deaths per 1000 live births in 2011 \[[@CIT0002], [@CIT0003]\]. However, the majority of deaths are still caused by infectious diseases that can be prevented through cost-effective prevention efforts \[[@CIT0004]\]. Childhood vaccination can reduce child mortality significantly and is a cost effective way to improve child health \[[@CIT0005]--[@CIT0007]\]. According to the World Health Organization (WHO), vaccination averts an estimated 2 to 3 million deaths every year from diphtheria, tetanus, pertussis, and measles \[[@CIT0008]\]. In 2011, about 83 percent of infants worldwide were vaccinated with three doses of diphtheria-tetanus-pertussis (DTP3) vaccine, and 162 countries have reached DPT3 coverage of 80 percent. Globally, in 2011, 22.4 million children under one year of age did not receive DTP3 vaccination. Ethiopia is one of ten countries with large proportions of children without DPT3 vaccination \[[@CIT0008]\]. The Ethiopian Health Policy emphasizes prevention and control of major communicable diseases \[[@CIT0009]\]. Strengthening the Extended Program on Immunization (EPI) is therefore one of the core activities in the recent Health Sector Development Program (HSDP) to reduce child mortality. The EPI policy calls for BCG vaccine at birth, three doses of DPT-HepB-Hib vaccine at approximately 4, 8, and 12 weeks of age, four doses of oral Polio vaccine at 0-2, 4, 8, and 12 weeks of age, and measles vaccine at or soon after reaching 9 months of age. A child is said to be fully vaccinated if all eight vaccinations have been received. Although the EPI was initiated in 1980 with the goal of universal coverage by 1990 \[[@CIT0010]\], achievements are still far below the international standard. According to the 2011 Ethiopian Demographic and Health Survey (EDHS), only 24% of children age 12-23 months were fully vaccinated. The coverage of measles vaccination was 56% \[[@CIT0003]\] although service statistics by the Ministry of Health and a national EPI coverage survey conducted in 2006 showed a higher coverage than the DHS reports. The 2006 national EPI coverage survey showed that 50% of children age 12-23 months were fully vaccinated \[[@CIT0011]\]. In addition to the low coverage, rural-urban and regional disparities in vaccination coverage are substantial \[[@CIT0003]\]. The 2011 EDHS also showed that urban children were more than two times as likely as rural children to have all basic vaccinations. Previous studies have documented several maternal, social and health care provision factors that influence completion of child vaccination in low and middle income countries. Maternal education and use of antenatal care services are consistently associated with completion of childhood vaccinations \[[@CIT0012]--[@CIT0016]\]. In studies from Ethiopia and other low and middle income countries, low access to services and inadequate awareness of the roles of vaccinations were found to be barriers to completion of child vacination series \[[@CIT0010], [@CIT0014], [@CIT0017]--[@CIT0019]\]. While the association between mother\'s age and parity and completion of childhood vaccination has been less consistent \[[@CIT0014], [@CIT0019], [@CIT0020]\], two recent studies from Africa (Kenya and Nigeria) found that younger maternal age and lower parity are associated with completion of child vaccination series \[[@CIT0015], [@CIT0019]\]. Gender differences in vaccination coverage have been reported in studies from India \[[@CIT0021]\], Bangladesh \[[@CIT0022]\] and Nigeria \[[@CIT0023]\] but not in previous studies from Ethiopia \[[@CIT0014], [@CIT0020]\]. Few studies in Sub-Saharan Africa have examined associations between vaccination rates and factors such as pregnancy intention, the number of under-five children in a family and women\'s autonomy within the household. However, these demographic factors are likely to be implicated in the coverage and completion of childhood vaccination. For instance, a large body of research highlights the negative consequences of unintended pregnancy on maternal health behavior. These studies have shown that unintended pregnancies are associated with delayed intiation and inadequate use of antenatal care services \[[@CIT0024]--[@CIT0026]\], maternal depression and anxiety during pregnancy \[[@CIT0027], [@CIT0028]\], and a shorter duration of breast feeding \[[@CIT0029], [@CIT0030]\]. On the other hand, studies that assessed the association between pregnancy intention and child preventive and curative care in high income countries found no effects \[[@CIT0031], [@CIT0032]\]. A study by Marston and Cleland (2003) that used DHS data from five low and middle income countries showed significantly higher risk of incomplete childhood vaccination for unintended births in three (Egypt, Kenya and Peru) of the five countries studied \[[@CIT0033]\]. Another study that assessed the effects of unwantedness on curative care using DHS data from Indonesia showed that unwanted children were less likely to receive treatment for illness compared with wanted children \[[@CIT0034]\]. Overall, there are few studies on the subject from low income countries and existing ones have shown mixed results. Although some studies have examined the association between women\'s autonomy and maternal health care \[[@CIT0035]--[@CIT0037]\], few have examined whether a similar relationship exists in the utilization of child health services including vaccination. In the only study from Ethiopia that used the 2005 EDHS data to assess the associations of women\'s autonomy with maternal and child health care, women\'s participation in decision making was found to be significantly associated with completion of childhood vaccination \[[@CIT0037]\]. Another study that assessed the associations of women\'s autonomy with under-five mortality from Central Ethiopia found that women\'s involvement in household decision making was significantly associated with under-five mortality \[[@CIT0038]\]. This study examined the association between childhood vaccination and demographic factors including pregnancy intention, women\'s autonomy and the number of under-five children in a family in southwestern Ethiopia. Methods {#S0002} ======= Study setting {#S20003} ------------- A cross-sectional survey was conducted in the Gilgel Gibe Health and Demographic Surveillance System (HDSS) which is located 260 kilometers to the southwest of Addis Ababa (the capital) in southwestern Ethiopia. The Gilgel Gibe HDSS, which is run by Jimma University, is used to collect vital events data. The HDSS covers more than 10,000 households and a population of over 55,000 people. Sample {#S20004} ------ Women residing in the demographic surveillance area who had a live birth in the two years before the survey served as a sampling frame for the present study. The data used for this study were collected as part of a larger study on the effects of unintended pregnancy and related socio-demographic factors on maternal and child health in the HDSS. A sample size of 1,456 women was estimated for the study. Participants were drawn from eleven kebeles (smallest administrative unit in Ethiopia) in the HDSS area using simple random sampling. There were 1,370 women interviewed in the main study who gave birth to 1,382 children in the two years before the survey. A sub-sample of 889 children of age 12-24 months were eligible for the present analysis. Procedures {#S20005} ---------- Data collection took place from March to May 2012. Data were collected by ten trained female data collectors who had a diploma-level training and data collection experience. They were closely supervised by supervisors who had similar or higher level of education and experience in supervision of data collection. The data collectors and supervisors participated in 5 days of training focusing on questionnaire administration and ethical considerations. After the training, a pre-test of the questionnaire was conducted. Information from the pre-test was used to finalize the questionnaire. Data were collected using a structured questionnaire originally developed in English and translated to Oromo. Vaccination data were recorded from cards if the mother was able to present a card or reported verbally. All study participants were interviewed at their home in private area. Ethical approval was obtained from the College of Health Sciences, Addis Ababa University. Support letters were obtained from regional and district health offices. Local (kebele) administrations were informed about the study. Participants were briefed on the study and provided informed consent. Measures {#S20006} -------- The main outcome variable was full vaccination coverage of children age 12-24 months. We used the WHO definition of full vaccination which states that children are considered to be fully vaccinated when they have received a vaccination against tuberculosis (BCG), three doses each of DPT-HepB-Hib vaccine and polio vaccines, and a measles vaccination by the age of 12 months. The main explanatory variables were women\'s pregnancy intention for the index child, number of under-five children in the household and women\'s participation in household decision making. Pregnancy intention was measured using the standard DHS approach, which asks women to recall their feelings at the time they became pregnant; "At the time you became pregnant, did you want to become pregnant then, did you want to wait until later, or did you not want to have any (more) children at all?" Women\'s participation in decision making was measured by asking the following questions; "who makes decisions in your household about: (1) obtaining health care for yourself; (2) large household purchases; (3) household purchases for daily needs; and (4) visits to family or relatives?" The responses were: (1) respondent alone, (2) respondent and husband/partner, (3) husband/partner alone, (4) someone else. Women were considered to participate in a decision if they usually make that decision alone or jointly with their husbands. A composite index was constructed by grouping women into two categories: women who participate in all four household decisions, indicating a higher level of autonomy, and women who do not have any say in one or more decisions. The internal consistency of the scale, as assessed using Cronbach\'s alpha, was 0.82. Socio-economic status was measured using a household assets index derived using principal components analysis. Maternal health seeking behaviour included antenatal care, place of delivery and postnatal check up. We also included several control variables including education, wealth index, parity, and distance from health facility. Data analysis Data were analyzed using STATA software version 11. Bivariate associations between child vaccination and the explanatory and control variables were assessed using Chi-square analyses. At the multivariate level, two logistic regression models were run to identify factors associated with complete versus incomplete vaccination and receipt of at least one vaccination versus no vaccination. Variables were entered into the models based on their association in the bivariate analysis (at p \< 0.20). Adjusted odds ratio and 95% confidence intervals are reported. Results {#S0007} ======= Sixty percent of mothers were aged 25-34 years, with a mean age of 27.5 years (SD±5.4) ([Table 1](#T0001){ref-type="table"}). ###### Description of study participants (N = 889), in Gilgel Gibe, southwestern Ethiopia, 2012 Socio-demographic Variables Percent Number of women -------------------------------------- --------- ----------------- **Mother\'s age** 15-24 24.6 219 25-34 59.5 529 35+ 15.9 141 **Mother\'s marital status** Married 97.8 869 Divorced or widowed 2.3 20 **Religion** Muslim 92.5 822 Christian 7.5 67 **Mother\'s Educational status** No formal education 75.8 674 Primary 21.2 188 Secondary and above 3.0 27 **Residence** Rural 75.9 675 Urban 24.1 214 **Use of ANC during pregnancy** No ANC Visit 57.7 513 1-3 ANC visits 24.2 215 4 or more ANC visits 18.1 161 **Place of delivery for last birth** Home 88.6 788 Health facility 11.4 101 Pregnancy Intention Intended 65.0 574 Unintended 35.0 311 **Participation in decision making** Low 44.9 400 High 55.1 489 Mean number of living children 3.83 889 Almost all mothers were married (98%), most had no formal education (76%), were Muslim (93%), and lived in rural areas (76%). Women reported an average of four living children. Forty two percent of mothers made at least one antenatal care visit during their last pregnancy, but only 18% reported the recommended number of four or more antenatal care visits. Nearly nine in ten women delivered their last child at home. For 35% of the births, the pregnancy was reported as unintended. Fifty-five percent of women said they participated in all household decisions and were categorized as having high participation in household decisions. Seventy-eight percent (n = 690) of children had ever been vaccinated. However, only 37% (95% CI, 33.5-39.9) of children age 12-24 months had received all basic recommended vaccinations ([Figure 1](#F0001){ref-type="fig"}). {#F0001} Among the 690 children who had ever received vaccination, 86% received BCG, 92% received DPT1, 88% received DPT2, 70% received DPT3, and 63% received measles vaccinations ([Table 2](#T0002){ref-type="table"}). ###### Vaccination history of children age 12-24 months in Gilgel Gibe, southwestern Ethiopia, 2012 Variable Frequency Percentage --------------------------- ----------- ------------ Any vaccination (n = 889) 689 77.6 BCG 596 86.4 DPT1 634 91.9 DPT2 604 87.5 DPT3 480 69.6 Measles 433 62.8 Vaccination card 283 40.9 Note: Percentages for the vaccinations and vaccination card variables are based on the number of those who had at least one vaccination Almost half (48%) of those who had at least one vaccination received all basic vaccinations. Vaccination cards were produced for only 283 (41%) of children with at least one vaccination. [Table 3](#T0003){ref-type="table"} shows bivariate associations between demographic and social factors and child vaccination (receiving at least one vaccination and full vaccination). ###### Child vaccination by demographic and social factors, Gilgel Gibe, southwestern Ethiopia, 2012 Socio-demographic Variables Received at least one vaccination p-value Fully vaccinated p-value ------------------------------------------------ ----------------------------------- --------- ------------------ --------- **Mother\'s Age** 0.714 0.896 15-24 78.5 48.8 25-34 78.7 46.7 35 + 72.5 47.0 **Sex of the child** 0.128 0.150 Female 76.0 49.4 Male 79.3 45.1 **Educational status** 0.010 0.012 No formal education 75.9 45.1 Primary 81.4 52.9 Secondary and above 96.3 57.7 **Father\'s education** 0.003 0.01 No formal education 75.0 42.1 Some education 84.0 55.8 **Wealth tertile** 0.257 0.328 Low 74.8 46.5 Middle 78.7 44.3 Upper 79.4 51.1 **Pregnancy Intention** 0.180 0.219 Intended 78.2 49.0 Unintended 76.6 44.1 **Participation in household decision making** 0.007 0.013 Low 72.7 43.9 High 81.7 49.8 **Number of under-five children** 0.008 0.023 1 79.5 52.5 2 77.1 45.0 3 + 70.4 31.6 **ANC use** 0.001 0.008 None 68.8 42.3 1-3 visits 87.0 48.7 4 or more visits 93.7 57.4 **Place of delivery** 0.001 0.033 Home 76.1 45.7 Health facility 90.0 57.8 **Walking distance from health facility** 0.001 0.001 ≤ 60 minutes 84.9 53.7 \> 60 minutes 70.1 39.3 The analysis showed that both receiving at least one vaccination and completion of vaccination varied by mother\'s education, women\'s participation in household decision making, number of under-five children in a family, antenatal care use, place of delivery and distance from health facility (p \< 0.05). However, there was no variation in immunization (both completion and receipt of at least one) by mother\'s age, child\'s sex, wealth tertile and pregnancy intention. [Table 4](#T0004){ref-type="table"} summarizes the results of the logistic regression analyses. Mother\'s age, wealth index and pregnancy intention were not significantly associated with vaccination status. ###### Odds ratio from logistic regression predicting factors associated with child vaccination in Gilgel Gibe, southwestern Ethiopia, 2012 Variables Receiving at least one vaccination[1](#TF0003){ref-type="table-fn"}, OR (95% CI) Completing all vaccinations[1](#TF0003){ref-type="table-fn"} OR (95% CI) ----------------------------------------- ---------------------------------------------------------------------------------- -------------------------------------------------------------------------- **Sex of the child** Female Ref Ref Male 1.32 (0.94-1.88) 1.35 (1.00-1.82)[\*](#TF0001){ref-type="table-fn"} **Mothers age** 15-24 Ref Ref 25-34 0.82 (0.54-1.24) 0.93 (0.61-1.44) 35 + 0.57 (0.28-1.14) 0.97 (0.52-1.78) Mothers educational status No education Ref Ref Primary 1.23 (0.86-1.69) 1.22 (0.87-1.55) Secondary and above 2.74 (1.34-5.80)[\*](#TF0001){ref-type="table-fn"} 1.77 (1.04-3.59)[\*](#TF0001){ref-type="table-fn"} **Wealth tertile** Poor Ref Ref Middle 0.82 (0.54-1.24) 0.75 (0.51-1.11) Rich 1.18 (0.71-1.95) 1.08 (0.72-1.63) **Pregnancy intention** Intended Ref Ref Unintended 0.81 (0.61-1.07) 0.93 (0.67-1.28) **Number of under-five children** 1 Ref Ref 2 0.98 (0.64-1.51) 0.97 (0.68-1.39) 3 + 0.70 (0.33-1.46) 0.45 (0.21-0.93)[\*](#TF0001){ref-type="table-fn"} **Father\'s education** Illiterate Ref Ref Literate 1.20 (0.79-1.82) 1.38 (0.98-1.92) **Participation in decisions** Low Ref Ref High 1.63 (1.15-2.31)[\*\*](#TF0002){ref-type="table-fn"} 1.35 (1.01-1.80)[\*](#TF0001){ref-type="table-fn"} **Distance from health facility** ≤ 60 minutes Ref Ref \> 60 minutes 0.55 (0.38-0.80)[\*\*](#TF0002){ref-type="table-fn"} 0.58 (0.41-0.81)[\*\*](#TF0002){ref-type="table-fn"} Antenatal care visits None Ref Ref 1-3 visits 2.79 (1.72-4.55)[\*\*](#TF0002){ref-type="table-fn"} 1.50 (1.06-2.13)[\*](#TF0001){ref-type="table-fn"} 4 or more visits 5.73 (2.77-11.84)[\*\*](#TF0002){ref-type="table-fn"} 2.27 (1.53-3.36)[\*\*](#TF0002){ref-type="table-fn"} **Place of delivery** Home Ref Ref Health institution 1.37 (0.63-2.98) 1.28 (0.80-2.03) Significant at P \< 0.05 Significant at P \< 0.01 Adjusted for mother\'s age, education, wealth index, father\'s education, place of delivery, antenatal care use and distance from health facility Male children were more likely than female children to be fully vaccinated. Among demographic factors, the number of under-five children in a family was significantly associated with completion of vaccination. Children are less likely to be fully vaccinated if there were three or more under-five children in the household (OR; 0.45, 95% CI, 0.21-0.96) compared with children living in households with only one under-five child. Women\'s participation in household decision making, maternal education, use of antenatal care services, and distance from the nearest health facility were associated with both completion of childhood vaccination and receiving at least one vaccination. Children were 1.35 times more likely to be fully vaccinated if their mothers participated in all household decisions than if they did not participate in all household decisions. Children with mothers who had completed secondary education were 1.77 times more likely to be fully immunized compared with children whose mothers had no formal education. Children were more likely to be fully vaccinated if their mother received any antenatal care during pregnancy. Children were 2.27 times more likely to be fully vaccinated if their mother reported four or more antenatal care visits than those whose mothers reported no antenatal visits while children whose mothers reported one to three visits were 1.5 times more likely to be fully vaccinated compared with those reporting no antenatal care visits. Proximity to health facility, measured by the time taken to reach to the nearest health facility, was associated with full vaccination. Children from households living within a 60-minute walking distance from any health facility were more likely to complete vaccination schedules than those located farther than a 60-minute walking distance. Discussion {#S0008} ========== This study examined associations between childhood vaccination and demographic factors such as pregnancy intention, women\'s autonomy and the number of under-five children in a family in rural southwestern Ethiopia. We found that 78% of children age 12-24 months have received at least one of the vaccinination series, although only 37% completed all basic vaccinations. The proportion of children age 12-24 months who were vaccinated with DPT3 was 54%. Considering the fact that DPT3 is an indicator of the global Universal Childhood Immunization initiative, this level of DPT vaccination in the study site is quite low compared with the global average of 83% coverage \[[@CIT0008]\]. The 2011 EDHS also showed that only 24% of children 12-23 months received all basic vaccinations and 37% received DPT3 vaccine \[[@CIT0003]\]. These results show that the coverage and completion of basic vaccinations in Ethiopia is low, particularly in rural areas where the majority of the population lives. We found that children from families with more than one under-five child were less likely to be fully vaccinated. This may be because women with many under-five children face a higher burden of care and may not be able to take their younger child(ren) for vaccination services. Other studies from low and middle countries have also found an association between parity and vaccination status \[[@CIT0015], [@CIT0019]\]. Pregnancy intention was not associated with vaccination status in this study showing that children from unintended pregnancies are no different from intended births in receiving full vaccination. Studies that assessed the association between pregnancy intention and child preventive and curative care in high income countries also found no effects \[[@CIT0031], [@CIT0032]\]. However, in a study by Marston and Cleland that used DHS data of five developing countries, unintended pregnancy was associated with incomplete childhood vaccination \[[@CIT0033]\]. Child\'s sex was associated with fully vaccinated status with male children being more likely to be fully vaccinated than female children. Although previous studies on child vaccination did not report significant differences by gender in Ethiopia \[[@CIT0014], [@CIT0020]\], other studies have reported that significant gender differences exist in food consumption and schooling \[[@CIT0039]\], and food insecurity and morbidity in Ethiopia \[[@CIT0040]\]. There is a tradition of son preference in Ethiopia (44) and the current finding may thus reflect this tradition of sex preference in providing proper care for male children including the decision to immunize a child. In Ethiopia, women are considered to be subordinate to men as evidenced by attitudes towards wife beating, women\'s participation in decision making, and women\'s financial autonomy \[[@CIT0003]\]. We found that women\'s participation in household decision making was associated with complete vaccination. Participation in decision making on health care use (a dimension of women\'s autonomy) may enable women to independently or jointly decide to have their child vaccinated. Previous studies from Ethiopia \[[@CIT0021]\] and Nigeria \[[@CIT0023]\] also found that women\'s autonomy is important in the utilization of child vaccination services. Our results show that other variables such as maternal education and use of antenatal care during pregnancy were significantly associated with full vaccination status. Children with mothers who have at least secondary level of education were more likely to complete the recommended vaccination series than children with mothers with no formal education. Education increases awareness on the role of vaccination services, and such awareness is important in influenceing use of vaccination services. This association is consistent with findings of several previous studies on maternal education and completion of vaccination series \[[@CIT0012], [@CIT0013], [@CIT0015], [@CIT0016], [@CIT0020]\]. As observed in previous studies \[[@CIT0013], [@CIT0014], [@CIT0020]\], use of antenatal care during pregnancy was significantly associated with completing childhood vaccination. Importantly, completing the recommended ANC visits (four or more visits) was strongly associated with full vaccination. The use of antenatal care encourages the use of subsequent maternal and child health services including vaccination \[[@CIT0041], [@CIT0042]\]. The use of delivery care was not significant in this study probably because the small number of women who delivered in health facilities meant the study was insufficiently powered for this analysis. Proximity to health facility, however, was associated with completion of the recommended vaccination series. This finding is consistent with previous studies from Ethiopia and other low and middle income countries, indicating that access to health facilities is an important factor for the utilization of child vaccination services \[[@CIT0017], [@CIT0043]\]. Study findings should be interpreted in light of several limitations. First, because this was a cross-sectional study, we cannot make causal inferences. In addition, although we sought to obtain vaccination data from actual records, not all women had vaccination cards for their children. As such, we had to rely on mothers' reports which are subject to recall bias. Conclusion {#S0009} ========== Our study adds to the existing body of literature regarding the factors that influence childhood vaccination in low and middle income countries. Study findings highlight several potential avenues to improve childhood vaccination rates. First, the association between women\'s decision making autonomy and vaccination highlights the need for initiatives that improve women\'s autonomy in order to attain both gender equality and improved child health service utilization. In addition, in contexts characterized by low literacy levels, providing information and education about the benefits of childhood vaccination may be important. Antenatal care provides provide a good opportunity to provide mothers with information about vaccination and other maternal and child health services. Finally, improving access to family planning information and services is also important because healthy timing and spacing of pregnancies may ease the burden of care and hence promote health care use including vaccination. This research was supported by a PhD fellowship funded in part by the STEP UP (Strengthening Evidence for Programming on Unintended Pregnancy) Research Programme Consortium through a sub-award grant to the African Population and Health Research Center (APHRC) for the African Doctoral Dissertation Research Fellowships (ADDRF). STEP UP is funded by UKaid from the Department for International Development. The ADDRF fellowships are offered by APHRC in partnership with the International Development Research Centre (IDRC). Competing interests {#S0010} =================== The authors declare no competing interests. Authors' contributions {#S0011} ====================== YDW, designed the study, monitored the data collection, analyzed the data, and wrote the first draft of the manuscript. MFA and MJH participated in the design of the study, supervised the whole process and reviewed and modified the drafts of the manuscript. All the authors have read and approved the final version of the manuscript.

 Executive Editors ================= *Senior Editors*: K.R.Fox, *Southampton, UK*; <[email protected]> B.Stoddard, *Seattle, WA, USA*; <[email protected]> ------------------------------------ ------------------------------------ ------------------------------------ M.Madan Babu, *Cambridge*, *UK* M.J.Gait, *Cambridge*, *UK* E.Westhof, *Strasbourg*, *France* <[email protected]> <[email protected]> <[email protected]> M.Churchill, *Aurora*, *CO*, *USA* A.D.Sharrocks, *Manchester*, *UK* J.A.Wise, *Cleveland*, *OH*, *USA* <[email protected]> <[email protected]> <[email protected]> W.Dynan, *Augusta*, *GA*, *USA* M.Wegner, *Erlangen*, *Germany* <[email protected]> <[email protected]> ------------------------------------ ------------------------------------ ------------------------------------ *Methods Editors*: G.Sczakiel, *Lübeck, Germany*; <[email protected]> A.R.Kimmel, *Bethesda, MD, USA*; <[email protected]> *Database Issue Editor*: M.Galperin, *Bethesda, MD, USA*; <[email protected]> *Web Server Issue Editor*: G.Benson, *Boston, MA, USA*; <[email protected]> Editorial Board =============== F.Allain, *Zurich, Switzerland* C.Bailly, *Lille, France* A. Bateman, *Hinxton, UK* P.B.Becker, *Munich, Germany* M.Belfort, *Albany, NY, USA* B.Berkhout, *Amsterdam, The Netherlands* E.H.Bresnick, *Madison, WI, USA* J.M.Bujnicki, *Warsaw, Poland* J.F.Caceres, *Edinburgh, UK* J.S.Carroll, *Cambridge, UK* X.Cheng, *Atlanta, GA, USA* J.J.Collins, *Boston, MA, USA* D.Corey, *Dallas, TX, USA* J.Dinman, *College Park, MD, USA* G.Divita, *Montpellier, France* S.Eddy, *Ashburn, VA, USA* M.Egli, *Nashville, TN, USA* D.Endy, *Cambridge, MA, USA* D.R.Engelke, *Ann Arbor, MI, USA* A.Engelman, *Boston, MA, USA* H.U.Göringer, *Darmstadt, Germany* T.Grange, *Paris, France* W.Gruissem, *Zurich, Switzerland* J.G.Hacia, *Los Angeles, CA, USA* P.Hsieh, *Bethesda, MD, USA* P.A.Jeggo, *Brighton, UK* A.Khvorova, *Lafayette, CO, USA* R.Kierzek, *Poznan, Poland* H.Klein, *New York, NY, USA* D.Klostermeier, *Basel, Switzerland* E.T.Kool, *Stanford, CA, USA* E.Koonin, *Bethesda, MD, USA* A.N.Lane, *Louisville, KY, USA* D.M.J.Lilley, *Dundee, UK* S.Linn, *Berkeley, CA, USA* L.J.Maher, *Rochester, MN, USA* D.Mann, *Newcastle-upon-Tyne, UK* A.Maxwell, *Norwich*, *UK* S.Mueller, *Greifswald, Germany* S.Neidle, *London, UK* R.Parker, *Tucson, AZ, USA* M.R.Parthun, *Columbus, OH, USA* R.Pillai, *Grenoble, France* N.Proudfoot, *Oxford, UK* R.J.Roberts, *Ipswich, MA, USA* D.B.Roth, *Philadelphia, PA, USA* T.D.Schneider, *Frederick, MD, SA* B.Schwer, *New York, NY, USA* B.Séraphin, *Gif sur Yvette, France* P.M.Sharp, *Edinburgh, UK* S.K.Silverman, *Urbana, IL, USA* H.Siomi, *Tokyo, Japan* I.Stancheva, *Edinburgh, UK* A.Stasiak, *Lausanne, Switzerland* N.Sugimoto, *Kobe, Japan* J.Vogel, *Würzburg, Germany* W.Wasserman, *Vancouver, BC, Canada* M.C.Williams, *Boston, MA, USA* S.H.Wilson, *Research Triangle Park, NC, USA* T.Wirth, *Ulm, Germany* R.D.Wood, *Smithville, TX, USA* S.A.Woodson, *Baltimore, MD, USA* S.Yokoyama, *Tokyo, Japan* M.Zavolan, *Basel, Switzerland* Founding Editors ================ R.T.Walker, *Birmingham*, *UK* D.Söll, *New Haven*, *CT*, *USA* A.S.Jones, *Birmingham*, *UK* Editorial and Production ======================== *Editorial ManagerProduction Team*Martine Bernardes-SilvaMichael Evans<[email protected]>Andrew MalvernKate Puttick<[email protected]>

Introduction ============ Heart failure (HF) is a complex clinical syndrome resulting from structural or functional cardiac disorders and characterized by symptoms such as shortness of breath, ankle swelling, and fatigue.[@suz235-B1] Heart failure in the developed world is primarily a disorder of the elderly[@suz235-B2] increasing from 1% among those aged 45--55 years to over 10% in the over 80 year olds. The effects of age ================== Comorbidities and mortality in HF both increase with age.[@suz235-B3]^,^[@suz235-B4] The typical pathophysiology of HF changes as we look at older patient populations. Younger HF patients typically have antecedent ischaemic heart disease or idiopathic or genetic -dilated cardiomyopathies. Elderly patients, in contrast, typically have a history of hypertension and multiple other comorbidities.[@suz235-B5] In the younger patient a dilated ventricle with significant remodelling and reduced ejection fraction is more common (HFrEF), whereas in the elderly the more common picture is a small hypertrophied ventricle with preserved ejection fraction (HFpEF). An older synonym for HFpEF was 'diastolic' HF , as opposed to the more 'systolic ' HF seen in HFrEF. However, the term HFpEF is now preferred to that of diastolic HF especially as HFrEF also frequently shows diastolic dysfunction[@suz235-B6] just as subtle abnormalities of systolic function may be also found in patients with HFpEF and many therapies may be tested in both conditions.[@suz235-B7] Heart failure with reduced ejection fraction has multiple proven therapies to reduce mortality and hospitalization rates, whereas HFpEF has none, perhaps helping somewhat to explain why hospitalization rates for HFpEF have stayed high whereas those of HFrEF have been reducing, an effect also explained by older populations at risk.[@suz235-B8] Epidemiology ============ An estimated 23 million individuals worldwide live with HF,[@suz235-B9] with overall prevalence rates in developed countries of 1--2% of the adult population.[@suz235-B10] The prevalence of HF has progressively increased for many years, both as a consequence of effective therapies keeping patients alive longer and the ageing many populations worldwide,[@suz235-B11] with the later explaining more of the increase. The prevalence of HF rises steeply with age.[@suz235-B12] Heart failure, mainly HFpEF, is one of the most common reasons for presentation to primary care for symptomatic dyspnoea, even when heart disease had not been diagnosed in the patient previously.[@suz235-B13]^,^[@suz235-B14] Prognosis ========= Heart failure is associated with a poor long-term prognosis with a 5-year survival rate still less than 50%.[@suz235-B15] Heart failure remains a major cause of disability and is still the most common cause of hospitalization in the developed world. An ESC HF registry[@suz235-B16] reported a 44% 12-month hospitalization rate and a 17% annual mortality rate for hospitalized HF patients; and 32% and 7%, respectively in stable/ambulatory patients. In this study, sudden death and worsening HF were the most common modes of death for the patients, although along with the ageing population there has a change in the distribution of the causes of death toward non-cardiovascular causes in older especially HFpEF patients, consistent with an increased number non-cardiovascular comorbidities.[@suz235-B17] The effects of age on therapy ============================= Most clinical trials in HFrEF have recruited younger patients,[@suz235-B18]^,^[@suz235-B19] and combined with a relative lack of trials for HFpEF this leaves us with a severe evidence imbalance in favour of younger HFrEF patients as opposed to older HFpEF patients. Although subgroup analyses in the major clinical trials have not consistently shown age to be a factor, these trials themselves have largely excluded real-world older patients with multiple comorbidities.[@suz235-B20] Very few studies specifically recruited older HF patients (notable exceptions being the SENIORS trial[@suz235-B25] and the CIBIS-ELD trial[@suz235-B26]) so that we still have a real evidence lack in elderly HF patients. Although guidelines proclaim that pharmacotherapy and other treatments[@suz235-B27] of HFrEF in elderly patients are recommended to be the same as for younger patients, in the real world, the use of guideline-directed medical therapy is notoriously lower in older patients. The effects of comorbidities ============================ Among older HF patients, women and comorbidities predominate, including hypertension,[@suz235-B28] diabetes,[@suz235-B29]^,^[@suz235-B30] lung disease,[@suz235-B31] coronary disease,[@suz235-B32] renal failure,[@suz235-B33] sleep-disordered breathing[@suz235-B36]^,^[@suz235-B37] anaemia,[@suz235-B38] and iron deficiency predominate.[@suz235-B39]^,^[@suz235-B40] This is despite, along with the elderly, in general, women and patients with comorbidities being selectively not recruited into major clinical trials in HF. There are some particular features in the older patient with HF, which potentially make them less tolerant to drug therapy and more prone to medication errors and side effects.[@suz235-B41] These include dementia and cognitive decline, anorexia,[@suz235-B42] muscle wasting,[@suz235-B43] and frailty[@suz235-B46] all of which more common in the elderly[@suz235-B49]^,^[@suz235-B50] and may have an impact on the risk/benefit relationship of treatments given.[@suz235-B51] Both HFpEF and HFrEF patients can have multiple comorbidities that can affect their treatments and their clinical course,[@suz235-B52] although the extent of this between the two patterns of HF is controversial.[@suz235-B53] The older HF patients' needs are not only just for mortality reducing therapies but also for quality of life enhancement, well-designed end-of-life care and the involvement of their carers in decision making thought-out the disease process.[@suz235-B54] Summary ======= The elderly patients with HF are becoming a major cause of health care expenditure in cardiology and yet our clinical trials have to date not told us accurately how they should be managed to bets effect. In this elderly population, therapeutic decisions should be tested for both effectiveness and applicability and we must recruit typical patients, not just 'trial ' patients. The management of elderly patients with HF requires a more holistic approach to recognizing individual needs and necessary support mechanisms. Funding ------- This paper is part of a supplement funded by the Heart Failure Association of the European Society of Cardiology. **Conflict of interest:** Nothing related to this work. Outside of this work, in the last 3 years, Professor Coats declares having received honoraria and/or lecture fees from: Astra Zeneca, Bayer, Menarini, Novartis, Nutricia, Servier, Vifor, Actimed, Cardiac Dimensions, CVRx, Enopace, Faraday, Gore, Impulse Dynamics, Respicardia, Stealth Peptides, V-Wave.

The molecular machinery of life is largely generated through the assembly of proteins into functional complexes. A particularly common form of protein selfassembly is that leading to linear filaments. These structures are widely used in nature, for instance as the basis of the cytoskeleton. Once formed, the vast majority of functional protein assemblies typically fulfil their biological function but do not directly catalyse the formation of further "daughter" complexes. However, certain protein structures possess the intriguing ability to promote their own replication. This phenomenon first came to prominence in the context of prions, where specific supra-molecular protein assemblies were observed to be able to effectively multiply once taken up into a variety of organisms, ranging from humans to yeast \[[@R1]--[@R3]\]. Such propensity to self-replicate has emerged as a more general feature of pathological protein self-assembly, observed in the context of sickle cell anemia \[[@R4], [@R5]\] as well as for amyloid fibrils implicated in medical disorders \[[@R6]--[@R8]\], such as Alzheimer's disease (A*β* peptide) \[[@R9], [@R10]\], type II diabetes (islet amyloid peptide, IAPP) \[[@R11]--[@R13]\], and Parkinson's disease (*α*-synudein) \[[@R14], [@R15]\]. Strikingly, all of these structures are able to catalyse the formation of their own copies under certain conditions. The initial fibrils are produced spontaneously from solution through primary nucleation, followed by proliferation via heterogeneous, fibril-dependent, secondary nucleation \[[@R12]\]. In this type of self-replication the information about the protein conformation is transferred to the replicas, but they are not necessarily exactly identical to the parent aggregates. Spontaneous fibril formation is inherently slow, while fibril self-replication is usually many orders of magnitude faster \[[@R10]\]; yet a detailed microscopic understanding of either processes is currently lacking. Autocatalytic replication intrinsically introduces positive feedback into the self-assembly process that renders it challenging to control once assembly has started. As such, most functional protein complexes and fibrils do not have self-replicating properties. This finding therefore motivates the question about the fundamental ingredients necessary for fibril self-replication to occur, or indeed to be avoided. Here, we develop a minimal computer model that is able to capture both spontaneous fibril formation in solution, and fibril-self replication. We study the necessary conditions required for self-replication to dominate over spontaneous formation, and find that strong bounds on inter-protein interactions exist for efficient self-replication that result in the high sensitivity of self-replication to environmental conditions. Indeed, it has been reported experimentally that the existence of secondary nucleation in *α*-synuclein, insulin, and A*β* peptide strongly depends on pH \[[@R14], [@R16], [@R17]\], while secondary nucleation in A*β* also varies dramatically with salt concentration \[[@R18]\]. The emergence of a narrow regime that supports self-replication sheds light on why it is relatively a rare property of protein self-assembly in vivo, and possibly provides a physical criterion to distinguish functional from pathological assembly. Moreover, these results suggest that even pathological self-assembly, in principle, can be suppressed by moderate changes to the system to move it from the narrow parameter space supporting self-replication. Our results further infer that the secondary nucleus has to be energetically different from the primary one, pointing to two distinctive pathways. Taking the aggregation of the Alzheimer's A*β* peptide into amyloid fibrils as a model for experimental comparison, in combination with kinetic and biosensing experiments, we show that the major characteristics of secondary nucleation can be explained by the adsorption of monomeric peptides onto the surface of fibrils, and the level of surface coverage. We then demonstrate, in simulations and in experiments, that self-replication can be modulated by controlling the fibril surface coverage. Through the powerful combination of coarse-grained simulations and physical measurements, our results offer microscopic insights into the mechanism of the autocatalytic replication of protein fibrils. Computer model {#S1} ============== As the basis for our model we take the aggregation of peptides and proteins into amyloid fibrils, which have a common structure enriched in *β*-sheet content. A minimal model that reproduces homogeneous fibril nucleation allows an amyloidogenic protein to exist in two states: a soluble state (denoted "*s*") that can form finite oligomers, and a higher free-energy state that can form the *β*-sheet enriched fibrils (denoted "*β*") \[[@R19], [@R20]\]. Simply considering the interaction of soluble proteins with the surface of existing fibrils captures the binding of monomers to the fibrils, but does not lower the free energy barrier for nucleation, thus does not result in catalysis. To achieve a self-replication rate that is significantly faster than spontaneous formation, the structure and energy of the involved species necessarily have to differ from those observed in the absence of fibrils ([Supplementary Section SI.C](#SD1){ref-type="supplementary-material"}). The self-replication cycle in the A*β* system has been shown to predominately generate small prefibrillar oligomers, whose structures differ from that of the mature fibrils ([Methods](#S8){ref-type="sec"}, \[[@R10], [@R21]\]). Although an ensemble of such intermediate structures could exist in reality, here we consider the simplest possible case: we include one additional, intermediate ("*i*"), conformation, which can take place on the fibril surface. This conformation is in-between the soluble and the *β*-state, and its selfinteraction is stronger than its interaction with the fibril, which leads to detachment of oligomers from the parent fibril, as observed in experiments. Amyloidogenic protein in our model are represented as hard spherocylinders with attractive patches ([Fig. 1](#F1){ref-type="fig"}). The attractive interactions account for generic features of inter-protein interactions, such as hydrophobic interactions, hydrogen bonding, and screened electrostatic interactions. The soluble state of the protein is modelled as a spherocylinder with an attractive tip ([Fig. 1a](#F1){ref-type="fig"}), whose self-attraction is given by the parameter *ϵ*~*ss*~. Such particles are able to make finite oligomers ([Fig. 1b](#F1){ref-type="fig"}) \[[@R20]\]. The attractive tip can also adsorb onto the outer surface of the fibril, with interaction strength *ϵ*~*sf*~ ([Supplementary Fig. S1](#SD1){ref-type="supplementary-material"}). The intermediate conformation *i* is modelled with the same potential as the soluble state, but possesses a stronger self-association parameter *ϵ*~*ii*~ and a vanishing adsorption onto the fibril ([Supplementary Fig. S1](#SD1){ref-type="supplementary-material"}). The fibril forming, *β*-sheet prone, configuration is a hard spherocylinder with an attractive side-patch ([Fig. 1a](#F1){ref-type="fig"}). The *β*-prone proteins pack parallel to one another with the maximal interaction strength *ϵ*~*ββ*~, leading to fibril-like aggregates ([Fig. 1b](#F1){ref-type="fig"}). We performed dynamic Monte Carlo (MC) simulations, allowing for the interconversion between the three protein conformations with a small probability at every MC step. The *s* → *i* → *β* conversion is thermodynamically unfavourable, reflecting the loss of the conformational entropy \[[@R22]\]. Further details are given in the [Methods](#S8){ref-type="sec"} Section. Spontaneous formation versus self-replication {#S2} ============================================= The first question we address involves the identification of those conditions that lead to secondary nucleation being dramatically dominant over spontaneous, primary, nucleation. We have performed a series of computer experiments, in which a capped preformed fibril (incapable of further growth) was inserted into a solution of monomeric proteins, and nucleation processes were monitored. Primary nucleation takes place in two steps, whereby protein oligomers first form in solution, and then convert into *β*-sheet nuclei, which continue growing by monomer addition ([Fig. 1c](#F1){ref-type="fig"}) \[[@R20], [@R23]\]. In the secondary nucleation process, proteins first adsorb onto the surface of the fibril, forming local clusters that keep growing and shrinking while still being attached to the fibril surface, as depicted in [Fig. 1d](#F1){ref-type="fig"}. Once the oligomer of a critical size is formed, the proteins within change their conformation into the intermediate form. The oligomer then detaches into the solution, converts into the *β*-sheet protofibril, and grows further by monomer addition ([Fig. 1d](#F1){ref-type="fig"}). To investigate possible scenarios for different aggregating proteins, under various solution conditions, we measured the rates of primary and secondary nucleation at different protein concentrations and inter-protein interactions. From these measurements we calculated the fraction of self-replication events in the system for a given set of external conditions ([Supplementary Sections SI.A and SI.B](#SD1){ref-type="supplementary-material"}), [Fig. 2a](#F2){ref-type="fig"}. Clearly, self-replication dominates over spontaneous fibril formation at low protein concentrations and low inter-protein interactions. Indeed, proteins are typically below their critical micelle concentration at physiological conditions, which corresponds to the regime of low inter-protein interactions and low protein concentrations, where self-replication can dominate. The reason for the dramatic dominance of self-replication in this regime is two-fold. The first contribution arises from the aided collocation of proteins on the one-dimensional surface of the fibril. This contribution is particularly important at low protein concentrations, where the probability of proteins meeting in solution and forming oligomers is very low. The second contribution lies in the decreased barrier for the secondary nuclei formation on the fibril surface, via the intermediate state ([Supplementary Section SI.C](#SD1){ref-type="supplementary-material"}). Essentially, for self-replication to dominate, the secondary nucleus has to be different from the primary one. Strong environmental bounds for self-replication {#S3} ================================================ Modulating environmental conditions and introducing protein mutations not only changes the properties of proteins interacting in solution, but also the strength of the adsorption of proteins onto the surface of fibrils, given by *ϵ*~*sf*~ in our simulations. We find that changing the protein-fibril affinity only by a few k*T*, the fraction of self-replication events changes non-monotonically, exhibiting a distinct region of optimal self-replication, [Fig. 2b](#F2){ref-type="fig"}. This result is in agreement with the high sensitivity of fibril self-replication to solution composition, and can explain why it is to date observed only in few systems. Comparably, in a recent simulation, secondary nucleation of Lennard-Jones particles at a crystalline surface, when exposed to mechanical agitation, was reported to take place only in the regime of intermediate supersaturation \[[@R24]\]. [Fig. 3a](#F3){ref-type="fig"}. analyses this effect in depth, at constant protein concentration. At low protein-fibril interaction strengths, proteins cover only a small fraction of the fibril surface, and the protein adsorption and oligomer formation on the fibril surface determine the reaction rate. [Fig. 3b](#F3){ref-type="fig"} depicts the Langmuir-type isotherm for the fibril surface coverage, *θ*, as a function of *ϵ*~*sf*~ ([Supplementary Section SI.D](#SD1){ref-type="supplementary-material"}), indicating that the increase in the surface coverage follows the increase in the rate of self-replication in [Fig. 3a](#F3){ref-type="fig"}. At high *ϵ*~*sf*~, the fibril is substantially covered by proteins, however, the oligomer detachment becomes unfavourable. Nucleation will happen only after the oligomer has reached a certain size, *N*\*, when the energy gain due to the stronger inter-protein interactions after the conformational change overcomes the loss in the protein-fibril adsorption energy. Stronger binding to the surface hence requires larger oligomers in order to overcome the loss in the favourable adsorption energy. For very large oligomers, due to the geometric constraints, this requirement cannot be satisfied. Therefore, the conformational change will become unfavourable as the binding to the surface increases further (inset in [Fig. 3b](#F3){ref-type="fig"}, [Supplementary Section SI.E](#SD1){ref-type="supplementary-material"}). In reality, in the regime of high adsorption, proteins are likely to distribute themselves evenly on fibrils in order to increase their contact area with the surface, and could form multiple layers, additionally hampering secondary nucleation. The narrow region of inter-protein interactions supporting self-replication is therefore the outcome of the balance between sufficient fibril coverage, and unhindered conformational change. Kinetics of self-replication and comparison with experimental measurements {#S4} ========================================================================== Our model makes a range of predictions that can be directly experimentally tested. Here, we seek to relate our simulations to kinetic measurements of self-replication of A*β*40 amyloid fibrils, one of the two major isoforms of the A*β* peptide associated with Alzheimer's disease. Kinetic experiments usually determine the dependence of the reaction rate on monomer concentration, *r* ∼ *c*^*γ*^, where the scaling exponent *γ* is the reaction order. It reflects the monomer dependence of the dominant aggregation processes, and is typically believed to be determined by the number of molecules reacting in the rate-limiting step, therefore carrying information about the reaction mechanism. [Fig. 4a](#F4){ref-type="fig"} depicts double logarithmic plot of the rate of secondary nucleation for the A*β*40 system, versus the initial monomer concentration, where the slope corresponds to the scaling exponent. Curiously, the scaling exponent is highly dependent on the concentration of the monomeric peptide in solution, suggesting a possible change in the nucleation mechanism over the concentration range \[[@R25]\]. [Fig. 4b](#F4){ref-type="fig"} shows the same quantities, collected in simulations, at a moderate peptide-fibril affinity. The reaction order varies with the protein concentration, with a high value at low monomer concentrations (*γ* ≈ 3.3), and low value at high monomer concentrations (*γ* ≈ 0.5), as with the A*β*40 experimental data. Due to our microscopic modelling we are able to pinpoint the processes underlying the switch in kinetic behaviour. [Fig. 4d](#F4){ref-type="fig"} shows that the change in the reaction order follows the trend in the change of fibril coverage. Hence, the non-linear increase in surface coverage, due to surface saturation, appears to be the cause of the continuous decrease in reaction order. It is beneficial to establish what controls the absolute value of the apparent reaction order (see [Methods](#S8){ref-type="sec"} and [Supplementary Section SI.F](#SD1){ref-type="supplementary-material"} for details). We find that the rate of self-replication follows the surface saturation as ln(*r*) ∼ *N*\*ln(*Kc*/(1 + *Kc*)), where *K* is the monomer-surface binding constant (*K* ∼ *ϵ*~*sf*~) and *N*\* is the size of the nucleating oligomer (found to be constant over the concentration range in our simulations, inset in [Fig. 4b](#F4){ref-type="fig"}). The reaction order then continuously changes between *γ* → *N*\*, at infinite dilution, and *γ* → 0 at full saturation. Since nucleation is possible within a finite time only when the surface coverage is non-negligible, observable values of *γ* will be necessarily smaller than *N*\*. Experimental verification of surface saturation {#S5} =============================================== To test experimentally the prediction that the change in the apparent reaction order is governed by the change in the surface coverage, and not by a change in the nucleation mechanism, we designed a series of surface plasmon resonance (SPR) biosensing experiments that allow direct measurement of the binding of monomeric peptide molecules to the surface of amyloid fibrils, under the same conditions as the kinetic experiments. This enabled us to obtain the Langmuir absorption isotherm of A*β*40 peptides onto their own fibrils ([Fig. 4c](#F4){ref-type="fig"} and [Supplementary Fig. S6](#SD1){ref-type="supplementary-material"}). Indeed, the surface saturation takes place in the micromolar regime (with an equilibrium binding constant of *K*^−1^ = 15*μ*M), which is exactly the regime where the change in the apparent reaction order takes place in aggregation experiments ([Fig. 4a](#F4){ref-type="fig"}). Furthermore, this value of *K* is of the same order of magnitude as the value obtained from the kinetic fit to the experimental aggregation data ([Methods](#S8){ref-type="sec"} and [Supplementary Section SII](#SD1){ref-type="supplementary-material"}), and therefore strongly supports the hypothesis that the change in exponent is due to surface saturation. Surface saturation controls the apparent reaction order {#S6} ======================================================= Finally, we show that by controlling the surface coverage via varying the strength of the inter-protein interactions, at constant monomer concentration, one can further modulate the kinetics of fibril self-replication. At constant protein concentration, the surface coverage is determined by the magnitude of protein-fibril affinity and inter-protein interactions. It is likely that both of these interaction strengths will be affected when altering experimental conditions, due to their similar physical origins. We observe that the surface coverage increases when both of these interactions are strengthened in simulations, resulting in a weaker dependence of self-replication on monomer concentration. The average scaling exponent *γ* from the simulations, as a function of *ϵ*~*ss*~ and *ϵ*~*sf*~, is shown in [Fig. 5a and Fig. 5b](#F5){ref-type="fig"}. We compare this behaviour to the aggregation of the A*β*42 at a range of NaCl salt concentrations \[[@R18]\], [Fig. 5c](#F5){ref-type="fig"}. In the context of our physical model, two isoforms of A*β* peptide, A*β*40 and A*β*42, share mechanistic similarities. An increase in ionic strength shields the electrostatic interactions and leads to an increased attraction between the negatively charged A*β*42 monomers and fibrils, as well as the monomers to each other. Hence a variation of ionic strength offers an experimental way to vary in a controlled way the value of *ϵ*~*ss*~ and *ϵ*~*sf*~. Indeed, the trend in the behaviour of the scaling exponents for the aggregation of A*β*42 with increasing salt concentration agrees well with that found in our simulations. Therefore the large effect of ionic strength on the aggregation behaviour is in agreement with a variation of the adsorption of peptides onto their fibrils, offering a direct way to influence the self-replication process in a controlled manner. Discussion and conclusions {#S7} ========================== By developing a minimal model of protein self-replication, we have identified its dominant physical determinant to be the adsorption of monomeric proteins onto the surface of protein fibrils. Strong limits on interprotein interactions are found for efficient self-replication, originating from the fact that changes in the interaction strength have opposing effects on the two parts of the nucleation mechanism: oligomer formation and oligomer detachment. A narrow region of "ideal" interaction values supporting self-replication ([Fig. 2b](#F2){ref-type="fig"}) results in its high specificity and sensitivity to environmental conditions. An additional conformational change taking place on the fibril surface is a minimal requirement for the catalysis and detachment of oligomers from the parent fibril, which, in the context of many amyloid diseases, is a crucial step in the proliferation of pathological species \[[@R26]--[@R28]\]. The conformational change is at the origin of the formation of amyloid fibrils; the aggregating protein necessarily undergoes a change from the soluble form into the characteristic *β*-hairpin conformation. Models which attempt to achieve self-replication in (nearly) minimal colloidal systems, require an external dynamical change to permit detachment of the replicas from the parents \[[@R29], [@R30]\]. Amyloidogenic proteins naturally possess this dynamic characteristic. A direct practical contribution from our analysis is the ability to relate the reaction order measured in experiments to the underlying microscopic mechanism. We have found that the changes in the reaction order can be related to the change in the fibril surface coverage byproteins, which we have confirmed by directly measuring the binding isotherm of monomers to the fibril surface. The characteristic concentration-dependence of the reaction order, observed in experiments, is consistent with a scheme where the rate-limiting step takes place on the surface, further confirming that primary and secondary nucleation are indeed different processes. Whether the change in the apparent reaction order will be experimentally measured will depend on the concentration range that can be explored, as the experiments might be limited to a concentration range where it appears locally constant. By measuring the fibril coverage and the apparent kinetic reaction order separately, the information about the critical size of oligomers produced via secondary nucleation becomes directly accessible, for any protein system which exhibits this behaviour. As a proof of principle, we have shown that by varying in a controlled manner the fibril surface coverage, by modulating the inter-protein interactions with ionic strength, one can control the kinetics of fibril self-replication. Hence the adsorption of monomeric proteins onto the surface of protein fibrils may pose a central target in limiting the proliferation of protein aggregates in a disease context. Methods {#S8} ======= The coarse-grained model and the choice of parameters {#S9} ----------------------------------------------------- We use the model developed in Ref. \[[@R20]\], extended to capture secondary nucleation. In spirit, this model is similar to the multistate Potts model of Zhang and Muthukumar \[[@R31]\], and the recent model of Ilie, Otter and Brils \[[@R32]\]. Recently, more rigorous schemes have been developed to map coarse-grained inter-peptide interactions onto patchy-colloids for the purpose of studying protein aggregation by Ruff et al. \[[@R33], [@R34]\]. In our model each spherocylinder is *σ* = 2*nm* wide and *L* = 4*σ* = 8*nm* long. The hard core repulsion forbids for any distance between any two spherocylinders to be smaller than *σ*. The interaction between two peptides in the soluble "*s*" form is implemented as: $$V_{ss}\left( r \right) = \begin{cases} {- \epsilon_{ss}\left( \frac{\sigma}{r} \right)^{6}} & {\text{if}\ r \leq 1.5\sigma} \\ 0 & {\text{if}\ r > 1.5\sigma} \\ \end{cases}$$ where *r* is the distance between the centers of the attractive tips located at the spherocylinders' ends. An attractive patch is added only at one spherocylinder pole to ensure formation of finite aggregates like those observed in experiments. This potential drives the formation of micellar-like oligomers, where tips of participating peptides are in contact in the oligomer center ([Fig. 1B](#F1){ref-type="fig"}). The parameter *ϵ*~*ss*~ controls the strength of the non-specific interactions between the soluble peptides. Using atomistic simulations we estimated *ϵ*~*ss*~ to be relatively small, on the order of 5k*T* \[[@R20]\]. To explore the influence of different solution conditions, we varied it between 3k*T* and 8k*T*, as indicated in the text. The interaction between two peptides in the intermediate conformation "*i*", and between the soluble and the intermediate conformation is implemented using the same potential as in [Eq. (1)](#FD1){ref-type="disp-formula"}, with *ϵ*~*ss*~ → *ϵ*~*ii*~ and *ϵ*~*ss*~ → *ϵ*~*si*~, respectively. The intermediate state is designed to be between the soluble and the *β*-sheet forming state, corresponding to a conformation with more *β*-content than the soluble state, but not yet a fully folded *β*-hairpin. Hence, the relative strength of interactions was always preserved, with *ϵ*~*ss*~ \< *ϵ*~*si*~ \< *ϵ*~*ii*~. Their values were chosen such that nucleation is achieved within a reasonable computer time (see [Supplementary Fig. S2](#SD1){ref-type="supplementary-material"}), while preserving their relative strength; *ϵ*~*ii*~ is kept constant at *ϵ*~*ii*~ = 16k*T*, and *ϵ*~*si*~ is kept constant at *ϵ*~*si*~ = 8k*T*. Throughout the article k denotes the Boltzmann's constant and *T* is the temperature. The attractive side-patch of the *β*-sheet forming configuration is *L*~*p*~ = 0.6*L* long and spans an angle of 180°. If two patches face each other their interaction is: $$V_{\beta\beta}\left( r \right) = \begin{cases} {- \epsilon_{\beta\beta}\textit{cos}^{2}\left( \phi \right) - \epsilon_{\beta\beta}\left( \frac{\sigma}{r} \right)} & {\text{if}\ d \leq 1.5\sigma} \\ 0 & {\text{if}\ d > 1.5\sigma} \\ \end{cases}$$ where *ϕ* is the angle between the axes of the particles, *d* is the shortest distance between the axes of the patches, and *r* is distance between the centers of the patches. The first term controls that peptides in the *β*-forms pack parallel to each other, mimicking the hydrogen-bond interactions between *β*-sheets, while the second term ensures compactness of the fibrils \[[@R22], [@R35], [@R36]\]. To drive the formation fo thermodynamically stable fibrils, *ϵ*~*ββ*~ has to be the strongest of all the interactions in the system. In this study we choose *ϵ*~*ββ*~ = 60k*T* \[[@R37], [@R38]\]. General aggregation of patchy-spherocylinders has been studied in details in our previous work \[[@R39]\]. The cross-interaction between the soluble and the *β-*sheet-forming configuration is designed as: $$V_{s\beta}\left( d \right) = \begin{cases} {- \epsilon_{s\beta}} & {\text{if}\ d < 1.5\sigma} \\ 0 & {\text{if}\ d > 1.5\sigma} \\ \end{cases}$$ where *d* is the shortest distance between the centre of the attractive tip and the axis of the *β*-patch, and ϵ~*sβ*~ = *ϵ*~*ss*~ + 1k*T*. The *i*-*β* interaction is described in the same way, with *ϵ*~*sβ*~ → *ϵ*~*iβ*~, and *ϵ*~*iβ*~ = *ϵ*~*ii*~ + 1k*T*. Peptide adsorption onto the preformed fibril is given by: $$V_{sf}\left( d \right) = \begin{cases} {- \epsilon_{sf}\left( \frac{\sigma}{d} \right)^{6}} & {\text{if}\ r \leq 1.5\sigma} \\ 0 & {\text{if}\ r > 1.5\sigma} \\ \end{cases}$$ where *d* is the shortest distance between the centre of the attractive tip of the soluble peptide and the body of the *β*-peptide (there is no other angular dependence). Adsorption of the intermediate "*i*" conformation onto the fibril is described in the same way ([Eq. (4)](#FD4){ref-type="disp-formula"}), with *ϵ*~*sf*~ → *ϵ*~*if*~, and *ϵ*~*if*~ = 1k*T*. The *β*-peptide interacts with the preformed fibril only via volume exclusion. The model parameters are summarized in [Supplementary Figure S1](#SD1){ref-type="supplementary-material"}. MC Scheme {#S10} --------- MC simulations were performed with small translational and rotational moves, to approach the realistic dynamics of the system. The interconversion between the three states was carried out with a small probability *P* = 0.0002, which mimics the slow conversion of the soluble peptide into fibril-forming *β*-sheet prone configuration. Every conversion from the soluble to the *β*-state is penalized with a change in the excess chemical potential of magnitude Δ*μ* = 20k*T*, and the *s* → *i* and the *i* → *β* with 0.5Δ*μ* ([Fig. 1a](#F1){ref-type="fig"}). These values are chosen to reflect the fact that amyloidogenic proteins with small-to mid-*β*-propensity, such as A*β*, are typically not found in the *β*-sheet prone conformation in solution \[[@R40], [@R41]\]. Simulations were performed in a periodic cubic box in a grand-canonical ensemble, where the chemical potential of non-adsorbed soluble peptides was kept constant. This scheme was chosen to avoid the depletion of monomers from the solution due to the adsorption onto the surface of the preformed fibril. For this purpose, we do not distinguish between the monomeric soluble species, and the soluble species that are part of an oligomer in solution. The number of soluble peptides in the beginning of each simulation was set to ∼ 600, and the box size was adjusted to match the targeted peptide concentration. Soluble peptides are added or removed from anywhere in the simulation box, according to the grand-canonical scheme \[[@R42]\], excluding the *r* = 5*σ* region around the capped preformed fibril. All simulations were performed with the same size of the preformed fibril, which consists of *N* = 92 *β*-peptides and is unable to grow further. We were monitoring only the first generation of replicas, and have allowed the soluble peptides to adsorb only onto the preformed fibril, and not onto its replicas. Kinetics of self-replication {#S11} ---------------------------- In bulk experimental systems, the overall kinetics are determined by the processes of spontaneous nucleation in solution, elongation and self-replication, that all alter the fibril population. To compare bulk kinetic measurements to the modelling of nucleation on a single, growth-incompetent fibril used in simulations, it is necessary to dissect the macroscopic behaviour into its constituent processes. This can be achieved by developing a theoretical kinetic model and global fitting to the experimental kinetic data. We have adapted a theoretical kinetic model for the aggregation of A*β*40 \[[@R25]\] to include the Langmuir-like adsorption of peptides onto the growing fibril, and fit it to bulk experimental kinetic data to obtain the rate of secondary nucleation at various peptide concentrations. The details of the kinetic model as well as the global fits used to obtain this rate of secondary nucleation are shown in the [Supplementary Section SII and Fig. S4](#SD1){ref-type="supplementary-material"}. Experimental exploration of intermediate oligomers in self-replication of A*β*42 {#S12} -------------------------------------------------------------------------------- If the oligomers generated through secondary nucleation were of the same structure as the fibrils, their concentration, \[*O*\], could be estimated from the known rate parameters for the fibrillar growth as $\left\lbrack O \right\rbrack = \frac{k_{2}m_{\text{tot}}^{n_{2}}}{2^{n_{2}} +^{1}k_{+}},$ where *k*~2~ is the rate constant for secondary nucleation, *k*~+~ is the fibril elongation rate constant and *m*~tot~ is the total protein concentration \[[@R43]\]. Using the values for the rate constants extracted from kinetic measurements of A*β*42 aggregation (*k*~2~ ≈ 10^4^ M^−2^s^−1^, *k*~+~ ≈ 3 × 10^6^ M^−1^s^−1^ and *m*~tot~ = 5*μ*M) \[[@R10]\], we find this concentration to be \[*O*\] ≈ 0.01 pM. This value is at least 5 orders of magnitude smaller than the experimentally measured concentration of oligomers in the same system (nanomolar range \[[@R10]\]), indicating that the structure of oligomers generated via such secondary pathway is necessarily different from that of the fibrils. Scaling of the rate of self-replication with surface coverage {#S13} ------------------------------------------------------------- We recall that the conformational change, and subsequent fibril nucleation, is favourable only for oligomers above a certain critical size *N*\*. The free energy of formation of such an oligomer on a finite surface scales as Δ*F*(*N*\*) ∼ −*N*\*ln(*Kc*/(1 + *Kc*)) where *K* is the monomer-surface binding constant (*K* ∼ *ϵ*~*sf*~) and *c* is the free monomer concentration ([Supplementary Section SI.G](#SD1){ref-type="supplementary-material"}). Since the rate of the process depends exponentially on the negative magnitude of the free energy change for the critical oligomer formation, we obtain: $$\text{ln}\left( r \right) \sim - \text{Δ}F\left( N^{*} \right) \sim N^{*}\text{ln}\left( {{Kc}/\left( {1 + Kc} \right)} \right).$$ [Supplementary Fig. S3](#SD1){ref-type="supplementary-material"} shows the free energy for oligomer formation on the fibril surface, Δ*F*(*N*), measured from the size-distribution of oligomers on the fibril in our simulations ([Supplementary Section SI.F](#SD1){ref-type="supplementary-material"}). As predicted, it decreases with increasing peptide concentration, reaching a plateau at high concentrations. An arrow in the [Supplementary Fig. S3](#SD1){ref-type="supplementary-material"} marks the lowest concentration range at which we observe nucleation (−9 \< ln(*c*) \< −8). The slope at that point (≈ 0.6), multiplied by the average critical oligomer size (*N*\* ≈ 6, inset in [Fig. 4b](#F4){ref-type="fig"}), should give us the expected apparent reaction order in the kinetic plot *γ* ≈ 3.6. The measured reaction order at the same concentration range in [Fig. 4b](#F4){ref-type="fig"} is *γ* ≈ 3.3, which agrees well with the predicted value within the error of our scaling theory and measurements. SPR Experiments {#S14} --------------- A*β*40 amyloid fibrils were attached to the surface of an SPR biosensor and exposed to a solution containing monomeric A*β*40. In this case, monomers simulta neously attach both to the fibril ends and to their surfaces. However, due to their very different kinetics and thermodynamics, the two processes can readily be distinguished ([Supplementary Section SIII](#SD1){ref-type="supplementary-material"}). The elongation of fibrils will lead to a linear increase in mass, while the rate of attachment of peptide to the surface of fibrils is expected to decrease exponentially with time as the available binding sites become occupied. Conversely, upon washing the fibrils, the surface-bound peptide molecules are expected to show an exponential detachment behaviour, at high rates due to their relatively low binding free energy, while the rate of loss from the fibril ends by monomer dissociation is expected to be linear and very slow due to the high thermodynamic stability of the *β*-sheet rich fibrils \[[@R44]\]. By following the kinetic data of monomer detachment, we can distinguish the fast exponential from the slow linear dissociation ([Supplementary Fig. S5](#SD1){ref-type="supplementary-material"}), and obtain the amplitude of the exponential signal resulting from attachment to the surface of the fibrils, at various concentrations of the free monomers. Supplementary Material {#SM} ====================== We acknowledge support from the Human Frontier Science Program and Emmanuel College (A.Š), Leverhulme Trust and Magdalene College (A.K.B), St. John's College (T.C.T.M), Biological Sciences Research Council (T.P.J.K. and C. M. D.), the Frances and Augustus Newman Foundation and the Biotechnology (T.P.J.K.), the European Research Council (T.P.J.K., S.L. and D.F), and the Engineering and Physical Sciences Research Council (D.F.). **Contributions** A.Š., T.P.J.K. and D.F. conceived the project; A.Š. designed and performed the computer simulations; A.K.B. performed the SPR measurements; G.M. performed the kinetic analysis; T.C.T.M. and S.L. contributed materials and/or analysis tools, and all authors participated in writing the paper. The data that support the plots within this paper and other findings of this study are available from the corresponding authors upon request. {#F1} {#F2} {#F3} ![Kinetics of fibril self-replication.\ (a) **Experimental results**: The rate of secondary nucleation for the A*β*40 system versus the initial concentration of soluble monomers, from Ref. \[[@R25]\]. (b) **Simulation results:** The rate of secondary nucleation of fibrils with a moderate affinity for soluble monomers (*ϵ*~*sf*~ = 6k*T*) as a function of the concentration of the monomeric proteins in solution. Inset: the average critical oligomer size stays constant over the entire concentration range; the solid line plots the linear fit over the concentration range, (c) **Experimental results**: Fraction of the peptides bound to the surface of A*β*40 fibrils, *θ*, under the same conditions as the kinetic experiments in (a), versus the concentration of the monomers. The dashed line is the fit to the Langmuir isotherm with *K*^−1^ = 15*μ*M. Inset: schematic representation of the adsorption of monomeric peptides (coloured in blue) to the surface of fibrils (coloured in magenta), measured via SPR. (d) **Simulation results**: Surface coverage *θ* versus the concentration of free monomers at *ϵ*~*sf*~ = 6k*T*. Inter-peptide interaction is kept constant at *ϵ*~*ss*~ = 4k*T* for all simulation data.](emss-68297-f004){#F4} ![The apparent reaction order is controlled by the surface saturation.\ **Simulation results:** (a) Scaling exponent for the kinetics of fibril self-replication, averaged over the range of concentrations (20*μ*M ≤ *c* ≤ 1mM), as a function of the interpeptide interaction between soluble monomers at constant peptide-fibril affinity *ϵ*~*sf*~ = 8k*T*, and (b) as a function of the peptide-fibril affinity at constant inter-peptide affinity *ϵ*~*ss*~ = 4k*T*. An increase in *ϵ*~*ss*~ and *ϵ*~*sf*~ increases the surface coverage, as shown by the representative snapshots in insets, taken at a monomer concentration *c* = 0.15*m*M. **Experimental results**: (c) The average scaling exponent for self-replication of *Aβ*42 fibrils at a range of NaCl concentrations, whose increase is expected to increase both *ϵ*~*ss*~ and *ϵ*~*sf*~ from Ref. \[[@R18]\].](emss-68297-f005){#F5}

{#sp1 .565}

INTRODUCTION ============ Postmenopausal osteoporosis (PMOP) is a prevalent metabolic bone disease characterized by bone loss and structural destruction, which increases the risk of fracture in postmenopausal women.[@B1] About 22 million women in the European Union and 8 million women in the United States have been diagnosed with osteoporosis.[@B2][@B3] Owing to the high morbidity and serious complications of PMOP, many efforts have been devoted to its prophylaxis and treatment.[@B4][@B5] However, women with osteoporosis still face high mortality due to the complications of fracture.[@B6] Thus, deeper understanding of the pathogenesis of PMOP is vital for the clinical treatment thereof. Osteoblast cells are specialized, terminally differentiated products of mesenchymal stem cells.[@B7] By targeting inflammatory cytokines, osteoblast cells contribute to the process of bone formation.[@B8] A previous study showed that osteoblast cells play a very important role in the progression of PMOP via influencing bone anabolic function.[@B9] Using primary cultured human osteoblast cells, Trošt, et al.[@B10] indicated that the differential expression of certain genes is associated with the progression of osteoporosis. As one of the regulatory genes in organisms, the long non-coding RNAs (lncRNA)-anti-differentiation noncoding RNA (ANCR) is essential for osteoblast differentiation.[@B11] Jia, et al.[@B12] showed that the down-regulation of lncRNA-ANCR promotes osteogenic differentiation of periodontal ligament stem cells. Actually, the function of lncRNA-ANCR on osteoblast is complicated. Importantly, lncRNA-ANCR is associated with enhancer of zeste homolog 2 (EZH2), and the association results in the inhibition of both runt related transcription factor 2 (RUNX2) expression and subsequent osteoblast differentiation.[@B13] This regulatory mechanism of lncRNA-ANCR has been observed in many diseases, such as breast cancer and colorectal tumor.[@B10][@B14] However, whether lncRNA-ANCR can affect the occurrence of PMOP by regulating the osteogenesis of osteoblast cells is still unclear. In the current study, osteoblast cells were isolated from a PMOP mice model and were transfected with siRNA-ANCR (si-ANCR). Subsequently, the proliferation, apoptosis, alkaline phosphatase (ALP) activity, and calcium nodules of osteoblast cells were analyzed. The expression of osteogenic differentiation related proteins was also investigated. Finally, the effect of si-ANCR on osteogenesis was evaluated in mice. In doing so, we aimed to explore the effects and mechanisms of lncRNA-ANCR on osteogenesis of osteoblast cells in PMOP to identify a new strategy for the treatment of PMOP. MATERIALS AND METHODS ===================== PMOP model construction ----------------------- A total of 45 male KM mice (20--25 g, 7--8 weeks) were fed in a SPF grade environment at 25--27℃ and 45--50% relative humidity. All mice cages were sterilized with 1:10000 perchloric acid. Meanwhile, bedding, water, and standard food were sterilized under high temperature and pressure. After 1 week of acclimatization, mice were randomly divided into three groups, including an osteoporosis group (n=15, undergo ovariectomy), a sham group (n=15, open/close abdomen operation without ovaries removed), and a control group (n=15, without any treatment). This study was approved by the local ethics committee, and all experiments were in accordance with the guide for the care and use of laboratory animals. Primary cell culture -------------------- Soft tissue on the bone surfaces of PMOP mice was rapidly cleaned under aseptic conditions, and the intima was separated and cut into fragments. Then, these tissues were digested by trypsinase and incubated in a CO~2~ incubator for 30 mins. After neutralizing trypsin, appropriate Roswell Park Memorial Institute (RPMI)-1640 medium was added. The suction tube was repeatedly blown into cell suspension and inoculated in a culture flask. When osteoblast cells spread over 90% of the bottom of the bottle, they were digested and passaged with 2.5 g/L of trypsinase. Finally, the medium was removed, and the cells were washed with phosphate belanced solution (PBS) twice and detached from the plates with 0.25% trypsin. RPMI-1640 culture medium containing a 10% volume fraction of bovine serum was used to blow osteoblast cells into a single-cell suspension for routine passage. Intracellular calcium ions measurement -------------------------------------- Osteoblast cells were cultured in serum-free medium for 24 h to synchronize the cell cycle, followed by culturing in 10% fetal bovine serum (FBS) medium for 24 h. Osteoblast cells were washed twice with PBS buffer solution (supernatant removed) after adherence. Fluo8 was added to the plate with a solubility of 5 µm/mL, followed by removing the supernatant. Then, Tg solution was added to the plate at a final concentration of 4 µm/mL. After 40 min at room temperature away from light, the plate was washed twice with calcium-free PBS. Laser confocal microscopy was used to scan cells for 30 seconds. Finally, changes in calcium ions in osteoblast cells were scanned continuously after CaCl~2~ solution (2 mM) injection. Primary cell culture -------------------- Soft tissue on the bone surfaces of PMOP rats was rapidly cleaned under aseptic conditions, and the intima was separated and cut into fragments. Then, these tissues were digested by trypsinase and incubated in a CO~2~ incubator for 30 mins. After neutralizing trypsin, appropriate RPMI-1640 medium was added. A suction tube was repeatedly blown into the cell suspension and inoculated in a culture flask. When osteoblasts spread over 90% of the bottom of the bottle, they were digested and passaged with 2.5 g/L of trypsinase. Finally, the medium was removed, and the cells were washed with PBS twice and detached from the plates with 0.25% trypsin. RPMI-1640 culture medium containing a 10% volume fraction of bovine serum was used to blow cells into a single-cell suspension for routine passage. LncRNA-ANCR siRNA transfection ------------------------------ For transfection, osteoblast cells were divided into three groups, including si-ANCR (experimental group, cell line was transfected with si-ANCR), NC-ANCR (negative control group, cell line was transfected with lncRNA-ANCR negative control), and BLANK (blank control group, cells without treatment). Transfection was performed using Lipofectamine 2000 transfection kits (Invitrogen, Carlsbad, CA, USA), and all transfection operations were performed in strict accordance with Lipofectamine 2000 transfection instructions. Cells that were successfully transfected were prepared into cell suspension using Dulbecco\'s Modified Eagle Medium with 10% FBS and were seeded at a density of 1×10^5^ cells per well in 24-well plates, followed by incubation in a constant temperature incubator (5% CO~2~, 37℃, 95% humidity). Then, the expression of lncRNA-ANCR was detected by quantitative real-time polymerase chain reaction (qRT-PCR) after transfection for 48 h. The qRT-PCR detection --------------------- Total RNA was extracted and detected using RNAprep Pure tissue kits (TIANGEN Biotech Co., Ltd, Beijing, China) and a NanoDrop ND-1000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA), respectively. Meanwhile, cDNA was synthesized according to the instructions of the Takara PrimeScript RT reagent kit gDNA Eraser (TaKaRa Biotech, Kyoto, Japan) in accordance with the manufacturer\'s instructions. Then, qRT-PCR was performed on an ABI 7500 quantitative polymerase chain reaction system (Life Technologies Inc., Carlsbad, CA, USA). In detection of ANCR expression (forward, 5′-GCCACTATGTAGCGGGTTTC-3′; reverse, 5′-CCTGCGCTAAGAACTGAGG-3′), Runx2 expression (forward, 5′-CGGGTCTCCTTCCAGGAT-3′; reverse, 5′-GGGAACTGCTGTGGCTTC-3′), Osterix expression (forward, 5′-GAAGCGACCACTTGAGCACAT-3′; reverse, 5′-TGTCCAAACTCATCAATGTATCT-3′), and EXH2 expression (forward, 5′-TTGTTGGCGGAAGCGTGTAAAATC-3′; reverse, 5′-TCCCTAGTCCCGCGCAATGAGC-3′), the PCR program included 95℃ for 10 min, 40 cycles of 95℃ for 10 s, 60℃ for 20 s, and 72℃ for 34 s. GAPDH was used as an internal control (forward: 5′-CGAGCCACATCGCTCAGACA-3′; reverse: 5′-GTGGTGAAGACGCCAGTGGA-3′). The experiment was repeated three times. The relative expression of target genes was calculated using the 2^−ΔΔCt^ method.[@B15] Western blot ------------ Osteoblast cells were lysed with 100 µL/50 mL protein extract lysates. Briefly, after optical density (OD) was measured by a microplate reader (Varioskan LUX, Thermo Fisher Scientific) at 490 nm, proteins were separated by sodium dodecyl sulfatepolyacrylamide gel electrophoresis on 10% polyacrylamide gels and transferred to polyvinylidene fluoride membranes. After being blocked with 5% skim milk/BSA, the membranes were incubated with primary antibodies for EZH2 (1:1000, sc-131324, Santa Cruz Biotechnology, Delaware Ave Santa Cruz, CA, USA), RUNX2 (1:1000, sc-7177, Santa Cruz Biotechnology), OSTERIX (1:1000, cs-9272, Cell Signaling Technology, Inc., Boston, MA, USA), and glyceraldehyde-3-phosphate dehydrogenase (GADPH, 1:1000, sc-131325, Santa Cruz Biotechnology) overnight at 4℃. Then membranes were washed with Tris Buffered Saline Tween (TBST) and incubated with horseradish peroxidase-conjugated secondary antibody (1:5000, Shanghai Solar Biological Reagent Co., LTD., Shanghai, China). Protein bands were visualized using a gel imaging system (E-Gel Imager, Thermo Fisher Scientific). MTT assay --------- The 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay was performed to detect the proliferation of osteoblast cells. Simply, 200 µL of osteoblast cells digested by trypsin were seeded in 96-well plates at a density of 6×10^3^/well. Then, 5 µg of MTT (Cell Proliferation Kit I, Sigma Chemical Co., St. Louis, MO, USA) was added into each well. After 4 h of culturing, the medium was removed, and 150 µL of dimethyl sulfoxide were added into each well. OD at 490 nm at different time points (0 d, 1 d, 2 d, and 3 d) was detected using a microplate reader (Varioskan LUX, Thermo Fisher Scientific), followed by MTT curve construction. Apoptosis detection by flow cytometry ------------------------------------- Flow cytometry was performed to detect the apoptosis of osteoblast cells. Simply, osteoblast cells were digested with trypsin, and then incubated with 200 µL of Annexin V-FITC (BD Biosciences, San Jose, CA, USA) for 25 min and 10 µL of propidium iodide (PI) for 15 min under darkness. Cell cycle progression was subsequently monitored based on flow cytometry (CytoFLEX, Cell Lab Quanta, Beckman Coulter, Kraemer Boulevard Brea, CA, USA), and data were analyzed using Multi-Cycle AV software, Version 6.0 (Phoenix Flow Systems, San Diego, CA, USA). ALP staining assay ------------------ ALP staining was used for the ALP activity assay of osteoblast cells. Osteoblast cells were digested by trypsin and seeded in 24-well plates. Propanol (200 µL, 15 min), incubation solution (200 µL, 6 h), cobalt nitrate (200 µL, 15 min), and ammonium sulfide (200 µL, 5 min) were added successively. OD at 490 nm was detected using a microplate reader (Varioskan LUX, Thermo Fisher Scientific). Alizarin red staining assay --------------------------- The formation of calcium nodules was detected by alizarin red staining. Osteoblast cells were digested and seeded in 24-well plates. Polyformaldehyde (5%, 500 µL, 10 min) and alizarin (200 µL, 37℃, 30 min) were added to the medium successively, followed by photomicrographs for calcium nodules. RNA pull-down assay ------------------- Transcription of lncRNA-ANCR was performed using T7 RNA polymerase (Thermo Fisher Scientific). LncRNA-ANCR was further labeled with biotin using biotin RNA Labeling Mix (Sigma Chemical Co.). Then the transcription product was purified using RNeasy Mini Kit (Qiagen, GmbH, Hilden, Germany) with DNase I (Qiagen). Totally, 1 µL of labeled RNA in RNA structure buffer (10 mmol/L^-1^Tris pH7, 0.1 mol/L KCl, 10 mmol/L^-1^MgCl~2~) was used for RNA secondary structure formation. Meanwhile, 3 µg of osteoblast cells were harvested and lysed with RNA binding protein immunoprecipitation (RIP) lysis buffer (4℃, 1 h). Then, after centrifugation (12000×g, 4℃, 10 min), the remaining supernatant was mixed with 400 ng of biotin-labeled lncRNA-ANCR and 500 µL of RIP buffer, followed by incubation for 1 h at 37℃. Afterwards, 50 µL of streptavidin agarose beads (Invitrogen) were added into the supernatant for 1-h incubation. The level of lncRNA-ANCR was examined by qRT-PCR analysis based on RIP buffer washed liquid. Finally, Western blot was performed to detect binding between lncRNA-ANCR and genes. Osteogenesis model construction ------------------------------- A total of 18 male immunocompromised mice (20--25 g, 4--6 weeks) was randomly divided into a si-ANCR group, NC-ANCR group, and BLANK group. Transfected cell lines (1×10^6^/100 µL) were injected subcutaneously into the posterior limb implanted with hydroxyapatite-tricalcium phosphate (HA-TCP) mixture. Mice were killed 5 weeks after inoculation. Then, samples were fixed in 10% (v/v) formalin solution, dehydrated in ethanol, and processed routinely by embedding in paraffin. Finally, sections were used for hematoxylin-eosin (HE) staining and microscopic observation. Morphological observation ------------------------- Bone tissue of mice was fixed with 4% paraformaldehyde (Tianjin Institute of Chemical Reagents, Tianjin, China) for 24 h and decalcified with 20% formic acid and 15% ethylenediamine tetraacetic acid for 14 days. Followed by being embedded in paraffin, sliced (6 µm), and stained with HE, pathological changes in bone tissue morphology were observed under a microscope. Statistical analyses -------------------- All experimental data were expressed as mean±standard deviation (SD). Statistical analyses were performed by SPSS version 17.0 (SPSS Inc., Chicago, IL, USA) and GraphPad Prism 5.0 (GraphPad Software Inc., San Diego, CA, USA). Comparison between two groups was determined by t test. One-way analysis of variance test was used for comparison among different groups. *p* values less than 0.05 were considered to be significantly different. RESULTS ======= LncRNA-ANCR expression analysis ------------------------------- The PMOP model was established by bilateral ovariectomy in female mice. Six weeks after operation, the mice in the PMOP group showed significant decreases in bone mineral density, compared with the control group and sham group ([Fig. 1A](#F1){ref-type="fig"}). Moreover, Fluo 8 staining and confocal laser microscopy were used to detect intracellular calcium ion levels in PMOP mice. Compared with the control group and sham group, intracellular calcium levels in the PMOP group were significantly lower ([Fig. 1B](#F1){ref-type="fig"}), suggesting that the PMOP mouse model was successfully constructed. Furthermore, qRT-PCR analysis showed that the expression levels of lncRNA-ANCR in the PMOP group were significantly higher than those in the control group and sham group (*p*\<0.001) ([Fig. 1C](#F1){ref-type="fig"}). The result of lncRNA-ANCR siRNA transfection -------------------------------------------- Osteoblast cells from PMOP mice were cultured and transfected with si-ANCR. qRT-PCR analysis showed that the expression levels of lncRNA-ANCR in the si-ANCR group was significantly lower than those in the NC-ANCR and BLANK groups (*p*\<0.001) ([Fig. 2](#F2){ref-type="fig"}), which further indicated that lncRNA-ANCR was silenced by si-ANCR in osteoblast cells. Si-ANCR transfection promoted the proliferation and osteogenesis of osteoblast cells ------------------------------------------------------------------------------------ The effect of lncRNA-ANCR inhibition on the proliferation of osteoblast cells was evaluated using MTT assay. Compared with the NC-ANCR group and BLANK group, the proliferation ability of osteoblast cells in the si-ANCR group was significantly increased (*p*\<0.015, [Fig. 3A](#F3){ref-type="fig"}). Moreover, flow cytometry was used to detect the effect of lncRNA-ANCR inhibition on the apoptosis of osteoblast cells. The results showed that the apoptosis of osteoblast cells in the si-ANCR group (2.3%±0.13) were significantly lower than those in both the NC-ANCR group (11.73%±0.15) (*p*\<0.001) and BLANK group (10.7%±0.12) (*p*\<0.001) ([Fig. 3B](#F3){ref-type="fig"}). The effect of lncRNA-ANCR inhibition on ALP activity in osteoblast cells was detected using ALP staining. The result showed that the ALP activity (OD 490 nm) of osteoblast cells was significantly higher in the si-ANCR group than in the NC-ANCR group and BLANK group at 14 and 21 days post-transfection (*p*\<0.004) ([Fig. 3C](#F3){ref-type="fig"}). Furthermore, alizarin red staining showed that the transfection of si-ANCR significantly promoted the calcium deposition of osteoblast cells at 14 and 21 days post-transfection ([Fig. 3D](#F3){ref-type="fig"}). Si-ANCR transfection promoted osteoid formation in mice ------------------------------------------------------- Osteoblast cells were injected into the posterior limb of mice implanted with HA-TCP. HE staining showed that HA-TCP were greatly absorbed in the si-ANCR group, and a large number of red-stained osteoid-like tissues were observed. However, more HA-TCP remained in the NC-ANCR and BLANK group, and less osteoid-like tissues were observed ([Fig. 4A](#F4){ref-type="fig"}). The osteoid/total area was significantly higher in the si-ANCR group than in the NC-ANCR and BLANK group (*p*\<0.001) ([Fig. 4B](#F4){ref-type="fig"}). Expression of EZH2, RUNX2, and OSTERIX after si-ANCR transfection ----------------------------------------------------------------- The expression of EZH2, RUNX2, and OSTERIX in osteoblast cells was detected by Western blot ([Fig. 5](#F5){ref-type="fig"}). Compared with the NC-ANCR group and BLANK group, the expression of EZH2 in osteoblast cells of the si-ANCR group decreased significantly (*p*\<0.001), while the expression of RUNX2 and OSTERIX increased significantly (*p*\<0.001). Interaction between lncRNA-ANCR and EZH2 ---------------------------------------- RNA pull-down assay showed that ANCR could specifically bind to EZH2. However, ANCR-Mut, the truncated mutation control group, could not bind to EZH2 (*p*\<0.001) ([Fig. 6A](#F6){ref-type="fig"}). Furthermore, antibodies binding to EZH2 exhibited significant ANCR enrichment, compared with non-specific antibodies (*p*\<0.001) ([Fig. 6B](#F6){ref-type="fig"}). DISCUSSION ========== In many individuals, bone mass homeostasis starts failing in midlife, leading to bone loss, osteoporosis, and debilitating fractures.[@B16] During this process, osteoblast cells play an important role in bone formation and differentiation.[@B17] A previous study showed that osteoblast cells are closely associated with the development of diseases, such as prostate cancer, via regulating cell proliferation.[@B18] In certain bone disease like multiple myeloma, the inhibition of osteoblast cells directly leads to the occurrence of bone lesions.[@B19] A previous study indicated that osteoblast cells play a very important role in the progression of PMOP via influencing bone anabolic function.[@B9] Actually, the dysregulation of lncRNA in osteoblast cells contributes to the progression of osteoporosis.[@B20] As one kind of lncRNA, lncRNA-ANCR is essential for osteoblast differentiation.[@B11] Jia, et al.[@B12] showed that the down-regulation of lncRNA-ANCR promotes osteogenic differentiation of periodontal ligament stem cells. However, the function of lncRNA-ANCR on the progression of PMOP is still unclear. In this study, qRT-PCR detection showed that lncRNA-ANCR is up-regulated in an osteoblast group based on the PMOP mice model, which further indicated that lncRNA-ANCR might take part in the development of PMOP. Moreover, siRNA can interfere with the expression of specific genes with complementary nucleotide sequences by degrading mRNA after transcription.[@B21] To verify the function of target lncRNA in diseases, RNA interference has been established as an important biological strategy for gene silencing.[@B22] A previous study showed that the silencing of tet methylcytosine dioxygenase 2 (TET2) by siRNA can be used to reveal the function of TET2 in bone marrow cells in myelodysplastic syndromes.[@B23] In the current study, compared with two control groups, osteoblast cells in the si-ANCR group showed a lower apoptosis rate, higher proliferation, and more stable calcium deposition. Thus, we speculated that silencing expression of lncRNA-ANCR might promote the osteogenic ability of osteoblast cells in PMOP. EZH2 is a gene that participates in transcriptional repression.[@B24] It has been shown to be involved in establishing and maintaining gene repression through cell division.[@B25] A previous study showed that chromatin organization regulated by EZH2 is required for the migration of bone marrow-derived mesenchymal stem cells.[@B26] Chen, et al.[@B27] indicated that EZH2 regulates the differentiation of age-dependent mesenchymal stem cells into osteoblast cells. In mice models, EZH2 controls the bone formation and cell cycle progression during osteogenesis [@B28] As a target gene, EZH2 is commonly regulated by lncRNA in cell proliferation in disease.[@B29] A previous study showed that the down-regulation of lncRNA-ANCR suppresses cell invasion and migration by regulating EZH2 expression.[@B14] In breast cancer, the degradation of EZH2 mediated by lncRNA ANCR attenuated the invasion and metastasis of tumor cells.[@B30] In bone diseases, lncRNA-ANCR can regulate cell growth by interacting with the expression of EZH2.[@B31] Moreover, RUNX2 is an important transcription factor necessary for osteoblast differentiation and bone formation.[@B32] A previous study proved that RUNX2 induces osteoblast and chondrocyte differentiation and enhances their migration by coupling with PI3K-Akt signaling.[@B33] Actually, lncRNA can sponge miRNA to promote osteoblast differentiation through upregulating RUNX2.[@B34] Zhu, et al.[@B13] indicated that downregulated lncRNA-ANCR promotes osteoblast differentiation by targeting EZH2 and regulating Runx2 expression. In this study, RNA pull-down analysis showed that lncRNA-ANCR could specifically bind to EZH2. Importantly, Western blotting in the current study showed that, compared with two other groups, the expression of EZH2 in osteoblast cells of the si-ANCR group decreased significantly, while the expression of RUNX2 and OSTERIX increased significantly. Based on these results, we speculated that the specific binding of lncRNA-ANCR with EZH2 inhibits the expression of RUNX2, thus inhibiting the osteogenesis of osteoblast cells in PMOP. However, there are some limitations in this study, such as the sample size and lack of gene targeting drugs investigated based on lncRNA-ANCR. In addition, since osteoclast cells are also key in PMOP, the regulatory roles of lncRNA-ANCR on osteoclast cells were not analyzed. Thus, further research into these areas is needed. In conclusion, lncRNA-ANCR may take part in the progression of PMOP. We found that silencing of lncRNA-ANCR promotes the proliferation and osteogenesis of PMOP osteoblast cells in vitro, as well as osteoid formation in vivo. Furthermore, the specific binding of lncRNA-ANCR with EZH2 inhibited the expression of RUNX2, thus inhibiting the osteogenesis of osteoblast cells in PMOP. This work was supported by The Medical Science Leadership Training Program of Yunnan Province (D201639). The authors have no potential conflicts of interest to disclose. **AUTHOR CONTRIBUTIONS:** **Conceptualization:** Nuoya Cai.**Data curation:** Nuoya Cai.**Formal analysis:** Nuoya Cai.**Funding acquisition:** Chao Li.**Investigation:** Chao Li.**Methodology:** Chao Li.**Project administration:** Nuoya Cai.**Resources:** Nuoya Cai.**Software:** Chao Li.**Supervision:** Chao Li.**Validation:** Fuke Wang.**Visualization:** Fuke Wang.**Writing---original draft:** Nuoya Cai, Chao Li, and Fuke Wang.**Writing---review & editing:** Nuoya Cai, Chao Li, and Fuke Wang. {#F1} {#F2} {#F3} {#F4} {#F5} {#F6}

{#sp1 .115} {#sp2 .116} {#sp3 .117} {#sp4 .118} [^1]: Since this was written a corps in connexion with the Edinburgh Medical School has been formed, and 215 students are being trained.

Introduction ============ Play is a natural mode of expression of children and constitutes a fundamental aspect of children's life ([@B22]). Play can be considered a window through which one can observe a child's cognitive, affective, and social functioning ([@B43]). In particular, pretend play is a complex behavior involving symbolic expression of thoughts and feelings ([@B8]; [@B35]). Recent evidence indicates that pretend play has important connections to the development of cognitive, affective, and social skills ([@B28]). These connections make pretend play an important phenomenon to study, due to its capacity to provide insight into children's abilities and internal representations ([@B23]) and its appropriateness for investigating the "architecture" of the child's mind ([@B45]). In order to grasp the complexity of play, it seems necessary to evaluate it in terms of the dimensions of pleasure expressed, contents/themes, and linguistic and structural aspects, through a meaningful theoretical and methodological model founded on evidence-based procedures ([@B15]; [@B16]). Russ developed a model of pretend play and a scale to measure the play of children from 6 to 10 years old, the Affect in Play Scale (APS) ([@B34], [@B35]). The APS qualitatively and quantitatively assesses affective and cognitive components of symbolic play, through the use of a standardized coding system that allows for measuring affective dimensions in fantasy and cognitive dimensions of play. In the present study, [@B35], [@B36]) model, and its related APS, were used to evaluate the implicit component of the child's internal representations of disability in the context of symbolic play. The APS had never before been used to evaluate fantasy and expression of affect in symbolic play involving characters with a disability. A modified version of the APS play task was used by introducing a toy wheelchair among the standard play task materials which could be worn by a puppet to represent a character with a disability. To date, the scientific literature has identified three main models of disability: individual (including the medical model), social, and biopsychosocial ([@B4]; [@B46]). The individual model posits that disability is a direct consequence of disease, trauma, or other health conditions and, hence, that disability is an individual problem. In other words, it represents an individual-level deviation from biomedical norms of structure or function that requires medical care and treatment from health professionals ([@B5], [@B6]; [@B4]; [@B32]). Since treatment by the medical profession is central to this model, it is also referred to as the medical model of disability ([@B11]; [@B24], [@B26]; [@B46]). Unlike the individual model, the social model portrays disability as a cultural construct, the product of a particular social environment ([@B25]; [@B33]). Physical barriers or social attitudes are considered the origin of disability, because they prevent individuals with a "disability" from gaining access to virtual and real spaces and make social participation difficult. In this model disability is not an individual attribute, but rather a complex collection of conditions which require social actions rather than just medical treatment ([@B46]). Finally, since 2001 the *ICF: International Classification of Functioning, Disability and Health* has used a biopsychosocial---also known as interactive ([@B3])---model of human functioning and health, which represents an attempt to integrate the conflicting medical and social models. To achieve this "synthesis" ([@B46]), disability is treated not as a consequence of disease but as the outcome of three variables related to human health: health status, environment, and personal factors. In **Table [1](#T1){ref-type="table"}**, the main concepts of the three models of disability were summarized. ###### Summary of the concepts of the tree main models of disability. Model of disability Causes of disablement The model's focus Interventions --------------------- -------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------ Medical/Individual Consequences of disease or disorder Sick individual who requires medical treatments provided by health professionals Rehabilitative to "fix" the sick individual Social Social barriers Cultural stereotypes, medical norms, and environmental hindrances Political and cultural to break down social barriers Biopsychosocial Components of health (health structures and functions, personal and environmental factors) Human beings functioning in their context Shaping the health system to meet the individual's functioning needs, and minimizing social barriers Disability is a complex phenomenon and models of disability allow people to make sense of emotional responses to disability, to process thoughts and structure knowledge about disability, and to make decisions and judgments to which it is relevant ([@B7]). Children's models of disability were first studied by [@B19], who reported that the youngest group of children (6--8 years old) thought of people with disabilities as being sick. This early representation of disability is consistent with the individual model of disability and was independent from parents' explanations and representations of disability. Older children (9--11 years old) had more knowledge of disability and endorsement of stereotypical beliefs tended to be lower, as children tended to espouse their parents' representations. Research based on Piagetian stage theory ([@B29], [@B30], [@B31]) claimed that young children's difficulty in conceptualizing disability was the result of their being too cognitively immature ([@B13]; [@B14]; [@B12]). According to these approaches, children can only process and structure knowledge about disability across a range of explanations for disabilities identifiable among the three main models of disability, which include physical, biological, and psychological causes of disablement and health, when they reach approximately 11 years of age. Although debates continue, studies of cognition in infancy demonstrate that knowledge begins to emerge early in life and constitutes part of humans' innate endowment ([@B2]; [@B41]; [@B1]), even including an early understanding of disease causality ([@B42]; [@B38]). In line with this work, challenging Piagetian framed research on children's knowledge of illness causation, [@B40] explored children's understanding of the causal origins of disabilities. They found that children of all ages showed a preference for physical and biological causes of disability (consistent with an individual/medical model) and rejected social--psychological causal explanations (consistent with a social model). The aim of the present paper was to explore the type of representation of disability and the effect of the presence of a disabled character on the expression of affect and on the cognitive dimensions of children's play. In accordance with previous findings ([@B40]; [@B19]), we expected that, when one of the two puppets in the modified APS play task was in a wheelchair (disabled puppet), children would tell stories in which the psychological constructs, sequence of actions, and affective expressions were consistent with the individual/medical model of disability. We also expected that pretend play with a disabled puppet would display a strong connection between affective categories such as nurturance/affective and sadness/hurt and actions such as providing and receiving care. Materials and Methods {#s1} ===================== Participants ------------ Thirty Italian 6- to 10-year-olds were recruited from a public primary school. All of them were Caucasian, attending mainstream classes and had no declared disability, and sixteen (53.3%) were female. Parents provided written, informed consent to their children's participation in the study and to the videotaping of their children in the play session. All the children also consented personally after the researcher explained that he or she would like to watch the child playing with two puppets for a few minutes. The Ethics Committee of the Department of Philosophy, Social and Human Sciences and Education at the University of Perugia reviewed and approved the study. The study presented "no more than minimal risk". Instrument ---------- *The standard APS* ([@B35]) is a standardized instrument for evaluating cognitive and affective dimensions in pretend play in children from 6 to 10 years of age, based on an observational procedure that focuses on different children's behaviors during a semi-structured, 5-min, evidence-based play task. The APS has been employed in numerous studies that have demonstrated its good psychometric characteristics. Good inter-rater reliability was achieved, with Cohen's kappa values ranging from 0.70 to 0.90 ([@B35]). The APS play task is video recorded and requires two neutral-looking hand-puppets, representing a boy and a girl, and some little wooden blocks. The instructions are standardized and facilitate the child playing freely, according to his or her skills, age, characteristics, and preferences. The researcher introduces the two puppets and the blocks to the child and asks him or her to play with them for five minutes. *The modified APS*, an adapted APS play task procedure, was developed for this study and involved changes to the APS materials and instructions. The modified APS play task included a wheelchair toy in addition to the two puppets and the blocks. The experimenter introduced the wheelchair toy (wearable by one puppet) and asked the child which of the two puppets (boy or girl) was disabled and which was not. Henceforth, the terms "standard APS" and "modified APS" are used to refer, respectively, to the original instrument ([@B35]) and to the version developed for this study. *A semantic discrimination task* ([@B19]) was administered to assess children's comprehension of the concept "disabled." The child was presented with six stimuli (2 photographs of people with disabilities; 2 photographs of people without disabilities; 2 words: "handicapped" and "normal") and asked sort them by placing them in one of two labeled baskets. The baskets were labeled "disabled" and "normal." A child was considered to have passed the test if he or she demonstrated understanding of the difference between the semantic categories represented by the stimuli. The test was repeated until the child had either demonstrated that he or she could correctly discriminate between the stimuli or it was clear that he or she was unable to do so. Measures -------- ### The APS Rating Scale The APS rating scale ([@B35]) was used to analyze the standard and the modified APS play tasks. The APS scores used in the present study belong to two domains: affective and cognitive. Affective domain: 1. *Total Frequency of Affective Expressions score*: is measured by the sum of eleven affective categories (happiness/pleasure, nurturance/affection, oral, sexual, competition, anxiety/fear, sadness/hurt, frustration/disappointment, aggression, anal, and oral aggression). The categories can be applied to verbal or nonverbal expressions, and can be an affect state ("This is fun") or an affect theme ("This bomb is going to explode"). 2. *Frequency of Positive Affect score*: sum of the five affect categories: happiness/pleasure, nurturance/affection, competition, oral, and sexual. 3. *Frequency of Negative Affect score*: sum of the six affect categories (aggression, sadness/hurt, anxiety/fear, frustration/disappointment, oral aggression, and anal). 4. *Variety of Total Affect Categories score*: is a count of affect expressions across the 11 possible categories. 5. *Variety of Positive Affect Categories*: is a count of affect expressions across the five positive categories. 6. *Variety of Negative Affect Categories*: is a count of affect expressions across the six negative categories. Cognitive domain, rated on a five-point Likert-type scale: 1. *Organization*: includes the quality and the complexity of the play plot. 2. *Elaboration*: measures the amount of embellishment in the play in terms of theme, facial expression, voice tones and character development. 3. *Imagination*: involves number of ideas, novelty, and fantasy of the play in terms of the presence of themes outside of everyday experience. 4. *Comfort*: rates the child's overall level of enjoyment engaging in pretend play and her ability to be involved in play. ### Representation of Disability Representation of disability was only scored for the modified APS. Expressions were classified assigning them to one of three categories of disability model (medical, social, and biopsychosocial), as follows. #### (i) Individual/Medical Model Statements in which the disability was related to the health of the disabled puppet. This category also included statements on impairments assigned to the disabled puppet. This category includes: (i) all statements implying that the disabled puppet (i.e., puppet in the wheelchair) was considered morally or ethically responsible for his or her disability; (ii) any judgment based on the appearance of the disabled puppet, e.g., beauty, or ugliness; (iii) any statement assigning responsibility for the disability to an external spiritual, vital, or religious force. #### (ii) Social Model All statements that attributed the disability to factors beyond control of the disabled puppet, such as architectural and cultural environmental factors (barriers, rules, regulations, etc.), or to human attitudes and prejudices. #### (iii) Biopsychosocial Model As the biopsychosocial model is a composite, we included in this category articulations attributing disability to a complex interaction of medical, environmental, and socio-relational factors, including a clear reference to individual functioning (health or disease). Procedure --------- ### Administration Procedure After a parent had provided written, informed consent for his or her child's participation, the researcher explained to the child that he or she would like to learn about play by watching the child play with the two puppets for a few minutes and asked for the child's own consent to this. All children were assessed individually. First, the semantic discrimination task was administered to assess comprehension of the concept "disabled." Then, children were invited to play using both the standard APS and the modified APS play task sequentially. The two sessions (standard and modified) were administered consecutively and were videotaped. The procedure lasted roughly fifteen minutes (5′ semantic discrimination task; 5′ standard APS; 5′ modified APS). ### Coding Procedures Recordings of the two play sessions (standard and modified) were transcribed verbatim and the video recordings scored, using the APS scoring system procedure ([@B35]), by two independent trained coders. The modified APS play task verbatim transcriptions were also scored by two independent trained coders. The score for each disability model (individual/medical, social, biopsychosocial) is obtained by summing the child's expressions attributable to each model. Inter-rater reliability was assessed using the Pearson correlation coefficient on 20 randomly selected protocols. The correlations between the two judges for all the scores ranged from 0.87 to 0.94. Data Analysis ------------- Descriptive statistics (mean, *M*; standard deviation, *SD*) were calculated to provide a profile of the sample. Inferential statistics (multivariate ANOVA and univariate ANOVA) were used to compare children's play performance on the standard APS and modified APS, and *t*-tests for unpaired samples and effect sizes for Student's *t-*test (Cohen's *d*) were calculated to compare children's play performance on the standard APS with normative Italian data. Chi-square tests were used to explore the association between children's gender and models of disability, and correlational analysis was used to explore the association between children's age and models of disability. Data were analyzed using IBM^®^SPSS Statistics 23. Results ======= Sample ------ Fifty-five out of 63 primary school pupils invited to play completed the experiment (male: *n* = 28, 50.9%; female: *n* = 27, 49.1%; *M* age = 8.10 years, *SD* = 1.45, range: 6--10) (**Table [2](#T2){ref-type="table"}**). Eight pupils stopped playing during one or both of the 5-min task periods, not playing after a 2-min period. For this reason, the people pupils were excluded from the analyses. ###### Sample profile: school grade and gender. Male Female Total ----------- -------- --------- -------- --------- -------- --------- I 4 14.3 6 22.2 11 18.2 II 5 17.9 7 25.9 12 21.8 III 5 17.9 5 18.5 10 18.2 IV 5 17.9 4 14.8 9 16.4 V 9 32.1 5 18.5 14 25.5 **Total** **28** **100** **27** **100** **55** **100** Twenty-five pupils played with the standard APS play task first. Affective and Cognitive Components in the Standard APS Play Task ---------------------------------------------------------------- Results for the APS standard condition are reported in **Table [3](#T3){ref-type="table"}**. Values available from the normative Italian sample are reported in parentheses ([@B17]). ###### Means and standard deviations for the sample. Frequency Variability ------------------------------------------ --------------- --------------- ------------- ------------- Total Frequency of Affective Expressions 14.83 (12.61) 12.99 (11.79) 3.38 (3.38) 2.64 (2.21) Frequency of Positive Affect 10.73 (7.63) 9.98 (8.53) 1.88 (1.74) 1.33 (1.28) Frequency of negative Affect 4.11 (4.96) 4.78 (6.31) 1.58 (1.64) 1.67 (1.32) Aggression 0.55 (1.50) 1.24 (3.71) 0.22 0.42 Nurturance/affection 1.35 (1.14) 2.06 (1.81) 0.42 0.49 Happiness/pleasure 6.50 (4.12) 5.36 (5.17) 0.75 0.44 Anxiety/fear 0.63 (0.76) 1.16 (1.87) 0.29 0.46 Sadness/hurt 0.78 (1.05) 1.62 (2.24) 0.33 0.47 Frustration/disappointment 1.22 (1.34) 2.27 (2.09) 0.39 0.49 Competition 0.36 (0.68) 1.42 (2.29) 0.09 0.29 Oral 2.11 (1.45) 4.53 (3.44) 0.36 0.48 Oral aggression 0.25 (0.08) 0.93 (0.47) 0.09 0.29 Anal 0.51 (0.25) 1.17 (0.91) 0.22 0.42 Sexual 0.36 (0.25) 1.11 (1.13) 0.13 0.34 Organization 2.42 (2.32) 1.28 (1.15) -- -- Elaboration 2.36 (2.21) 1.27 (1.03) -- -- Imagination 2.35 (2.21) 1.04 (1.03) -- -- Comfort 2.73 (2.96) 1.31 (1.05) -- -- In parenthesis are data from the normative Italian sample. In order to compare children's play performance on the standard APS with normative Italian data, *t*-tests for unpaired samples and effect sizes in terms of Cohen's *d* were calculated. According to [@B10], effect size values of 0.2, 0.5, and 0.8 are considered small, medium, and large. The comparison with data from the Italian normative sample indicated that, for the standard APS play session, children in the present sample displayed typical play, similar to the normative samples in all of the APS scores. The only two exceptions to this trend were the Frequency of Positive Affect (*t* = 2.61, *df* = 1264, *p* \< 0.01, *d* = 0.333) and the happiness/pleasure category (*t* = 3.33, *df* = 1264, *p* \< 0.01, *d* = 0.452). Both Cohen's *d* results were of a medium size. Affective and Cognitive Components in the Modified APS Play Task ---------------------------------------------------------------- A one-way ANOVA of the frequency of the 11 affect categories revealed effects of APS condition on one category of positive affect, "nurturance/affection" \[*F*(1,110) = 11.98, *p* = 0.01, η^2^ = 0.100\], and one category of negative affect, "sadness/hurt" \[*F*(1, 110) = 9.82, *p* \< 0.01, η^2^ = 0.083\]. An effect size measured using partial η2 of 0.01 constitutes a small effect, 0.06 a medium effect, and 0.14 a large effect. Both affect categories were more frequent in the modified APS condition (nurturance/affection: *M* = 3.93, *SD* = 5.13; sadness/hurt: *M* = 2.49, *SD* = 3.71). There was no effect of APS condition on cognitive components (Organization, Elaboration, Imagination and Comfort) and on variety of affect. Children's Models of Disability ------------------------------- During the modified APS session, 29 out of 55 pupils expressed concepts relevant to a disability model (i.e., occurrence of a disability model). Twenty-six out of those 29 pupils only represented disability through the medical model, two through the medical and social models, and one through only the social model. None of the pupils used the biopsychosocial model. Amongst the 29 pupils who referred to disability in the modified APS session, the mean frequency of statements related to the medical model was 2.07 (*SD* = 1.28), while the mean frequency of mentions of the social model was 0.27 (*SD* = 0.92). Chi-square tests indicated that there was no relationship between gender and mentions of a disability model \[χ^2^(1, *N* = 29) = 0.31, *p* = 0.58\] and no relationship between gender and the relative frequency of the various disability models \[χ^2^(5, *N* = 29) = 5.11, *p* = 0.40\]. There was a correlation between age and mentioning at least one model of disability \[*r*(55) = 0.27, *p* \< 0.05\]. There was no relationship between puppet gender and child's assignment of puppet to the wheelchair in the modified APS. Finally, **Table [4](#T4){ref-type="table"}** shows which puppets were worn by children before starting timing of the two APS sessions (the child must put on the puppets before timing the session; [@B35]). Only one child wore the puppet with the wheelchair: 54 children never wore the puppet with the wheelchair freely (i.e., before the experimenter invited him or her to put on) (**Table [4](#T4){ref-type="table"}**). ###### Puppet wearing in the standard and modified APS play tasks. Puppet(s) worn Standard APS Modified APS -------------------- -------------- ------------------------------------- Neither 16 20 Both 37 1 One (not-disabled) 2 (Males) 34 (male *n* = 15; female *n* = 18) One (disabled) -- 0 A one-way ANOVAs on puppet wearing (1 = wearing neither; 2 = wearing both; 3 = wearing only one (non-disabled)) revealed an effect of APS condition (standard APS; modified APS) on puppet wearing, *F*(1,110) = 11.86, *p* = 0.01. Discussion ========== In line with previous studies ([@B21]; [@B39]), the present paper claims that pretend play is a fundamental tool for investigating children's affective, cognitive, and social functioning. Studying children's pretend play can also provide insight into human cognitive architecture and its development ([@B45]). With regard to the affective and cognitive components, children showed a play profile in line with Italian norms, playing in a typical way for their age ([@B17]). Children showed more expressions of empathy or sympathy and helping or supporting with another character (nurturance/affection) and more expressions of pain, sadness, or loneliness (sadness/hurt) when playing with the wheelchair toy (modified APS play task). The organization and elaboration of the play plot and the child's imagination and comfort in play, namely the cognitive components of the APS ([@B9]), remained quantitatively unchanged across the experimental conditions (standard and modified APS play tasks). Disability---suggested by the introduction of the wheelchair toy in the modified APS---did not affect the construction of the structural components of the play plot and children remained available to symbolically play as usual. At the same time, children were shown to be very sensitive to the presence of the wheelchair toy, showing a general attitude of compassion and sadness toward disability. We infer that, from the point of view of the play plot, a wheelchair toy has the same weight as all the other play elements (hand-puppets and wooden blocks), by demonstrating that for a child one toy is like another. On the other hand, the child proves to know the emotional value of the wheelchair, since it is immediately associated with the disease and, therefore, with feelings of care, compassion, and sadness; this explains the variations measured in the affective components of play, but not the cognitive ones. With regard to the children's models of disability, data confirm the finding of [@B19] that, for children, individuals with disabilities are mainly thought of as being sick. When children are involved in pretend play, from which concepts of disability emerge, these concepts are almost exclusively related to the individual/medical model of disability. As we expected, the children imagine the disabled puppet as sick (e.g., disabled puppet says: "I'm sick because my legs are broken"), or requiring medical treatment (e.g., disabled puppet says: "The physician told me that within few days I'll be better"), and the non-disabled puppet as a provider of health care (e.g., non-disabled puppet says to disabled one: "The physician told that you must take these pills that make you feel better"). These findings were reflected in the higher frequency of nurturance/affection and sadness/hurt in the modified APS condition, as expected. As a cognitive organizer, a model of disability helps people to identify and understand the causal origins of disability ([@B19]). The individual/medical model directs understanding of disability to the physical and biological condition of an individual, rejecting social and cultural determinants of disability. Therefore, our findings are in line with those of [@B40], who found that 4- to 11-year-old children showed a preference for physical and biological causes of disability and rejected social--psychological causal explanations. The 26 pupils out of 55 who did not express concepts referable to a disability model were the youngest. It suggests that the capacity to tell stories in which disability is salient develops with age. At an early age, disability does not seem to attract children's attention and is not featured in their stories. When disability is mentioned in a story, however, it emerges as the most salient element and drives the narrative. The disability element in children's stories tends to conform mainly, if not exclusively, to 'schemata' ([@B7]) from the individual/medical model of disability. That the youngest did not express concepts referable to a disability model highlights what [@B40] suggested with regard to open-ended verbal methods: young children may have been so concerned with spontaneously generating a causal explanation that they were unable to verbalize a cause. In fact, when a forced-choice paradigm is adopted, as in [@B40] and [@B19], young children show some causal knowledge of disabilities. Another interesting finding related to puppet wearing behavior before starting timing of the two APS sessions. In the standard APS, the majority of children wore both puppets, whereas in the modified APS most children only wore the non-disabled puppet freely. This behavior seems consistent with studies by [@B27] and [@B18], which demonstrated that disability elicits disgust and, hence, avoidance behavior. As the medical model of disability schematizes an aspect of human diversity by providing a cognitive organizer, such as a "frame" ([@B20]), the avoidance behavior provides a "script" ([@B37]) as well. No wonder, then, that, despite voicing caring attitudes toward the puppet in the wheelchair, almost none of the children elected to wear this puppet freely. Conclusion ========== Findings from this study support the validity of the APS to evaluate differences in children's play in different situations, confirming the validity of the scale in showing the affective nuances of the play of school-aged children. Use of the APS also allowed for confirmation of what previous research on disability representations and attitudes in children aged 6--10 years suggested. In particular, the perceptual salience of disability increases as the age of children grows. In the present study, all children who paid attention to disability tended to only describe and explain it in its biological and physical dimensions, neglecting any social and cultural determinants of disability. Therefore, the perspective of disability emerging from the children's attitudes in the APS play task was fully compatible with the individual/medical model of disability as defined in the scientific literature. This compatibility was further confirmed by the fact that sadness/hurt affect categories prevailed in children's storytelling when they interacted with the disabled character (puppet in the wheelchair), along with the nurturance/affection category. In fact, the prevalence of sadness/hurt feelings is consistent with the personal tragedy view ([@B44]) on which the medical model of disability is grounded. The results obtained also suggest that disability is strongly and stereotypically associated with a negative and unpleasant dimension of existence, providing evidence for a cognitive mechanism underpinning the cultural construction of the individual/medical model of disability. Moreover, according to [@B40] research on children's understanding of the causal origin of disability, our results challenge Piagetian's assumption ([@B29], [@B30], [@B31]) that young children conceptualize disability with difficulty. Children, indeed, were not surprised by diversity of disability, and demonstrated possession of cognitive schemata to elaborate on it and congruent emotions to respond to it. On the basis of these results, we claim that the child's education about disability should not have the aim to introduce the concept of diversity, but rather to enhance their views on disability, modeling the social and cultural dimensions of diversity among the disabled that seem to be lacking in children. Future research might overcome some limitations of the present study. These include, for example, increasing the sample size, given that the sample of children recruited in the present study prevents us from generalizing the results as representative of the Italian child population. In addition, in order to better investigate the evolution of the concept of disability, a longitudinal study would be advisable. Finally, the present research reflects the Italian socio-cultural context, where the pressure to include disabled people in social life is a political and legal norm but not yet widespread in the social fabric. It would be interesting to see how the disability models feed children's narratives on disability in other socio-cultural contexts. Author Contributions ==================== Conceived and designed the experiments: SF and FM. Performed the experiments: AC. Analyzed the data: SF, FM, AC, and CM. Contributed analysis tools: SF, FM, AC, and CM. Wrote the paper: SF, FM, AC, and CM. Conflict of Interest Statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. [^1]: Edited by: *Gianluca Castelnuovo, Università Cattolica del Sacro Cuore, Italy* [^2]: Reviewed by: *Gianluca Serafini, University of Genoa, Italy; Michelle Dow Keawphalouk, Harvard University -- Massachusetts Institute of Technology, United States* [^3]: This article was submitted to Psychology for Clinical Settings, a section of the journal Frontiers in Psychology

INTRODUCTION {#sec1-1} ============ Ameloblastic carcinoma (AC) is characterized by malignant cytological features in combination with the overall histological pattern of ameloblastoma. The term malignant ameloblastoma is confined to those ameloblastomas that metastasize despite an apparently typical benign histology in both the primary and the metastatic lesions.\[[@ref1]\] The incidence ratio of AC to malignant ameloblastoma is 2:1.\[[@ref2]\] In this article, we report two cases of AC, one involving left maxilla (primary type), which would be an addition to existing 35 cases previously reported and the other one affecting the left side of mandible arising ex ameloblastoma. CASE REPORTS {#sec1-2} ============ Case 1 This paper reports a case of a 21-year-old male patient who reported to our department with the complaint of a painless, rapidly growing swelling, over the left side of mid-face since 4 months \[[Figure 1](#F1){ref-type="fig"}\]. He had difficulty in chewing on the left side and three episodes of epistaxis for the last 1 month. On extraoral examination, a firm diffuse, non-tender swelling with normal overlying skin over a left side of mid face measuring 5 cm × 6 cm × 4 cm was observed. The swelling was seen extending superiorly from left infraorbital region inferiorly up to 1 cm below the commissure, medially from the philtrum region obliterating the nasolabial fold and extending laterally up to 3 cm anterior to the left ear lobule. Intraorally, a firm swelling involving left maxilla extending from distal aspect left maxillary central incisor up to the second molar region with more buccal cortical expansion, obliterating left buccal vestibule and displacing lateral incisor, canine, premolars and maxillary first molar palatally, but without any mobility was observed. All associated teeth were vital. Swelling was negative on aspiration. On neck examination, no cervical lymphadenopathy was found. Orthopantomograph showed a unilocular radiolucent lesion affecting the left maxilla displacing lateral incisor, canine, both premolars and molars \[[Figure 2](#F2){ref-type="fig"}\]. The patient\'s vital signs were as Followes: Heart rate -- 76/min, blood pressure -- 126/78 mm Hg and respiratory rate -- 14/min. His medical condition was not suggestive of any systemic diseases and the biochemical and hematological investigations were all within the normal limits. {#F1} {#F2} On performing incisional biopsy, it was found that the outer cortical plate was lost, the growth being attached to the mucoperiosteum. The gross appearance of the specimen was slimy in character. The initial histopathology report was that of follicular variant of ameloblastoma with a, suspicion of AC. The patient was planned for wide resection of tumor with normal margins through the modified Weber Ferguson approach under general anesthesia \[[Figure 3](#F3){ref-type="fig"}\]. Resection was done with adequate margins, orbital floor was spared and proper hemostasis achieved. The surgical procedure was uneventful. The defect was closed with a surgical obturator. The resected specimen was sent for histopathologic examination \[[Figure 4](#F4){ref-type="fig"}\]. The histopathological study revealed that the tumor was composed of follicles of odontogenic epithelial islands in a dense, mature connective stroma. Islands resembled dental organs with peripheral layer of ameloblast- like cells and centrally stellate reticulum-like cells. Furthermore seen was acanthomatous change (squamous metaplasia with keratin pearl formation) within the follicle, which also showed comedo necrosis \[[Figure 5](#F5){ref-type="fig"}\]. These cells showed nuclear hyperchromatism, nuclear pleomorphism, increased nuclear-cytoplasmic ratio and mitotic figures or abnormal mitosis \[[Figure 6](#F6){ref-type="fig"}\]. The histopathological examination also showed clear cells and fascicles of plump spindle cells merging with other cells of the follicular ameloblastomatous epithelium. This confirmed the diagnosis as AC (spindle cell variant). The patient was then referred to radiotherapy department, where he received 25 fractions of 1.8 Gy each for a period of 5 weeks (45 Gy). Post-radiation mucositis was reported. {#F3} {#F4} {#F5} {#F6} The patient was regularly followed-up for 1 month interval for a period of 6 months. To rule out distant metastasis, he was sent to the Department of Pulmonary Medicine and Gastroenterology of this institution, but they have cleared for any distant metastasis. Presently, after the 15 month postoperative period, there has been no recurrence at the primary site and no distant metastasis or regional lymphadenopathy has been observed. Patient is regular to his follow-up and maintaining his oral hygiene \[[Figure 7](#F7){ref-type="fig"}\]. {#F7} Case 2 {#sec2-1} ------ This was a case of a 70-year-old male patient who reported to our department with the complaint of rapidly growing painless swelling of the left side of lower jaw with difficulty in chewing, multiple mobile teeth over left side of both jaws and progressive loss of teeth for last 6 months \[[Figure 8](#F8){ref-type="fig"}\]. He had the past history of a similar lesion over the left lower jaw, which was treated by en-bloc resection 20 years back. The histopathology then was suggestive of ameloblastoma. {#F8} On extraoral examination, a firm diffuse non-tender swelling over left side of face extended superiorly from the left infraorbital region to the lower border of mandible inferiorly. Medially, it extended up to one-third of both upper and lower lips with an elevated angle of mouth and laterally 1 inch anterior to the angle of mandible with more medio-lateral expansion measuring about 10 cm × 8 cm × 6 cm. Neither paresthesia was associated with the swelling nor was any regional lymph node palpable. Intraoral examination revealed a painless firm proliferative growth from midline up to the left anterior border of ramus with an obliterated left side buccal sulcus and extended lingually with expansion of the lingual cortex and a medially displaced tongue \[[Figure 9](#F9){ref-type="fig"}\]. The swelling was negative on aspiration. No cervical lymphadenopathy was found. Orthopantomogram showed a multilocular radiolucent lesion involving left angle and ramus of mandible with multiple areas of resorption over anterior border of ramus and maxillary alveolus distal to premolar region \[[Figure 10](#F10){ref-type="fig"}\]. {#F9} {#F10} His medical history was suggestive of Type 2 diabetes mellitus, under irregular medication. His vitals were as followes: Blood pressure -- 148/96 mm Hg, heart rate -- 76/min, respiratory rate -- 14/min and afebrile. Fasting blood glucose (FBG) came out with 202 mg/dl. Endocrinology consultation was done and he was advised injection Human insulin 28 U in three divided doses 6/8/14 U subcutaneously. After 3 days, the FBG level reported to be 92 mg/dl. The patient was then planned for an incisional biopsy under the local anesthesia. It was suggestive of follicular variant of ameloblastoma. He was then taken up for surgical resection of the lesion with 2 cm of bony margins on the symphysis region. After hemi-mandibulectomy, the defect was reconstructed with the placement of 2.4 mm titanium reconstruction plate with condyle and surgical wound was closed primarily in layers \[[Figure 11](#F11){ref-type="fig"}\]. Postoperative recovery was uneventful and the surgical wound healed well. The resected specimen was sent for histopathologic examination \[[Figure 12](#F12){ref-type="fig"}\]. To our surprise, the histopathology report came out with diagnosis of AC. He was sent to radiotherapy department for further intervention, but he refused to undergo it. {#F11} {#F12} After about 8 months, the patient again turned up with a firm, tender swelling over the left side mid and lower face extending to the neck \[[Figure 13](#F13){ref-type="fig"}\]. The computed tomography (CT) scan showed a hypodense image of approximately 12 cm × 16 cm size present over left side mandibular and submandibular region. He was planned for resection of tumor under general anesthesia. The submandibular incision was given and the tumor appeared firm in consistency, surrounding the reconstruction plate, it was in close approximation to the great vessels and while removing it, the left internal carotid artery got injured; the vessel was clamped and repaired with 6-0 prolene suture. The tumor was removed with normal margins along with the reconstruction plate. Hemostasis achieved and closure done in layers. The postoperative recovery was delayed, with unstable blood pressure. Patient was shifted to intensive care unit and placed on ventilator support. The next day right side hemiparesis of whole body was detected and the CT scan showed left parietal infarct. Serum urea and creatinine were also marked elevated. After 3 days, the patient died due to multiorgan failure followed by cardiorespiratory arrest. {#F13} DISCUSSION {#sec1-3} ========== Odontogenic tumors are rare tumors of tooth forming apparatus involving jaw bones. In North America, between 11% and 24% of odontogenic tumors are ameloblastomas,\[[@ref2]\] while among sub-Saharan Africans it accounts for 66-99%.\[[@ref3]\] The tumor is frequently quoted as comprising 1% of all oral tumors among Caucasians,\[[@ref4]\] approximately 80% are found in the mandible and 20% in the maxilla. Mostly it is asymptomatic and the symptoms appear with the expansion of the jaw. In 1955, Small and Waldron showed that 47% of maxillary ameloblastomas occur in the molar region, 33% occur in the antrum and floor of the nose, 9% occur in both the premolar and the canine regions and 2% occur in the palate.\[[@ref5]\] There are three variants of this tumor, including the solid or multicystic variant, the unicystic variant and the peripheral variant. A review of the international literature by Reichart *et al*. in 1995 found the solid or multicystic variant to be the most common, comprising 92% of the 3677 cases of ameloblastoma, while the unicystic and peripheral variants accounted for 6% and 2% of the cases, respectively.\[[@ref6]\] AC is an extremely rare malignant odontogenic epithelial neoplasm that may arise *de novo* or from a pre-existing odontogenic lesion.\[[@ref1][@ref7]\] There are about 92 cases reported from 1984 to 2012 in scientific literature. The age of presentation ranges from 7 to 91 years. Males are more frequently affected with M:F ratio of 2.3:1. The most common site is mandible with about 56 reported cases, 35 cases in maxilla and in only one case arising from anterior skull base.\[[@ref8]\] Posterior portion of jaw is more commonly involved. Clinically, AC causes expansion of the jaw, grows rapidly, frequently causes pain and often results in perforation of the cortex. Involvement of the nasal cavity is usually related to local invasion of the maxillary ACs. Although regional and distant metastasis is the feature of malignant ameloblastoma, but in few cases of AC, these lesions have been known to metastasize mostly to the lung or regional lymph nodes.\[[@ref9]\] Few cases reported with metastasis to the brain, bone marrow and liver.\[[@ref10][@ref11][@ref12][@ref13][@ref14]\] MacIntosh pointed out that the first site of reported ameloblastic metastasis was the lung and originally believed to be due to aspiration from the oral lesion rather than a true hematogenous or lymphatic spread. Indeed, enucleation and curettage surgeries might liberate neoplastic cells into the upper airway that could find their way into the lower airway.\[[@ref15]\] Zwahlen *et al*. reported a case with histologically proven myocardial metastasis of a maxillary malignant ameloblastoma.\[[@ref16]\] The pathogenesis for malignant transformation of ameloblastoma may occur either spontaneously or because of induction following chemotherapy or post-surgical radiation.\[[@ref6][@ref17]\] Aggressive behavior and metastatic potential associated with ameloblastoma has been noted in two of its variants -- one is the granular cell type and the other clear cell type.\[[@ref18][@ref19]\] Pulmonary metastasis was reported in two of the 20 granular cell ameloblastomas reviewed by Hartman.\[[@ref20]\] Furthermore, the cases of malignant ameloblastoma reported by Tsukada *et al*.,\[[@ref21]\] Hoke and Harrelson\[[@ref22]\] were of granular cell type. Histologically it shows features of ameloblastoma with cytologic atypia, high mitotic index, reverse polarization, peripheral palisading and necrosis, neural and vascular invasion. There are various tumors which may mimic AC histologically or clinically which includes primary intra-alveolar epidermoid carcinoma,\[[@ref23]\] squamous cell carcinoma arising in the lining of an odontogenic cyst,\[[@ref24]\] acanthomatous ameloblastoma and kerato-ameloblastoma,\[[@ref25]\] squamous odontogenic tumor,\[[@ref26]\] calcifying epithelial odontogenic tumor, salivary gland neoplasms such as pseudoadamantine adenocarcinoma, ductal carcinoma, high-grade mucoepidermiod carcinoma and metastatic carcinoma to the jaws from lung, breast and gastrointestinal tract. Yoon *et al*. compared the immunohistochemical markers and found that the significant expression of cytokeratin 18, parenchymal matrix metalloproteinases-2 (MMP-2), stromal MMP-9 and Ki-67 differentiated AC from ameloblastoma.\[[@ref27]\] The management of the AC is similar to that for ameloblastoma along with radiotherapy as reported by various authors. Literature also suggests only few case reports in which neck was addressed. The neck dissection was not performed in both the cases because no neck nodes were detected on clinical examination. Addressing no neck dissection in cases of squamous cell carcinoma is still in controversy. As our diagnosis was AC with no palpable neck nodes, we have not addressed the neck. Prior to the mid-1980s, the literature was lacking with well-documented evidence regarding the relative radioresponsiveness of the ameloblastoma. Until this point, it was believed that ameloblastoma was radioresistant. In 1984, Atkinson *et al*. published their review of 10 patients undergoing the delivery of megavoltage irradiation. Nine of the patients responded, three of whom underwent surgical salvage. Of the 10 patients, seven showed no evidence of disease after surgery and/or radiation therapy, with follow-up ranging from 1 to 10 years.\[[@ref28]\] Additional evidence suggesting the value of radiation therapy in treating the ameloblastoma has been reported by Gardner.\[[@ref29]\] Radiotherapy and chemotherapy seem to have limited value in AC and although primary radiotherapy was expected to be useful in the cases with perineural or massive soft-tissue invasion and in positive surgical margins.\[[@ref13][@ref30]\] Dhir *et al*. reviewed 18 patients with maxillary AC. In 11 of 18, radiotherapy was used either as primary or secondary treatment in the case of metastasis and/or recurrence.\[[@ref30]\] Jensen *et al*. in 2011 introduced carbon ion therapy for multiple recurring AC.\[[@ref31]\] Horváth *et al*. reported a case of mandibular AC with pleura-pulmonary and bone marrow metastasis in a 8-year-old girl. They prescribed five cycles of chemotherapy with vincristine, endoxane, adriamycin, carboplatin and etoposide. It resulted in tumor shrinkage but no significant effect upon pulmonary site and the patient died after 8 months.\[[@ref32]\] Hence, the role of chemotherapy is still unpredictable. [@ref77] Survival rate depends upon local recurrence, regional or distant metastasis. Infante-Cossio *et al*. reported a maximum 5 years of follow-up in two patients with no evidence of local recurrence, regional or distant metastasis. Third patient showed local recurrence and he died after few months of it. Hence, after surgery and radiotherapy, 5 years of survival rate can be accepted as a standard if no recurrence or metastasis occurs.\[[@ref11]\] Case reports of AC affecting both the jaws, which have been published at the time of writing this article, have been listed in [Table 1](#T1){ref-type="table"}. ###### Total review of literature of AC from 1984 to 2012  CONCLUSION {#sec1-4} ========== Based on the above study although, ameloblastoma is a slowly growing, locally aggressive tumor. Any change in its growth pattern should cause suspicion of malignancy. AC is likely to metastasize. Therefore, complete history and general systemic examination are required to rule out any distant metastasis. Radiotherapy has definite role in cure of AC as in our cases; recurrence was seen in the patient who refused to undergo it. **Source of Support:** Nil **Conflict of Interest:** None declared.

The mycroarray data are under protocolo E-MEXP-3973 at <http://www.ebi.ac.uk/arrayexpress/>. The experiments were performed at JDL\'s laboratory and any further information can be obtained from Dr. Perez (<[email protected]>). Introduction {#sec001} ============ B-1 lymphocytes are part of the innate immune system and have distinct roles in inflammation, infection, and resistance to tumors. They constitute the central B-cell population in the mice peritoneal and pleural cavities but rarely occur in spleen and lymph nodes \[[@pone.0187333.ref001], [@pone.0187333.ref002]\]. These cells contribute toward the maintenance of natural IgM levels and are the major source of IL-10, which is a regulatory cytokine involved in the downregulation of immune responses. In response to the activation of the innate immunity, B-1 lymphocytes tend to increase both natural IgM and IL-10 levels, which are very important for the development of resistance to pathogens and also the modulation of several immune-mediated inflammatory responses \[[@pone.0187333.ref003], [@pone.0187333.ref004]\]. Despite an ever-growing interest in the role of B-1 cells in various immunological responses, their participation in resistance or susceptibility to tumors has been neglected. Thus, the aim of this study was to determine the involvement of B-1 cells in the behavior of B16F10 melanoma cells. Using a heterotypic co-culture system, we demonstrated that upon contact with B-1 lymphocytes, the B16F10 melanoma cells increased the activation of the ERK signaling pathway and up-regulated the expression of MMP-9, CXCR4, and MUC18, which in turn, led to increased tumor growth and metastatic spreading \[[@pone.0187333.ref005]--[@pone.0187333.ref007]\]. We also showed that the functional and phenotypic alterations that are observed in melanoma cells post contact with B-1 lymphocytes were dependent on the IL-10 production by the latter \[[@pone.0187333.ref006]\]. As B-1 lymphocytes produce and utilize IL-10 as an autocrine growth factor \[[@pone.0187333.ref004]\], we hypothesized that this cytokine plays a significant role in the interplay between B-1 and B16F10 cells. In the current work, the differential gene expression between B-1 lymphocytes isolated from the wild-type (B-1WT) or the IL-10 knockout (B-1IL-10^-/-^) mice was determined by the microarray analysis. Among seven genes that displayed different expression patterns, claudin-10, a component of tight junction strands, was chosen for functional analysis because it has been reported to be associated with lung adenocarcinoma \[[@pone.0187333.ref008]\]. To this end, we used small interfering RNA (siRNA) to silence the claudin-10 expression in B-1 lymphocytes and evaluated the effects on the metabolism of the B16F10 cells. The silencing of claudin-10 expression in B-1 lymphocytes impairs the ability of the cells to induce the ERK phosphorylation in the B16F10 cells and to increase the aggressiveness of the tumor. Thus, our results demonstrate that the axis IL-10/claudin-10 is a suitable target to control the behavior of melanoma cells. Materials and methods {#sec002} ===================== Mice {#sec003} ---- Female wild-type (WT), B cell-deficient (BKO) and IL-10 knockout (IL-10-/-) C57BL/6 mice, 6--8 weeks old, were obtained from Centro de Desenvolvimento de Modelos Experimentais para Medicina e Biologia (CEDEME), Brazil. All mice were housed in a pathogen-free facility and fed with autoclaved water and solid standard chow. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. This study was approved by the Ethics Committee of the Universidade Federal de São Paulo, Brazil (Permit Number: 0799/08). All euthanasia was performed using cervical dislocation method and all efforts were made to minimize suffering. Cell culture {#sec004} ------------ B16F10 melanoma cells were cultured in RPMI-1640 (Sigma, St Louis, MO, USA) containing 10% of fetal bovine serum (Cultilab, Campinas SP, Brazil) antibiotics and supplements. B-1 lymphocytes enriched fraction from wild-type (B-1WT) or IL-10^-/-^ (B-1IL-10^-/-^) C57BL/6 mice were obtained by culture of five days of total peritoneal cells. As previously described \[[@pone.0187333.ref006]\], free-floating cells correspond mostly to B-1 lymphocytes. The total peritoneal cells were stained with anti-CD19, and anti-CD23 antibodies (BD Biosciences, USA) and the B-1 lymphocytes were separated by electronic cell sorting of CD19+CD23- cells (FACS Aria II cell sorter, BD Biosciences, USA). Purification of B-1 lymphocytes {#sec005} ------------------------------- Purified B-1 lymphocytes ex vivo were obtained from the peritoneal cavity of WT mice by successive washed with RPMI medium. The cell suspension was labeled with anti-mouse CD19 APC and anti-mouse CD23 FITC (both from BD Biosciences, Mountain View, CA, USA) and then submitted to electronic cell sorting using a FACSAria II cell sorter (BD Biosciences). Finally, single B-1 lymphocytes (CD19+CD23-, pB-1WT) and other cells (non-B-1WT) were collected separately. Microarray sample preparation, Gene-chip hybridization and analysis {#sec006} ------------------------------------------------------------------- Trizol reagent (Invitrogen Corporation, Carlsbad, CA, USA) was used to isolate total RNA from enriched cultures of B-1 WT or IL-10-/- lymphocytes. Three biological replicates of each sample were used for the array hybridizations. The sample preparation, hybridization, and microarray data analysis on GeneChip® mouse genome 430 2.0 array (Affymetrix, Santa Clara, CA, USA) were performed and analyzed as described \[[@pone.0187333.ref009]\]. Briefly, images were first examined for visible defects using GeneChip ® Operating Software (GCOS). Signal intensities for each gene and data normalization were performed using RMA method (Robust Multi-Array Analysis) \[[@pone.0187333.ref010]\] with RMA express software \[[@pone.0187333.ref011]\]. Differentially expressed genes were identified using SAM (Significance Analysis of Microarray) with MultiExperiment Viewer (MeV), software deployed by The Institute of Genomic Research--TIGR, Rockville, MD. The criteria for selecting differentially expressed genes between B-1 lymphocytes from WT and IL-10^-/-^ mice was performed using fold change ≥2 and false selection rate (FDR--false discovery rate) ≤4%. The microarray data are available at the ArrayExpress under protocol accession E-MEXP-3973. Transfection with siRNA {#sec007} ----------------------- Three Stealth RNAi™ siRNA sequences were synthesized commercially by Invitrogen Life Technologies Inc. (USA) with the help of tools available online ([http://www.invitrogen.com](http://www.invitrogen.com/)). Stealth RNAiTM siRNA1 (RNA)-`AUCAUUAGUCCUCUACAUGCCUGGA` (cldn-10 siRNA1, primer number 148163E05), Stealth RNAiTM siRNA2 (RNA)-`UCCAGGCAUGUAGAGGACUAAUGAU` (cldn-10 siRNA2, primer number 148163E06), Stealth RNAiTM siRNA3 (RNA)-`UUUGCAUACAGGGAACAGCCUGUCA` (cldn-10 siRNA3, primer number 148163E07) were designed to target claudin-10 mRNA. The basic local alignment search tool (BLAST) was carried out to verify the target of the three sequences. All sequences have the ability to suppress any of six claudin-10 isoforms\[[@pone.0187333.ref012]\]. Stealth RNAi™ siRNA Negative Control (negative siRNA 12935--300, Ambion®/Life Technologies Corporation, CA, USA) was used to check nonspecific cellular events caused by the introduction of the oligonucleotide into cells. For transfection, B16F10 cells or B-1 lymphocytes enriched culture were collected and seeded at 0.5x10^6^ cells per well in a 12-well plate. Then, cells were transfected or not with 100 nM of claudin-10 siRNAi1, siRNAi2 or siRNAi3 or with negative control siRNA, according to the manufacturer's protocol for siPORT™ NeoFX™ Transfection Agent (InvitrogenTM, Life Technology Inc. CA, USA). The transfection efficiency of each Stealth RNAi™ siRNA sequence was confirmed by flow cytometry using Silencer® Cy™3 Labeled GAPDH (AM 4623, Ambion®/ Life Technologies Inc. CA, USA). The western blot analyses confirmed the silencing of claudin-10 expression. Co-cultures of B16F10 cells and B-1 lymphocytes {#sec008} ----------------------------------------------- As previously described 1x10^5^ B16F10 cells were co-cultured with 1x10^6^ B-1 lymphocytes in 6-well plate \[[@pone.0187333.ref006]\]. After 48 hours without changing the medium, non-adherent cells (B-1 lymphocytes) were removed, and adherent cells (B16F10 cells) were washed for three times with phosphate-buffered saline (PBS) to remove all the remaining B-1 lymphocytes. B16F10 cells were collected using trypsin (Cultilab, Brazil) for subsequent assays. To determine the purity of B16F10 cells after co-cultures, they were incubated with CD11b and IgM antibodies (BD Biosciences, USA) for flow cytometry analysis. The data acquisition was performed in a FACSCalibur using CELLQUEST software (BD Biosciences, USA), and data were analyzed using FlowJo® software (Tree Star Inc., USA). Experimental metastasis assays {#sec009} ------------------------------ B16F10 melanoma cells recovered from culture were injected into the lateral tail vein of syngeneic C57BL/6 mice (1x10^5^cells/animal). After fourteen days, mice were euthanized, and the number of lung tumor colonies was determined. Total cellular protein extracts and western blot {#sec010} ------------------------------------------------ Total protein extraction was performed as previously described \[[@pone.0187333.ref006]\]. Anti-phospho-ERK (E-4, sc-7383), total ERK (K-23, sc-94) from Santa Cruz Biotechnology (Santa Cruz, CA, USA), anti-claudin-10 (38--8400) from Invitrogen (Invitrogen Corporation, CA), anti-β actin (Sigma, St Louis, MO) or anti-GAPDH (AB9485) both from Abcam® (Abcam®, Cambridge, USA) were used as primary antibodies. Peroxidase-conjugated goat anti-mouse IgG (Sigma, St Louis, MO) and anti-rabbit (BioRad, Hercules, CA, USA) were used as secondary antibodies. The reactions were developed using a chemiluminescence ECL kit (Amersham Pharmacia, Uppsala, Sweden). Bands were visualized using UVItec (UVItec Limited, Cambridge, UK) and analyses were performed using Uviband 1D analysis software (UVItec Limited, Cambridge, UK). Statistical analysis {#sec011} -------------------- All data represent at least three independent experiments, expressed as the mean ± the standard deviation. Statistical comparisons were made by ANOVA followed by Turkey Multiple Comparison Test. Analyses were performed with GraphPad Prism 4.0 software (GraphPad Software Inc., San Diego, CA, USA). Values of p\<0.05 were considered statistically significant. Results {#sec012} ======= B-1 lymphocytes, but not the other peritoneal cell population, increase the metastatic behavior of melanoma cells in an IL-10-dependent manner {#sec013} ---------------------------------------------------------------------------------------------------------------------------------------------- To prove the non-engagement of the peritoneal cells other than the B-1 lymphocytes in the pro-metastatic effect on the B16F10 melanoma cells, we took advantage of the B cell-deficient (BKO) C57/BL6 mice. The BKO mice are deficient in both the B-1 and the B-2 lymphocytes. Briefly, total peritoneal cells were isolated from the wild-type (WT) or the BKO mice and were cultured for 40 min followed by the removal of non-adherent cells. The WT adherent peritoneal cells (WTapc) were collected and co-cultured with the B16F10 melanoma cells at 10:1 ratio for 48 h. The mice which received B16F10 co-cultivated with WTapc exhibited more metastatic nodules than the ones which were injected with B16F10 cells, either cultivated alone or with the BKO mice adherent peritoneal cells (BKOapc, [Fig 1A](#pone.0187333.g001){ref-type="fig"}). To further confirm that B-1 lymphocytes are responsible for the aggressiveness of the tumor, B16F10 melanoma cells were co-cultivated with purified peritoneal B-1 lymphocytes (CD19^+^CD23^-^; [S1 Fig](#pone.0187333.s001){ref-type="supplementary-material"}) or the non-B-1 peritoneal cells. As depicted in the [Fig 1B](#pone.0187333.g001){ref-type="fig"}, only the purified B-1 lymphocytes (B16+pB-1) enhanced the metastatic potential of the B16F10 melanoma cells. {#pone.0187333.g001} Endogenous IL-10 drives the pro-metastatic effect of B-1 lymphocytes on B16F10 melanoma cells {#sec014} --------------------------------------------------------------------------------------------- Although earlier studies have reported that the addition of exogenous IL-10 to melanoma cultures has no effect on the B16F10 behavior \[[@pone.0187333.ref006]\], we hypothesized that endogenous IL-10 could influence the metabolism of B-1 lymphocytes; thereby indirectly affecting the B16F10 metastatic potential. To test this, we checked by flow cytometry analysis phenotypic differences between B-1 lymphocytes from WT and IL-10^-/-^ mice. [Fig 2A](#pone.0187333.g002){ref-type="fig"} illustrates that there was no significant difference, neither in the phenotype nor in the total number of cells that were recovered after the culture of peritoneal cells from the WT or the IL-10^-/-^ mice. In contrast, the animals that were inoculated with the B16F10 cells co-cultured with B-1WT lymphocytes presented a higher number of lung melanoma colonies in comparison to the mice that received B16F10 cells cultivated alone or co-cultured with the B-1IL-10^-/-^ lymphocytes ([Fig 2B](#pone.0187333.g002){ref-type="fig"}). Consequently, such data supported the notion that the influence of B-1 lymphocytes in the behavior of melanoma cells depends upon the endogenous IL-10 levels. {#pone.0187333.g002} IL-10 deficiency impairs the expression of claudin-10 in B-1 lymphocytes {#sec015} ------------------------------------------------------------------------ Since the flow cytometry analysis did not reveal any phenotypic differences between the B-1WT and the B-1IL-10^-/-^ cells ([Fig 2A](#pone.0187333.g002){ref-type="fig"}), we used the high throughput microarray technique to obtain the gene expression profiles of these cells. The comparison analyses were done using the MEV software, and the statistically significant transcripts were selected by the unpaired two-class SAM analysis (FDR and Q-value 4%) by applying 500 random permutations. Using these parameters, the chip analysis revealed the variations in the expression of eleven (11) genes between the WT and the IL10^-/-^ B-1 lymphocytes ([Fig 3A](#pone.0187333.g003){ref-type="fig"}). The analysis of the regulated gene clusters by the DAVID Functional Annotation Chart tool displayed relevant differences between the WT and the IL-10^-/-^ B-1 lymphocytes in a group of five genes: three up-regulated (cldn-10, mid-1, ldlrap1) and two down-regulated (cnr2 and nnt). Among these five (5) genes, we validated the expression of claudin-10 (cldn-10) transcript because the claudin family of proteins has been associated, either positively or negatively, with the progression of various tumors \[[@pone.0187333.ref013]--[@pone.0187333.ref016]\]. In accordance with the microarray results, the western blot analysis also showed the down-regulation of claudin-10 in the B-1IL-10^-/-^ lymphocytes as compared to the B-1WT cells ([Fig 3B](#pone.0187333.g003){ref-type="fig"}). {#pone.0187333.g003} Claudin-10 mediates the interplay between B-1 lymphocytes and melanoma cells {#sec016} ---------------------------------------------------------------------------- In an attempt to verify the role of claudin-10 gene in our model, we used the Stealth RNAi^TM^ siRNA technology. The B-1WT lymphocytes were transfected with claudin-10 siRNA1, siRNA2 or siRNA3 sequences and subjected to western blotting to verify the abrogation of transcription of the claudin-10 gene. Among the three siRNAs, only the siRNA1 exhibited significant inhibition of the claudin-10 gene expression in B-1 cells ([Fig 4A](#pone.0187333.g004){ref-type="fig"}). To determine the role of claudin-10 in the pro-metastatic effect of B-1 lymphocytes, the B16F10 cells were co-cultured for 48 h with B-1 lymphocytes, untransfected or transfected, with claudin-10 siRNAs followed by experimental metastasis assay. The inhibition of claudin-10 expression by siRNA1 impaired the B-1 lymphocyte mediated increase of the metastatic potential of B16F10 cells ([Fig 4B](#pone.0187333.g004){ref-type="fig"}). In contrast, the mice, which were inoculated with the B16F10 cells co-cultured with the siRNA2 or siRNA3 transfected or untransfected B1 cells, demonstrated the highest number of lung melanoma metastases ([Fig 4B](#pone.0187333.g004){ref-type="fig"}). In addition, the suppression of the claudin-10 expression in B-1 lymphocytes inhibited the ERK pathway activation in B16F10 cells post contact with B-1 lymphocytes ([Fig 4C](#pone.0187333.g004){ref-type="fig"}). Similarly, increase in claudin-10 expression in the B16F10 cells upon contact with B-1 lymphocytes was also inhibited, indicating that the ERK pathway is involved in the expression of claudin-10 in the melanoma cells. Of note, B-1 lymphocytes that were co-transfected with the negative control siRNA displayed the expression of claudin-10 gene and the pro-metastatic influence on the B16F10 cells ([S2 Fig](#pone.0187333.s002){ref-type="supplementary-material"}). This finding signified that only the claudin-10 gene silencing specifically led to the observed phenomenon. {#pone.0187333.g004} Finally, to further confirm the role of claudin-10 in bringing out the changes that are observed in the melanoma cells after contact with B-1 lymphocytes, the claudin-10 expression was inhibited in the B16F10 cells. As hypothesized, in the absence of claudin-10, B-1 lymphocytes were unable to modulate the metastatic behavior of the B16F10 melanoma cells ([Fig 5](#pone.0187333.g005){ref-type="fig"}). Thus, our findings demonstrate that claudin-10 is a key mediator in the interplay between B-1 lymphocytes and B16F10 melanoma cells and is involved in the enhancement of melanoma aggressiveness upon contact with the B-1 cells. {#pone.0187333.g005} Discussion {#sec017} ========== Like the other cells of the immune system, B-1 lymphocytes can be associated with protective or deleterious responses depending on the context in which they are activated \[[@pone.0187333.ref003], [@pone.0187333.ref004], [@pone.0187333.ref017]--[@pone.0187333.ref019]\]. Concerning the involvement of B-1 lymphocytes in tumor behavior, we found that analogous to tumor-associated macrophages (TAMs), the B-1 lymphocytes seem to promote growth and spread of melanoma cells \[[@pone.0187333.ref005], [@pone.0187333.ref006], [@pone.0187333.ref020], [@pone.0187333.ref021]\]. Previously, we reported that the pro-metastatic effect of B-1 lymphocytes on the B16F10 melanoma cells was dependent upon the cell-to-cell contact and the ERK signaling, however, the mechanisms involved in this phenomenon remained unclear \[[@pone.0187333.ref005], [@pone.0187333.ref006], [@pone.0187333.ref009]\]. Since IL-10 is known to play an important role in B-1 biology, we hypothesized that the cytokine could be involved in the behavior alteration that is observed in the case of the B16F10 melanoma cells. In the present study, we demonstrated that the peritoneal cells other than the B-1 lymphocytes were not able to boost the melanoma aggressiveness. Although a previous study demonstrated that IL-10, itself, was unable to modulate the B16F10 behavior, its influence on the cross-talk between the B-1 lymphocytes and the melanoma cells was not clear \[[@pone.0187333.ref006]\]. Thus, we co-cultured the B16F10 cells with B-1 lymphocytes that were isolated from IL-10^-/-^ mice. The IL-10-deficient B-1 lymphocytes were unable to induce alterations in the metastatic melanoma behavior, indicating that the IL-10 production by B-1 lymphocytes plays a pivotal role in the interplay between these cells. Furthermore, in an attempt to investigate the influence of IL-10 in B-1 cell biology, we evaluated the effect of IL-10 in gene expression profile of the B-1 lymphocytes. The microarray analysis revealed differential expression of eleven (11) transcripts among the B-1WT and B-1IL-10^-/-^ lymphocytes, which was in concordance with the hypothesis that the endogenous IL-10 plays crucial role in the B-1 physiology. Among the 11 transcripts, claudin-10 was downregulated in the IL-10^-/-^ B-1 lymphocytes and seemed to be the most important mediator in the BF16/B-1 interaction. Firstly, the claudin family of proteins presents the major integral proteins that constitute the backbone of tight junctions, being involved in cell-to-cell contact \[[@pone.0187333.ref022]\] and; secondly, the claudin gene expression may be regulated by the ERK signaling pathway \[[@pone.0187333.ref006], [@pone.0187333.ref023]\], which is up-regulated in both the B16F10 cells and the B-1 lymphocytes after contact with each other \[[@pone.0187333.ref006], [@pone.0187333.ref023]\]. To determine the influence of claudin-10 in the interaction between the B-1-lymphocytes and the melanoma cells, claudin-10 expression was inhibited using RNAi approach. The suppression of claudin-10 levels in the B-1 lymphocytes impaired the ability of the latter to increase the aggressiveness of melanoma cells, which was accomplished by the down-modulation of the ERK signaling pathway ([Fig 4C](#pone.0187333.g004){ref-type="fig"}). Thus data from the present work corroborated the previous findings that established an association between the ERK signaling and the tumor aggressiveness \[[@pone.0187333.ref006]\] and, supporting the notion that claudin-10 is involved in the changes in melanoma cells upon contact with B-1 lymphocytes. In addition, similar results were recorded when claudin-10 was inhibited in the B16F10 cells, validating the role of claudin-10 as a mediator to increase the metastatic potential of B16F10 melanoma after contact with B-1 lymphocytes. Together, these data support the hypothesis that homotypic interactions mediated by claudin-10 are required to trigger the more metastatic behavior on B16F10 melanoma cells after contact with B-1 lymphocytes. Conclusions {#sec018} =========== In summary, our findings demonstrate that endogenous IL-10 influences the expression of claudin-10 in B-1 lymphocytes and implicate the axis IL-10/claudin-10 to augment the aggressive behavior of the B16F10 melanoma cells upon contact with B-1 lymphocytes. Thus, the identification of claudin-10 as a mediator of the pro-metastatic effect of B-1 lymphocytes on melanoma cells reveals a potential target for the development of molecular therapeutics to control the melanoma metastasis. Supporting information {#sec019} ====================== ###### Gate strategy for B-1 lymphocyte purification by flow cytometry after cell culture. The gate strategy resulted in single cells with purity of \> 99.5% after electronic cell sorting. (TIFF) ###### Click here for additional data file. ###### The siRNA methodology has no effect on B1-lymphocyte ability to promote B16F10 aggressiveness. B-1 lymphocytes were transfected for 24 h with claudin-10 siRNA or negative siRNA to determine the impact of stealth methodology in their activity on B16F10 cells. A) The inhibition of claudin-10 expression depends on the specific siRNA used. B) The negative stealth siRNAi has no effect on B-1 lymphocytes\' ability to promote further the metastatic behavior in B16F10 melanoma cells. Data are the mean ± SD of three independent experiments. \*\*\*p \< 0.0001, using one-way ANOVA with Tukey's post hoc test. (TIFF) ###### Click here for additional data file. Thanks to Dr. Renata Pellegrino, Dr. Viviane Bernardo, Daniela Teixeira and Geova Pereira for technical support. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: † Deceased.

{#sp1 .470}

Introduction {#S0001} ============ Le syndrome de résistance complète aux androgènes (SICA) ou anciennement appelé testicule féminisant (TF) est une affection rare qui correspond à la forme complète des pseudohermaphrodismes androgynoïdes. C\'est une maladie génétiquement déterminée récessive liée à l\'X, en rapport avec des mutations au niveau de Xq11-q12 des gènes du récepteur des androgènes \[[@CIT0001]\]. Elle est caractérisée par un trouble de la réceptivité périphérique aux androgènes secrétés par les testicules. Ces sujets ont un sexe génétique et gonadique masculin, mais phénotypique féminin \[[@CIT0002]\]. Ce syndrome est souvent diagnostiqué au cours de la période pubertaire quand la patiente consulte pour aménorrhée primaire. Le SICA est accompagné d\'un développement anormal des testicules et d\'un risque élevé des tumeurs malignes des cellules germinales débutant après la puberté. Les tumeurs à cellules de Sertoli-Leydig sont des tumeurs gonadiques très peu répandues, mais elles sont bénignes dans la majorité des cas \[[@CIT0001]\]. Les auteurs rapportent deux observations de deux jeunes filles présentant le SICA illustrant les particularités cliniques, anatomopathologiques et biologiques du syndrome avec certaines particularités. Patient et observation {#S0002} ====================== Observation 1 {#S20003} ------------- O.K. est une jeune fille de 17 ans, issue d\'un mariage non consanguin, Troisième d\'une fratrie de quatre. Aucune de ses trois sœurs n\'a présenté une aménorrhée primaire. Elle a bénéficié à l'âge de 5 ans d\'une biopsie d\'une tuméfaction inguinale droite dont l\'examen histologique a objectivé un testicule immature. Le diagnostic d\'un testicule féminisant a été posé et le médecin traitant a expliqué à la famille la nécessité de suivi dans un centre spécialisé mais la famille a démentis les propos énoncés par l'équipe médicale. Vers l'âge de 12 ans, un développement normal des seins est apparu avec un morphotype féminin harmonieux. Une aménorrhée primaire a été le motif de sa consultation à l\'unité de gynécologie-obstétrique. L\'examen clinique constate une patiente en bon état général, de phénotype parfaitement féminin qui pèse 50 kg pour 1,74 m, soit un indice de masse corporelle à 17,30 kg/ m^2^. Le développement mammaire normal (stade 3 de Tanner) ([Figure 1](#F0001){ref-type="fig"}), une pilosité pubienne P1 faite d\'un fin duvet et une pilosité axillaire absente A0. Il n\'y avait pas de masses palpables au niveau inguinal. Les organes génitaux externes sont typiquement féminins avec présence des grandes et petites lèvres, pas de clitoromégalie, pas de fusion postérieure des grandes lèvres ni de gonades palpables au niveau des lèvres. On avait bien identifié deux orifices distincts: un orifice urétral et un autre vaginal. L\'examen du vagin a montré un vagin perméable, de 8 cm de profondeur, d\'aspect normal à la vaginoscopie, mais borgne. Au toucher rectal, l\'utérus n\'est pas palpable. Une échographie et une IRM abdomino-pelviennes ont montré l\'absence de l\'utérus et des ovaires. En plus, l\'IRM a objectivé la présence de deux formations oblongues bilatérales de 44 mm de grand axe, tissulaires en discret hypersignal T2 de siège rétro-péritonéales siégeant en regard des vaisseaux iliaques externes, bilatérales et associées à des formations kystiques polaires inferieures de 17 mm de grand axe ([Figure 2](#F0002){ref-type="fig"}). {#F0001} {#F0002} Ces formations tissulaires arrivent au contact du pole supérieur de la vessie sans anomalie de cette dernière avec un canal vaginal en place. A la biologie elle présentait un taux de testostérone élevé 13,68 ng/ml dans les normes masculines, un taux de FSH bas à 1,2 UI/ l, un taus élevé de LH à 21 UI/l et un 17 bêta-œstradiol élevé à 58,6 pg/ml. Les marqueurs tumoraux (la βHCG était élevé à 114,5 mUI/ml, l\'AFP et la LDH étaient normaux). Le caryotype a révélé une formule chromosomique 46, XY. Devant les antécédents de la patiente ainsi que ces constatations cliniques, biologiques et radiologiques, Le diagnostic du syndrome de résistance complète aux androgènes a été retenu. La patiente a bénéficie ainsi d\'une orchidectomie bilatérale par laparotomie qui a permis la résection des deux testicules situés en rétro-péritonéal en regard des vaisseaux iliaques externes ([Figure 3](#F0003){ref-type="fig"}). A l\'examen macroscopique, la gonade droite mesure 6×4×3 cm et la gonade gauche 7×4×3 cm renfermant sur leurs pôles inférieurs des kystes séreux. L\'examen histologique a objectivé un parenchyme testiculaire immature fait de tubes séminifères souvent atrophiques d\'aspect sertolien bordés par des cellules de Sertoli et parfois avec des très rares cellules de la lignée germinale sans spermatides ni spermatozoïdes visibles. Les tubes sont entourés par un tissu interstitiel tantôt fibreux tantôt lâche avec quelques cellules de Leydig par endroits et des remaniements kystiques sans signes histologiques de malignité ([Figure 4](#F0004){ref-type="fig"}). Un traitement à base d'œstro-progetaifs a été instauré en post opératoire. Après 6 mois de traitement hormonal substitutif, la patiente a maintenu un bon développement de ses caractères sexuels secondaires (développement mammaire stade 3 de Tanner). {#F0003} {#F0004} Observation 2 {#S20004} ------------- H.H est une jeune fille de 15 ans, fille unique, issue d\'un mariage non consanguin. Elle a consulté à l'âge de 4 ans pour hernie inguinale bilatérale. Vu l\'indisponibilité de l' IRM à ce moment, une cœlioscopie a été réalisée qui a objectivé une migration incomplète des deux testicules au niveau de l\'orifice profond de chaque canal inguinal avec absence de visualisation de l\'utérus et des ovaires. Un caryotype réalisé a révélé une formule chromosomique 46 XY. La patiente n´a pas été traitée et il a été décidé d\'attendre la puberté pour programmer une gonadectomie. La puberté est survenue chez cette patiente à l'âge de 13 ans avec un développement mammaire normal et un morphotype féminin harmonieux malgré l\'aménorrhée. Elle pèse 70 kg et mesure 1,68 m soit un indice de masse corporelle à 24,80 kg /m^2^. La patiente a été admise à l\'unité de gynécologie obstétrique pour une gonadectomie en période post pubertaire. L\'examen physique retrouve un développement mammaire normal (stade 3 de Tanner) avec une aréole pale, une pilosité pubienne P1 faite d\'un fin duvet et une pilosité axillaire absente A0. Les organes génitaux externes sont de type féminin sans ambiguïté: les grandes et petites lèvres sont présentes, sans clitoromégalie, les orifices urétral et vaginal sont distincts et un vagin perméable de 4 cm mais borgne. L\'examen abdominal retrouve deux tuméfactions inguinales bilatérales. Une échographie pelvienne ainsi qu\'une IRM pelvienne réalisées retrouvent l\'absence d\'organes génitaux internes (utérus et ovaires) et la présence de formations inguinales bilatérales faisant 46×14 mm à droite et 46×28 mm à gauche évoquant des testicules avec présence d\'un nodule faisant 15×10mm au niveau du pôle supérieur du testicule gauche bien limité. Le bilan biologique retrouve un taux élevé de testostérone à 2,34 ng/ml mais qui était dans les normes masculines, un taux de FSH et de LH normaux respectivement de 4,3 mUI/ml et 4,72 mUI/ml, et un taux d'œstradiol à 26 pg/ml; les marqueurs tumoraux à savoir l\'AFP, la LDH et βHCG étaient normaux. La patiente a bénéficié d\'une orchidectomie bilatérale. A l\'examen macroscopique, la gonade droite mesure 4×2×2 cm et la gonade gauche 4×3×2 cm renfermant sur la périphérie de son pole supérieur un nodule mesurant 1,3×1,3×1 cm ([Figure 5](#F0005){ref-type="fig"}). A l\'examen histologique, le nodule bien limité, circonscrit par une fine capsule est composé de tubes sertoliens atrophiques avec un tissu interstitiel très réduit comportant de rares cellules de Leydig. Ce nodule, correspond à une tumeur bien différenciée à cellules de Sertoli-Leydig. A l\'immunohistochimie, ces cellules expriment l\'Inhibine et le Melan A ([Figure 6](#F0006){ref-type="fig"}). Le reste du parenchyme testiculaire atrophique est composé de cordons pleins remplis de cellules de Sertoli, sans spermatogonie. Le tissu interstitiel renferme des cellules de Leydig hyperplasiques associé à une ébauche de la trompe de Fallope sur la périphérie de la gonade gauche. Un traitement à base d'œstradiol a été instauré en post opératoire. Après 2 ans de traitement hormonal substitutif, la patiente a maintenu un bon développement de ses caractères sexuels secondaires (développement mammaire stade 3 de Tanner). {#F0005} {#F0006} Discussion {#S0005} ========== Le syndrome d\'insensibilité complète aux androgènes est une entité rare; son incidence est en fait très variable, allant, selon les auteurs de 1/20000 à 1/60000 naissances \[[@CIT0003]\]. Il est caractérisé par un déficit complet de l\'action des androgènes au niveau des organes cibles lié à un dysfonctionnement du récepteur des androgènes en rapport avec une absence, ou un déficit, fonctionnel du récepteur cellulaire des androgènes, à un défaut qualitatif du récepteur, ou éventuellement, à un défaut touchant soit la transcription du gène, soit la liaison nucléaire du complexe stéroïde récepteur \[[@CIT0004]\]. La conséquence est un morphotype typiquement féminin chez des sujets de caryotype masculin 46 XY. C´est une maladie héréditaire récessive liée au chromosome X, donc transmise par les femmes à une partie de leur descendance mâle. Cependant dans presque la moitié des cas, il s\'agit d\'une mutation de novo vu l\'absence d\'antécédents familiaux \[[@CIT0005]\]. La recherche de la mutation causale s\'effectue par séquençage du gène du récepteur des androgènes (RA). Les mutations au niveau du gène du RA sont très nombreuses, la substitution est l\'anomalie la plus fréquente. Plus de 350 mutations responsables de syndrome de résistance aux androgènes sont répertoriées. Dans la majorité des cas, il s\'agit d\'une simple substitution d\'acides aminés \[[@CIT0005]\]. Au niveau de l\'exon 1, Gottlieb et al. ont rapporté 65 mutations \[[@CIT0006]\]. Concernant les mutations responsables d\'un codon stop au niveau de l\'exon 1, elles sont retrouvées dans 18 cas de formes complètes, dont 17 à type de substitutions et un seul cas de délétion. Bel Hadj et al. ont décrit une délétion de quatre bases de 1029 à 1032 de l\'exon 1 du récepteur aux androgènes, occasionnant un codon stop en 476, et conduisant à une protéine tronquée non fonctionnelle \[[@CIT0005]\]. Rarement, le diagnostic est établi en pré-pubertaire, à l´occasion de l´exploration d´une hernie inguinale comme c\'est le cas de nos deux patientes. Dans la plupart des cas, les sujets consultent à l´adolescence ou après la puberté pour une aménorrhée primaire, voire une stérilité ou un syndrome tumoral à l'âge adulte. L\'examen physique retrouve souvent un développement morphologique féminin et harmonieux; une disposition gynoïde du tissu adipeux et une voix féminine; les seins sont bien développés et les cheveux sont normalement implantés, à l\'opposé d\'une absence de pilosité: les aisselles sont parfaitement glabres et le pubis ne présente aucun poil noir et dur. Ce signe doit attirer l\'attention et faire évoquer le diagnostic. Cependant, un fin duvet clairsemé est possible et ne doit pas faire éliminer le diagnostic \[[@CIT0007]\]. Il est classique de dire que les sujets atteints de ce syndrome se présentent comme de jolies femmes plus grandes que la moyenne féminine comme ce fut le cas de nos deux patientes et en général minces (le cas de notre première patiente) \[[@CIT0005]\]. Les organes génitaux externes sont féminins avec un petit clitoris et des grandes et des petites lèvres bien développées, le vagin est petit et perméable se terminant en cul-de-sac. Les organes génitaux internes, dérivés des canaux de Wolff, sont souvent absents, il n\'y a ni utérus, ni trompes, ni pavillon, ni prostate, mais la présence de résidus müllériens ne doit pas faire éliminer le diagnostic. La gonade est un testicule comportant un épididyme et un canal déférent; elle peut siéger en position inguinale, labiale ou abdominale \[[@CIT0008]\]. Les résultats des dosages hormonaux doivent être interprétés en fonction des valeurs de références masculines et selon l'âge du patient. En période post pubertaire, le SICA est classiquement caractérisée par des taux de testostérone dans la zone masculine normale ou élevés vu la réponse exagérée à la stimulation par LH et des taux de LH sériques normaux ou élevés du fait de l\'insensibilité hypothalamo-hypophysaire à la testostérone et des taux de FSH normaux \[[@CIT0005]\]. L'échographie identifie aisément les testicules ectopiques lorsqu\'ils sont dans les canaux inguinaux ou dans la région des grandes lèvres; sa performance est moindre lorsqu\'ils sont en position abdomino-pelvienne. L\'IRM pelvienne localise facilement les testicules ectopiques, qu\'ils soient en intra-abdominal ou dans les canaux inguinaux ou les grandes lèvres; elle permet également de mettre en évidence de potentielles lésions focales en rapport avec une tumeur ou une inflammation et une éventuelle extension ganglionnaire \[[@CIT0009]\]. La cœlioscopie comme ce fut le cas de l\'une de nos patientes confirme l\'absence d\'utérus et d\'ovaires. Le caryotype est masculin 46 XY. Histologiquement, le testicule du syndrome d\'ISCA est formé de tubes séminifères immatures contenant de rares cellules germinales et bordés uniquement de cellules de Sertoli. Le nombre de cellules germinales peut être normal chez les enfants de moins de cinq ans et tend à diminuer voire s´annuler avec l´âge. Dans le tissu interstitiel, il existe une hyperplasie des cellules de Leydig associée dans 70% des cas \[[@CIT0005]\]. Des ébauches de trompe de Fallope dérivant de l´extrémité des canaux de Müller sont présentes dans un tiers des cas en périphérie des testicules dû à un retard de synthèse de l\'hormone anti-müllérienne par les cellules de Sertoli ou à son inactivation comme c'était le cas pour notre deuxième patiente. Le principal risque est la cancérisation des testicules ectopiques. Ce risque est estimé à 5 - 10%, il est rare jusqu'à l'âge de 25 ans, puis augmente avec l'âge pour atteindre 33% après 50 ans \[[@CIT0009]\]. Comme c´est le cas pour la cryptorchidie, la plupart des néoplasies chez les patientes atteintes de syndrome d\'ISCA proviennent des cellules germinales \[[@CIT0009]\]; en particulier le carcinome non invasif in situ, gonadoblastome et leurs homologues invasifs seminome/dysgerminome ainsi que les tumeurs non seminomateuses. Les tumeurs bénignes provenant des cellules non-germinales englobent les adénomes Sertoliens et les hamartomes \[[@CIT0010]\]. La tumeur à cellules de Sertoli-Leydig est une tumeur rare du cordon sexuel et de stroma qui représente moins de 1% de l\'ensemble des tumeurs ovariennes, survenant souvent à un âge jeune. Elle est également rarement développée chez le patient avec syndrome d\'ISCA, il n\'existe que deux cas rapportés dans la littérature \[[@CIT0001]\]. Les tumeurs bénignes, bien différenciées représentent 10% de l\'ensemble des tumeurs à cellules de Sertoli-Leydig et sont souvent associées au syndrome de testicule féminisant \[[@CIT0011]\], c\'est le cas de notre deuxième observation. Vu le risque de transformation maligne, l\'orchidectomie bilatérale préventive doit être systématique par la laparotomie ou de préférence par coelioscopie lorsque les gonades sont de localisation intra-abdominale \[[@CIT0012]\]. Elle est recommandée en période poste pubertaire afin de permettre un développement normal des organes génitaux externes et des seins durant la puberté quand la transformation maligne des cellules germinales est relativement rare et tardive \[[@CIT0001]\]. L\'orchidectomie peut également être pratiquée dans l\'enfance quand le diagnostic est posé précocement lors d\'une cure d\'hernie inguinale. Cela, après une discussion des modalités thérapeutiques avec les parents qui peuvent choisir soit une gonadectomie avec induction de la puberté par des estrogènes ou de différer le geste en post pubertaire \[[@CIT0010]\]. Pour nos deux patientes, la castration a été réalisée en post-pubertaire. La prise en charge du syndrome d\'ISCA nécessite après la castration, l\'instauration d\'un traitement substitutif à base d'œstrogène, visant à prévenir les conséquences d\'une ménopause par privation ostrogénique et à éviter la régression des caractères sexuels secondaires. Cette oestrogénothérapie peut être associée à un progestatif intermittent pour diminuer le risque de carcinogenèse mammaire \[[@CIT0013]\]. Un soutien psychologique pourrait être également nécessaire aussi bien pour la patiente que pour sa famille surtout à l\'annonce de la stérilité en évitant de révéler le sexe génétique. Conclusion {#S0006} ========== Le syndrome de résistance complète aux androgènes reste une entité rare. L\'insensibilité des organes cibles aux androgènes est à l\'origine du syndrome. Les tumeurs bénignes ont été décrites dans 80% des cas; elles sont représentées essentiellement par les adénomes sertoliens. Le risque de dégénérescence est assez faible mais augmente avec l'âge. Le traitement essentiel reste la castration pour éviter la dégénérescence gonadique et ce de préférence dès l'établissement du diagnostic en instaurant en pré-pubertaire une œstrogénothérapie substitutive pour assurer un développement harmonieux des caractères sexuels secondaires. Conflits d\'intérêts {#S0007} ==================== Les auteurs ne déclarent aucun conflit d\'intérêt. Contributions des auteurs {#S0008} ========================= Tous les auteurs ont contribué à la conduite de ce travail. Tous les auteurs déclarent également avoir lu et approuvé la version finale du manuscrit.

1. Introduction {#sec1} =============== Misregulation of plasma triglyceride metabolism and fatty acid delivery has been implicated in several metabolic diseases, including metabolic syndrome, diabetes mellitus, and atherosclerosis [@bib1], [@bib2]. The enzyme lipoprotein lipase (LPL) is positioned at the nexus of plasma triglyceride delivery, hydrolyzing plasma lipoprotein triglycerides and releasing fatty acids for uptake into heart, skeletal muscle, and adipose tissue. LPL-mediated hydrolysis normally occurs on the luminal surface of capillary endothelial cells where LPL is anchored by GPIHBP1, an endothelial cell GPI-anchored protein responsible for transporting LPL across endothelial cells [@bib3], [@bib4]. Angiopoietin-like 4 (ANGPTL4), also known as fasting-inducible adipose factor (FIAF), is a fasting induced inhibitor of LPL and a regulator of triglyceride metabolism [@bib5], [@bib6], [@bib7]. ANGPTL4 is most highly expressed in adipose tissue and liver, but it is also expressed at lower levels in muscle, heart, kidney, and intestine [@bib5], [@bib8], and circulates in plasma [@bib5]. Plasma triglycerides are elevated in mice overexpressing ANGPTL4, whereas *Angptl4*^*−/−*^ mice display reduced plasma triglyceride levels [@bib7]. ANGPTL4 inactivates LPL by accelerating the dissociation of active LPL dimers to inactive monomers [@bib9]. ANGPTL4 expression increases markedly upon fasting leading to the hypothesis that ANGPTL4 is involved in regulating fatty acid delivery in the fasted state [@bib5]. In humans, inactivating mutations in ANGPTL4 are associated with lower plasma triglycerides [@bib10], [@bib11], [@bib12] and lower incidence of coronary artery disease [@bib11], [@bib12], [@bib13]. Thus, it has been proposed that targeting ANGPTL4 activity may be a useful way to therapeutically increase LPL-driven triglyceride clearance, lower plasma triglycerides, and lower the risk of coronary disease [@bib11], [@bib12], [@bib14], [@bib15]. However, as LPL-driven ectopic lipid deposition can potentially lead to detrimental effects, including skeletal muscle insulin resistance [@bib16], cardiac lipotoxicity [@bib17], and severe inflammatory responses [@bib18], understanding where and when ANGPTL4 normally acts is essential. In this study, we investigate the physiological mechanisms by which ANGPTL4 regulates plasma triglycerides. We analyze the tissue-specific kinetics of fasting-induced *Angptl4* gene expression, and examine the effects of ANGPTL4 deficiency on tissue-specific lipase activity and triglyceride delivery. 2. Materials and methods {#sec2} ======================== 2.1. Mice {#sec2.1} --------- All animal procedures were approved by the Institutional Animal Care and Use Committee at the University of Iowa and were carried out in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Mice were group housed in a controlled environment with a 12/12 light/dark cycle, with food and water provided *ad libitum* during non-fasting conditions. During fasting conditions, water was provided *ad libitum*. C57BL/6J mice were obtained from Jackson Laboratories. *Gpihbp1*^*−/−*^ mice (B6;129S5-*Gpihbp1*^*tm1Lex*^/Mmucd) [@bib19], [@bib20], and *Angptl4*^*−/−*^ mice (B6;129S5-*Angptl4*^*Gt(OST352973)Lex*^/Mmucd) [@bib19], [@bib21] were generated by breeding from strains obtained from the Mutant Mouse Resource and Research Center ([mmrrc.org](http://mmrrc.org){#intref0010}). Age-matched wild-type littermates were used as controls for *Angptl4*^*−/−*^ mice. 2.2. Plasma triglyceride assays {#sec2.2} ------------------------------- *Angptl4*^*−/−*^ mice and wild-type littermates were fasted for 4 h (fasted group) or fasted for 6 h and then fed normal chow for 2 h (refed group). Blood was collected via tail-nick into EDTA-coated collection tubes. After centrifugation at 1500×*g* for 15 min at 4 °C to pellet the cells, plasma from each mouse was combined with Infinity™ Triglyceride Reagent (Thermo Scientific TR22421) according to the manufacturer\'s instructions. Samples were incubated at 37 °C for 10 min and absorbance was measured at 500 nm. Triglyceride concentrations were determined by comparison to a standard curve prepared from a triolein standard (Nu-Chek Prep, Lot T-235-N13-Y). 2.3. RNA isolation and qPCR analysis {#sec2.3} ------------------------------------ Mouse tissue samples were frozen in liquid nitrogen and pulverized using a Bessman tissue pulverizer. Crushed tissue was resuspended in Trizol (Ambion, 15596-018) and processed according to the manufacturer\'s instructions. After assessing mRNA concentration and quality using a Nanodrop spectrophotometer (ThermoScientific), cDNA was prepared using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems, Part No. 4368813). qPCR was performed (Invitrogen, SYBR GreenER qPCR Supermix, 11762100) according to the manufacturer\'s specifications using an Applied Biosystems 7900HT Fast Real-Time PCR System (Iowa Institute of Human Genetics). Relative expression was calculated with the ΔΔct method [@bib22] using *CycloA* as the reference gene. 2.4. Lipase activity assay {#sec2.4} -------------------------- Frozen tissue samples were crushed and resuspended in LPL assay buffer (25 mM NH~4~Cl, 5 mM EDTA, 0.01% SDS, 45 U/mL heparin, 0.05% 3-(N,N-Dimethylmyristylammonio) propanesulfonate zwittergent detergent (Acros Organics, 427740050)) containing Mammalian ProteaseArrest (GBiosciences, cat no. 786-331). The tissue suspension was mixed by vortexing and incubated on ice for 30 min, with intermittent disruption with surgical scissors. The resulting lysate was centrifuged at 15,000×*g* for 15 min at 4 °C to pellet cellular debris. Lipase activity assays were performed on the supernatants as previously described [@bib23]; supernatants were combined with a working buffer composed of 0.6 M NaCl, 80 mM Tris--HCl pH 8, 6% fatty-acid free BSA and an EnzChek lipase fluorescent substrate (Molecular Probes, E33955). Fluorescence was measured over 30 min at 37 °C on a SpectraMax i3 plate reader (Molecular Devices). Relative lipase activity was calculated by subtracting background (calculated by reading fluorescence of a sample with no LPL) and then calculating the slope of the curve between the 5 and 13 min reads. The data were graphed as the average of slopes for each group. 2.5. Preparation of ^3^H-Labeled intralipid {#sec2.5} ------------------------------------------- ^3^H-Intralipid was prepared by mixing \[9,10-3H(N)\]-Triolein (Perkin Elmer, NET431001MC) with 5% Intralipid (prepared fresh from Intralipid^®^ 20% (NDC 0338-0519-03)) in a ratio of 1 μCi triolein to 10 μL of Intralipid. The mixture was sonicated briefly on low power 3 × 20 s, with 1 min on ice between each round of sonication. The mixture was centrifuged briefly and diluted 10-fold in saline to prepare a 0.5% stock. 2.6. Preparation of ^3^H-Labeled chylomicrons {#sec2.6} --------------------------------------------- *Gpihbp1*^−/−^ mice were fasted 4 h and then gavaged with 100 μCi of \[9,10-3H(N)\]-Triolein (Perkin Elmer, NET431001MC) suspended in olive oil. After 4 h, mice were anesthetized, and blood was collected by cardiac puncture. Blood was diluted 1:10 with 0.5 M EDTA (pH 8.0) and centrifuged 1500×*g* for 15 min at 4 °C to pellet blood cells. The plasma was then transferred to ultracentrifuge tubes and mixed 1:1 with PBS. After centrifugation at 424,000×*g* for 2 h at 10 °C, the chylomicrons form an upper layer. The chylomicron layer was resuspended in fresh PBS and the centrifugation was repeated. Following the second centrifugation, the chylomicron layer was resuspended in PBS to the original plasma volume. Protein content was assayed using the BioRAD DC Protein Assay (BioRAD, 5000116). Radioactivity was determined in BioSafe II scintillation fluid (RPI, 111195) on a Beckman-Coulter Liquid Scintillation Counter (BCLSC6500). 2.7. Triglyceride clearance assay {#sec2.7} --------------------------------- Wild-type and *Angptl4*^*−/−*^ mice were fasted 4 h (fasted group) or were fasted for 6 h and then returned to chow for 2 h (refed group). Mice were anesthetized with isoflurane and injected retro-orbitally with 200 μL 0.5% ^3^H-Intralipid (see Section [2.5](#sec2.5){ref-type="sec"}) or 100 μL of ^3^H-chylomicron suspension (see Section [2.6](#sec2.6){ref-type="sec"}). Proparacaine hydrochloride ophthalmic solution, USP 0.5% (AKORN, 17478-263-12) was used to minimize discomfort both during and after injection. Blood samples were taken via tail-nick at 1, 5, 10, and 15 min after injection. Blood samples were assayed in BioSafe II scintillation fluid on a Beckman-Coulter Scintillation Counter. After the last blood draw, the mice were anesthetized with isoflurane, and tissues were harvested and weighed. A portion of each tissue was then weighed and placed in 2:1 chloroform:methanol overnight at 4 °C. 1 mL of 2 M CaCl~2~ was then added to each sample to separate organic and aqueous layers. The samples were centrifuged for 10 min at 1500 rpm, and the upper aqueous layer was mixed with BioSafe II scintillation fluid and assayed on a Beckman-Coulter Scintillation Counter. The lower organic layer was evaporated overnight to remove chloroform, and the remaining sample was resuspended in scintillation fluid and assayed in BioSafe II scintillation fluid on a Beckman-Coulter Liquid Scintillation Counter. Aqueous and organic fractions were combined to obtain the total uptake CPM. CPM were measured for an aliquot representing 10% (by volume) of the chylomicrons injected into each mouse. This value was used to normalize the radiolabel data across mice. 2.8. Triglyceride uptake after gavage with ^3^H-Triolein {#sec2.8} -------------------------------------------------------- Wild-type and *Angptl4*^*−/−*^ mice were fasted 4 h, beginning at the onset of the light cycle. After 4 h, the mice were gavaged with 2 μCi of \[9,10-3H(N)\]-Triolein (Perkin Elmer, NET431001MC) in 100 μL olive oil. After 4 h, the mice were anesthetized with isoflurane, and tissues were harvested and weighed. Tissues were then analyzed for radiolabel as described in Section [2.7](#sec2.7){ref-type="sec"}. 2.9. Statistics and outlier identification {#sec2.9} ------------------------------------------ Statistics and outlier identification were performed using Graphpad Prism. Statistical significance was tested using Student\'s T-test unless otherwise indicated. Outliers were identified using Grubbs test and were excluded from graphs and from statistical analysis. The number of mice analyzed for each experiment ranged from 4 to 10 and is specified in each figure legend. 3. Results {#sec3} ========== 3.1. *Angptl4* expression is induced early in fasting {#sec3.1} ----------------------------------------------------- Consistent with previous reports [@bib5], we found that *Angptl4* expression in mice after a 24 h fast was upregulated in several tissues including heart, liver, and adipose tissue ([Figure 1](#fig1){ref-type="fig"}A). Because longer fasts induce a catabolic state in mice [@bib24], we asked at what point during the fast *Angptl4* was upregulated. We began fasting mice at the beginning of their light cycle and measured *Angptl4* expression after 0, 2, 4, 6, 8, 10, and 12 h. After 12 h of fasting, food was restored to the remaining mice and *Angptl4* expression was measured 2 or 4 h after this refeeding ([Figure 1](#fig1){ref-type="fig"}B). Surprisingly, *Angptl4* expression was already highly induced at the 2 h time point in all tissues ([Figure 1](#fig1){ref-type="fig"}C). In most tissues, expression gradually declined after the 2 h time point and returned to baseline or lower after refeeding.Figure 1Induction of *Angptl4* expression by fasting. (**A**) Male C57/Bl6 mice (n = 5/group were fed *ad lib* or fasted for 24 h. *Angptl4* expression in BAT, WAT, heart, liver, and hypothalamus (mean ± SEM; \*p \< 0.05, \*\*p \< 0.01, \*\*\*p \< 0.001). (**B**) Experimental design for fasting time course; male C57/Bl6 mice (n = 5/time point) were fasted at the onset of the light cycle and sacrificed after 0, 2, 4, 6, 8, 10, or 12 h of fasting, or after a 12 h fast, and 2 or 4 h of refeeding. (**C**) Expression of *Angptl4* in liver, brown adipose tissue (BAT), heart, quadriceps, subcutaneous white adipose tissue (sWAT), and gonadal WAT (gWAT) in mice subjected to the fasting time course depicted in (B). Points show expression relative to time 0 (mean ± SEM; \*p \< 0.05, \*\*p \< 0.01, \*\*\*p \< 0.001 when compared to time 0).Fig. 1 To investigate the possibility that the increased *Angptl4* expression at 2 h was the result of circadian regulation at the onset of the light cycle, rather than the result of actual fasting, we compared *Angptl4* expression at the same time points during the light cycle for mice that were fasted or allowed to feed *ad lib* ([Figure 2](#fig2){ref-type="fig"}A). For fasted mice, food was withdrawn at the beginning of the light cycle, and *Angptl4* expression was assessed for both fasted and fed mice after 2 and 12 h. Although mice typically eat less during the light cycle, the mice with access to food showed no significant upregulation of *Angptl4* expression at either 2 or 12 h ([Figure 2](#fig2){ref-type="fig"}B). However, consistent with our previous results, *Angptl4* expression was upregulated after 2 h in the fasted mice ([Figure 2](#fig2){ref-type="fig"}B). These data strongly suggest that fasting itself, not circadian cycles, drives increased *Angptl4* expression.Figure 2Expression of *Angptl4* is fasting induced, rather than circadian. (**A**) Diagram depicts mouse light cycle and experimental design for (B); tissues from wild-type mice (n = 6/group) were harvested at 0, 2, or 12 h, either after fasting or *ad lib* access to chow. (**B**) *Angptl4* expression in liver, brown adipose tissue (BAT), heart, quadriceps (quad), subcutaneous white adipose tissue (sWAT), and gonadal white adipose tissue (gWAT). Points show expression relative to time 0 (mean ± SEM; \*p \< 0.05, \*\*p \< 0.01, \*\*\*p \< 0.001 when comparing fasted to fed mice at the same time point).Fig. 2 3.2. *Angptl4*^−/−^ mice have reduced plasma triglycerides in the fasted state {#sec3.2} ------------------------------------------------------------------------------ To examine the physiological role of ANGPTL4 in triglyceride metabolism and fatty acid delivery, we used *Angptl4*^*−/−*^ mice. As expected, *Angptl4*^*−/−*^ mice did not express *Angptl4*, whereas in wildtype littermates, *Angptl4* expression was easily detectable, and was higher after a 4 h fast than after refeeding ([Figure 3](#fig3){ref-type="fig"}A). *Angptl4*^*−/−*^ mice had lower serum triglyceride levels than wild-type litter mates after a 4 h fast ([Figure 3](#fig3){ref-type="fig"}B), consistent with the established idea that ANGPTL4-deficiency results in increased LPL activity and thus, increased triglyceride clearance in the fasted state. Notably, no significant difference was observed in triglyceride levels between wild-type and *Angptl4*^*−/−*^ mice in refed mice (6 h fast, 2 h refeed), a time when *Angptl4* expression is reduced ([Figure 3](#fig3){ref-type="fig"}B).Figure 3Fasted plasma triglycerides are lower in *Angptl4*^*−/−*^ mice than in wild-type mice. ANGTPL4 expression in liver, gonadal white adipose tissue (gWAT) and liver (**A**) and plasma triglyceride levels (**B**) of female *Angptl4*^*−/−*^ and wild-type littermate mice that were fasted 4 h (fasted group) or fasted 6 h and allowed to feed *ad lib* for 2 h (refed) (n = 5--8/group). Bar graphs show mean ± SEM (\*p \< 0.05).Fig. 3 Gene expression of the LPL transporter GPIHBP1 also increases after an overnight fast [@bib25]. It had previously been reported that ANGPTL4-defiency disrupts this increase [@bib26], we did not observe this to be the case. After a 16 h fast, *Gpihbp1* expression responded similarly in both *Angptl4*^*−/−*^ and wild-type mice ([Supplemental Figure 1A and B](#appsec2){ref-type="sec"}). 3.3. Triglyceride clearance is increased in the adipose depots of *Angptl4*^−/−^ mice {#sec3.3} ------------------------------------------------------------------------------------- Given that ANGPTL4 inhibits LPL activity [@bib6], [@bib7], [@bib9], the reduced plasma triglycerides in *Angptl4*^*−/−*^ mice are likely the result of increased LPL-mediated triglyceride hydrolysis and fatty acid uptake. To determine the fate of triglyceride-derived fatty acids in *Angptl4*^*−/−*^ mice, and thus the likely location of ANGPTL4 action, we performed triglyceride clearance assays in wild-type and *Angptl4*^*−/−*^ mice using Intralipid spiked with radiolabeled triolein. We reasoned that the tissues in which there is the greatest ANGPTL4 inhibitory activity would manifest the greatest increase in radiolabel uptake in *Angptl4*^*−/−*^ mice. Thus, if circulating endocrine ANGPTL4 is primarily responsible for LPL inhibition, tissues with the greatest LPL activity (heart, adipose tissue, skeletal muscle) would have increased radiolabeled uptake in *Angptl4*^*−/−*^ versus wild-type mice. Whereas, if ANGPTL4 acts locally to inhibit LPL, we would expect tissues with the greatest expression of ANGPTL4, particularly adipose tissue, to have the greatest increase in radiolabel uptake in *Angptl4*^*−/−*^ mice. Our data strongly supported the latter possibility, as we observed increased radiolabel uptake exclusively in the adipose tissues of *Angptl4*^*−/−*^ mice ([Figure 4](#fig4){ref-type="fig"}). Uptake into heart was decreased and was unchanged in liver, kidney, and muscle ([Figure 4](#fig4){ref-type="fig"}).Figure 4*Angptl4*^−/−^ mice show increased uptake of triglyceride to adipose tissue compared to wild-type mice. Radiolabel uptake into tissues after 15 min in male wild-type and *Angptl4*^−/−^ mice that had been fasted 4 h and injected intravenously with Intralipid spiked with ^3^H-triolien (n = 6/group). Bars show relative radiolabel uptake per mg tissue normalized to wild-type (mean ± SEM; \*\*p \< 0.01, \*\*\*p \< 0.001).Fig. 4 3.4. Chylomicron clearance is increased in the white adipose tissue of *Angptl4*^−/−^ mice primarily in the fasted state {#sec3.4} ------------------------------------------------------------------------------------------------------------------------ Because Intralipid emulsions lack many properties of triglyceride-rich lipoproteins, including the presence of apolipoproteins, we performed triglyceride clearance assays using radiolabeled chylomicrons. Radiolabeled chylomicrons were isolated from the plasma of *Gpihbp1*^−/−^ mice fed ^3^H-triolein (see Section [2.6](#sec2.6){ref-type="sec"}). Fasted (4 h) and refed (6 h fast, 2 h refeed) *Angptl4*^*−/−*^ and wild-type littermate mice were injected intravenously with radiolabeled chylomicrons. Triglyceride clearance from the circulation was measured by taking blood samples 1, 5, 10, and 15 min after injection. After 15 min, tissues were harvested and the amount of radiolabel was measured to determine uptake into individual tissue. Experiments were performed in both male and female mice. In female mice, radiolabel clearance from the plasma was faster in *Angptl4*^*−/−*^ mice than in wild-type mice, but only in the fasted state ([Figure 5](#fig5){ref-type="fig"}A). In wild-type female mice, triglyceride uptake into white adipose tissue was significantly elevated in the refed state, whereas uptake into the heart decreased ([Figure 5](#fig5){ref-type="fig"}B--I). Consistent with the Intralipid uptake data ([Figure 3](#fig3){ref-type="fig"}B), radiolabel uptake into adipose tissue in the fasted state was significantly higher in female *Angptl4*^*−/−*^ mice than in wild-type mice ([Figure 5](#fig5){ref-type="fig"}D--F). Unlike wild-type mice, radiolabel uptake into the white adipose tissue of *Angptl4*^*−/−*^ mice did not significantly increase in the refed state. Interestingly, radiolabel uptake into adipose tissue after refeeding was not significantly different between wild-type and *Angptl4*^*−/−*^ mice, supporting the idea that ANGPTL4 regulates triglyceride metabolism primarily in the fasted state ([Figure 5](#fig5){ref-type="fig"}D--F). In female mice, radiolabel uptake into liver, kidney, brown adipose tissue, and skeletal muscle, did not differ significantly between wild-type and *Angptl4*^*−/−*^ mice ([Figure 5](#fig5){ref-type="fig"}A, B, H, I).Figure 5Chylomicron clearance and uptake in female wild-type and *Angptl4*^−/−^ mice. Fasted (4 h) and refed (6 h fast 2 h fed) female mice were injected intravenously with ^3^H-triglyceride containing chylomicrons (n = 6/group). (**A**) Clearance of radiolabel from the plasma 1, 5, 10, and 15 min after injection. Points represent counts per minute (CPM) in 10 μl plasma at the indicated time points (mean ± SEM; \*p \< 0.05, \*\*p \< 0.01). (**B--I**) Uptake of radiolabel after 15 min (% injected dose/mg tissue) into the tissues of wild-type and *Angptl4*^−/−^ mice (mean ± SEM; \*p \< 0.05, \*\*p \< 0.01).Fig. 5 In male mice, no significant differences in the rate of plasma triglyceride clearance were observed ([Figure 6](#fig6){ref-type="fig"}A). As in female mice, fasted *Angptl4*^*−/−*^ mice had reduced radiolabel uptake into heart and increased uptake into gonadal white adipose tissue (gWAT) and subcutaneous white adipose tissue (sWAT) when compared to fasted wild-type mice ([Figure 6](#fig6){ref-type="fig"}D, E, G). Interestingly, in the gWAT and sWAT of male mice, the differences in uptake between *Angptl4*^*−/−*^ and wild-type mice persisted in the refed state ([Figure 6](#fig6){ref-type="fig"}D, E). As in female mice, there were no differences between *Angptl4*^*−/−*^ and wild-type mice in radiolabel uptake into liver, kidney, skeletal muscle, or brown adipose tissue ([Figure 6](#fig6){ref-type="fig"}).Figure 6Chylomicron clearance and uptake in male wild-type and *Angptl4*^−/−^ mice. Fasted (4 h) and refed (6 h fast 2 h refed) male mice were injected intravenously with ^3^H-triglyceride containing chylomicrons (n = 6/group). (**A**) Clearance of radiolabel from the plasma 1, 5, 10, and 15 min after injection. Points represent counts per minute (CPM) in 10 μl plasma at the indicated time points (mean ± SEM). (**B--I**) Uptake of radiolabel after 15 min (% injected dose/mg tissue) into the tissues of wild-type and *Angptl4*^−/−^ mice (mean ± SEM; \*p \< 0.05).Fig. 6 We also tested triglyceride clearance after an oral gavage of ^3^H-triolein. Male mice were fasted 4 h and then administered ^3^H-triolein in olive oil by oral gavage. Tissues were harvested 4 h after gavage, and the amount of radiolabel was measured to determine uptake into individual tissue. Consistent with our previous observations, we found reduced radiolabel uptake in heart and increased uptake in adipose tissue ([Figure 7](#fig7){ref-type="fig"}). Together these data indicate that the lower plasma triglyceride levels in *Angptl4*^*−/−*^ mice are a result of increased uptake into adipose tissue.Figure 7Triglyceride clearance after oral gavage in male wild-type and *Angptl4*^−/−^ mice. Radiolabel uptake into tissues after 4 h in male wild-type and *Angptl4*^−/−^ mice that had been fasted 4 h and orally gavaged with olive oil spiked with ^3^H-triolien (n = 4--6/group). Bars show relative radiolabel uptake per mg tissue normalized to wild-type (mean ± SEM; \*p \< 0.05, \*\*p \< 0.01).Fig. 7 3.5. Lipase activity is increased in the adipose depots of *Angptl4*^−/−^ mice {#sec3.5} ------------------------------------------------------------------------------ To verify that the increased uptake of triglyceride-derived fatty acids into adipose tissue in *Angptl4*^*−/−*^ mice was the result of decreased inhibition of LPL, we also measured heparin-releasable lipase activity in the tissue of fasted wild-type and *Angptl4*^*−/−*^ mice. Lipase activity was greater in the white adipose tissue of *Angptl4*^*−/−*^ mice, consistent with the idea that the absence of ANGPTL4 in adipose tissue leads to increased LPL activity and great triglyceride uptake in these tissues ([Figure 8](#fig8){ref-type="fig"}). Interestingly, lipase activity was unchanged in the hearts of *Angptl4*^*−/−*^ mice, suggesting that the reduction of fasting triglyceride uptake in these mice is primarily a result of increased uptake in adipose tissue rather than a change in cardiac lipase activity.Figure 8Lipase activity in male and female *Angptl4*^−/−^ mice. Female (**A**) and male (**B**) wildtype and *Angptl4*^−/−^ mice (6--10/group) were fasted for 4 h. Heart, brown adipose tissue (BAT), subcutaneous adipose tissue (WAT), and quadriceps (quad) were harvested and lipase activity was measured. Bars show relative lipase activity in each tissue normalized to wild-type (mean ± SEM; \*p \< 0.05, \*\*p \< 0.01).Fig. 8 4. Discussion {#sec4} ============= Although the induction of *Angptl4* expression by fasting has been demonstrated by several previous studies [@bib5], [@bib7], [@bib26], the rapidity of *Angptl4* induction has not been fully appreciated. Initial reports looked at *Angptl4* expression after 24 h [@bib5], [@bib27] or overnight [@bib7] fasts. Here we demonstrated that *Angptl4* expression is induced after only 2 h of fasting, long before the mice enter a catabolic state. This contrasts with other fasting-induced factors, such as FGF21 or PPARα, which are induced after fasts of 8 h or longer [@bib28], [@bib29], [@bib30]. Somewhat surprisingly, *Angptl4* was not induced during the light cycle if mice had access to food, despite the fact that mice generally eat far less during the light cycle. These data suggest that while ANGPTL4 is induced rapidly upon true fasting, even a low level of feeding is sufficient to abolish this induction. A limitation of our study is that we primarily measure gene expression. We attempted to also measure protein levels but of the many commercial antibodies we tested none were specific for ANGPTL4 in mouse tissues (as judged by using *Angptl4*^−/−^ mice as a negative control). In this study, we used chylomicrons collected from *Gpihbp1*^*−/−*^ mice to assess chylomicron clearance and fatty acid uptake. Doing so allowed us to use physiologically accurate chylomicrons rather than triglyceride emulsions [@bib31], [@bib32], [@bib33] or "reconstituted" lipoproteins [@bib34], [@bib35]. Our data from this and other assays suggest that ANGPTL4 mediates its effects on plasma triglyceride levels and fatty acid delivery primarily in the fasted state. Female *Angptl4*^*−/−*^ mice had lower plasma triglycerides than wild-type, but only in the fasted state. Likewise, fatty acid delivery to tissues in female *Angptl4*^*−/−*^ mice was only significantly different from that in wild-type mice in the fasted state. This was also the case in male mice; however, for these mice, a trend towards increased triglyceride uptake into adipose tissue persisted two hours after refeeding, indicating that there may be gender specific differences in the kinetics of the refeeding response. Additional studies are necessary to determine whether ANGPTL4 kinetics may also be altered in females during other metabolic states including pregnancy. Deficiency in ANGPTL4 increased lipase activity and triglyceride-derived fatty acid uptake primarily in adipose tissue, indicating that this is the primary location of ANGPTL4 action in regulating triglyceride metabolism. ANGPTL4 is expressed in adipose tissue at much high levels than in other tissues [@bib5]. It is also important to note that vascular LPL is partially protected from ANGPTL4 by its endothelial cell transporter GPIHBP1 [@bib23], [@bib36], but that adipose-expressed ANGPTL4 could act on LPL before LPL binds GPIHBP1. Indeed, recent studies have shown that adipocyte-expressed ANGPTL4 can inhibit LPL even before LPL is secreted [@bib37]. Together these observations strongly suggest that the ANGPTL4 acting in adipose tissue is locally expressed. Although generation of an ANGPTL4 conditional allele and specific knockout of ANGPTL4 in adipose tissue will likely be needed to completely settle the issue, nonetheless, we predict that adipocyte-expressed ANGPTL4 is the major driver of the plasma triglyceride phenotypes observed in ANGPTL4 knockout mice as well as the shifts in fatty acid delivery observed during fasting. Our studies support a model in which ANGPTL4 primarily acts locally in adipose tissue to inhibit LPL, and thus triglyceride uptake, during fasting. This inhibition would redirect triglyceride-derived fatty acids to tissues such as heart and muscle. Upon refeeding, ANGPTL4 inhibition of LPL ceases in adipose tissue, and at the same time inhibition of LPL by ANGPTL3 and ANGPTL8 increases in heart and skeletal muscle [@bib38], [@bib39]. Thus, delivery of triglyceride-derived fatty acids to heart and skeletal muscle would slow and more fatty acids would be delivered to adipose tissue for storage. Although *Angptl4* is highly expressed in adipose tissue, it is also expressed at lower levels in other tissues and cell types, including liver, skeletal muscle, heart, and macrophages [@bib38]. If, as we propose, adipose ANGPTL4 is primarily responsible for plasma triglyceride and tissue fatty-acid uptake phenotypes in *Angptl4*^−/−^ mice, what is the role of ANGPTL4 in other tissues? It seems likely that the ability of ANGPTL4 to inhibit LPL is still at work in these tissues. ANGPTL4 has been shown to be important in preventing macrophage foam cell formation [@bib40], in shifting fatty acid delivery from unexercised muscle to exercised muscle [@bib41], and in preventing lipotoxicity in skeletal muscle and heart [@bib42], [@bib43]. ANGPTL4 has also been shown to perform roles outside of triglyceride metabolism, and these roles may also be in play [@bib44]. Again, generation of tissue-specific *Angptl4* knockout mice may be necessary to completely address the role of ANGPTL4 in specific tissues. Conflict of interest {#appsec1} ==================== None declared. Appendix A. Supplementary data {#appsec2} ============================== The following is the supplementary data related to this article: This work was supported by grants from the National Institutes of Health (R01HL130146 \[BSJD\], R01DK106104 \[MJP\], and T32GM082729 \[EMC\]) and an American Heart Association Scientist Development Grant (12SDG8580004 \[BSJD\]). Supplementary data related to this article can be found at [http://dx.doi.org/10.1016/j.molmet.2017.06.007](10.1016/j.molmet.2017.06.007){#intref0015}

1. Introduction =============== Spider bites are common worldwide, and spider bite epidemics have been reported in countries such as Italy, Spain, and the USA ([@b1-epj-09-4703]). Some of these epidemics were caused by spiders that live in some provinces of Iran ([@b2-epj-09-4703]--[@b5-epj-09-4703]). In recent years, cases of spider bite have been reported from suburban areas of Shiraz, which could not have been diagnosed if the patients had not seen the spiders. Spider bites are relatively common in Khorasan Razavi province ([@b6-epj-09-4703]--[@b9-epj-09-4703]). There are also only two other case reports of spider bite from the south of the country ([@b10-epj-09-4703]) and Bandar-Abbas ([@b11-epj-09-4703]). Lack of reports on spider bites from other areas of Iran does not mean that there are no dangerous spiders or no events of spider bites in these areas; failure to diagnose spider bites may be one of the reasons. If the patient had not seen the spider, the clinical manifestations of latrodectism would have been easily mistaken for other types of bite or sting, an infectious disease, or severe withdrawal symptoms (for example, due to the chronic abuse of opioid substances in combination with buprenorphine), and the clinical manifestations of loxoscelism may also be easily misdiagnosed as cellulitis, a particular skin infection, or Gadim scorpion (*Hemiscorpius lepturus*) sting. ([@b11-epj-09-4703]). Based on the importance and prevalence of spider bites in Iran, the Razi Vaccine and Serum Research Institute has been conducting investigations to produce an anti-venom serum against spider bites. It is important that physicians in Iran are familiar with the geographical distribution of dangerous spiders, the clinical manifestations and management of their bites. Accordingly, based on spider bites reports from Iran and personal experiences, in this editorial, the subject of spider bites in Iran is reviewed and discussed in detail, which may be beneficial for Iranian physicians. 2. Discussion ============= So far, more than 46,000 spider species have been identified worldwide, about 200 species' venom of which is dangerous to humans ([@b5-epj-09-4703]). With the exception of two spider families, they all have venomous glands, but they are not usually dangerous to humans, due to many reasons, such as the spider's small fang size, low amount of injected venom, little damage effect of venom, and spiders' behavior and life style. ([@b5-epj-09-4703]). More than 600 species have been registered in Iran, out of which six species are dangerous to humans, five species belong to the widow spider group (*Latrodectus* spp.), and one belongs to the recluse spider group (*Loxosceles* spp.) ([@b5-epj-09-4703]). [Table 1](#t1-epj-09-4703){ref-type="table"} demonstrates the names, characteristics, and the provinces in which the spiders were observed, and [Figure 1](#f1-epj-09-4703){ref-type="fig"} depicts their images. Reports on spider bites from Iran include only Mediterranean widow and recluse spiders, and there is no information regarding bites from other species ([Table 1](#t1-epj-09-4703){ref-type="table"}) ([@b6-epj-09-4703]--[@b11-epj-09-4703]). 2.1. Clinical manifestations and management of Mediterranean widow spider bite (latrodectism) --------------------------------------------------------------------------------------------- The Mediterranean widow spider (*Latrodectus tredecimguttatus*) is called *Dolmak* in Khorasan Razavi province. Bites by this species have been only reported from Mashhad. According to Afshari et al. ([@b6-epj-09-4703], [@b7-epj-09-4703]), this spider lives close to the ground, in farms, or around human habitation, and usually bites people outdoors. This spider commonly bites forearms and lower legs for reasons that it hides in shoes and because wheat can fall on the farmers' exposed forearms when harvesting. The common clinical manifestations of bites are mild burning sensation at the bite site, local pain, tingling feeling which begins after a while, red color halo around the bite site, diffused pain in the involved extremity, abdominal and pelvic pain, profuse sweating, chills, restlessness, dyspnea, flushing, muscular spasm, nausea, vomiting, and vertigo. It is also reported that there would sometimes be no obvious signs of a spider bite. In twenty-five percent of the cases, ST depression was reported in at least two precordial leads. Dysrhythmia and myocarditis may also occur. Increase of the creatine phosphokinase (CPK) level was also reported in some cases. There is an interesting phenomenon called washerwoman's hand's syndrome and happens following several days of profound sweating which can cause wrinkles in the skin of bitten patients' hands. Death, due to this type of spider bite, is rare, but it may lead to death following pulmonary edema, cardiac complications, and disseminated intravascular coagulation (DIC). In recent years, there have been two cases of *L. tredecimguttatus* spider bites from suburban areas of Shiraz, one of whom was presented with burning sensation at the site of bite, abdominal, chest, and back pain with severe agitation, hypertension, tachycardia, and increased CPK. The clinical presentations of the patient were at first very similar to that of someone who had severe withdrawal symptoms (due to the chronic use of methadone in combination with buprenorphine). The patient showed fever and profound sweating approximately 24 hours after the bite. However, there was no evidence of a bite at the bite's location. The biting spider was immature with orange-red spots on the abdomen, similar to species found in other countries ([Figure 2](#f2-epj-09-4703){ref-type="fig"}). The other case had burning sensation at the bite site, diffused pain in the involved extremity, abdominal and pelvic pain, profound sweating, severe agitation, headache, tachycardia, and increased CPK. Similarly, no sign of a bite was observed at the bite site. The spider was completely black. Although there is no information regarding other *Latrodectus spp.* bites in Iran, it can be expected that the clinical manifestations of biting by other spiders of this group are similar to those of Mediterranean widow spiders, with lower or higher severity. 2.2. Treatment -------------- The most useful treatment for latrodectism is anti-venom administration. Producing spider bite anti-venom in the Razi Vaccine and Serum Research Institute is under investigation. Therefore, the treatment of latrodectism is currently supportive including observation, administration of fluids, analgesics (morphine, pethidine, apotel or nonsteroidal anti-inflammatory drugs \[NSAIDs\]), muscle relaxant (methocarbamol and diazepam), calcium gluconate, and anti-histamine and steroids in the event of allergic reactions ([@b7-epj-09-4703]). 2.3. Clinical manifestations and management of Mediterranean recluse spider bite (loxoscelism) ---------------------------------------------------------------------------------------------- The Mediterranean recluse spider (*Loxosceles rufescens)* mostly lives in residential areas and houses. This spider keeps itself hidden among clothes, bed sheets, and beneath objects. It is active during the night and is usually non aggressive ([@b5-epj-09-4703], [@b6-epj-09-4703]). So far, several cases of Mediterranean recluse spider bites have been reported from Bandar-Abbas, Mashhad ([@b8-epj-09-4703]), and southern areas of the country ([@b10-epj-09-4703], [@b11-epj-09-4703]). Clinical manifestations in bitten patients at first included mild burning sensation at the site of bite; papules, pain, erythema, and itching occur after several hours. The cytotoxic reactions usually occur in the second to sixth days after the bite, which lead to cellulitis, severe pain, and swelling accompanied by systemic manifestations such as fever, nausea, and sweating. Skin necrosis occurs at the site of bite from one to two weeks after the bite. The systemic toxicity rarely leads to coagulopathy, hemolysis, and renal failure (from the third day to the seventh day after bite) ([@b8-epj-09-4703], [@b11-epj-09-4703]). In recent years, two cases of *L. rufescens* bite have been reported from suburban areas of Shiraz. The first case presented with mild burning sensation and swelling at the bite site. The swelling increased after a couple of days. A slight skin necrosis had occurred at the site of bite after several days. Other clinical manifestations were not observed in the patient. The other case presented with mild burning sensation and pain at the site of the bite. After five to six hours, the patient developed generalized flushing and rash. The swelling and pain at the bite site increased in two to three days, and slight skin necrosis was developing. If the patient had not seen the spider (*L. rufescens*), the spider bite may not have been diagnosed or may have been mistaken for a Gadim scorpion sting, which is painless, skin infection (such as the infection caused by methicillin-resistant *Staphylococcus aureus*), or dermatitis ([@b11-epj-09-4703]). 2.4. Treatment -------------- Anti-venom administration is the exact treatment for loxoscelism, but is not available in Iran. Thus, as latrodectism, the treatment of loxoscelism is supportive at present including observation, washing the site of bite with water and soap, applying ice pack to reduce erythema, swelling, itching, and pain, avoiding debridement of the skin lesion, avoiding use of topical steroids, administration of tetabulin, tetanus vaccine if necessary, administration of analgesics, anti-histamine, and antibiotics. Systemic steroids, dapsone, and topical tetracycline are also recommended ([@b11-epj-09-4703]--[@b13-epj-09-4703]). 3. Conclusions ============== Some of the world's most dangerous spiders have been certified in some geographical areas of Iran. Spider bites are relatively common in Khorasan Razavi province, but there are some other reports from other areas. Given the geographical distribution of dangerous spiders in Iran and nonspecific presentation of spider bites, when the patient had not seen the spider, one must keep the diagnosis in mind and question patients regarding possible exposure to spiders. Spider bites can be presented in two forms of latrodectism and loxoscelism. In spider bite cases, the patients should be hospitalized, and the serum level of calcium (Ca), potassium (K), sodium (Na), blood urea nitrogen (BUN), creatinine (Cr), complete blood count (CBC), platelet count (PLT), prothrombin time (PT), partial thromboplastin time (PTT), international normalized ratio (INR), creatine phosphokinase (CPK), and urine analysis (U/A) should be checked. At present, the treatment of spider bites in Iran is supportive. In addition, tetabulin and, if necessary, the tetanus vaccine should also be administered. It is important that Iranian physicians are familiar with the geographical distribution of dangerous spiders, (particularly the areas in which they practice), clinical manifestations, and management of their bites. Administration of anti-venom is the most useful treatment for a spider bite, the production of which is under investigation in the Razi Vaccine and Serum Research Institute. The author cheerfully acknowledges his gratitude to Mr. Alireza Zamani, the M.Sc. candidate of Animal Biosystematics, Department of Animal Biology, School of Biology, College of Sciences, University of Tehran, Tehran, Iran, for his invaluable comments, reviewing and editing of the current article, and also for providing images of the spiders. iThenticate screening: October 01, 2016, English editing: November 22, 2016, Quality control: April 14, 2017 **Conflict of Interest:** There is no conflict of interest to be declared. {#f1-epj-09-4703} {#f2-epj-09-4703} ###### Different species of dangerous spiders reported from different geographical areas of Iran Scientific name General name General characteristics The province(s) in which the spider was observed -------------------------------- ------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------ *Loxosceles rufescens* Mediterranean recluse spider Brown color with a violin-shaped marking on the dorsum of the cephalothorax Tehran, Hormozgan, Khorasan Razavi, Fars, Mazandaran, Alborz, Qom *Latrodectus tredecimguttatus* Mediterranean widow spider, *Dolmak* The Iranian mature species is completely black which may be mistaken for a Dahl widow spider at first sight. The immature species have orange-red spots (mostly 13 spots) on the abdomen Alborz, East and West Azerbaijan, Alborz, Bushehr, Tehran, Khorasan Razavi, North Khorasan, Semnan, Qom, Golestan, Mazandaran, Hormozgan, Fars, Khuzestan, *Latrodectus dahli* Dahl widow spider The female is completely black and the male is in light black color East Azerbaijan, Bushehr, North Khorasan, South Khorasan, Khorasan Razavi, Fars, Hormozgan *Latrodectus geometricus* Brown widow spider The color varies from white to brown. Leg straps have dark and light colors with different geometric shapes on dorsal part of abdomen Khorasan Razavi *Latrodectus pallidus* White widow spider Brilliant white color of the abdomen; legs are light brown Alborz, Bushehr, Khorasan Razavi, South Khorasan, Semnan, Qazvin, Hormozgan *Latrodectus cinctus* Red-back widow spider, Round widow spider Black color with an orange to red longitudinal marking on the dorsum of the abdomen Bushehr, Hormozgan, Khuzestan

Introduction {#s1} ============ Epidemiological research shows that psychiatric disorders, particularly behavioral problems, are quite common in early childhood. The most frequent behavioral disorder in preschoolers is oppositional defiant disorder (ODD), the prevalence of which is around 10% in the United States community samples (Bufferd et al., [@B8]) and 7% in Spain (Ezpeleta et al., [@B18]). ODD is conceptualized as the result of the interaction of individual characteristics (such as a high negative affectivity, low effortful control, difficulties in social learning or emotional dysregulation) with environmental adversities (such as dysfunctional parenting style, parental psychopathology, socioeconomic problems or family conflict). The main objective of this work is to value the mediating mechanisms of ODD in the general population of preschool children, through pathways analysis including the family socioeconomic status, the parenting style and the children\'s executive functioning (EF). Socioeconomic status (SES) as a risk factor for ODD --------------------------------------------------- In the group of environmental adversities, lower SES has showed a strong association with children\'s behavioral problems. Reiss ([@B45]) describes a clear relationship between socioeconomic deprivation and mental health problems in childhood and adolescence, with low household income and low parental education being the strongest predictors. In studies using samples of typically developing preschoolers, strong associations between family SES and child behavioral problems were found, with a possible mediating role of the neighborhood the family is living in (Kalff et al., [@B31]; Boyle and Lipman, [@B6]). Davis et al. ([@B13]) found negative associations between SES and parental report of child emotional symptoms, conduct problems, hyperactivity problems, and/or peer relationship problems. They also found teacher-reported hyperactivity and conduct problems to be associated with lower SES, and with low parental education in particular. Parental style as a potential mediator for ODD ---------------------------------------------- Into the environmental adversities domain, parenting style has also showed a strong association with children\'s behavioral problems, especially with ODD. A clear relation between parenting and problem behavior in young children has been reported with, for example, harsh parenting practices, inconsistent parental discipline, and negative parent emotional expressiveness having unique and cumulative effects on child disruptive behavior (Hughes and Ensor, [@B28]; Duncombe et al., [@B16]). Examining the role of early parenting behaviors in predicting later ODD symptoms, Harvey and Metcalfe ([@B24]) found a negative association between maternal warmth and ODD symptoms. They also found that paternal laxness predicted ODD symptoms one year later. Tung et al. ([@B52]) found that harsh punishment was positively associated with parent-rated aggressive behavior and ODD symptomatology. Parenting behavior and SES seems also related to each other. Conger et al. ([@B12]) examined the link between economic stress in family life and adolescent internalizing and externalizing behavioral symptoms. In their model they found for both parents a significant path from economic pressure to parental hostility through marital conflict or parent-adolescent financial conflict. Pinderhughes et al. ([@B43]) found a significant, direct relation between SES and discipline responses. Indirect relations suggested that these SES differences in discipline responses were due to differences in parenting beliefs and levels of stress. Dodge et al. ([@B15]) found that children from families with lower SES are more likely to receive harsh discipline. Furthermore, their mothers are less warm in their behavior toward the children (Dodge et al., [@B15]). Executive functioning as a potential mediator for ODD ----------------------------------------------------- EF are the cognitive processes that enable purposeful and goal-directed behavior and flexible, adaptive responses to changes in the environment (Anderson, [@B1]; Hughes and Ensor, [@B29]). They include mental processes such as planning, working memory, inhibition of inappropriate responses, flexibility in adaptation to environmental changes, and decision making (Nigg, [@B41]). The literature shows a systematic relation between EF and problem behavior in typically developing preschoolers (Espy et al., [@B17]). Studies in clinical samples usually have a mixture of children with attention deficit/hyperactivity disorder (ADHD) and/or ODD. Given the frequent comorbidity between these disorders, it is hard to distinguish differences in executive functioning associated with each diagnostic type. Some studies show deficits in working memory, planning and organizing and inhibition, but when controlled for ADHD these associations disappear for ODD (Thorell and Wåhlstedt, [@B51]; Schoemaker et al., [@B47]; Skogan et al., [@B49]). However, developmental theorists have proposed the distinction between more abstract-cognitive, "cool" and "hot" EF (Zelazo and Müller, [@B54]). Cold cognition is thought to be associated with executive functions elicited by abstract and deconceptualized tasks (such as motor response inhibition, attention and cognitive flexibility), and the area of the brain that is utilized for these tasks is the dorsolateral prefrontal cortex. Hot cognition refers to the tasks that require the regulation of emotion or motivation (as well as reevaluating the motivational significance of a stimulus), and it is executed by the activation of the ventral and medial areas of the prefrontal cortex. Some studies have related the association of disruptive disorders such as ODD, conduct disorder (CD) and attention-deficit/hyperactivity disorder (ADHD), with deficits in cool abstract-cognitive functions and hot reward-related EF tasks, but it is unclear what the specific association of ODD with these neuropsychological deficits is, independent of ADHD. A recent study in a clinical sample of *N* = 59 children with a history of early-onset ODD problems (Hobson et al., [@B26]) observed that ODD with no ADHD is associated with deficits in both hot EF performance (particularly with risky decision-making tasks) and cool EF (slower speeds of inhibitory responding). They concluded that these EF deficits specifically related to ODD should be implicated in theories of antisocial behavior development. Previous research also shows that in typically developing preschoolers, maternal positive parenting behaviors are related to better child performance in tasks assessing working memory, conflict and impulse control and categorization (Bernier et al., [@B4]) and that these parental influences already start in infancy (Bernier et al., [@B3]; Kraybill and Bell, [@B35]). Furthermore, a relationship between EF and SES has been found, indicating an impact of early access to learning resources on the acquisition of these skills (Clark et al., [@B11]). Rhoades et al. ([@B46]) found effects of family ecological risk profiles and children\'s earliest environment, both during infancy, on the development of EF 2 years later. They also imply that this link between SES and EF is at least partially explained by associated parenting behaviors. Children\'s gender and ODD -------------------------- The literature indicates that the aforementioned relationships of ODD with SES, parenting practices and EF may be different according to the child\'s gender. At a basic epidemiological level, problem behaviors have obtained higher prevalences in boys (Chen, [@B10]; Ullebø et al., [@B53]; Searle et al., [@B48]). This differentiation, however, was not corroborated for ODD at very young ages (Ezpeleta et al., [@B18]). Children\'s gender differences related to parenting practices have reported conflicting outcomes. Tung et al. ([@B52]), for instance, found that inconsistent discipline was positively associated with aggressive behavior and rule breaking only for boys. Other studies have suggested that the influence of parenting behaviors on psychological functioning is probably stronger in girls than in boys (Javo et al., [@B30]). Kim et al. ([@B33]) found associations between overreactive parenting and externalizing behavior in girls, and between lax parenting and externalizing behavior in boys, while for internalizing behavior the associations were inverse. The association between SES and gender also seems more ambiguous. Some researchers have observed evidence of an equal distribution of externalizing behavior in male and female preadolescents. Mrug and Windle ([@B40]) found more children\'s externalizing behavior in preadolescents living in neighborhoods with higher poverty rates, but the distribution of these externalizing behaviors were equally distributed between boys and girls. However, Henninger and Luze ([@B25]) concluded an effect of time spent in poverty on the occurrence of externalizing behaviors over time only for girls. Kim et al. ([@B33]) also found a different effect depending on the SES level: the association between overreactive parenting and internalizing behavior in boys was specific to families from low socioeconomic levels. For EF, gender differences may depend on the kind of task executed. Several researchers showed that girls outperform boys in inhibition tasks (Kochanska et al., [@B34]; Olson et al., [@B42]). Raaijmakers and colleagues found more EF deficits in boys, specifically in inhibition, verbal fluency, working memory, and set shifting. These differences were found irrespective of aggression or attention problems (Raaijmakers et al., [@B44]). Objectives ---------- Although EF, parenting behaviors and SES have been strongly associated with child behavioral problems, particularly with ODD, no studies have analyzed the underlying processes including these factors and their mediating dependencies in predicting ODD symptomatology levels in preschoolers. Structural-causal modeling is required to assess the direct and indirect associations among the variables explaining the early onset and development of ODD in the context of developmental psychopathology. The objective of this study is to investigate the mediating mechanisms of the ODD levels in preschool age through pathways analysis, using a model that includes as potential predictors family SES, parenting behaviors, children\'s EF and sex. Based on available empirical evidence, it is hypothesized that: (1) low SES will be directly associated with high ODD scores; (2) the association of SES and ODD will also be mediated by parenting style and executive functioning, and parents from low SES families will score higher on negative parenting behaviors (poor monitoring/supervision, inconsistent discipline, and corporal punishment) which in turn will lead to their children scoring higher on EF deficits; and (3) children from families with low SES, children from parents with negative parenting behaviors, and children with deficits in EF score higher on ODD symptomatology. Regarding children\'s gender, it is expected that this variable may affect the pathways including SES, EF and parenting, but the lack of empirical evidence prevents us from hypothesizing its specific contribution to the structural modeling. Methods {#s2} ======= Participants ------------ The data are part of a longitudinal study concerning risk factors of behavioral problems in preschoolers (Ezpeleta et al., [@B18]). They were collected through a multisampling design, and corresponded to the in-school population of preschoolers P3-grade (age 3 years old) in Barcelona, Spain. A random sample was considered, inviting 2283 families to be included in the study. The participation rate was 58.7%, which is fairly high for community studies, leading to a sample in the first phase of 1341 families. In this phase semi-public schools were significantly more likely to refuse to participate than public ones (*p* \< 0.001), and families with high SES participated more than families with low SES (*p* \< 0.001). The children were screened for symptoms of ODD by having their parents complete the Strengths and Difficulties Questionnaire for 3- and 4-year-olds (SDQ^3−4^) (Goodman, [@B22]). Children with mental retardation and/or pervasive developmental disorders, and families with difficulties with Spanish or Catalan, without a primary caregiver who could report about the child, or who were moving to another city within a year were excluded. In the second phase all children with a positive screen score (indicating high ODD symptomatology, and defined as a raw score above 4 in the SDQ^3−4^ conduct scale or at least one symptom of the ODD symptoms list) (*n* = 522) were invited to participate, and from the children with negative screen score (indicating low symptomatology, for children who did not meet the positive screen criterion) (*n* = 756) a random 30% was asked to continue. In this phase 10.6% of the families invited refused continued participation. The final sample consisted of *N* = 622 preschoolers and their families, of whom 417 had high ODD symptomatology and 205 were controls. No differences were found when participants and refusals were compared by gender (*p* = 0.82) or by type of school (*p* = 0.85). The data used in this work correspond to this first assessment at preschool P3 grade, when the participating children were approximately 3 years of age. Final sample included *N* = 604 children with data available in all the measures of the study. No statistical differences emerged on comparing children with all the measures available (*n* = 604) and with missing values in the measures of the study (*n* = 18) in gender (*p* = 0.64), SES (*p* = 0.83) or type of school (*p* = 0.45). Instruments ----------- The *Strengths and Difficulties Questionnaire*^3−4^ (SDQ^3−4^) (Goodman, [@B22]), in its parental version, measured children\'s psychopathology. This is a brief screening questionnaire with 25 items codified through a 3-point ordered scale (0 = *not true*, 1 = *sometimes true*, and 2 = *certainly true*). The items are factorized in 5 scales: hyperactivity, emotional symptoms, conduct problems, peer problems and prosocial behavior. The conduct problems scale includes four items that correspond to the DSM-IV ODD symptoms "often has temper tantrums or hot tempers," "generally disobedient, usually does not do what adults request," "often argumentative with adults," "can be spiteful to others." Four additional items were included in the Spanish version to measure the remaining DSM-IV ODD symptoms: "often deliberately annoys others," "often blames others for his/her mistakes or bad behavior," "is easily offended by things others say" and "is often angry and resentful." These eight items allowed the definition of an additional scale, labeled ODD in this work, which measured the DSM-IV ODD symptomatology level. The scores analyzed in this work corresponded to the ODD and hyperactivity (labeled ADHD next) scales, which achieved moderate internal consistency (alpha equal to α = 0.74 and α = 0.67). The *Alabama Parenting Questionnaire* (APQ) (Frick, Unpublished Rating Scale) measured parenting style through 42 items reported by the parents themselves and grouped into 5 scales: parental involvement, positive parenting, poor monitoring/supervision, inconsistent discipline and corporal punishment. Items are codified in a 5-point Likert scale (1 = *never* to 5 = *always*). The Spanish version of this questionnaire, adapted and validated for preschoolers by de la Osa et al. ([@B14]) in the same sample of this study obtained adequate psychometric evidence (moderate to high internal consistency and good associations with external measures of psychopathology). In this work, the APQ was answered by 604 participating children. The *Behavior Rating Inventory of Executive Functioning--Preschool version* (BRIEF-P) (Gioia et al., [@B21]) measured EF reported by children\'s teachers. It consists of 63 items referring to behaviors with a 3-point ordered scale (0 = *never*, 1 = *sometimes* and 2 = *very often/always*). Items are structured in five clinical scales (the higher the score in each scale, the higher the level of impairment in the construct): inhibition (assessing problems in inhibitory control), shift (difficulties in moving freely among situations, activities, or aspects of a problem), emotional control (difficulties for controlling emotional response), working memory (difficulties in holding information in mind for completing a task), and plan/organize (problems for anticipating future events and taking appropriate steps, and for tapping information to achieve a goal) (Gioia et al., [@B21]). The Spanish version of the BRIEF-P used in this work achieved adequate psychometric properties (Ezpeleta et al., [@B19]) in the study\'s sample itself: high internal consistency (α ≥ 0.87) and moderate convergent validity with other measures of psychopathology and temperament. A total of 94 teachers (96.8% females) from 54 schools completed the BRIEF-P for 620 participating children. Participating teachers had known their students for a mean of 7.6 months (*SD* = 2.2). The *Hollingshead\'s Four Factor Index of Social Status* (Hollingshead, [@B27]) was used to measure families\' socioeconomic status. This index corresponds to a family\'s composite score based on the parents\' occupational level (rated on a 9-point scale, in which 9 = higher executive, proprietor of large businesses or major professional and 1 = farm laborers, menial service workers, students or housewives) and educational level (rated on a 7-point scale, in which 7 = graduate/professional training and 1 = less than primary school). The composite family index is computed by multiplying the occupational code by a weight of 5 and the educational code by a weight of 3, and summing the products. Hollingshead Index raw scores range from 8 to 66 (with higher scores reflecting higher SES) and allows placing the families in one of five social classes: 1 = High, 2 = Medium-High, 3 = Medium, 4 = Medium-Low, and 5 = Low. The numbers labeling these five social classes (1--5) represent a range of classes from high to low. For example, a medium-class family has a primary breadwinner who has completed college and could be employed as a school teacher or as an insurance salesperson. Procedure --------- The longitudinal project was approved by the ethics review committee of the institution to which the original authors belonged. The heads of the participating schools, as well as the children\'s parents, received a complete description of the study. All parents of children from P3-grade were invited to participate and required to complete the screening questionnaire. Families who agreed to be included into the longitudinal study and who met the screening criteria were contacted by telephone and gave written consent. A diagnostic interview with the parents took place at the school by previously trained interviewers who were blind to the children\'s screening group. After the interview, the parents completed the APQ and were asked about demographic characteristics. The children\'s teachers were asked to complete the BRIEF-P before the end of the academic year. All the measures for the current study were completed during the P3-grade academic year, when the children were around 3 years old. Statistical analysis -------------------- Data was analyzed with Stata13 for Windows. Due to the double phase sampling, all the analyses included sample weights to correct for the unequal probabilities of selection: each child was weighted with the reciprocal of its probability of selection in the second phase of the sampling (this allows the generalization of the results to the original general population). Structural equation modeling (SEM) was conducted to test the hypothesized mediation model that specifies the relationship between SES, parenting behaviors, EF measures and ODD symptoms level. Children\'s gender was defined as a group variable, since it was expected that boys and girls could obtain different structural coefficients in the pathway. The Maximum Likelihood method of parameter estimation was used and goodness-of-fit was evaluated using the usual statistics: the chi-square test (χ^2^), the Root Mean Square Error of Approximation (RMSEA), the Bentler\'s comparative Fit Index (CFI), the Tucker-Lewis Index (TLI), and the Standardized Root Mean Square Residual (SRMR). Adequate model fit was considered for non-significant χ^2^ test, RMSEA \< 0.08, TLI \> 0.9, CFI \> 0.9 and SRMR \< 0.1. The global predictive capacity of the model was measured with the Coefficient of Determination (CD). According to the procedures defined by Baron and Kenny ([@B2]) and Sobel ([@B50]), the mediational pathway is considered to occur when the following criteria are met: (a) the independent variable (IV) is statistically associated with the dependent variable (DV); (b) the IV is statistically associated with the hypothesized mediator variable (MV); (c) the hypothesized MV is statistically associated with the DV; and (d) the effect of the IV on the DV diminishes on the addition of the MV to the model. In this study, the mediation test was carried out through Sobel\'s method ("*estat teffects*" command in Stata13), which breaks total effects down into direct and indirect effects. Due to the strong comorbid association between ODD and ADHD symptom levels, and the common relationships of both disorders with the variables analyzed in the study, the SDQ-ADHD score was included as a covariate in the SEM. Results {#s3} ======= Description of the sample ------------------------- The mean age of the children analyzed in the study was 3.77 years old (*SD* = 0.33), 302 participants (50%) were boys, 586 (97.0%) were born in Spain and 539 (89.2%) were Caucasian-European (6.3% were Hispanic from South-America and 4.5% pertained to other ethnic groups). Twenty-six children (4.3%) lived in a single parent family and 9 in a reconstructed family (1.5%). The first rows of Table [1](#T1){ref-type="table"} contains the distribution of the quantitative variables analyzed in this work (range, mean, and standard deviation) for the total sample and stratified by the children\'s gender. Differences between boys and girls emerged for the mean comparison of BRIEF inhibit, working memory and plan-organize scales, and also for the APQ monitoring scale. The final part of Table [1](#T1){ref-type="table"} contains the distribution of the SES, also for the total sample and for the cohorts defined by children\'s gender (boys and girls did not differ in the SES levels). ###### **Distribution of the variables of the study**. **Total sample (*N* = 604)** **Girls (*N* = 302)** **Boys (*N* = 302)** ***t~df~*~=\ 602~** ***p*** ---------------------- ------------------------------ ----------------------- ---------------------- --------------------- --------- ------- ------ ------ --------- ------- ------ ------ ---------------------- --------- SDQ-ODD score 0 9 3.8 2.12 0 9 3.7 2.12 0 9 3.8 2.13 0.73 0.469 BRIEF-Inhibit 16 48 23.0 6.91 16 42 21.6 6.20 16 48 24.4 7.30 5.00 \<0.001 BRIEF-Shift 10 29 13.2 3.51 10 25 12.9 3.40 10 29 13.4 3.61 1.85 0.064 BRIEF-Emot.Control 9 27 12.1 3.52 9 24 12.0 3.39 9 27 12.3 3.64 0.97 0.333 BRIEF-Working.Mem 17 48 22.5 6.77 17 47 21.4 6.07 17 48 23.7 7.21 4.22 \<0.001 BRIEF-Plan/organize 10 30 13.4 3.79 10 27 12.6 3.18 10 30 14.1 4.18 4.95 \<0.001 APQ-Positive 5 24 18.4 2.96 10 24 18.5 3.04 5 24 18.3 2.88 0.92 0.358 APQ-Involvement 7 32 20.5 5.83 8 32 20.8 5.79 7 32 20.2 5.85 1.23 0.219 APQ-Monitoring 5 22 12.8 2.02 6 19 12.5 1.93 5 22 13.1 2.07 3.35 0.001 APQ-Corporal 0 9 0.98 1.13 0 5 0.9 1.00 0 9 1.05 1.24 1.60 0.110 APQ-Inconsistent 0 15 4.17 2.33 0 15 4.23 2.25 0 13 4.11 2.41 0.59 0.554 SDQ-ADHD *covariate* 0 10 3.89 2.49 0 10 3.71 2.52 0 10 4.07 2.45 1.74 0.083 ***n*** **%** ***n*** **%** ***n*** **%** **χ^2^~*df*\ =\ 4~** ***p*** SES High 199 32.9 98 32.5 101 33.4 1.57 0.813 Medium-high 188 31.1 93 30.8 95 31.5 Medium 86 14.2 40 13.2 46 15.2 Medium-low 96 15.9 51 16.9 45 14.9 Low 35 5.8 20 6.6 15 5.0 BRIEF, Behavior Rating Inventory of Executive Functioning--Preschool version; APQ, Alabama Parenting Questionnaire; SDQ, Strengths and Difficulties Questionnaire; SD, Standard deviation; SES, Socioeconomic Status. Structural equation model ------------------------- Table [2](#T2){ref-type="table"} shows the correlation-matrix for the variables of the study. Since one of the objectives of this work is to value the potential differences in the pathways due to sex, correlation matrixes have been computed stratified by children\'s gender. The APQ and BRIEF-P scores which achieved the strongest associations with the SES and/or the ODD levels (in any of the correlation matrices, for boys and/or girls) were considered for an initial SEM: (a) EF factors of inhibit, emotional control, working memory and plan-organize; and (b) parental style dimensions of corporal punishment, inconsistent discipline, and involvement. During the modeling process, the BRIE-P working memory and plan-organize scores were excluded since these measures obtained high correlations with the other two BRIEF-P scales considered for the SEM and entailed worse or lack of fitting. The APQ involvement scale score was also excluded of the final SEM because it worsened goodness-of-fit. ###### **Correlations between the variables of the study**.   **2** **3** **4** **5** **6** **7** **8** **9** **10** **11** **12** **13** ----------- ------------------------- --------------------------------------- --------------------------------------- --------------------------------------- --------------------------------------- --------------------------------------- --------------------------------------- ---------------------------------------- --------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- **Girls** 1 SDQ-ODD score 0.17[^\*^](#TN1){ref-type="table-fn"} −0.10 0.10 0.11 0.14[^\*^](#TN1){ref-type="table-fn"} −0.07 −0.18[^\*^](#TN1){ref-type="table-fn"} 0.03 0.18[^\*^](#TN1){ref-type="table-fn"} 0.22[^\*^](#TN1){ref-type="table-fn"} 0.20[^\*^](#TN1){ref-type="table-fn"} 0.28[^\*^](#TN1){ref-type="table-fn"} 2 BRIEF-Inhibit 0.24[^\*^](#TN1){ref-type="table-fn"} 0.54[^\*^](#TN1){ref-type="table-fn"} 0.59[^\*^](#TN1){ref-type="table-fn"} 0.57[^\*^](#TN1){ref-type="table-fn"} 0.06 −0.01 0.02 0.03 0.07 0.17[^\*^](#TN1){ref-type="table-fn"} 0.27[^\*^](#TN1){ref-type="table-fn"} 3 BRIEF-Shift 0.56[^\*^](#TN1){ref-type="table-fn"} 0.35[^\*^](#TN1){ref-type="table-fn"} 0.37[^\*^](#TN1){ref-type="table-fn"} 0.01 0.05 0.02 −0.05 −0.05 −0.02 −0.06 4 BRIEF-Emotional Control 0.42[^\*^](#TN1){ref-type="table-fn"} 0.42[^\*^](#TN1){ref-type="table-fn"} −0.04 −0.07 −0.03 0.04 0.04 0.01 0.01 5 BRIEF-Working memory 0.89[^\*^](#TN1){ref-type="table-fn"} 0.13[^\*^](#TN1){ref-type="table-fn"} 0.05 −0.04 0.03 0.03 0.16[^\*^](#TN1){ref-type="table-fn"} 0.18[^\*^](#TN1){ref-type="table-fn"} 6 BRIEF-Plan/organize .11 0.01 −0.04 0.02 0.02 0.09 0.14[^\*^](#TN1){ref-type="table-fn"} 7 APQ-Positive 0.72[^\*^](#TN1){ref-type="table-fn"} 0.11 −0.08 −0.35[^\*^](#TN1){ref-type="table-fn"} 0.09 0.05 8 APQ-Involvement 0.14[^\*^](#TN1){ref-type="table-fn"} −0.12[^\*^](#TN1){ref-type="table-fn"} −0.39[^\*^](#TN1){ref-type="table-fn"} −0.10 −0.04 9 APQ-Monitoring 0.06 0.08 0.10 0.04 10 APQ-Corporal 0.18[^\*^](#TN1){ref-type="table-fn"} 0.14[^\*^](#TN1){ref-type="table-fn"} 0.12[^\*^](#TN1){ref-type="table-fn"} 11 APQ-Inconsistent 0.14[^\*^](#TN1){ref-type="table-fn"} 0.13[^\*^](#TN1){ref-type="table-fn"} 12 Socioeconomic status 0.33[^\*^](#TN1){ref-type="table-fn"} 13 SDQ-ADHD *(covariate)* **BOYS** 1 SDQ-ODD score 0.21[^\*^](#TN1){ref-type="table-fn"} 0.09 0.20[^\*^](#TN1){ref-type="table-fn"} 0.21[^\*^](#TN1){ref-type="table-fn"} 0.19[^\*^](#TN1){ref-type="table-fn"} −0.09 −0.13[^\*^](#TN1){ref-type="table-fn"} 0.10 0.10 0.10 0.15[^\*^](#TN1){ref-type="table-fn"} 0.31[^\*^](#TN1){ref-type="table-fn"} 2 BRIEF-Inhibit 0.21[^\*^](#TN1){ref-type="table-fn"} 0.54[^\*^](#TN1){ref-type="table-fn"} 0.65[^\*^](#TN1){ref-type="table-fn"} 0.61[^\*^](#TN1){ref-type="table-fn"} 0.02 −0.05 0.10 0.08 −0.03 0.01 0.38[^\*^](#TN1){ref-type="table-fn"} 3 BRIEF-Shift 0.55[^\*^](#TN1){ref-type="table-fn"} 0.42[^\*^](#TN1){ref-type="table-fn"} 0.46[^\*^](#TN1){ref-type="table-fn"} 0.14[^\*^](#TN1){ref-type="table-fn"} 0.03 0.18[^\*^](#TN1){ref-type="table-fn"} −0.02 0.00 0.06 −0.04 4 BRIEF-Emotional Control 0.41[^\*^](#TN1){ref-type="table-fn"} 0.44[^\*^](#TN1){ref-type="table-fn"} 0.06 −0.06 0.11 0.06 −0.08 −0.01 0.13[^\*^](#TN1){ref-type="table-fn"} 5 BRIEF-Working memory 0.90[^\*^](#TN1){ref-type="table-fn"} 0.09 −0.03 0.10 0.03 0.02 0.11 0.34[^\*^](#TN1){ref-type="table-fn"} 6 BRIEF-Plan/organize 0.08 −0.03 0.11 0.08 0.04 0.06 0.30[^\*^](#TN1){ref-type="table-fn"} 7 APQ-Positive 0.67[^\*^](#TN1){ref-type="table-fn"} 0.06 −0.01 −0.27[^\*^](#TN1){ref-type="table-fn"} 0.02 −0.05 8 APQ-Involvement −0.08 −0.19[^\*^](#TN1){ref-type="table-fn"} −0.36[^\*^](#TN1){ref-type="table-fn"} −0.18[^\*^](#TN1){ref-type="table-fn"} −0.13[^\*^](#TN1){ref-type="table-fn"} 9 APQ-Monitoring 0.17[^\*^](#TN1){ref-type="table-fn"} 0.22[^\*^](#TN1){ref-type="table-fn"} 0.12[^\*^](#TN1){ref-type="table-fn"} 0.10 10 APQ-Corporal 0.11 0.05 0.04 11 APQ-Inconsistent 0.12[^\*^](#TN1){ref-type="table-fn"} 0.16[^\*^](#TN1){ref-type="table-fn"} 12 Socioeconomic status 0.19[^\*^](#TN1){ref-type="table-fn"} 13 SDQ-ADHD *(covariate)* BRIEF, Behavior Rating Inventory of Executive Functioning--Preschool version; SES, Socioeconomic Status; APQ, Alabama Parenting Questionnaire; SDQ, Strengths and Difficulties Questionnaire. Significant correlation (0.05 level). Figure [1](#F1){ref-type="fig"} shows the final SEM measuring the underlying process between SES, EF, parenting style and ODD at age 3, adjusted by SDQ^3−4^-ADHD score (Table [3](#T3){ref-type="table"} details the structural coefficients). Goodness-of-fit of the new model was achieved: χ^2^ = 1.835 (*p* = 0.399), RMSEA = 0.001, CFI = 0.999, TLI = 0.999, and SRMR = 0.012. To assess the potential differences in the SEM due to gender, measurement invariance across the groups has been tested by comparing the previous unconstrained model with a new model in which loadings and intercepts were constrained to be equal across boys and girls. Since the χ^2^ difference statistic reveals a significant difference between models (χ^2^ = 54.02, *p* \< 0.001), a lack of invariance across gender is assumed. ###### **Structural coefficients of the final SEM diagram (*N* = 604)**. **Coefficient** **SE** ***z*** ***p*** **95% CI (coefficient)** ------------------- ------------------ ----------------- --------- --------- --------- -------------------------- --------- -------- APQ_Corporal Socioecon.Status *Girls* 0.1347 0.0604 2.23 0.026 0.0164 0.2530 *Boys* 0.0499 0.0540 0.92 0.355 −0.0559 0.1557 APQ_Inconsistent Socioecon.Status *Girls* 0.1309 0.0642 2.04 0.042 0.0050 0.2568 *Boys* 0.1096 0.0602 1.82 0.049 0.0083 0.2275 BRIEF_Inhibit APQ\_ Corporal *Girls* −0.0358 0.0549 −0.65 0.515 −0.1433 0.0718 *Boys* 0.0639 0.0483 1.32 0.186 −0.0308 0.1587 APQ_Inconsistent *Girls* 0.0542 0.0481 1.13 0.260 −0.0400 0.1485 *Boys* −0.0310 0.0581 −0.53 0.594 −0.1449 0.0829 Socioecon.Status *Girls* 0.1762 0.0615 2.86 0.004 0.0556 0.2968 *Boys* 0.0006 0.0606 0.01 0.992 −0.1183 0.1194 BRIEF_EmotControl APQ_Corporal *Girls* 0.0351 0.0623 0.56 0.573 −0.0870 0.1571 *Boys* 0.0621 0.0561 1.11 0.268 −0.0478 0.1720 APQ_Inconsistent *Girls* 0.0402 0.0544 0.74 0.461 −0.0665 0.1469 *Boys* −0.0891 0.0573 −1.55 0.120 −0.2014 0.0232 Socioecon.Status *Girls* −0.0018 0.0629 −0.03 0.978 −0.1251 0.1216 *Boys* 0.0013 0.0628 0.02 0.984 −0.1218 0.1243 SDQ_ODD APQ_Corporal *Girls* 0.1128 0.0542 2.08 0.037 0.0066 0.2190 *Boys* 0.0553 0.0407 1.36 0.174 −0.0244 0.1350 APQ_Inconsistent *Girls* 0.1658 0.0600 2.76 0.006 0.0481 0.2835 *Boys* 0.0710 0.0503 1.41 0.158 −0.0276 0.1697 BRIEF_Inhibit *Girls* 0.1246 0.0658 1.89 0.058 −0.0045 0.2536 *Boys* 0.1225 0.0715 1.71 0.047 0.0176 0.2626 BRIEF_EmotCont *Girls* 0.0225 0.0601 0.37 0.708 −0.0953 0.1403 *Boys* 0.1343 0.0724 1.85 0.044 0.0076 0.2761 Socioecon.Status *Girls* 0.1302 0.0540 2.41 0.016 0.0244 0.2360 *Boys* 0.1488 0.0544 2.74 0.006 0.0423 0.2554 SES, Socioeconomic Status; APQ, Alabama Parenting Questionnaire; BRIEF, Behavior Rating Inventory of Executive Functioning--Preschool version; SDQ, Strengths and Difficulties Questionnaire. {#F1} The pathway diagram for girls showed that high ODD level was directly predicted by lower SES levels, high scores in the parenting style variables of corporal punishment and inconsistence, and high scores in the EF measure of inhibition. Parenting style measures were potential partial mediator variables between SES and ODD: lower SES predicted higher scores in corporal punishment and inconsistence, and higher scores in both parenting scales were related to higher ODD level. The BRIEF-P-inhibit score was also a potential partial mediating factor in the pathway between SES and ODD: lower SES related to greater impairments in inhibition and higher impaired inhibition related to higher ODD level. No statistical association was found between SES and the BRIEF-P-emotional control score, nor between emotional control and ODD level. EF measures contemplated in the pathway were not mediating factors between parenting style and ODD level. The global predictive capacity of the girls\' structural pathway model was *CD* = 0.15. Structural coefficients for boys showed few statistical associations. Higher ODD levels were predicted by lower SES and higher scores on the EF inhibition and emotional control scales. Parenting style measured by corporal punishment and inconsistence dimensions did not influence EF or ODD levels. The global predictive capacity of the boys\' structural model was lower than for girls (*CD* = 0.06). Table [4](#T4){ref-type="table"} contains the global mediation tests for the complete pathway showed in Figure [1](#F1){ref-type="fig"} differentiating between direct and indirect effects for the dependent variable ODD level. Results show that the socioeconomic status achieved a significant direct effect (*z* = 2.40, *p* = 0.017) and also a significant indirect effect (*z* = 2.60, *p* = 0.009) on the ODD score for the girls\' pathway, which suggest that the relationship between SES and ODD level is partially mediated by parenting and/or EF. To identify what were the specific mediators of the ODD levels, individual-specific mediation tests were carried into the girl\'s model. Results showed that EF variables (inhibit and emotional control dimensions) did not mediated into the relationships between the parenting style variables (corporal punishment and inconsistent discipline) and ODD (that is, non-mediation mechanism was found for the paths APQ-\>BRIEF-\>ODD). However, the parenting style variables obtained significant direct effects (punishment: *z* = 2.95, *p* = 0.003; inconsistence: *z* = 3.14, *p* = 0.002) and quasi-significant indirect effects (punishment: *z* = 1.74, *p* = 0.082; inconsistence: *z* = 1.75, *p* = 0.080) into the relationships between SES-\>APQ-\>ODD. The BRIEF inhibit score also achieved a significant direct effect (*z* = 2.66, *p* = 0.008) and a significant indirect effect (*z* = 1.96, *p* = 0.050) into the relationship between SES-\>BRIEF-\>ODD. ###### **Sobel mediation tests for the global SEM: direct and indirect effects of APQ and BRIEF on the dependent variable ODD level**. **Global tests** **Direct effects** **Indirect effects** **Total effects** ---------------------- -------------------- ---------------------- ------------------- ------- -------- ------- ----------------------------------- ----------------------------------- ----------------------------------- ----------------------------------- ----------------------------------- ----------------------------------- ------- ------- ------ ------- --------- ------- **APQ-corporal** Girls 0.246 0.117 2.11 0.035 0.018 0.475 −0.008 0.017 −0.48 0.629 −0.040 0.024 0.238 0.116 2.05 0.040 0.011 0.465 Boys 0.093 0.069 1.35 0.178 −0.042 0.227 0.027 0.020 1.35 0.176 −0.012 0.066 0.120 0.072 1.66 0.096 −0.021 0.260 **APQ-inconsistent** Girls 0.159 0.056 2.83 0.005 0.049 0.270 0.007 0.006 1.14 0.255 −0.005 0.020 0.167 0.057 2.93 0.003 0.055 0.279 Boys 0.062 0.044 1.42 0.156 −0.024 0.149 −0.014 0.011 −1.22 0.221 −0.036 0.008 0.049 0.045 1.08 0.280 −0.0040 0.137 **BRIEF-inhibit** Girls 0.042 0.022 1.91 0.056 −0.001 0.085 [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} 0.042 0.022 1.91 0.058 −0.001 0.085 Boys 0.035 0.021 1.70 0.089 −0.005 0.076 [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} 0.035 0.021 1.70 0.047 0.005 0.076 **BRIEF-emot.cont**. Girls 0.014 0.038 0.38 0.708 −0.060 0.089 [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} 0.014 0.038 0.38 0.708 −0.060 0.089 Boys 0.078 0.042 1.85 0.065 −0.005 0.160 [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} [^\*^](#TN2){ref-type="table-fn"} 0.078 0.042 1.85 0.044 0.005 0.160 **SES** Girls 0.232 0.097 2.40 0.017 0.042 0.422 0.106 0.041 2.60 0.009 0.026 0.186 0.338 0.099 3.40 0.001 0.143 0.533 Boys 0.262 0.098 2.67 0.008 0.070 0.455 0.017 0.029 0.59 0.553 −0.040 0.075 0.280 0.104 2.70 0.007 0.077 0.483 Not applicable. Discussion {#s4} ========== Final SEM analyses showed that high ODD symptom levels were predicted by low SES levels, high scores in parenting behaviors characterized by corporal punishment and inconsistent discipline, and high scores in EF inhibit and emotional control dimensions. Children\'s gender differences emerged in the pathway: (a) the ODD level score for boys was directly predicted by SES and EF inhibit and emotional control dimensions, while parenting scores did not obtain significant associations with any other variables analyzed; (b) the ODD level score for girls was directly predicted by SES, parenting practices (corporal punishment and inconsistent scores) and EF inhibit scale, the parenting practices analyzed obtained quasi-significant indirect effects into the relationship between SES and ODD, and EF inhibit dimension achieved a significant indirect effect into the relationship between SES and ODD. These results suggest that parenting style (defined by the dimensions corporal punishment and consistence) and EF inhibit level should achieve a partial mediation role between SES and ODD for preschooler girls from the general population. The direct association between lower SES and ODD symptomatology is congruent with the literature (Brooks-Gunn and Duncan, [@B7]; Mistry et al., [@B39]). In the final SEM model of this work, boys are slightly more sensitive to the effect of lower SES, showing no influence of parenting practices, often associated with SES. This may be due to a combination of boys showing more externalizing behavior in general (Miner and Clarke-Stewart, [@B38]; Searle et al., [@B48]) and parents with low SES perceiving their children exhibiting more at-risk levels of aggression than parents with higher SES (Graves et al., [@B23]). One might suspect that boys from lower SES families would be at greater risk of exhibiting externalizing behavior than girls from lower SES families or boys from higher SES families. However, no significant difference was found between ODD scores for boys and girls in our study, and correlations did not suggest a stronger association between SES and ODD scores for boys. For girls the association between SES and ODD is more complex than for boys, with both negative parenting behaviors and EF deficits acting as separate partially mediators. Lower SES is associated with more corporal punishment and more use of inconsistent discipline, which in turn are both related to higher levels of ODD symptoms. Several studies have reported less maternal warmth and more harsh discipline in families with low SES (Pinderhughes et al., [@B43]; Callahan and Eyberg, [@B9]), and this association has been corroborated in our study for boys considering the inconsistency score. For girls, lower SES was associated not only with inconsistent discipline but also with corporal punishment. A possible explanation for this gender difference could be that boys, based on gender stereotypes, are more likely to receive harsh physical discipline in general (Mahoney et al., [@B36]), regardless of ODD. Parents may still believe that boys need more physical discipline in order to change their behavior than girls do (McKee et al., [@B37]). Therefore, it would be less remarkable that boys in lower SES families experience harsh discipline: boys in families with higher SES experience the same amount. Girls in lower SES families however, endure more harsh discipline than girls from high SES families, resulting in a clear association between lower SES and more corporal punishment. In our study, the mean APQ-corporal score was statistically higher for girls in low SES level compared to girls in high SES (1.19 vs. 0.84, *p* = 0.016), and also for boys compared to girls when the comparison was carried out in higher SES levels (means equal to 1.06 vs. 0.84, *p* = 0.033). Besides a consistent relationship between lower SES and negative parenting practices, harsh parenting practices have been found to have a unique effect on child disruptive behavior, even in early childhood (Duncombe et al., [@B16]; Harvey and Metcalfe, [@B24]). Although the association is not clear in the literature, some studies confirm a greater influence of parenting practices on problem behavior in girls (Javo et al., [@B30]). Kim et al. ([@B33]) found that girls\' externalizing behavior is associated with overreactive parenting, whereas for boys this relation did not occur in their study. The authors attributed this to gender stereotypes as well. With aggressive behavior more acceptable in boys than in girls and shy and dependent behavior more acceptable in girls than in boys, Kim and colleagues offer the possibility that parents tend to be overreactive to behaviors that are incongruent with these gender stereotypes by punishing stereotype-inconsistent behavior. The association between SES and ODD in girls was also partially mediated by the EF aspect inhibition, given the associations between lower SES and deficits in inhibition and between deficits in inhibition and ODD symptoms. The relation between SES and EF seems clear in the literature, with more EF deficits in children from lower SES families (Rhoades et al., [@B46]; Clark et al., [@B11]). The associations between EF inhibit and ODD symptoms in the SEM model are remarkable though, for both boys and girls, since the model is controlled for ADHD. Several studies have found relations between EF and ODD, but as soon as ADHD was taken into account, the associations disappeared (Thorell and Wåhlstedt, [@B51]; Schoemaker et al., [@B47]). For girls, inhibition acted as a partial mediator in the association between SES and ODD symptoms. As mentioned before, girls tend to perform better on inhibition tasks than boys (Kochanska et al., [@B34]; Olson et al., [@B42]). In our study girls scored significantly lower than boys on BRIEF-Inhibit (21.62 vs. 24.40, *p* \< 0.001), indicating fewer deficits in inhibition. Especially in early childhood this could have to do with girls maturing faster than boys. They develop some self-regulatory and communicative skills earlier than boys and are therefore better in regulating their emotions and behavior (Keenan and Shaw, [@B32]; Bornstein et al., [@B5]). With these abilities developed to a further stage than boys at the same age, they develop more positive relationships with peers and parents, communicate their needs more effectively and their needs are therefore met more often. This in turn results in less frustration and fewer conflicts and thus less externalizing behavior (Keenan and Shaw, [@B32]). In contrast, girls who show deficits in the development of inhibitory and regulatory abilities miss the communication advantages and may therefore exhibit more externalizing behavior. The role of EF emotional control in the SEM models is relevant. Emotional control is considered as a hot EF task, which may be more associated with ODD than cool EF (Zelazo and Müller, [@B54]). Finding this relationship in the model for boys supports the hypothesized association between hot EF and ODD. Research on this topic is scarce in preschoolers, and therefore the current finding is important in the research concerning ODD in young children. However, it is present in boys only, and not in girls. The strengths of this work include a large community sample of young preschoolers (3 years old), the analysis of ODD differentiated from ADHD (many studies combine ODD + ADHD, making it hard to distinguish between findings for the different disorders) and the modeling through SEM to obtain a *causal* model of the underlying process of ODD. The main limitations of the study include: (a) the cross-sectional nature of the data, which formally prevents causal interpretations of the pathways analyses in light of potential common reciprocal associations between ODD and parenting; (b) the analysis of data gathered only through questionnaire (the BRIEF-P answered by the teachers and the APQ answered by parents); and (c) the limited range of families\' SES (particularly from low SES categories). This study contributes some insights which may be helpful for understanding the mechanisms by which SES affects the development of ODD and in developing interventions for young children at risk for ODD. The study suggests that SES is a good indicator for selecting children at high risk and it confirms the need to involve parents in preventive and intervention programs. Parents can adjust their behavior toward their child in a more constructive way and perhaps prevent their child\'s behavior from escalating or help to turn it around. Parental involvement seems particularly important for girls. Another important insight of this study is that intervention programs for boys and girls may have different components. Whereas for girls an important focus should be on parental involvement and the development of inhibitory abilities, for boys programs should focus on the development of emotional control as well. Future studies are required in order to increase the knowledge of the underlying process of ODD during preschool age, including new constructs (and their reciprocal relationships) and multi-informant measures. Developmental differences should also be considered in future pathways analyses, to assess if the underlying processes between SES, parenting, EF and ODD could be dependent on children\'s age. In addition, due to the strong association between ODD, CD, and ADHD, it should be also relevant to obtain empirical evidence of the pathways for the CD and the ADHD levels controlling for the presence of ODD. Author contributions {#s5} ==================== All authors have made substantial contributions to the work reported in the manuscript: LE designed the study and was responsible for assessment instruments. LL conducted literature searches and provided summaries of previous research studies. She also wrote the first draft of the manuscript. RG designed and conducted the statistical analysis. All the authors reviewed, contributed to and have approved the final manuscript. Funding sources {#s6} =============== Funding for this study was provided by Spanish Ministry of Science and Innovation PSI2009-07542, the Spanish Ministry of Economy and Competiveness PSI2012-32695 and the Secretaria d′Universitats i Recerca del Departament d′Economia i Coneixement de la Generalitat de Catalunya (2014 SGR 312). These funding sources had no role in the study design, collection, analysis or interpretation of the data, writing the manuscript, or the decision to submit the paper for publication. The terms of this arrangement have been reviewed and approved by the Autonomous University of Barcelona in accordance with its policy on research. Conflict of interest statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. We would like to thank the participating schools and families. [^1]: Edited by: Hwajin Yang, Singapore Management University, USA [^2]: Reviewed by: Ayelet Lahat, McMaster University, Canada; Irene Tung, University of California, Los Angeles, USA [^3]: This article was submitted to Developmental Psychology, a section of the journal Frontiers in Psychology

Introduction {#eji4726-sec-0010} ============ When resting, an average adult inhales about 11 000 L of ambient air per day. The inhaled air is neither sterile nor free of particles/macromolecules, and it potentially contains pathogens as well as allergens. Accordingly, the respiratory system has developed several safety mechanisms (e.g. mucus production by goblet cells, secretion of a plethora of antimicrobials, mucociliary transport by ciliated epithelial cells, coughing, etc.), which all together protect the body from harmful intruders at the alveolar--air interface \[[1](#eji4726-bib-0001){ref-type="ref"}\]. Besides immune cells homing to this important organ, the cells lining the airways, that is airway epithelial cells, significantly contribute to host defence by expressing, amongst several other classes of antimicrobials, a multitude of pattern recognition receptors (PRR), which enables them to quickly respond to dangerous cues. In the literature, PRRs are classically described as cytoplasmic (nucleotide‐binding oligomerization domain‐like receptors/retinoic acid‐inducible gene‐like receptors) or cell surface expressed (i.e. C‐type lectin‐like receptors and Toll‐like receptors) receptors. Less well known is the role of soluble pattern recognition receptors (sPRRs), which represent a group of several other evolutionarily ancient but secreted molecules. Soluble PRRs are key players of the humoral arm of innate immunity \[[2](#eji4726-bib-0002){ref-type="ref"}, [3](#eji4726-bib-0003){ref-type="ref"}\]. Under physiological conditions, these molecules are expressed constitutively at low levels in the liver by hepatocytes \[[4](#eji4726-bib-0004){ref-type="ref"}\] and by a variety of immune cells, but their production rapidly increases in response to pathogens, initiating innate immune responses that modulate mucosal immunity in extracellular compartments such as the bronchial and alveolar airspace \[[2](#eji4726-bib-0002){ref-type="ref"}, [5](#eji4726-bib-0005){ref-type="ref"}, [6](#eji4726-bib-0006){ref-type="ref"}\]. Notably, in the respiratory tract of mice and men, airway epithelial cells (AECs) are a major source of sPRRs (Fig. [1](#eji4726-fig-0001){ref-type="fig"}) and protect this barrier tissue from potential injuries (Table [1](#eji4726-tbl-0001){ref-type="table"}). Moreover, cells of the macrophage lineage \[7\] but also T cells, monocytes, dendritic cells (DCs) and granulocytes contribute to protection of the respiratory tract, by locally secreting sPPRs \[[8](#eji4726-bib-0008){ref-type="ref"}, [9](#eji4726-bib-0009){ref-type="ref"}\] (Table [1](#eji4726-tbl-0001){ref-type="table"}). Soluble PRRs share the capacity to bind various microbial and environmental proteins and eliminate them through common mechanisms including agglutination, neutralization, opsonization followed by phagocytosis, with some of them having the capacity to activate complement \[[10](#eji4726-bib-0010){ref-type="ref"}, [11](#eji4726-bib-0011){ref-type="ref"}, [12](#eji4726-bib-0012){ref-type="ref"}\]. As such, sPRRs possess antibody‐like features and might, in fact, represent their functional ancestors \[[13](#eji4726-bib-0013){ref-type="ref"}\]. In addition, they also interact with and regulate the function of membrane‐bound PRRs, which cooperate at the level of the recognition of environmental patterns and the regulation of inflammation \[[12](#eji4726-bib-0012){ref-type="ref"}\]. Accordingly, innate humoral immune responses at the respiratory mucosa are shaped by the presence of (i) sPRR with the ability to activate complement, that is, the pentraxins PTX3, CRP; the collectin mannose‐binding lectin (MBL); the ficolins ficolin‐2 and ‐3; (ii) sPRR unable to activate complement (SAA) and (iii) complement components themselves (C3). {#eji4726-fig-0001} ###### Cellular source of soluble pattern recognition receptors in the lung +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | sPRRs | Cellular source[a](#eji4726-tbl1-note-0001){ref-type="fn"} | Evidence | Implication in allergic responses | +:================================+:===========================================================+:==========================================================================================================================================================================================+:==========================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================+ | ***Pentraxins*** | | | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | Pentraxin 3 (PTX3) | Airway epithelial cells | Secreted protein and mRNA \[[41](#eji4726-bib-0041){ref-type="ref"}, [66](#eji4726-bib-0066){ref-type="ref"}\] | Increased BALF and serum levels in allergic asthma; levels correlate with disease severity \[[14](#eji4726-bib-0014){ref-type="ref"}, [21](#eji4726-bib-0021){ref-type="ref"}, [22](#eji4726-bib-0022){ref-type="ref"}, [41](#eji4726-bib-0041){ref-type="ref"}, [70](#eji4726-bib-0070){ref-type="ref"}\]. | | | | | | | | Airway smooth muscle cells | Secreted protein and mRNA \[[41](#eji4726-bib-0041){ref-type="ref"}, [66](#eji4726-bib-0066){ref-type="ref"}\] | | | | | | | | | Dendritic cells | Secreted protein and mRNA \[[64](#eji4726-bib-0064){ref-type="ref"}\] | | | | | | | | | Macrophages | Secreted protein and mRNA \[[64](#eji4726-bib-0064){ref-type="ref"}\] | | | | | | | | | Neutrophils | Store and secrete protein but no mRNA expression \[[65](#eji4726-bib-0065){ref-type="ref"}\] | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | C‐reactive protein (CRP) | Airway epithelial cells | Secreted protein from cell lines and primary cells, mRNA \[[105](#eji4726-bib-0105){ref-type="ref"}, [106](#eji4726-bib-0106){ref-type="ref"}, [107](#eji4726-bib-0107){ref-type="ref"}\] | Serum levels significantly elevated in asthmatics; correlate with reduced lung function and eosinophil numbers; associated with allergic sensitization and elevated risk of concomitant allergic airway inflammation \[[15](#eji4726-bib-0015){ref-type="ref"}, [34](#eji4726-bib-0034){ref-type="ref"}, [40](#eji4726-bib-0040){ref-type="ref"}, [98](#eji4726-bib-0098){ref-type="ref"}, [99](#eji4726-bib-0099){ref-type="ref"}, [103](#eji4726-bib-0103){ref-type="ref"}, [104](#eji4726-bib-0104){ref-type="ref"}\]. | | | | | | | | Alveolar macrophages | Secreted protein and mRNA \[[108](#eji4726-bib-0108){ref-type="ref"}\] | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | Serum amyloid P component (SAP) | Liver | | Increased sputum levels in asthmatics \[[14](#eji4726-bib-0014){ref-type="ref"}\] | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | ***Collectins*** | | | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | Mannose Binding Lectin (MBL) | Liver, foetal lung | Secreted protein and mRNA \[[116](#eji4726-bib-0116){ref-type="ref"}\] | Directly binds allergen extracts \[[131](#eji4726-bib-0131){ref-type="ref"}\]; increased plasma levels in asthmatics that correlate with eosinophil numbers \[[132](#eji4726-bib-0132){ref-type="ref"}, [133](#eji4726-bib-0133){ref-type="ref"}\] | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | Surfactant protein A | Alveolar type II cells | Secreted protein and mRNA \[[136](#eji4726-bib-0136){ref-type="ref"}\] | Both SP‐A and SP‐D can bind allergenic components including pollen, house dust mite and *A. fumigatus* \[[150](#eji4726-bib-0150){ref-type="ref"}, [151](#eji4726-bib-0151){ref-type="ref"}, [152](#eji4726-bib-0152){ref-type="ref"}\]; SP‐D serum and BALF increased in allergic asthmatics \[[160](#eji4726-bib-0160){ref-type="ref"}, [161](#eji4726-bib-0161){ref-type="ref"}\]. | | | | | | | (SP‐A) | club cells | Secreted protein and mRNA \[[136](#eji4726-bib-0136){ref-type="ref"}\] | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | Surfactant protein D | Alveolar type II cells | Secreted protein and mRNA \[[137](#eji4726-bib-0137){ref-type="ref"}\] | | | | | | | | (SP‐D) | club cells | Secreted protein and mRNA \[[138](#eji4726-bib-0138){ref-type="ref"}\] | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | ***Ficolins*** | | | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | L ficolins (ficolin‐2) | Liver | Deficiency is associated with allergic and infectious diseases \[[179](#eji4726-bib-0179){ref-type="ref"}\] | Direct involvement in allergic inflammation currently unknown. | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | H ficolins (ficolin‐3) | Airway epithelial cells | Secreted protein and mRNA \[[167](#eji4726-bib-0167){ref-type="ref"}\] | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | ***Serum Amyloid A*** | | | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | Serum amyloid A (SAA) | Alveolar epithelial lining | Secreted protein and mRNA \[[181](#eji4726-bib-0181){ref-type="ref"}\] | Elevated sputum SAA levels correlate with CRS and asthma prevalence and disease severity \[[23](#eji4726-bib-0023){ref-type="ref"}, [24](#eji4726-bib-0024){ref-type="ref"}, [212](#eji4726-bib-0212){ref-type="ref"}\] | | | | | | | | Alveolar macrophages | Secreted protein and mRNA \[[181](#eji4726-bib-0181){ref-type="ref"}\] | | | | | | | | | monocytes | secreted protein and mRNA \[[182](#eji4726-bib-0182){ref-type="ref"}\] | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | ***Complement components*** | | | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | C3 | Airway epithelial cells | Secreted protein from cell lines and primary cells \[[226](#eji4726-bib-0226){ref-type="ref"}, [227](#eji4726-bib-0227){ref-type="ref"}, [237](#eji4726-bib-0237){ref-type="ref"}\] | Allergens possess complement convertase activity and lead to C3a and C5a formation \[[243](#eji4726-bib-0243){ref-type="ref"}\]; C3a levels are found upregulated in individuals with respiratory Allergies \[[250](#eji4726-bib-0250){ref-type="ref"}, [251](#eji4726-bib-0251){ref-type="ref"}\]; positive correlation between C3 levels and asthma \[[252](#eji4726-bib-0252){ref-type="ref"}\] | | | | | | | | Dendritic cells | Secreted protein and mRNA \[[241](#eji4726-bib-0241){ref-type="ref"}\] | | | | | | | | | Monocytes | Secreted protein and mRNA \[[241](#eji4726-bib-0241){ref-type="ref"}\] | | | | | | | | | Neutrophils | Secreted protein and mRNA \[[241](#eji4726-bib-0241){ref-type="ref"}\] | | | | | | | | | T cells | Secreted protein and mRNA \[[241](#eji4726-bib-0241){ref-type="ref"}\] intracellularly stored protein \[[8](#eji4726-bib-0008){ref-type="ref"}\] | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | C5 | Airway epithelial cells | Secreted protein from cell lines \[[228](#eji4726-bib-0228){ref-type="ref"}, [237](#eji4726-bib-0237){ref-type="ref"}\] | | | | | | | | | Dendritic cells | Secreted protein and mRNA \[[241](#eji4726-bib-0241){ref-type="ref"}\] | | | | | | | | | T cells | Secreted protein and mRNA \[[241](#eji4726-bib-0241){ref-type="ref"}\] intracellularly stored protein \[[8](#eji4726-bib-0008){ref-type="ref"}\] | | +---------------------------------+------------------------------------------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ Different cells are listed in alphabetical order. John Wiley & Sons, Ltd. Due to their dramatic systemic upregulation upon infection (i.e. by 500‐ to 1000‐fold), sPRRs have been termed 'inflammation markers', with some of them being extremely valuable in daily clinical practice (e.g. CRP, high‐sensitivity CRP and SAA) \[[14](#eji4726-bib-0014){ref-type="ref"}, [15](#eji4726-bib-0015){ref-type="ref"}, [16](#eji4726-bib-0016){ref-type="ref"}\]. Airway mucosal inflammation in response to foreign antigens is an essential effector arm of innate host defence; however, the involvement of sPRRs in the regulation of the intensity and persistence of airway inflammation remains poorly understood. For instance, some of these sPRR, such as the collectin MBL \[[17](#eji4726-bib-0017){ref-type="ref"}\] and the pentraxin PTX3, seem to inhibit, while others, like complement \[[18](#eji4726-bib-0018){ref-type="ref"}, [19](#eji4726-bib-0019){ref-type="ref"}\], appear to aggravate infectious lung diseases triggered by deadly coronaviruses well known to induce severe acute respiratory syndrome and middle east respiratory syndrome \[[20](#eji4726-bib-0020){ref-type="ref"}\]. Moreover, recent evidence suggests that sPRRs might also be causally involved in allergic airway inflammation \[[21](#eji4726-bib-0021){ref-type="ref"}, [22](#eji4726-bib-0022){ref-type="ref"}, [23](#eji4726-bib-0023){ref-type="ref"}, [24](#eji4726-bib-0024){ref-type="ref"}, [25](#eji4726-bib-0025){ref-type="ref"}\]. In the following, we review the function of sPRRs present in the AEC mucus lining of mice and men and discuss their potential involvement in allergic inflammation in the lungs (Table [1](#eji4726-tbl-0001){ref-type="table"}). Pentraxins {#eji4726-sec-0020} ========== Pentraxins belong to a superfamily of cyclic multimeric sPRRs (Fig. [2](#eji4726-fig-0002){ref-type="fig"}) that can be divided into short and long pentraxins based on their primary structure \[[11](#eji4726-bib-0011){ref-type="ref"}, [12](#eji4726-bib-0012){ref-type="ref"}, [13](#eji4726-bib-0013){ref-type="ref"}, [26](#eji4726-bib-0026){ref-type="ref"}, [27](#eji4726-bib-0027){ref-type="ref"}\]. Prototypes of short pentraxins are the serum amyloid P component (SAP) and C‐reactive protein (CRP) \[[26](#eji4726-bib-0026){ref-type="ref"}\] that constitute the main acute phase proteins in mouse and human, respectively \[[27](#eji4726-bib-0027){ref-type="ref"}\]. Short pentraxins show a pentameric radial symmetry consisting of five or ten identical subunits of 25 kDa in size \[[13](#eji4726-bib-0013){ref-type="ref"}, [28](#eji4726-bib-0028){ref-type="ref"}, [29](#eji4726-bib-0029){ref-type="ref"}, [30](#eji4726-bib-0030){ref-type="ref"}\]. While CRP and SAP are mainly produced systemically in the liver in response to inflammatory stimuli such as IL‐6, pentraxin 3 (PTX3), the most commonly known long pentraxin, is induced locally at sites of inflammation or cell damage where it plays a similar role as CRP and SAP in the circulation \[[11](#eji4726-bib-0011){ref-type="ref"}, [12](#eji4726-bib-0012){ref-type="ref"}, [31](#eji4726-bib-0031){ref-type="ref"}\]. In contrast to CRP, measurement of systemic levels of PTX3 seems to be a much faster (peak levels reached at 7.5 h versus 24 h after hospital admission, respectively) and more reliable (faster regression to normal values) marker for acute myocardial infarction \[[32](#eji4726-bib-0032){ref-type="ref"}, [33](#eji4726-bib-0033){ref-type="ref"}\]. Increasing evidence now also suggests that the expression of pentraxins is increased in asthmatic patients, contributing to allergic airway inflammation \[[14](#eji4726-bib-0014){ref-type="ref"}, [21](#eji4726-bib-0021){ref-type="ref"}, [22](#eji4726-bib-0022){ref-type="ref"}, [28](#eji4726-bib-0028){ref-type="ref"}, [34](#eji4726-bib-0034){ref-type="ref"}, [35](#eji4726-bib-0035){ref-type="ref"}, [36](#eji4726-bib-0036){ref-type="ref"}, [37](#eji4726-bib-0037){ref-type="ref"}, [38](#eji4726-bib-0038){ref-type="ref"}, [39](#eji4726-bib-0039){ref-type="ref"}, [40](#eji4726-bib-0040){ref-type="ref"}, [41](#eji4726-bib-0041){ref-type="ref"}\]. Accordingly, pentraxins can be regarded as markers for (allergic) inflammation in the airways. {#eji4726-fig-0002} Pentraxin 3 (PTX3) {#eji4726-sec-0030} ------------------ ### PTX3: structure/function relationship {#eji4726-sec-0040} Thirty years ago, PTX3 was identified as a member of the pentraxin family and is now classified as the prototypic long pentraxin \[[42](#eji4726-bib-0042){ref-type="ref"}, [43](#eji4726-bib-0043){ref-type="ref"}\]. PTX3 exists as a homo‐oligomer formed by eight identical, covalently linked subunits \[[44](#eji4726-bib-0044){ref-type="ref"}\]. Protein oligomerization seems to be essential for PTX3 activity as other oligomeric forms of PTX3 have decreased functionality \[[45](#eji4726-bib-0045){ref-type="ref"}\]. In addition, PTX3 function can be regulated by elastase and *Aspergillus fumigatus* proteases \[[46](#eji4726-bib-0046){ref-type="ref"}\]. This indicates that PTX3 is capable to serve pleiotropic functions depending on its oligomeric and proteolytic state. As a sPRR, PTX3 has opsonic activity with a nonredundant protective role against selected pathogens, such as fungi, bacteria and viruses \[[2](#eji4726-bib-0002){ref-type="ref"}, [10](#eji4726-bib-0010){ref-type="ref"}, [13](#eji4726-bib-0013){ref-type="ref"}, [47](#eji4726-bib-0047){ref-type="ref"}\]. PTX3 binds with high affinity to the outer membrane protein A from *Klebsiella pneumoniae* (KpOmpA) but does not interact with cellular receptors that bind this bacterial molecule. PTX3^−/−^ mice show reduced local inflammation in response to KpOmpA, indicating that PTX3 might amplify KpOmpA‐induced responses \[[48](#eji4726-bib-0048){ref-type="ref"}\]. ### PTX3: epidemiologic data‐lung immunity {#eji4726-sec-0050} Most recently, PTX3 was shown to be a critical positive regulator for inflammatory responses during pneumococcal pneumonia \[[49](#eji4726-bib-0049){ref-type="ref"}\]. While PTX3 deficiency renders mice more susceptible to pulmonary aspergillosis due to a defective recognition of conidia by antigen presenting cells and an inappropriate induction of an adaptive type 2 response \[[50](#eji4726-bib-0050){ref-type="ref"}\], administration of exogenous PTX3 is able to rescue antifungal resistance early in infection \[[51](#eji4726-bib-0051){ref-type="ref"}\]. Similarly, transgenic overexpression of PTX3 is responsible for resistance to viruses (e.g. influenza virus, cytomegalovirus), pathogenic (e.g. *Neisseria meningitidis*) and opportunistic (e.g. *Klebsiella pneumoniae* and *Pseudomonas aeruginosa*) bacteria, including preclinical models of bacterial sepsis induced by cecal ligation puncture \[[52](#eji4726-bib-0052){ref-type="ref"}, [53](#eji4726-bib-0053){ref-type="ref"}, [54](#eji4726-bib-0054){ref-type="ref"}\]. PTX3 also takes part in extracellular matrix formation \[[45](#eji4726-bib-0045){ref-type="ref"}, [47](#eji4726-bib-0047){ref-type="ref"}\] and interacts with recognition molecules of the classical and lectin complement pathway, Fcγ receptor, fibroblast growth factor 2 and P‐selectin \[[10](#eji4726-bib-0010){ref-type="ref"}, [50](#eji4726-bib-0050){ref-type="ref"}, [55](#eji4726-bib-0055){ref-type="ref"}, [56](#eji4726-bib-0056){ref-type="ref"}\]. A number of clinical studies have pointed to the important role of PTX3 in lung and bone marrow transplant recipients \[[57](#eji4726-bib-0057){ref-type="ref"}, [58](#eji4726-bib-0058){ref-type="ref"}, [59](#eji4726-bib-0059){ref-type="ref"}, [60](#eji4726-bib-0060){ref-type="ref"}, [61](#eji4726-bib-0061){ref-type="ref"}, [62](#eji4726-bib-0062){ref-type="ref"}\]. In fact, patients with elevated pretransplant PTX3 plasma levels are more likely to suffer from ischemia reperfusion injury leading to primary graft dysfunction after lung transplantation \[[57](#eji4726-bib-0057){ref-type="ref"}, [58](#eji4726-bib-0058){ref-type="ref"}, [59](#eji4726-bib-0059){ref-type="ref"}\]. In contrast, genetic deficiency of PTX3 predisposes to invasive lung aspergillosis in patients who underwent bone marrow transplantation \[[61](#eji4726-bib-0061){ref-type="ref"}\]. Elevated bronchoalveolar lavage fluid (BALF) PTX3 levels might be indicative of invasive aspergillosis in lung transplant patients \[[62](#eji4726-bib-0062){ref-type="ref"}\] and represent a more reliable biomarker than determination of galactomannan, which gives a high rate of false positive results \[[60](#eji4726-bib-0060){ref-type="ref"}\]. Moreover, serum PTX3 levels were found to be higher in patients with lung cancer as compared to cancer‐free heavy smokers \[[63](#eji4726-bib-0063){ref-type="ref"}\]. In the airways, several innate immune cells have been found to produce PTX3 locally including macrophages, DCs \[[64](#eji4726-bib-0064){ref-type="ref"}\] and neutrophils that store and secrete PTX3 as part of neutrophil extracellular traps \[[65](#eji4726-bib-0065){ref-type="ref"}\], highlighting its critical role in the humoral arm of innate immunity. Besides infiltrating inflammatory cells, also structural cells, particularly bronchial and alveolar epithelial cells \[[66](#eji4726-bib-0066){ref-type="ref"}, [67](#eji4726-bib-0067){ref-type="ref"}\] and smooth muscle cells, are main sites of PTX3 expression, which is absent in resting and activated T or B lymphocytes and NK cells \[[41](#eji4726-bib-0041){ref-type="ref"}, [68](#eji4726-bib-0068){ref-type="ref"}, [69](#eji4726-bib-0069){ref-type="ref"}\]. ### PTX: role in allergic diseases {#eji4726-sec-0060} In human patients, PTX3 levels in BALF and sputum significantly correlate with disease severity \[[21](#eji4726-bib-0021){ref-type="ref"}, [22](#eji4726-bib-0022){ref-type="ref"}\] and have been associated with allergic asthma in adults \[[14](#eji4726-bib-0014){ref-type="ref"}, [21](#eji4726-bib-0021){ref-type="ref"}, [22](#eji4726-bib-0022){ref-type="ref"}, [41](#eji4726-bib-0041){ref-type="ref"}, [70](#eji4726-bib-0070){ref-type="ref"}\] and children \[[14](#eji4726-bib-0014){ref-type="ref"}\]. Moreover, patients with allergic asthma show significantly higher PTX3 expression in airway smooth muscle cells as compared to healthy donors \[[41](#eji4726-bib-0041){ref-type="ref"}\]. PTX3 enhances smooth muscle cell eotaxin‐1 release, which contributes to eosinophilic airway inflammation and airway remodelling \[[41](#eji4726-bib-0041){ref-type="ref"}\]. In children, a positive correlation was identified between high sputum PTX3 levels and decreased pulmonary function. Similarly, significant positive correlations were identified between PTX3 sputum concentrations and eosinophil counts in sputum and blood as well as total serum immunoglobulin E (IgE) levels \[[22](#eji4726-bib-0022){ref-type="ref"}\]. These findings have been confirmed recently in an independent study in a group of over 100 children with allergic asthma \[[71](#eji4726-bib-0071){ref-type="ref"}\]. Notably, there is also a role for PTX3 in children with allergic rhinitis \[[37](#eji4726-bib-0037){ref-type="ref"}\]. Children monosensitized to seasonal or perennial allergens and those with allergic rhinitis presented higher PTX3 plasma levels which also correlated with symptom severity \[[37](#eji4726-bib-0037){ref-type="ref"}\]. Interestingly, the human PTX3 gene is located on chromosome 3q24--28 \[[42](#eji4726-bib-0042){ref-type="ref"}\], a region associated with sensitization to Der p 1, a major allergen from house dust mites, and total serum IgE levels in humans \[[72](#eji4726-bib-0072){ref-type="ref"}\]. A connection between PTX3 and type 2 immune responses was also established in an experimental model of ovalbumin‐induced allergic asthma in mice \[[21](#eji4726-bib-0021){ref-type="ref"}\]. PTX3‐deficient mice exhibited enhanced type 2 inflammation in the lungs after ovalbumin sensitization and challenge with increased airway hyperresponsiveness, mucus plugging of the airways as well as BALF and lung eosinophilia \[[21](#eji4726-bib-0021){ref-type="ref"}\]. The allergic phenotype was associated with a Th17‐dominant inflammatory response with an enhanced recruitment of granulocytes, particularly eosinophils and neutrophils, and increased IgE/IgG2a secretion \[[21](#eji4726-bib-0021){ref-type="ref"}\]. Mechanistically, the Th17‐dominant allergic response in these mice was supported by the Th17‐polarizing cytokines IL‐6 and IL‐23 from DCs \[[21](#eji4726-bib-0021){ref-type="ref"}\]. In summary, serum PTX3 levels might be a valuable marker to demonstrate the presence of low‐grade inflammation in asthma \[[71](#eji4726-bib-0071){ref-type="ref"}\]. However, serum PTX3 levels seem to be inadequate as marker for asthma management, and it remains a matter of debate whether PTX3 might be directly involved in the etiology of allergic airway inflammation. C‐reactive protein {#eji4726-sec-0070} ------------------ CRP has been named for its ability to bind and precipitate the 'C' polysaccharide contained in the pneumococcal cell wall in a Ca^2+^‐dependent manner \[[73](#eji4726-bib-0073){ref-type="ref"}\]. Later on, this antigen has been identified as phosphocholine and related molecules on other microorganisms \[[74](#eji4726-bib-0074){ref-type="ref"}\] but CRP also binds to phosphocholine exposed on the surface of damaged cells \[[75](#eji4726-bib-0075){ref-type="ref"}\]. This involves CRP in the clearance of apoptotic cells \[[75](#eji4726-bib-0075){ref-type="ref"}\]. Additional CRP ligands have been identified in recent years including nuclear antigens such as histones and chromatin as well as proteins from small nuclear RNP particles, fibronectin, laminin and polycations \[[76](#eji4726-bib-0076){ref-type="ref"}\], which points to an important regulatory role of CRP in the development of autoimmunity \[[77](#eji4726-bib-0077){ref-type="ref"}\]. CRP is an ideal marker for acute inflammation, showing a more than 1000‐fold increase in concentration under inflammatory conditions \[[75](#eji4726-bib-0075){ref-type="ref"}\]. In recent years, more sensitive assays have been developed and high‐sensitivity CRP determination is considered as an indicator of low‐grade systemic inflammation that may be a useful predictor for the risk of atherosclerosis \[[78](#eji4726-bib-0078){ref-type="ref"}\] and myocardial infarction \[[79](#eji4726-bib-0079){ref-type="ref"}\]. It can also serve as an indicator of airway inflammation for many noncommunicable lung diseases including asthma \[[14](#eji4726-bib-0014){ref-type="ref"}, [34](#eji4726-bib-0034){ref-type="ref"}, [39](#eji4726-bib-0039){ref-type="ref"}, [40](#eji4726-bib-0040){ref-type="ref"}, [80](#eji4726-bib-0080){ref-type="ref"}\]. ### CRP: structure/function relationship {#eji4726-sec-0080} The strongest activator for CRP transcription is interleukin‐6 \[[81](#eji4726-bib-0081){ref-type="ref"}\], which requires the synergistic interaction of CCAAT/enhancer‐binding protein (C/EBP)‐beta and Rel p50 \[[82](#eji4726-bib-0082){ref-type="ref"}\]. By contrast, interferon α significantly inhibits IL‐6‐induced CRP expression in hepatocytes \[[83](#eji4726-bib-0083){ref-type="ref"}\], explaining why viral infections cause only modest elevations of CRP. Notably, the small molecule hypolipidemic drug Gemcabene not only lowers cholesterol but also CRP levels alone and synergistically with statins \[[84](#eji4726-bib-0084){ref-type="ref"}\], and this is accomplished by interference with C/EBP‐delta and NF‐kB \[[85](#eji4726-bib-0085){ref-type="ref"}\]. Once translated, monomeric CRP (mCRP; 21 kDa) assembles to pentamers in human plasma, which, in fact, might be in equilibrium with a decamer form under physiological conditions \[[86](#eji4726-bib-0086){ref-type="ref"}\]. Mechanistically, CRP has homologous regions found in immunoglobins G (IgGs) and can therefore bind to FcγRI and FcγRIIa but also activate the complement cascade via C1q \[[87](#eji4726-bib-0087){ref-type="ref"}\]. Moreover, an interesting dissociation mechanism has been described for pentameric CRP in the presence of bioactive lipids recently, suggesting that pentameric CRP which gives rise to mCRP \[[88](#eji4726-bib-0088){ref-type="ref"}, [89](#eji4726-bib-0089){ref-type="ref"}, [90](#eji4726-bib-0090){ref-type="ref"}, [91](#eji4726-bib-0091){ref-type="ref"}\]. mCRP, formed on the surface of activated platelets, shows increased inflammatory properties and is rarely found in circulation, suggesting its predominant role in local inflammation that likely serves as a contributing factor atherosclerotic plaque formation \[[92](#eji4726-bib-0092){ref-type="ref"}\]. However, mCRP also mitigates distinct immune reactions, for example, when bound to neutrophils via the FcγRIII, mCRP was shown to inhibit fMLP‐triggered chemotaxis \[[93](#eji4726-bib-0093){ref-type="ref"}\], while aggregated CRP appeared to enhance reactive oxygen species production in response to aggregated IgG but not other stimuli, such as phorbol myristate acetate or opsonized zymosan \[[94](#eji4726-bib-0094){ref-type="ref"}\]. Moreover, by recruiting factor H, mCRP also seems to be involved in blunting C3b activity, thereby contributing to the safe removal of opsonized bodily material \[[95](#eji4726-bib-0095){ref-type="ref"}\]. ### CRP: epidemiologic data‐lung immunity {#eji4726-sec-0090} CRP is a well‐established biomarker of lung infection. In fact, CRP was first described in patients suffering from acute pneumonia due to *Streptococcus pneumoniae* infections \[[73](#eji4726-bib-0073){ref-type="ref"}\]. Subsequently, increased CRP levels have been established as a systemic marker of reduced lung function also in chronic lung diseases such as cystic fibrosis \[[96](#eji4726-bib-0096){ref-type="ref"}\] and chronic obstructive pulmonary disease \[[97](#eji4726-bib-0097){ref-type="ref"}\]. The exact role of mCRP in the lung is still unknown. ### CRP: role in allergic diseases {#eji4726-sec-0100} Serum CRP levels, as determined by high‐sensitivity CRP assays, have been found to be significantly elevated in asthma patients as compared to healthy controls \[[14](#eji4726-bib-0014){ref-type="ref"}, [34](#eji4726-bib-0034){ref-type="ref"}, [40](#eji4726-bib-0040){ref-type="ref"}, [98](#eji4726-bib-0098){ref-type="ref"}, [99](#eji4726-bib-0099){ref-type="ref"}\], suggesting an association between plasma CRP levels and the severity of asthma \[[39](#eji4726-bib-0039){ref-type="ref"}, [100](#eji4726-bib-0100){ref-type="ref"}\]. Indeed, elevated serum CRP levels have been correlated with decreased pulmonary function and increased sputum eosinophilic cationic protein levels and eosinophil numbers in steroid naïve asthma patients \[[34](#eji4726-bib-0034){ref-type="ref"}, [35](#eji4726-bib-0035){ref-type="ref"}, [39](#eji4726-bib-0039){ref-type="ref"}, [98](#eji4726-bib-0098){ref-type="ref"}, [99](#eji4726-bib-0099){ref-type="ref"}, [101](#eji4726-bib-0101){ref-type="ref"}, [102](#eji4726-bib-0102){ref-type="ref"}\]. These studies were the first to indicate an association of systemic inflammation with airway inflammation. Similarly, studies in Danish school children as well as cohorts of U.S. children and adolescents reported that increased CRP levels were significantly associated with allergic sensitization and an elevated risk of concomitant allergic airway inflammation (rhinitis, asthma), however, independently of allergen type or clinical allergy symptoms \[[103](#eji4726-bib-0103){ref-type="ref"}, [104](#eji4726-bib-0104){ref-type="ref"}\]. Yet, CRP cannot be used as a predictive biomarker in allergy as no associations were observed between CRP levels at 6 months of age with later development of allergic diseases \[[104](#eji4726-bib-0104){ref-type="ref"}\]. Therefore, it was concluded that the low‐grade inflammation seen in children with allergic sensitization may be host intrinsic and part of a systemic inflammatory disorder rather than an isolated type 2 immune response or indirectly linked to sensitization through shared environmental risk factors (diet, microbiome) \[[104](#eji4726-bib-0104){ref-type="ref"}\]. Apart from systemic alterations of serum CRP levels during lung diseases, there is clear evidence that CRP is also produced and secreted locally in the airways by epithelial cells such as nasal epithelial cells and the alveolar basal epithelial cell line A549 \[[105](#eji4726-bib-0105){ref-type="ref"}, [106](#eji4726-bib-0106){ref-type="ref"}, [107](#eji4726-bib-0107){ref-type="ref"}\] and alveolar macrophages \[[108](#eji4726-bib-0108){ref-type="ref"}\] Therefore, CRP may have direct and local effects on airway inflammation \[[109](#eji4726-bib-0109){ref-type="ref"}, [110](#eji4726-bib-0110){ref-type="ref"}\] and contribute to bacterial clearance in the human respiratory tract \[[105](#eji4726-bib-0105){ref-type="ref"}\]. With regard to the short pentraxin SAP, very little is known about its expression in allergic disease. Recently, Gao and coworkers found that sputum SAP levels in patients with eosinophilic asthma were significantly higher than in individuals suffering from noneosinophilic asthma or healthy controls \[[14](#eji4726-bib-0014){ref-type="ref"}\]. Collectins {#eji4726-sec-0110} ========== Collectins are collagen containing C‐type lectins comprising MBL, surfactant proteins A and D as well as organ‐specific collectins. Mannose‐binding lectin {#eji4726-sec-0120} ---------------------- ### MBL: structure/function relationship {#eji4726-sec-0130} Human mannose (also mannan)‐binding lectin recognizes a large array of pathogens and activates the most ancient pathway of complement \[[111](#eji4726-bib-0111){ref-type="ref"}\], very similar to C1q, which activates the classical pathway of complement \[[112](#eji4726-bib-0112){ref-type="ref"}\]. MBL is composed of three identical 32 kDa polypeptide chains, which associate to a hexamer \[[113](#eji4726-bib-0113){ref-type="ref"}\]. The individual polypeptide chains are composed of a carbohydrate binding and a neck domain, which is followed by a collagen domain. In circulation, MBL is complexed to the MBL‐associated serine proteases MASP‐1 and ‐2 \[[114](#eji4726-bib-0114){ref-type="ref"}, [115](#eji4726-bib-0115){ref-type="ref"}\]. MBL is mainly expressed and produced in the liver \[[116](#eji4726-bib-0116){ref-type="ref"}\]. ### MBL: epidemiologic data‐lung immunity {#eji4726-sec-0140} Studies in mouse and man have indicated that MBL is absent from the BALF of healthy individuals \[[117](#eji4726-bib-0117){ref-type="ref"}, [118](#eji4726-bib-0118){ref-type="ref"}\], but can be clearly detected in the BALF of patients suffering from acute (pneumonia) \[[119](#eji4726-bib-0119){ref-type="ref"}\] or chronic (protracted bacterial bronchitis) \[[120](#eji4726-bib-0120){ref-type="ref"}\] respiratory diseases. Thus, detectable MBL levels in BALF can be regarded as a sign of inflammatory immune cell infiltration during infection \[[119](#eji4726-bib-0119){ref-type="ref"}\]. With regard to its protective activity, MBL seems to be especially important during early childhood and in situations of immunosuppression. Some MBL variants fail to produce stable multimeric forms and are thus dysfunctional \[[116](#eji4726-bib-0116){ref-type="ref"}\]. Indeed, MBL protects from infections and prolongs life expectancy in patients with congenital lung diseases such as cystic fibrosis, in which low MBL levels reduce life expectancy by 8 years \[[121](#eji4726-bib-0121){ref-type="ref"}\]. The disease‐modifying role of MBL2 has been also correlated with the frequency of respiratory infection‐associated hospital admissions in chronic obstructive pulmonary disease patients \[[122](#eji4726-bib-0122){ref-type="ref"}\], implicating a putative beneficial role for MBL replacement therapy in this disease entity \[[123](#eji4726-bib-0123){ref-type="ref"}\]. Paradoxically, a recently performed long‐term follow‐up study showed that MBL‐deficient COPD patients had more polymorphic microbiota and less risk for infectious exacerbations when compared to patients with fully functional MBL. This was attributed to the fact that no oxidized and thus no macrophage‐inhibitory MBL could be formed in the collective of MBL‐deficient patients studied \[[124](#eji4726-bib-0124){ref-type="ref"}\]. Hence, lung MBL can be regarded a double‐edged sword in situations of chronic inflammation, since oxidized forms of MBL might interfere with MBL‐oligomer formation and thus clearance of distinct microorganisms \[[125](#eji4726-bib-0125){ref-type="ref"}\] Apart from opsonizing pathogens and thus facilitating phagocytosis of microorganisms, MBL also largely contributes to the removal of apoptotic cells, especially within the lining of the lung alveoli, a process commonly referred to as efferocytosis \[[126](#eji4726-bib-0126){ref-type="ref"}\]. Target cell bound MBL is recognized by calreticulin (the cC1qR), which is associated with CD91 (the α2 macroglobulin receptor), an endocytic receptor expressed on many cell types (hepatocytes, neurons, fibroblasts) including blood monocytes \[[127](#eji4726-bib-0127){ref-type="ref"}\]. While the role of calreticulin for the uptake of MBL immune complexes is undisputed, some reports question a firm association of endocytosing calreticulin with CD91 \[[128](#eji4726-bib-0128){ref-type="ref"}\]. This might indicate that other cell surface receptors are involved in calreticulin‐dependent endocytosis and thus require a redefinition of the clearance pathway for MBL bound. Intriguingly, the high frequency of MBL deficiency, amounting to 5% in the general population, represents a constant matter of debate, since only a selective advantage is able to establish such a high degree of 'deficiency' for a single protein playing otherwise an important role in humans. One could speculate that MBL deficiency potentially protects from autoimmunity, because in the absence of MBL the likelihood that the immune system becomes primed against constituents of apoptotic cell bodies or serum components is much lower \[[129](#eji4726-bib-0129){ref-type="ref"}\]. Moreover, the absence of MBL might protect from intracellular parasitism/infections, since MBL might favour infection with mycobacteria (*Mycobacterium tuberculosis and Mycobacterium leprae*) \[[111](#eji4726-bib-0111){ref-type="ref"}\]. However, also the exact opposite has been hypothesized, by showing that coincubation of T cells with MBL inhibits T cell activation, while the absence of functional MBL has been implicated to promote T cell driven autoimmune processes such as systemic lupus erythematosus and rheumatoid arthritis \[[130](#eji4726-bib-0130){ref-type="ref"}\]. ### MBL: role in allergic diseases {#eji4726-sec-0150} MBL has been shown to bind to distinct allergen extracts and to activate complement via the lectin pathway, however, only in the presence of allergen‐specific IgG \[[131](#eji4726-bib-0131){ref-type="ref"}\]. Moreover, significantly increased MBL plasma levels have been detected in children with asthma and in adults with asthma and associated allergic rhinitis \[[132](#eji4726-bib-0132){ref-type="ref"}\], which correlated with peripheral blood eosinophil levels in children \[[133](#eji4726-bib-0133){ref-type="ref"}\]. Similarly, patients suffering from *Aspergillus fumigatus*‐associated bronchopulmonary aspergillosis presented with elevated MBL serum levels \[[132](#eji4726-bib-0132){ref-type="ref"}\], with the generated C3a and C5a fragments contributing to the bridging of innate and adaptive immune responses in asthma \[[134](#eji4726-bib-0134){ref-type="ref"}\]. The possible involvement of MBL in the pathogenesis of airway hyperresponsiveness has been substantiated in preclinical studies in mMBL‐A knockout mice, which revealed a mitigated airway response upon challenge with *Aspergillus fumigatus* in the absence of mMBL‐A \[[135](#eji4726-bib-0135){ref-type="ref"}\]. Surfactant proteins‐A and ‐D {#eji4726-sec-0160} ---------------------------- ### Surfactant proteins‐A and‐D: structure/function relationship {#eji4726-sec-0170} In the lungs, surfactant protein‐A (SP‐A) and surfactant protein‐D (SP‐D) are expressed and produced by alveolar type II cells and club cells of the distal bronchioles \[[136](#eji4726-bib-0136){ref-type="ref"}, [137](#eji4726-bib-0137){ref-type="ref"}, [138](#eji4726-bib-0138){ref-type="ref"}\]. Similar to C1q, SP‐A proteins form a bouquet‐like structure \[[139](#eji4726-bib-0139){ref-type="ref"}\]. Within the alveolar compartment, SP‐A is tightly associated with phospholipids. SP‐A opsonizes *Staphylococcus aureus* but also *Herpes simplex* particles for uptake by alveolar macrophages \[[140](#eji4726-bib-0140){ref-type="ref"}, [141](#eji4726-bib-0141){ref-type="ref"}\]. In contrast, SP‐D obtains a cruciform structure, which links its trimeric subunits \[[142](#eji4726-bib-0142){ref-type="ref"}\]. In addition, SP‐D also exists as a nonlipid bound free form, especially in alveoli. Apart from saccharides, SP‐D also binds to phosphatidyl‐inositol and glucosylceramide \[[143](#eji4726-bib-0143){ref-type="ref"}, [144](#eji4726-bib-0144){ref-type="ref"}, [145](#eji4726-bib-0145){ref-type="ref"}\]. SP‐D plays an important role in the microbicidal activity within the alveolar space, since it opsonizes a number of Gram‐negative bacteria including *Escherichia coli*, activates the respiratory burst in alveolar macrophages \[[146](#eji4726-bib-0146){ref-type="ref"}\] and leads to uptake of *E. coli* particles by DCs followed by T cell activation \[[147](#eji4726-bib-0147){ref-type="ref"}\]. ### Surfactant proteins A and D: epidemiologic data‐lung immunity {#eji4726-sec-0180} Lung diseases, which increase alveolar‐capillary leakage, such as lung fibrosis or interstitial pneumonia, are associated with increased serum levels of SP‐A and SP‐D, while the respective BALF levels are reduced \[[148](#eji4726-bib-0148){ref-type="ref"}\], making SP‐A and SP‐D markers for alveolar integrity. In addition, increases in especially of SP‐D levels have been observed to correlate with the degree of lung pneumonitis upon irradiation therapy \[[149](#eji4726-bib-0149){ref-type="ref"}\]. ### Surfactant proteins A and D: role in allergic diseases {#eji4726-sec-0190} SP‐A has been shown to bind to a number of grass pollen grains in a Ca^2+^‐ and carbohydrate‐specific manner \[[150](#eji4726-bib-0150){ref-type="ref"}\]. Moreover, both SP‐A and SP‐D bind to house dust mite \[[151](#eji4726-bib-0151){ref-type="ref"}\] and *Aspergillus fumigatus* \[[152](#eji4726-bib-0152){ref-type="ref"}\] antigens, binding to the latter inhibits the allergen\'s recognition by specific IgE antibodies and consequently blunts IgE‐dependent effector cell \[[153](#eji4726-bib-0153){ref-type="ref"}\] and T cell \[[154](#eji4726-bib-0154){ref-type="ref"}\] activation. In line with these findings, intranasal application of SP‐D in mice inhibited airway hypersensitivity induced by *Aspergillus fumigatus* \[[155](#eji4726-bib-0155){ref-type="ref"}, [156](#eji4726-bib-0156){ref-type="ref"}\] and *Dermatophagoides pteronyssinus* \[[157](#eji4726-bib-0157){ref-type="ref"}\], while challenge with *Aspergillus fumigatus* of SP‐A or SP‐D knockout mice promoted type 2 immunity involving hypereosinophilia and a high IL‐13/IFN‐γ‐ratio, which, again, could be abrogated by SP‐A and/or SP‐D application \[[158](#eji4726-bib-0158){ref-type="ref"}\]. Of relevance, once IL‐13 becomes produced, it reduces SP‐D production by alveolar type II cells \[[159](#eji4726-bib-0159){ref-type="ref"}\]. Irrespective of this regulatory mechanism, SP‐D serum \[[160](#eji4726-bib-0160){ref-type="ref"}\] and BALF \[[161](#eji4726-bib-0161){ref-type="ref"}\] levels were found to be increased in patients with allergic asthma. Ficolins {#eji4726-sec-0200} ======== Ficolins: structure/function relationship {#eji4726-sec-0210} ----------------------------------------- Ficolins represent a class of opsonins that are characterized by the presence of a C‐terminal fibrinogen‐like and an N‐terminal collagen‐like domain (fi‐col‐ins) \[[162](#eji4726-bib-0162){ref-type="ref"}\] that have distinct ligand specificities (GlcNAc GalNAc, sialic acid, [d]{.smallcaps}‐fucose). The fibrinogen‐like domain specifically binds to pathogen‐associated carbohydrates, while the collagen‐like domain signals ligand binding to the MBL‐associated serine proteases leading to their activation, eventually leading to the cleavage of C2 and C4 and the formation of C3 convertase activity \[[116](#eji4726-bib-0116){ref-type="ref"}\]. Ficolins are lectins, which assemble to form oligomeric structures that look like a 'bunch of flowers' and have high, calcium‐dependent binding specificity for *N*‐acetyl‐glucosamine (GlcNAc), which is an important component of the cell wall of *Aspergillus fumigatus*. Although primarily functioning as opsonins, orchestrating the phagocytosis of fungi, ficolins also contribute to the activation of the lectin pathway by interacting and activating the MBL‐associated serine proteases \[[163](#eji4726-bib-0163){ref-type="ref"}\] and trigger the production of inflammatory cytokines and nitric oxide by macrophages \[[164](#eji4726-bib-0164){ref-type="ref"}\]. Ficolins: Epidemiologic data‐lung immunity {#eji4726-sec-0220} ------------------------------------------ Two serum soluble ficolins produced by hepatocytes exist, that is, L‐ficolin (ficolin‐2) \[[165](#eji4726-bib-0165){ref-type="ref"}\] and H‐ficolin (ficolin‐3) \[[166](#eji4726-bib-0166){ref-type="ref"}\]. Notably, the latter can be also produced and secreted by bronchial and alveolar type II epithelial cells \[[167](#eji4726-bib-0167){ref-type="ref"}\]. Ficolin‐3 becomes increasingly produced by AEC upon systemic challenge with lipopolysaccharides (LPS) \[[168](#eji4726-bib-0168){ref-type="ref"}\]. In addition, a membrane‐bound ficolin, M‐ficolin, for which binding to *Escherichia coli* has been described \[[169](#eji4726-bib-0169){ref-type="ref"}\], is primarily expressed on monocytes \[[170](#eji4726-bib-0170){ref-type="ref"}\]. While there is clear‐cut binding of ficolin‐3 to *Aspergillus fumigatus* conidia, its binding affinity is much lower than that of serum‐derived ficolin‐2. Moreover, low ficolin‐2 serum levels identify patients suffering from common variable immunodeficiency who are at risk to develop bronchiectases, a serious life‐shortening sequela of the disease \[[171](#eji4726-bib-0171){ref-type="ref"}\]. Ficolin‐3 bound to sialylated glycans scavenges influenza A virus, which leads to the block of infectivity and hemagglutination activity of influenza A virus and its complement mediated destruction \[[172](#eji4726-bib-0172){ref-type="ref"}\]. Ficolin‐3 deficiency due to mutations at position 1637 of the gene (1637delC) leads to immunodeficiency with recurrent lower respiratory infections, septicemia and warts \[[173](#eji4726-bib-0173){ref-type="ref"}\], which can even lead to sudden death due to acute meningitis \[[174](#eji4726-bib-0174){ref-type="ref"}\]. Interestingly, individuals heterozygous for the 1637delC mutation present with lower serum ficolin‐3 concentrations \[[175](#eji4726-bib-0175){ref-type="ref"}\]. Moreover, ficolin‐3 has been shown to be responsible for the removal of late apoptotic cells \[[176](#eji4726-bib-0176){ref-type="ref"}\]. This salient function might also be one of the reasons why ficolin‐3 is also a prominent target for autoantibodies (with DNA and other factors acting as adjuvant). Of note, anti‐ficolin‐3 antibodies are frequently (in one third) observed in systemic lupus erythematosus patients and currently have the strongest association of all autoantibodies with active lupus nephritis \[[177](#eji4726-bib-0177){ref-type="ref"}\]. Ficolins: Role in allergic diseases {#eji4726-sec-0230} ----------------------------------- The direct involvement of ficolins in allergic inflammation is currently unknown. However, a retrospective study suggests that low ficolin‐2 serum levels are associated with atopic disorders \[[178](#eji4726-bib-0178){ref-type="ref"}\], which subsequently was confirmed in a prospective study indicating that ficolin‐2 deficiency is associated with allergic and infectious respiratory diseases \[[179](#eji4726-bib-0179){ref-type="ref"}\]. The authors speculated that low ficolin levels might lead to a lack of protection from microorganisms potentially complicating (driving) allergic diseases \[[178](#eji4726-bib-0178){ref-type="ref"}\]. Serum amyloid A {#eji4726-sec-0240} =============== Similar to the short pentraxin CRP, the serum amyloid A (SAA) proteins, have been classically viewed as highly inducible, liver‐derived factors in response to infection or trauma \[[180](#eji4726-bib-0180){ref-type="ref"}\]. Today, baseline expression of SAA is well documented for many extrahepatic human tissues including the alveolar epithelial lining \[[181](#eji4726-bib-0181){ref-type="ref"}\], alveolar macrophages \[[181](#eji4726-bib-0181){ref-type="ref"}\] but also monocytes \[[182](#eji4726-bib-0182){ref-type="ref"}\]. SAA: structure/function relationship {#eji4726-sec-0250} ------------------------------------ SAA is a sPRR for Gram‐negative bacteria \[[183](#eji4726-bib-0183){ref-type="ref"}, [184](#eji4726-bib-0184){ref-type="ref"}\] and blocks entry of viruses (e.g. hepatitis C virus) into cells \[[185](#eji4726-bib-0185){ref-type="ref"}\]. More recently, SAA1 was found to directly bind LPS to form a complex that promotes LPS clearance by macrophages, which offers partial protection against LPS‐induced inflammation and acute lung injury \[[186](#eji4726-bib-0186){ref-type="ref"}\]. In addition, SAA proteins can bind and transport retinol during bacterial infection \[[187](#eji4726-bib-0187){ref-type="ref"}\], thereby restricting their optimal growth and expansion. Moreover, mice lacking *Saa3* express lower levels of retinoic acid receptors than their wild‐type littermates, which inhibits their ability to respond to retinoic acid and may lead to an incapacity to develop immune tolerance \[[188](#eji4726-bib-0188){ref-type="ref"}\]. In humans, there are three SAA proteins that are encoded by the inducible genes, *SAA1* and *SAA2*, and the *SAA4* gene that is constitutively expressed \[[180](#eji4726-bib-0180){ref-type="ref"}, [189](#eji4726-bib-0189){ref-type="ref"}\]. *SAA3*, which is a pseudogene in humans, encodes an additional inducible form of SAA in mice with high structural similarities to other SAA proteins \[[180](#eji4726-bib-0180){ref-type="ref"}, [189](#eji4726-bib-0189){ref-type="ref"}\]. Locally produced, lipid‐free SAA1 can act through a number of potential SAA receptors \[[190](#eji4726-bib-0190){ref-type="ref"}\] and has pleiotropic effects on the immune system including the recruitment of immune cells \[[191](#eji4726-bib-0191){ref-type="ref"}, [192](#eji4726-bib-0192){ref-type="ref"}, [193](#eji4726-bib-0193){ref-type="ref"}\], epithelial wound repair \[[194](#eji4726-bib-0194){ref-type="ref"}\] and proinflammatory cytokine production \[[185](#eji4726-bib-0185){ref-type="ref"}, [195](#eji4726-bib-0195){ref-type="ref"}, [196](#eji4726-bib-0196){ref-type="ref"}, [197](#eji4726-bib-0197){ref-type="ref"}, [198](#eji4726-bib-0198){ref-type="ref"}, [199](#eji4726-bib-0199){ref-type="ref"}, [200](#eji4726-bib-0200){ref-type="ref"}\]. Considering the dramatic increase in SAA levels at sites of inflammation, the biological activities of SAA1 need to be tightly controlled, for example, by factors present in specific tissue microenvironments or by the ability of SAA1 to adopt different oligomeric states \[[187](#eji4726-bib-0187){ref-type="ref"}, [201](#eji4726-bib-0201){ref-type="ref"}, [202](#eji4726-bib-0202){ref-type="ref"}, [203](#eji4726-bib-0203){ref-type="ref"}, [204](#eji4726-bib-0204){ref-type="ref"}\]. The structure of SAA, however, is different from chemokines and cytokines, and SAA seems to require ligand‐induced activation, which is not observed under physiological conditions \[[186](#eji4726-bib-0186){ref-type="ref"}, [205](#eji4726-bib-0205){ref-type="ref"}, [206](#eji4726-bib-0206){ref-type="ref"}\]. For example, binding to plasma high‐density lipoprotein abrogates the cytokine‐like activities of SAA proteins \[[206](#eji4726-bib-0206){ref-type="ref"}\]. SAA: epidemiologic data‐lung immunity {#eji4726-sec-0260} ------------------------------------- *SAA3* knockout mice develop intrinsic airway hyperresponsiveness and show increased mortality to influenza A virus infection \[[188](#eji4726-bib-0188){ref-type="ref"}\]. In mice, *Saa3* may therefore have key homeostatic functions in lung development and inflammation \[[188](#eji4726-bib-0188){ref-type="ref"}\]. The sequence similarity of mouse Saa3 and human SAA1/2 suggests that their functions during inflammatory processes might be similar. In the lungs, SAA1 has been established as a mediator of local effector Th17 responses \[[197](#eji4726-bib-0197){ref-type="ref"}, [207](#eji4726-bib-0207){ref-type="ref"}, [208](#eji4726-bib-0208){ref-type="ref"}, [209](#eji4726-bib-0209){ref-type="ref"}\]. This is consistent with the role for SAAs in retinol shuttling that is required to induce Th17 responses in the presence of infection and inflammation \[[187](#eji4726-bib-0187){ref-type="ref"}\]. SAA‐induced Th17 responses are associated with robust airway neutrophilia \[[207](#eji4726-bib-0207){ref-type="ref"}\] and more severe asthma phenotypes that are less responsive to corticosteroids \[[210](#eji4726-bib-0210){ref-type="ref"}\]. SAA: role in allergic diseases {#eji4726-sec-0270} ------------------------------ SAA is by far the most widely studied acute phase protein in the airways and has been repeatedly associated with chronic rhinosinusitis but also asthma symptoms \[[23](#eji4726-bib-0023){ref-type="ref"}, [24](#eji4726-bib-0024){ref-type="ref"}, [36](#eji4726-bib-0036){ref-type="ref"}, [186](#eji4726-bib-0186){ref-type="ref"}, [188](#eji4726-bib-0188){ref-type="ref"}, [207](#eji4726-bib-0207){ref-type="ref"}, [211](#eji4726-bib-0211){ref-type="ref"}, [212](#eji4726-bib-0212){ref-type="ref"}, [213](#eji4726-bib-0213){ref-type="ref"}, [214](#eji4726-bib-0214){ref-type="ref"}\]. One such study analysed a fraction of the SARP‐3 study population (i.e. adult participants suffering from severe and nonsevere asthma), an NIH‐sponsored multisite cohort study conducted to investigate mechanisms of severe asthma. Notably, it was found that patients with severe asthma showed increased pruning of the pulmonary vasculature that was associated with decreased lung function, greater peripheral and sputum eosinophilia, and higher BAL SAA/lipoxin A~4~ ratio \[[87](#eji4726-bib-0087){ref-type="ref"}\]. Moreover, there is a strong correlation between elevated blood as well as sputum SAA levels and asthma prevalence and/or allergic rhinitis and disease severity \[[23](#eji4726-bib-0023){ref-type="ref"}, [24](#eji4726-bib-0024){ref-type="ref"}, [212](#eji4726-bib-0212){ref-type="ref"}\]. As described above, the ability of SAA1 to adopt different oligomeric states (monomers and oligomers) \[[202](#eji4726-bib-0202){ref-type="ref"}\] or to transport retinols and related molecules \[[187](#eji4726-bib-0187){ref-type="ref"}, [188](#eji4726-bib-0188){ref-type="ref"}\] might serve to scavenge and sequester essential vitamins (or even nutrients) to deter pathogen growth at mucosal surfaces \[[215](#eji4726-bib-0215){ref-type="ref"}\] and may thus have important functional implications for promoting local inflammation. Considering that SAA is increased in allergic asthma \[[23](#eji4726-bib-0023){ref-type="ref"}, [24](#eji4726-bib-0024){ref-type="ref"}\], it remains to be established whether SAA functions as an adjuvant and/or interacts directly with environmental airborne proteins to promote allergic sensitization. Preliminary evidence of the authors (Smole et al., in revision) suggests that a component within house dust mite extracts can lead to active forms of SAA which impact on AEC and the induction of type 2 immunity, thus contributing to the priming of exposed individuals for sensitization and allergic disease (Smole et al.). A better understanding of the diverse immune protective mechanisms of acute phase SAA at mucosal surfaces may help to develop novel therapies for infectious but also inflammatory diseases such as allergic asthma and rhinitis \[[186](#eji4726-bib-0186){ref-type="ref"}\]. Complement components {#eji4726-sec-0280} ===================== The complement system represents a central and very effective innate protection mechanism, which received its name from *Paul Ehrlich*, who intended to indicate that it enhances (complements) the activity of antibodies and phagocytes. It can be regarded as one of the centrepieces of humoral innate immunity and critically contributes to defence mechanisms in the airways. An emerging theme is, however, the recently discovered noncanonical functions of complement \[[216](#eji4726-bib-0216){ref-type="ref"}, [217](#eji4726-bib-0217){ref-type="ref"}\], which indicate that this system has several important functions beyond being 'the guardian of the extracellular space' and being an 'important player within the innate sensor network'. These functions include but are not restricted to its important contributions to neuronal cell and organ development and brain function \[[218](#eji4726-bib-0218){ref-type="ref"}\], tissue repair, regeneration \[[219](#eji4726-bib-0219){ref-type="ref"}\] and metabolic programming (also referred to as complosome) \[[220](#eji4726-bib-0220){ref-type="ref"}, [221](#eji4726-bib-0221){ref-type="ref"}, [222](#eji4726-bib-0222){ref-type="ref"}, [223](#eji4726-bib-0223){ref-type="ref"}\], including the modulation of immune cell effector functions. Complement: structure/function relationship {#eji4726-sec-0290} ------------------------------------------- The complement system encompasses more than 30 serum‐soluble and membrane‐expressed molecules \[[224](#eji4726-bib-0224){ref-type="ref"}, [225](#eji4726-bib-0225){ref-type="ref"}\]. All three complement activation pathways (classic, nonclassic and lectin) eventually lead to C3‐ followed by C5‐convertase activity. Serine protease activity converts the evolutionary conserved C3 molecule, which is molecularly related to alpha‐2‐macroglobulin and highly abundant in human serum (1.2--1.3 g/L), into the small (10 kDa) and diffusible anaphylatoxin C3a and the larger opsonin fragment C3b \[[224](#eji4726-bib-0224){ref-type="ref"}, [225](#eji4726-bib-0225){ref-type="ref"}\]. Similarly, C5 is converted into the small anaphylatoxin C5a (11 kDa) and the larger C5b fragment. The latter associates with C6 and C7, attaches to membranes (microbial or bodily) and completes the formation of the membrane attack complex by binding of C8 and C9 \[[223](#eji4726-bib-0223){ref-type="ref"}\]. C3b‐dependent opsonization occurs upon conformational change of C3b and exposure of the hidden internal thioester bond which is highly reactive and covalently binds to nucleophils such as OH‐ or NH~2~‐groups in its immediate surroundings. The anaphylatoxins C3a and C5a are potent inflammatory mediators, which target a number of inflammatory and noninflammatory cell types. Complement: epidemiologic data‐lung immunity {#eji4726-sec-0300} -------------------------------------------- The central components of complement, that is, circulating C3 and C5, are largely produced in the liver. However, it has been clearly demonstrated that apart from other nonliver cells, such as endothelial cells and immune cells (granulocytes, monocytes, DCs, T cells), also airway epithelial cells, such as BEAS‐2B cells, constitutively synthesize and release C3 under serum‐free conditions which can be significantly upregulated by IL‐1 or TNF‐α but not IFN‐γ \[[226](#eji4726-bib-0226){ref-type="ref"}, [227](#eji4726-bib-0227){ref-type="ref"}\]. In addition, lung cancer cells substantially generate and release C5a, which is believed to create a milieu favourable for tumour cells and thus contributing to lung cancer progression \[[228](#eji4726-bib-0228){ref-type="ref"}\]. Intratracheal instillation of C3a or peptides thereof induces respiratory distress, which is accompanied by C3‐dependent bronchial smooth muscle contraction in bronchioles and hemodynamic effects due to C3a‐induced platelet aggregation in pulmonary arteries \[[229](#eji4726-bib-0229){ref-type="ref"}, [230](#eji4726-bib-0230){ref-type="ref"}\]. AEC (and airway smooth muscle cells) can sense complement, since both human and murine AEC express C3a‐ and C5a‐receptors under steady‐state conditions, which become upregulated in situations of endotoxinemia and allergen‐induced asthma \[[231](#eji4726-bib-0231){ref-type="ref"}\]. Targeting of these receptors on AEC by intratracheal instillation of antibodies has been shown to reduce pulmonary oedema and lung injury \[[232](#eji4726-bib-0232){ref-type="ref"}, [233](#eji4726-bib-0233){ref-type="ref"}\]. Apart from soluble complement components, AEC under steady‐state conditions also express transmembrane complement regulatory proteins such as Membrane cofactor protein (CD46), Decay accelerating factor (CD55) and protectin (CD59), with IFN‐γ able to significantly upregulate DAF levels \[[226](#eji4726-bib-0226){ref-type="ref"}\]. It is therefore not surprising that complement proteins can also be detected in the BALF of healthy individuals \[[234](#eji4726-bib-0234){ref-type="ref"}\] and that complement levels are increased upon exposure to bacterial danger signals such as LPS \[[235](#eji4726-bib-0235){ref-type="ref"}\]. Some of these complement regulatory proteins might also contribute to the overall reduction of inflammation, such as CD46, which was shown to downregulate IL‐1β levels induced by H~2~O~2~‐activated AEC due to induction of autophagy and associated downregulation of inflammasome components like pro‐IL‐1β and NLRP3 \[[236](#eji4726-bib-0236){ref-type="ref"}\]. Moreover, AEC were shown to scavenge (actively take up) C3 from their surroundings, very similar to CD4 T cells. In its intracellularly stored form, C3 is supposed to play a cytoprotective role by potentially influencing oxidative stress of AEC \[[237](#eji4726-bib-0237){ref-type="ref"}\]. Production but also storage of C3 by AEC might have important implications also for end‐stage lung diseases such as cystic fibrosis and chronic obstructive pulmonary disease in which C3 stores in AEC are found to be increased. C3 is stored within the endoplasmic reticulum and late endosomes in its mature form, ready to become activated by cleavage. Other cell types known to not only synthesize but also store large amounts of C3 are neutrophils \[[238](#eji4726-bib-0238){ref-type="ref"}\] and monocytes \[[239](#eji4726-bib-0239){ref-type="ref"}\]. Moreover, T cells are not only activated by complement split products but also contain C3 and C5 intracellularly \[[240](#eji4726-bib-0240){ref-type="ref"}\]. Whether T cells express and are able to activation‐dependently upregulate C3 and C5, or rather accumulate preformed C3 and C5 proteins present in large amounts in bodily fluids is a matter of intense discussion and might critically depended on APC‐T cell interaction \[[8](#eji4726-bib-0008){ref-type="ref"}, [241](#eji4726-bib-0241){ref-type="ref"}\]. However, it has been clearly shown that T cells have complement convertase activity and can activate intracellular C3 by virtue of T cell‐expressed cathepsin L \[[9](#eji4726-bib-0009){ref-type="ref"}\]. Secretion of such activated C3 (C3a) contributes to T cell‐mediated inflammation in target tissues including the lung. This also fits to the recent finding that intracellular pools of C3 become reduced upon T cell activation \[[8](#eji4726-bib-0008){ref-type="ref"}\]. Complement: role in allergic diseases {#eji4726-sec-0310} ------------------------------------- Early reports by Nagata and Glovsky have shown that different airborne allergens contain complement convertase activity and lead to C3a formation when incubated with serum \[[242](#eji4726-bib-0242){ref-type="ref"}\]. Later on, these findings were corroborated by studies demonstrating that the serine protease from the house dust mite *Dermatophagoides farinae* generates the anaphylatoxins C3a and C5a \[[243](#eji4726-bib-0243){ref-type="ref"}\], which already suggested a possible involvement of house dust proteases in the pathogenesis of allergic diseases \[[243](#eji4726-bib-0243){ref-type="ref"}\]. Moreover, the mould component chitin, which is an important component of *Aspergillus fumigatus*, is a strong activator of the alternative complement pathway and thereby shapes the ensuing immune response upon inhalation \[[244](#eji4726-bib-0244){ref-type="ref"}\]. Preclinical data confirmed the importance of anaphylatoxins in asthma pathogenesis. Along those lines, it was demonstrated that genetic deletion of C3aR on mouse AEC protects mice from lung dysfunction after allergen challenge \[[245](#eji4726-bib-0245){ref-type="ref"}\]. Consistent with these findings, BALF of human patients with asthma \[[246](#eji4726-bib-0246){ref-type="ref"}, [247](#eji4726-bib-0247){ref-type="ref"}\] but also respiratory distress syndrome \[[248](#eji4726-bib-0248){ref-type="ref"}\] contain increased levels of C3a and C5a which might, at least in part, originate from AEC stores. Notably, the C3 pathway has been shown to be involved in the development of Th2 responses in allergic inflammation \[[245](#eji4726-bib-0245){ref-type="ref"}, [249](#eji4726-bib-0249){ref-type="ref"}\], which is also reflected by the fact that C3a levels are found upregulated in individuals with respiratory allergies \[[250](#eji4726-bib-0250){ref-type="ref"}, [251](#eji4726-bib-0251){ref-type="ref"}\] and asthma \[[252](#eji4726-bib-0252){ref-type="ref"}\]. In addition, a positive correlation between C3 levels and asthma severity has been described \[[252](#eji4726-bib-0252){ref-type="ref"}\]. Apart from lungs, also upper airway diseases, such as chronic rhinosinusitis, have been shown to be associated with increased local production of C3. Indeed, its neutralization has therapeutic effects in this difficult to treat condition \[[253](#eji4726-bib-0253){ref-type="ref"}\]. In summary, the complement system and especially C3 is pivotal for the innate humoral protection of the airways. Humoral factors that are directly released by airway epithelial cells or immune cells associated with the epithelial cell lining diffuse into the mucus lining, and the majority of them eventually take advantage of the C3 convertase activity to signal the presence of danger. Summary {#eji4726-sec-0320} ======= Mucosal surfaces represent very critical barrier tissues of our body that ensure immune homeostasis in response to microorganism that ranges from host defence to active tolerance and symbiosis. Consequently, and among other protection mechanisms, sPRR play an important role in local immune surveillance of mucosal surfaces and can quickly neutralize, opsonize or induce phagocytosis of potentially harmful intruders. Ligand binding by sPRR very effectively activates and amplifies canonical pathways of humoral immunity such as the complement system. The central component of it, that is C3, is also found in the airway mucosal lining. As such, sPRRs within the airway epithelial lining represent a very effective form of 'forward defence', especially against highly virulent and/or invasive strains of microorganisms characterized by strain‐specific sugar or lipid moieties (e.g. *Streptococcus pneumoniae*, *Aspergillus fumigatus*, etc.), and thereby prevent early and devastating epithelial damage to occur. Although seemingly redundant, sPRRs have largely diverging ligand specificities, several of them converge by activating the complement cascade. A large body of evidence suggests that sPRR might also sense the presence of usually innocuous proteins (allergens), setting the stage for allergic sensitization and subsequent allergic disease. Future research will focus on dissecting the underlying cellular and molecular complexity of potential sPRR--allergen interactions and highlight approaches of how to interfere with such undesired, allergen‐driven activation mechanism. Author contributions {#eji4726-sec-0340} ==================== U.S., B.K. and W.F.P. wrote the paper. Conflict of interest {#eji4726-sec-0350} ==================== Winfried F. Pickl holds stocks of Biomay AG and receives honoraria from Novartis and Roche. All other authors have no additional commercial or financial conflict of interest. AEC : airway epithelial cell BALF : bronchoalveolar lavage fluid CRP : C‐reactive protein DCs : dendritic cells MBL : mannose‐binding lectin mCRP : momomeric C‐reactive protein PTX : pentraxin SAA : serum amyloid A SAP : serum amyloid P SP‐A : surfactant protein‐A SP‐D : surfactant protein‐D sPRR : soluble pattern recognition receptors This work was supported by grants DK‐W1248 and SFB F4609 of the Austrian Science Foundation (FWF).

Introduction ============ Human evaluation or appraisal processes concern the assignment of positive and negative valence to things and events (Arnold, [@B6]; Lazarus, [@B44]; Scherer et al., [@B69]; Storbeck and Clore, [@B74]). Such appraisal processes are very important for human ontogenetic and phylogenetic fitness (Arnold, [@B6]; LeDoux, [@B45]). The appraisal of things and events is made with respect to an organism's vital goals. Appraisal can thus guide the selection of crucial stimulus information and goal-directed actions for utility maximization because utility is based on the achievement of merited or positively evaluated outcomes and the avoidance of negative consequences (Posner and Dehaene, [@B64]; Pessoa et al., [@B62]). In fact, humans carry out evaluations so routinely and automatically (Kissler et al., [@B35]) that the corresponding appraisal processes can be triggered by irrelevant words, and even by stimuli remaining outside of the awareness of humans, such as subliminal or unconscious words or tokens (Naccache et al., [@B54]; Galliard et al., [@B21]; Klauer et al., [@B36]; Pessiglione et al., [@B61]). These studies suggest a fast evaluation system that operates independent of conscious perception. Despite their enormous significance, however, to date we do not fully understand how such automatic evaluations relate to other non-evaluative dimensions of semantic processes. Here, we refer to semantic processes as the processes concerned with the assignment of meaning in general (Osgood et al., [@B60]; Meyer and Schvanefeldt, [@B51]; Kintsch and van Dijk, [@B34]; Rey, [@B66]). According to some theories, valence would be one integral dimension of all semantic meaning (Osgood et al., [@B60]). Even if this is true, however, it remains to be specified whether evaluative meaning is as quickly available as non-evaluative meaning or whether non-evaluative meaning precedes evaluative meaning. One possibility is that evaluations are special and carried out so quickly (Zajonc, [@B78]; LeDoux, [@B45]) as to precede and ground visual recognition and subsequent processing of non-evaluative forms of meaning. In line with this, for example, visual stimuli with a conditioned valence can elicit valence-dependent changes of visually evoked potentials of the human EEG with a latency of less than 100 ms (Stolarova et al., [@B73]). Related to this, Naccache et al. ([@B54]), for example, demonstrated that visually masked subliminal words elicited valence-dependent activation of the human amygdala, a brain structure involved in emotional processing (LeDoux, [@B45]). On the other hand, some emotion theories argued that visual recognition and specific non-evaluative semantic inferences have to precede appraisal and evaluative processes (Storbeck and Clore, [@B74]; Nummenmaa et al., [@B59]). Specifically, according to the embodied cognition view, word meaning is always grounded in more basic representations, such as sensory or sensorimotor representations (Barsalou, [@B9], [@B10]; Glenberg and Kaschak, [@B22]; Zwaan et al., [@B79]). Thus, valence could be accessible only after some other forms of non-evaluative sensory meaning have been successfully extracted from a word. In line with this, participants in a study of Nummenmaa et al. ([@B59]), for example, were able to saccade toward one of two scenes based on a scene's semantic content before they were able to saccade to a scene based on a scene's valence or affective content. Likewise, Niedenthal ([@B56]) argued that emotions could in general be grounded in preceding basic sensory, perceptual, and sensorimotor representations of postures and facial expressions. These divergent claims are the point of departure for the present research. We address the directed interaction of valence evaluation, semantic processing, and conscious perception. As follows from the above, a key issue is how quickly the valence of a word becomes available. Is valence accessible before non-evaluative forms of meaning, especially sensory-related semantic meaning? Or is the sensory-related and non-evaluative meaning of a word accessible before its valence? For our test, we used an across-category priming effect of sensory representations on valence recognition: the influence of spatial verticality on valence classification where spatially high corresponds to positive valence, and spatially low to negative valence. Of course, spatial location is but one of the sensory representations that nurtures the embodied representation of emotions. For instance, facial expressions are also among the powerful sensory representations that link up to emotions (Niedenthal et al., [@B57]). However, the association between valence and space is strong, robust, and already reflected in the very meaning of some words (Lakoff and Johnson, [@B41]; Melara and O'Brien, [@B50]). For example, the spatial word "*high*" can be used almost synonymously to denote an affective state of "*euphoria*" and the spatial word "*low*" for an affective state of "*depression*." In their seminal experiments, Meier and Robinson ([@B49]) demonstrated this modulating influence of spatial verticality on valence in a *space-valence congruence effect*: their participants were faster when discriminating the valence of positive words above screen center and of negative words below screen center. This advantage in space-valence congruent conditions was found in comparison to incongruent conditions, in which the negative words were presented above screen center and the positive words below screen center (for related results, see Crawford et al., [@B13]; Weger et al., [@B76]; Horstmann, [@B26]; Horstmann and Ansorge, [@B27]; Santiago et al., [@B68]; Ansorge and Bohner, [@B1]; Gozli et al., [@B23]). So far, this space-valence congruence effect has not been tested with subliminal words. This, however, is important. If we can demonstrate an across-category interaction with subliminal words (Forster, [@B20]; Ansorge et al., [@B3]) then the space-affect congruence interaction during lexical access reflects a form of early automatic impact of either quick semantic processes on evaluations or of quick evaluative appraisal on semantic processing. For example, Lamme ([@B42]) and Lamme and Roelfsema ([@B43]) estimated that unaware processing occurs during the first 100 ms post-stimulus. The exact duration might be slightly longer (Mulckhuyse and Theeuwes, [@B52]) but authors share the view that awareness-independent processing occurs early, during the feed-forward phase of stimulus processing, whereas awareness-dependent processing takes time and occurs later because it depends on feedback from processes higher up the hierarchy (Lamme, [@B42]). This implies limitations on subliminal processing. For example, subliminal processing could sometimes be conditional on top-down task sets. To be precise, subliminal processing could be "conditionally automatic," meaning that it depends on the task-relevance of the unaware stimulus (e.g., of a semantic dimension of a subliminal word; Bargh, [@B8]). Yet, importantly, this also means that a subliminal word does not elicit a fitting intention for its own processing in and by itself (Forster, [@B20]; Kinoshita et al., [@B33]). On the contrary, before the processing of subliminal stimuli can take place the prior set-up of a top-down or goal-directed intention would have to be firmly established (Klinger et al., [@B38]; Kunde et al., [@B39]). Only once such an intention, goal, or task set has been firmly established, a stimulus meaning of which a person remains unaware would be able to elicit a stimulus processing in the feed-forward processing phase (e.g., Norris and Kinoshita, [@B58]; Ansorge et al., [@B3], [@B5]). Note that this general assumption also holds true for the processing of a subliminal word's valence meaning (Klinger et al., [@B38]; De Houwer et al., [@B15]; Klauer and Musch, [@B37]; Eckstein and Perrig, [@B17]; see also Spruyt et al., [@B72]; Spruyt et al., [@B71]). Here, to test the predictions, and to understand whether sensory semantics quickly and (conditionally) automatically affected evaluations during the feed-forward phase and/or whether this is the other way round, we used (1) subliminal spatial words as primes and valence words as targets (Experiments 1 and 3), or we used (2) subliminal valence words as primes and spatial words as targets (Experiments 2 and 3). If valence meaning is based on (connotative) spatial representations, we expected a congruence effect of space primes on target evaluations (in Experiment 1). In the congruent condition, a valence judgment should be facilitated. For example, the prime word "up" should facilitate classifying the emotional target "happy" as positive. Facilitation was expected compared to the incongruent condition, for example, when the prime word "down" was presented prior to the emotional target "happy." If the expected congruence effect reflected a quick, mandatory (or conditionally automatic) process, this space-valence congruence effect should be found with subliminal space primes, too (Ansorge et al., [@B3]). In addition, if word valence is as quickly available as a word's spatial meaning, we should find a congruence effect in the reversed situation (in Experiment 2) in which subliminal valence primes were presented before spatial targets. When related predictions have been tested concerning the symmetry versus asymmetry of the space-valence congruence effect in the original paradigm of Meier and Robinson ([@B49]), researchers observed congruence effects based on irrelevant spatial positions during word evaluation but no influence of irrelevant word valence on the discrimination of word locations (Santiago et al., [@B68]; but see [Experiment 3](#s1){ref-type="sec"} of Meier and Robinson, [@B49]). However, Santiago et al. achieved their results with clearly visible words and locations, leaving it open whether a similar asymmetry of space-valence congruence originates during the feed-forward phase of processing. In addition, when more attention was shifted to the irrelevant valence of the clearly visible words, the space-valence congruence effect was reestablished even during spatial discrimination (Experiment 6 of Santiago et al., [@B68]). In our study, as control conditions, we therefore used (1) blocks with clearly visible, so-called "supraliminal" primes and targets from different categories (and from the same categories) and (2) trials with subliminal primes and targets from the same category (e.g., a valence prime before a valence target) instead of primes and targets from different categories in both Experiments 1 and 2. In the control conditions, several hypotheses predict a congruence effect. With the supraliminal primes, a congruence effect could be based on strategic rather than (only) quick, obligatory (or conditionally automatic) processing of the visible prime because the longer perceptual trace of the visible primes affords also more attentional dwelling on visible than masked primes: the clearly visible prime will be seen and can elicit its strategic processing in and by itself, so that processing of the priming word would no longer be dependent on a preceding fitting task set (Cheesman and Merikle, [@B11]; Forster, [@B20]). For example, in the study of Cheesman and Merikle, participants strategically used the predictive power of the categories of the visible prime words of male and female gender names for the target word categories of female and male gender names, respectively. This was evident in the efficient preparation of the most likely target word responses. This strategic effect was found when the prime words were visible but not when the prime words were invisible (see also, e.g., Kinoshita et al., [@B33]). It is possible that this boosting influence of strategies on priming is due to a fundamental difference between feed-forward processing and recurrent processing: due to the participants' awareness of the visible prime, this prime word might be broadcasted to different processing modules throughout the mental sphere and could thus be used for multiple new purposes, including the strategic assessment of the fit of its meaning relative to that of the target (Baars, [@B7]; Dehaene and Naccache, [@B16]). Therefore, the standard space-valence congruence effect should be found in the supraliminal conditions and a difference might be expected between supraliminal and subliminal conditions: if one of the prime word meaning dimensions (spatial or evaluative) contributing to the space-valence association is not also quickly and automatically processed (and thus could not be processed during the early feed-forward processing phase), then we expect that this dimension should become effective in the supraliminal but not in the subliminal conditions. Also, in the second type of control conditions, with the subliminal primes of the same category, a congruence effect of the masked primes is predicted based on the prime's potential of priming a motor response (Neumann, [@B55]; Kunde et al., [@B39]; Ansorge and Neumann, [@B4]). Say, one needs to press the left button for a positive target word and a right button for a negative target. A positive prime before a positive target would then indicate giving the same response but a negative prime before a positive target would cause response conflict. Because this response-activation effect has already been demonstrated for subliminal prime words (Damian, [@B14]) -- a principle termed "action triggering" (Kunde et al., [@B39]) -- at least in the within-category priming conditions a congruence effect should be found with the subliminal primes. In the subliminal conditions with primes and targets from the same category, we can expect a congruence effect, thus, making sure that our methods are sensitive enough to reveal a subliminal priming effect even if this priming effect happens to fail in the subliminal across-category (space-valence or valence-space) priming conditions. To conclude, two origins of the congruence effects are conceivable in the present context: semantic priming (or category priming) -- that is, decision priming in favor of one meaning or one category (Collins and Loftus, [@B12]; Plaut and Booth, [@B63]); and response priming (Klinger et al., [@B38]; Kunde et al., [@B39]). Semantic or category priming (Greenwald et al., [@B25], [@B24]; Naccache and Dehaene, [@B53]; Kiefer, [@B30]; Norris and Kinoshita, [@B58]; Martens et al., [@B48]) was expected in the across-category priming conditions. Response priming (and maybe semantic or category priming) was expected in the within-category priming conditions only (Damian, [@B14]; Kunde et al., [@B39]). Prior research has shown that the different origins of the congruence effect can be discriminated on the basis of their development over time \[i.e., over the reaction time (RT) distribution\]. According to Kinoshita and Hunt ([@B32]), the priming of a decision for or against one category of semantically defined targets would be reflected in a temporally stable congruence effect. This congruence effect would be present in about similar strength across all of the RT distribution, from fast to slow responses. Thus, a congruence effect based on category priming should be found in fairly equal amounts for all RTs, from the fastest to the slowest. By contrast, according to Kinoshita and Hunt ([@B32]), a congruence effect based on response priming should decrease across time (i.e., across the RT distribution) in a manner different from a category priming effect. A response priming effect should be stronger among the faster responses and it should decrease among the slower responses. To understand the origin of the expected congruence effects in the current study, we therefore tested the congruence effects as a function of the RT. Experiment 1 {#s2} ============ In Experiment 1, our participants had to categorize the clearly visible valence targets as either positive (e.g., the target "*joyful*") or negative (e.g., the target "*sad*") in a 2 (within/across-category) × 2 (congruent/incongruent) × 2 (visible/masked) design. They had to press one of two keys (left key versus right key) to classify each target. Prior to every target, a prime word was shown. In half of the trials of the across-category priming condition, the prime was a spatial up-word (e.g., the word "*above*"), and in the other half it was a down-word (e.g., the word "*below*"). Together, these were the spatial priming conditions. In the within-category priming condition, in half of the trials the prime was a word of positive valence (e.g., the prime "*happy*"), and in the remaining trials it was of negative valence (e.g., the word "*frustrated*"). Together these were the valence priming conditions. In the congruent conditions, primes and targets had associated meanings. For example, in the within-category condition a positive prime could have been presented before a positive target, while in the across-category condition an up-prime could have been presented before a positive target. In the incongruent conditions, primes and targets had less associated meanings. For example, in the within-category condition a negative prime could have been presented before a positive target, while in an across-category condition an up-prime could have been shown before a negative target. In one block, the primes were presented as clearly visible words. In another block, the primes were presented subliminally, here: masked. In this context, masking denotes an experimental procedure where a visual stimulus replaces a preceding word so as to suppress the word's visibility (Marcel, [@B47]). To ensure that the masked primes were truly subliminal, prime visibility was individually tested, and participants that were suspiciously good during the discrimination of the masked primes were excluded[1](#fn1){ref-type="fn"}. On the basis of prior research (Meier and Robinson, [@B49]), we expected an across-category, space-valence congruence effect with the visible primes. Responses should be faster in congruent than incongruent conditions. Critically, if sensory (here spatial) meaning can be extracted swiftly and (conditionally) automatically from the priming words, we might find a space-valence congruence effect in the visible *and* in the subliminal priming conditions -- that is, regardless of awareness. Method ------ ### Participants The participants of Experiment 1 and of the other experiments had normal or corrected-to-normal vision, were mostly university students and given course credit for participating. Two participants had to be excluded because of too high a number of correct prime-target judgments in the masked condition (see text footnote 1), and two further participants had to be excluded because of chance performance in the prime-target judgments of the unmasked condition, indicating that they were unable or unwilling to discriminate the unmasked prime-target relations. The remaining 40 participants (31 female, *M*~age~ = 22.0 years, age range: 18--28 years) were analyzed. ### Apparatus, stimuli, and procedure Prime and target stimuli were German words denoting emotional adjectives or prime stimuli were spatial words denoting directions or positions on the vertical axis. We used the following prime and targets that were all high frequency words because there were more than 60 instances in 1 million words (Jescheniak and Levelt, [@B28]), with frequencies word counts in parentheses according to the Wortschatz Lexikon of the University of Leipzig, <http://dict.uni-leipzig.de/>, and frequencies calculated relative to 400,000 entries, contained in the Wortschatz Lexikon on the date of retrieval, December 5, 2012). As positive valence words, we used: "*lustig*" (*jolly*; 2,521), "*glücklich*" (*lucky*; 6,097), "*freudig*" (*cheerful*; 558), "*vergnügt*" (*happy*; 483), "*spaßig*" (*funny*; 129), "*mutig*" (*brave*; 1,328), "*stolz*" (*proud*; 4,315), "*verliebt*" (*loving*; 2,300), "*fröhlich*" (*merry*; 1,862), and "*froh*" (*joyful*; 5,424), with a mean word length of Ø = 6.5 letters (range 4--9 letters) and an average frequency of Ø = 2,502. We used the following negative valence prime and target words: "*furchtsam*" (*fearful*; 59), "*ängstlich*" (*anxious*; 775), "*bekümmert*" (*worried*; 154), "*traurig*" (*sad*; 2,646), "*zornig*" (*furious*; 447), "*hasserfüllt*[^2^](#fn2){ref-type="fn"}" (*full of hate*; 339), "*wütend*" (*enraged*; 1,549), "*frustriert*" (*frustrated*; 899), "*beschämt*" (*ashamed*; 202), and "*schuldig*" (guilty; 4,753), with a mean word length of Ø = 8.3 letters (range 6--11 letters) and an average frequency of Ø = 1,182. The spatial primes that we used as up-words were: "*oben*" (*on top*; 21,453), "*darüber*" (*above*; 45,943), "*hinauf*" (2,214), "*aufwärts*" (1,652), "*empor*" (*upward*; 714), "*hoch*" (*high*; 27,559), "*gehoben*" (1,555), "*erhöht*" (*elevated*; 14,891), "*aufsteigend*" (143), "*steigend*" (*rising*; 750), with a mean word length of Ø = 6.6 letters (range 4--11 letters) and an average frequency of Ø = 11,687. Finally, the spatial primes that we used as down-words were: "*unten*" (*down*; 11,971), "*darunter*" (*below*; 22,589), "*hinab*" (1,024), "*abwärts*" (767), "*herab*" (*downward*; 1,624), "*niedrig*" (*low*; 3,529), "*gesenkt*" (*lowered*; 5,027), "*abfallend*" (see text footnote 2; 143), "*sinkend*" ("*declining*"; 60), and "*tief*" (*deep*; 10,331), with a mean word length of Ø = 6.3 letters (range 4--9 letters) and an average frequency of Ø = 5,707. These words were selected because of their relatively similar distributions in text corpora, a relatively similar length, and on the basis of their easy and equal discriminability of category-membership (which was empirically tested during pre-testing). Each of the 20 valence words was presented as a target equally often. For the creation of the within-category priming condition, each target was randomly combined with each of the nine remaining valence words. For the across-category priming condition, each of the valence targets was randomly combined with each of the 10 spatial words as a prime. Across trials, the different prime words were equally likely and the resulting prime-target pairs were equally likely to be congruent or incongruent. In a trial, prime and target were never identical even in congruent conditions. This was done to rule out repetition priming (Forster, [@B20]). All stimuli were presented in black (\<1 cd/m^2^) on a gray background (24 cd/m^2^). Each trial started with the presentation of a fixation cross centered on the screen for 750 ms (see Figure [1](#F1){ref-type="fig"}). In masked trials, a forward mask was shown next for 200 ms. It consisted of 10 randomly drawn uppercase letters. The prime word was shown for 34 ms immediately after the forward mask or after the blank screen in the case of visible trials. The prime was depicted in lowercase letters. In masked trials, the prime preceded a backward mask. The backward mask also consisted of 10 randomly drawn capital letters that were shown for 34 ms. In visible trials, both forward and backward masks were omitted and the masking screens were replaced by blank screens. Next, the target word was shown for 200 ms. In masked trials, all words and the masks were shown centered on the screen directly one after the other. Timing of all stimuli was adapted from prior studies that exhibited little prime visibility in masked and good prime visibility in unmasked conditions (Kiefer and Brendel, [@B31]; Ansorge et al., [@B3], [@B2]). {#F1} The experiment consisted of two blocked conditions, one block with masked primes, and a second block with unmasked primes. The order of the blocks was either masked block before unmasked block, or vice versa, with different block orders balanced across participants. Each block lasted about 30 min. In each trial of both the masked and the unmasked conditions, participants had two tasks, first a blocked target-response task and subsequently a blocked prime-discrimination task. During the target-response task, participants discriminated the meaning of the target word. Half of the participants pressed the right key for positive target words and the left key for negative target words. The other half of the participants pressed the left key for positive targets and the right key for negative targets. The second task was a prime visibility task. This task required that one key (say the right key) be pressed in trials, in which the prime was congruent to the target and the other key (say the left key) if the prime was incongruent to the target. The levels of the variable prime-target congruence that had to be judged were carefully explained to the participants with relevant examples in the instructions. This task was conducted in every trial, directly after the target-discrimination task, so that conditions in this task were exactly the same as in the target-response task. This task of discriminating between congruent and incongruent trials has two advantages as compared to a task of discriminating the prime's meaning (e.g., its valence) alone. First, the task of discriminating congruent from incongruent trials requires processing of prime *and* target. It thus necessitates processing of the targets not only in the target-discrimination task but also in the prime visibility test. Because we are interested in understanding whether prime visibility in the target-discrimination task might account for any priming effect in these conditions, the congruence-incongruence discrimination task is thus more apt to answer the research question that we asked. Second and related, the congruence effect in the target-discrimination task can only be explained on the basis of supraliminal word processing if the critical characteristics of the corresponding conditions that created the priming effect can be correctly discriminated by the participants. This critical characteristic that is decisive for the priming effect -- that is, whether a quick or a slow response can be given, is whether a trial is congruent or whether it is incongruent. Therefore, a fitting prime visibility test needs to assess the participants' awareness of this critical difference rather than the participants' awareness of a difference between the primes that is only related to this critical difference, such as the exact meaning of the prime word alone. In the prime visibility task, mappings of judgments to alternative (left and right) key presses were also fixed and balanced across participants. After each incorrect response to the target and if the target-discrimination RT exceeded 1,250 ms, participants received feedback about their error or their too slow responses. Feedback took 750 ms. Thus, keeping a high accuracy and a fast response was mildly rewarded (i.e., saved 750 ms per trial). No feedback was given concerning the prime visibility task. Each block consisted of 240 trials. In total (across blocks), this involved 60 trials of each combination of the 2 prime types (spatial primes; valence primes) × 2 prime-target congruence relations (congruent; incongruent). Prior to the first block, participants were carefully instructed about the target-response task and the prime-discrimination task. Also, prior to both blocks, the participants practiced the task for a minimum of 20 trials but they could also practice for another 20 trials if they wanted to practice more. During these practice phases, the procedure was explained verbatim in addition to the preceding written instructions, and in more detail if necessary (i.e., if there were questions). Results ------- Of all correct responses, 3.8% were excluded because these RTs deviated by more than 2 SDs from a respective condition's and individual participant's mean RT (with SD and mean RT computed separately for each of the conditions and individuals). An ANOVA of the medians of the correct responses, with the within-participant variables congruence (congruent; incongruent), prime type (valence; spatial), visibility (masked; unmasked), and quintile of RT distribution (first to fifth quintile) led to the following results. Here and in the subsequent analyses, results were adjusted by Greenhouse--Geisser coefficients and the ε values are reported, if Mauchly tests indicated a deviation from sphericity. A significant main effect of congruence, *F*(1, 39) = 83.81, *p* \< 0.01, partial η^2^ = 0.68, was found, reflecting faster RTs in congruent (652 ms) than incongruent (669 ms) conditions. There was also a significant main effect of prime type, *F*(1, 39) = 14.35, *p* \< 0.01, partial η^2^ = 0.27, indicating that responses after spatial primes were slightly faster (RT = 656 ms) than with valence primes (RT = 664 ms). Critically, we also found significant interactions between congruence and prime type, *F*(1, 39) = 13.93, *p* \< 0.01, partial η^2^ = 0.26, and between congruence and prime visibility, *F*(1, 39) = 10.15, *p* \< 0.01, partial η^2^ = 0.21. Splitting up the data during follow-up ANOVAs, we confirmed a congruence effect with both types of primes, a strong across-category priming effect of the spatial primes, *F*(1, 39) = 81.07, *p* \< 0.01, partial η^2^ = 0.68 (congruent RT = 644 ms; incongruent RT = 669 ms), and a smaller within-category priming effect of the valence primes, *F*(1, 39) = 7.83, *p* \< 0.01, partial η^2^ = 0.17 (congruent RT = 660 ms; incongruent RT = 668 ms). Follow-up ANOVAs split up for visible and masked primes confirmed a stronger congruence effect with visible primes, *F*(1, 39) = 76.10, *p* \< 0.01, partial η^2^ = 0.66 (congruent RT = 650 ms; incongruent RT = 673 ms), than with masked primes, *F*(1, 39) = 16.63, *p* \< 0.01, partial η^2^ = 0.30 (congruent RT = 653 ms; incongruent RT = 664 ms). There was also a trivial main effect of the variable quintile, *F*(4, 156) = 270.59, *p* \< 0.01, partial η^2^ = 0.87 (ε = 0.26), and the variable quintile also interacted significantly with congruence, *F*(4, 156) = 6.97, *p* \< 0.01, partial η^2^ = 0.15 (ε = 0.51), with prime type, *F*(4, 156) = 5.80, *p* \< 0.01, partial η^2^ = 0.13 (ε = 0.45), and in a marginally significant three-way interaction with congruence *and* prime type, *F*(4, 156) = 3.02, *p* = 0.07, partial η^2^ = 0.07 (ε = 0.38). As it can be seen in Figure [2](#F2){ref-type="fig"}, the within-category congruence effect (incongruent RT -- congruent RT) of the valence primes (depicted as circles) decreased across RTs. It was significant only in the faster RTs \[first quintile: 22 ms, *t*(39) = 5.96, *p* \< 0.01, second quintile: 15 ms, *t*(39) = 4.43, *p* \< 0.01, third quintile: 12 ms, *t*(39) = 3.61, *p* \< 0.01\] but it was absent among the slower response (fourth quintile: 0 ms, fifth quintile: −5 ms, both *t*s \< 1.00). By contrast, the across-category congruence effect of the spatial primes (depicted as crosses in Figure [2](#F2){ref-type="fig"}) was fairly stable across RT. Across the RT distribution the congruence effect varied slightly in size between 20 and 30 ms (all *t*s \> 2.70, all *p*s \< 0.01). Together, these results are perfectly in line with the assumption that response-activation was responsible for (within-category) valence priming effects but categorization lay aground of spatial (across-category) priming because the valence primes were also response-relevant but the spatial primes were not. {#F2} An ANOVA of the error rates (ERs) with the variables congruence, prime type, and visibility confirmed the picture. The slower median correct reactions in incongruent than congruent conditions were accompanied by a lower mean accuracy in incongruent (ER = 6.5%) than congruent (ER = 5.3%) conditions. This was reflected in a significant main effect of congruence, *F*(1, 39) = 7.75, *p* \< 0.01, partial η^2^ = 0.17. There was also a significant main effect of prime type *F*(1, 39) = 17.77, *p* \< 0.01, partial η^2^ = 0.31 -- with higher ERs for valence (6.6%) than spatial primes (5.2%), and a significant interaction between congruence and prime type, *F*(1, 39) = 20.62, *p* \< 0.01, partial η^2^ = 0.35. Follow-up ANOVAs split up for the type of prime confirmed the existence of a significant main effect of congruence (3.4%) for spatial primes, *F*(1, 39) = 34.11, *p* \< 0.01, partial η^2^ = 0.47, but not for valence primes (−0.7%, *F* \< 1.00). ### Supplementary information concerning polarity-correspondence effects We also made sure that the within-category congruence effect did not merely reflect a polarity-correspondence effect (Lakens, [@B40])[^3^](#fn3){ref-type="fn"}. According to this interpretation, as compared to the so-called "minus poles" (or "−poles") of a meaning dimension (e.g., negative concepts in the valence dimension and down concepts in the spatial dimension), the so-called "plus poles" (or "+poles") of meaning dimensions (e.g., positive concepts and up concepts) are processed faster and additionally facilitate +pole responses. The polarity-correspondence hypothesis would thus predict that prime-target combinations of concepts with similar polarities (e.g., +/+ prime-target combinations) are facilitated as compared to combinations with dissimilar polarities (e.g., −/+ prime-target combinations). Although this is true of the −pole concepts, too, the slower −pole processing and responses should lead to reduced congruence effects, with a congruent but very slow −/− prime-target combination being not so different of an incongruent but slightly facilitated +/− prime-target combination. As can be seen in Table [1](#T1){ref-type="table"} though, there was the expected facilitation for +targets but no strong difference between the congruence effects of +targets versus −targets. ###### **Reaction times (in ms) as a function of prime-target combination in Experiment 1 and 2**. Experiment 1 (ms) Experiment 2 (ms) ------------------ ------------------- ------------------- +/+ Prime-target 634 617 −/− Prime-target 655 623 −/+ Prime-target 645 643 +/− Prime-target 679 649 ### Prime visibility tests To test whether participants consciously identified the unmasked primes but failed to see the masked primes, we computed *d*′, a sensitive index of stimulus visibility (Reingold and Merikle, [@B65]). Individual *d*′ was computed separately for masked and unmasked primes and for spatial primes and valence primes. For our measure of *d*′ congruent trials counted as signals and incongruent trials as noise. Accordingly correct (i.e., "congruent") judgments in congruent trials figured as hits, and incorrect (i.e., "congruent") judgments in incongruent trials as false alarms (FAs). The participants were not able to discriminate the masked prime-target relations with better than chance accuracy. For the masked spatial primes, *d*′ was −0.10, *t*(39) = 1.08, *p* = 0.29, and for the masked valence primes it was 0.19, *t*(39) = 1.70, *p* = 0.10. In addition, the correlation between individual *d*′ values and individual Cohen's *D* indices of congruence effects \[calculated as (incongruent RT -- congruent RT)/SD (pooled over congruent and incongruent) RT\] indicated that there was no significant influence of residual prime visibility on the RT congruence effect for masked spatial primes, *r*(40) = 0.003, *p* = 0.99, and for masked valence primes, *r*(40) = 0.03, *p* = 0.86. The unmasked prime-target relations were successfully discriminated for the spatial primes, with *d*′ = 2.34, *t*(39) = 14.77, *p* \< 0.01, and the valence primes with *d*′ = 2.82, *t*(39) = 15.26, *p* \< 0.01. Discussion ---------- In line with a quick influence of sense-related word meaning, an across-category space-valence congruence effect was found even with masked spatial primes[^4^](#fn4){ref-type="fn"}. This across-category congruence effect was created by subliminal words because the masked primes could not be seen by the participants. The effect is in line with the predictions of the embodied cognition view that assumes that a swift and (conditionally) automatic extraction of sensory meaning from a word could occur so fast as to influence emotions and evaluations (Niedenthal, [@B56]). The effect is also in line with the observed fast and automatic extraction of non-evaluative meaning before stimulus evaluation (Nummenmaa et al., [@B59]). In line with these observations, the participants' processing of the spatial primes was so fast and efficient that an awareness of the primes was not a necessary precondition of the across-category congruence effect (Lamme, [@B42]). The present experiment's across-category priming effect of spatial words on valence discriminations might appear surprising in light of the finding of Meier and Robinson ([@B49]) that discriminating between the spatially upper or lower position of a string of crosses on the computer screen had no influence on the discrimination of the valence of a subsequent word in the center of the screen: whether the discriminated position was congruent to a word's valence or not had no systematic influence on valence discrimination in Meier and Robinson's Experiment 3. However, Meier and Robinson asked their participants to respond first to the cross positions and only then to the valence targets whereas we asked our participants to first quickly respond to the valence targets. The use of the prime and the target in different tasks can be critical for the across-category congruence effect. For example, according to an explanation developed by Gozli et al. ([@B23]), sorting prime and target into different tasks is one precondition that can lead to a reverted word-location congruence effect with short SOAs but not with long SOAs. Assuming that different RTs to the spatial primes in Experiment 3 of Meier and Robinson led to a mixture of short and long prime-target intervals, it could thus be that their lacking congruence effect reflected a mixture of straight and reverted congruence effect that averaged to zero. Whereas Meier and Robinson thus clearly sorted the cross positions and valence targets into separate tasks and created different prime-target intervals, we used a single task of discriminating a word's valence and used a fix prime-target interval. Thus conditions for a straight across-category congruence effect were better in the current experiment than in Experiment 3 of Meier and Robinson. Against our findings in Experiment 1, one might want to argue that the spatial primes were just implied by the set of task-relevant affective target words. According to this argument, the task of the participants to discriminate valenced emotional words, such as "sad" and "happy," would have led the participants to also judge spatial words, such as "above" and "below," by their respective connotative valence. In general agreement with this hypothesis, participants judge words, such as "up" as more positive, and words such as "down" as more negative (e.g., Eder and Rothermund, [@B19]). Although this alternative explanation is a theoretical possibility it would be difficult to reconcile this alternative explanation with the obvious differences between the congruence effects of the emotional primes and the spatial primes in the current study. Whereas the emotional primes led to a congruence effect that decreased across the RT distribution, this was not the case for the spatial primes. Therefore, the across-category congruence effect of the spatial primes was probably due to the priming of the target's categorization. This was reflected in the development of the across-category congruence effect over time. The RT distribution reflected a fairly stable congruence effect of the spatial primes. According to Kinoshita and Hunt ([@B32]) this would be typical of a category priming effect. In addition, the spatial words were also response-irrelevant in the first place because these words were not used as response-relevant targets anyway. We also observed a weaker within-category congruence effect of the valence primes. This within-category priming effect probably reflected response-activation. To note, the valence words were used as primes and targets. Therefore, the valence primes were response-relevant. In line with this interpretation, the congruence effect of the valence primes decreased over RTs. It was stronger among the faster than the slower responses. This is typical of the response-activation effects of masked primes (Kinoshita and Hunt, [@B32]; Ansorge et al., [@B3]). Finally, a significantly stronger congruence effect was found for visible than masked primes. The stronger influence of the visible primes probably reflected that their perceptual traces lingered longer and opportunities for attentional dwelling were thus higher with visible primes than masked primes. In line with this interpretation, when Santiago et al. ([@B68]) directed their participants' attention to the task-irrelevant valence meaning of words, they found a significant space-valence congruence effect that was absent when attention was not directed toward word valence. Experiment 2 ============ Experiment 2 was our second test of the across-category priming effect. Some evidence suggests that the semantic content from images is extracted before an image can be evaluated (Nummenmaa et al., [@B59]). According to this line of thinking, the quick awareness-independent across-category congruence effect could be abolished when the roles of valence words and spatial words as primes and targets are reversed (as compared to Experiment 1). We therefore reversed the roles of valence words and spatial words as across-category primes and targets. In contrast to Experiment 1, spatial words were now used as visible targets. The participants had to discriminate between up targets and down targets. As in Experiment 1, valence words as well as spatial words were used as primes. In this way, we were able to test whether masked valence primes created an automatic across-category space-valence congruence effect as would be predicted by a quick and automatic valence assessment of the words. Method ------ ### Participants Four participants had to be excluded based on their above-chance discrimination of the masked primes^1^, and one because of very low performance even in the unmasked condition. The remaining 39 participants (29 female, *M*~age~ = 26.3 years, age range: 19--42 years) were analyzed. ### Apparatus, stimuli, and procedure These were the same as in Experiment 1, except for the changed targets and instructions. In Experiment 2, the participants were presented with space targets and they had to discriminate between up and down targets in the target-response task. Half of the participants responded to up targets by a right-hand key press and to down targets by a left-hand key press. The other half of the participants got the opposite mapping. Results ------- ### Target-response task See also Figure [3](#F3){ref-type="fig"}. Of all correct responses, 4.0% were discarded by the same criterion as was used in Experiment 1. We ran an ANOVA of the correct RTs with the within-participant variables congruence, prime type, prime visibility, and quintiles. This ANOVA confirmed the significant main effects of congruence, *F*(1, 38) = 90.87, *p* \< 0.01, partial η^2^ = 0.71, and an almost significant effect of prime type, *F*(1, 38) = 3.79, *p* = 0.06, partial η^2^ = 0.09, as well as the important interactions between (1) congruence and prime type, *F*(1, 38) = 23.21, *p* \< 0.01, partial η^2^ = 0.38, (2) congruence and prime visibility, *F*(1, 38) = 5.34, *p* \< 0.05, partial η^2^ = 0.12, and (3) congruence, prime type, and prime visibility, *F*(1, 38) = 4.62, *p* \< 0.05, partial η^2^ = 0.11. Splitting up the data for follow-up analyses for different combinations of the steps of the variables congruence, visibility, and prime type (while collapsing across quintiles), we found significant congruence effects (incongruent RT -- congruent RT) only for visible primes of either type, across-category valence primes \[18 ms, *t*(38) = 6.25, *p* \< 0.01\] and within-category space primes \[30 ms, *t*(38) = 6.63, *p* \< 0.01\], as well as for masked within-category space primes \[26 ms, *t*(38) = 7.72, *p* \< 0.01\]. However, there was no significant congruence effect with the masked across-category valence primes \[4 ms, *t*(38) = 1.17, *p* = 0.25\]. {#F3} Crucially, and in line with different origins of the congruence effects in the different prime type conditions, the ANOVA also revealed significant two-way interactions between quintile and congruence, *F*(4, 152) = 5.39, *p* \< 0.01, partial η^2^ = 0.12, and between quintile and prime type, *F*(4, 152) = 6.26, *p* \< 0.01, partial η^2^ = 0.14, as well as a significant three-way interaction between quintile, congruence, and prime type, *F*(4, 152) = 6.61, *p* \< 0.01, partial η^2^ = 0.15. Figure [3](#F3){ref-type="fig"} depicts these results. As it can be seen by looking at the cross symbols (depicting the spatial primes), in line with a motor priming effect of the spatial primes their congruence effect now decreased over RTs. In the spatial priming conditions, congruence effects were 37, 34, 30, 25, and 12 ms, all *t*s(38) \> 2.20, all *p*s \< 0.05, from the first to the fifth quintile, respectively. By contrast, looking at the circular symbols (depicting the valence primes), especially in the visible conditions, in line with a category priming effect, congruence effects of the visible valence primes were relatively similar over RTs. From the first to the fifth quintile the across-category valence priming effect fluctuated between 9 and 12 ms, all *t*s(38) \> 2.20, all *p*s \< 0.05. Figure [3](#F3){ref-type="fig"} also indicates that a four-way interaction between congruence, prime type, visibility, and quintile should have obtained -- basically because the absence of the congruence effect with masked valence primes only -- but this interaction fell short of significance, *F*(4, 152) = 2.10, *p* = 0.08, partial η^2^ = 0.05. In addition we observed an interaction between quintile and prime visibility, *F*(4, 152) = 2.89, *p* = 0.07, reflecting a shallower slope (i.e., less variance) of the RT distribution in masked than in visible conditions, as well as a trivial main effect of quintile, *F*(4, 152) = 853.30, *p* \< 0.01, partial η^2^ = 0.96. In an ANOVA of the mean ERs, with the variables congruence, prime type, and visibility the main effects of congruence, *F*(1, 38) = 6.18, *p* \< 0.05, partial η^2^ = 0.14 (congruent ER = 3.6%; incongruent ER = 4.3%), and prime type, *F*(1, 38) = 9.58, *p* \< 0.01, partial η^2^ = 0.20 (spatial prime: ER = 4.5%; valence prime: ER = 3.4%), were also significant. These effects made clear that the congruence effect was not due to a speed-accuracy trade-off, whereas the faster RTs to spatial primes (compared to valence primes) came at the expense of higher ERs (for spatial primes than for valence primes). In addition, the two-way interaction of prime type and congruence was significant, *F*(1, 38) = 9.55, *p* \< 0.01, partial η^2^ = 0.20, and there was a significant three-way interaction, *F*(1, 38) = 4.45, *p* \< 0.05, partial η^2^ = 0.11. Follow-up *t*-tests to compare congruent with incongruent ERs, conducted separately for the different combinations of prime types and prime visibility, revealed that the three-way interaction reflected the same tendencies that we observed in the RTs, standard congruence effects (with advantages in congruent relative to incongruent conditions) in all priming conditions (unmasked/spatial primes: 1.1%; masked/spatial primes: 2.3%; unmasked valence primes: 0.3%) but a reverse congruence effect for masked valence primes (−1.7%). ### Supplementary information concerning polarity-correspondence effects This time, we did not even find the expected facilitation of +pole targets as compared to −pole targets and no difference in the respective congruence effects of these targets alike. For the results see Table [1](#T1){ref-type="table"}. ### Prime visibility tests The masked primes were invisible as demonstrated by the participants' mean chance performance. Mean *d*′ was not significantly different from zero. It was 0.02, *t* \< 1.00, with the masked spatial primes, and it was 0.10, *t*(38) = 1.19, *p* = 0.24, with the masked valence primes. In addition, the correlation between individual *d*′ values and individual Cohen's *D* indices of congruence effects demonstrated that the residual visibility of masked primes did not significantly affect RT congruence effects of masked spatial primes, *r*(39) = −0.07, *p* = 0.67, and of masked valence primes, *r*(39) = 0.07, *p* = 0.67. The same participants were able to discriminate between the unmasked prime-target relations. Mean *d*′ was 2.70, *t*(38) = 18.82, *p* \< 0.01, with the unmasked spatial primes, and it was 3.43, *t*(38) = 30.87, *p* \< 0.01, with the unmasked valence primes. Discussion ---------- In Experiment 2, we found no significant across-category congruence effect based on masked valence primes. This is in contrast to the predictions based on quick (conditionally) automatic affective processes influencing semantic analysis and it is also in contrast to Experiment 1 in which we found an across-category congruence effect of the masked spatial primes. Together, the results point to an asymmetry between the sensory and the affective processing of the word meanings. Also, in line with prior research, an across-category effect of the visible valence primes showed that if strategic processing (or other forms of awareness-dependent processing) was allowed, we replicated the standard across-category congruence effect with the valence primes, too. Evidently, only the awareness-independent aspect of the valence-based across-category priming was prevented. This result fits well with recent findings from vision sciences, where it was found that image content is partly available before its evaluation (Nummenmaa et al., [@B59]). In addition, our RT distribution analysis was suggestive of a category-based congruence effect of the valence primes and of a response-activation effect of the spatial primes. The congruence effect of the spatial primes decreased with an increasing RT which is typical of a response-activation effect (Kinoshita and Hunt, [@B32]; Ansorge et al., [@B3]). By contrast, the congruence effect of the visible valence primes was approximately the same for the different quintiles of the RT distribution which is the finger print of a category congruence effect (Kinoshita and Hunt, [@B32]). Despite this qualitative similarity of the result patterns in Experiments 1 and 2, there were also a few important differences. First, the congruence effect of the visible spatial primes in the present experiment did not decrease to zero with an increasing RT, whereas the congruence effect of the visible valence primes in Experiment 1 was completely eliminated among the slowest responses. Second and related, in the current experiment, the residual congruence effect of the visible spatial primes in the slowest responses was of about the same size as that of the visible valence primes. These differences might reflect more average semantic congruence between different spatial words than between different emotional words, and could reflect unique sources of meaning differences between the congruent emotional words. Even valence congruent emotions, such as sadness and anger (both negative) or pride and loving (both positive) vary drastically according to further word meaning dimensions, like arousal (Wundt, [@B77]; Russell, [@B67]; which would be low for sadness but high for anger). These differences might have counteracted the valence-based congruence effect but no such diminishing influence seems to have been present with the spatial words. Experiment 3 {#s1} ============ It is also possible that the across-category priming effect of the spatial primes in Experiment 1 reflected a type of intention-independent or truly stimulus-driven priming effect instead of a conditionally automatic across-category congruence effect. Experiment 3 was therefore an additional control experiment. The control experiment was conducted to experimentally rule out an interpretation of Experiment 1's across-category congruence effect in terms of a strongly automatic, bottom-up priming effect. As explained above, we assumed that Experiment 1's across-category congruence effect probably reflected that the prime's sensory meaning affected the task-relevant categorical evaluation of the valence targets as negative versus positive. If this was the case, it should be possible to abolish the across-category congruence effect. Past research has shown that participants can flexibly change their prime analysis in accordance with the instructions and the changing target categorization requirements (Klinger et al., [@B38]; Eckstein and Perrig, [@B17]; Norris and Kinoshita, [@B58]). For example, Klinger et al. ([@B38]) asked their participants to classify the same visible targets as either positive versus negative targets in one condition but as animate versus inanimate targets in a second condition. These authors found that only the prime word meaning that was currently relevant for classifying the targets created an awareness-independent congruence effect. For example, if the participants classified the target words on the basis of the target's valence, a target-congruently evaluated prime facilitated responses as compared to a target-incongruently evaluated prime. By contrast to this, it did not matter whether both prime and target were of the same animate or inanimate category or whether one denoted an animate object and the other an inanimate object. This pattern of results was reversed when the participants had to classify the targets on the basis of the targets' category-membership to the categories of animate versus inanimate objects. Now the valence-based congruence effect was eliminated but a congruence effect on the basis of the status of the primes as names for animate versus inanimate objects was found. From this it follows that we should be able to abolish a conditionally automatic across-category congruence effect when we no longer require categorization of the positive versus negative targets as two different categories. Here, we achieved this by changing the instructions and asking the participants to categorize both kinds of valence targets, negative and positive words, as belonging into the same class of objects. To that end, we used both valence and space words as primes *and* targets and asked our participants to categorize the targets into emotional adjectives on the one hand and into spatial prepositions on the other. Under these conditions, both negative and positive words belong to the same category. Hence, a conditionally automatic priming effect of the different spatial primes (i.e., up- versus down-words) on the categorization into negative and positive target words should be abolished because the difference between the emotional valences would no longer be relevant for the task at hand (Klinger et al., [@B38]; Eckstein and Perrig, [@B17]). Method ------ ### Participants For the new experiment, one participant with epilepsy was not tested, and another eight participants failed on the prime-discrimination criterion^1^ as in Experiments 1 and 2. The remaining thirty participants (25 female, *M*~age~ = 24.1 years, age range: 21--38 years) were analyzed. ### Apparatus, stimuli, and procedure These were the same as in Experiment 1, except for the following differences. First, in the target-response task, participants had to discriminate between valence targets and spatial targets. Thus, in contrast to Experiment 1, all words were used as targets and as primes. However, to prevent combinatorial explosion, we used only the across-category prime-target combinations of major interest. (During prime-discrimination, participants had to judge whether prime-target pairs were space-valence congruent or whether they were space-valence incongruent.) ### Target-response task For the results, see also Figure [4](#F4){ref-type="fig"}. Of all responses, 4.2% were excluded by the same criterion as was used in Experiments 1 and 2. {#F4} In contrast to the preceding experiments, in an ANOVA with the variables congruence (space-valence congruent versus incongruent), prime type/target type (valence primes/spatial targets; spatial primes/valence targets), prime visibility (masked prime versus visible prime), and quintiles of the RT distribution (first to fifth quintile) there was no significant main effect of congruence, *F*(1, 29) = 1.08, *p* = 0.31, partial η^2^ = 0.04, and there was neither a significant interaction of congruence and prime type, *F* \< 1.00, nor of congruence and prime visibility, *F* \< 1.00. The same was true of the other interactions with the variable congruence, all *p*s \> 0.15. What we found were significant main effects of prime type, *F*(1, 29) = 12.65, *p* \< 0.01, partial η^2^ = 0.30, of prime visibility, *F*(1, 29) = 4.75, *p* \< 0.05, partial η^2^ = 0.14, and of quintile, *F*(4, 116) = 298.31, *p* \< 0.01, partial η^2^ = 0.91, ε = 0.27. Responses were faster for spatial primes/valence targets (RT = 695 ms) than for valence primes/spatial targets (RT = 715 ms) and for visible primes (RT = 697 ms) than for masked priming conditions (RT = 714 ms). Also, RT increased over the RT distribution. In addition, there was a significant Prime Type/Target Type × Quintile interaction, *F*(4, 116) = 4.84, *p* \< 0.05, partial η^2^ = 0.14, ε = 0.47, that was due to a steeper increase of the curves for space primes/valence targets than for valence primes/spatial targets. To put it differently, the faster RTs for spatial primes/valence targets that we observed in a main effect of prime type/target type were stemming from the fastest responses, whereas there were no large differences between the RTs in the different prime type conditions among the slower responses. In the ANOVA of the ERs, we observed a significant interaction of congruence and prime type, *F*(1, 29) = 4.68, *p* \< 0.05, partial η^2^ = 0.14. *Post hoc* *t*-tests showed that this was due to a non-significant "standard" congruence effect (0.7%), *t*(29) = 1.66, *p* = 0.1, for the spatial priming/valence target conditions and an almost significant "reverse" congruence effect (−1.1%) in the valence prime/spatial target conditions, *t*(29) = 1.89, *p* = 0.07. In addition, a significant interaction of prime type and visibility, *F*(1, 29) = 7.64, *p* \< 0.01, partial η^2^ = 0.21, reflected that at least with the valence primes the faster RTs in unmasked than masked conditions (see RT ANOVA above) came at the expense of a higher ER in unmasked (7.2%) than masked (5.1%) conditions, *t*(29) = 2.50, *p* \< 0.05, whereas with spatial primes no such difference was found \[spatial primes: unmasked ER = 5.1%; masked ER = 5.4%, *t*(29) = 1.66, *p* = 0.11\]. The main effects, all non-significant *F*s \< 2.60, all *p*s \> 0.19, all partial η^2^s \< 0.09, and the remaining interactions, all non-significant *F*s \< 1.10, all *p*s \> 0.32, all partial η^2^s \< 0.04, were not significant. ### Prime visibility tests Again, the participants were unable to discriminate the masked prime-target pairs. For the masked spatial primes, *d*′ amounted to 0.07, *t* \< 1.00. For the masked valence primes, *d*′ was 0.13, *t*(29) = 1.51, *p* = 0.14. Participants were capable of discriminating the prime-target relations with the unmasked primes. For the unmasked spatial primes, *d*′ was 1.53, *t*(29) = 6.65, *p* \< 0.01. For the unmasked valence primes, *d*′ amounted to 2.08, *t*(29) = 7.80, *p* \< 0.01. Discussion ---------- In Experiment 3, we wanted to rule out that the masked priming effect reflected a strongly automatic stimulus-driven effect. This was tested in this control experiment, in which all valence targets required one response and all spatial targets the alternative response. Under these conditions, if the masked priming effect was conditionally automatic, the spatial meaning of the primes should not have facilitated a categorization of the valence targets into negative versus positive words because valence discrimination was no longer required. As a consequence, the typical conditionally automatic categorization-congruence effect was expected to disappear because the categorization-congruence effect critically depends on an appropriate top-down categorization criterion on the side of the participants (Klinger et al., [@B38]; Eckstein and Perrig, [@B17]; Norris and Kinoshita, [@B58]). An alternative prediction, however, was made for the control experiment if the congruence effect was due to stimulus-driven priming. If the spatial primes activated particular valence meaning in the preceding experiments in a stimulus-driven way, the congruence effect of the primes should have been found in the control conditions of Experiment 3, too, because the same primes and targets as in the preceding experiments were used. At variance with this prediction, however, our results indicated that no congruence effect could be found in the present experiment. Thus, the data were much better in line with an origin of the across-category congruence effect in Experiments 1 and 2 via facilitation of the task-dependent target-category classification into positive versus negative words than via the stimulus-driven priming of one particular meaning. Note also that the current experiment ruled out that the across-category priming effect in the supraliminal control conditions of Experiments 1 and 2 was an artifact of our procedure to ask the participants for a prime-target congruence judgment in the prime visibility test. One might want to argue that asking the participants to categorize primes and targets as congruent or incongruent in the across-category priming conditions created the supraliminal across-category priming effect (in the second blocks -- that is after the first prime visibility test at the end of the first block) of the preceding Experiments 1 and 2 in the first place. If this would have been the case, the same across-category congruence effect should have been found in the supraliminal conditions of the present experiment because the same prime-target visibility judgment as in Experiments 1 and 2 was also required in the present Experiment 3. However, in contrast to this prediction, no across-category congruence effect could be found in the supraliminal conditions of the present experiment either. These results are better in line with an explanation of the across-category congruence effect in terms of the facilitating and/or interfering influence of spatial meaning during the participants' discrimination between different positive and negative word valences. General Discussion ================== Since about 20 years, an increasing number of experiments firmly established the existence of a space-valence congruence effect (Lakoff and Johnson, [@B41]). The space-valence congruence effect reflects an advantage for the classification of and responses to combinations of associated (or congruent) spatial and affective meaning, such as the classification of positive words at elevated locations, as compared to less associated (or incongruent) meaning, such as negative words at elevated locations (Meier and Robinson, [@B49]). Such across-category congruence effects are potentially very informative with respect to the connection between cognition and emotion (Eder et al., [@B18]). One particular question that haunts researchers in this domain is the sequence of events during semantic analysis of words in general and the meaning-connected evaluations in particular. On the one hand, evaluations of words and objects are very swift and automatic (Arnold, [@B6]) and they can occur before or outside of awareness (Naccache et al., [@B54]). On the other hand, some semantic information also seems to lay aground of subsequent evaluations and therefore some forms of non-evaluative meaning extraction might have to precede evaluations (Storbeck and Clore, [@B74]). In line with this view, non-evaluative semantic classifications are sometimes faster than evaluative classifications (Nummenmaa et al., [@B59]) and non-evaluative and evaluative semantic classifications can both occur independently of awareness, too (Kiefer, [@B30]). To investigate these issues during lexical access to word meaning in general and the case of the space-valence association in particular, we used subliminal priming with words as primes and targets. Subliminal priming of words allows measuring of an awareness-independent quick and (conditionally) automatic congruence effect based on the degree of congruence or fit between subliminal priming word and to-be-classified target word. Here, this method was used to investigate one particular source of non-evaluative semantic impact on valence assignments that has been emphasized as important by the defenders of an embodied cognition view on emotions: the impact of sense-related (or sensory) non-evaluative meaning on emotions (Niedenthal, [@B56]). If the quick extraction of sensory non-evaluative meaning impacts on emotions, we expected an across-category priming effect of subliminal (masked) spatial prime word meaning on the classification of the valence of the visible target words. A corresponding awareness-independent across-category congruence effect was found in the present Experiment 1. In addition, if evaluations also occur so quickly as to influence non-evaluative semantics used in sense-related classifications of words, we expected a similar congruence priming effect of subliminal valence words on the classification of the spatial elevation-meaning of visible target words. In contrast to this prediction, however, this across-category congruence effect was not found with subliminal valence word primes (Experiment 2). The across-category congruence effect of the valence prime words on the categorization of the spatial target words was only found with clearly visible supraliminal valence primes. The latter effect presumably reflected a strategic processing elicited by the prime words themselves and was therefore dependent on the use of clearly visible valence prime words (Forster, [@B20]). Such strategic processing of the visible prime words concerns the alteration of the prime processing that owes to their conscious recognition. For example, visible primes might have offered more opportunities for the participants' attentional dwelling on their meaning than masked primes and the amount of attention indeed seems to be an important mediator for space-valence congruence effects (Santiago et al., [@B68]). Related, once the participants can see the primes, participants might want to learn whether the prime categories predict the target categories. If the participants adopt such a strategy, they will be willingly processing both the meaning of the prime words and that of the target words. As a consequence, the prime meaning can then influence processing of the target meaning even in valence-space across-category prime-target word pairings where this kind of influence would otherwise not be possible. Together, the significant across-category congruence effect of the supraliminal primes and the absent congruence effect of the subliminal primes in the conditions with valence primes and spatial targets also made clear that a qualitative difference existed between aware and unaware processing modes. In addition, in Experiments 1 and 2, we also found motor-activation effects of the primes in the within-category priming conditions. Here, the prime words were from the same set as the target words so that the prime words had the power to elicit a target-associated response alternative (Klinger et al., [@B38]; Kunde et al., [@B39]). In these conditions, a quickly dissipating congruence effect was found, also for the subliminal valence words (see [Experiment 1](#s2){ref-type="sec"}). The fact that this congruence effect decreased over RT was in line with its assumed origin on a response-activation level (Kinoshita and Hunt, [@B32]). Note that this kind of response-activation effect probably reflected the task-dependent motor meaning of the words. Therefore, these motor priming effects could not account for the across-category congruence effects in Experiments 1 and 2 in which the priming words were from a different category than the response-relevant target and were never used as response-relevant stimuli (Naccache and Dehaene, [@B53]; Klauer et al., [@B36]). In addition, no stimulus-driven automatic across-category priming effect could be found once the valence discrimination was no longer required (Experiment 3). Moreover, two further observations suggested that the awareness-independent across-category congruence effect reflected an influence of the primes on the targets' semantic categorization. First, the fact that the across-category congruence effect in Experiment 3 was absent is in line with its origin on the level of a category priming effect because subliminal semantic category priming effects are conditional on a fitting task set (Kunde et al., [@B39]; Eckstein and Perrig, [@B17]; Norris and Kinoshita, [@B58]). Second, in Experiments 1 and 2 the across-category congruence effect in the subliminal and supraliminal priming conditions were more or less of a similar strength across the RT distribution (Kinoshita and Hunt, [@B32]). Jointly, our data were thus in line with a swift and awareness-independent influence of semantic processes on evaluations in general (Storbeck and Clore, [@B74]; Nummenmaa et al., [@B59]), and of sense-related non-evaluative meaning on emotions in particular (Niedenthal, [@B56]). This is not to say, however, that awareness-independent processing of non-evaluative meaning always has to precede evaluative processing. To note, we have studied the sequence of non-evaluative versus evaluative meaning extraction with regard to only one particular class of stimuli (words) and one particular type of non-evaluative meaning (spatial meaning). We can therefore not tell whether a similar sequence holds with other stimuli and alternative types of meaning. It is possible for example that specific valence stimuli, such as emotional facial expressions, are processed as quickly or even quicker than certain stimuli with a non-evaluative meaning. In line with this assumption, for example, participants are able to process subliminal facial expressions (Jolij and Lamme, [@B29]; Smith, [@B70]) even without a specific categorization task (Naccache et al., [@B54]). In fact, sensory or sensorimotor representations of affective facial expressions could be one valence-specific way of the embodiment of evaluative meaning (Niedenthal, [@B56]) -- that is, the distinction of evaluative and non-evaluative meaning might not be feasible with regard to stimuli, such as human faces. One further aspect that requires a brief discussion is our visibility measure. One might argue that with the current procedure of asking the participants to classify prime-target pairs after the preceding target responses, we could have underestimated the masked primes' visibility and prime judgments might have been influenced by the fluency of the target responses (e.g., prime judgments might have occurred earlier in congruent than incongruent conditions, allowing for more forgetting of priming information in incongruent conditions). However, when exactly the same timing, sequence, size, luminance, and positioning of the primes and masks were used with different prime visibility tasks in a preceding study, the same results were found: prime visibility is also zero if the prime judgments are given in separate blocks of trials without preceding target responses, and if the prime judgment concerns the classification of the primes alone instead of the prime-target categorization (Ansorge et al., [@B3]). Thus, we are confident that the present finding of prime invisibility in the masked priming conditions is not just an artifact of the particular method that we have chosen. As a final cautionary remark, however, an asymmetry of the masked priming effect (of space concepts on valence but not of valence on space) as we have found it might also be in line with some characteristics of language use, such as frequent metaphorical reference to abstract concepts (here: valence) by more concrete concepts (here: spatial meaning; Lakoff and Johnson, [@B41]; Santiago et al., [@B68]), whereas the asymmetry of masked priming effects might be at variance with some particular explanations of the space-valence congruence effect (Walsh, [@B75]; Meier and Robinson, [@B49]). Conclusion ========== The present study shows that subliminal spatial words affected classification of associated valence word meaning but that there was no corresponding influence of subliminal valence words on the classification of spatial word meaning. This was different with supraliminal words where the influences of space on valence and of valence on space were reciprocal. Jointly, these data suggest that spatial word meaning can precede access to valence word meaning to create space-valence associations in word understanding, and that space-valence meaning associations with aware and unaware words are owing to partly different processing strategies. Author Note =========== The authors declare shared first-authorship of Ulrich Ansorge, Faculty of Psychology, University of Vienna, Vienna, Austria, and Shah Khalid, Institute of Cognitive Science, University of Osnabrück, Osnabrück, Germany. Supported by Deutsche Forschungsgemeinschaft Grant An 393/5-1 to Ulrich Ansorge, Werner Klotz and Ingrid Scharlau, and by European Research Council's ERC-2010-AdG \#269716 -- MULTISENSE to Peter König. Thanks to Daniela Egerer, Katharina Sehling, and Caroline Rott for help with the data collection. Conflict of Interest Statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ^1^To ensure that the primes were not seen, individual *d*′ scores were computed as a participant's difference between her/his *z*-transformed hit rate and the *z*-transformed FA rate during prime-discrimination. The *d*′ score becomes zero for chance performance and it can infinitely increase with an increasing number of correct judgments. Participants with an individual *d*′ score above the confidence interval, calculated according to equation (1) (Macmillan and Creelman, [@B46]) were excluded from the analysis because of their potential capability to see the masked prime: $$\Phi\left( p \right) = \left( {2\pi} \right)^{- {1/2}}\exp\left\lbrack {- 0.5z\left( p \right)^{2}} \right\rbrack.$$ ^2^Word counts of "*hasserfüllt*" included the word "*hass*" ("*hatred*") and word counts of "*abfallend*" included the word "*fallend*" ("*falling*"). ^3^We are grateful to Daniël Lakens for pointing these predictions out to us. ^4^We were concerned that the congruence effect of the subliminal spatial primes could have reflected transfer of priming effects from preceding unmasked to masked blocks. Therefore, a complementary ANOVA was run with the additional between-participants variable block sequence (masked primes first vs. unmasked primes first), and the within-participant variables congruence, prime type, and visibility, as before. In addition to the significant results of the primary ANOVA, we observed a significant three-way interaction of sequence, prime type, and prime-target congruence, *F*(1, 38) = 4.19, *p* \< 0.05, and partial η^2^ = 0.10. This three-way interaction reflected that a significant congruence effect (incongruent RT minus congruent RT) was found for all prime types \[all significant congruence effects \> 11 ms; all significant *t*s(19) \> 2.40, all significant *p*s \< 0.05) but for the affective primes when unmasked primes were presented in the first block \[congruence effect = 8 ms; *t*(19) = 1.35, *p* = 0.19\]. Thus, the across-category priming effect of the masked spatial primes was not due to transfer from preceding visible priming blocks only. [^1]: Edited by: Mattie Tops, VU University Amsterdam, Netherlands [^2]: Reviewed by: Daniel Lakens, Eindhoven University of Technology, Netherlands; Carolin Dudschig, University of Tübingen, Germany [^3]: This article was submitted to Frontiers in Cognition, a specialty of Frontiers in Psychology.

**DEAR EDITOR** We would like to describe a catastrophic case of skin necrosis in a heavy smoker patient after facelift surgery which was managed with non-surgical wound care and conservative treatment. Our patient is a 57-year-old smoker woman presented with bilateral extensive ulcers on the both sides of the face with sharp and distinct margins. She had undergone face lift procedure bilaterally 2 months earlier in another center. In the early postoperative period she started smoking again, which then ischemic blisters and discoloration with undermined skin happened, which was a sign of impending necrosis to occur. Eventually complete necrosis and skin loss happened, which was subsequently debrided by the surgeon 3 weeks postoperatively. Although she was seeking a rapid magic solution for her problem, the lesion was a dramatic subtotal skin loss. She did not have any relevant past history, including rheumatologic or collagen vascular disease. The lesion was painless and had a clean red base. She was a heavy smoker 30 pack/year of 30 years of 1 packs a day of cigarette smoking. The evaluations were pointing towards the diagnosis of flap necrosis which has happened because of insufficient blood supply due to continuous cigarette smoking. During face lift procedures, the facial perforating arteries are frequently cut, hence the blood is only reached from subdermal plexus of the adjacent intact skin.^[@B1]^ In patients with poor oxygenation, like this case, flap circulation can be compromised and due to stagnation of blood flow and thrombosis of microcirculation of venous and arterial systems, subsequent skin necrosis may occur. Deciding about which treatment option to consider for these patients is not an easy one, as various options are available, including skin grafting, tissue expansion, flap transfer, keratinocyte graft with skin substitute coverage and healing by secondary intension, but the last one seems to be the most efficacious one in these patients, as chosen in this patient.^[@B2],[@B3]^ In performing the procedures which require good healing, careful attention to the history of cigarrete smoking and other medical conditions that compromise vascular supply including the vasculitides and vascular insufficiencies are important for proper wound healing. As a surgeon, it is imperative to discuss the possibility of occurrence of necrosis like this one in patients with insufficient blood supply. Direct questioning of the history of cigarrete smoking and its quantity should never be hesitated before performing these surgeries. Written consent should be obtained from the patient and to signify the possible complications that may occur postoperatively. The patient was advised to stop smoking to help the skin to be healed by wound contraction. She decreased her smoking from more than 20 cigarettes a day to almost 3, and after 4 months the lesion shrinked in size considerably ([Figure 1](#F1){ref-type="fig"}). Downsizing of the lesion has had a good positive reinforcement for her to stop smoking. However, after almost 4 weeks of not smoking, she reinitiated smoking but to a lesser degree, up to 5 cigarettes a day. {#F1} CONFLICT OF INTEREST ==================== The authors declare no conflict of interest.

Background {#Sec1} ========== Salts, particularly NaCl, can be toxic to plants through inhibition of important biochemical and physiological processes, such as protein synthesis, photosynthesis, and enzymatic reactions, after moving into the cytosol from soils \[[@CR1]\]. While salt stress can inhibit plant growth and development, many types of plants are able to grow in high salinity environments because they have complex mechanisms that facilitate adaptation to salinity stress \[[@CR2]\]. Of these mechanisms, the ability to transport excess Na^+^ out of cells is critical to salt tolerance. SOS1 (salt overly sensitive 1) is a Na^+^/H^+^ antiporter and the only Na^+^ efflux protein present in plant plasma membranes (PMs) characterized to date. SOS1 mediates extrusion of Na^+^ through a proton gradient generated by the H^+^-ATPase (AHA) in the PM \[[@CR3]\]. Therefore, SOS1 and AHA are two key plant halotolerance factors. PM H^+^-ATPase is encoded by a large family of genes \[[@CR4], [@CR5]\]. Bioinformatics analyses of *Arabidopsi*s and genomic sequences of rice revealed the presence of 11 and 10 PM AHAs, respectively \[[@CR6], [@CR7]\]. Of these AHAs, NaCl treatment induced expression of three*, AtAHA1, AtAHA2,* and *AtAHA3*, in *Arabidopsis* \[[@CR8]\]. The transcript levels of PM AHA were found to be higher in a salt-tolerant poplar than a salt-sensitive poplar \[[@CR9]\]. In addition, PM AHA mRNA is more abundant in halophytes than glycophytes \[[@CR10], [@CR11]\]. Salinity causes upregulation of PM *AHA* gene expression, as well as accelerates protein biosynthesis and H^+^-pumping activity in some plants \[[@CR12]--[@CR14]\]. AHA in a salt-tolerant rice species has higher activity than in a salt-sensitive rice species \[[@CR15]\]. An *Arabidopsis* PM AHA4 mutant has dramatically reduced growth when exposed to salt stress compared to WT \[[@CR16]\]. Expression of a constitutively activated PM AHA lacking the autoinhibitory domain in transgenic tobacco plants increases salt tolerance compared to untransformed plants \[[@CR17]\]. *SOS1* genes have been found in many plants \[[@CR18]--[@CR25]\]. Of these, *Arabidopsis* SOS1 (AtSOS1) was the first PM Na^+^/H^+^ antiporter to be thoroughly physiologically, biochemically, and molecularly characterized \[[@CR18], [@CR26]\]. Exposure to salinity stress increases *SOS1* transcript abundance in wheat plants \[[@CR19]\], induces the accumulation of *SOS1* mRNA in rice plants \[[@CR27]\], and causes upregulation of *SOS1* transcription in *Arabidopsis* \[[@CR28]\]. Under high salt conditions, *SOS1* mRNA levels are higher in *Thellungiella salsuginea* (a halophytic *Arabidopsis*-relative plant) than *Arabidopsis* \[[@CR20]\]. Mutant *Arabidopsis* plants lacking SOS1 are extremely sensitive to salt stress \[[@CR18], [@CR29]\]. *Thellungiella salsuginea* lines expressing SOS1-RNAi (RNA interference) are sensitive to salt \[[@CR20]\]. The salt sensitivity of an *Arabidopsis sos1* mutant can be overcome by transforming in native or other plant *SOS1* genes \[[@CR27], [@CR28]\]. *Arabidopsis* overexpressing *AtSOS1* is more salt tolerant than WT plants \[[@CR30]\]. Expression of wheat SOS1 (*TaSOS1*) in transgenic tobacco plants improves their growth following NaCl treatment \[[@CR31]\]. SOS1 uses the proton gradient established by PM AHA to exchange Na^+^ for H^+^ across the PM \[[@CR3], [@CR27]\]. The aforementioned data indicate the PM Na^+^/H^+^ antiporter SOS1 and H^+^-ATPase AHA are involved in plant salt tolerance, where an Na^+^/H^+^ antiporter utilizes the proton gradient generated by H^+^-ATPase to move Na^+^ from the cytoplasm to the external medium and help plant cells maintain non-toxic cytosolic concentrations of Na^+^. Therefore, theoretically, coordinating SOS1 and AHA could enhance Na^+^ extrusion, where co-expression of these two genes should confer better tolerance to salinity to transgenic plants. However, it has not been reported whether SOS1 and AHA1 function cooperatively in transgenic plants to more efficiently improve salinity tolerance. *Sesuvium portulacastrum* is a halophyte that grows optimally in the presence of 200--300 mM NaCl \[[@CR32]\]. When growing in a saline environment, *S. portulacastrum* cells accumulate large amounts of Na^+^ despite salt glands and bladders not being present in all tissues \[[@CR33]--[@CR35]\], suggesting *S. portulacastrum* may have a unique ability to remove Na^+^ from cells. The SOS1 protein functions as a PM Na^+^/H^+^ antiporter driven by the proton gradient that is produced by the PM H^+^-ATPase AHA, so they are considered as superior salt tolerance determinants \[[@CR3], [@CR36]\]. The *SpAHA1* and *SpSOS1* genes encode a PM H^+^-ATPase and Na^+^/H^+^ antiporter, respectively, and are more highly transcribed in *S. portulacastrum* plants exposed to salt stress. SpSOS1 more efficiently mediates Na^+^ removal using a proton gradient created by SpAHA1 in *SpAHA1*-*SpSOS1* co-transgenic yeast cells, where yeast cells co-expressing *SpSOS1* and *SpAHA1* grow better following NaCl treatment than cells transformed with only *SpSOS1* or *SpAHA1* \[[@CR3]\]. Over-expression of *SpAHA1* conferred salt tolerance to transgenic *Arabidopsis* \[[@CR37]\]. SpSOS1 complemented the salt sensitivity of transgenic *Arabidopsis sos1* mutant plants \[[@CR38]\]. These results suggest that SpSOS1 and SpAHA1 are involved in salt tolerance of *S. portulacastrum*, and co-expression of *SpAHA1* and *SpSOS1* may improve transgenic plant salt tolerance. To test this hypothesis, *SpAHA1* and *SpSOS1* genes were co-transformed into *Arabidopsis* plants. Functional analyses indicate that *Arabidopsis* plants co-expressing *SpSOS1* and *SpAHA1* had better salt tolerance than plants expressing either gene alone due to efficient Na^+^ removal mediated by SpSOS1 using the extra proton gradient generated by SpAHA1. Therefore, genetic evidence may significantly guide development of more salt tolerant crops using PM-localized Na^+^/H^+^ antiporters and H^+^-ATPases. Results {#Sec2} ======= Transgenic plant identification {#Sec3} ------------------------------- *SpSOS1* and *SpAHA1* were transformed alone or together into *Arabidopsis* plants using *Agrobacteria* carrying pCAMBIA1304-*SpSOS1*, pCAMBIA1304-*SpAHA1,* or pCAMBIA1304-*SpSOS1*-*SpAHA1*. PCR analyses of genomic DNA performed using *SpAHA1/SpSOS1* and *hygB* gene-specific primers revealed 12 *SpSOS1*-, 11 *SpAHA1*-, and 10 *SpSOS1*-*SpAHA1*-transgenic lines were obtained (Additional file [1](#MOESM1){ref-type="media"}: Figure S1**).** Total RNA was isolated from the above transgenic plant lines and RT-PCR analyses were used to study the *SpAHA1* and *SpSOS1* expression levels. The *SpAHA1* gene was significantly expressed in all single *SpAHA1*-transgenic lines, except for SpAHA1- lines 5 and 8. Of the *SpSOS1*-expressing single transgenic plants, *SpSOS1*-line 1 had the highest *SpSOS1* expression of the *SpSOS1*-transgenic lines. In *SpAHA-SpSOS1* co-expressing plants, the clearest expression of both *SpAHA1* and *SpSOS1* was observed in line 10 (Additional file [2](#MOESM2){ref-type="media"}: Figure S2). Therefore, the T3 generation transgenic plants of the homozygous *SpSOS1*-line 1, *SpAHA1*-line 1, and *SpAHA1-SpSOS1*-line 10 were used to characterize the functions of SpSOS1 and SpAHA1. SpSOS1 and SpAHA1 functioned together to more efficiently improve transgenic plant salt tolerance {#Sec4} ------------------------------------------------------------------------------------------------- In plant cells, the PM Na^+^/H^+^ antiporter SOS1 mediates Na^+^ excretion using a proton gradient created by PM H^+^- ATPases. Therefore, both of these proteins are involved in plant salt tolerance. Much evidence indicates that overexpressing *SOS1* or *AHA* increases the salt tolerance of transgenic plants \[[@CR39]\]. In addition, our recent investigation found SpSOS1 and SpAHA1 function cooperatively in transgenic yeast cells, where yeast cells co-expressing *SpSOS1* and *SpAHA1* are better growers than cells transformed with only *SpAHA1* or *SpSOS1* \[[@CR3]\]. Therefore, we hypothesized co-expression of *SpSOS1* and *SpAHA1* would increase the salt tolerance of transgenic plants compared to plants transformed with only *SpSOS1* or *SpAHA1*. To examine the influence of *SpSOS1*-*SpAHA1* co-expression on the salt tolerance of transgenic plants, 5-day-old *Arabidopsis* WT, *SpSOS1-*expressing, *SpAHA1*-expressing, and *SpSOS1*-*SpAHA1* co-expressing seedlings were grown on MS plates containing 0, 50, 75, or 100 mM NaCl. Two weeks post-NaCl treatment, the seedlings were photographed and their fresh weight, root length, and lateral root number were measured**.** Upon exposure to salinity stress, the growth of all tested plants decreased, but expression of either *SpSOS1* or *SpAHA1* ameliorated this growth inhibition from NaCl treatment compared to WT plants. Furthermore, among all the transgenic plants, salt tolerance improved the most in plants co-expressing *SpSOS1* and *SpAHA1* based on growth in MS medium containing different concentrations of NaCl (Fig. [1](#Fig1){ref-type="fig"}).Fig. 1Growth of transgenic and WT seedlings under salt stress. Five-day-old seedlings grown on MS plates were transferred to MS plates containing 0, 50, 75, and 100 mM NaCl. a The seedlings were photographed after 2 weeks of growth. The growth was assessed based on fresh weight (b), root length (c), and number of lateral roots (d). Data are presented as mean ± SE of 12 replicates, where the different letters above the columns indicate statistically significant differences at a *p* \< 0.05 level between the experimental cohorts. SpSOS1, *SpSOS1*-overexpressing plants; SpAHA1, *SpAHA1*-overexpressing plants; SpSOS1-SpAHA1, *SpSOS1* and *SpAHA1* co-expressing plants; WT, wild-type plants Similarly, the growth of transgenic and WT plants was inhibited in soil supplemented with 200 mM NaCl. However, *Arabidopsis* plants expressing both *SpSOS1* and *SpAHA1* grew the best among the different experimental cohorts under these conditions (Fig. [2](#Fig2){ref-type="fig"}a). *SpAHA1-SpSOS1*-line 10 displayed 26, 33, and 67% greater fresh weights than *SpSOS1*-line 1, *SpAHA1*-line 1, and WT plants, respectively (Fig. [2](#Fig2){ref-type="fig"}b). The percent reduction in growth of plant lines treated with NaCl was ordered: *SpSOS1*-*SpAHA1* co-expressing plants \<*SpSOS1*-expressing plants\<*SpAHA1*-expressing plants \<WT plants (Fig. [2](#Fig2){ref-type="fig"}c). These findings indicate the PM-localized Na^+^/H^+^ antiporter SpSOS1 and H^+^-ATPase SpAHA1 function cooperatively to improve the salt tolerance of transgenic plants.Fig. 2Growth of transgenic and WT seedlings. Seven-day-old WT and transgenic seedlings were transferred from MS plates into soil (4 plants/pot) grown for 4 weeks. The plants were then treated with 0 or 200 mM NaCl. Ten days post-treatment, the plants were photographed (a) and their fresh weight was measured (b). The two preparations were performed at different times to more accurately assess the salt tolerance of the transgenic plants and (c) the relative change (percentage reduction) in fresh weight in the presence of salt stress relative to the nonstressed control was determined. Data are presented as mean ± SE of nine replicates. Different letters above the columns indicate statistically significant differences at a *p* \< 0.05 level among the different experimental cohorts. SpSOS1, *SpSOS1*-overexpressing plants; SpAHA1, *SpAHA1*-overexpressing plants; SpSOS1 + SpAHA1, *SpSOS1* and *SpAHA1* co-expressing plants; WT, wild-type plants *SpSOS1*-*SpAHA1* co-expressing *Arabidopsis* plants had higher H^+^ efflux rates than *SpSOS1-* or *SpAHA1-*expressing plants under high saline conditions {#Sec5} ----------------------------------------------------------------------------------------------------------------------------------------------------------- Net H^+^ flux at the roots of WT plants was close to the mock control (Fig. [3](#Fig3){ref-type="fig"}A), which is in full agreement with the recent report that both transient and long-term salinity exposure did not induce H^+^ efflux from *Arabidopsis* roots \[[@CR8]\]. These results suggested that H^+^ efflux might be balanced by H^+^ influx at the roots exposed to salinity stress. PM H^+^-ATPase activity is a major factor in H^+^ excretion at the PM \[[@CR40]\]. It was recently reported that SpAHA1 can function as an H^+^-ATPase on vesicles isolated from yeast cells expressing *SpAHA1* \[[@CR3]\]. Roots expressing SpAHA1 had a faster net H^+^ efflux than the WT plants under saline conditions, suggesting SpAHA1 is responsible for the extra H^+^ efflux, i.e., SpAHA1 pumped more protons out of the cells. It is not expected that protons were extruded faster in roots transformed with *SpSOS1* relative to WT plants and the phenomenon might be from feedback regulation of Na^+^ extrusion mediated by SpSOS1. This hypothesis is also supported by the H^+^ flux in the roots of *SpSOS1*-*SpAHA1* co-expressing transgenic plants (Fig. [3](#Fig3){ref-type="fig"}) being the highest among all the transgenic plants, where the H^+^ efflux rates in the roots co-expressing *SpSOS1*-*SpAHA1* were 49 and 52% greater than *SpSOS1*- and *SpAHA1*-expressing roots, respectively.Fig. 3H^+^ flux in roots of NaCl-treated *Arabidopsis* plants. Seedlings were grown for 3 days on MS plates supplemented with 100 mM NaCl and then H^+^ flux in the roots was measured using the non-invasive micro-test technology (NMT) technique described in the Methods section. (a) Changes in the NMT signals are expressed as arbitrary units. (b) H^+^ flux is expressed as the amount of efflux per second per square centimeter (pmol•cm^− 2^•s^− 1^). Data are presented as mean ± SE of six replicates. Different letters above the columns indicate statistically significant differences at a *p* \< 0.05 level among the different experimental cohorts. SpSOS1, *SpSOS1*-overexpressing plants; SpAHA1, *SpAHA1*-overexpressing plants; SpSOS1-SpAHA1, *SpSOS1* and *SpAHA1* co-expressing plants; WT, wild-type plants; Control, mock controls Plants co-expressing *SpSOS1-SpAHA1* had higher Na^+^ efflux in roots and less Na^+^ accumulation after NaCl treatment {#Sec6} ---------------------------------------------------------------------------------------------------------------------- SOS1 mediates Na^+^ excretion from cells and is a key halotolerance factor. SpSOS1 has been shown to be a PM-localized Na^+^/H^+^ antiporter and capable of improving the growth of transgenic yeast cells under salt stress by decreasing the cellular Na^+^ content \[[@CR3]\]. In this scenario, the roots from all tested plants grown in medium without NaCl displayed Na^+^ uptake characteristics, but no significant differences in Na^+^ flux activities at roots of transgenic and untransformed plants under unstressed condition were observed (Additional file [3](#MOESM3){ref-type="media"}: Figure S3). On the contrary, NaCl treatment stimulated Na^+^ effluxes at all tested roots. *SpSOS1*-expressing roots had faster Na^+^ efflux relative to WT plants in saline conditions (Fig. [4](#Fig4){ref-type="fig"}a, b), suggesting the extra Na^+^ extrusion may be mediated by SpSOS1, which would result in the observed lower Na^+^ content in the *SpSOS1*-transgenic plants than WT plants under salt stress (Fig. [5](#Fig5){ref-type="fig"}a). SOS1-mediated Na^+^/H^+^ exchange is powered by a proton gradient generated by an H^+^-ATPase. Therefore, a proton gradient generated by SpAHA1 (Fig. [3](#Fig3){ref-type="fig"}) might catalyze native SOS1 (AtSOS1) to transport more Na^+^ out of cells, which may be one reason for the higher Na^+^ efflux rate in roots transformed with *SpAHA1* compared to WT plants in saline conditions (Fig. [4](#Fig4){ref-type="fig"}). Roots co-expressing *SpAHA1* and *SpSOS1* had the highest Na^+^ efflux rates among all the transgenic plant lines tested, where the Na^+^ efflux rate in *SpSOS1*-*SpAHA1* co-transgenic roots was 53 and 72% greater than the plants expressing *SpSOS1* or *SpAHA1* singly, respectively (Fig. [4](#Fig4){ref-type="fig"}b). Correspondingly, the Na^+^ levels in the transgenic plants were lower than in WT plants (Fig. [5](#Fig5){ref-type="fig"}a). Therefore, it is reasonable that *SpSOS1*-*SpAHA1* co-expression quickened Na^+^ extrusion in the roots of and decreased Na^+^ accumulation in transgenic plants compared to *SpSOS1-* or *SpAHA1-*expressing plants under saline conditions (Fig. [4](#Fig4){ref-type="fig"}, [5](#Fig5){ref-type="fig"}a). These results indicate SpAHA1 produced an additional proton gradient and, thus, promoted SpSOS1-mediated Na^+^ extrusion in *Arabidopsis* plants co-expressing both the *SpAHA1* and *SpSOS1* genes.Fig. 4Na^+^ flux in roots of NaCl-treated *Arabidopsis* plants. Seedlings were grown for 3 days on MS plates containing 100 mM NaCl. Na^+^ flux in the roots was then measured using the NMT technique described in the Methods section. (a) Changes in the NMT signals are expressed as arbitrary units. (b) Na^+^ flux is expressed as the amount of efflux per second per square centimeter (pmol•cm^− 2^•s^− 1^). Data are presented as mean ± SE of six replicates. Different letters above the columns indicate statistically significant differences at a *p* \< 0.05 level among the different experimental cohorts. SpSOS1, *SpSOS1*-overexpressing plants; SpAHA1, *SpAHA1*-overexpressing plants; SpSOS1-SpAHA1, *SpSOS1* and *SpAHA1* co-expressing plants; WT, wild-type plants; Control, mock controlsFig. 5Na^+^ and K^+^ content in *Arabidopsis* plants. WT and transgenic plants were grown for 10 days in soil containing 0 or 200 mM NaCl. Na^+^ (a) and K^+^ (b) in the leaves of these plants were measured as described in the Methods section. Data are presented as mean ± SE of three replicates. Different letters above the columns indicate statistically significant differences at a *p* \< 0.05 level among the different experimental cohorts. SpSOS1, *SpSOS1*-overexpressing plants; SpAHA1, *SpAHA1*-overexpressing plants; SpSOS1-SpAHA1, *SpSOS1* and *SpAHA1* co-expressing plants; WT, wild-type plants *SpSOS1-SpAHA1* co-transgenic plants had higher K^+^ retention under saline conditions {#Sec7} -------------------------------------------------------------------------------------- Among the physiological and biochemical processes in plant cells influenced by high salinity, nutrient imbalance is among the most deleterious resulting effects \[[@CR41]\]. The chemical and physical characteristics of sodium most resemble potassium among the nutrient elements. Therefore, excess Na^+^ inhibits plant growth by interfering with cytosolic K^+^ homeostasis. No differences in K^+^ content was found among all the tested plants under normal conditions. However, upon salinity stress, the transgenic plants displayed less of a decrease in K^+^ content than the WT plants. Furthermore, co-expression of *SpSOS1* and *SpAHA1* most efficiently alleviated K^+^ loss among the transgenic plants exposed to NaCl, where the K^+^ content was highest in the leaves of plants co-expressing *SpSOS1* and *SpAHA1* (85 mg/g dry weight), followed by those from *SpSOS1*-transgenic plants (73 mg/g dry weight), and then *SpAHA1*-expressing plants (69 mg/g dry weight). WT plants had the lowest K^+^ content (47 mg/g dry weight; Fig. [5](#Fig5){ref-type="fig"}b). *SpSOS1-SpAHA1* co-expression decreased malondialdehyde accumulation in transgenic plants {#Sec8} ----------------------------------------------------------------------------------------- Salinity creates oxidative stress and excess reactive oxygen species can interfere with metabolism in the cytoplasm, such as by damaging membrane structures and destroying membrane integrity through lipid peroxidation \[[@CR42]\]. An indicator of membrane lipid oxidation, malondialdehyde (MDA) represents membrane lipid damage to some extent. Upon exposure to NaCl, the amount of MDA in the leaves of all tested plants increased, but MDA accumulation in the *SpSOS1*-*SpAHA1* co-expressing leaves was the lowest (Fig. [6](#Fig6){ref-type="fig"}) at only 84, 74, and 61% of that in *SpSOS1*-expressing, *SpAHA1*-expressing, and WT plants under saline conditions. These results indicate *SpSOS1* and *SpAHA1* coordination could more efficiently reduce oxidative damage to membranes from salinity stress in transgenic plants.Fig. 6Malondialdehyde content in leaves. Two-week-old WT and transgenic *Arabidopsis* seedlings were treated with 200 mM NaCl for 7 days and then their leaves were harvested. Malondialdehyde content in the leaves was measured as described in the Methods section. Data are presented as mean ± SE of three replicates. Different letters above the columns indicate statistically significant differences at a *p* \< 0.05 level among the different experimental cohorts. SpSOS1, *SpSOS1*-overexpressing plants; SpAHA1, *SpAHA1*-overexpressing plants; SpSOS1-SpAHA1, *SpSOS1* and *SpAHA1* co-expressing plants; WT, wild-type plants Discussion {#Sec9} ========== Plants grown under K^+^ deficiency substitute it by Na^+^, especially some halophyte species can use Na^+^ for stomata operation instead of K^+^ \[[@CR43]\]. However, when there is excess sodium in the cytosol, it interferes with some key metabolic processes and eventually inhibits the growth and development of the plant. To tolerate high Na^+^ levels, plant cells must be capable of removing Na^+^ from the cytoplasm through some physiological processes. In one of these processes, cytoplasmic Na^+^ can be imported into vacuoles through tonoplast Na^+^/H^+^ antiporter NHXs using an electrochemical gradient established by a vacuolar H^+^-ATPase and H^+^-PPase (AVP/VP) using protons. The sequestration of Na^+^ into vacuoles not only prevents the deleterious effects resulting from Na^+^ in the cytoplasm, but also lets the plants use Na^+^ as an osmoticum, which helps maintain the osmotic potential that drives water into the cells \[[@CR39], [@CR44]\]. Therefore, tonoplast NHX antiporters and H^+^-pumps have important functions in plant responses to salt stress. In transgenic plants, overexpression of genes encoding vacuolar Na^+^/H^+^ antiporters or H^+^-PPases enhances salt tolerance \[[@CR39], [@CR42]\]. Furthermore, co-expression of AVP and NHX may better improve the growth of transgenic plants exposed to salt stress through more efficient compartmentalization of Na^+^ into vacuoles than when NHX or AVP are expressed alone. Co-expression of *ZxNHX* and *ZxVP1--1* confers better salt tolerance to transformed sugar beet and lotus plants \[[@CR42], [@CR45]\]. *NHX1-AVP1* co-transgenic rice grows better under salt stress than rice plants expressing only one of these genes \[[@CR46], [@CR47]\]. Tobacco plants co-expressing *TNHXS1* and *TVP1* have higher salt tolerance than transgenic plants expressing *TNHXS1* or *TVP1* alone \[[@CR48]\]. Another mechanism contributing to Na^+^ extrusion is the PM Na^+^/H^+^ antiporter SOS1. The Na^+^ extrusion mediated by SOS1 is also driven by electrochemical gradients of protons generated by a PM H^+^-pump (H^+^-ATPase, AHA). The overexpression of *SOS1* significantly improves the salt tolerance of transgenic grapevine compared to WT plants \[[@CR49]\]. Overexpression of the *SOS1* gene in tobacco plants increases salt tolerance by maintaining a lower Na^+^ content \[[@CR50]\] and the growth of *Arabidopsis* plants overexpressing *SOS1* is better than that of WT plants under salt stress \[[@CR51]\]. It has been reported that overexpression of *PeHA1*(H^+^-ATPase 1)*,* a poplar gene encoding a PM-localized H^+^-ATPase, enhances the salt tolerance of transgenic *Arabidopsis* \[[@CR52]\]. These studies suggest that co-expression of both *SOS1* and *AHA* in transgenic plants should more effectively increase salinity tolerance just as co-expression of vacuolar NHX and AVP results in higher salt tolerance. In the present investigation, *SpAHA1*-transgenic roots had faster H^+^ efflux than WT plants under salt stress (Fig. [3](#Fig3){ref-type="fig"}), suggesting SpAHA1 enhanced proton efflux and generated an additional proton gradient that acted as a driving force for Na^+^/H^+^ exchange mediated by SpSOS1. More interestingly, the magnitude of net H^+^ flux is in 3 to 4 pmol•cm^− 2^•s^− 1^ range (Fig. [3](#Fig3){ref-type="fig"}), while Na^+^ flux is around 1000 pmol•cm^− 2^•s^− 1^; the stoichiometric ratio for Na^+^/H^+^ exchange of SOS1 protein is 1H^+^:1Na^+^, so such tiny increase in H^+^ flux (from 2 to 4 pmol•cm^− 2^•s^− 1^) may not cause such a massive flux of Na^+^. Net H^+^ fluxes at roots was determined in the present study, that is, the data for net H^+^ efflux is equal to total H^+^ efflux minus total H^+^ influx at the roots, suggesting that H^+^-ATPase mediated H^+^ efflux is likely balanced by H^+^ uptake through SOS1 transporters. These suggest that SpAHA1 might provided more H^+^ gradient than the shown data of net H^+^ efflux in transgenic plant exposed to salinity stress, resulting in faster Na^+^ efflux from cells, consequently Na^+^ content in transgenic plants was lower than that in wild type plants (Fig. [5](#Fig5){ref-type="fig"}a). Overexpression of *SpSOS1* in transgenic *Arabidopsis* accelerated Na^+^ efflux in the roots (Fig. [4](#Fig4){ref-type="fig"}a, b), resulting in decreased Na^+^ content in the transgenic plants compared to the WT plants (Fig. [5](#Fig5){ref-type="fig"}a). Interestingly, the rate of H^+^ efflux in the *SpSOS1-*expressing roots was faster than in the WT or *SpAHA1*-transgenic roots (Fig. [3](#Fig3){ref-type="fig"}), suggesting that increased Na^+^ extrusion mediated by SpSOS1 might regulate H^+^-pumping activity via feedback at the PM. This is because the Na^+^/H^+^ exchange mediated by SOS1 is dependent on energy and driven using a PM-localized H^+^-ATPase-driven proton motive force \[[@CR53]\]. The H^+^ and Na^+^ efflux rates in the roots of *Arabidopsis* plants co-expressing *SpAHA1* and *SpSOS1* were highest among all the transgenic plants (Figs. [3](#Fig3){ref-type="fig"} and [4](#Fig4){ref-type="fig"}), leading to the lowest Na^+^ content in the co-transformed plants relative to other transgenic plants expressing only *SpSOS1* or *SpAHA1* (Fig. [5](#Fig5){ref-type="fig"}a). In response to NaCl treatment, the biomass of the transgenic *Arabidopsis* plants co-expressing *SpSOS1* and *SpAHA1* was greater than the biomasses of the single-gene expressing plants (Fig. [2](#Fig2){ref-type="fig"}). Taken together, the higher rate of Na^+^ extrusion, lower Na^+^ levels, and better growth of the *SpAHA1*-*SpSOS1* co-expressing plants compared to the single *SpAHA1* or *SpSOS1* gene transgenic plants provides direct genetic evidence that SOS1 and AHA function in a cooperative manner to inhibit Na^+^ accumulation in the cytosol and play important roles in plant adaption to highly saline conditions. High soil salinity is characterized by high soluble salt concentrations, of which sodium salt is the most soluble and widespread salt \[[@CR44]\]. Excessive sodium ions in soils can enter into plant cells and then interference with some critical biochemical and physiological processes. The most deleterious effect of salinity is ion toxicity \[[@CR41]\]. K^+^ is a necessary macronutrient that has a critical role in the growth and development of plants \[[@CR54]\]. Due physicochemical similarities between Na^+^ and K^+^, Na^+^ can compete with K^+^ for binding sites important in critical cytoplasmic physiological and biochemical processes \[[@CR55]\]. In particular, Na^+^ inhibits the activity of many K^+^-dependent enzymes \[[@CR56]\] and, therefore, excess Na^+^ can inhibit K^+^-associated activities in the cytosol \[[@CR55]\]. It is hypothesized that plant survival in the presence of salt stress requires a high K^+^/Na^+^ ratio in the cytoplasm \[[@CR57]\]. Therefore, limiting Na^+^ influx into cells may facilitate plant growth under salt stress \[[@CR58], [@CR59]\]. Under high salt conditions, the PM potential becomes depolarized, which encourages passive Na^+^ influx into cells and K^+^ efflux out of cells. H^+^-ATPase-generated electrochemical potential gradients across PMs can repolarize PMs following NaCl-induced depolarization \[[@CR39]\]. Therefore, maintenance of the PM potential using H^+^-ATPases can reduce the Na^+^ influx via depolarization-activated non-selective cation channels (NSCCs) and K^+^ efflux via K^+^ outward rectifiers (KORs) and NSCCs \[[@CR60]\]. The net H^+^ efflux from root cells in *SpAHA1*-transgenic plants occurred at a higher rate relative to WT plant roots, suggesting SpAHA1 increased the H^+^-ATPase activity and electrochemical potential and, thus, the H^+^ gradient across the PM. This may reduce Na^+^ influx and K^+^ efflux and, correspondingly, the *SpAHA1*-transgenic plants had less Na^+^ and more K^+^ than WT plants (Fig. [5](#Fig5){ref-type="fig"}a, b). In addition to Na^+^ extrusion, SOS1 is also involved in K^+^ homeostasis in plant cells under high salt conditions. Transgenic tobacco plants expressing *SbSOS1* contain less Na^+^, but more K^+^, in their roots than WT plants under high salt stress \[[@CR50]\]. Horie et al. \[[@CR61]\] suggested SOS1 plays a primary role facilitating high-affinity absorption of K^+^ into roots. SOS1 is necessary for protecting K^+^ uptake and is involved in K^+^ homeostasis maintenance in cells under salinity stress \[[@CR19], [@CR62]\]. Overexpression of *TaSOS1* confers salt tolerance to transgenic tobacco plants by decreasing the Na^+^ and increasing the K^+^ levels \[[@CR31]\]. *Arabidopsis* roots expressing *SpSOS1* displayed faster Na^+^ efflux than WT plant roots under saline condition (Fig. [4](#Fig4){ref-type="fig"}), suggesting SpSOS1 was responsible for the extra Na^+^ extrusion. The faster H^+^ efflux in the roots of plants expressing *SpSOS1* may aid repolarization following NaCl-induced depolarization of the PM, thus decreasing Na^+^ influx and K^+^ efflux \[[@CR60]\]. These actions may have led *SpSOS1*-transgenic plants to contain less Na^+^ and more K^+^ relative to WT plants under salt treatment. Therefore, faster H^+^ and Na^+^ efflux in the roots also resulted in retention of more K^+^ and a lower Na^+^ concentration in cytosol of *SpSOS1*-*SpAHA1* co-transgenic *Arabidopsis* plants compared to plants expressing *SpSOS1* or *SpAHA1* alone (Figs. [3](#Fig3){ref-type="fig"} and [4](#Fig4){ref-type="fig"}). This led to *Arabidopsis* plants co-expressing *SpSOS1* and *SpAHA1* to have higher a K^+^/Na^+^ level than the transgenic plants with only *SpSOS1* or *SpAHA1*, which is strong evidence of salt tolerance. These results suggest SOS1 and AHA1 facilitate more efficient prevention of K^+^ loss and enhance Na^+^ extrusion and thereby contribute to better salt tolerance. Another deleterious effect of salinity stress in plants is associated with oxidative stress \[[@CR39]\]. Accumulation of ROS (reactive oxygen species) is toxic in cells. Therefore, intracellular ROS levels are tightly regulated under normal conditions through a number of intracellular peroxidative and antioxidative reactions within the cell. Salinity can disrupt the ROS production and scavenging balance, resulting in ROS accumulation, which can negatively affect cellular structures and metabolism \[[@CR63], [@CR64]\]. In order to protect cells from salinity-induced oxidative damage, excess ROS is scavenged by antioxidant molecules and enzymes. RCD1 (Radical-induced cell death) is a regulator of responses to oxidative stress and protects cells from oxidative damage caused by H~2~O~2~, diamide, and tert-butyl peroxide \[[@CR65]--[@CR67]\]. SOS1 functions in tolerance to oxidative stress by interacting with RCD1 and regulating expression of certain genes associated with oxidative-stress tolerance in *Arabidopsis* \[[@CR66]\]. Haem oxygenase (HO) is an important factor in plant antioxidant defense systems. Overexpression of the *AtHO* gene enhances *Arabidopsis* tolerance to salt by increasing PM H^+^-ATPase activity and expression \[[@CR68]\]. Excess ROS can damage membrane structures by oxidizing lipids in the PM, leading some key metabolites abnormally leak out of cells. ROS could disturb ion homeostasis in cells by inducing the efflux of several cations \[[@CR69]--[@CR71]\]. Coskun et al. \[[@CR53]\] found NaCl-induced efflux of K^+^ was a result of a lack of PM integrity in rice. This indicates the maintenance of PM stability has a key role in plant salt tolerance. In the present investigation, both SpSOS1 and SpAHA1 prevented the accumulation of MDA in transgenic plants following NaCl treatment, but plants co-expressing *SpSOS1* and *SpAHA1* had a more drastic decrease in MDA content under salt stress than plants expressing only one of these genes. This suggests SpSOS1 and SpAHA1 coordinate in transgenic *Arabidopsis* and ameliorate salt toxicity by more efficiently alleviating oxidative damage to the PM generated by salinity stress. Conclusions {#Sec10} =========== *Arabidopsis* plants co-expressing *SpSOS1* and *SpAHA1* had higher K^+^ and lower MDA levels than plants transformed with only *SpSOS1* or *SpAHA1* and, thus, grew better under salt stress. The coordinated action of these genes might be a novel and effective method for increasing the salt tolerance of crops. Methods {#Sec11} ======= Plasmid construction {#Sec12} -------------------- The *SpAHA1* and *SpSOS1* genes were separately cloned from *S. portulacastrum* and inserted into plasmids p414 (p414-*SpAHA1*) and p416 (p416-*SpSOS1*) in our recent investigation \[[@CR3]\]. Plant vectors expressing *SpAHA1* or *SpSOS1* alone or together bicistronicly were constructed as described in Additional file [4](#MOESM4){ref-type="media"}: Figure S4. (1) Amplification of the *SpSOS1* gene was performed using the p416-*SpSOS1* plasmid as the template and the primers SpSOS1-F and SpSOS1-R (Additional file [5](#MOESM5){ref-type="media"}: Table S1). The amplified gene was inserted into the pCAMBIA1300 vector between *Sal* I and *Kpn* I restriction sites, generating pCAMBIA1300-*SpSOS1*. (2) A fragment containing a constitutive promoter (cauliflower mosaic virus 35S promoter), the *SpSOS1* gene, and the NOS terminator was excised from the pCAMBIA1300-*SpSOS1* plasmid using *Pst* I and *Eco*R I restriction enzymes and transferred into the plant expression vector pCAMBIA1304 between the same restriction sites. The resulting plasmid was named pCAMBIA1304-*SpSOS1*. (3) The *SpAHA1* gene was amplified using the p414-*SpAHA1* plasmid as the template and the primers SpAHA1-F and SpAHA1-R (Additional file [5](#MOESM5){ref-type="media"}: Table S1) and inserted into the pCAMBIA1304 and pCAMBIA1304-*SpSOS1* plasmids between the *Spe* I and *Eco*72 I restriction sites to replace the *GUS* (β-glucuronidase) gene. The resulting plasmids were designated pCAMBIA1304-*SpAHA1* and pCAMBIA1304-*SpSOS1*-*SpAHA1*, respectively. The plasmids were all verified by sequencing. *Arabidopsis* transformation and identification {#Sec13} ----------------------------------------------- The three recombinant plasmids described above were added to a 100 mM CaCl~2~ solution containing competent *Agrobacterium tumefaciens* GV3101 cells. The plasmids were then introduced into the *Agrobacterium* cells via heat shock (42 °C). Finally, the above three expression cassettes were transformed into *Arabidopsis thaliana* (Col-0) by infecting flower buds with the *Agrobacterium* cells containing the recombinant plasmids \[[@CR72]\]. T0-generation seeds were screened initially on MS (Murashige & Skoog) medium supplemented with 50 μg/mL hygromycin B. DNA was purified from the candidate lines and used as template for PCR (polymerase chain reaction) amplification with specific primer pairs to identify the different transgenic plants. The primers are listed in Additional file [5](#MOESM5){ref-type="media"}: Table S1. All the transgenic lines were furthermore verified by PCR using primers specific for the *hygB* marker gene, HygB-TF and HygB-TR (Additional file [5](#MOESM5){ref-type="media"}: Table S1). Total RNA was purified from the transgenic lines and *SpAHA1* and *SpSOS1* expression was assessed by RT-PCR (reverse transcription PCR) with a housekeeping gene, *Actin,* as an internal control. The primers for the *SpAHA1* (SpAHA1-RT-F and SpAHA1-RT-R), *SpSOS1* (SpSOS1-RT-F and SpSOS1-RT-R), and *Actin* (Actin-RT-F and Actin-RT-F) genes are listed in Additional file [5](#MOESM5){ref-type="media"}: Table S1**.** The PCR conditions were as follows: 94 °C for 2 min, followed by 28 cycles of 94 °C for 30 s, 55 °C for 30 s, and 72 °C for 30 s, and a final extension at 72 °C for 5 min. The resulting PCR products were assessed by agarose gel electrophoresis. Cultivation and salt treatment of transgenic and WT plants {#Sec14} ---------------------------------------------------------- To analyze the salt tolerance of transgenic and WT plants, seeds from T3 homozygous transgenic lines (expressing single *SpSOS1*, single *SpAHA1,* and both *SpSOS1* and *SpAHA1*) and untransformed plants were germinated on MS plates in a growth chamber (22 °C with a 16 h light / 8 h dark cycle and a light intensity of 100 μmol•m^− 2^•s^− 1^). After 5 days, the seedlings grown on MS plates were transferred onto MS plates containing 0, 50, 75, and 100 mM NaCl and allowed to grow for 2 weeks. Then the root length, number of lateral roots, and fresh seedling weights were measured. In addition, 10-day-old seedlings were transferred to a mixture of organic soil and sand (3:1, *v*/v) in pots (4 seedlings/pot) and grown in a greenhouse with long-day conditions (16 h light/8 h dark at 22 °C and a light strength of 150 μmol•m^− 2^•s^− 1^) for 4 weeks. The pots containing the plants were then put into water containing 0 or 200 mM NaCl. Ten days post-NaCl treatment, the treated plants were photographed and their fresh weights were determined. Determination of Na^+^ and K^+^ content in *Arabidopsis* plants {#Sec15} --------------------------------------------------------------- At the end of the NaCl treatment, the WT and transgenic plants were separately collected. The Na^+^ and K^+^ in the samples were measured using atomic absorption spectrometry as described in a previous work \[[@CR31]\]. Measurement of Na^+^ and H^+^ flux in roots {#Sec16} ------------------------------------------- Seven-day-old uniform T3 seedlings, which had been grown on MS plates, were transferred to MS medium containing 100 mM NaCl and grown for 3 days, and then the roots of salt stressed seedlings were put into measurement buffer to balance for 10 min, after that net H^+^ and Na^+^ fluxes were measured in the YoungerUSA Xuyue (Beijing) BioFunction Institute by using Non-invasive Micro-test Technology (NMT100 Series, Xuyue (Beijing) Sci. & Tech. Co., Ltd., Beijing, China) Software. H^+^, Na^+^-selective microsensors were prepared as described previously \[[@CR60]\]. Pre-pulled and silanized microsensor (Φ4.5 ± 0.5 μm, XY-CGQ-01, YoungerUSA) were first filled with a backfilling solution (H^+^: 15 mM NaCl + 40 mM KH~2~PO~4~, pH 7.0; Na^+^: 250 mM NaCl) to a length of approximately 1.0 cm from the tip. Then the microsensors were front filled with 40--50 μm columns of selective liquid ion-exchanger (LIX) (H^+^: XY-SJ-H; Na^+^: XY-SJ-Na; all from YoungerUSA). An Ag/AgCl wire microsensor holder YG003-Y11 (YoungerUSA) was inserted in the back of the microsensor to make electrical contact with the electrolyte solution. YG003-Y11 (YoungerUSA) was used as the reference microsensor. Prior to the flux measurement, the microsensor were calibrated with cultural media having different concentrations of H^+^ (pH 6.0 and pH 7.0) or Na^+^ (5 mM and 0.5 mM), respectively. Only microsensor with a Nernstian slope \>50 mV/decade were used in our study. The same microsensors were calibrated again according to the same procedure and standards after each test. After that, net ion fluxes were recorded on the root meristematic zones, 120 μm from the tip where SOS1 activity was the highest \[[@CR73]\], in 5 mL measurement buffer (0.1 mM KCl, 0.1 mM CaCl~2~, 0.1 mM MgCl~2~, 0.5 mM NaCl, 0.3 mM MES (2-(N-Morpholino) ethanesulfonic acid) and 0.2 mM Na~2~SO~4~, pH 6.5). Net H^+^ and Na^+^ flux was calculated by Fick's law of diffusion \[[@CR60]\]. Six biological repeats were performed for each analysis. Assays of malondialdehyde content {#Sec17} --------------------------------- Two-week-old T3 transgenic and untransformed *Arabidopsis* seedlings were grown in the presence of 200 mM NaCl for 7 days. Malondialdehyde (MDA) content in the leaves was measured using the thiobarbituric acid method previously described by Dhindsa and Matowe \[[@CR74]\]. Statistical analysis {#Sec18} -------------------- Two-tailed Student's t-tests were used to analyze the data. The results are expressed as mean ± SE and differences with a *P*-value \< 0.05 were considered statistically significant. At least three biological replicates were performed for each experiment. Additional files ================ {#Sec19} Additional file 1:**Figure S1.** Molecular identification of transgenic plants. DNA was purified from transgenic and WT plant leaves. (a) PCR identification of *SpSOS1* transgenic plants. M: DL2000 marker (Sangon Biotech, China; No. B600335); 1--12, different transgenic lines (lines 1--12); 13, negative control (WT plants). (b) PCR identification of *SpAHA1* transgenic plants. M: DL2000 marker; 1--11, different transgenic lines (lines 1--11); 12, negative control (WT plants). (c) PCR identification of *SpSOS1* and *SpAHA1* co-expressing plants. M: DL2000 marker; 1--10, different transgenic lines (lines 1--10); 11, negative control (WT plants). PCR amplification was performed using primers specific for *SpSOS1*, *SpAHA1,* or *hygB* gene (expected sizes of 980, 916, and 750 bp, respectively) with the corresponding DNA serving as the template. The PCR products were assessed by agarose gel electrophoresis. (TIF 835 kb) Additional file 2:**Figure S2.** Expression of *SpSOS1* and *SpAHA1* genes in transgenic *Arabidopsis* lines. Total RNA was purified from leaves from the T3 generation of transgenic plants and used for RT-PCR analysis. The *Arabidopsis Actin* gene served as an internal control. (a) Expression of the *SpSOS1* gene in *SpSOS1-*transgenic plants was analyzed by RT-PCR. 1--12, different transgenic lines (lines 1--12). (b) Expression of *SpAHA1* gene in *SpAHA1-*transgenic plants as analyzed by RT-PCR. 1--11, different transgenic lines (lines 1--11). (c) Expression of *SpSOS1* and *SpAHA1* genes in *SpSOS1-SpAHA1* co*-*expressing plants as analyzed by RT-PCR. 1--10, different transgenic lines (lines 1--10). (TIF 531 kb) Additional file 3:**Figure S3.** Na^+^ flux in roots of *Arabidopsis* plants grown in media without NaCl. Na^+^ flux in the roots of seven-day-old seedlings was measured using the NMT technique described in the Methods section. (a) Changes in the NMT signals are expressed as arbitrary units. (b) Na^+^ flux is expressed as the amount of efflux per second per square centimeter (pmol•cm^− 2^•s^− 1^). Data are presented as mean ± SE of three replicates. Same letter above the columns indicate that the differences at a *p* \< 0.05 level among the different experimental cohorts are not significant statistically. SpSOS1, *SpSOS1*-overexpressing plants; SpAHA1, *SpAHA1*-overexpressing plants; SpSOS1-SpAHA1, *SpSOS1* and *SpAHA1* co-expressing plants; WT, wild-type plants. (TIF 456 kb) Additional file 4:**Figure S4.** Schematic of T-DNA region in the binary vectors. (a) The pCAMBIA1300-*SpSOS1*, (b) T pCAMBIA1304-*SpSOS1*, (c) pCAMBIA1304-*SpSOS1*-*SpAHA1*, and (d) pCAMBIA1304-*SpAHA1* plasmids. (TIF 239 kb) Additional file 5:**Table S1.** Sequences of primers used in this study. Small letters indicate restriction enzyme sites. (XLS 18 kb) AHA : H^+^-ATPase AVP/VP : Vacuolar H^+^-PPase GUS : β-glucuronidase HA : H^+^-ATPase HO : Haem oxygenase KOR : K^+^ outward rectifier MDA : Malondialdehyde MES : 2-(N-Morpholino) ethanesulfonic acid MS : Murashige & Skoog NHX : Na^+^/H^+^ antiporter NSCC : Non-selective cation channels PCR : Polymerase chain reaction PM : Plasma membrane RCD : Radical-induced cell death RNAi : RNA interference ROS : Reactive oxygen species RT-PCR : Reverse transcription polymerase chain reaction SOS : Salt overly sensitive WT : Wild-type We thank Boston Professional Group (BPG) Editing for English language editing. We also thank the reviewers and editors for helpful comments on earlier drafts of the manuscript. Funding {#FPar1} ======= This work was supported by the National Natural Science Foundation of China (31660253 to XJ), the Scientific and Technological Foundation of Hainan Province (HNGDhs201502 to XJ), the Foundations of Hainan University and Shandong Normal University (hdkytg201706 to XJ and JS), and Startup funding from Hainan University (KYQD(ZR)1845 to YZ). The above funding was used for the design of the study and collection, analysis, and interpretation of data in writing the manuscript. Availability of data and materials {#FPar2} ================================== The datasets used and/or analyzed during the current study are available from corresponding authors on reasonable request. XJ, YZ and ZW conceived and designed the experiments. YF, XY, QX and YX performed the experiments. YZ and JS made substantial contributions to the data analysis. XJ and JS supervised, wrote and revised the manuscript. All authors read and approved the final manuscript. Ethics approval and consent to participate {#FPar3} ========================================== Not applicable. Consent for publication {#FPar4} ======================= Not applicable. Competing interests {#FPar5} =================== The authors declare that they have no competing interests. Publisher's Note {#FPar6} ================ Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

{#sp1 .154}

SARS-Like Virus Can Probably Pass From Person-to-Person {#s1} ======================================================= 13 May 2013 (Reuters Health \[Angus McDowall\])---A second man in France was diagnosed with a severe acute respiratory syndrome (SARS)--like disease after sharing a hospital room with France\'s only other sufferer. World Health Organization (WHO) Assistant Director-General Keiji Fukuda told reporters in Saudi Arabia, the site of the largest cluster of infections, there was no evidence so far the virus was able to sustain "generalized transmission in communities"---a scenario that would raise the specter of a pandemic. But he added: "Of most concern ... is the fact that the different clusters seen in multiple countries ... increasingly support the hypothesis that when there is close contact, this novel coronavirus can transmit from person to person." The first French patient was confirmed as suffering from the disease after traveling in the Gulf. The second patient was transferred to intensive care after the two men shared a room in a hospital in Lille. Copyright © 2013 Reuters Limited. All rights reserved. ***Editorial comment.*** As of 18 May 2013, the number of cases of infection with the novel coronavirus, newly named "Middle East Respiratory Syndrome Coronavirus" (MERS-CoV), was 41, with 20 deaths. The breakdown of cases of severe acute respiratory infection (SARI) associated with MERS-CoV (and deaths) by country of report was as follows: Saudi Arabia: 31 (15 deaths); Jordan: 2 (2 deaths); United Kingdom: 4 (1 patient from Qatar, 3 patients from the United Kingdom, 1 patient with history of travel to Saudi Arabia and Pakistan) (2 deaths); Germany: 2 (1 patient from Qatar, 1 patient from the United Arab Emirates \[UAE\]) (1 death); France: 2 (1 with a history of travel to the UAE prior to illness, 1 healthcare facility contact with another case). Of the total 41 cases confirmed as of 18 May, 32 were probably infected in Saudi Arabia, 2 in Jordan, 2 in Qatar, 2 in the UAE, 2 in the United Kingdom, and 1 in France. Among the most recently infected were 2 hospital workers in Saudi Arabia. No animal host has been identified. As with SARS, nosocomial transmission is clearly a risk factor. In addition, it is recommended to use lower respiratory tract specimens rather than nasopharyngeal swabs for laboratory diagnosis of the MERS-CoV infection as lower respiratory tract specimens have a higher yield. It is likely that, as with SARS, the aerosolization of lower respiratory tract specimens during bronchoscopy would carry a high risk for transmission. Most patients have been male (79%), and the patients have almost all been adults (median age, 56 years). All of the laboratory-confirmed cases had respiratory disease as part of the illness, and most had severe acute respiratory disease requiring hospitalization. Reported clinical features include acute respiratory distress syndrome, renal failure requiring hemodialysis, consumptive coagulopathy, and pericarditis. Many patients have also had gastrointestinal symptoms including diarrhea. China Reports 3 New Bird Flu Deaths, Toll Hits 35 {#s2} ================================================= 13 May 2013 (Reuters Health \[Koh Gui Qing\])---Three more people have died in China from the new strain of H7N9 bird flu virus, raising the death toll to 35 while the total number of infections rose to 130 (plus 1 in Taiwan), state media said. Without giving details of the deaths, Xinhua News Agency said a new case of H7N9, described by the World Health Organization (WHO) as one of the most lethal flu viruses around, was found in China\'s eastern Jiangxi province. There has so far been no evidence of human-to-human transmission of the virus, a point reiterated by a spokesperson for the agency on Monday, citing health authorities. It noted that 57 of those infected have recovered. Chinese scientists say the virus has been transmitted to humans from chickens, although according to the WHO, 40% of people infected with H7N9 had no contact with poultry. The US Centers for Disease Control and Prevention has said the current strain of bird flu cannot start a pandemic but notes there is no guarantee it will not mutate and gain the potential to cause a serious pandemic. Copyright © 2013 Reuters Limited. All rights reserved. ***Editorial comment.*** As of 21 May, the count is 131 cases with 35 deaths. The good news is that although the epidemic is continuing, it seems to be at a diminishing rate with no new cases since early May. If the decrease in rates continues, it is very likely related to the enormous efforts in China to reduce the potential of exposure to H7N9 by the massive closing of live bird markets. Other factors such as changes in weather could also be playing a role. Let\'s hope that the decrease in cases continues. Spinal and Paraspinal Infections Associated With Contaminated Methylprednisolone Acetate Injections---Michigan, 2012--2013 {#s3} ========================================================================================================================== (MMWR 62:377, 2013)---As of 6 May 2013, the outbreak of fungal meningitis and other fungal infections (mainly *Exserohilum rostratum*) resulting from injections of contaminated methylprednisolone acetate (MPA) had resulted in 741 reported cases and 55 deaths in 20 states. The total case count in Michigan was 261 and included 16 deaths. During the first 4 weeks of the outbreak, 7 September--5 October 2012, nearly all of the reported cases nationally met the Centers for Disease Control and Prevention (CDC) case definition solely for meningitis. However, at outbreak week 5, certain states, including Michigan, began reporting cases of localized spinal and paraspinal infections, including epidural abscesses, phlegmon, arachnoiditis, discitis, or vertebral osteomyelitis. Michigan has reported the highest number of spinal and paraspinal infection cases (167), accounting for 52% of the 320 cases reported nationally. Michigan also had reported an additional 43 spinal and paraspinal infection cases with meningitis. Four pain management facilities in Michigan received 2225 of the approximately 17 000 vials of MPA that came from the contaminated lots distributed nationally. As of 29 January 2013, epidemiologic or clinical data were available for 180 patients in Michigan: 141 of the 165 patients (167 as of 6 May) had spinal or paraspinal infections alone, and 39 of the 43 patients had spinal or paraspinal infections along with meningitis. Of the 180 patients, 160 (89%) received care for their infections from St Joseph Mercy Hospital in Ann Arbor. The 160 patients treated for their infections at St Joseph included 113 (80%) who had diagnoses only of spinal or paraspinal infection and not meningitis. Four (2%) of the 180 patients died. Among patients with available information, median number of days from the last injection to the first positive magnetic resonance imaging (MRI) finding was 50 (range: 12--121 days) for all patients with a spinal or paraspinal infection, 52 (range: 12--121) for patients who received 1 injection, and 43 (range: 18--116) for patients who received 1 or more injections. Median number of days from the first positive lumbar puncture finding to the first positive MRI finding for patients with spinal and paraspinal infections and meningitis was 21. Several reasons might explain the higher number and percentage of patients with spinal or paraspinal fungal infection in Michigan compared with other states including increased case finding, possible higher contamination of vials of MPA shipped to Michigan, or differences in injection technique (a transforaminal rather than translaminar approach preferred by clinicians at St Joseph Mercy Hospital). CDC guidelines urge clinicians to maintain a higher index of suspicion for patients who have unrecognized localized spinal or paraspinal infections and to embark on an assertive clinical management approach. However, because both voriconazole and liposomal amphotericin B, the most widely used therapies, can be toxic and MRI findings might be equivocal, a strategy of waiting 2--4 weeks for repeat MRIs while watching for signs of progression might be a reasonable alternative to immediate initiation of treatment. ***Editorial comment.*** As of 24 May 2013, the US Food and Drug Administration has received reports of infections in patients who received MPA injections compounded by a pharmacy in Tennessee. At least 1 of these infections appears to be fungal in nature. Impact of a Shortage of First-line Antituberculosis Medication on Tuberculosis Control---United States, 2012--2013 {#s4} ================================================================================================================== (MMWR 62:398, 2013)---In November 2012, the United States began to experience a severe interruption in the supply of isoniazid (INH). A survey showed that 79% of the responding health departments reported difficulties with procuring INH within the last month, with 15% reporting that they no longer had INH and 41% reporting that they would no longer have a supply within 1 month of the survey. Because of local interruptions in INH supply, responding tuberculosis programs were changing INH suppliers (69%), prioritizing only high-risk patients for treatment of latent tuberculosis infection (72%), delaying latent tuberculosis treatment (68%), and changing to alternative latent tuberculosis treatment regimens (88%). Treatment of tuberculosis disease with second-line drugs can be less effective, more toxic, and more costly than treatment with first-line drugs. Interruptions in the supply of second-line antituberculosis medications have been ongoing in the United States for several years, but since November 2012, tuberculosis control programs have experienced the first sustained generalized supply interruption of a first-line antituberculosis medication. The INH shortage was unexpected, has affected US tuberculosis control efforts, and has lasted months longer than predicted. How the increased use of alternative regimens and the rising cost of INH driven by increased demand might affect the future supply of INH and other first-line antituberculosis medications is uncertain.

INTRODUCTION ============ Methicillin-resistant *Staphylococcus aureus* (MRSA) is a major cause of skin, wound, soft tissue, respiratory, and endovascular infections as well as a leading cause of hospital-acquired infections. Many outbreaks of *S. aureus* have been reported, and some have been traced to environmental sources \[[@B1]-[@B3]\]. The transmission of MRSA is presumed to occur through contact with a common reservoir of environmental contamination, including computer keyboards, faucets, door handles, floors, bed linens, patient gowns, and blood pressure cuffs \[[@B1], [@B2], [@B4]\]. In many hospitals, portable radiological studies are routinely performed on patients in wards or intensive care units. Even before the emergence of the antimicrobial resistant organisms, there were reports of contaminated patients coming into contact with medical devices used in radiology departments \[[@B5], [@B6]\]. A study from the late 1960s showed that staphylococci, coliform bacteria, and molds contaminated the surfaces of X-ray equipment contacted by patients, such as chest racks, chin supports, and X-ray tables \[[@B5]\]. However, the contamination of X-ray cassette surfaces by antimicrobial-resistant microorganisms, such as MRSA, has not yet been examined. The present study was performed to determine whether the surfaces of X-ray cassettes contacted by patients were contaminated with MRSA. In addition, the clonal relationships of isolates were analyzed using pulsed-field gel electrophoresis (PFGE). While screening for MRSA, we also identified heavy contamination of X-ray cassette surfaces with methicillin-resistant *Staphylococcus haemolyticus* (MRSH). MRSH is a mannitol-fermenting, coagulase-negative staphylococcus whose colonies resemble MRSA on mannitol salt agar containing oxacillin. METHODS ======= 1. Bacterial isolation ---------------------- Thirty-seven X-ray cassettes from the radiology department of a tertiary-care hospital were evaluated for MRSA contamination. Vertically stacked cassettes in the radiology department were sequentially sampled for cultures. For each X-ray cassette, the entire patient-contacting surface, which was made of polyvinyl chloride, was wiped with a cotton swab moistened with sterile normal saline. The screening swabs were inoculated on mannitol salt agar containing 6 µg/mL of oxacillin (MSO), and the plates were incubated at 35℃ for 48 hr. From each MSO plate, 1-3 yellow colonies suspicious for MRSA were subcultured on blood agar plates. Gram stains were performed from pure cultures of the isolates. All gram-positive cocci were tested for coagulase, DNase, and growth on mannitol salt agar. Isolates were identified using the MicroScan WalkAway system (Dade Behring, Sacramento, CA, USA). Antimicrobial susceptibility was determined by disk diffusion testing, and the results were interpreted using the Clinical and Laboratory Standards Institute breakpoint criteria \[[@B7]\]. In addition, PCR-based tests on the *mecA* gene were performed to confirm the identification of MRSA. 2. PFGE ------- For each cassette yielding positive cultures, 1 isolate was selected for analysis by PFGE. PFGE was performed as previously reported \[[@B8]\] with some modifications, including prolonged times for plug digestion and washing. Plugs were digested in a total of 100 µL of restriction enzyme buffer and 20 U of *Sma*I (New England Biolabs, Beverly, MA, USA) at 25℃ for 6 hr. Electrophoresis was performed using the CHEF Mapper system (Bio-Rad Laboratories, Hercules, CA, USA) in 0.5×Tris-borate-EDTA buffer. Migration conditions were as follows: temperature, 14℃; voltage, 6 V/cm; switch angle, 120°; and one linear switch ramp of 5-40 sec for 20 hr. PFGE patterns were analyzed using BioNumerics software (Applied Maths BVBA, Sint-Martens-Latem, Belgium). RESULTS ======= MRSA was isolated from 6 of the 37 X-ray cassettes (16.2%). Of the 6 contaminated X-ray cassettes, 4 (samples 1, 2, 3, 4) were adjacent to each other, and another 2 (samples 22, 23) were adjacent to each other while stored in the radiology department. Of the 4 adjacent X-ray cassettes, 2 cultures (samples 3, 4) yielded more than 100 colonies, and the remaining 2 cultures (samples 1, 2) yielded 3 and 20 colonies, respectively. The cultures of the other 2 adjacent X-ray cassettes each yielded less than 5 colonies. During strain identification, some colonies suspicious for MRSA on MSO were determined to be coagulase-negative staphylococci. Nearly all of these coagulase-negative staphylococci were identified as MRSH. MRSH was isolated from 19 of the 37 X-ray cassettes (51.4%). Fourteen of the cultures yielded less than 10 colonies; 2 cultures yielded 10-50 colonies, and the remaining 3 cultures yielded 50 or more colonies of MRSH. One cassette was contaminated with 2 colonies of *S. capitis*, a mannitol-fermenting coagulase-negative staphylococci. Among the MRSA contaminants (N=6), PFGE analysis revealed 2 major pulsotypes ([Fig. 1](#F1){ref-type="fig"}). An identical PFGE pattern was observed in 4 isolates (samples 1, 3, 4, and 22), and a second identical pattern was observed in 2 isolates (samples 2 and 23). Among the MRSH contaminants (N=19), PFGE analysis revealed 2 major pulsotypes ([Fig. 2](#F2){ref-type="fig"}); a single identical PFGE pattern was observed in 11 isolates (samples 5-9, 13, 14, 17, 19, 24, and 25) and another was observed in 5 isolates (samples 15, 16, 20, 21,and 26). DISCUSSION ========== The principal mode of MRSA transmission within a hospital is from patient to patient via transient colonization of hospital personnel\'s hands, which acquire the organism from direct patient contact or handling contaminated materials \[[@B9]\]. Furthermore, during radiology procedures, contact can occur between contaminated X-ray cassettes and a patient\'s clothing and skin or the radiology technician\'s hands. The role of inanimate objects in clinical infections is controversial, and there is little evidence demonstrating that decreasing environmental MRSA contamination lowers the rates of patient infection \[[@B10], [@B11]\]. However, the patient environment is considered to be a potential fomites for infection, and contamination of the hospital environment with multi-resistant bacteria can be controlled \[[@B10], [@B12]\]. The importance of reducing contamination in the hospital environment remains unclear. However, 2 of the 37 X-ray cassettes were heavily contaminated with MRSA, which could increase the risk of transmission within the hospital, such as to elastic bandages for wound care, patient\'s skin, and the hands of radiology personnel. The 6 MRSA isolates and 19 MRSH isolates each showed 2 major pulsotypes. For each organism, the pulsotypes were closely related. These findings suggest that cross-contamination of X-ray cassettes by MRSA and MRSH can result from a few common sources of contamination. Notably, we did not screen the technicians in the department of radiology for MRSA during the study period. Therefore, it is not clear whether the X-ray cassettes became contaminated by the medical personnel\'s hands or by surface-to-surface contact between the stacked X-ray cassettes. Although this study identified X-ray cassettes contaminated by methicillin-resistant staphylococci, we could not detect non-mannitol fermenting staphylococci, such as *Staphylococcus epidermidis*, or other pathogenic bacteria, because we only performed screening using MSO. As such, we cannot exclude contamination with another genus or coagulase-negative staphylococcus. The use of chromogenic medium can provide fast and accurate detection of MRSA contamination in the presence of coagulase-negative staphylococci producing yellow colonies on MSO. The best method for disinfecting contaminated X-ray cassettes is unclear. In our hospital, the surfaces of X-ray cassettes were disinfected once weekly with a gauze soaked in 70% ethanol. In order to prevent the transmission of microorganisms from contaminated X-ray cassettes to patients, an X-ray cassette cover could be used for patients who are infected or colonized with MRSA and for patients with oozing skin lesions. However, this study did not establish a role for the contaminated X-ray cassettes in the spread of MRSA, and the role of preventive measures requires further investigation. In addition, contamination rates might differ according to the prevalence of MRSA; for instance, the rate of MRSA in bloodstream infections is as high as 70% in Korea, a highly endemic country \[[@B13]\]. With the exception of MRSA isolates from 6 X-ray cassettes, many of the yellow colonies on MSO were identified as MRSH, another mannitol fermenting staphylococcus. Although less virulent than *S. aureus*, *S. haemolyticus* has been associated with septicemia and skin and soft-tissue infections \[[@B14], [@B15]\]. PFGE analysis demonstrated that most MRSH isolates in this study were identical or clonally related. This finding suggests that *S. haemolyticus* can spread in the hospital environment via medical devices, as reported in other studies using PFGE \[[@B16], [@B17]\]. In conclusion, we demonstrated that patient-contact surfaces of X-ray cassettes were contaminated with MRSA and MRSH. PFGE patterns indicated that these methicillin-resistant staphylococci were clonally related. Because X-ray cassettes are possible fomites for the transmission of MRSA and other pathogens, the application of contamination barriers or decontamination procedures should be considered. No potential conflict of interest relevant to this article was reported. {#F1} {#F2}

There are amount of requirements to control wave fields with desired patterns, such as non-diffracting[@b1], twisted[@b2] wave front. One of the most intriguing phenomena which attracted considerable research interest recently is the notably self-accelerating beam since the concept of Airy beam was introduced for optical wave[@b3][@b4][@b5][@b6][@b7]. These realization of self-accelerating beams in paraxial and non-paraxial domains propagating along designed trajectories indicates amount of potential applications, such as guiding micro-particles[@b8], producing curved plasma channels[@b9], and so on. In principle, these self-accelerating beams are formed based on the special solutions of wave equations or caustic theory[@b10]. As another classic wave, acoustic wave obeys the Helmholtz wave equation, indicating the possibility that it can be designed to propagate along desired trajectories. Recently, acoustic self-accelerating beam were demonstrated both numerically and experimentally with *active* phased arrays[@b11][@b12]. However, the sources in the *active* way require to be operated individually with electric techniques, resulting in the high cost and complexity. To avoid these significant limitations, considerable efforts have been dedicated to exploring the *passive* control of sound by means of the metasurface[@b13][@b14][@b15][@b16][@b17][@b18][@b19][@b20] or metascreen[@b21], which can be regarded as ultra-thin metamaterials[@b22][@b23][@b24][@b25]. To form self-accelerating beam in non-paraxial domain with excellent performance, the *passive* structures should possess the capacities of transmitting sound energy effectively, shifting the phase of incident wave covering 2*π* range, and holding a subwavelength feature to avoid the spatial aliasing effect[@b26]. These conditions are rarely realized simultaneously by the previous models, resulting in the fact that the non-paraxial self-accelerating beams and their physical features and potential applications were rarely explored. Furthermore, all the previous models are designed in two-dimensional space, which inevitably hinder the real applications. Actually, three-dimensional acoustic self-accelerating beams, if realized, could open a new degree of freedom for acoustic wave manipulations and have deep implications in acoustical applications where special control of sound is needed. For instance, the unique self-healing behavior of the beam could provide a promising solution to the narrow "acoustic window" resulting from the obstruction of the rib cage in ultrasonic ablation of liver tumors. Here we present the generation of a three-dimensional acoustic collimated self-accelerating beam with sourceless metascreen. By imposing a fine local phase shift profile on the metascreen, the sound energy could be delivered along a designed curved trajectory and then focused at a spot even with existing blocking obstacles in front of the spot and along the trajectory. Results ======= Illustration ------------ The desired three-dimensional acoustic collimated self-accelerating beam in non-paraxial domain is illustrated in [Fig. 1(a)](#f1){ref-type="fig"}. The metascreen possess the abilities of providing a local phase shift *ϕ*(*r*) on the incident sound field, consequently shaping the transmitted sound propagating along a desired trajectory *r* = *f*(*z*), and finally forming a focusing spot at a the intersectional region of the trajectory. The relationship between the phase shift profile and the desired trajectory could be retrieved from tracing each individual caustic ray and expressed as[@b11][@b12][@b27] where *k* is the sound wave number in the medium and *θ*(*z*) is the angle of the path \[cf. [Fig. 1(a)](#f1){ref-type="fig"}\]. Using this relation, the desired phase profile *ϕ*(*r*) can be calculated by finding the inverse tangent of the path slope As an example of the self-accelerating beam beyond the paraxial approximation, we employ a circle trajectory with center at (*r*, *z*) = (0, *r*~*b*~). The desired phase shift profile for forming such a bending beam from a normally plane wave is [Figure 1(b)](#f1){ref-type="fig"} shows the desired phase profile for the forming of a circle beam with *r*~*b*~ = 2.5*λ*. This phase profile illustrates the requirement of the phase shift profile provided by the metascreen for the desired beam with good performance. The requirement is the ability of providing a phase shift that can span a full 2*π* range in a controllable manner and rapidly varies along the metascreen in *r* direction[@b11][@b21]. The variation is in a subwavelength scale so that the metascreen needs to hold a fine spatial resolution when using discrete structures along *r* directions to avoid spatial aliasing effect. To illustrate the performance of the self-accelerating beam, we will place a spherical and a ring-like obstacle in front of the metascreen and along the propagating trajectory to obstruct the formation of the desired focused wave field. The big circle and the two small ones \[cf. [Fig. 1(a)](#f1){ref-type="fig"}\] in the *r* − *z* plane refer the spherical and the ring-like obstacle, respectively. Design ------ To realize the desired phase profile shown in [Fig. 1(b)](#f1){ref-type="fig"}, we use a three-dimensional subwavelength hybrid elements to construct the metascreen. [Figure 2(a)](#f2){ref-type="fig"} illustrates two adjacent elements in three-dimensional space to demonstrate the configuration of the metascreen. [Figure 2(b)](#f2){ref-type="fig"} shows an individual element in *r* − *z* plane consisting of four Helmholtz resonator (HRs) and a straight pipe. Here the series connection of the HRs acts as acoustic resistance to shift the phase of the incident wave[@b28]. The cavity series has a tunable width *w*~3~ to span the phase shift over a full 2*π* range. The functionality of the straight pipe with fixed length of *λ*/2 supporting Fabry-Pérot resonance[@b29][@b30][@b31] is to provide coupling resonances keeping relatively impedance matching[@b21]. Considering the fact that the transmission coefficient is determined by the coupling resonances between the Fabry-Pérot and Helmholtz resonance, the number of the HRs is selected to be four in order to provide enough coupled resonances so that transmission coefficient can keep high value while covering 2*π* range[@b21]. Distinct to the models in two-dimensional cartesian coordinates where the transmission coefficient, \|*p*~*t*~/*p*~*i*~\|, and phase shift, *ϕ*/(2*π*), is independent of the position of the individual element, these variables in cylindrical coordinates are related to the distance from the element to the axis, *s*~*n*~. This difference stems from the different volumes of the elements locating at different *s*~*n*~ even with identical geometrical parameters (such as *w*~1~, *w*~3~). [Figure 2(b,c)](#f2){ref-type="fig"} illustrate the simulated phase shift and transmission coefficient map as functions of the straight pipe width ratio, *w*~1~/*w* and the distance ratio, *s*~*n*~/*w*. Due to the symmetric geometry, only the positive part where *r* ≥ 0 is illustrated. The phase shift could cover 2*π* range for each elements (viz., 0 ≤ *s*~*n*~/*w* ≤ 99) when tuning 0.2 ≤ *w*~1~/*w* ≤ 0.8, even the one for individual elements changed slightly with different *s*~*n*~. Furthermore, the transmission coefficient is considerably high where the phase shifts covering 2*π*. According to the [Fig. 2(b)](#f2){ref-type="fig"}, it is readily to obtain a required phase shift profile for a corresponding beam. For example, we choose the radius of the circular trajectory as *r*~*b*~ = 2.5*λ*. In order to obtain a good shaping of the desired beam, the number of the individual elements composed of the whole metascreen is fixed to be 100. Then the *w*~1~/*w* needed for the desired phase shift profile \[cf. [Fig. 1(b)](#f1){ref-type="fig"}\] is illustrated as hollow black point \[cf. [Fig. 2(b,c)](#f2){ref-type="fig"}\]. It can be found that the metascreen can transmit sound with high efficiency greater than 91% and shift the incident phase covering 2*π*. The spatial resolution of the metascreen, viz., *w*, is as small as *λ*/10, which is fine enough to avoid the spatial aliasing effect. Collimated self-bending beam ---------------------------- The realization of our screen allows effective control of sound propagation along desired trajectory. The desired collimated self-bending beam is shown in [Fig. 3(a)](#f3){ref-type="fig"}. A boundary with a unity amplitude and a continuous phase profile is employed to form the self-bending beam. We construct the metascreen with 100 elements along *r* direction with desired geometrical parameters, *w*~1~/*w* and *s*~*n*~/*w*, shown in [Fig. 2(b,c)](#f2){ref-type="fig"}. The transmitted wave fields through the metascreen is shown in [Fig. 3(b)](#f3){ref-type="fig"} with a normally plane incident wave propagating along +*z* direction. The screen yields a discrete desired phase shift profile on the incident wave with spatial resolution *w* = *λ*/10. The self-bending beam is well established \[cf. [Fig. 3(b)](#f3){ref-type="fig"}\] and in a good shape of the desired propagating trajectory and then focused at the spot. Excellent agreement could be obtained by comparing the wave fields in [Fig. 3(a,b)](#f3){ref-type="fig"}. The excellent performance of the proposed metascreen owes to the fine spatial resolution, the high transmission and the fully controlled phase shift. The self-bending beam possesses the capacity to bypass solid obstacle due to the curved trajectory. From [Fig. 3(c)](#f3){ref-type="fig"}, one can observe that transmitted field pattern nearly keeps identical even with the existing scattering from the solid spherical obstacle (diameter 3*λ*) located in the region surrounded by the main lob. Additionally, the metascreen holds its own self-healing feature. A ring-like obstacle (diameter *λ*) located along the trajectory that blocks the main lob of the beam is added. The beam restores to its shape \[cf. [Fig. 3(d)](#f3){ref-type="fig"}\] after passing the obstacles and forms the desired focused spot. In order to qualify the observed features, a comparison of the normalized sound pressure level (SPL) along the *z* direction \[cf. [Fig. 3(e)](#f3){ref-type="fig"}\] shows that, even when both obstacles simultaneously occupy the space, the beam endows extremely robust against perturbations, owing to its caustic nature. Our metascreen not only can transmit normally incident plane waves but also any sound fields to form desired beams in homogeneous medium. Due to the fine spatial resolution of the metascreen, the width of the inlet and outlet of the individual element, *w*~1~, is in deep subwavelength scale so that the pressure along the *r* direction in these regions could be regarded as a uniform value. The designed metascreen should provide another phase shift profile to compensate the phase difference along the boundary in the incident side. As an example, a point source located at (*r*~*s*~, *z*~*s*~) = (0, −10*λ* − *h*) is employed to radiate a spherical wave. To form the same non-paraxial self-accelerating beam, the local phase shift provided by the metascreen can be expressed as where the second part with *z*=-*h* compensates the arrival phase difference of the point source along the boundary of the metascreen at the incident side. While the first part is same to Eq. [4](#eq4){ref-type="disp-formula"}. The realized collimated self-bending beam from the point source \[cf. [Fig. 4(a)](#f4){ref-type="fig"}\] propagating along the designed trajectory closely resembles the desired beam illustrated in [Fig. 3(a)](#f3){ref-type="fig"}, providing a solid support for the great capacity of our presented screen. It is also not surprising to observe that the non-paraxial accelerating beam can convincingly bypass solid obstacle due to the curved trajectory and hold its own self-healing feature \[cf. [Fig. 4(b)](#f4){ref-type="fig"}\]. A comparison of the SPL along the *z* direction for these cases \[cf. [Fig. 4(c)](#f4){ref-type="fig"}\] indicates that, even if both obstacles block the formation of the desired wave field, the self-bending beam could be reconstructed to propagate along the desired trajectory and focused behind the solid obstacles. Discussion ========== In conclusion, we have proposed a three-dimensional acoustic metascreen constructed by combining a series connection of four Helmholtz resonators with a straight pipe supporting Fabry-Pérot resonance. The elements of the metascreen can effectively transmit sound energy, steer the phase shift covering a full 2*π* range and hold a fine spatial resolution in *r* direction as small as *λ*/10 to avoid the spatial aliasing effect. With these great capacities, acoustic metascreen composed of 100 individual elements along the *r* direction was implemented to generate collimated non-paraxial self-bending beams, whose self-healing and bypassing behaviors were further demonstrated. The realization of the three-dimensional collimated self-accelerating beams should open a new degree of freedom for wave manipulations and have deep implications for various potential applications, especially in the fields of ultrasonic imaging, diagnosis and treatment. For instance, the beams may be used to generate negative radiation force to manipulate micro-particles. In additional, the metascreen may be employed to design novel ultrasonic transducers to overcome the "acoustic window" issue or deliver acoustic energy along designed arbitrary curvatures bypassing organs. Methods ======= Simulations are conducted with a commercial software based on finite elements method, COMSOL Multiphysics Version 5.1, in frequency domain with a fixed *λ* = 0.2 m. Considering the symmetry of the metascreen, two-dimensional axisymmetric models rather than three dimensional models are built for the simulations for reducing the calculating time. The HRs and the solid obstacles are made of steel with a density of 7800 kg/m^3^ and sound speed of 6100 m/s. The surrounding medium is air with its density 1.21 kg/m^3^ and sound speed 343 m/s. Perfectly matched layers are employed to mimic infinite space to obtain the sound fields shown in [Figs 3](#f3){ref-type="fig"} and [4](#f4){ref-type="fig"}. A plane wave with unit amplitude is employed as the incident wave in [Fig. 3](#f3){ref-type="fig"}. A point source located at (*r*, *z*) = (0, −10*λ* − *h*) radiates a spherical wave in [Fig. 4](#f4){ref-type="fig"}. The thermal dispassion and viscous loss are neglected in our simulations due to the fact the minimum width of the channels, *h*~2~, is \~61 times greater than the the thickness of the viscous boundary layers, , with *ω* and *μ* referring to angular frequency and the coefficient of dynamic viscosity. For higher frequencies, such as, 20000 Hz, *h*~2~ is just \~17 times bigger than *d*~*v*~ so that these effects need to be considered. The geometrical parameters of the elements should be re-optimized for good performance. Additional Information ====================== **How to cite this article**: Li, Y. and Assouar, M. B. Three-dimensional collimated self-accelerating beam through acoustic metascreen. *Sci. Rep.* **5**, 17612; doi: 10.1038/srep17612 (2015). Y.L. wish to thank Dr. Likun Zhang at The University of Texas at Austin for fruitful discussions. This work was supported by the FEDER "Fonds Européen de Développement Régional" (project "MASTER") and by the "Région Lorraine". **Author Contributions** Y.L. conceived the concept and performed simulated computations. M.B.A. contributed to the analysis of the results and supervised the project. Y.L. and M.B.A. contributed to the writing of the manuscript. {#f1} {ref-type="disp-formula"} with *r*~*b*~ = 2.5*λ*.](srep17612-f2){#f2} {#f3} {#f4}

Introduction {#section5-2050312117726196} ============ Behavioral and psychological symptoms of dementia (BPSD) reduce the quality of life (QOL) of dementia patients and their care providers.^[@bibr1-2050312117726196]^ They also increase a sense of burden on the care providers,^[@bibr2-2050312117726196]^ as well as the necessity for patients to be institutionalized.^[@bibr3-2050312117726196],[@bibr4-2050312117726196]^ BPSD are caused by genetic, neurobiological, psychological, and social aspects.^[@bibr5-2050312117726196]^ The first choice for BPSD is non-pharmacological intervention. The targets of non-pharmacological intervention are psychological and social aspects for BPSD. Various forms of non-pharmacological intervention, including reminiscence, music, and aroma therapies, have been implemented. However, the results of a recent comprehensive review suggest that the reliability of their effects has not been established or that the effects do not continue.^[@bibr6-2050312117726196][@bibr7-2050312117726196][@bibr8-2050312117726196][@bibr9-2050312117726196][@bibr10-2050312117726196]--[@bibr11-2050312117726196]^ This result is obvious. If those therapies are conducted once a week on Alzheimer's patients, their long-term effects are not expected, particularly when the predominant symptom is memory impairment. Even if the therapies are implemented every day for 40 min, the interval between the therapies is 23 h and 20 min. Furthermore, it is very difficult for experts and staff members to conduct those non-pharmacological therapies every day. Dementia does not involve impairment of a single category of cognitive ability, such as impaired memory, judgment, or orientation, and may include apraxia and agnosia.^[@bibr12-2050312117726196]^ Dementia is the general disturbance of these abilities,^[@bibr12-2050312117726196]^ and its essence is a decrease in self-awareness.^[@bibr13-2050312117726196][@bibr14-2050312117726196][@bibr15-2050312117726196]--[@bibr16-2050312117726196]^ Since humans have the ability to be self-aware, unlike other primates, they can be conscious of themselves as well as other people and the surrounding environment. Consciousness is and must be oriented toward something (intentionality).^[@bibr17-2050312117726196]^ Therefore, dementia patients who are conscious of themselves live in fear of and with anxiety over losing themselves. Bryden,^[@bibr18-2050312117726196]^ an Alzheimer's patient, wrote a book to express her feelings about the awareness of herself collapsing. She identified a spot (mole) on her leg one day and stated that "Maybe it's cancer, and if I do not get it treated, I could die as 'me', ..." She was afraid of losing her self-awareness more than her actual death. Therefore, non-pharmacological intervention for dementia patients with awareness of themselves collapsing needs to fulfill the following three requirements: (1) change the direction of patients' consciousness, which is oriented toward fear and anxiety, by stimulating and expanding their self-awareness, (2) apply stimulus on a continuous basis, and (3) needing only a small number of collaborators. We focused our attention on finger rings. Rings were a symbol of the succession to the throne in ancient Egypt, earlier than 1000 BC. In medieval Europe (the 13th century), rings symbolized knights' love for noble ladies.^[@bibr19-2050312117726196]^ Throughout human history, a significant number of legends about rings have emerged based on fierce battles for power and love, and the brilliance of rings has fascinated females. When a woman wears a ring, it attracts attention of herself, such as while eating meals and talking with other people seated at a table. Even when she is not looking at her hand wearing the ring, she will probably notice it if she moves her hand on top of the other one. Thus, once a female has put a ring on her finger, it will continuously promote her femininity without the support of care providers and change the direction of her consciousness. Therefore, rings worn by females are considered to fulfill the above-mentioned requirements. Thus, the objective of this study was to explore whether wearing a ring might afford reduction of BPSD in Japanese female patients with dementia. Methods {#section6-2050312117726196} ======= Subjects {#section7-2050312117726196} -------- The study was conducted in seven female patients with dementia, primarily those having BPSD and judged to be aware themselves of their intellect collapsing. The patients were staying at five small-scale nursing homes for the aged located apart from each other. All of them had been diagnosed as having Alzheimer's disease according to the nursing home records. Medications were kept unchanged during the study period. The patient sample size to investigate the efficacy of the ring on the BPSD and explore the inter-individual differences in the efficacy was set at 7. The length of stay at the nursing home prior to the start of this study was at least 10 months for all the seven patients so that change of the environment associated with accommodation into the home was unlikely to be the cause of the BPSD. Study design {#section8-2050312117726196} ------------ The study design was a single-case experimental study (A-B design). The subjects were first asked not to wear a ring during a baseline period of one week (A), and instructed to wear rings during a one-week intervention period (B). In addition, the study did not adopt A-B-A design. This was because the subjects with dementia might become confused, thinking that their rings had been stolen during a non-intervention period (A). Interventions {#section9-2050312117726196} ------------- The lead author and a coauthor as an observer visited the subjects 1 week before the initiation of the study (baseline period). The subjects were asked to choose 2 rings from among 12 different sample rings with variously colored stones displayed in a black jewelry case. Crystal glass from Swarovski Jewelry was used as the stones. The price of one set of crystal glass and a metal ring was approximately eight dollars. We asked the subjects to choose two rings because we wanted them to become strongly interested in rings. In the elderly nursing home, the lead author adjusted the circumferences of the rings using ring gauges and ring gauge sticks. The subjects were allowed to choose which fingers of the left hand they put the two rings on: the "fourth and fifth" or "third and fourth" fingers. We had received advice on the adjustment of the circumferences of rings from a clerk at a jewelry shop in advance. The glass stones selected by the subjects were put on the adjusted metal rings, and we brought them to the elderly nursing home on the first day of the intervention period. The subjects put on the rings at 9:00 and removed them at 19:00. They were left in an office and taken care of by staff members at night. The researchers conducted one-on-one observation of the subjects, keeping an appropriate distance from them, to prevent them from putting the rings into their mouth or pulling them off from their fingers forcibly. Assessments {#section10-2050312117726196} ----------- ### Overall profiles {#section11-2050312117726196} The lead author asked the nursing care providers about the overall profiles of the subjects. ### Interest in rings {#section12-2050312117726196} The subjects' interest in rings was recorded by the coauthor in charge of observation. ### Change in BPSD {#section13-2050312117726196} Changes in the BPSD were evaluated on the basis of the Japanese version of the Neuropsychiatric Inventory (NPI),^[@bibr20-2050312117726196]^ a scale for assessment of the psychological symptoms of dementia patients, and information about the BPSD recorded by the care providers. Nursing care providers who had worked for at least 4 days during both the one-week baseline and the intervention periods conducted assessment using the NPI. This assessment was conducted after work on the last day of each period or the following day. In addition, pocket-sized notepads were distributed to all care providers so that they could record when BPSD appeared during the two-week period. Most of the care providers, who served as the evaluators, were female. The NPI evaluation method and BPSD recording method were explained in advance to each evaluator. According to the NPI manual, a semi-structured interview is used for the assessment. In this study, the care providers were asked to conduct the NPI assessment after work on the last day of each period or the following day so as to avoid hazing of the evaluators' memory about their subjects. During the interview, knowledge is reconstructed from the interactions between the listener and speaker rather than by direct reporting of facts.^[@bibr21-2050312117726196]^ Thus, what each speaker said could vary depending on the listener. It was impossible for the same author to interview many care providers working at five facilities located apart from each other and differing in the time at which the daily duty hours ended. For this reason, in order to avoid interview of the care providers by different interviewers (an approach which could reduce data reproducibility), we adopted the care provider self-entry in which no interviewer was involved in the NPI evaluation process. ### Stage of self-awareness ability {#section14-2050312117726196} This represents a method we have developed for assessing the stage of self-awareness of patients with dementia and consists of theory-of-mind tasks, self-evaluation tasks, and self-consciousness tasks.^[@bibr13-2050312117726196][@bibr14-2050312117726196][@bibr15-2050312117726196]--[@bibr16-2050312117726196]^ The theory-of-mind tasks inquire the ability of the patients to conjecture their own and others' minds, the self-evaluation tasks inquire the ability of the patients to understand their own present conditions, and the self-consciousness tasks inquire the ability of the patients to be sensitive of a psychological border between the patient's self and others. This staging was carried out by the lead author and the nursing care provider in charge. ### Clinical Dementia Rating {#section15-2050312117726196} The Clinical Dementia Rating (CDR)^[@bibr22-2050312117726196]^ was used to assess the severity of general cognitive dysfunction. Assessment using the CDR was conducted by the nursing care provider in charge. Criteria for judgment of the effects {#section16-2050312117726196} ------------------------------------ The NPI consists of the following 10 categories: delusions, hallucinations, agitation/aggression, dysphoria, anxiety, euphoria, apathy, disinhibition, irritability/lability, and aberrant motor activity. The NPI score in each category was calculated by multiplying the frequency (0--4 points) by the severity (0--3 points). The higher the NPI score, the severer the level of a specific psychological symptom; an NPI score of 0 suggested that the absence of symptoms. The care providers conducting the NPI assessment were confined to those who had worked for at least 4 days of the week during both the baseline period and the intervention period. At the nursing homes, the care providers worked in shifts, and each subject was evaluated by four or five care providers. Therefore, for the same subject, while some care providers encountered BPSD, others did not. To improve the validity of the evaluation under such circumstances, where different evaluations of the same subject might be made by multiple care providers, the efficacy judgment criteria adopted two requirements for making the judgment "effective": (1) no evaluator judged that the NPI score increased during the intervention period from the baseline and (2) a majority of evaluators judged that the NPI score decreased from the baseline to 0 during the intervention period. In addition, triangulation (combining the rating by evaluators with the scenes of BPSD recorded by all care providers) was adopted. Ethical consideration {#section17-2050312117726196} --------------------- Informed consent was obtained from the subjects and their families. The study was conducted with the approval of the ethics committee of the university. Results {#section18-2050312117726196} ======= Of the seven subjects, two discontinued wearing the ring during the intervention period, as they were afraid of being forced to buy the ring. Therefore, the results are presented based on data from the remaining five subjects. The survey items included the age of the five subjects, their CDR, overall profiles, interest in rings during the intervention period, changes in the NPI scores and scenes of BPSD during the baseline and intervention periods, and stage of self-awareness ability. In the three patients who displayed interest in wearing rings (Cases 1, 2, and 4), scenes of BPSD related to the items of NPI that showed responses were described as recorded by the nursing care providers. Case 1: a female in her 70s, CDR: moderate {#section19-2050312117726196} ------------------------------------------ ### Overall profile {#section20-2050312117726196} She walked by herself only between the living room and dining hall, which is a few meters away, in the nursing home. The distance was relatively short, and she was able to walk safely. She engaged in daily conversation; however, she had developed a loss of memory and had some difficulty communicating with other people. Although she never used violence toward other people, she easily became agitated and yelled frequently. ### Interest in rings {#section21-2050312117726196} Although she told us that "a ring would not suit me" at first, she started to enjoy wearing a ring since other residents said to her, "you look so beautiful." When a staff member said to her, "Mrs. XX, the rings look beautiful on you," she answered, "Yes, it looks nice on me because I am beautiful." She often put up her hand and looked up at the ring. She also talked to another subject (Case 2), sitting next to her to describe "how beautiful the ring was." Comments of care providers: she previously used to laugh with her mouth open, but after she began to wear the ring, she began to smile gracefully, placing her hand on her mouth. She has also started to lock the door of her room before she uses the bathroom (the act of a female covering her mouth with her hand implies shyness and modesty in Japan). ### Changes in the NPI scores {#section22-2050312117726196} Of the four evaluators who judged that there was "agitation/aggression" during the baseline period, three answered that no "agitation/aggression" was observed during the intervention period (see [Table 1](#table1-2050312117726196){ref-type="table"}). Also, of the three evaluators who judged that there was "irritability/lability" during the baseline period, two answered that no "irritability/lability" was observed during the intervention period. ###### Changes in the NPI scores for Case 1 between the baseline and intervention periods.  Evaluators ------------------------- --------------------- -------------- -------------- --- Delusions --- 2(1 × 2) − 2(1 × 2) Agitation/aggression ◎ ○ ◎ ◎ 1(1 × 1) − 0 3(3 × 1) − 1(1 × 1) 2(1 × 2) − 0 2(1 × 2) − 0 Disinhibition ◎ 1(1 × 1) − 0 Irritability/lability ◎ ○ ◎ 1(1 × 1) − 0 3(3 × 1) − 1(1 × 1) 1(1 × 1) − 0 Aberrant motor activity ◎ 1(1 × 1) − 0 NPI: Neuropsychiatric Inventory. Symbols indicate changes in NPI scores between baseline and intervention periods. ◎ indicates decreased to 0. ○ indicates lower than that in baseline period. --- indicates no change. ○(○ × ○) − ○(○ × ○) indicates the NPI scores (frequency × severity) in the baseline and intervention periods. The higher the NPI score, the severer the psychological symptom. ### Scenes in which BPSD were identified {#section23-2050312117726196} Although she easily became agitated and yelled four times during the baseline period, such behavior was not observed during the intervention period (see [Table 2](#table2-2050312117726196){ref-type="table"}). ###### Scenes in which the BPSD of Case 1 were identified.  ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Period in which the rings were not worn ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- **Day 4**\ Agitation/aggression and irritability/lability (in the daytime): After eating lunch, she was folding waste newspaper and suddenly started yelling, "I haven't had a meal," when she had finished folding it. After a while, she came to the office and yelled at staff members again: "I haven't had a meal." Later on that day, she yelled again: "I haven't had a meal."\ Agitation/aggression and irritability/lability (in the evening): She yelled in the dining hall: "I hate that old woman."\ **Day 6**\ Agitation/aggression and irritability/lability (in the daytime): When she saw another resident removing name stickers off the tables in the dining hall, she yelled at that person: "Do not remove them."\ Agitation/aggression and irritability/lability (in the daytime): She shouted, "I am tired," in the dining room. Period in which the rings were worn The records include no descriptions of agitation/aggression and irritability/lability. ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- BPSD: behavioral and psychological symptoms of dementia. ### Stage of self-awareness ability {#section24-2050312117726196} Passed the theory-of-mind tasks. Case 2: a female in her 80s, CDR: moderate {#section25-2050312117726196} ------------------------------------------ ### Overall profile {#section26-2050312117726196} She moved around the nursing home by herself using a wheelchair and walked around her room while holding bed rails and other railings. She engaged in daily conversation; however, she had developed a loss of memory and had some difficulty communicating with other people. Although she never used violence toward other people, she easily became agitated and yelled frequently. Furthermore, she frequently complained of migraine and occasionally muttered the words, "I'll die soon." ### Interest in rings {#section27-2050312117726196} She did not hesitate to wear a ring. Wearing the ring on her finger, she looked at it over and over again. Every time the staff members talked to her, saying, "Mrs. XX, you look so beautiful" she flashed a pleasant smile. When she was told, "Mrs. XX, this ring looks very becoming on you," she behaved shyly. Every morning while putting a ring on her finger, she thanked a coauthor, and in the evening while removing it, she asked if she could keep it. ### Changes in the NPI scores {#section28-2050312117726196} All three evaluators who judged that there was "agitation/aggression" during the baseline period and both evaluators who judged that there was "dysphoria" and "irritability/lability" during the baseline period answered that they were not observed during the intervention period (see [Table 3](#table3-2050312117726196){ref-type="table"}). ###### Changes in the NPI scores for Case 2 between the baseline and intervention periods.  Evaluators ------------------------- -------------- --------------------- -------------- --- Delusions ◎ 2(1 × 2) − 0 Agitation/aggression ◎ ◎ ◎ 2(1 × 2) − 0 6(3 × 2) − 0 6(3 × 2) − 0 Dysphoria ◎ ◎ 1(1 × 1) − 0 4(2 × 2) − 0 Anxiety ◎ 6(3 × 2) − 0 Disinhibition ◎ 2(1 × 2) − 0 Irritability/lability ◎ ◎ 1(1 × 1) − 0 2(1 × 2) − 0 Aberrant motor activity --- ◎ --- 4(4 × 1) − 4(4 × 1) 6(3 × 2) − 0 4(4 × 1) − 4(4 × 1) NPI: Neuropsychiatric Inventory. Symbols indicate changes in NPI scores between baseline and intervention periods. ◎ indicates decreased to 0. --- indicates no change. ○(○ × ○) − ○(○ × ○) indicates the NPI scores (frequency × severity) in the baseline and intervention periods. The higher the NPI score, the severer the psychological symptom. ### Scenes in which BPSD were identified {#section29-2050312117726196} Although this subject easily became agitated and yelled eight times during the baseline period, such behavior was observed only once during the intervention period (see [Table 4](#table4-2050312117726196){ref-type="table"}). ###### Scenes in which the BPSD of Case 2 were identified.  ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Period in which the rings were not worn ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- **Day 1**\ Agitation/aggression and irritability/lability (in the morning): She asked a staff member: "I want to visit Ms. XX near the sweet shop. Let me go out." The staff member said to her, "You cannot go out alone," and she yelled, "Are you telling me that I am not allowed to go anywhere by myself? It drives me crazy."\ Agitation/aggression and irritability/lability (in the morning): She kicked the entrance door repeatedly while yelling, "Open the door. I can hardly breathe staying here all day."\ **Day 2**\ Agitation/aggression (in the daytime): She kicked a vending machine twice and put her fingers into the slot to look for coins.\ Agitation/aggression and irritability/lability (in the evening): She yelled at a staff member: "You gave the sweet that I had bought earlier today to the man in a wheelchair, didn't you? The staff here always annoys me."\ Agitation/aggression and irritability/lability (in the evening): She came to the office with her paper diapers in her hands and yelled at staff members: "Who took this out of the bag without my approval? I just can't believe it. Change it to a new one."\ **Day 4**\ Agitation/aggression and irritability/lability (at night): She yelled at a staff member: "Give the teeth (dentures) back to me. I do not want you to help me put them on."\ **Day 7**\ Agitation/aggression and irritability/lability (in the evening): She became angry and yelled saying that someone had taken paper diapers out of the bag in the living room without her approval.\ Agitation/aggression and irritability/lability (in the evening): When a staff member talked to her in a wheelchair, "Excuse us. Let us pass through," to let another resident walk through, she yelled, "There is enough space." Period in which the rings were worn **Day 2**\ Agitation/aggression and irritability/lability (in the morning): When a staff member talked to her, saying, "Let's go shopping," she yelled at the staff member: "I do not want to go," "I have nothing to buy," and "I do not want to buy anything." ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- BPSD: behavioral and psychological symptoms of dementia. ### Stage of self-awareness ability {#section30-2050312117726196} Passed the theory-of-mind tasks. Case 3: a female in her 80s, CDR: moderate {#section31-2050312117726196} ------------------------------------------ ### Overall profile {#section32-2050312117726196} She used a walker to move around the nursing home. She walked around her room while holding bed rails and other railings. She engaged in daily conversation; however, she had developed a loss of memory and had some difficulty communicating with other people. She went back and forth between the second floor, on which the living room was located, and the first floor using an elevator, presumably due to anxiety. She came to the office frequently and asked the same question to the staff members. She sometimes became paranoid. ### Interest in rings {#section33-2050312117726196} On the first day of the intervention period, she said, "This does not suit me" and "Give it to someone more beautiful." She hardly looked at her ring from the second day onwards. She only had a glimpse of her ring at 7:00 in the evening while removing it. When she was told by a person sitting next to her in the day-service center located in the vicinity of the nursing home, "That ring is beautiful," she answered proudly, "Yes, it is beautiful." ### Changes in the NPI scores {#section34-2050312117726196} There was no obvious change in the NPI score (see [Table 5](#table5-2050312117726196){ref-type="table"}). ###### Changes in the NPI score for Case 3 between the baseline and intervention periods.  Evaluators ------------------------- --------------------- --------------------- --------------------- ----------------------- ----- Delusions ◎ ◎ --- ∆ --- 6(3 × 2) − 0 8(4 × 2) − 0 6(3 × 2) − 6(3 × 2) 1(1 × 1) − 2(1 × 2) 12(4 × 3) − 12(4 × 3) Hallucinations ◎ 2(1 × 2) − 0 Agitation/aggression --- × --- 1(1 × 1) − 1(1 × 1) 0 − 6(3 × 2) 1(1 × 1) − 1(1 × 1) Dysphoria --- --- 1(1 × 1) − 1(1 × 1) 1(1 × 1) − 1(1 × 1) Anxiety --- --- --- --- 6(3 × 2) − 6(3 × 2) 3(3 × 1) − 3(3 × 1) 3(3 × 1) − 3(3 × 1) 2(2 × 1) − 2(2 × 1) Apathy ◎ 1(1 × 1) − 0 Disinhibition ◎ --- --- --- 2(1 × 2) − 0 3(3 × 1) − 3(3 × 1) 3(3 × 1) − 3(3 × 1) 3(3 × 1) − 3(3 × 1) Irritability/lability --- --- --- --- 2(2 × 1) − 2(2 × 1) 3(3 × 1) − 3(3 × 1) 6(3 × 2) − 6(3 × 2) 3(3 × 1) − 3(3 × 1) Aberrant motor activity --- --- --- --- 12(4 × 3) − 12(4 × 3) 4(4 × 1) − 4(4 × 1) 8(4 × 2) − 8(4 × 2) 12(4 × 3) − 12(4 × 3) NPI: Neuropsychiatric Inventory. Symbols indicate changes in NPI scores between baseline and intervention periods. ◎ indicates decreased to 0. --- indicates no change. ∆ indicates higher than that in baseline periods. × indicates increased from 0. ○(○ × ○) − ○(○ × ○) indicates the NPI scores (frequency × severity) in the baseline and intervention periods. The higher the NPI score, the severer the psychological symptom. ### Stage of self-awareness ability {#section35-2050312117726196} Passed the theory-of-mind tasks. Case 4: a female in her 90s, CDR: severe {#section36-2050312117726196} ---------------------------------------- ### Overall profile {#section37-2050312117726196} She used a wheelchair to move around the nursing home. She walked to the bathroom in her room with the support of a staff member. Although she engaged in daily conversation, she only had fragmentary memories and almost always kept on talking without listening. She often talked to an imaginary person when she was in her room. Although she was usually in a state of euphoria and enjoyed singing nursery rhymes and popular songs, she sometimes became grumpy suddenly. She often talked to herself. ### Interest in rings {#section38-2050312117726196} On the first day of the intervention period, she said, "I do not want this ring because I have no money." However, she stopped saying that after she had received an explanation a few times and understood that she did not have to pay for it. When a nursing care provider spoke to her, saying, "Mrs. XX, you look so beautiful," she answered pleasantly, "Yes, it is beautiful." She often put up her hand and looked up at the ring. She also talked to an imaginary person: "This ring is fancy" and "They gave me this ring." ### Changes in the NPI scores {#section39-2050312117726196} Of the five evaluators who judged that there was "irritability/lability" during the baseline period, three answered that it was not observed during the intervention period (see [Table 6](#table6-2050312117726196){ref-type="table"}). ###### Changes in the NPI score for Case 4 between the baseline and intervention periods.  Evaluators ------------------------- --------------------- --------------------- --------------------- --------------------- ----- Hallucinations ◎ ◎ --- --- --- 3(3 × 1) − 0 3(3 × 1) − 0 4(4 × 1) − 4(4 × 1) 3(3 × 1) − 3(3 × 1) 3(3 × 1) − 3(3 × 1) Agitation/aggression ◎ ○ --- 1(1 × 1) − 0 8(4 × 2) − 6(3 × 2) 3(3 × 1) − 3(3 × 1) Euphoria ◎ ○ --- --- 2(2 × 1) − 0 3(3 × 1) − 1(1 × 1) 6(3 × 2) − 6(3 × 2) 3(3 × 1) − 3(3 × 1) Apathy --- --- 6(3 × 2) − 6(3 × 2) 1(1 × 1) − 1(1 × 1) Disinhibition × --- 0 − 3(3 × 1) 3(3 × 1) − 3(3 × 1) Irritability/lability ◎ ◎ ◎ --- --- 2(2 × 1) − 0 3(3 × 1) − 0 6(3 × 2) − 0 3(3 × 1) − 3(3 × 1) 3(3 × 1) − 3(3 × 1) Aberrant motor activity × --- --- 0 − 2(2 × 1) 1(1 × 1) − 1(1 × 1) 3(3 × 1) − 3(3 × 1) NPI: Neuropsychiatric Inventory. Symbols indicate changes in NPI scores between baseline and intervention periods. ◎ indicates decreased to 0. ○ indicates lower than that in baseline period. --- indicates no change. × indicates increased from 0. ○(○ × ○) − ○(○ × ○) indicates the NPI scores (frequency×severity) in the baseline and intervention periods. The higher the NPI score, the severer the psychological symptom. ### Scenes in which BPSD were identified {#section40-2050312117726196} She experienced hallucinations and yelled four times during the baseline period. However, she yelled only once while removing the ring during the intervention period (see [Table 7](#table7-2050312117726196){ref-type="table"}). ###### Scenes in which the BPSD of Case 4 were identified.  ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Period in which the rings were not worn ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ **Day 1**\ Irritability/lability (in the daytime): She yelled, "Come out, Sesuo (name of a male)" and "What are you afraid of? How can you say that when you have never lived with me" while rattling the bedrail in the living room.\ **Day4**\ Irritability/lability (in the morning): She yelled at the wall: "Who are you, stupid? Go away."\ Irritability/lability (in the evening): As a staff member walked in front of her room, the staff member heard her yelling: "That's enough," "What are you talking about? Are you a fool? Don't be stupid."\ **Day 7**\ Irritability/lability (in the evening): She yelled, "You are all stupid" and "Get arrested." Period in which the rings were worn **Day 7**\ Irritability/lability (at midnight): As she did not take the rings off her fingers at 19:00, a staff member persuaded her to remove them when the staff member took her to the bathroom at midnight, and she held out her hand. The staff member removed the rings and put them in the case. When the staff member was leaving the room with the case, she yelled, "Do not take the rings," "Put them there," and "You are stealing my rings." ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ BPSD: behavioral and psychological symptoms of dementia. ### Stage of self-awareness ability {#section41-2050312117726196} The patient passed the self-evaluation tasks with ease, but she did not even try to think about the inquiries in the theory-of mind tasks, saying, "I just can't understand what these mean; I'm a punk!" Case 5: a female in her 70s, CDR: severe {#section42-2050312117726196} ---------------------------------------- ### Overall profile {#section43-2050312117726196} She had no difficulty walking. She often walked around the dining hall and talked to a mirror on a wall in the hall. She easily became angry when she called a staff member and the staff did not respond to her promptly. Although she engaged in some daily conversation, she almost always kept on talking and never listened. She told stories of her miserable life, which appeared to be fictitious, to the staff members with tears in her eyes. ### Interest in rings {#section44-2050312117726196} Even though she was told by her caregivers, "Mrs. XX, you look so beautiful," she was scarcely concerned about the ring. She did not look at her ring while she was putting it on or removing it. Even when she put the ring on or removed it by herself, she showed no expression. When she was having meals, she kept her ring hand on the table, but she hardly turned her eyes toward it. She said, "This ring is beautiful," only once during the observation period. ### Changes in the NPI scores {#section45-2050312117726196} There were no changes in the NPI scores (see [Table 8](#table8-2050312117726196){ref-type="table"}). ###### Changes in the NPI scores for Case 5 between the baseline and intervention periods.  Evaluators ------------------------- --------------------- ---------------------- --------------------- ----- Delusions --- --- --- ∆ 8(4 × 2) − 8(4 × 2) 4(4 × 1) − 4(4 × 1) 3(3 × 1) − 3(3 × 1) 3(3 × 1) − 4(4 × 1) Hallucinations --- --- --- ∆ 8(4 × 2) − 8(4 × 2) 4(4 × 1) − 4(4 × 1) 4(4 × 1) − 4(4 × 1) 3(3 × 1) − 4(4 × 1) Agitation/aggression --- ◎ ∆ 8(4 × 2) − 8(4 × 2) 1(1 × 1) − 0 3(3 × 1) − 8(4 × 2) Dysphoria --- ◎ --- --- 8(4 × 2) − 8(4 × 2) 1(1 × 1) − 0 1(1 × 1) − 1(1 × 1) 1(1 × 1) − 1(1 × 1) Anxiety --- ○ 4(4 × 1) − 4(4 × 1) 3(3 × 1) − 1(1 × 1) Euphoria ○ --- × ∆ 6(3 × 2) − 3(3 × 1) 4(4 × 1) − 4(4 × 1) 0 − 2(1 × 2) 1(1 × 1) − 4(4 × 1) Apathy ○ --- --- --- 3(3 × 1) − 2(2 × 1) 3(3 × 1) − 3(3 × 1) 9(3 × 3) − 9(3 × 3) 3(3 × 1) − 3(3 × 1) Disinhibition ○ --- ○ ○ 8(4 × 2) − 4(4 × 1) 1(1 × 1) − 1(1 × 1) 3(3 × 1) − 2(2 × 1) 6(3 × 2) − 1(1 × 1) Irritability/lability ○ --- --- --- 8(4 × 2) − 4(4 × 1) 2(2 × 1) − 2(2 × 1) 6(3 × 2) − 6(3 × 2) 2(1 × 2) − 2(2 × 1) Aberrant motor activity --- --- ○ --- 8(4 × 2) − 8(4 × 2) 3(3 × 1) − 3(3 × 1) 12(4 × 3) − 8(4 × 2) 9(3 × 3) − 9(3 × 3) NPI: Neuropsychiatric Inventory. Symbols indicate changes in NPI scores between baseline and intervention periods. ◎ indicates decreased to 0. ○ indicates lower than that in baseline period. --- indicates no change. ∆ indicates higher than that in baseline periods. × indicates increased from 0. ○(○ × ○) − ○(○ × ○) indicates the NPI scores (frequency × severity) in the baseline and intervention periods. The higher the NPI score, the severer the psychological symptom. ### Stage of self-awareness ability {#section46-2050312117726196} Not passed the theory-of-mind tasks, but passed the self-evaluation tasks. Discussion {#section47-2050312117726196} ========== The effect of wearing a ring was evident in the three patients who displayed interest in wearing a ring (Cases 1, 2, and 4), whereas in the two patients who showed no interest in wearing a ring (Cases 3 and 5), no effect of wearing a ring was seen. As for specific NPI items, rings effectively alleviated "agitation/aggression" and "irritability/lability" in Case 1, "agitation/aggression," "dysphoria," and "irritability/lability" in Case 2, and "irritability/lability" in Case 4; rings reduced "irritability/lability" in all three patients. Once a ring was worn on the finger, the "irritability/lability" vanished. Then, under what situation did the anger surge up? Aristotle^[@bibr23-2050312117726196]^ defined anger as a strong desire for revenge on someone who has explicitly slighted the person, which is accompanied by pain. In other words, anger is an emotion felt by a person in relation to ego (self-awareness) when his/her self-esteem has been reduced.^[@bibr24-2050312117726196]^ Cases 1 and 2 passed the theory-of-mind tasks in the evaluation of self-awareness. In Case 4, the patient passed the self-evaluation tasks with ease, but she did not even try to think about inquiries in the theory-of mind tasks, saying, "I just can't understand what these mean; I'm a punk!" The theory-of-mind tasks inquire the ability of the patients to conjecture their own and others' minds, while the self-evaluation tasks inquire the ability of the patients to understand their own present conditions. Case 4 did not try to tackle the theory-of mind tasks, yet she caught on to her own mind ("I'm a punk!"). In other words, these three women were aware themselves of their intellect collapsing. Meanwhile, we who take care of these individuals with dementia are liable inadvertently to make light of them. We lie just to suit the occasion in case we are at a loss for answering an obstinate complaint by a subject with dementia. We are in such a situation just because we generally have in mind that patients with dementia can forget everything after all. It seems to be an undeniable fact that nursing care providers tend to slight subjects with dementia. Women with dementia who are still capable of conjecturing others' minds, however, are sensitive to being slighted. The self-esteem of those women who are aware of their own collapsing intellect is hurt and diminishes when they are slighted by other persons. Individuals with low self-esteem are inclined to get angry and display aggression.^[@bibr25-2050312117726196]^ In women with low self-esteem, anger readily surges up when they are slighted by others. This low self-esteem is considered to have been elevated upon wearing a ring. It seems that narcissism may be involved therein. At first, they said to us, "These rings do not suit me" and "You should give them to someone more beautiful." Their remarks suggest that Japanese females of this generation believe that rings should be worn by beautiful people. The three females who responded to rings did not necessarily wear them happily at first. However, the remarks of the nursing care providers and other residents, such as, "Mrs. XX, you look so beautiful," motivated them to wear them. As they continued to wear rings every day, the rings became increasingly integrated into their lives. Now that the rings, accessories only worn by beautiful people from their point of view, had become an integrated part of them, they became increasingly assured that they themselves were beautiful every time people said to them, "Mrs. XX, you look so beautiful," and their self-esteem was improved based on their self-evaluation. It did not take long for them to be able to believe that they were beautiful. Narcissism is "the libidinal complement to the egoism of the instinct of self-preservation."^[@bibr26-2050312117726196]^ That is to say, people cannot continue to live without self-love. For women, "beautiful" is a word of conjuration that improves their self-esteem. After putting rings on her hand, Case1 laughed gracefully while covering her mouth. She also started to lock the door of her room before she used the bathroom. These behaviors are typical examples of the power of the word of conjuration "beautiful" to enhance the sense of femininity and self-esteem. Those with high self-esteem do not really care when other people do not pay much attention to them.^[@bibr25-2050312117726196]^ Since their self-esteem had been improved by other people's remarks, such as, "Mrs. XX, you look so beautiful," they were more tolerant of acts of slight. As a result, "agitation/aggression" and "irritability/lability" are alleviated. In Case 1 and Case 2, "agitation/aggression" and "irritability/lability" subsided when the patients wore rings. "Irritability/lability" and "agitation/aggression" originally represent similar concepts to each other. The scenes where these two combined concepts arise are conjectured to be often identical so that it would appear probable that "agitation/aggression" subsided in association with diminishing emergence of "irritability/lability." In Case 2, ring-wearing led to the subsidence of "dysphoria." This patient frequently complained of migraine and occasionally muttered the words, "I'll die soon." Depressed mood is not uncommon among patients with Alzheimer's disease and is characterized by dysthymia, while the sense of sorrow and feeling of guilt/self-accusation are rather inconspicuous in these patients, unlike tとin patients with depression. Such depressed mood in patients with Alzheimer's disease readily improves in many instances in response to a favorable environment and optimized personal relations.^[@bibr27-2050312117726196]^ The heightened self-esteem of a woman who is told, "Mrs. XX, you look so beautiful," by the staff members made her feel happy. As a result, the "dysphoria" subsides in such a case. The use of rings has thus proved to have particular efficacy in these three cases, although this is a rather small number of patients. Rings led the nursing care providers and other people to say, "beautiful" to the residents of the nursing home. The researchers did not ask the care providers to say, "beautiful" to the residents. Thus, wearing of rings, which has fascinated females in general, was found to attract the attention of the female care providers in this study, even when it was a low-priced one. The power of rings led them to say that, and existing non-pharmacological interventions do not have such power. The results of this study suggest that an effective non-pharmacological intervention for patients with dementia is an approach that helps people involved in dementia care to improve the self-esteem of dementia patients, who tend to have their feelings hurt easily, albeit unintentionally. In other words, the target of non-pharmacological interventions is the surrounding people, not the dementia patients themselves. During this study, the coauthor, who had been monitoring the study, observed and recorded the interest of the subjects in the rings. That is, the author also played the role of an evaluator, in part. This was aimed at allowing the care providers to focus on observation and recording of the scenes of the BPSD. It was considered that if the recording of the subjects' interest in the rings was assigned to the care providers, the descriptions of the care providers would be confined to the interest of the subjects in the rings. In such an event, it would have been impossible to observe communication related to the ring between the subject and the care provider, and it would have been difficult to find, for example, that when the care provider said to her, "Mrs. XX, you look so beautiful," the subject's self-esteem would have been elevated and her "agitation/aggression" and "irritability/lability" reduced. We believe that the recording of subjects' interest in the rings by the coauthor worked positively in approaching the main aspect of the ring's efficacy, with its benefit outweighing the risk of biases. Limitations and open issues of the study {#section48-2050312117726196} ======================================== The results of this study suggest that in subjects with dementia manifesting BPSD, wearing of rings can alleviate "agitation/aggression" and "irritability/lability." These are included as items in the Cohen-Mansfield Agitation Inventory (CMAI)^[@bibr28-2050312117726196]^ focusing on agitation. CMAI allows quantitative evaluation of the frequencies of 29 agitation-related items based on responses provided on a 7-grade scale. Because each of these items is expressed with verbs corresponding to concrete behaviors, the interviewee can answer the questions easily, without being required to interpret the sentences written in the questionnaire. This study was designed as an exploratory study in seven subjects. It is necessary to verify the results in a larger number of subjects and to confirm the indications of the "ring" approach quantitatively using the CMAI. An essential remedy for alleviating the BPSD in persons with dementia who are aware of their own collapsing intellect appears to be to elevate the self-esteem of these individuals. Self-esteem is not heightened unless it is taken good care of by other persons. Another open issue is to discover and develop other means, in addition to rings, by which care providers can show their appreciation to persons with dementia. **Declaration of conflicting interests:** The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. **Ethical approval:** Ethical approval for this study was obtained from Ethics Committee of Osaka Yukioka University (Approval number: 0008 (10/10/2013)). **Funding:** The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by JSPS KAKENHI (grant number 25350653). **Informed consent:** Written informed consent was obtained from all subjects before the study. Written informed consent was obtained from legally authorized representatives before the study.

Introduction ============ Stickler syndrome is a genetically and clinically heterogeneous condition first described in 1965 by Gunnar Stickler as an inherited progressive arthro-ophthalmopathy \[[@r1]\]. Although Stickler syndrome is the most frequently inherited cause of retinal detachment in childhood, this disorder is rare, and the reported prevalence in the United States is 1:10,000 \[[@r2]-[@r4]\]. The autosomal dominant (AD) Stickler syndrome subtypes include Stickler syndrome type I (STL1, OMIM [108300](http://omim.org/entry/108300)), Stickler syndrome type II (STL2, OMIM [604841](http://omim.org/entry/604841)), and Stickler syndrome type III (STL3, OMIM [184840](http://omim.org/entry/184840)). STL1 results from altered type II collagen molecules encoded by the collagen type II alpha 1 (*COL2A1*) gene while STL2, caused by mutations in the collagen type XI alpha 1 (*COL11A1*) gene, and STL3, caused by mutations in the collagen type XI alpha 2 (*COL11A2*) gene, affect the type XI collagen molecules \[[@r5]\]. Since 2006, mutations in the collagen type IX alpha 1 (*COL9A1*) and collagen type IX alpha 2 (*COL9A2*) genes, encoding for type IX collagen, have been reported as causal for an autosomal recessive form of Stickler syndrome \[[@r6],[@r7]\]. Type II, IX, and XI collagen molecules are major extracellular matrix components in the hyaline cartilage and the vitreous, and are structurally and functionally associated \[[@r8]-[@r10]\]. Affected individuals with mutations in STL1 and STL2 demonstrate a combination of ocular and systemic manifestations. In contrast, STL3 involves only non-ocular manifestations; affected individuals present with systemic malformations such as deafness and cleft palate \[[@r7],[@r10]-[@r13]\]. Ocular features associated with AD Stickler syndrome consist of myopia, vitreous degeneration, radial perivascular retinal degeneration, presenile cataract, and high incidence of retinal detachments \[[@r14]\]. The vitreous phenotype depends on the AD Stickler syndrome type, since STL1 vitreous demonstrates a congenital retrolental membrane while patients with STL2 demonstrate vitreous with fibrillar or beaded aspect \[[@r15]-[@r17]\]. Non-ocular manifestations of AD Stickler syndrome include arthropathy, hearing impairment, flat midface, midline clefting, and skeletal abnormalities, but these findings can be highly variable, even within a family \[[@r5]\]. Autosomal recessive forms of Stickler syndrome may present with ocular features, such as high myopia and vitreoretinal degeneration but no presenile cataract, and systemic features may include but are not limited to short stature, hearing impairment, facial structural abnormalities but no midline clefting, and causal gene-dependent spondyloepiphyseal dysplasia \[[@r7]\]. The expression of the *COL2A1* gene, associated with STL1, is tissue-dependent, due to an alternative splicing of exon 2, which results in two isoforms of type II collagen \[[@r18]\]. Type IIB is the shorter form with 53 exons (exon 2 spliced), and is mainly expressed in the cartilage while type IIA is the longer form with 54 exons (including exon 2) and is predominantly present in the vitreous \[[@r18]\]. Thus, mutations in *COL2A1* exon 2, such as a premature stop codon, may lead to the ocular variant of the STL1. This ocular condition shares typical STL1 ocular features, such as membranous vitreous and radial perivascular retinal degeneration, but shows none or few systemic manifestations \[[@r3],[@r19]-[@r21]\]. To date, several families harboring *COL2A1* exon 2 mutations have been reported \[[@r3],[@r19],[@r20],[@r22]-[@r25]\]. In these families, the reported penetrance of vitreoretinal degeneration is more than 90% in affected patients at the age of 20 \[[@r19],[@r20]\]. Wagner syndrome (OMIM [14200](http://omim.org/entry/14200)) is also associated with the extracellular matrix component gene versican (*VCAN*; alternatively called *CSPG2*), identified in 2005 by Miyamoto et al., which encodes for a large chondroitin sulfate proteoglycan versican \[[@r26]-[@r30]\]. We confirmed the involvement of this gene in 2009 by describing an intronic base pair splice site substitution in *VCAN* segregating with the disease in a multigenerational family resulting in a truncated protein affected by the splicing \[[@r31]\]. Wagner syndrome clinical features involve ocular manifestations close to those found in Stickler syndrome. However, patients also complain of nyctalopia, where the vitreous phenotype is different as characterized by an optically empty aspect with avascular strands and veils or fibrillary condensations \[[@r32]\], retinal detachments occur rarely, and retinal features demonstrate retinitis pigmentosa--like "bone spicule" pigmentary atrophy \[[@r20]\] with accompanying electroretinogram abnormalities. Nevertheless, in some cases, distinguishing Wagner syndrome from the ocular-only variant of STL1 may be difficult. Erosive vitreoretinopathy syndrome (OMIM [143200](http://omim.org/entry/143200)) is another ocular-only disorder, first reported by Brown et al. \[[@r33]\], that can also lead to indeterminate diagnosis based on clinical features only \[[@r30]\]. The clinical features of erosive vitreoretinopathy syndrome include night blindness, visual field defects, and chorioretinal atrophy. The variability in the clinical presentations of vitreoretinal disease phenotypes underscores the importance of careful clinical assessment. We clinically and genetically report a three-generation Caucasian family from the southeast United States demonstrating AD vitreous degeneration with variable phenotypes among affected members. Over the years, one affected member was initially diagnosed clinically with Stickler syndrome and then Wagner syndrome by her ophthalmologist. We conducted Sanger sequencing of the *COL2A1* ([NM_001844](http://www.ncbi.nlm.nih.gov/nuccore/NM_001844)) and *VCAN* ([NM_004385](http://www.ncbi.nlm.nih.gov/nuccore/NM_004385)) genes to delineate the genetic etiology of disease in this family. Methods ======= Study subjects -------------- Ten individuals (six affecteds, four unaffecteds) from a three-generation Caucasian family was recruited. All consenting family members (four males, six females) were recruited under the approval of the Duke University Institutional Review Board according to the principles of the Declaration of Helenski, under the research protocol entitled "Clinical and Molecular Analysis of Genetic Eye Disorders", to include molecular genetic testing (protocol number Pro00008040). Individuals underwent ophthalmic examinations that included health histories regarding systemic issues such as cleft palate, midline defects, skeletal or joint abnormalities, and early onset arthritis. The clinical evaluation included assessment tests of Early Treatment Diabetic Retinopathy Study visual acuity (Snellen equivalent) and intraocular pressure, slit-lamp inspection of the anterior segment, and indirect ophthalmoscopy to inspect the fundus \[[@r34],[@r35]\]. Genomic DNA was extracted using AutoPure LS® DNA Extractor and PUREGENE™ reagents (Gentra Systems Inc., Minneapolis, MN) from blood or saliva samples. DNA samples were also collected from 1,142 unrelated ethnically matched Caucasian healthy control participants. Gene screening and sequence analysis of collagen type II alpha 1 and versican genes ----------------------------------------------------------------------------------- Primers for PCR and sequencing were designed to cover coding and untranslated gene regions, including intron--exon boundaries, using the ExonPrimer and [Primer3](http://frodo.wi.mit.edu/primer3/) programs ([Helmholtz Zentrum, Munich](http://ihg2.helmholtz-muenchen.de/ihg/ExonPrimer.html), Germany). Primers were selected to produce amplified product sizes not to exceed 900 bp for optimal sequence output and reading. Large exons or untranslated gene regions were covered with overlapping amplicons, with a minimal 50 bp of overlapped sequence. All 54 *COL2A1* exons and 15 *VCAN* exons were examined. Appendix 1 displays the optimized primer sequences used for *VCAN* and *COL2A1* screening. Genomic DNA of two affected individuals (II:2 and II:3; [Figure 1](#f1){ref-type="fig"}) of the study family was initially screened for sequence variations in the *COL2A1* and *VCAN* genes. The DNA of the remaining family members was subsequently screened to determine and confirm sequence variants segregation. {#f1} PCR was conducted using an Eppendorf Mastercycler Pro S® with a standard touchdown PCR protocol. PCR amplicons were visualized with 2% agarose gel electrophoresis. BigDye™ Terminator 3.1 was used to perform sequencing reactions, and ABI3730XL robotics was used to process the DNA fragments (Applied Biosystems Inc. \[ABI\], Foster City, CA). The Sequencher® 5.0 Software (Gene Codes, Ann Arbor, MI) was used to analyze the base pair calls. Sequences of affected and unaffected individuals were aligned to a known reference genomic sequence ([UCSC Genome Browser](http://genome.ucsc.edu/)) and compared for sequence variation. Sorting Intolerant From Tolerant ([SIFT](http://sift.jcvi.org/)) \[[@r36]\] and Polymorphism Phenotyping ([PolyPhen2](http://genetics.bwh.harvard.edu/pph2/index.shtml)) \[[@r37]\] software tools were used to predict mutational consequence of all *COL2A1* and *VCAN* variations segregating with disease. Genotyping ---------- Applied Biosystems (ABI) TaqMan® SNP Genotyping assays were designed and employed to measure the allelic frequencies in 1,142 ethnically matched control DNA samples. Variants of interest were screened with PCR assay technology using TaqMan probes according to the manufacturer's protocol (Applied Biosystems). Alleles were detected and allelic discrimination were analyzed with ABI Prism® 7900HT Sequence Detection System and ABI Sequence Detection Systems 2.4 software, respectively (Applied Biosystems). For quality control, positive and negative controls were run in the same experiment. Complementary deoxyribonucleic acid tissue expression ----------------------------------------------------- We investigated the expression of the *COL2A1* messenger ribonucleic acid (mRNA) construct(s) in fetal ocular tissues to verify the presence of type IIA and/or type IIB isoforms. Fetal ocular tissue panels not affected with disease were established internally by acquisition of whole eye globes from Advanced Bioscience Resources (Alameda, CA). Twenty-four-week fetal eyes were obtained and preserved in RNAlater® Foster City, CA within minutes of abortion and shipped overnight on ice. Whole globes were dissected the same day as they arrived, and specific ocular tissues were isolated by snap-freezing the samples and storing at −80 °C until RNA extraction. RNA was extracted from each tissue sample independently using the Ambion Foster City, CA mirVana [Total RNA Extraction Kit](http://www.ambion.com/techlib/prot/fm_1560.pdf) per the protocol. The tissue samples were homogenized in Ambion's lysis buffer using an [Omni Bead Ruptor 24 Homogenizer](http://www.omni-inc.com) per protocol. Reverse transcription reactions were performed with Invitrogen's SuperScript™ III First-Strand Synthesis kit to obtain cDNA (Life Technologies, Grand Island, NY). In-house fetal eye cDNA was amplified using primers that spanned multiple exons, not to exceed 600 bp (Appendix 1). PCRs were run using a standard protocol. Visualization of the PCR products was done on a 2% agarose gel through electrophoresis at 120 V for 50 min. Products with exon 2 were expected to amplify at 510 bp, whereas those that did not contain the exon 2 were expected to produce a product of 303 bp. Band extraction, purification, and sequencing of the bands were conducted to verify the amplicon. Results ======= Clinical features ----------------- A three-generation family with affected member variable ocular-only diagnosis of either Stickler or Wagner syndrome was ascertained. The disease appeared to be transmitted in an AD inheritance pattern and showed variable expressivity with 100% penetrance ([Figure 1](#f1){ref-type="fig"}). Six affected and four unaffected individuals participated in the study. Clinical data were obtained where available. The proband, patient III:3, was a 40-year-old woman with moderate myopia and history of retinal detachment in the left eye (OS) at the age of 18 years. Periodically, she underwent prophylactic peripheral laser photocoagulation treatment bilaterally. She had cataract extraction surgery of both eyes at the age of 37 (right eye, OD) and 40 (OS). Fundus examination showed bilateral vitreous syneresis, with a vitreous membrane in the left eye, and bilateral radial extensive lattice degeneration. She had no history of systemic manifestations. Patient III:2 was a 40-year-old woman and the monozygotic twin sister of III:3. Patient III:2 presented with moderate myopia and underwent retinal surgeries at the age of 18 (OD) and 33 (OS) for retinal detachment (OD) and epiretinal membrane and large retinal tears (OS). Subsequently, she was treated with extensive peripheral prophylactic photocoagulation. She underwent cataract extractions bilaterally at the age of 37 (OS) and 38 (OD). Fundus examination revealed bilateral radial lattice degeneration and bilateral vitreous syneresis. She had no history of systemic manifestations. Patient II:2 was the 65-year-old mother of the twins. She had a history of bilateral congenital cataract extractions. Her fundus examination demonstrated bilateral vitreous syneresis with vitreous membranes. Patient II:3 presented with a cleft palate at an early age. Patients IV:3 and IV:4 presented with a history of retinal detachments but no systemic manifestations. We were not able to collect clinical data from the deceased individuals I:1 and I:2, although historical accounts from II:1 revealed her mother (I:2) had a history of retinal problems. Molecular genetic analysis -------------------------- Initial genomic DNA sequencing of two affected individuals (II:2 and II:3) was conducted in the *VCAN* and *COL2A1* genes. For *VCAN*, no sequence variants segregated with the disease status were identified compared to the control DNA and to published reference sequences. For the *COL2A1* gene, we identified 18 single nucleotide variations: one nonsense variant, one known missense variant, two known coding-synonymous variants, and 14 intronic single nucleotide polymorphisms (Appendix 1). The nonsense mutation was a C to A change ([Figure 2](#f2){ref-type="fig"}) in exon 2 (c.258C\>A; [NM_001844.4](http://www.ncbi.nlm.nih.gov/nuccore/NM_001844.4)), converting codon TGC for cysteine at position 86 to codon TGA for a premature stop codon (Cys86Stop; [NP_001835.3](http://www.ncbi.nlm.nih.gov/nuccore/NP_001835.3)). Subsequent sequencing of the remaining family members showed mutation cosegregation with the disease phenotype ([Figure 2](#f2){ref-type="fig"}). , +1 in cDNA numbering corresponding to the A of the methionine translation initiation codon) converting a cysteine codon to a stop codon in two affected individuals while the mutation is not present in two unaffected individuals. DNA analysis of all other affected family members cosegregated with this mutation while the mutation was not present in all unaffected family individuals.](mv-v19-759-f2){#f2} Genotyping was performed for the c.258C\>A mutation in the *COL2A1* gene for an additional 1,142 unrelated ethnically matched controls (2,284 chromosomes). The mutation was not present in the control samples. Tissue expression ----------------- We examined *COL2A1* expression across normal fetal eye tissues to verify the presence or absence of type IIA and/or type IIB isoforms ([Figure 3](#f3){ref-type="fig"}). Both *COL2A1* mRNA isoforms (type IIA and type IIB) were expressed in the fetal retina/retinal pigment epithelium and choroid (Appendix 1). The *COL2A1* mRNA type IIB isoform (excluding exon 2) was expressed in the sclera, optic nerve, and cornea. Gel extraction and Sanger sequencing of the specific product bands confirmed our findings of the isoforms except for the optic nerve, where the IIA isoform was not sequenced. {#f3} Discussion ========== We report a nonsense mutation in a large Caucasian family consisting of six affected individuals variably diagnosed with Stickler and Wagner syndromes. A base pair change at c.258C\>A leading to a premature stop codon in exon 2 of *COL2A1* was cosegregated with the disease status. Sequencing of ocular tissues confirmed the presence or absence of exon 2, demonstrating that isoforms may be ocular tissue specific. The mutation was not present in more than 2,000 chromosomes, validating the rarity of this mutation and confirming Stickler syndrome has a predominant ocular-only phenotype. Two striking features of Stickler syndrome are, as in our reported family, the high penetrance and variable expressivity. In the literature, the same mutation as in our family (c.258C\>A; [NM_001844.4](http://www.ncbi.nlm.nih.gov/nuccore/NM_001844.4)) demonstrated high penetrance \[[@r3],[@r19],[@r20]\]. In families harboring alternative *COL2A1* exon 2 mutations, the ocular manifestation penetrance was also high---from 90% \[[@r20]\] to 100% \[[@r24],[@r38]\]. The variable expressivity, even within the same family \[[@r39]\], contrasts with the high disease penetrance: In most reported *COL2A1* exon 2 mutations, ocular features were variable as either myopia, retinal detachment \[[@r22]\], or retinal degeneration \[[@r23]\] could be absent in affected patients. Furthermore, two common ocular features in *COL2A1* exon 2 mutations are vitreous degeneration and radial perivascular retinal degeneration \[[@r3],[@r19],[@r20]\]. Systemic manifestations were rarely associated with *COL2A1* exon 2 mutations, as manifestations were present in few cases \[[@r3],[@r19],[@r20],[@r22]\]. In our family, only one affected individual presented with cleft palate. Underlying causes of variable expressivity in ocular-only STL1 are still undetermined. However, in recent years the phenotypic variability of exon 2 mutations has been hypothesized to be due to degradation of mRNA by nonsense-mediated decay (NMD) or synthesis of alternatively spliced protein \[[@r21]\]. NMD is a regulation pathway involving the targeted degradation of mRNA that contains a premature stop codon. In this way, NMD may play a role in phenotype variability by minimizing the potential damage caused by premature termination codons \[[@r40],[@r41]\]. In achondrogenesis and hypochondrogenesis caused by *COL2A1* mutations, a relationship has been proved between the severity of the phenotype and the amount of type II collagen within the cartilage extracellular matrix \[[@r42]\]. This implies that not only qualitative but also quantitative factors likely modulate phenotypes linked to the *COL2A1* gene \[[@r43]\], as in haploinsufficiency due to NMD. Haploinsufficiency due to NMD was reported by Kaarniranta et al. \[[@r44]\], who found that heterozygous inactivation of *COL2A1* gene in the murine model led to structural defects and alterations that resulted from haploinsufficiency in ocular tissues containing type II collagen. These alterations included vitreous changes similar to those seen in patients with Stickler syndrome, which included reduced immunostaining of type II collagen in the vitreous and retina, in addition to reduced density of vitreous filaments in *COL2A1*+/− mutant mice \[[@r45]\]. Furthermore, in *COL2A1* exon 2 mutant mice with mutant allele encoding *COL2A1* mRNA without exon 2 \[[@r46]\], IIA+/− mutant embryos demonstrated, at an early stage, craniofacial abnormalities of truncated frontonasal structures and hypoplasia of the midface tissues. These malformations were more frequent in IIA−/− mutants. These findings are consistent with *COL2A1* type IIA mRNA expression in regions of active recruitment of cells for chondrogenesis and in areas of skeletal growth \[[@r41]\]. Alternatively, a second hypothesis is that nonsense-mediated altered splicing can be caused by disruption within the splicing *cis* element \[[@r21]\]. Minigene constructs created by McAlinden et al. demonstrated that disruptions in the enhancer sites in *COL2A1* exon 2 favor the production of the procollagen type IIB isoform. The decrease in the ratio of type IIA compared to type IIB leads to variance in expression levels, perhaps one isoform predominating over another, but the less expressed is not completely absent. These studies highlight the alternative imbalance between the two isoforms, which may have adverse effects during ocular embryogenesis \[[@r21]\]. Systemic manifestations associated with STL1, particularly facial development abnormalities and midline clefting as reported here (individual II:3; [Figure 1](#f1){ref-type="fig"}), have been observed in some cases of *COL2A1* exon 2 mutations, with a frequency depending on the series, 1%, 4%, and 43% in the Donoso \[[@r3]\], Parma \[[@r19]\], and Richards \[[@r22]\] series, respectively. These findings are not inconsistent with exclusive expression of the longer type IIA isoform in the adult vitreous \[[@r47]\], as embryonic expression of this isoform has been demonstrated in chondroprogenitor tissues \[[@r48]\]. The Cys86Stop mutation has previously been reported in four families with Stickler syndrome whose genealogy was traced to the 16^th^ century. Subsequently, Donoso et al. identified a member of the branch that migrated to the southeastern and mid-southern United States during the 19^th^ century \[[@r3],[@r19],[@r20]\]. Some members of these families included direct descendants from the passengers of the 1620 *Mayflower* voyage \[[@r20]\] who arrived on the northeastern coast and then migrated to the south. Interestingly, an affected individual (III:2) reviewed her genealogy and traced her ancestry to the state of Georgia. If descendants from Donoso's family indeed migrated south due to the Cherokee Land Grant of 1813, the similar geographical region of their cohort and ours would not exclude the possibility that the families could be related \[[@r20]\]. The apparent founder effect coupled with literature estimations of 50,000 to 100,000 descendants that could be related to the original family with this reported mutation demonstrates the importance of the genotype-phenotype relationship in patients with Stickler syndrome with this particular mutation \[[@r3]\]. The overarching similarities in phenotypes among vitreoretinal diseases make accurate diagnosis difficult clinically. Clinicians must understand and be updated on all allied conditions associated with Stickler syndrome to properly diagnose patients \[[@r47]\]. The characteristics of vitreous and retinal degeneration may guide molecular testing, but in case of doubt, a retinal specialist should be referred \[[@r47]\]. Although concentrations of exon 2 mutations are for predominantly ocular-only phenotypes, family members with mutations can still have systemic manifestations, seen in individual II:3, who presented with a cleft palate at a young age. The broad phenotypic variation seen in families with Stickler syndrome underscores the importance of using clinical and genetic testing to properly diagnose and treat patients with Stickler and Wagner syndromes. The authors would like to thank all family members for participation in the study. This research effort was supported by the National Institutes of Health Grant EY014685, The Lew Wasserman Award from Research To Prevent Blindness Inc., and the Duke-National University of Singapore core grant. (TLY) This research was also supported by the Toulouse Hospital Young Researcher Fellowship, the Fondation pour la Recherche Médicale and Fondation de France (VS). Supplemental section contains 4 tables and 1 figure. Table S1 contains all variants identified in COL2A1. Table S2 and S3 contains primers used for COL2A1 and VCAN gene screening. Table S4 contains primers used for cDNA amplification. Figure S1 contains PCR products of COL2A1 cDNA tissue expression. To access the data, click or select the words "[Appendix 1](http://www.molvis.org/molvis/v19/appendices/mv-v19-759-app1.pdf)." [^1]: The first two authors contributed equally to this work.

The Editor, The modern era of heart surgery began when the cardiopulmonary bypass (CPB) technique was introduced in the early 1950s. CPB is essential for the majority of cardiac surgery, but an undesirable inflammatory reaction occurs because of its use. There is a frequent misconception that methylene blue (MB) is a vasopressor drug. The MB effect appears in conditions of nitric oxide (NO) upregulation; it is not a vasoconstrictor and it is receptors-independent. MB blocks the cyclic guanosine monophosphate (cGMP) pathway through a "cross-talk" mechanism, thereby facilitating the cyclic adenosine monophosphate-dependent epinephrine vasoconstrictor effect. In an excellent systematic review of pharmacologic agents for acute hemodynamic instability, Morozowich and Ramakrishna included MB as a "pure vasoconstrictor" and mentioned "an passant" that the MB use is justified as "rescue therapy."\[[@ref1]\] Targeting MB for vasoplegic syndrome (VS) in a personal statement including questions, answers, doubts, and certainties, the following conclusions were drawn:\[[@ref2][@ref3]\] (1) The recommended doses are safe (the lethal dose is 40 mg/kg); (2) MB did not cause endothelial dysfunction; (3) the MB effect appears in cases of NO upregulation; (4) MB is not a vasoconstrictor; (5) it is possible that the MB acts through this "crosstalk" mechanism; (6) the most used dosage is 2 mg/kg as intravenous bolus followed by the same continuous infusion because the plasma concentrations strongly decay in the first 40 min; (7) there are no definitive multicentric studies, MB, at present, is the best, maybe it is the safest and cheapest option, but (8) there is a possible "window of opportunity" for the MB\'s effectiveness.\[[@ref2][@ref3]\] Many times, hemodynamic and metabolic MB effects were not observed because there is a potential "window of opportunity" for MB\'s effectiveness.\[[@ref4]\] In the first 8 h, there was an increased nitric oxide synthase (NOS) activity and guanylyl cyclase (GC) upregulation. In the next 8 h, there was a lack of GC expression and downregulation of NOS. In the third 8-h window, there was an upregulation of GC and NOS. Other authors and we have been emphasizing, over and over, two practical aspects: (1) The disclosure in using MB treatment without considering the window of opportunity and (2) the need for the establishment of this window in humans, perhaps choosing cGMP as a biomarker because the attempt to use nitrite/nitrate, measured by chemiluminescence was frustrating. In summary, there are two opposing concepts: (1) The use of MB as a rescue therapy to treat VS and (2) the use of MB as an early adjuvant drug (window 1). There is a possibility that MB does not act (second window) or acts too late (third window) when the circulatory shock is metabolically irreversible, presenting high lactate levels and uncontrollable metabolic acidosis. Maybe, it would be more sensible to consider MB, not a late rescue treatment, but as an adjuvant drug to be used early (window 1). In the absence of a protocol for the MB use as an adjuvant therapy, maybe the persistent vasoplegia and the increased needs of vasoconstrictors would be indicative signals for the MB use. By sharing again the discussed concepts, we hope that they are complimentary to the Morozowich and Ramakrishna\'s review. Financial support and sponsorship {#sec2-1} ================================= Nil. Conflicts of interest {#sec2-2} ===================== There are no conflicts of interest.

INTRODUCTION {#sec1-1} ============ Inaugural attempts of bone grafting were made by Lexer (1908) and Dratcher (1914) in growing cleft patients.\[[@ref1]\] Since then, opinions continue to differ on the indications and timing of maxillary bone grafting.\[[@ref2]\] Primary alveolar bone grafting was first suggested by Schrudde and Stellmach. In the 1960\'s, numerous reports about the subject appeared where long-term evaluation of treatment results had been made, showing serious growth disturbances in the maxilla as a sequelae to primary bone grafting. This has resulted in a search for new methods.\[[@ref3]\] Secondary bone grafting of maxilla and residual alveolar clefts at the stage of transitional dentition was first introduced by Boyne and Sands (1972). The treatment results after secondary alveolar bone grafting have been reported by several authors.\[[@ref4]--[@ref10]\] In 1981, Frank E Abyholm and Bergland *et al*.\[[@ref6]\] developed and established a semi-quantitative evaluation, to determine the success of grafting into the cleft maxilla. This semi-quantitative evaluation along with evaluation of other clinical variables such as eruption of canine through the grafted site, periodontal status of teeth adjacent to the graft, and root resorption were later adopted and adapted by many investigators to determine the success of alveolar bone grafting.\[[@ref5][@ref10]\] The aim of this prospective case control study was to clinically and radiologically evaluate the success rate of anterior iliac crest graft in primary alveolar cleft. MATERIALS AND METHODS {#sec1-2} ===================== A series of 10 patients who underwent secondary and late secondary alveolar bone grafting between 2008 and 2010 were included in this study. The inclusion criteria for the study were patients with complete unilateral or bilateral cleft lip and palate and patients with no other craniofacial abnormalities. Eight patients had complete unilateral cleft lip, palate, and alveolus (right side 3, left side 5). Two patients had complete bilateral cleft lip, palate, and alveolus \[[Table 1](#T1){ref-type="table"}\]. All the patients were females, with an average age of 12.6 years (range 9--16 years) at the time of surgery. Iliac crest (right side) was used as a graft material in all cases. Five patients had undergone orthodontic treatment with fixed orthodontic appliance (five unilateral). The mean evaluation period was 10.5 months. A detailed case history and preoperative radiographs were taken for all the patients. Intraoral periapical radiographs were taken with the film placed against the palate with its long axis parallel to the long axis of the canine. The central ray was directed through the canine eminence, and the point of entry was around the intersection of the distal and inferior border of the ala of the nose \[[Figure 1](#F1){ref-type="fig"}\]. The upper occlusal radiographs were taken, with the patient seated upright with the sagittal plane perpendicular to the floor and the occlusal plane horizontal. The film was placed cross-wise into the mouth. The film was gently pushed until it contacted the anterior border of the rami. Film stabilization was achieved by the patient gently biting the film. The central ray was directed at a vertical angulation of +65 degrees and a horizontal angulation of 0 degrees toward the middle of the film. Lateral trapdoor technique \[[Figure 2](#F2){ref-type="fig"}\] was used for harvesting the bone graft from iliac crest and then the graft was placed at the cleft site \[Figures [3](#F3){ref-type="fig"}--[6](#F6){ref-type="fig"}\]. Evaluation of treatment results included assessment of inter alveolar septum height on the teeth adjacent to the cleft by means of intraoral IOPA films \[[Figure 7](#F7){ref-type="fig"}\] and 4 scores were given based on the type of interdental septum (4) \[Figure 8\] (type I---septum height normal, type II---septum height less than normal but more than 3/4^th^ of the normal height, type III---septum height less than 3/4^th^ of the normal height, and type IV---when no bone was present across the gap), the tooth with the lowest marginal bone level was given a score of 4. In patients with bilateral clefts, separate evaluation was done for each side, eruption and migration of teeth into the grafted area, periodontal status of the teeth adjacent to the cleft was assessed by measuring gingival recession from the cementoenamel junction and periodontal pockets from the gingival margin by probing, and presence/absence of external root resorption on the teeth adjacent to the cleft by means of intraoral periapical radiographs. ###### Diagnosis of type of cleft, orthodontic treatment, age and gender of patient included in the study  {#F1} {#F2} {#F3} {#F4} {#F5} {#F6} {#F7} RESULTS {#sec1-3} ======= Postoperative radiographic evaluation revealed type I inter alveolar septum in 7 cases (87.5%), with complete unilateral cleft lip, palate, and alveolus. One case (case IV) showed type IV interalveolar septum after a period of 8 months. One patient (50%) in complete bilateral cleft lip, palate, and alveolus showed type IV (case VI) interalveolar septum after a period of 4 months \[[Table 2](#T2){ref-type="table"}\]. Noneruption of canine occurred in five patients (50%). In four cases (40%), canine had erupted completely. In one case canine had erupted 1/3^rd^ \[[Table 3](#T3){ref-type="table"}\]. Periodontal examination revealed, presence of pocket formation (less than 4 mm) and grade II mobility in two cases (20%) (case IV and case VI). None of the patients showed any external root resorption. ###### Sequential radiographic evaluation of patients in the study group  ###### Clinical evaluation of the study group pre-operatively and at the end of post-operative period  DISCUSSION {#sec1-4} ========== Secondary alveolar bone grafting is a procedure where results can be predicted with reasonable accuracy.\[[@ref7]\] Osseous healing of the transplant evaluated on intraoral periapical radiographs may be regarded as terminated within 6 months postoperatively in most cases.\[[@ref9][@ref11]\] Evaluation of interalveolar septal bone height on an intraoral periapical radiograph is a well-accepted method for determining results of secondary bone grafting.\[[@ref1][@ref2][@ref7][@ref12][@ref13]\] Abyholm *et al*. described interalveolar septal bone height as four types.\[[@ref6]\] Presence of type I or type II inter alveolar septal height after a period of 6 months was considered to be a success.\[[@ref2]\] In our series, 87.5% in unilateral cases and 50% in the bilateral cases showed type I interalveolar septal height, after a minimum observation period of 6 months. Abyholm *et al*.\[[@ref6]\] in their study had a success rate of 85.3% in unilateral cases. The success rate in bilateral cases was 76.3%. Johanson *et al*. reported a success rate of 90%.\[[@ref2]\] Enemark *et al*. had 98.65% of type II interalveolar septal height after a period of 4 years.\[[@ref9]\] Amanat *et al*.,\[[@ref7]\] had success rate of 82%. To achieve an adequate interalveolar septal height, various technical details are important. Of utmost importance is the separation of the nasal cavity from the grafted site by careful suturing of nasal mucoperiosteum. A mucogingival flap on the buccal aspect is the most widely recommended and practiced flap design as opposed to mucobuccal and mucolabial flap. The third important technical prerequisite being complete filling of the alveolar cleft with purely cancellous bone.\[[@ref6][@ref14]\] Bergland *et al*. recommended immobilization of the premaxilla in bilateral cleft cases for a period of 3 months postoperatively to allow consolidation of the graft free from shearing forces, which may be produced by otherwise mobile premaxilla.\[[@ref6]\] Another factor that appears to have a bearing on healing of the graft is the amount and extent of previous surgery, which leads to the formation of scar tissue which compromises the blood supply to the area.\[[@ref7]\] The two failures in our cases (case IV and case VI) could be due to loss of the graft from the secretions of the nasal cavity in the unilateral case and inadequate soft tissue coverage in the bilateral case, as it was a wide cleft. Eruption/retention of canine {#sec2-1} ---------------------------- The complication occurs to a variable extent in most reported series.\[[@ref15][@ref16]\] Retention of canine occurred in 50% of cases in the present study. In contrast, 10% - 30% of the cases considered showed retention. Enemark *et al*., reported an incidence of canine retention in 21% of cases.\[[@ref14]\] 10% of patients in a study conducted by Paulin *et al*., showed retention.\[[@ref17]\] Enemark *et al*., in a separate study showed canine retention in 30% of his cases.\[[@ref9]\] External root resorption {#sec2-2} ------------------------ External root resorption, although rare, has been reported in the literature.\[[@ref8][@ref14]\] Enemark *et al*. reported an incidence of 3.3%.\[[@ref14]\] Bergland *et al*. in two separate studies reported cervical root resorption in 5% and 2% of their cases.\[[@ref6][@ref8]\] Abyholm *et al*. had no incidence of external root resorption.\[[@ref4]\] Amanat *et al*., reported an incidence of 3.3%.\[[@ref7]\] In the present study no evidence of external resorption was noted radiographically. Periodontal status of teeth adjacent to the graft {#sec2-3} ------------------------------------------------- Jia *et al*., in their postoperative study on periodontal status, found no pathological pockets.\[[@ref18]\] Sindet-Petersen *et al*., in their study demonstrated an incidence of periodontal pocketing in 0.3% of cases.\[[@ref11]\] The results achieved in the present study are comparable with that of emerging from other centers. Evaluation of success outcome for secondary alveolar bone grafting in terms of radiographic interpretation of the inter-alveolar septum, the eruption/retention of canine, presence/absence of external root resorption, and periodontal status of the teeth adjacent to the graft is a well-accepted method. However, the really critical factors are timing of the procedure, proper closure of the grafted site with gingival and palatal mucoperiosteal flaps, and complete filling of the alveolar cleft with cancellous bone. CONCLUSION {#sec1-5} ========== Secondary alveolar bone grafting done during the time of transitional dentition, before the eruption of permanent canine, is an excellent treatment modality, whose long-term success rate can be predicted after a short period of evaluation. **Source of Support:** Nil **Conflict of Interest:** None declared.

{#F0001} {#F0002} {#F0003} {#F0004} {#F0005} Protein aggregation {#S0001} =================== Protein misfolding and aggregation is associated with an increasing number of highly debilitating human diseases, including Alzheimer\'s disease (AD), diabetes and some types of cancer \[[@B1]\]. The proteins involved in these disorders are different and not related in evolutive, sequential or structural terms \[[@B2]\]. However, in all the cases, misfolded conformers of these polypeptides establish non-native intermolecular contacts that result in their deposition into insoluble amyloid aggregates in the intra- or extra-cellular space \[[@B3]\]. Amyloid fibrils are highly ordered and repetitive structures where all polypeptides adopt a common fold \[[@B4],[@B5]\]. They are thread-like protein aggregates with a core region formed by repetitive arrays of β-sheets oriented perpendicularly to the fibril axis forming a structure known as cross-β because it displays two sets of characteristic X-ray diffraction signals, forming a 'cross' pattern \[[@B6]\]. These particular fibrils have diameters of approximately 10 nm and often consist of multiple proto-filaments twisted around the fibril axis. Importantly, the ability to form amyloid-like structures is not restricted to a subset of disease-linked proteins and this conformation may be accessed by most, if not all, human proteins, irrespective of their native fold \[[@B7]\]. In fact, the molecular interactions leading to the formation of amyloids are similar to those promoting the folding and functional assembly of proteins. As a result, folding and aggregation pathways are continuously competing in the cell \[[@B8],[@B9]\]. This implies that the proteins involved in conformational disorders might me much larger than previously thought \[[@B10]\], neuronal receptors among them. Prediction of protein aggregation {#S0002} ================================= Although any polypeptide has the potential to self-assemble into β-sheet-enriched amyloid-like conformations \[[@B7]\], the composition and the primary structure of a protein strongly influences its propensity to aggregate \[[@B11]\]. Not all the regions in a protein contribute equally to its aggregation behavior and there exist short specific amino acid stretches with defined physicochemical properties able to initiate the self-assembly process by nucleating the aggregation reaction \[[@B12]\]. Small sequential changes inside or close to these sequences usually have a large impact on protein solubility \[[@B13],[@B14]\]. Moreover, short peptides corresponding to these regions are able to form amyloids, in the absence of the rest of the protein sequence \[[@B15],[@B16]\]. The main intrinsic features that modulate the protein-aggregation propensity of proteins have been identified. They have been exploited by different prediction algorithms to detect β-aggregating motifs in protein sequences and to infer their experimental aggregation rates \[[@B17]\]. We will use several of these algorithms to forecast the aggregation propensities of different neuronal receptors. Aβ peptide aggregation in AD {#S0003} ============================ AD is a complex neurodegenerative multifactorial pathology characterized by a progressive and irreversible decline of cognitive function. It is the most common form of dementia in individuals over 65 years of age; and the incidence of AD doubles every 5 years beyond the age of 65. According to the World Alzheimer Report 2011, it is estimated that 36 million people worldwide suffer from dementia, and patients are expected to double every 20 years to 66 million by 2030, and 115 million by 2050 \[[@B18]\]. A majority of AD is sporadic, although several genetic linkages have also been identified. A hallmark of AD is the extracellular accumulation of aggregated amyloid-β (Aβ). Aβ is a 37--43 residues peptide that results from multiple proteolytic cleavage of a large transmembrane precursor, the amyloid precursor protein \[[@B19]\]. AD pathological features result from a chronic imbalance between Aβ production and clearance. The self-assembly of initially soluble but sticky Aβ molecules into oligomers and large amyloid fibrils initiate the pathogenic cascade. The process of Aβ amyloid fibril formation is a multistep process that follows a nucleation-elongation mechanism in which the formation of the first oligomeric assemblies is the rate limiting step of the reaction and is followed by a rapid fibril elongation phase to form protofibrils and fibrils \[[@B20]\]. Despite the toxic effect was initially associated to the presence of mature fibrils, it is now clear that the smaller intermediate species are also toxic for neurons. In fact it is becoming evident that AD is initiated by alterations in hippocampal synaptic function caused by diffusible oligomeric assemblies of Aβ. These early extracellular aggregates exhibit hydrophobic aggregation-prone regions to solvent that can interact anomalously with different cellular components, and specifically with cellular membranes promoting oxidative stress and Ca^2+^ homeostasis deregulation \[[@B21]\]. It has been proposed recently that apart from interacting with the lipidic membrane components, aggregated Aβ can interact with proteins anchored in such membranes, including neuronal receptors \[[@B22]\]. Such interactions might result in signaling dysfunction or even in the sequestration of the receptor into the amyloid aggregate with deleterious effects for neurons. In fact, since Aβ and receptors share the endocytic pathways it is also possible that wrong interactions would occur during traffic. Prediction of aggregation prone regions in neuronal receptors {#S0004} ============================================================= To address if human neuronal receptors display aggregation-prone regions susceptible of interacting with Aβ oligomeric assemblies, we have selected five different receptors for which the 3D structure were available. We have used human structures in most cases, when these were not available we used homolog proteins with a high sequential identity. In all cases we have analyzed the extracellular domains of the receptors, which include the ligand-binding sites, because it is more likely that Aβ diffusible oligomeric species will interact with them than with domains already embedded in the membrane. We have used PDBePISA (Protein Interfaces, Surfaces and Assemblies) \[[@B23]\] to deconstruct oligomeric receptors into their monomers in such a way that we can identify which residues are located at the interface and therefore might become exposed to solvent when the domain dissociates. To have a complete description of the aggregation properties of the selected neuronal receptors we have used four different predictors based on different concepts and calculation schemes: Waltz \[[@B24]\] uses a position-specific scoring matrix deduced from the biophysical and structural analysis of the amyloid properties of a large set of hexapeptides to distinguish between amyloid and amorphous aggregating sequences \[[@B25]\].Tango \[[@B26]\] is based on the analysis of the physico-chemical principles underlying β-sheet formation together with the assumption that the core regions of an aggregated polypeptide are protected from the solvent \[[@B27]\].Aggrescan \[[@B28]\] is based on an aggregation propensity scale for natural amino acids derived from *in vivo* experiments in bacteria and on the assumption that short and specific sequence stretches modulate protein aggregation \[[@B29],[@B30]\].Amylpred \[[@B31]\] is a consensus predictor of amyloid propensity that integrates the results of five different methods in a single output to detect the regions in a protein with the highest intrinsic propensity to form amyloids \[[@B32]\]. As a stringent criterion for identification of aggregation-prone regions we have considered as positive hits only those regions recognized simultaneously by at least three of these algorithms. Aggregation-prone regions in GRM5 {#S0005} ================================= Human GRM5 (P41594) is a G-protein coupled receptor for glutamate. The binding of the ligand to the receptor promotes a conformational change that initiates the signaling pathway through guanine nucleotide-binding proteins, which further modulate down-stream effectors. Signaling involves a phosphatidylinositol-calcium second messenger system, which triggers a calcium-activated chloride current. GMR5 has been shown to be a co-receptor for Alzheimer Aβ oligomers bound to cellular prion protein \[[@B33]\]. GRM5 is a receptor of 1212 residues located at the cell membrane, residues 22 to 579 correspond to the extracellular region, exhibiting an α β 3-layer (aba) sandwich structure, with 32% of helical structures (16 helices; 159 residues) and 18% of β-sheets (22 strands; 91 residues) (PDB 3LMK). Four aggregation regions comprising residues 35--45 (IIIGALFSV), 161--166 (LLQLFN), 183--198 (TLFKYFM) and 398--410 (GFVINAIYSMAYG) are detected with the consensus of three predictors ([Figure 1](#F0001){ref-type="fig"}). The second and fourth sequence stretches are located at two helices (α-helices 3 and 14), the first one in β-strand 2 and the third one in a turn. The second and third aggregation-prone regions overlap with the interface between domains in the receptor, with 83 and 42% of their residues being involved in interdomain contacts. The first and fourth stretches are totally hidden in the inner core in the native structure of the receptor. The four predictors coincide that the sequences 35-IIIGALFS-42 and 399-FVINAIYSMA-408 are amyloidogenic. Aggregation-prone regions in human glutamate receptor 2 {#S0006} ======================================================= Human glutamate receptor 2 (AMPA 2) (P42262) plays a central role in excitatory synaptic transmission at the CNS. AMPA 2 is a receptor for glutamate that acts as ligand-gated ion channel. Binding of the excitatory neurotransmitter [l]{.smallcaps}-glutamate to the receptor promotes a structural change that forces the opening of the cation channel. This action transforms the chemical input into an electrical signal, upon which the receptor desensitizes and adopts a transient inactive conformation, in which it remains bound to an agonist. This receptor is activated by the presence of Aβ oligomers at the synapsis \[[@B34]\]. AMPA 2 is a 883 residues receptor which sequence is further processed into a mature form (25--883). The crystallized extracellular region corresponds to residues 25 to 395, showing an α β 3-layer (aba) sandwich structure, with 34% helical structure (13 helices; 135 residues) and 21% of β sheet (18 strands; 82 residues) (PDB 2WJW) \[[@B35]\]. Three hot spots of aggregation comprising residues 85--92 (VYAIFGFY), 108--113 (HVSFIT) and 136--144 (LLSLIEYYQ) are detected in the crystallized receptor ([Figure 2](#F0002){ref-type="fig"}). The first and the second sequence stretches are located at two parallel strands (β-strand 3 and 4), the third one at the α-helix 4. This last region is exposed and overlaps significantly with the interface between extracellular domains in their dimer structure ([Figure 2](#F0002){ref-type="fig"}). The sequence 85-VYAIG-90 is an amyloidogenic region according to the four predictors, but is essentially buried inside the native structure. Other three aggregation-prone regions comprising residues 415--420 (TVVVTT), 500--503 (LTIT), 534--539 (GVFSFL) are predicted outside the crystallized domain. In agreement with data on GRM5, the glutamate binding sites do not correspond to any of the above-described aggregation-prone regions. Aggregation-prone regions in acetylcholine receptor {#S0007} =================================================== After binding to acetylcholine the mouse acetylcholine receptor (AChR) responds with an extensive change in conformation that affects all their subunits and leads to opening of an ion-conducting channel across the cellular membrane. It exhibits a pentameric structure with two α chains (P04756), and three other structurally different subunits. This receptor interacts directly with Aβ oligomers \[[@B36]\]. The α chain of AChR is 457 residues long. After signal peptide processing the AChR sequence mature form comprises residues 21--457, with residues 21--230 located at the extracellular domain. Extracellular mature α AChR shows 8% helical structure (three helices; 18 residues) and 47% of β sheets (14 strands; 100 residues) and is a mainly β distorted sandwich (PDB: 2QC1). Mouse α AChR shares 83.6% identity with the human AChR α receptor chain \[[@B37]\]. Four regions are predicted as aggregation prone with the consensus of three predictors, comprising residues 49--62 (VQVTVGLQLIQLIN), 147--155 (YCEIIVTHF) ([Figure 3](#F0003){ref-type="fig"}). In this mainly β protein the first, second and fourth sequences are located at β-sheets (β-sheets 2, 8 and 12) and the third corresponds to a β turn. The solvent exposed degree of the first and second aggregation-prone regions is 28.57 and 44.44%, and interestingly, regions 120--126 (FAIVKFT) and region 206--212 (HWVFYSC) are fully exposed to solvent. It is likely that in native conditions these dangerous regions are covered by interactions with other domains of the receptor. The four predictors coincide that 54-GLQLIQLI-61 is amyloidogenic and exposed to solvent. Aggregation-prone regions in human insulin receptor {#S0008} =================================================== Human insulin receptor subunit α (P06213) is one of the subunits of the insulin receptor (IR). This transmembrane receptor is activated by insulin and belongs to the tyrosine receptors superfamily. Binding of insulin results in the phosphorylation of several intracellular substrates, which act as docking proteins for other signalling polypeptides. From a metabolic point of view, the IR is responsible for the regulation of glucose homeostasis and its involved in a wide range of pathologies, including diabetes and cancer. Aβ oligomers have been shown to reduce responsiveness to insulin in synapsis \[[@B38]\]. Human IR is a 1382 amino acid residues receptor which sequence is further processed into IR subunit α (28--758). The crystallized subunit α comprises residues 25--457, which show 11% helical structure (9 helices; 56 residues) and 29% of β-sheets (43 strands; 141 residues) (PDB 2HR7). Two different structural domains are evident in the crystallized chain: an α β horseshoe (4--191, 310--468) and a mainly β ribbon domain (192--309). Several hot spots of aggregation are detected with the consensus of three predictors, 54--69 (LIMITDYLLLFRVYGL), 87--99 (LFFNYALVIFEMV), 338--344 (GSLIINI), 502--508 (GFMLFYK), 561--571 (QYAIFVKTLVT), 628--632 (GVFSFL) ([Figure 4](#F0004){ref-type="fig"}). The three first sequence stretches correspond to parallel β-strand structures, with 43.75, 69.23 and 28.57% of their residues exposed to solvent, respectively. The other three aggregation-prone regions are located in structurally undefined protein segments. Interestingly enough, in this receptor the binding site for insulin overlaps with the first and second aggregation sequence stretches. The sequence 87-LFFNYALVIF-96 in the second stretch is predicted as highly amyloidogenic. Aggregation-prone regions in IGF1R {#S0009} ================================== Human IGF1R (P08069) is, as IR, a member of the tyrosine kinase receptors. The activated form of this receptor is involved in cell growth and survival control. In addition, it is an important determinant of tumor transformation and the survival of malignant cells in different cancers. Binding of the ligand results in activation of the receptor kinase and receptor autophosphorylation, as well as the phosphorylation of tyrosine residues in different protein substrates, that function as signaling adapter proteins. Aβ oligomers have been shown to mediate IGF-1R activation in the hippocampus of AD brain during the early stages of disease development \[[@B39]\]. IGR1R complete sequence has 1367 residues, which further processing renders the α subunit located at the extracellular environment comprising residues 31--736. The crystallized IGR1R α subunit exhibits three structural domains, (1--183) α/β horseshoe, (224--299), mainly β ribbon domain and, (300--478) another α/β horseshoe domain (PDB 1IGR). Seven aggregation prone regions are detected with the consensus of three predictors, comprising residues 27--35 (GYLHILLIS), 48--60 (LTVITEYLLLFRV), 81--92 (LFYNYALVIFEM), 416--420 (MYFAF), 490--498 (LISFTVYY), 546--551 (QYAVYV) and 695--701 (FLHNSIF) ([Figure 5](#F0005){ref-type="fig"}). The crystal structure includes residues 1--478, being the first, second, third and fourth sequence stretches located at β-strands 3, 5 and 8 in the first β-sheet and at β-strand 8 in the ninth β-sheet, respectively. The main property that differentiates IGF1R and other members of the IR family from other receptor families is their presence at the cell surface as disulfide-linked dimers and require domain conformational changes rather than oligomerization for signaling. The first four aggregation-prone regions show a 55.55, 23.08, 66.67 and 20% of their residues exposed to solvent and might be therefore involved in inter-domain contacts in the dimeric structure or in functional binding. In fact, residues Y28, H30 and L33 in the first region and L696, H697, N698, Y100 and Y101 in the last aggregation stretch have been shown to be crucial for ligand binding. Sequences 81-LFYNYALVIF-90 and, 493-FTVYY-497 are predicted as amyloidogenic with the consensus of the four predictors. Conservation of aggregation prone-regions in neuronal receptors among different species {#S0010} ======================================================================================= To decipher whether the detected aggregation-prone sequences stretches might play a conserved role in different organisms we made multiple alignments of the sequences of the five analyzed receptors in vertebrates. As an example we show the results in [Table 1](#T1){ref-type="table"} for the α chain of the acetylcholine receptor. As a general rule, independently, of the structural location of the aggregating stretches, they are highly conserved among species, and most of the substitutions have a conservative character, with isosteric and isopolar changes in most cases. Provided that we know now that there is a strong selective pressure against the maintenance of aggregation-prone regions in protein sequences along evolution, especially when they become exposed to solvent in the folded structure, the strict conservation of the detected stretches along vertebrate sequences clearly indicate that they play an important structural and/or functional role in the respective receptors. Conclusion & future perspective {#S0011} =============================== Despite the work presented here is based on *in silico* predictions, which should be further validated experimentally, our data suggest that, as previously demonstrated for other globular proteins, the extracellular domains of neuronal receptors cannot skip the presence of aggregation-prone regions in their sequences. However, the number and size of these stretches is rather small if we compare them with the average in globular proteins. This is likely a consequence of their important physiological role, since it has been shown that essential proteins are under higher selective pressure against the accumulation of aggregating sequences that nonessential ones. The detected regions are highly conserved, pointing out to an important role in the function or the structure of these receptors. Importantly, and contrary to what one could expect, many of these regions appear to be exposed to solvent in the monomeric forms. A structural inspection indicates that they are part of the interface between subunits or even overlap with the binding site for proteinic substrates. These data confirm that protein--protein interaction surfaces and regions with high aggregation propensity overlap significantly in proteins. The result is that formation of native receptor oligomers or native complexes with substrates and self-aggregation reactions or anomalous interactions with other aggregating ensembles such as Aβ oligomers are probably constantly competing in the extracellular face of neuronal cells, which might result in signaling problems. Moreover, neurons are vulnerable to perturbations in the endocytosis process in which they recycle their receptors. It is likely that Aβ oligomers might interact promiscuously with the aggregation-prone regions in conformational flexible, monomeric and unliganded receptors during this process, impeding them to arrive to their proper destinations in the neuronal membrane. ###### **Sequence alignment of detected aggregation-prone regions in the α chain of the acetylcholine receptor in different species.** ***Mus musculus*** **VQVTVGLQLIQLIN** **FAIVKFT** **YCEIIVTHF** **HWVFYSC** -------------------------- -------------------- ------------- --------------- --------------------- *Bos Taurus* V**E**VTVGLQLIQLIN FAIVKFT YCEIIVTHF HWVFY**A**C *Canis lupus familiaris* VQVTVGLQLIQLIN FAIVKFT YCEIIVTHF HWVFY**A**C *Cricetulus griseus* VQVTVGLQLIQLIN FAIVKFT YCEIIVTHF HWVFYSC *Heterocephalus glaber* VQVTVGLQLIQLIN FAIVKFT YCEIIVTHF HWVFYSC *Homo sapiens* VQVTVGLQLIQLIN FAIVKFT YCEIIVTHF H**S**V**T**YSC *Pan troglodytes* VQVTVGLQLIQLIN FAIVKFT YCEIIVTHF H**S**V**T**YSC *Sus scrofa* VQVTVGLQLIQLIN FAIVKFT YCEIIVTHF H**R**V**L**Y**A**C *Torpedo marmorata* V**DI**TVGLQLIQLIN FAIV**HM**T YCEIIVTHF HWVYY**T**C Residues that diverge from the mouse sequence, for which the crystal structure has been solved, are bold. ###### Executive summary 1. The formation of amyloid-like structure is an intrinsic property of most proteins and polypeptides. 2. The presence of aggregation-prone regions can be read out directly from the primary protein sequence. 3. Neuronal receptors exhibit aggregation-prone regions exposed to the solvent that might interact with hydrophobic diffusible amyloid-β peptide oligomers. 4. Aggregation-prone regions in neuronal receptors often overlap with the interface between monomers or domains or coincide with the ligand-binding sites, illustrating a competition between functional and anomalous interactions. **Financial & competing interests disclosure** The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties. No writing assistance was utilized in the production of this manuscript. **Open access** This work is licensed under a Creative Commons Attribution 4.0 License. To view a copy of this license, visit <http://creativecommons.org/licenses/by/4.0/>

In this third volume of *RBMS*, I am thrilled to introduce a new section of the journal called *What's Out There* which will present reviews of books, films, plays, art exhibitions, conferences and other events of interest to our readers. Reproductive biomedicine is definitely 'out there' in the popular imagination -- whether in BBC documentaries about non-invasive prenatal testing and CRISPR-Cas9, newspaper articles about treating mitochondrial disease or growing embryos beyond fourteen days, adverts for fertility tracker apps and egg freezing, or soap opera storylines about surrogacy. *What's Out There* will give a sense of what kind of representations of reproductive biomedicine are currently on the cultural landscape. It reflects the editors' ambition that *RBMS* be a truly interdisciplinary journal and so we will be taking a broad approach in commissioning reviews for this section, encompassing literature, culture, public events and the arts. I am personally delighted to take up the position of editor of *What's Out There*. I am a Senior Research Associate at the University of Cambridge where I am currently working on a study of media representations of IVF in the UK, focusing particularly on the birth of Louise Brown in 1978. As a social anthropologist working in a sociology department on a project that also draws on cultural studies, media studies and history, I strongly support this journal's interdisciplinary approach to reproduction, so it is an honour to become part of the editorial team. I believe it is important to get people talking across the disciplines in order to augment our knowledge and have our assumptions questioned, but also because reproduction is an inherently interdisciplinary topic and we can only hope to understand it if we take an approach that eschews traditional boundaries and conventional precepts. In this spirit of interdisciplinarity and boundary-crossing, who better to have as the author of our first piece in *What's Out There* than *RBMS* Co-Editor-in-Chief, Sarah Franklin ([@bb0005])? Sarah has written a fascinating review of Louise Brown's memoir of her life as the world's first 'test-tube baby', which gives an insight into her very particular experience of being famous for how she was conceived -- or, to put it another way, of being miraculous yet ordinary. As Sarah reminds us, in the wake of the development of a global fertility industry, going back to the early history of IVF is an important corrective to certain assumptions that have become commonplace as this technology has become rapidly normalised. In the British media in 1978, IVF was at least as much about creating families and making parents as it was about scientific discoveries or technological innovations and the experience of the Brown family, which was central to the media narrative of the day, is a crucial reminder of this. 1978 was also a very interesting year in British history as the country teetered on the brink of the \'Winter of Discontent\' and the subsequent election of Margaret Thatcher as Prime Minister. This was a time in which the post-war socialist and liberal consensus of the 1950s, '60s and early '70s was being rejected in favour of neoliberalism and social conservatism. As Sarah writes, Louise Brown was born to parents from a very 'ordinary' background, who nonetheless epitomised many of the social, political and economic changes of the late '70s and early '80s. Her parents, Lesley and John Brown, had gone from deprived childhoods in 'broken homes', to homelessness and indigence in their young adulthood, to buying their council house (before Thatcher even came to power) and working overtime to pay for private medical treatment by their early thirties. Yet despite their social mobility, which was further boosted by the money they received from Associated Newspapers for selling the exclusive rights to their story, they wanted nothing more for themselves or their family than to be normal and ordinary. In a profile in the *Daily Mail* newspaper in August 1978, John Brown explained that Louise would not be spoiled or treated differently from any other child and expressed his only ambition for her: 'She's going to be an ordinary girl brought up in an ordinary house as an ordinary person'. Strangely, parenting is often overlooked in contemporary accounts of IVF. Like happily-ever-after romances that tell the story of how a couple got together and end at their first kiss, many of the stories we hear of IVF are about quests to conceive. IVF made Lesley and John Brown parents, certainly, but only up to a point. Once they had their longed-for children, the rest was up to them. With Louise Brown's memoir we now have an insight into just what sort of parents they were and of how the real legacy they left her was an ordinary, loving childhood. Steptoe and Edwards may have perfected the culture medium in which the embryo that became Louise spent its first days, but her parents created the stable environment in which she would grow up to be a warm, thoughtful and caring adult. As editor of *What's Out There*, I look forward to bringing many further contributions to fruition, including from authors who might not normally contribute to a journal like this. *What's Out There* is an important part of the on-going development of this young journal and I hope that this section will also attract new readers. This is one reason that I will be encouraging reviews of popular, as well as scholarly, work. Initially, we envisage having a single *What's Out There* contribution in each volume. Contributions will be by invitation, but I am also happy to receive recommendations from authors, producers, publishers and would-be review-writers. Contributions can be up to 3000 words long and illustrations are welcome. And, best of all, there will be no charge to the authors, as Elsevier has kindly agreed for the usual publication fee to be waived for one contribution in each volume. I look forward to hearing more from you about what's out there!

An important aspect of gene regulation is the proper control of mRNA levels, which is the product of both synthesis and decay. Although the earliest studies placed the greatest attention on how mRNA levels are controlled through transcription, interest in the role of mRNA decay is increasing. In yeast, the canonical mode of degradation starts with the shortening of the poly(A) tail by [Pan2](http://www.yeastgenome.org/locus/S000003062/overview)/[3p](https://www.yeastgenome.org/locus/S000001508) and the main cytoplasmic deadenylase, [Ccr4p](http://www.yeastgenome.org/locus/S000000019/overview), which is part of the [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not complex ([@bib11]; [@bib97]; [@bib14]; [@bib37]; [@bib100]). Deadenylation is followed by the removal of the 5′ cap by [Dcp1](http://www.yeastgenome.org/locus/S000005509/overview)/[2p](https://www.yeastgenome.org/locus/S000005062), which leads to exonucleolytic cleavage of the mRNA in the 5′ to 3′ direction by [Xrn1p](http://www.yeastgenome.org/locus/S000003141/overview) ([@bib30]). Alternatively, 3′ to 5′ decay is catalyzed by the exosome complex in a regulated manner ([@bib88]). Additional proteins bind specific regions on the mRNA, such as the 3′ UTR (untranslated region), to mediate decay and the localization of the mRNA ([@bib37]). Remarkably, even among the most well-studied factors like [Ccr4p](http://www.yeastgenome.org/locus/S000000019/overview), it remains unclear how and what regulates their recruitment to mRNAs across the transcriptome. [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not is composed of nine conserved core subunits that form a 0.9--1.2 MDa protein complex: [Not1](http://www.yeastgenome.org/locus/S000000689/overview) through [Not5](http://www.yeastgenome.org/locus/S000006276/overview), [Caf1](http://www.yeastgenome.org/locus/S000005335/overview)/[Pop2](http://www.yeastgenome.org/locus/S000005335/overview), [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), [Caf40](http://www.yeastgenome.org/locus/S000005232/overview), and [Caf130](http://www.yeastgenome.org/locus/S000003366/overview) ([@bib68]; [@bib19]). The [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not complex interacts with a number of other proteins, such as the RNA-binding proteins (RBPs) [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) and the RNA helicase [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) ([@bib44]; [@bib21]; [@bib64]; [@bib38]; [@bib93]; [@bib1]). [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) is a member of the Fem-binding family of proteins ([@bib80]) that binds to a motif in the 3′ UTR of mRNAs and enhances deadenylation by [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not ([@bib38], [@bib39]). Recent crystallography studies verified the interaction between [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) and the N-terminus of [Not1](http://www.yeastgenome.org/locus/S000000689/overview) ([@bib15]; [@bib66]). Although [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) binds to [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not, it is much more abundant, and additionally inhibits translation and promotes the decapping of mRNAs ([@bib36]; [@bib20]; [@bib91]; [@bib81]). Although these three proteins physically and genetically interact with each other, they play distinct functions in mRNA regulation ([@bib68]; [@bib79]); thus, it is unclear to what extent their mRNA targets overlap. One of the features of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not that makes it an intriguing complex to study is that it regulates both transcription and mRNA degradation ([@bib68]; [@bib19]). The deletion of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not subunits affects the abundance, decay, and synthesis of many mRNAs ([@bib40]; [@bib22]; [@bib71]; [@bib89], [@bib90]). However, the interpretation of gene expression changes in deletion mutants, while valuable, can be influenced by complex genetic interactions and secondary effects. In the aforementioned studies, it remains unknown which changes in mRNA expression are direct and which result from perturbed mRNA synthesis or decay. Therefore, identifying the direct mRNA targets of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not could shed light on the molecular underpinnings behind how RNA abundance, decay, and synthesis are controlled across the transcriptome. Here, a modified RIP-seq procedure was used to identify transcripts associated with [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview). In addition to showing that these factors are recruited to many of the same mRNAs, the analysis of the mRNA targets suggests that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) imparts its greatest influence on gene regulation at the level of decay *vs.* synthesis. Our study has illuminated the interplay between the two cytoplasmic deadenylases in mRNA decay. Additionally, the recruitment of these three factors negatively correlated with ribosome density, suggesting new insights into the relationship between mRNA decay and translation. Finally, we show that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) bind mRNAs that fluctuate during the yeast metabolic cycle (YMC), suggesting a role for post-transcriptional regulation in reshaping the transcriptome in response to changing environmental conditions. Materials and Methods {#s1} ===================== Strain construction {#s2} ------------------- All strains were constructed in the BY4741 background using published protocols ([@bib62]). A list of strains is contained in Supplemental Material, Table S5 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf). RIP {#s3} --- RIP-seq was based upon a previous version of the procedure ([@bib26]), but incorporating changes for high-throughput sequencing of mRNAs. Cells were grown in 1 L of YPAD (2% peptone, 1% yeast extract, 0.02 mg/ml adenine sulfate, and 2% dextrose) at 30° to an OD~600~ ∼0.9, and then formaldehyde was added to 1% (v/v) for 15 min. Cross-linking was quenched with glycine (final = 136 mM) for 5 min. Subsequent steps used buffers prepared with diethyl polycarbonate-treated (DEPC) water. Cells were harvested and washed in ice-cold STE (10 mM Tris-HCl pH 8, 1 mM EDTA, 50 mM NaCl, 0.5 mM PMSF, and 1 mM benzamidine-HCl) and frozen at −80°. Cells were resuspended in FA-lysis buffer (50 mM HEPES/KOH pH 7.5, 150 mM NaCl, 1% Triton X-100, and 0.1% sodium deoxycholate) containing protease inhibitors (2 μg/ml leupeptin, 3 μg/ml aprotinin, 2 μg/ml pepstatin A, 1 μg/ml chymostatin, 1 mM benzamidine-HCl, and 0.5 mM PMSF). Cells were disrupted by vortexing in the presence of glass beads for 45 min at 4°, and then 200 µl FA-lysis buffer was added to each tube and mixed for 30 sec, twice. Cell lysate was then transferred to two 15 ml polystyrene tubes (∼1.8 ml lysate in each), sonicated with two 30 sec pulses using a Bioruptor sonicator (Diagenode, Philadelphia, PA), and then clarified by centrifugation twice at 4° at 14,000 rpm. Protein content was measured by Bradford assay (Bio-Rad) using BSA as a standard. Samples with \> 7 mg/ml protein concentration were used in RIP. Whole-cell extract (2.5 ml) was diluted with an equal volume of FA-lysis buffer (containing protease inhibitors). MgCl~2~ and CaCl~2~ were added to concentrations of 25 and 5 mM, respectively, and then RNase-free DNase I (Worthington, Lakewood, NJ) was added to 174 units per ml. The samples were incubated for 90 min at 30°. EDTA was added to 50 mM and the samples were placed on ice. Samples were spun for 10 min at 14,000 rpm at 4°, and then 1 ml of supernatant was transferred to four tubes containing 15 µl of protein A--sepharose (GE Healthcare) slurry containing 3.5 µl 9E10 (anti-myc) monoclonal antibody from ascites fluid (Biolegend). In experiments where RT-qPCR of RNA was used for validation, Protein A beads bound with 7 µl of purified 9E10 antibody were used. The samples were incubated overnight at 4° with rotation. Beads were washed three times with FA-lysis buffer, two times with FA-wash buffer 2 (50 mM HEPES/KOH pH 7.5, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.5 M NaCl, 0.5 M PMSF, and 1 mM benzamidine-HCl), three times with FA-wash buffer 3 (0.25 M LiCl, 1% NP-40, 1% sodium deoxycholate, 1 mM EDTA, 10 mM Tris-HCl pH 8.0, 0.5 M PMSF, and 1 mM benzamidine-HCl), and two times with TE buffer (10 mM Tris-HCl and 1 mM EDTA, pH 8). RNA was eluted off the beads at 65° for 20 min in 450 µl elution buffer \[25 mM Tris-HCl (pH 7.5), 1 mM EDTA, 0.2 M NaCl, and 0.5% SDS\]. To reverse cross-links, proteinase K was added to 70 μg/ml and incubated at 65° for 5 hr. Input samples (100 µl of extract) were supplemented with 350 μl elution buffer, 10 μl 10% SDS, and then incubated at 65° for 5 hr. The RNA was purified using an acid--phenol (pH 4.8)/chloroform (1:1) extraction followed by ethanol precipitation in the presence of 20 μg glycogen. Samples were resuspended in DEPC water, and then treated with DNase once again. RNA was extracted with acid--phenol (pH 4.8)/chloroform/isoamyl-alcohol (25:24:1) and ethanol-precipitated in the presence of 10 μg glycogen. The pellet was resuspended in DEPC water and the nucleic acids quantified using a Nanodrop instrument. Library construction, sequencing, and read mapping {#s4} -------------------------------------------------- Each sample was performed in biological triplicate except for the input, which was performed in duplicate. First, 0.5 μg of RNA from each replicate was incubated with RNase III (Ambion cat\#2290) for 10 min at 37° to fragment the RNA and prepare the ends for linker addition. RNA was then concentrated using a concentration module (Invitrogen). The size and quality of the RNA was examined on a 2100 Agilent Bioanalyzer. cDNA libraries were then prepared using the SOLiD Total RNA-Seq Kit (PN 4445374). In brief, adapters were ligated onto the fragmented RNA and reverse transcribed using the SOLiD Total RNA kit. After two rounds of purification and size selection using the Agencourt AMPure XP Reagent, the library was PCR amplified and then purified using the Invitrogen PureLink PCR Micro Kit. Yield and size of amplified DNA was assessed using a 2100 Agilent Bioanalyzer. Each amplified DNA library was then clonally amplified in an emulsion PCR reaction. RIP and total RNA samples were sequenced on a SOLiD 4 platform. All samples were downloaded separately by barcode and reads were mapped to the 2007 *Saccharomyces cerevisiae* reference genome (sg7) using SHRiMP version 2.2.2 ([@bib87]). Reads were trimmed 15 bp from the 3′ end as a quality control measure. During mapping, SHRiMP2 calculated a score for each read based upon mismatches, and reads with \< 90% of the maximum possible score were filtered out (a 90% threshold is similar to allowing for three mismatches). Composite plots {#s5} --------------- Python scripts were used to create composite plots. The purpose of the scripts can be conceptually described as follows. First, the reads were aligned to a TSS (Transcription Start Site) or a TTS (Transcription Termination Site) from a published source ([@bib23]), in a strand-specific manner. The read density was summarized around TSS or TTS mRNAs, then corrected for differences in total uniquely mappable reads so that the final output would be average read density per 100 million reads. Finally, the data were binned into 15 bp bins and smoothed using a 6 bp sliding window average. Enrichment calculation {#s6} ---------------------- See Figure S1 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf) for a diagram of the enrichment calculation. The reads per gene were tabulated after aligning reads to TSSs and TTSs identified from a published source ([@bib23]). Counts per gene were used to measure reproducibility. The abundant reads from ribosomal RNAs were filtered out prior to measuring the reproducibility between the samples. To calculate enrichment, two of three replicates were processed at a time in all three combinations (*i.e.*, rep 1 *vs.* rep 2, rep 2 *vs.* rep 3, and rep 1 *vs.* rep 3) then averaged if the false discovery rate (FDR) was \< 1%. Enrichment was calculated in edgeR for each gene by dividing reads in the immunoprecipitation (IP) sample by reads in the mock IP, while controlling for differences in library size among samples, ([@bib86]). One count was added to all genes using the prior.count argument in order to avoid infinite log~2~ values resulting from 0 reads in the mock IP. Log~2~ and p-values were calculated using the edgeR exacTest (assuming the Poisson model) and the Benjamini--Hochberg method was used to correct p-values for multiple testing ([@bib6]). To identify the regions of the mRNA bound by each factor, enrichment within the 5′ UTR, coding region, and 3′ UTR, reads were counted in a strand-specific manner using BEDOPS ([@bib74]) and published gene annotations ([@bib73]). Following this step, enrichment was calculated in the same manner as for the whole RNA, as described in Figure S1 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf). Enrichment was also calculated across the length of mRNAs based on gene length as follows. Each replicate was first split into reads that aligned to the first-third, middle-third, and last-third of each gene before calculating enrichment. Following this step, enrichment was calculated in the same manner as for the whole RNA. For intronic analysis, HOMER was used to count reads per gene or per intron ([@bib46]). Enrichment was calculated separately for intronic regions using the pipeline described in Figure S1 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf), with the exception of using the saccer2 reference genome. A twofold enrichment cutoff was used to select for enriched intronic regions. Gene ontology (GO) terms analysis {#s7} --------------------------------- For GO terms analysis, the percent ranks for enrichment values were first calculated using the =PERCENTRANK() function in Microsoft Excel. The genes among the top 20th percentile of enrichment values along with a list of all genes ("background population") were submitted to an online GO term finder (go.princeton.edu; [@bib8]), after which p-values were corrected using the Bonferroni method. The following terms were not included in the figures representing the GO terms output due to their redundant nature: "regulation of cellular process," "regulation of biological process," "biological regulation," and "biological_process." The 232 mRNAs connected to "organelle organization" in the [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) GO Term Finder output were submitted to GO Slim Mapper using default settings (<http://www.yeastgenome.org/cgi-bin/GO/goSlimMapper.pl>). Motif analysis {#s8} -------------- Sequence motifs were identified using FIRE (finding informative regulatory elements), which was performed either locally or on the iGET website (<https://iget.c2b2.columbia.edu>) using default settings to analyze "discrete" or "continuous" data on [Puf5](http://www.yeastgenome.org/locus/S000003146/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) unstressed enrichment values using 6, 7, and 8 bins, respectively ([@bib28]). Statistical and visual analysis {#s9} ------------------------------- Correlations were calculated in R using the cor, cor.test, or corr.test (from the "psych" package) functions. Heatscatter graphs were made using the R package "LSD." Circles and ellipses for Venn Diagrams were made using the program eulerAPE ([@bib67]) or using the package "VennDiagram" in R. Other graphs were generated using base R graphics or in Microsoft Excel. Screen shots were generated in IGV after adjusting tag count for differences in library size ([@bib85]; [@bib95]). Average reads per kilobase of transcript per million reads mapped (RPKM) was calculated by first adding 0.5 to each gene so as to not eventually take the log of zero, then calculating RPKM for each input separately, and finally taking the average of the two inputs. ### External data sources: {#s10} The Comparative Dynamic Transcriptome Analysis (cDTA) data ([@bib89], [@bib90]) can be downloaded from the Cramer laboratory website <http://www.cramer.genzentrum.lmu.de/movies/>, or can be found in the supplemental information associated with the respective publications. The RIP-Chip ([@bib35]; [@bib47]), Ribo-seq ([@bib34]), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) CLIP-seq ([@bib70]), [Not1](http://www.yeastgenome.org/locus/S000000689/overview) RIP-seq ([@bib41]), gene run-on (GRO-Chip) ([@bib71]), and genome-wide ChIP ([@bib99]) data sets were downloaded from their respective online supplemental files associated with the publication. Gene expression array data ([@bib9]; [@bib2]) was also found in the online supplemental files associated with their respective publication. The [@bib33]) microarray data were located online at <http://downloads.yeastgenome.org/published_datasets/Expression_connection_data/>. YMC gene expression values can be found at <http://moment.utmb.edu/cgi-bin/dload.cgi>, while processed data were obtained via E-mail (see *Acknowledgments*). ### RT-qPCR verification: {#s11} RNA was isolated from biological triplicate (cells were grown overnight in separate flasks before harvesting) samples. Equivalent amounts of RNA from input and IP samples (∼350 ng) were used to produce cDNA using random hexadeoxynucleotide primers (\[final\] = 0.01 μg/μl, Promega \#C1181) and AMV reverse transcriptase (0.4 unit/µl) for 1 hr at 42° in a 15 μl volume. Real-time PCR was performed using Quanta PerfeCTa SYBR Green SuperMix in a 96-well plate (in technical triplicate) in an Applied Biosystems StepOnePlus machine. For each input and IP, the average Ct value was calculated from the technical triplicates. Then the %IP was calculated using the equation 2\^(Ct~Input(avg)~−Ct~IP(avg)~). The average and SD of three biological replicates was reported using the AVERAGE and STDEV functions in Microsoft Excel. See [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf) for a list of primers. Data availability {#s12} ----------------- The data from this study will be deposited with the Gene Expression Omnibus, accession no. GSE72366. Results {#s13} ======= Factors involved in different aspects of mRNA decay specifically target the 3′ UTR of the same collection of mRNAs {#s14} ------------------------------------------------------------------------------------------------------------------ RNA CLIP procedures are often used to map RNA--protein interactions, but the harvesting, processing, and UV cross-linking steps can introduce stress and change the RNA--protein interaction landscape ([@bib70]; [@bib84]). We used formaldehyde (FA) cross-linking of cells in culture, which rapidly traps protein--nucleic acid interactions in their physiological state with minimal-to-no stress response. All three proteins under study here contact RNA directly ([@bib14]; [@bib16]; [@bib38]; [@bib26]), but FA can potentially cross-link a protein to RNA indirectly via another protein that is in direct contact with the RNA. Percentages of cross-linking efficiency are protein-dependent and hard to estimate *in vivo*, but it is widely accepted that a protein is cross-linked to a nucleic acid only a fraction of the time; therefore, "double-hit" cross-linking is expected to occur at a low frequency. Furthermore, even if the recruitment of a protein to an RNA is through a bridging protein, this in itself is desirable and informative because it indicates that it could regulate the fate of that RNA. Finally, RNA isolation and library construction used in this procedure do not target the poly(A) tail; therefore, potential technical artifacts caused by variations in the poly(A) tail length of messages under the control of the decay machinery are reduced. We identified RNAs associated with [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) as representatives of the mRNA decay machinery because they play different roles in the regulation of mRNAs. [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) is the major mRNA deadenylase; thus, its binding can estimate the decay potential of transcripts in the same way that the binding of RNA polymerase II to a gene is used to determine its transcription level. [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) was chosen for study, as opposed to [Pan2](http://www.yeastgenome.org/locus/S000003062/overview)/[3](https://www.yeastgenome.org/locus/S000001508), because [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) plays a more prominent role in regulating poly(A) tail length and degradation rates ([@bib97]; [@bib90]). Until this study, the recruitment of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) to mRNAs has not been reported on a transcriptome-wide scale in *S. cerevisiae*. RNAs isolated from the immunoprecipitates of extracts from cells containing a myc epitope-tagged version of each protein (*e.g.*, [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) IP), and from extracts of untagged cells (from here on referred to as UT IP), were identified by high-throughput sequencing ([Figure 1A](#fig1){ref-type="fig"}). The assays were carried out in triplicate and the reproducibility among all samples was very high (*R*^2^ values \> 0.93) (Table S1 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf) and [File S1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS1.xlsx)). Enrichment was calculated using the workflow described in the *Materials and Methods* and as illustrated in Figure S1 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf). The vast majority (\> 95%) of the transcripts enriched by each protein were mRNAs (Figure S2 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf), [File S6](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS6.pdf)). We compared the enrichment values of each decay factor to mRNA abundance ([@bib89]) to address the question of whether these proteins are recruited to mRNAs (targeted) or if the enrichment is simply a reflection of their relative abundance in the cell (nonspecific). The mRNA enrichment to all three proteins displayed a negative correlation with abundance ([Figure 1B](#fig1){ref-type="fig"}), with the [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) data showing the strongest anticorrelation (−0.41 Pearson correlation). The negative correlation between RIP-seq enrichment and abundance was observed using the input RNA-seq samples from this study and other published measurements of mRNA abundance (Table S2 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). Thus, these observations are robust and applicable to multiple data sets. Enrichment of a transcript did not correlate with its length on a global scale either (Table S3 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). Thus, the data suggest that these proteins are recruited to specific mRNAs, rather than associating with them indiscriminately based on abundance or length. ![Identification of Ccr4, Dhh1, and Puf5 RNA targets. (A) An illustration of the RNA immunoprecipitation high-throughput sequencing (RIP-seq) procedure (see *Materials and Methods* for details). (B) Heat scatter plots of RIP-seq enrichment values correlated with mRNA abundance levels (mRNA/cell) on a log~2~ scale \[measured in [@bib89]\]. The *R* values within the graphs represent Pearson correlation values. The color refers to the relative concentration of data points in a single area, *i.e.*, red refers to greater density and blue to lower density. The dotted black trend line represents the least squares regression.](315f1){#fig1} [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) display physical and genetic interactions with each other; however, they play different roles in mRNA decay. For example, [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) deadenylates the poly(A) tail located at the 3′ end, and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) regulates translation and decapping occurring at the 5′ end of mRNAs ([@bib97]; [@bib32]; [@bib20]). Furthermore, [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) is 15 and 34 times more abundant than either [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) or [Puf5](http://www.yeastgenome.org/locus/S000003146/overview), respectively ([@bib36]); thus, it was unclear how well the targets of these proteins would overlap and if they would be found at the same location on mRNAs. A pairwise comparison of the enrichment values of RNAs bound to each protein revealed a high level of correlation (*R* \> 0.6, [Figure 2A](#fig2){ref-type="fig"} and Figure S3, A--C in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). Additionally, we validated several targets using an RT-PCR approach, which revealed a good-to-strong correlation between RIP-seq values and those obtained by RT-PCR (Figure S4, A and B in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). ![RNA immunoprecipitation high-throughput sequencing (RIP-seq) reveals that Ccr4, Dhh1, and Puf5 target the same collection of mRNAs (A). Pearson correlations of pairwise RIP-seq sample enrichment values. (B) Venn diagram of targets of Puf5 (green), Dhh1 (red), and Ccr4 (blue) with enrichment values \> 1 log~2~ in two of the three biological replicates using a false discover rate \< 1%. (C) Composite plots of mRNA sequencing read density of all mRNAs with a length of \> 400 nucleotides (nt) (*n* = 4565). The reads were aligned relative to the TSS \[transcription Start Site (left)\] and TTS \[Transcript Termination Site (Right)\] in nt. The TSS and TTS were identified from another study ([@bib23]). Tag count per gene was normalized to 100 million reads to account for differences in library sizes. (D) The enrichment of each factor was calculated over the 5′ untranslated region (UTR), coding sequence (CDS), and 3′ UTR of mRNAs. The number of mRNAs (\> 400 nt) with more than fourfold enrichment is displayed on the *y*-axis.](315f2){#fig2} Next, we examined the overlap of RNA targets of the three factors with each other. For this purpose, we identified RNAs that were enriched at least twofold (\> 1 log~2~) with an FDR \< 1% and present in at least two of the three replicates. Using these criteria, 3157, 2262, and 1241 RNAs were found to be associated with [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview), respectively ([Figure 2B](#fig2){ref-type="fig"}). There was a strong overlap among the targets of these three proteins. Interestingly, most mRNAs bound by [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) or [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) were also targets of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) ([Figure 2B](#fig2){ref-type="fig"}). These results suggest that the physical interactions between these proteins that have been identified in previous studies reflect their coregulation of mRNAs *in vivo*. Furthermore, the overlap of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) targets is consistent with studies that have shown that [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) recruits the [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not complex to the 3′ UTR of the *[HO](http://www.yeastgenome.org/locus/S000002386/overview)* mRNA ([@bib38], [@bib39]). Moreover, motifs (H[UGUA]{.ul}NH[A]{.ul}D) with likeness (underlined portion is similar) to both [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) ([UGUA]{.ul}AY[A]{.ul}WUA) and the related RBP [Puf4](http://www.yeastgenome.org/locus/S000002982/overview) (WH[UGUA]{.ul}HA[W]{.ul}UA) were overrepresented among the RNAs associated with all three proteins (p-value \< 10^−4^, Figure S3D in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)) ([@bib35]). The discovery of motifs similar to both Puf4- and Puf5-binding sites in the enriched mRNAs may be explained by the ability of [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) to adapt to variations in the RNA motif and recognize related sequences ([@bib103]). Although [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) was detected with the vast majority of mRNAs associated with [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) and/or [Puf5](http://www.yeastgenome.org/locus/S000003146/overview), an additional ∼1000 mRNAs were exclusive to the [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) data set. These messages were depleted of the [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) and/or [Puf4](http://www.yeastgenome.org/locus/S000002982/overview) motifs ([Figure 2B](#fig2){ref-type="fig"} and Figure S3D in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf), p-value \< 10^−5^), and [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) is likely recruited to this group of mRNAs by other RBPs (see below). The 5′ and 3′ UTRs of mRNAs contain sequences that control the translation, localization, and decay of mRNAs ([@bib52]; [@bib47]; [@bib60]). RIP sample preparation required a sonication step to solubilize cross-linked RNA--protein complexes, resulting in the shearing of RNAs to ∼200--600 nucleotides (nt). Thus, information on the spatial location of these proteins on mRNAs can be obtained. Composite plots were generated by aligning the sequencing reads to the TSS and the transcript TTS identified from a published high-resolution mapping of the transcriptome ([@bib23]). A strong enrichment of [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) was observed over the 3′ end of mRNAs ([Figure 2C](#fig2){ref-type="fig"}), which agrees with studies showing that [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) binds to the 3′ UTR of mRNAs ([@bib35]; [@bib38]). Where [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) binds on an mRNA has not been examined on a global scale, but its function in deadenylation predicts that it would bind near the poly(A) tail. The mapping of reads from the [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) IP revealed that it is preferentially recruited to 3′ ends of mRNAs ([Figure 2C](#fig2){ref-type="fig"}). Likewise, [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) showed a stronger enrichment to the 3′ end of mRNAs. While the cross-linking of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) was strongest at the 3′ end of mRNAs, reads above background were also observed in the middle and 5′ end. The detection of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) in the body of the mRNA is unlikely to result from incompletely fractionated mRNAs during sample preparation, because all three samples were processed identically and equivalent enrichment of [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) along the length of the mRNA was not observed. While it was unexpected that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) would be enriched at both ends of mRNAs, our results with [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) agree with a CLIP-seq study conducted in stressed cells, which observed it cross-linking to both the 5′ and 3′ ends of transcripts ([@bib70]). The cross-linking of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) at multiple locations on the transcript could result from the packaging of mRNAs into mRNPs, and the juxtaposition of the 5′ and 3′ ends of the RNA caused by associations between factors binding to the 5′ and 3′ ends. Indeed, [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) interacts with both [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not and factors associated with the 5′ cap of mRNAs ([@bib44]; [@bib64]; [@bib75]). Composite plots can be biased by a large number of reads from a few mRNAs or can underestimate the extent of recruitment to a smaller number of transcripts. Therefore, we calculated the enrichment of each factor over the 5′ UTR, the coding sequence (CDS), and the 3′ UTR of mRNAs with a length \> 400 nt. The frequency (number of mRNAs) that showed strong enrichment over these regions was calculated and displayed in [Figure 2D](#fig2){ref-type="fig"}. Similar to what was observed in the composite plots, the results indicate that each protein was associated predominantly with the 3′ UTR of mRNAs, and that [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) and [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) were detected at the 5′ UTR and CDS of more mRNAs than [Puf5](http://www.yeastgenome.org/locus/S000003146/overview). An alternative way to analyze this is to separate the mRNAs into thirds, based on lengths, and calculate the number of RNAs where enrichment was detected at the 5′ ends, middle, or 3′ ends of the mRNA. Here too, the results indicate that each protein predominantly associated with the 3′ ends of the mRNA (Figure S5 in [File S3](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS3.pdf)). The recruitment of [Puf5](http://www.yeastgenome.org/locus/S000003146/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) to the 3′ ends of mRNAs was observed at individual mRNAs (Figure S6 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). This is the first study to map mRNAs bound by [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview); however, targets of [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) have been identified using RIP-Chip and CLIP-seq methods. We next compared our [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) targets to published data sets. A native RIP-CHIP procedure conducted on [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) identified far fewer RNA targets than we did (only 224; [@bib35]), which may be explained by a lack of a cross-linking step and/or a high background of native RIP procedures. Nonetheless, we identified many of the same targets, and a significant positive correlation was observed between the RIP-Chip study and our own data (Figure S7, A and B in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). The number of transcripts bound to [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) in our study differs significantly from two other studies that used native (no cross-linking) conditions. When native RIP-Chip and RIP-seq methods were employed, \< 80 Dhh1-enriched RNAs were detected ([@bib47]; [@bib13]). Considering the high abundance of [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) and its known functions in mRNA regulation, one can surmise that omitting a cross-linking step greatly underestimated the number of targets. [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) targets were identified using CLIP-seq in cells undergoing glucose deprivation stress ([@bib70]). This study identified 299 high- and 1838 low-confidence mRNAs (twofold cutoff and 2% FDR). A comparison of the targets identified in our study (2262 targets) to those identified in the CLIP-seq study (1838 targets) found relatively little overlap between the two data sets (548 targets; Figure S7C in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). Since the CLIP-seq study was conducted in cells deprived of a carbon source (stressed), and ours was carried out in rich media, the relatively weak overlap between the two studies is not surprising, considering that glucose starvation causes strong processing body (p-body) formation and presumably widespread remodeling of RNA--protein interactions ([@bib94]; [@bib70]). The mRNA decay machinery shares a regulatory network with other RBPs {#s15} -------------------------------------------------------------------- Decay factors are recruited to mRNAs by RBPs that recognize sequence motifs located within the 3′ UTR ([@bib37]; [@bib80]). Information on how [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) are recruited to mRNAs can be obtained by identifying sequence motifs in the bound mRNAs and by identifying overlap with publicly available RNA--protein interaction data sets. A search for sequence motifs among our highly enriched [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) targets using FIRE ([@bib28]) identified a motif resembling the [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) motif, as defined by the FIRE program ([Figure 3A](#fig3){ref-type="fig"}, p-value = 1.3 × 10^−12^). Identification of the [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) motif in our data set, and the correlation between our targets and those identified in other studies ([Figure 3A](#fig3){ref-type="fig"} and Figure S7, A and B in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)), validate our RIP-seq procedure. [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not is not known to display sequence-specific RNA binding, but instead is recruited to mRNAs via RBPs. The overlap of mRNA targets of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) suggests that [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) is a candidate ([Figure 2, A and B](#fig2){ref-type="fig"}), and we also discovered that the binding motif for [Puf3](http://www.yeastgenome.org/locus/S000003936/overview) was overrepresented in the most highly enriched [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) targets (p-value = 6.5 × 10^−4^, [Figure 3B](#fig3){ref-type="fig"}). It should be noted that the [Puf3](http://www.yeastgenome.org/locus/S000003936/overview) motif emerging from the FIRE analysis of our data is a more restricted version than the motif identified in another study ([@bib35]). Finding the [Puf3](http://www.yeastgenome.org/locus/S000003936/overview) motif overrepresented in [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) targets is in good agreement with a study suggesting that [Puf3](http://www.yeastgenome.org/locus/S000003936/overview) interacts with [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) to regulate deadenylation of *[COX17](http://www.yeastgenome.org/locus/S000003932/overview)* mRNA ([@bib58]). This, and our own data, suggests that [Puf3](http://www.yeastgenome.org/locus/S000003936/overview) plays a significant role in recruiting [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) to mRNAs across the genome. Neither our nor the published CLIP-seq study identified a motif significantly enriched in [Dhh1](https://www.yeastgenome.org/locus/S000002319)-bound transcripts ([@bib70]). ![The decay machinery is directed to mRNAs through a network of RNA-binding proteins (RBPs). Enrichment values were submitted to FIRE (finding informative regulatory elements; [@bib28]) using default setting options for Puf5 (A) and Ccr4 (B). Values were binned into six and eight bins, respectively. The black and red heatmap represents the distribution of enrichment values contained within each bin (see range of enrichment values to the left in red font). Yellow bins represent groups of genes that have a significant overrepresentation of the motif, while blue represent bins where the motif is significantly underrepresented. See [@bib28] for a detailed description of how the optimized motifs and location were attained. (C) Pearson correlations between RNA immunoprecipitation high-throughput sequencing (RIP-seq) enrichment values and enrichment values of multiple RBP targets from another study ([@bib47]). Multiple testing correction was performed and only correlations that met a false discover rate \< 1% threshold are included.](315f3){#fig3} To find other RBPs that coregulate targets of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview), correlations were calculated between the enrichment values from this study and those of other RIP-Chip studies ([@bib47]) ([Figure 3C](#fig3){ref-type="fig"}). Generally, [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) enrichment positively correlated with targets of sequence-specific RBPs that have been implicated in mRNA decay or translational repression, including [Vts1](http://www.yeastgenome.org/locus/S000005886/overview), [Puf4](http://www.yeastgenome.org/locus/S000002982/overview), and [Khd1](http://www.yeastgenome.org/locus/S000000128/overview). [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not, [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) have genetic and physical interactions with [Puf4](http://www.yeastgenome.org/locus/S000002982/overview) and [Vts1](http://www.yeastgenome.org/locus/S000005886/overview) ([@bib48]; [@bib83]; [@bib93]), and [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Khd1](http://www.yeastgenome.org/locus/S000000128/overview) regulate the mRNAs of *[LRG1](http://www.yeastgenome.org/locus/S000002399/overview)* and *[ROM2](http://www.yeastgenome.org/locus/S000004363/overview)* \[also identified in this work ([File S2](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS2.xlsx))\] ([@bib51]). Furthermore, there was a strong correlation between the mRNA targets identified here and those bound to [Nab2](http://www.yeastgenome.org/locus/S000003090/overview) and [Npl3](http://www.yeastgenome.org/locus/S000002840/overview), two nuclear poly(A)-binding proteins connected to mRNA export ([@bib104]; [@bib59]; [@bib45]; [@bib65]). Physical interactions between [Nab2](http://www.yeastgenome.org/locus/S000003090/overview) and [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not complexes have been identified, suggesting that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not regulates some aspect of mRNA transport ([@bib53]). [Nab2](http://www.yeastgenome.org/locus/S000003090/overview) binds poly(A) tails, and while thought to protect mRNAs from degradation, may aid in the recruitment of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not to mRNAs in the nucleus. In contrast, enrichment of the decay factors examined here negatively correlated with the targets of factors involved in ribosomal biogenesis (*i.e*., [Cbf5](http://www.yeastgenome.org/locus/S000004165/overview), [Nop56](http://www.yeastgenome.org/locus/S000004187/overview), and [Sof1](http://www.yeastgenome.org/locus/S000003934/overview)). Ccr4-Not directly regulates both mRNA synthesis and decay, but identification of its targets suggests that its role in decay predominates in gene expression control {#s16} -------------------------------------------------------------------------------------------------------------------------------------------------------------------- The function of a gene is often inferred from gene expression profiling of mutants. However, interpretation of this type of data is difficult if the gene of interest has multiple functions in the cell. For example, [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) associate with RNA polymerase II and regulate transcription at the initiation and elongation stages of mRNA synthesis, and also function in decay ([@bib24]; [@bib54]; [@bib68]). Deleting [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not subunits result in both increased and decreased expression of many genes ([@bib22]; [@bib3]), but which effects result from changes in transcription *vs.* decay are not known. RIP-seq enrichment data were compared to the decay and synthesis rates of mRNAs determined by cDTA, which measures these parameters with minimal perturbations by metabolic labeling of mRNAs ([@bib89]). [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) enrichment showed a positive correlation with mRNA decay rates (*R* = 0.43 and 0.38, respectively) ([Figure 4](#fig4){ref-type="fig"}), which is consistent with their role in mRNA decay ([@bib97], [@bib98]). A positive correlation with decay rates, albeit not as strong, was observed also for [Puf5](https://www.yeastgenome.org/locus/S000003146)-enriched mRNAs (*R* = 0.28). When the correlation between enrichment and decay rate was calculated using ranks (Spearman), the correlation was even stronger \[0.53 ([Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)), 0.47 ([Dhh1](http://www.yeastgenome.org/locus/S000002319/overview)), and 0.34 ([Puf5](http://www.yeastgenome.org/locus/S000003146/overview))\]. In contrast, correlations between the RIP-seq data from all three proteins and mRNA synthesis rates were weaker, and showed a modest negative-to-no correlation ([Figure 4](#fig4){ref-type="fig"}). ![Recruitment of Ccr4, Dhh1, and Puf5 correlates with decay and not synthesis rates. Pair-wise comparisons of enrichment values (*y*-axis) *vs.* decay rates (1/min) and synthesis rates (mRNA/Cell/Cell Cycle, Log~2~) from [@bib89] and ribosome footprints \[reads per kilobase of transcript per million reads mapped (RPKM), Log~2~\] from [@bib34]. Numbers within the graphs represent Pearson correlation values. The dotted black trend line represents the least squares line of regression.](315f4){#fig4} The synthesis rates measured by cDTA analysis were extrapolated from measured abundance and turnover rates ([@bib89]). GRO-Chip analysis has been performed in yeast, which directly measures the transcription frequency across the genome ([@bib71]). To further verify that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) recruitment negatively correlated with mRNA synthesis rates, we examined the correlation between [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) recruitment and mRNA synthesis rates measured by GRO-Chip and found a similar anticorrelation (Pearson = −0.25, Figure S8 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). Therefore, using two different estimates of mRNA synthesis, [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) recruitment to an mRNA negatively correlated with its synthesis rate. Since [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) associate with elongating RNAPII ([@bib54]; [@bib27]), we considered that the recruitment that we detected may be linked to the process of transcription. To examine this possibility, we first looked for enrichment of these factors within intronic sequences. Yeast has 282 genes with introns (<http://intron.ucsc.edu/yeast4.1/>), and less than nine of these mRNAs showed enrichment of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), or [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) over both the exon and intron (Table S4 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). Additionally, neither [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) (*R* = −0.03) nor [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) (*R* = −0.10) recruitment to mRNAs correlated with a published genome-wide ChIP-seq experiment that mapped the association of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not subunits with genes ([@bib99]). Hence, few RNA--protein interactions detected here are likely to occur cotranscriptionally, and they are more likely to occur on cytoplasmic RNAs. Taken together, these results suggest that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not recruitment to mRNAs reflects its function in decay as opposed to synthesis. A recent report suggested that the synthesis of ribosomal protein mRNAs may be regulated by [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not in a Not5-dependent manner during stress ([@bib41]). However, we did not find a strong correlation with synthesis across the genome. The differences may be related to the subunits examined (NOTs *vs.* [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview)), the RNA enrichment strategy (native *vs.* cross-linked), or the positive correlation with transcription being restricted to a particular class of genes (see *Discussion*). mRNA decay factor recruitment anticorrelates with ribosome occupancy {#s17} -------------------------------------------------------------------- A long-standing model posits that mRNAs are either partitioned into the translatable or nontranslatable pool, with the former category encompassing mRNAs that are associated with ribosomes and undergoing translation ([@bib94]; [@bib76]). We next explored the relationship between decay factor recruitment and the translation of mRNAs measured by Ribo-seq ([@bib50]; [@bib34]). [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) recruitment displayed an anticorrelation with ribosome footprint data (R = −0.52, −0.42 and −0.33, respectively) ([Figure 4](#fig4){ref-type="fig"}). Thus, mRNAs bound to decay factors, especially [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), are less likely to be translated at a high rate. This observation supports the long-standing model that deadenylation and translation are in competition ([@bib76]), and is also consistent with the role of [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) in translation repression ([@bib79]). Identification of Ccr4-associated mRNAs suggests that the two deadenylase complexes in yeast regulate different groups of mRNAs {#s18} ------------------------------------------------------------------------------------------------------------------------------- [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Pan2](http://www.yeastgenome.org/locus/S000003062/overview)/[3](https://www.yeastgenome.org/locus/S000001508) are the main mRNA deadenylases in yeast, and two models have been proposed to explain how they cooperate in the decay of messages. The first model suggests that [Pan2](http://www.yeastgenome.org/locus/S000003062/overview)/[3](https://www.yeastgenome.org/locus/S000001508) initiates poly(A) shortening, which is then followed by complete deadenylation by [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not ([@bib10]; [@bib97]; [@bib100]). The second model, based on a study that found no correlation between the changes in decay rates of mRNAs in a *[ccr4](http://www.yeastgenome.org/locus/S000000019/overview)*∆ *vs.* a *[pan2](http://www.yeastgenome.org/locus/S000003062/overview)*∆ mutant, suggests that they each target different collections of mRNAs ([@bib90]). Little is known about the interplay between these two complexes on the transcriptome; thus, we used the [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) RIP-seq data to shed light on the contributions of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Pan2](http://www.yeastgenome.org/locus/S000003062/overview)/[3](https://www.yeastgenome.org/locus/S000001508) to mRNA decay. The [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-enriched mRNAs were divided equally into five percentile ranks, and boxplots were created using the relative abundance values in a *[ccr4](http://www.yeastgenome.org/locus/S000000019/overview)*∆ mutant *vs.* a wild-type strain ([Figure 5A](#fig5){ref-type="fig"}). This plot illustrates that the most highly enriched mRNAs increased in abundance in a *[ccr4](http://www.yeastgenome.org/locus/S000000019/overview)*∆ deletion (top 20%, p-value \< 2.2 × 10^−16^), and the lowly (essentially no recruitment) enriched mRNAs actually decreased in abundance (bottom 20%, p-value \< 2.2 × 10^−16^). Although almost all mRNAs had decreased decay rates in the *[ccr4](http://www.yeastgenome.org/locus/S000000019/overview)*∆ mutant, mRNAs that were highly enriched by [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) had the largest decrease in their decay rate (p-value \< 2.2 × 10^−16^, [Figure 5B](#fig5){ref-type="fig"}), and those that were lowly recruited had the smallest decrease in their decay rate ([Figure 5B](#fig5){ref-type="fig"}). Similar correlations were also observed between [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) enrichment and changes in abundance and decay in a *[dhh1](http://www.yeastgenome.org/locus/S000002319/overview)*∆ mutant (Figure S9, A and B in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). ![Identification of Ccr4 targets suggests a division of labor between the two cytoplasmic deadenylases. (A) Ccr4 RNA immunoprecipitation high-throughput sequencing (RIP-seq) enrichment values were equally divided into five groups based upon their percent rank (*i.e.*, 80--100 is the top 20th percentile). Each boxplot represents the log2 change in abundance in a *ccr4*∆ strain relative to a wild-type strain ([@bib90]). Statistical significance was calculated using a Wilcoxon rank sum test with continuity correction. The red line indicates the median of all values. (B) The same as (A) but the boxplots represent the log2 change in decay rates in a *ccr4*∆ strain relative to wild-type. (C) The same as (A) but the boxplots represent the log2 change in expression in a *pan2*∆ strain relative to wild-type. (D) The same as (A) but the boxplots represent the log2 change in decay rates in a *pan2*∆ strain relative to wild-type. In order to clearly visualize the boxplots in (A), (C), and (D), two outlier data points were removed from the *y*-axis. \* p-value \< 0.001, \*\* p-value \< 0.0001, and \*\*\* p-value \< 2.2 × 10^−16^.](315f5){#fig5} mRNAs that were stabilized in a *[pan2](http://www.yeastgenome.org/locus/S000003062/overview)*∆ mutant were either not strongly affected or were turned over more rapidly in a *[ccr4](http://www.yeastgenome.org/locus/S000000019/overview)*∆ strain ([@bib90]), suggesting that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Pan2](http://www.yeastgenome.org/locus/S000003062/overview)/[3](https://www.yeastgenome.org/locus/S000001508) may regulate different mRNAs. Consistent with this, we found that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-enriched transcripts displayed reduced abundance (p-value \< 2.2 × 10^−16^, [Figure 5C](#fig5){ref-type="fig"}) and increased decay rates (p-value \< 0.001, [Figure 5D](#fig5){ref-type="fig"}) in the *[pan2](http://www.yeastgenome.org/locus/S000003062/overview)*∆ mutant. Also, mRNAs that were lowly enriched increased in abundance (p-value \< 2.2 × 10^−16^, [Figure 5C](#fig5){ref-type="fig"}) and displayed reduced decay rates (p-value \< 0.001, [Figure 5D](#fig5){ref-type="fig"}) in the *[pan2](http://www.yeastgenome.org/locus/S000003062/overview)*∆ mutant. In other words, the abundance and stability of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-bound mRNAs are most affected by the deletion of *[CCR4](http://www.yeastgenome.org/locus/S000000019/overview)*, while the lowly enriched targets are more strongly affected by the deletion of *[PAN2](http://www.yeastgenome.org/locus/S000003062/overview)*. The decrease in abundance of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) targets in the *[pan2](http://www.yeastgenome.org/locus/S000003062/overview)*∆ mutant is interesting. The cause of this is not known but it suggests that, in the absence of [Pan2](http://www.yeastgenome.org/locus/S000003062/overview), either [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) compensates for its loss by increasing its activity or another redundant pathway, such as the exosome, becomes more prevalent. Ccr4 and Dhh1 bind mRNAs that respond to the metabolic state of the cell {#s19} ------------------------------------------------------------------------ To identify the pathways regulated by [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview), lists of the most highly enriched mRNAs (top 20%) were submitted to GO terms analysis. The [Puf5](https://www.yeastgenome.org/locus/S000003146)-bound mRNAs had terms associated with transcription and RNA biogenesis overrepresented, among others (Figure S10 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). This is consistent with a previous native RIP-Chip study, which also revealed that [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) targets transcriptional and chromatin regulators ([@bib35]). The mRNAs that showed the highest level of recruitment to [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) code for proteins involved in many of the same processes, such as RNA metabolism, nitrogen compound metabolism, transcription, and several biosynthetic processes ([Figure 6, A and B](#fig6){ref-type="fig"}). The identification of many of the same GO terms in the [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) data sets is consistent with the high degree of overlap between the mRNA targets of these two proteins ([Figure 2, A and B](#fig2){ref-type="fig"}). [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) was also recruited to 232 mRNAs important for organelle organization processes ([Figure 6B](#fig6){ref-type="fig"}). This collection of mRNAs was further analyzed using GO Slim Mapper to narrow down the organelle(s), which revealed that 26% of the messages were specifically connected to mitochondrial organization ([File S4](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS4.xlsx)). {#fig6} Since mRNAs involved in multiple metabolic processes were identified in the RIP-seq data sets, we examined the correlation between our recruitment data and gene expression changes during metabolic stress ([@bib9]). We found that the most highly enriched mRNAs (top 10th percentile) bound to [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) overlapped significantly with genes that are induced during nitrogen and carbon starvation (Figure S11, A and B in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf); p-values from \< 0.001 to \< 1 × 10^−14^). In order to establish if this relationship reached beyond the top 10th percentile of targets, we calculated the correlation between RIP-seq data and the changes in gene expression over time during carbon and nitrogen starvation ([Figure 7](#fig7){ref-type="fig"} and Figures S11 and S12 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) showed positive correlations between its enrichment values and gene expression levels during carbon starvation ([Figure 7A](#fig7){ref-type="fig"}). Furthermore, the Pearson correlation values increased as cells progressed deeper into the starvation program, which most likely results from the accumulation of more genes with expression changes over time ([@bib9]). Although [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) enrichment did not correlate with gene expression changes during nutritional stress as strongly as [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) did, the same trends between recruitment and gene expression changes were still observed ([Figure 7B](#fig7){ref-type="fig"} and Figures S11 and S12 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). ![Ccr4 and Dhh1 are recruited to mRNAs that respond to nutrient and metabolic signals. (A) Pearson correlation values were calculated between Ccr4 RNA immunoprecipitation high-throughput sequencing (RIP-seq) enrichment and expression values from the carbon depletion experiment in [@bib9]. Each column represents a different time point after carbon depletion relative to the unstressed condition. (B) The same as in (A), except that Dhh1 RIP-seq enrichment values were used. (C and D) Scatter plots comparing RIP-seq enrichment to the change in decay rates in a deletion strain relative to wild-type. The plots for Ccr4 (C) and Dhh1 (D) are shown. The color of each data point represents how the expression changed at the 240 min time point after carbon depletion from [@bib9]. Genes that were upregulated (log2 expression \> 1), downregulated (log2 expression \< −1), or unchanged were colored red, blue, and gray, respectively. The dashed black line trend line represents the least squares line of regression. (E) A χ^2^ test was performed between the highly enriched Ccr4 targets (top 10th percentile) against the upregulated genes \[top 10th percentile of genes (red area)\] and downregulated genes \[bottom 10th percentile (blue area)\] from [@bib96]. The periods when oxygen consumption was increasing (Inc) or decreasing (Dec) is displayed above the graph.](315f7){#fig7} We found that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) recruitment correlated with changes in gene expression during carbon source deprivation stress. We next attempted to determine how many of these differentially expressed mRNAs are dependent upon [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and to what extent the correlations are driven by changes in the decay of these messages. We constructed scatter plots that displayed mRNA enrichment (*y*-axis) *vs.* changes in decay rates (*x*-axis) in the mutants. Each point was then colored red, blue, or gray to indicate if the mRNA expression increased, decreased, or remained unchanged in stressed cells, respectively. The plot shows that the upregulated mRNAs were predominantly those that showed the highest [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) recruitment and the largest decrease in decay rate in the *[ccr4](http://www.yeastgenome.org/locus/S000000019/overview)*∆ mutant ([Figure 7C](#fig7){ref-type="fig"}, *R* = −0.31, p-value \< 2.2e−16). Conversely, the genes that were downregulated more often had transcripts that displayed a modest increase in decay rate and were less enriched with [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) compared to upregulated genes. The same trend was observed for [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) ([Figure 7D](#fig7){ref-type="fig"}), but to a lesser extent (*R* = −0.16). Finally, repeating the same analysis in cells undergoing nitrogen depletion stress also revealed the same pattern (Figure S12 in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). In summary, the genes that are differentially expressed during metabolic stress are regulated during unstressed conditions by [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), which likely involves changes in mRNA stability. Methods of carbon and nitrogen depletion differ between studies, as do other variables. For example, the study by [@bib9]), which was used as the data set for our analysis, depleted nutrients by transferring cells on membranes between different media conditions (in this case, nitrogen base plates replete of depleted of nutrients). To demonstrate that our conclusions are robust and not unique to any depletion technique or condition, we repeated our analysis using data sets from two other studies measuring gene expression changes during nutrient limitations (Figure S11C in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). One study that measured the response to nitrogen depletion shifted cells from liquid culture in nitrogen-containing to nitrogen-depleted medium ([@bib33]). Another measured changes in gene expression by shifting cells in rich (YPD) liquid media to medium lacking dextrose ([@bib2]). The latter condition is most like those used to grow cells in this work. Repeating the analysis using these other data sets lead to the same conclusion: the RIP-seq data correlated with the changes in gene expression during carbon and nitrogen starvation (Figure S11C in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). Thus, we found similar correlations between the RIP-seq targets and changes in gene expression from three different studies, suggesting that the correlations we identified are robust and not unique to any one depletion protocol, data set, or strain used to generate the expression data. The identification of many mRNA targets associated with metabolic processes prompted us to investigate the relationship between mRNA decay machinery recruitment and changes in gene expression during the YMC. The YMC is a process in which budding yeast cells oscillate between the storage (reductive, nonrespiratory phase) and consumption (oxidative, respiration) of metabolites, which produces waves of gene expression that correlate with bursts of O~2~ utilization ([@bib96]; [@bib12]). The overlap between the most highly enriched [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) targets (top 10%) and mRNAs that were highly or lowly expressed during the YMC was calculated, a χ^2^ test was performed, and the p-values of the overlap were plotted *vs.* time ([Figure 7E](#fig7){ref-type="fig"}). The top [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) targets significantly overlapped (p-value \< 10^−4^) with the most highly expressed genes when cells decreased their O~2~ consumption during the nonrespiratory phase ([Figure 7E](#fig7){ref-type="fig"}, see time points 2, 14--15, and 26--28). Strikingly, as cells progressed out of a nonrespiratory phase and into the respiratory phase (increasing O~2~ consumption), the top [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) targets significantly overlapped (p-value \< 10^−4^) with the strongly repressed genes (bottom 10%) ([Figure 7E](#fig7){ref-type="fig"}, see time points 8--12, 19--24, 32, and 34--36). These patterns suggest that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) may target specific transcripts in response to metabolic signals that arise from a change in the redox or metabolic state of the cell. Comparing the overlap between [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) targets and fluctuations in gene expression produced similar trends, although the level of overlap, based on p-values, was less extensive than those observed for [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-associated transcripts (Figure S12A in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). In contrast, there was no significant overlap between [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) targets and the genes fluctuating during the YMC (Figure S12B in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). Interestingly, the [Puf3](http://www.yeastgenome.org/locus/S000003936/overview) motif was overrepresented both among the top [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) targets identified in our study ([Figure 3B](#fig3){ref-type="fig"}) and among the genes that showed peak expression during the transition from increased to decreased oxygen consumption ([@bib96]). This suggests that [Puf3](http://www.yeastgenome.org/locus/S000003936/overview), a protein shown to recruit [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not to a single mRNA encoding a protein involved in regulating mitochondrial respiration, [COX17](http://www.yeastgenome.org/locus/S000003932/overview), ([@bib58]), regulates the association of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not with many transcripts during the YMC. Discussion {#s20} ========== The integration of mRNA expression data (*e.g.*, microarrays) with transcription factor recruitment throughout the genome (ChIP-Chip or ChIP-seq) has been an important advance in our understanding the mechanisms of transcription regulation. For instance, correlating the binding of RNA polymerase II to genes with changes in mRNA levels has identified direct effects and provided deeper mechanistic insights into the process of gene transcription. In contrast to transcriptional regulators, which have been studied extensively, much less is known about the relationship between the binding of mRNA regulatory factors and the fate of mRNAs on a global scale. Our study identified the targets of three characterized mRNA decay factors, [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview), and these data were used to address their roles in the control of gene expression. Identifying mRNA targets of Ccr4 provides insights into its role in transcription and decay {#s21} ------------------------------------------------------------------------------------------- The [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not complex regulates mRNA levels by directly regulating synthesis in the nucleus and mRNA decay in the cytoplasm ([@bib68]). Expression profiling in [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not mutants has been performed ([@bib22]; [@bib3]; [@bib90]), but disentangling direct *vs.* indirect effects and attributing a change in mRNA levels to altered synthesis or decay is not trivial without a global map of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)--mRNA interactions. We utilized existing genome-wide mRNA synthesis and decay rate data to illustrate that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) recruitment, and that of its binding partners [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) and [Puf5](http://www.yeastgenome.org/locus/S000003146/overview), correlated better with decay rather than synthesis. In fact, the highly enriched mRNAs displayed a lower synthesis rate overall. Considering that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not directly promotes elongation *in vitro* ([@bib54]; [@bib4]; [@bib27]), cross-links to active genes ([@bib54]; [@bib99]; [@bib41]), and was first identified as a transcription factor ([@bib18]; [@bib68]), its anticorrelation to synthesis rates was surprising on the surface. However, an explanation for this observation is that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not is utilized more at lowly synthesized genes that are under greater elongation control. [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) is the main cytoplasmic deadenylase in eukaryotes and plays a global role in decay. Therefore, it might be expected that it is bound to nearly every mRNA. We found this not to be the case. There are a number of factors, some technical and others biological, which can explain this. Technically, the stringent 1% FDR filtering and imposition of a twofold cut off would exclude some mRNAs from the data set. A biological explanation is redundancy with the other deadenylase, [Pan2](http://www.yeastgenome.org/locus/S000003062/overview). Comparison of the level of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) recruitment to the changes in gene expression in *[ccr4](http://www.yeastgenome.org/locus/S000000019/overview)*∆ and *[pan2](http://www.yeastgenome.org/locus/S000003062/overview)*∆ mutants showed that mRNAs that are not/lowly bound to [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) are more strongly affected by the deletion of *[PAN2](http://www.yeastgenome.org/locus/S000003062/overview)* ([Figure 5](#fig5){ref-type="fig"}). This is highly suggestive of redundancy between these two deadenylases. There may be a contribution from deadenylation-independent decay pathways as well, such as exosome-mediated decay ([@bib88]). Analysis of decay factor targets does not support a widespread imprinting mechanism as a major mode of action to regulate mRNA levels {#s22} ------------------------------------------------------------------------------------------------------------------------------------- As stated above, decay factors such as [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not, [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), [Xrn1](http://www.yeastgenome.org/locus/S000003141/overview), [Lsm1](http://www.yeastgenome.org/locus/S000003660/overview), and [Dcp2](http://www.yeastgenome.org/locus/S000005062/overview) cross-link to the promoters and open reading frames of highly transcribed genes ([@bib61]; [@bib54]; [@bib43]; [@bib31]). These findings, along with others, have led to speculation that mRNAs could be imprinted with decay factors during transcription in the nucleus, where they mark mRNAs for post-transcriptional regulation and decay in the cytoplasm ([@bib42]; [@bib82]). [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not is a good candidate to carry out this function, since it participates in both the synthesis and destruction of mRNAs ([@bib68]; [@bib82]; [@bib19]; [@bib41]). If the imprinting of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) onto mRNAs during transcription was an obligate step in the targeting of mRNAs for regulation in the cytoplasm, one might expect that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) would be loaded onto the vast majority of newly synthesized RNAs proportional to their synthesis; thus, [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) recruitment would correlate positively with transcription rates. As discussed above, this was not the case. These results do not rule out the possibility that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not is loaded onto mRNAs cotranscriptionally and escorts them out into the cytoplasm, but once in the cytoplasm its binding to mRNAs may be subject to dynamic regulation driven by normal cellular cues and/or redistribution to specific mRNAs by sequence-specific mRNA-binding proteins. A recent paper described a genome-wide map of [Not1](http://www.yeastgenome.org/locus/S000000689/overview) targets ([@bib41]). The authors of that paper concluded that mRNAs are imprinted with [Not1](http://www.yeastgenome.org/locus/S000000689/overview) in the nucleus, which affects the ultimate translation of the mRNA. This appears to apply to a subset of genes, such as those encoding ribosomal proteins. We found some similarities between the [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) targets described here and [Not1](http://www.yeastgenome.org/locus/S000000689/overview) targets identified in the other study. For example, both [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Not1](http://www.yeastgenome.org/locus/S000000689/overview) associated with mRNAs with a [Puf3](http://www.yeastgenome.org/locus/S000003936/overview) motif and were bound by other RBPs known to genetically or physically interact with [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not ([@bib41]). Nonetheless, the correlation between the [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Not1](http://www.yeastgenome.org/locus/S000000689/overview) RIP-seq data was weak (Spearman's *R* = 0.11, Figure S13A in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)), and our [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) recruitment data more strongly correlated with global mRNA decay rates (*R* = 0.53) than the [Not1](http://www.yeastgenome.org/locus/S000000689/overview) data did (*R* = 0.06) (see [Figure 4](#fig4){ref-type="fig"} and Figure S13B in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). Additionally, we found a smaller number of overlapping (562 mRNAs) targets than anticipated, considering that the two proteins reside in the same complex and results indicate that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) must bind to [Not1](http://www.yeastgenome.org/locus/S000000689/overview) to carryout mRNA decay ([@bib5]). Having said that, technical differences between the studies may explain this. Most notably, we used rapid formaldehyde cross-linking and detergents in the isolation step, which allows for more stringent determination of targets, and would prevent the redistribution of RBPs and inhibit degradation of mRNAs during isolation ([@bib84]). As we have found to be the case for [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), studies not using a cross-linking step ([@bib47]; [@bib13]) identified far fewer targets than those that did \[ours and that of [@bib70])\]. Another significant difference is the proteins examined. We examined [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), proteins that form the nuclease module of the complex. It is possible that [Not1](http://www.yeastgenome.org/locus/S000000689/overview) has a different behavior than the nuclease submodule subunits analyzed here. [@bib41]) demonstrated by immunofluorescence that [Not1](http://www.yeastgenome.org/locus/S000000689/overview) staining is largely nuclear, while multiple studies have shown that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) are predominantly cytoplasmic ([@bib97]; [@bib32]; [@bib70]). Furthermore, [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) (and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview)) bind the N-terminus of [Not1](http://www.yeastgenome.org/locus/S000000689/overview) and display different genetic interactions than those that bind to the C-terminus of [Not1](http://www.yeastgenome.org/locus/S000000689/overview), such as [Not5](http://www.yeastgenome.org/locus/S000006276/overview) ([@bib18]). Since there is the potential for heterogeneity among [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not complexes in yeast and metazoans \[for review see [@bib68]\], it is possible that separate pools of [Not1](http://www.yeastgenome.org/locus/S000000689/overview) (with [Not5](http://www.yeastgenome.org/locus/S000006276/overview) and others) are dedicated to global decay and imprinting. mRNAs highly enriched with decay factors are less likely to have high ribosome occupancy {#s23} ---------------------------------------------------------------------------------------- A long-standing model for the fate of mRNAs is that they are partitioned into either the translatable or nontranslatable pool, with the latter category encompassing mRNAs that are stored in mRNPs or undergoing degradation ([@bib94]; [@bib76]). However, recently, the lines have become blurred. The processes of translation and decay is more intertwined than was initially thought, and some mRNA decay factors have also been implicated in regulating translation ([@bib20]; [@bib17]; [@bib77]; [@bib19]; [@bib41]). In addition, decay factors and 5′ to 3′ decay intermediates have been detected among polyribosomes ([@bib49]; [@bib91]; [@bib77]; [@bib78]), suggesting that RNAs undergoing translation can be subjected to decay. We found that decay factor recruitment negatively correlated with ribosome density. If deadenylation mostly occurred on polyribosomes, it would be expected that the recruitment of the major initiator of decay, [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), would show a positive correlation with ribosome density. This was not the case. Furthermore, the mRNA targets of the decay factors analyzed here correlated with those targets of multiple proteins involved in decay and translational repression, and not those that promote translation. The relationship between decay factor recruitment and ribosome occupancy described here, and the correlations between the deadenylase and other repressors of translation, supports a model that deadenylation and translation of transcripts are in competition. It is important to note that our results do not eliminate the existence of cotranslational decay pathways. Additional studies are required to fully understand the relationship between translation and decay. Targeting of Ccr4 and Dhh1 to nutrient-regulated transcripts suggests that mRNA decay and/or translational repression are important for cells to respond to nutrient availability and environmental conditions {#s24} -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- *[CCR4](http://www.yeastgenome.org/locus/S000000019/overview)* was initially identified as a regulator of *[ADH2](http://www.yeastgenome.org/locus/S000004918/overview)* and other nonfermentative growth genes ([@bib25]). This discovery provided the first evidence that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) plays an important role in the cell's ability to adjust to carbon sources. Genetic studies have revealed that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) are required for cells to survive nutrient starvation ([@bib101], [@bib102]; [@bib7]), and the [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) homolog from *Schizosaccharomyces pombe*, ste13+, regulates N~2~ starvation-induced entry into meiosis ([@bib63]). Furthermore, [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) have been tied to the TOR pathway ([@bib92]; [@bib55]). Thus, there is previous genetic evidence that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not is an important regulator of the nutrient response. Our work has added to these observations and provided additional insights into how [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) regulate gene expression in response to nutrient levels. [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) associate with mRNAs that are normally repressed in rich media and induced by nutrient starvation. Significantly, these mRNAs are under the tightest control of these factors at the level of decay (*i.e.*, highly enriched and stabilized by deletion of the gene) ([Figure 7, C and D](#fig7){ref-type="fig"}). This suggests a logical model whereby [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) are recruited to these mRNAs when nutrients are rich to repress their expression by mediating the decay of these messages. However, as stated in other parts of this manuscript, [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not regulates gene expression at the level of transcription, binds transcription factors, and is recruited to active genes. Therefore, its functions in the nutritional response are unlikely to be restricted to mRNA decay. However, analysis presented here, and that of others, presents a solid case that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) are important for the cell to control nutrient-regulated genes, and that this involves regulation of the decay of the messages. Typically, experiments are carried out under exaggerated conditions (*e.g.*, nutrient-rich *vs.* complete starvation), comparing two "static" conditions to amplify effects. Although contrasting nutrient-rich and -poor conditions is useful, understanding how yeast behave during the YMC has been suggested as a better way to capture the reprograming of metabolic pathways ([@bib12]). Thus, the more interesting result of our mapping study is the uncovering of the previously unknown relationship between [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) recruitment and the fluctuations of mRNAs during the YMC. We found that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) are highly recruited to mRNAs that are repressed as cells move from the nonrespiratory phase into the respiratory phase (increasing O~2~ consumption). As mentioned in other sections of this manuscript, [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not is targeted to specific mRNAs via RBPs. The [Puf3](http://www.yeastgenome.org/locus/S000003936/overview) motif is overrepresented in the [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-bound mRNAs ([Figure 3](#fig3){ref-type="fig"}) and also in those that fluctuate during the yeast YMC ([@bib96]; [@bib57]). [Puf3](http://www.yeastgenome.org/locus/S000003936/overview) immunoprecipitates with [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), suggesting that a physical interaction between [Puf3](http://www.yeastgenome.org/locus/S000003936/overview) and [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not is important for deadenylase recruitment ([@bib58]). Although [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) are bound to messages undergoing nutrient control, levels of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview), [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview), or [Puf5](http://www.yeastgenome.org/locus/S000003146/overview) proteins do not change under nutrient starvation conditions or when cells are grown in nonfermentable carbon sources (Figure S14A in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). Whether nutrient signals affect the recruitment of [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not to mRNAs or regulate its activity remains to be tested. [Not1](http://www.yeastgenome.org/locus/S000000689/overview), [Not4](http://www.yeastgenome.org/locus/S000000870/overview), and [Caf1](http://www.yeastgenome.org/locus/S000005335/overview) are phosphoproteins, and [Not4](http://www.yeastgenome.org/locus/S000000870/overview) becomes phosphorylated in stressed cells ([@bib56]) and [Caf1](http://www.yeastgenome.org/locus/S000005335/overview) undergoes phosphorylation in carbon-starved cells ([@bib72]). We noticed that the mobility of [Not4](http://www.yeastgenome.org/locus/S000000870/overview), a subunit of the complex, changed when cells were starved for carbon or grown in nonfermentable carbon sources (Figure S14B in [File S5](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1/FileS5.pdf)). [Puf3p](http://www.yeastgenome.org/locus/S000003936/overview) phosphorylation during the YMC and glucose starvation correlated with increased accumulation and decreased turnover of [Puf3](https://www.yeastgenome.org/locus/S000003936)-targeted mRNAs ([@bib96]; [@bib69]; [@bib57]). Taken together, these observations raise the intriguing possibility that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview)-Not, and/or the proteins that recruit it to messages, are regulated by the nutritional status of the cell. Testing this hypothesis is worthy of comprehensive future studies, which are well beyond the scope of the work reported here. The association of decay factors with specific messages in response to metabolic fluxes may be required to rapidly "clear" the cell of these mRNAs, allowing the cell to transition between phases of fermentation and respiration. In addition, recruitment of decay factors could sharpen peaks of gene expression to impart tighter temporal control over the YMC, similar to how changes in the decay and synthesis rates of cell cycle-regulated messages are important during the mitotic cell cycle ([@bib29]). Previous genetic studies indicate that [Ccr4](http://www.yeastgenome.org/locus/S000000019/overview) and [Dhh1](http://www.yeastgenome.org/locus/S000002319/overview) are required for cells to survive stresses that affect the cell cycle ([@bib101], [@bib102]; [@bib7]). Thus, the regulated targeting of decay factors to transcripts is likely to be important for cells to respond to and recover from stressful conditions. Supplementary Material {#s25} ====================== Supplemental material is available online at [www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1](http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.117.300415/-/DC1). ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. We thank the members of the Center for Eukaryotic Gene Regulation and the Center for RNA Molecular Biology for advice and support. SOLiD sequencing was performed at the Penn State Huck Institutes of the Life Science Genomics Core Facility at University Park. We acknowledge Craig Praul for advice on library preparation and DNA sequencing, Zhenhai Zhang for sharing a computational pipeline that the read alignment and composite plot code would be based upon, Björn Schwalb and Patrick Cramer for providing cDTA data, and Benjamin Tu for providing processed YMC data. This work was supported by National Institutes of Health grants GM-58672 (J.C.R.) and ES-013768 (B.F.P.). Author contributions: J.E.M. and J.C.R. conceived the study and cowrote the manuscript. J.E.M. and H.J. conducted the experiments. J.E.M. analyzed the data. Y.L., L.Z., and B.F.P. conducted read mapping and construction of part of the bioinformatics pipeline. Communicating editor: B. Gregory [^1]: Present address: Biomedical and Translational Informatics Institute, Geisinger Health System, Danville, PA 17822. [^2]: Present address: School of Life Science and Technology, 100 Haike Rd., Shanghai, 201210, China. [^3]: Present address: Life Technologies, 200 Oyster Point Blvd., South San Francisco, CA 94080.

Tao Jiang, Qingzhu Jia, Wenfeng Fang, and Shengxiang Ren contributed equally to this work. **Dear Editor,** Immune checkpoint inhibitors (ICIs) targeting cytotoxic T lymphocyte antigen‐4 (CTLA‐4), or programmed cell death protein 1 (PD‐1) and its ligand (PD‐L1) interaction achieve a significant improvement in overall survival (OS) and revolutionize treatment paradigms in many types of cancers.[^1^](#ctm2109-bib-0001){ref-type="ref"}, [^2^](#ctm2109-bib-0002){ref-type="ref"} Despite of the durable antitumor effect, ICIs could only benefit a minority of patients (∼20%) without effective predictive biomarkers.[^3^](#ctm2109-bib-0003){ref-type="ref"} Therefore, there is an urgent need to develop novel biomarkers for the majority of patients, who do not respond to ICIs monotherapy. Telomerase reverse transcriptase (*TERT*), as the catalytic subunit of telomerase, plays a critical role in modulating telomerase activity and immortalization of cancer cells.[^4^](#ctm2109-bib-0004){ref-type="ref"}, [^5^](#ctm2109-bib-0005){ref-type="ref"}, [^6^](#ctm2109-bib-0006){ref-type="ref"}, [^7^](#ctm2109-bib-0007){ref-type="ref"} It plays a crucial role in the tumorigenesis, cancer cell proliferation, invasion, and DNA damage response.[^4^](#ctm2109-bib-0004){ref-type="ref"}, [^7^](#ctm2109-bib-0007){ref-type="ref"} Recently, an elegant study found that short telomere length could cause a primary T cell immunodeficiency,[^8^](#ctm2109-bib-0008){ref-type="ref"} suggesting the contribution of telomere length to T cell apoptosis and function. Therefore, it is valuable to investigate the predictive value of *TERT* alterations for ICIs treatment outcome in multiple cancers. We identified a cohort of 43 910 cancer patients with 46 237 sequenced tumor samples, together with sequenced data and collected clinical information from the cBioPortal online database (<https://www.cbioportal.org>) (Figure S1). The prevalence of *TERT* alterations was 6.7%, with patients with thyroid cancer having the highest levels of *TERT* alterations (60.2%, 139/231; Figure S2). Most of the alterations were somatic mutations (73.1%, 2258/3091), especially promoter mutations (75.0%, 1693/2258). We then investigated the prevalence and spectrum of *TERT* alterations in early‐stage (TCGA cohort, N = 10 967) and advanced‐stage cancers (MSK‐IMPACT cohort, N = 10 945). The results showed that advanced‐stage cancers (15.2%, 1659/10 945) had significantly higher frequency of *TERT* alterations than early‐stage cancers (5.5%, 604/10 967; *P* \< .0001). Most detected *TERT* alterations were amplifications in early‐stage cohort (Figure S3A), whereas most were *TERT* somatic mutations including promoter mutations in advanced‐stage cancers (Figure S3B). In MSK‐IMPACT cohort,[^9^](#ctm2109-bib-0009){ref-type="ref"} we found that TMB of patients with *TERT* alterations was significantly higher than those without these alterations (17 vs 6 mut/Mb, *P* \< .0001; Figure S4A). This was validated in the ICI‐treated cohort that included 1661 patients (TMB of *TERT* alterations vs wild type: 20 vs 9 mut/Mb, *P* \< .0001; Figure S4B). Notably, cancers with multiple *TERT* alterations had the highest TMB level in both cohorts (Figures S4C and S4D). Next, we evaluated the association between *TERT* alterations and clinical outcome. We first found that patients with *TERT* alterations showed a significantly shorter OS (38 vs 113 months; HR = 1.90; 95% CI, 1.73‐2.09; *P* \< .0001; Figure [1A](#ctm2109-fig-0001){ref-type="fig"}) than those without in whole group. In the ICI treatment cohort,[^10^](#ctm2109-bib-0010){ref-type="ref"} we first identified 1661 patients with different cancers receiving ICI therapy and 521 of them with *TERT* alterations. Although clinicopathological features, including age, sex, sample type, and tumor purity, were not well balanced (Table S1), patients with *TERT* alterations had a substantially prolonged OS of 24 versus 17 months in the wild‐type group (HR = 0.78; 95% CI, 0.68‐0.91; *P* = .0016; Figure [1B](#ctm2109-fig-0001){ref-type="fig"}). Although *TERT* alterations were associated with higher level of TMB, multivariate analysis revealed that *TERT* alterations were associated with markedly longer OS than wild type independent of TMB (HR = 0.77; 95% CI, 0.67‐0.91; *P* = .0020; Table S2). Notably, we also observed the association between *TERT* alterations and prolonged OS in patients with microsatellite‐stable solid tumors (not reached vs 20 months; HR = 0.35; 95% CI, 0.14‐0.88; *P* = .0248; Figure S5). Subgroup analysis revealed that patients with *TERT* promoter mutations also had the better OS than those with *TERT* wild type (22 vs 17 months, *P* = .0225; Figure [1C](#ctm2109-fig-0001){ref-type="fig"}). Interestingly, in nonsmall‐cell lung cancer (NSCLC) treated with ICI, patients with *TERT* promoter mutations had the longest progression‐free survival (PFS) than other alterations and wild‐type groups (7.9 vs 2.6 vs 3.2 months; *P* = .0765; Figure [1D](#ctm2109-fig-0001){ref-type="fig"}). To validate the abovementioned predictive value of *TERT* subtypes in NSCLC, we independently identified 158 patients with advanced NSCLC who received ICI monotherapy from three medical centers. Clinicopathological parameters were well balanced between *TERT* alterations and wild‐type groups (Figure [1E](#ctm2109-fig-0001){ref-type="fig"}). Similarly, we found that patients with *TERT*‐altered NSCLC had significantly better objective response rate (35.7% vs 17.4%; *P* = .044) and prolonged PFS than wild type (8.8 vs 3.1 months; *P* = .0248; Figure [1F](#ctm2109-fig-0001){ref-type="fig"}). Moreover, patients with *TERT* promoter mutations had the longest PFS than other alterations and wild‐type groups (9.7 vs 6.3 vs 3.1 months; *P* = .0333; Figure [1G](#ctm2109-fig-0001){ref-type="fig"}). {#ctm2109-fig-0001} To depict the tumor immune microenvironment of *TERT*‐altered tumors, we surveyed the relationship between *TERT* alterations and eight common immune infiltrates including total CD3^+^ T cells, CD8^+^ T cells, cytotoxic lymphocytes, B lineage, NK cells, monocytic lineage, neutrophils, and myeloid dendritic cells across different cancer types (Figure [2](#ctm2109-fig-0002){ref-type="fig"}). The results showed that tumor‐infiltrating CD8^+^ T cells, especially cytotoxic lymphocytes, were generally more abundant in the *TERT*‐altered tumors compared with those in the wild‐type tumors across multiple cancer types (Figure [2A‐D](#ctm2109-fig-0002){ref-type="fig"}), whereas neutrophils were lower in *TERT*‐altered tumors than in wild type (Figures [2A](#ctm2109-fig-0002){ref-type="fig"} and [2I](#ctm2109-fig-0002){ref-type="fig"}). The results of signaling pathways in HALLMARK gene set collection showed that several signatures associated with antitumor immunity including DNA repair, unfolded protein response, E2F targets, cholesterol homeostasis, and so on were significantly higher in *TERT*‐altered tumors, whereas hallmarks associated with immune inhibitory function including hedgehog pathway, NOTCH signaling, transforming growth factor‐β (TGF‐β) signaling, angiogenesis, hypoxia, and so on were obviously higher in *TERT* wild‐type tumors (Figure [2J](#ctm2109-fig-0002){ref-type="fig"}). {#ctm2109-fig-0002} In summary, the current study first provides the evidence that *TERT* alterations were associated with enhanced tumor immunogenicity and inflamed antitumor immunity, which result in prolonged OS in cancer patients treated with ICIs. The predictive value of *TERT* alterations was independent of tumor mutational burden and microsatellite status, suggesting that *TERT* alterations could be considered as a potential pan‐cancer predictive biomarker for ICI treatment. For the future, we still need to investigate the exact molecular mechanism, and large‐scale, prospective studies are also warranted. CONFLICT OF INTEREST {#ctm2109-sec-0030} ==================== The authors declare no conflict of interest. Supporting information ====================== ###### Supporting Information ###### Click here for additional data file. This study was supported in part by grants from the National Natural Science Foundation of China (No. 81871865 and 81972167), the Backbone Program of Shanghai Pulmonary Hospital (No. FKGG1802), and the Shanghai Municipal Science and Technology Commission Medical Guidance Project (No. 17411969200). Tao Jiang and Caicun Zhou designed this study. Tao Jiang, Qingzhu Jia, Wenfeng Fang, Shengxiang Ren, Xiaoxia Chen, and Chunxia Su collected the clinical and sequenced data. Tao Jiang, Qingzhu Jia, and Shengxiang Ren performed the statistical analyses. Tao Jiang, Shengxaing Ren, Li Zhang, and Caicun Zhou drafted the manuscript. Li Zhang and Caicun Zhou provided critical comments and suggestions and revised the manuscript. All authors read and approved the final version of the manuscript. All the data and materials are available. All the authors consent for publication.

MNNG : N‐methyl‐N\'‐nitro‐N‐nitrosoguani‐dine SHR : spontaneously hypertensive rats WKY : Wistar Kyoto rats BrdU : bromodeoxyuridine

Background ========== Previously limited to case series, \[[@B1]-[@B4]\] information on the patient characteristics and course of human immunodeficiency virus or acquired immunodeficiency syndrome associated nephropathy (HIVAN) after the onset of end stage renal disease (ESRD) has been reported for the national population of US ESRD patients. \[[@B5],[@B6]\] Substantial improvements in the survival of dialysis patients with HIV infection have been noted after 1995,\[[@B6]\] and have been attributed to treatment with highly active antiretroviral therapy (HAART). \[[@B7]\] Despite these encouraging reports, the morbidity and mortality of these patients remains high compared with age-matched patients with ESRD from other causes. \[[@B6]\] Despite the improving information on the pharmacokinetics of these drugs in dialysis patients, a recent report suggested that only 58% of the 62 HIV-infected patients with ESRD were on antiretroviral therapy. \[[@B8]\] Use of angiotensin converting enzyme inhibitors (ACE) in patients with HIVAN is associated with delayed progression to ESRD. \[[@B9],[@B10]\] Some other reports have suggested that use of ACE in dialysis patients may improve survival. \[[@B11]\] It is possible the use of ACE might be particularly beneficial in HIV infected patients with ESRD particularly those with HIV associated nephropathy as heavy proteinuria often continues after the onset of dialysis \[[@B12]\]. Alternatively, risk of hyperkalemia from ACE in patients with HIVAN due to associated tubulointerstitial inflammation and renal tubular acidosis may increase mortality. \[[@B13]\] Although a recent single-center report found no benefit of ACE in dialysis patients with HIVAN \[[@B14]\], it is not clear whether this represents national experience. Recent anecdotal reports of remission of HIVAN with use of HAART are encouraging. However, the prevalence of patients with HIVAN who develop ESRD may not decrease, because of improvement in survival of these patients, allowing them to live long enough to reach ESRD. \[[@B15]-[@B17]\] Such patients may even be more difficult to manage, since the onset of HIVAN could be delayed until these patients develop resistance to the therapy. Therefore, HIVAN is likely to continue as a major cause of ESRD in young African-Americans. Better understanding of factors associated with poor outcomes in patients with HIVAN and ESRD would help in improving outcomes in these patients. More detailed information on the metabolic, cardiovascular and chronic kidney disease specific characteristics of HIVAN patients on dialysis, as well as the use and possible benefit of certain medications, especially ACE and antiretroviral drugs in this population, would supplement previous reports. Further, use of a national database would allow the comparison of HIVAN patients in a large at-risk population and allow comparison with single center studies that may reflect practices that may not be representative of the national as a whole. Therefore, we analyzed data from the standard analysis files of the 2000 United States Renal Data System (USRDS) Dialysis Morbidity and Mortality (DMMS) Wave 2 database. The primary objective of the study was to examine whether patients with HIVAN who present to chronic dialysis had important differences in clinical and laboratory parameters compared to patients with other causes of ESRD and to assess factors associated with better survival such as laboratory parameters and medication use. Methods ======= A historical cohort study of the USRDS DMMS Wave 2 was performed. Details on the inception, limitations, validity, variables and questionnaires used in the study are available online at the USRDS researcher\'s guide website, <http://www.usrds.org/research.htm>. This database has been used in many previous cross-sectional and longitudinal studies including one by our own institution. \[[@B18]\]. Briefly, DMMS WAVE 2 was a prospective cohort study of a random sample of 20% of all U.S. hemodialysis patients and virtually all peritoneal dialysis patients starting treatment in 1996 and early 1997. Because patients who started dialysis in 1997 comprised only a small proportion of the study population, we limited analysis to patients who started dialysis in 1996. Also, because outcomes such as dates of death and follow-up had to be merged with this study from other USRDS files, we did not consider it a true prospective cohort study for the purposes of this analysis. Characteristics of hemodialysis and peritoneal dialysis patients (abstracted from prospective surveys conducted specifically for DMMS WAVE 2) were matched and weighted to allow more appropriate comparisons between modalities. Baseline and follow-up data used in the study are shown in Table 1 \[see [Additional file: 1](#S1){ref-type="supplementary-material"}\]. All results are reported as N (%) or the mean ± the standard deviation. Patients with ESRD due to HIVAN were determined from the variable PDIS (primary cause of renal failure) from the file SAF.PATIENTS and merged with the DMMS WAVE 2 files using unique patient identifiers. Causes of renal failure selected as HIVAN were 0429A, 0429Z, 0439Z, or 0449Z (various codes for HIV or AIDS associated nephropathy), from the variable PDIS in the file SAF.PATIENTS, Core CD and were coded as 1 in statistical analysis, with all other patients coded as 0. AIDS associated nephropathy was listed as a diagnosis in the years studied and therefore that term is still used in the present study. In addition, a maximum of 15 medications prescribed to each patient at the study start date (day 60 of dialysis) were recorded. From this list, the use of beta-blockers (both cardioselective and non-selective), angiotensin-converting enzyme (ACE) inhibitors, calcium channel blockers (subcategorized as dihyropyridine and non-dihydropyridine), 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitors (statins), and aspirin was determined. Carvedilol, which has been independently associated with improved survival in certain populations, was approved for use by the FDA in February 1997, and was therefore not assessed. Blood pressure levels, systolic and diastolic, were obtained as the mean of three readings before and after dialysis, respectively. Of the study cohort, 2198 (65.1%) of patients with the variable HIV and 2164 (64.1%) with the variable AIDS, respectively, were either missing, unknown, or coded as \"cannot disclose,\" and these variables were therefore not used in analysis. Survival status was linked to the DMMS Wave 2 data from the 2000 USRDS Patients Standard Analysis File (SAF.PATIENTS) via unique patient identifiers assigned by the USRDS. The date and cause of death listed in a patient\'s SAF was obtained from a form submitted to the USRDS by the patient\'s nephrologist (form HCFA 2746). Descriptions of these files are available under the \"USRDS Researcher\'s Guide\" at <http://www.usrds.org/research.htm>. Patient survival status was complete through March 2000. Survival time was defined as the time from 90 days after the date of the first dialysis session until the date of death, censored for receipt of renal transplant, loss to followup or the end of the study period. For comparison, models that did not censor for the date of receipt of renal transplant were also used. Statistical analysis -------------------- Continuous variables are presented as mean ± standard deviation unless otherwise specified. Univariate analysis was performed with Chi-Square testing for categorical variables (Fisher exact test used for violations of Cochran\'s assumptions, meaning fewer than 5 cases per cell) and student\'s t-test for continuous variables (Mann Whitney test or logarithmic transformation was used for variables without a Gaussian distribution, verified by visual inspection and goodness of fit tests (Chi Square for linear trend). Variables with borderline significance (p value \< 0.1 in univariate analysis) and those thought likely to have a clinical relationship with HIVAN were entered into the multivariate models. Stepwise (forward step likelihood ratio method) Cox proportional hazards analysis was used to assess the association between baseline factors and HIVAN, independent of other factors. This method used the most computationally intensive method for calculating mean hazard ratios (approximations of relative risk) entering each variable into the model and removing those that were no longer significant after adjustment for all other variables. Log-log plots were inspected to verify the existence of proportional hazards. Estimated hazard ratios (HR) along with corresponding 95% confidence intervals (CI) and p values are reported for all regression covariates. The association of HIVAN with mortality was also tested by Cox regression analysis using adjusted hazard ratios (AHR). Hierarchically well-formed models were used in the assessment of interaction terms. Collinearity diagnostics were evaluated both from the output of the statistical procedures and by manual generation of models including only one of several variables with close relationships (for example, malnutrition, body mass index and serum albumin). To assess for possible confounding, logistic regression was also performed to assess for significant associations with a history of HIVAN using the same covariates as for Cox Regression as above. Multiple methods were used to assess goodness of fit with all variables, including Nagelkerke r^2^(an attempt to quantify the proportion of explained \"variation\" in the logistic regression model, although values for r^2^in logistic regression are much lower than for linear regression) and the c-statistic, which is the receiver operating curve for the cumulative predicted probability of the logistic regression model (0.5 being equivalent to chance and 1.0 equivalent to 100% agreement). Results ======= A total of 4065 patients were included in the DMMS Wave 2 cohort. Of these, 3621 patients had valid dates for starting dialysis in 1996. From this cohort, 3374 had sufficient information to calculate follow-up times. Thirty-six of these patients had ESRD as a result of HIVAN. Mean follow-up was 2.19 ± 1.14 (standard deviation) years. Characteristics of all the patients and patients with HIVAN are summarized in Table 1 (see Additional file 1). As shown, in this file factors positively associated with HIV in univariate analysis included male gender, African American race, younger age, lower body mass index, decreased comorbidity in general, increased rates of malnutrition and smoking, lower blood pressure and cholesterol, higher serum creatinine, lower serum albumin bicarbonate, and decreased use of aspirin and calcium channel blockers. The distribution of serum PTH levels in the study population was severely skewed, with a skewness of 2.49 and a kurtosis of 9.08. We elected to use non-parametric tests of association (the Mann-Whitney test) and a t-test of logarithmically transformed values of PTH. Although the difference in serum PTH levels between patients with HIVAN and patients with other causes of renal disease appeared large, it was not significant in Mann-Whitney testing (p = 0.34) or t-tests of logarithmically transformed values of serum PTH (p = 0.64). Specific antiretroviral agents are also shown in this file. The most common agents used in HIVAN patients were zidovudine and lamivudine. Only 22 (61%) of patients with HIVAN received antiretroviral agents, and only nine patients (25%) received combination antiretroviral therapy. Five patients received both zidovudine and lamivudine (a combination later called Combivir^TM^). Other combinations included one each for zidovudine-zalcitabine, stavudine-saquinavir, stavidine-lamivudine, and indinavir-lamivudine. Table [2](#T1){ref-type="table"} shows factors independently associated with HIVAN in logistic regression. Factors included antiretroviral therapy, male gender, African American race, younger age, lower serum albumin, higher rates of malnutrition and smoking, and lower use of calcium channel blockers. There were no significant differences in ACE use by either HIV or HIVAN status, either unadjusted or adjusted. There was no significant interaction between ACE and antiretroviral therapy. ###### Logistic Regression of Factors Associated with HIVAN, entire cohort **P value** **Adjusted odds ratio for HIVAN** **95% CI** --------------------------------------- ------------- ----------------------------------- ------------- Age \<48 (vs. \> = 71)\* 0.005 9.35 1.99, 43.48 Serum albumin (per higher quartile) \<0.001 0.31 0.15, 0.69 Male (vs. female) 0.001 6.58 2.39, 18.11 African American \<0.001 11.03 4.09, 27.78 CCB use (vs. other medications) 0.02 0.23 0.07, 0.75 Malnourished (vs. adequate nutrition) 0.001 4.09 1.56, 10.72 Smoking (vs. non active smoking) 0.02 2.89 1.16, 7.18 HIVAN=Human immunodeficiency virus associated nephropathy. \*Other age categories were not statistically significant when compared to age \> = 71 years Nagelkerke r^2^was 0.50, c-statistic was 0.97 (95% CI, 0.96, 0.99) Patients with HIVAN had significantly lower survival compared to patients with other causes of ESRD (Table [3](#T2){ref-type="table"}, Figure [1](#F1){ref-type="fig"}). Plots of survival by ACE use are shown for HIVAN patients (Figure [2](#F2){ref-type="fig"}) and for the entire study cohort (Figure [3](#F3){ref-type="fig"}). In Cox Regression analysis adjusted for other factors known to be associated with mortality, HIVAN was independently associated with increased risk of mortality, adjusted hazard ratio, 4.74, 95% CI, 3.12--7.32, p \< 0.0001. ACE use was not significantly associated with mortality in the entire cohort in Cox Regression, and it was not significant as an interaction term with HIVAN (p = 0.38, adjusted hazard ratio, 1.64, 95% CI, 0.55--4.87). In stratified analysis limited to patients with HIVAN, ACE use was not significantly associated with mortality (p = 0.48, AHR = 1.88, 95% CI, 0.33--10.83). Although the risk of death of patients with HIVAN and ESRD treated with single antiretroviral drug was higher than those not taking antiretroviral agents, patients on multiple antiretroviral drugs had a lower hazard ratio in adjusted analysis (0.62, 95% CI, 0.10--3.86 by Cox Regression), however this was not statistically significant (Figure [4](#F4){ref-type="fig"}). Among patients with HIVAN, the leading causes of death were AIDS (50%), missing/unknown (10.7%), and cardiac arrest of unknown cause (10.7%). Factors independently associated with survival in HIVAN patients are shown in Table [4](#T3){ref-type="table"}. The only factor independently associate with improved survival was African American race, compared to all other races. {#F1} {#F2} {#F3} {#F4} **Percent survival** **1-year** **2-year** **3-year** ------------------------------------------------------ ------------ ------------ ------------ HIVAN^A^ 53 31 28 Patients at risk 19 11 10 Dialysis Patients with other causes of renal failure 83 66 49 Patients at risk 2707 2180 1722 ^A^Unadjusted cumulative percent survival, censored for receipt of renal transplant HIVAN=ESRD due to HIV/AIDS associated nephropathy AIDS=acquired immunodeficiency syndrome ###### Cox Regression of Factors Associated with Mortality, DMMS WAVE 2 patients with HIVAN only One-Year Unadjusted Survival Factor Present One-Year Unadjusted Survival Factor Absent P Value AHR 95% CI ----------------------------------------------- --------------------------------------------- -------------------------------------------- --------- ------ ------------ African American Race\* (vs. all other races) 54% 40% 0.05 0.20 0.04, 0.99 Two-Year Unadjusted Survival Two-Year Unadjusted Survival Antiretroviral therapy (yes/No) 15% 39% 0.21 1.92 0.69, 5.37 Multiple antiretroviral agents (yes/No) 22% 33% 0.60 0.62 0.10, 3.86 ACE inhibitors (vs. nonuse) 13% 29% 0.28 2.26 0.55, 9.31 CCB (vs. nonuse) 42% 25% 0.36 0.56 0.16, 1.96 Male (vs. female) 33% 17% 0.31 0.44 0.09, 2.16 Peritoneal Dialysis (vs. Hemodialysis) 38% 26% 0.47 0.58 0.13, 2.53 N = 36, N in final model = 30 \*Because so few patients with HIVAN were not African American, only one-year survival could be compared Only factors significant in the final model are shown. Discussion ========== The present study found that only 61% of a national sample of US dialysis patients with HIVAN received antiretroviral therapy, consistent with previous regional reports of 58% use of anti-retroviral agents, \[[@B8]\] suggesting that underutilization of antiretroviral in dialysis patients with HIVAN is a national problem. HAART use, which the present study could not directly measure, is also low in dialysis patients with HIVAN, at 33% according to a recent single-center study \[[@B14]\]. We found no beneficial effect of ACE on survival in HIVAN, also consistent with previous studies \[[@B14]\]. The survival of HIVAN patients on dialysis in the preset study is also consistent with previous reports \[[@B5],[@B6]\]. One-year survival of patients with HIVAN was only 53% compared to 83% for patients with all other causes of ESRD. This survival is lower than reported recently by Ahuja et al. \[[@B6]\] in HIV infected patients starting dialysis in the United States after 1997, likely from the fact that patients included in the DMMS wave 2 were recruited in 1996. Although data on CD4 counts and viral loads were unavailable from the database (the importance of which is illustrated in two reviews.\[[@B19],[@B20]\] HIVAN is generally a late manifestation of HIV infection. \[[@B21]\] Therefore, use of antiretroviral therapy would be expected in patients with HIVAN in the absence of definite contraindications, even disregarding the previously cited reports of a possible beneficial effect of HAART on the course of HIVAN itself. However, in the present study only 61% of patients with HIVAN were receiving antiretroviral therapy and 41% of these were only on single antiretroviral therapy. Although not statistically significant, there was a trend towards better survival of patients on two antiretroviral drugs. None of the patients was on more then two antiretroviral drugs. This underutilization of antiretroviral therapy could be due hesitancy on the part of the infectious disease specialists and nephrologists for using these drugs due to unavailability of data on the pharmacokinetics of these drugs in patients on dialysis. Providers may also be concerned about the nephrotoxicity of certain antiretroviral agents, notably indinavir-associated crystal induced renal failure, \[[@B22]\] and other antiretroviral agents that require dose adjustment in renal failure. \[[@B22]\] Urgent studies are therefore needed to understand factors responsible for this underutilization of HAART in these patients. Similar underutilization of medications known to be beneficial in patients with chronic kidney disease has also been reported for cardiovascular disease. \[[@B18],[@B23]\], suggesting such practice may not be an isolated phenomenon. However, the long term tolerability and efficacy of HAART has not been studied in patients on dialysis. Although a benefit of ACE use in HIV infected patients has been reported previously, \[[@B9]\] the benefits of ACE in retarding progression of renal failure in patients with HIVAN did not translate into improvement of survival of these patients on dialysis. This is in agreement with some recommendations that ACE use, to be effective, must start early in the course of HIVAN. As the number of patients with HIVAN who were treated with ACE was small in the present cohort, further prospective studies are required to validate our initial observation. In comparison to patients with other causes of ESRD, the patients with HIVAN included in the DMMS wave 2 study were younger, the majority were African-Americans and had lower serum albumin and more severe acidosis. Some reports have suggested that HIV can infect parathyroid cells and these patients have low parathyroid hormone (PTH) and 1,25 dihydroxyvitamin D3 levels. \[[@B24]\] However, this did not prevent development of secondary hyperparathyroidism in patients with HIVAN and ESRD. The mean PTH level of the patients with HIVAN was 225 ± 42.6 pcg/L, although lower was not statistically different from patients with causes of ESRD. The present study has several limitations: it is observational and not randomized, therefore, confounding might persist despite adjustment. It is a random sample of all hemodialysis patients, not the total population, although the peritoneal dialysis population is almost fully captured. Information bias could have arisen due to mistakes in coding, especially for medications. We did not know the duration of use of medications prior to the study start. We were also unable to follow changes in variables over time. Therefore, we could not follow changes in blood pressure, laboratory values, or dialysis adequacy. This most especially applied to possible changes in medication use and changes in patient dry weight. These same limitations apply to other published studies using this database. In addition the information on CD4 count and plasma viral load in patients with HIVAN was not available and this could be an important uncorrected confounding factor. We could not assess factors associated with HIV seropositivity or AIDS due to the high percentage of patients with missing values for these variables. In summary, we conclude that patients with HIVAN have a lower survival compared to patients with ESRD from other causes. Antiretroviral therapy is underutilized and the use of ACE does not improve survival of these patients. Future prospective studies are required to determine efficacy and tolerability of HAART and ACE inhibitors in patients with HIVAN and ESRD. Competing interests =================== None declared. Authors\' contributions ======================= KCA performed the primary data analysis and collaborated in the writing of the manuscript. FCT derived the algorithms for the medication variables and collaborated in the writing of the manuscript. LYA provided the USRDS CD data files and supervised the project, and collaborated in the writing of the manuscript. TSA collaborated in the writing of the manuscript. Note ==== The opinions are solely those of the authors and do not represent an endorsement by the Department of Defense or the National Institutes of Health. This is a U.S. Government work. There are no restrictions on its use Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2369/4/5/prepub> Supplementary Material ====================== ###### Additional file 1 ###### Click here for file

The authors confirm that all data underlying the findings are fully available without restriction. Data are from the Atherosclerosis Risk in Communities (ARIC) Study, whose investigators can be contacted at <http://www.cscc.unc.edu/aric/ladindex4.php>. Introduction {#s1} ============ Atrial fibrillation (AF) is the most common cardiac arrhythmia in clinical practice, and its prevalence in the population is increasing [@pone.0110111-Majeed1]--[@pone.0110111-Naccarelli1]. Along with traditional risk factors, underlying structural and functional cardiac abnormalities such as chamber dilatation, diastolic dysfunction, and valvular problems are associated with higher AF risk [@pone.0110111-Tsang1]--[@pone.0110111-Heeringa1]. Although blacks have a higher prevalence of AF risk factors than whites [@pone.0110111-Sharp1]--[@pone.0110111-Berry1], they paradoxically have a lower risk for developing AF [@pone.0110111-Go1], [@pone.0110111-Alonso1]. While previous research has suggested that genetic factors might explain some of this racial disparity [@pone.0110111-Marcus1], the mechanisms by which these factors affect AF incidence still remain undetermined. Exploring AF risk factors in blacks is important given that the lower incidence of AF in this racial group despite higher risk factors is not completely explained. Understanding racial differences in risk factors for AF may help improve understanding of mechanisms and pathophysiology of the disease, prediction of AF in blacks, and prevention and treatment of AF in all populations. Echocardiography is a useful modality in evaluating underlying structural and functional heart disease, and some studies have shown differences in echocardiographic parameters between whites and blacks [@pone.0110111-Marcus2]. The link between AF and various echocardiographic markers of cardiac structure and function in whites has been previously assessed [@pone.0110111-Rosenberg1]--[@pone.0110111-Jons1]. To our knowledge, no study has examined this association in a large cohort of blacks with long term follow up for AF. Therefore, we examined standardized echocardiographic measures performed at the Jackson site of the Atherosclerosis Risk in Communities (ARIC) study. We hypothesized that echocardiographic measures of cardiac structure and function, namely left atrial (LA) diameter, left ventricular (LV) internal diameter, LV mass index, depressed LV ejection fraction (LVEF), percent fractional shortening of 2D left ventricular diameter (%FS), and early-to-late (E/A) diastolic filling velocity ratio, are independently associated with incident AF in blacks. Methods {#s2} ======= Study Population {#s2a} ---------------- The ARIC Study has been previously described [@pone.0110111-The1]. Briefly, the ARIC study is a population-based prospective study of cardiovascular disease (CVD) in a cohort of 15,792 individuals aged 45 to 64 years at enrollment (1987--1989) sampled from four US communities: Minneapolis, Minnesota; Washington County, Maryland; Forsyth County, North Carolina; and Jackson, Mississippi. Only blacks were recruited in Jackson. The baseline visit (visit 1) and three triennial visits - 1990--1992 (visit 2), 1993--1995 (visit 3), and 1996--1998 (visit 4), included interviews, laboratory measurements, and clinic examinations. Participants were also contacted annually by phone to obtain updated history. Echocardiograms were only performed at visit 3 (1993--1995) in the Jackson cohort (N = 2,622 black participants). Visit 3 was used as the baseline for these analyses. The ARIC study protocols were approved by the institutional review boards of each participating center, and informed consent was obtained from each study participant. Echocardiography {#s2b} ---------------- Details of the design and quality control measures for echocardiography have been described previously [@pone.0110111-Skelton1]. Echocardiographic parameters including 2D parasternal long axis LA diameter, LV internal diameter, LV posterior wall thickness (PWT), and interventricular septal thickness (IVST) were estimated at diastole. Relative wall thickness (RWT), defined as the ratio of the PWT multiplied by 2 and the LVID in diastole, was also calculated. The individual dimensions were used to estimate LV mass according to the American Society of Echocardiography (ASE) simplified cubed equation and were indexed by height^2^ to normalize heart size to body size [@pone.0110111-Devereux1]--[@pone.0110111-deSimone1]. This permitted categorization of individuals with increased LV mass index (\>95 g/m^2^ in females and \>115 g/m^2^ in males), as either concentric (RWT\>0.42) or eccentric (RWT≤0.42) hypertrophy. Individuals with normal LV mass index were categorized as either concentric remodeling (RWT\>0.42), or normal geometry (RWT≤0.42). Qualitative two-dimensional left ventricular systolic function (depressed {\<50%} and normal) and Doppler mitral valve inflow consisting of early (E) and late (A) peak velocity were measured. Given that regional wall motion abnormalities were unlikely to be present in this population-derived cohort, %FS was used to better evaluate the relationship between LV function and AF. Doppler mitral inflow velocity time integral (VTI) of E wave and A wave were also measured. Intra- and intersonographer and intrareader correlation were excellent as assessed on a randomly selected 2% sample of participants (for example, 0.94, 0.88 and 0.98 respectively, for LV mass). Ascertainment of Incident Atrial Fibrillation {#s2c} --------------------------------------------- Only incident AF events occurring from visit 3 through December 31, 2009 were considered. Individuals with AF identified on or before visit 3 were excluded (N = 19). The methods used for AF ascertainment have previously been described in detail [@pone.0110111-Alonso1], [@pone.0110111-Chamberlain1]. AF was identified through hospital discharge codes (N = 182), electrocardiograms (ECG) performed at visit 4 (N = 2), death certificates (N = 1), and both hospital discharge codes and death certificates (N = 6). AF was identified when *International Classification of Diseases*, ninth revision, clinical modification (ICD-9-CM) code 427.31 or 427.32, or ICD-10 code I48, was listed. Previous analyses within this population have validated the diagnosis of incident AF based on hospital discharges [@pone.0110111-Alonso1]. Standard supine 12-lead resting ECGs were recorded after a 12-hour fast followed by a light snack and at least one hour after smoking tobacco or ingestion of caffeine. All ECGs automatically coded as AF were visually re-checked by a trained cardiologist to confirm the diagnosis. Left ventricular hypertrophy (LVH) based on Cornell voltage criteria [@pone.0110111-Casale1], PR interval, and p wave indices [@pone.0110111-Macruz1] were also recorded. Measurement of other Covariates {#s2d} ------------------------------- At baseline (visit 3), weight, height, alcohol intake, cigarette smoking, blood pressure, and antihypertensive medication use were assessed using standardized methods [@pone.0110111-The1]. Blood levels of calcium, magnesium, phosphorus, and uric acid were measured at visit 2. Definition of covariates are stated in the supplemental material ([File S1](#pone.0110111.s001){ref-type="supplementary-material"}). Statistical Analysis {#s2e} -------------------- Of 2,622 blacks in the ARIC Jackson cohort at visit 3, we excluded 339 individuals (87 with missing or unreadable ECGs at baseline, 19 with prevalent AF, and 233 with missing covariates). The final sample included 2283 individuals. Due to technical limitations, different echocardiographic measurements were available for diverse sets of individuals, resulting in a variable sample size for each analysis ([**figure 1**](#pone-0110111-g001){ref-type="fig"}). {#pone-0110111-g001} Cumulative incidence of AF by categories or quintiles of echocardiographic parameters were estimated using Kaplan-Meier curves. To estimate the association of echocardiographic measures with time to incident AF, we calculated hazard ratios (HRs) and 95% confidence intervals (CIs) using Cox proportional hazards models. Person-years at risk were calculated from visit 3 until date of development of AF, death, loss to follow up, or end of follow up (December, 2009), whichever occurred first. Initially, the shape of the association between each continuous echocardiographic measure and AF was explored using age- and sex-adjusted restricted cubic spline models (Figure S1 in [File S1](#pone.0110111.s001){ref-type="supplementary-material"}). In order not to assume linearity, LA diameter, LVID, %FS and LV mass index were modeled as quintiles. Individuals were dichotomized into two groups for LVEF: depressed (\<50%) or normal. Mitral early-to-late (E/A) diastolic filling velocity ratio was divided into 3 categories (\<0.7, 0.7--1.5, and \>1.5), with 0.7--1.5 used as the reference point for all analyses. In an initial model, we adjusted for age and sex. A second model additionally adjusted for the recently described CHARGE-AF risk score [@pone.0110111-Alonso2], which consists of known AF risk factors including age, height, weight, current smoking, systolic blood pressure, diastolic blood pressure, anti-hypertensive medications, diabetes, prevalent HF, and prevalent MI; and estimates cumulative risk of AF over 5 years. The proportional hazards assumption was tested using time interaction terms and inspection of log-negative log survival curves. The assumption was satisfied for all parameters except the E/A ratio. To address this issue, follow-up for the E/A ratio analysis was split at the midpoint into two time periods (0--8 years and \>8 years to the end of follow-up) at the midpoint of average follow up time, and HRs were calculated for each period. Interactions by sex and HTN (a major risk factor for AF) with these echocardiographic parameters were examined. Interaction between LVH (by voltage criteria on ECG) and LV diameter; and between LV mass index and RWT was also examined. Furthermore, we determined whether echocardiographic measures (modeled as continuous variables) improve the predictive ability of the CHARGE AF risk score measuring improvement in C-statistic for survival analysis over ten years [@pone.0110111-Pencina1]. To determine whether individuals were reclassified correctly rather than due to chance, we calculated the Net Reclassification Improvement (NRI), using \<5.0%, 5.0%--10.0%, \>10.0% risk as cutoff points over ten years [@pone.0110111-Pencina2]. To avoid associations resulting from reverse causation (AF leading to changes in echocardiographic parameters) and to avoid including possible undetected prevalent AF cases as new incident cases, we conducted a sensitivity analysis excluding cases of AF occurring in the first 2 years after echocardiographic measurements. Finally, in secondary analyses, we added ECG measures of increased LA size (defined as P wave duration ≥120 ms) and ECG-LVH (Cornell voltage criteria) to models with echocardiography measurements to determine whether ECG measures added predictive value. The risk of AF based on the presence of concentric hypertrophy, eccentric hypertrophy, or concentric remodeling compared to normal geometry was determined. We further adjusted covariate analyses models for PR interval, calcium, magnesium, phosphorus, uric acid and alcohol use. Analyses were performed using SAS, version 9.2; SAS Institute Inc. All p values were 2-sided. Results {#s3} ======= [**Table 1**](#pone-0110111-t001){ref-type="table"} shows selected clinical, echocardiographic and electrocardiographic measures of ARIC Jackson cohort participants at visit 3 by incident AF status. In this population, individuals who developed AF were more likely to be older males, to have higher CHARGE-AF risk scores, larger LA, heavier LV, and more systolic or diastolic dysfunction. They were also more likely to have LVH based on Cornell voltage criteria and LA enlargement by P wave indices. 10.1371/journal.pone.0110111.t001 ###### Baseline characteristics with incident atrial fibrillation (AF) events through 2009 follow-up, Atherosclerosis Risk in Communities Study Jackson Cohort, 1993--1995. {#pone-0110111-t001-1} No AF during follow-up Incident AF --------------------------------------------------------------------- ------- ------------------------ ------------- ---------- Clinical Measures[\*](#nt101){ref-type="table-fn"} **Age, years** 59±5.6 61±5.7 \<0.0001 **Sex, male** 729 (34.9) 82 (42.9) 0.03 **Body Mass Index, kg/m^2^** 30.4±6.3 32.1±6.8 0.0004 **Current Smoking** 399 (19.1) 54 (28.3) 0.002 **High School Graduate** 578 (27.6) 61 (31.9) \<0.0001 **High-Density Cholesterol, mg/dL** 57±18.5 50±15.6 \<0.0001 **Systolic Blood Pressure, mmHg** 130±19.9 136±22.2 \<0.0001 **Diastolic Blood Pressure, mmHg** 77±10.4 76±12.4 0.17 **Anti-Hypertensive Medications** 1041 (49.8) 138 (72.3) \<0.0001 **Digoxin** 30 (1.4) 11 (5.8) \<0.0001 **Prevalent Diabetes** 462 (22.1) 79 (41.4) \<0.0001 **Prevalent Heart Failure** 124 (5.9) 27 (14.1) \<0.0001 **Prevalent Coronary Heart Disease** 84 (4.0) 19 (10.0) 0.0002 **Current alcohol use** 622 (29.7) 63 (33) 0.15 **CHARGE-AF Risk Score, % (5 years)** 1.5±1.6 2.9±2.4 \<0.0001 **Serum Calcium, mg/dL** 9.5 (0.5) 9.5 (0.4) 0.92 **Serum Magnesium, mEq/L** 1.6 (0.2) 1.6 (0.2) 0.82 **Serum Phosphorus, mmol/L** 4.2 (0.4) 4.1 (0.5) 0.04 **Uric Acid, mg/dL** 6.6 (1.7) 7.2 (2.0) \<0.0001 Echocardiographic Measures[\*](#nt101){ref-type="table-fn"} **Left Atrial Diameter, cm** 3.3±0.5 3.6±0.6 \<0.0001 **Left Ventricular Diameter (Diastole), cm** 4.4±0.6 4.4±0.8 0.07 **Left Ventricular Mass Index, g/m^2^** 84.4±24.8 99.4±38.2 \<0.0001 **Depressed Left Ventricular Ejection Fraction (EF \<50%)** 39 (1.9) 13 (6.8) \<0.0001 **Mitral Early-to-Late (E/A) Diastolic Filling Velocity Ratio** \<0.0001 \<0.7 154 (7.4) 28 (14.7) \>1.5 125 (6.0) 17 (8.9) **% Fractional Shortening of Left Ventricular Diameter** 34.6±8.9 33.4±10.5 0.10 **Relative Wall Thickness** 0.56 (0.13) 0.61 (0.16) \<0.0001 **Electrocardiographic Measures** [\*](#nt101){ref-type="table-fn"} **Left Atrial Enlargement, P Wave Index** 569 (27.8) 69 (37.1) 0.01 **Left Ventricular Hypertrophy, Cornell Voltage Criteria** 119 (5.8) 23 (12.3) 0.001 **PR Interval, msec** 172 (28) 171 (40) 0.69 **\* Data given as mean±SD or n (%).** During a total 30,826 person-years of follow-up (mean follow-up 13.5 years), we identified 191 incident cases of AF. After adjustment for age and sex, longer LA diameter was associated with an increased risk of AF in a J-shaped pattern (Figure S1 in [File S1](#pone.0110111.s001){ref-type="supplementary-material"}). Compared to those in the 1^st^ quintile, the HR (95% CI) of AF for those in the 5^th^ quintile was 2.46 (1.51--4.02), after adjustment for the CHARGE-AF score. Similarly, LV mass index was linearly associated with AF risk: HR (95% CI) comparing the 5^th^ to the 1^st^ quintile was 2.81 (1.61--4.91) after adjustment for the CHARGE-AF score ([**Table 2**](#pone-0110111-t002){ref-type="table"}). There was no significant interaction between LV mass index and RWT (p = 0.54). On the other hand, LV diameter showed a U-shaped relationship with incident AF (Figure S1 in [File S1](#pone.0110111.s001){ref-type="supplementary-material"}), with increased risk among individuals with small and those with large diameter. Risk was lowest at 2^nd^ quintile (HR 0.49, 95% CI 0.29--0.83 compared to the 1^st^ quintile) ([**Table 2**](#pone-0110111-t002){ref-type="table"} **, Model 2**). %FS also showed a U-shaped relationship with incident AF (Figure S1 in [File S1](#pone.0110111.s001){ref-type="supplementary-material"}), with risk lowest among individuals in the third quintile (32.41--36.55) of %FS. Reduced LVEF was associated with increased risk of AF (HR 3.46, 95% CI 1.88--6.36; [**Table 2**](#pone-0110111-t002){ref-type="table"} **, Model 2**). Finally, diastolic dysfunction based on an E/A ratio \<0.7 and \>1.5 was also associated with increased AF risk compared to an E/A ratio 0.7--1.5, though the magnitude of the association differed over the entire follow-up ([**Table 2**](#pone-0110111-t002){ref-type="table"}). Additionally adjusting for educational level, HDL cholesterol and digoxin use and incident HF and MI as time-dependent covariates did not change the results (Table S1 in [File S1](#pone.0110111.s001){ref-type="supplementary-material"}). Analyzing hazard ratios for AF by sex-specific quintiles did not significantly change the results (Table S2 in [File S1](#pone.0110111.s001){ref-type="supplementary-material"}). Kaplan-Meier curves showing cumulative incidence of AF by quintiles and categories of echocardiographic parameters are shown in [**figure 2**](#pone-0110111-g002){ref-type="fig"}. Due to a small number of AF cases, quintiles were divided into 3 groups in the Kaplan-Meier curves. {#pone-0110111-g002} 10.1371/journal.pone.0110111.t002 ###### Hazard ratios (HR) and 95% confidence intervals (CI) for the association of echocardiographic parameters in quintiles with time to incident atrial fibrillation (AF), ARIC Jackson Cohort, 1993--2009. {#pone-0110111-t002-2} Parameter AF Cases \| N Group/Quintile Model 1[†](#nt103){ref-type="table-fn"} HR (95% CI) Model 2[‡](#nt104){ref-type="table-fn"} HR (95% CI) ------------------------------------------------------------------------- --------------- -------------------- ----------------------------------------------------- ----------------------------------------------------- Left Atrial Diameter, cm 23 \| 431 1.54--2.93 Reference Reference 28 \| 432 \>2.93--3.21 1.21 (0.70--2.10) 1.14 (0.66--1.98) 26 \| 432 \>3.21--3.46 1.11 (0.63--1.94) 1.05 (0.60--1.85) 40 \| 431 \>3.46--3.78 1.76 (1.05--2.94) 1.62 (0.97--2.71) 65 \| 431 \>3.78--5.18 3.06 (1.90--4.93) 2.46 (1.51--4.02) **P for Trend** [\*](#nt102){ref-type="table-fn"} \<0.0001 \<0.0001 Left Ventricular Mass Index, g/m^2^ 17 \| 402 27.91--65.03 Reference Reference 26 \| 403 65.04--75.77 1.42 (0.77--2.61) 1.40 (0.76--2.58) 30 \| 403 75.78--86.96 1.70 (0.94--3.08) 1.66 (0.91--3.00) 33 \| 403 86.97--102.86 1.87 (1.04--3.36) 1.70 (0.95--3.06) 55 \| 402 102.87--316.60 3.52 (2.04--6.08) 2.81 (1.61--4.91) **P for Trend** [\*](#nt102){ref-type="table-fn"} \<0.0001 0.0001 Left Ventricular Diameter (Diastole), cm 39 \| 402 2.30--3.87 Reference Reference 21 \| 403 \>3.87--4.21 0.50 (0.29--0.85) 0.49 (0.29--0.83) 28 \| 403 \>4.21--4.48 0.72 (0.44--1.17) 0.68 (0.42--1.11) 34 \| 403 \>4.48--4.81 0.80 (0.50--1.28) 0.72 (0.45--1.15) 39 \| 403 \>4.81--7.19 1.03 (0.65--1.62) 0.78 (0.48--1.24) **P for Trend** [\*](#nt102){ref-type="table-fn"} 0.42 0.72 \% Fractional Shortening of Left Ventricular Diameter 44 \| 400 1.51--27.46 Reference Reference 32 \| 401 27.47--32.40 0.69 (0.44--1.08) 0.79 (0.50--1.25) 23 \| 401 32.41--36.55 0.45 (0.27--0.75) 0.50 (0.30--0.83) 29 \| 401 36.56--41.75 0.62 (0.38--0.99) 0.71 (0.44--1.15) 33 \| 401 41.76--78.21 0.65 (0.41--1.02) 0.72 (0.45--1.15) **P for Trend** [\*](#nt102){ref-type="table-fn"} 0.06 0.14 Left Ventricular Ejection Fraction (LVEF) 175 \| 2223 *Normal (≥50%)* Reference Reference 13 \| 52 *Low LVEF (\<50%)* 5.16 (2.91--9.14) 3.46 (1.88--6.36) Mitral Early-to-Late (E/A) Diastolic Filling Velocity Ratio: 0--8 years 8 \| 182 \<0.7 1.37 (0.64--2.93) 1.09 (0.50--2.37) 45 \| 1775 0.7--1.5 Reference Reference 10 \| 142 \>1.5 3.56 (1.78--7.14) 3.59 (1.79--7.20) \>8 years to the end of follow-up 20 \| 141 \<0.7 2.81 (1.69--4.67) 2.47 (1.47--4.15) 73 \| 1603 0.7--1.5 Reference Reference 7 \| 120 \>1.5 1.51 (0.69--3.31) 1.51 (0.69--3.30) **\*** Linear trend in quintile number. Model 1 - adjusted for age and sex. Model 2 -- adjusted for Model 1 + CHARGE risk score. After adjustment for the CHARGE-AF risk score, LA enlargement by P wave index on ECG was associated with a 35% increased risk for incident AF {HR (95% CI) = 1.35 (1.00--1.82)}. Similarly, LVH using Cornell voltage criteria was also associated with increased AF risk {HR (95% CI) = 2.28 (1.47--3.53)} (Tables S3 and S4 in [File S1](#pone.0110111.s001){ref-type="supplementary-material"}). For all models examined, there was no significant evidence of interaction by HTN or sex. There was no significant interaction between LVH (Cornell voltage criteria) and LV diameter (P = 0.48). In a sensitivity analysis, the above models were further adjusted for phosphorus, magnesium, uric acid, calcium, alcohol use and PR interval on ECG both individually and all in the same model. The results were unchanged from those from the above models (data not shown). Addition of individual echocardiographic variables to the CHARGE AF score resulted in a modest, non-statistically significant improvement in the C-statistic for prediction of AF over a 10-year follow-up ([**Table 3**](#pone-0110111-t003){ref-type="table"}). NRI for adding LA diameter to the CHARGE-AF risk score was 0.12 (p = 0.01); other echocardiographic variables did not improve reclassification over ten years ([**Table 3**](#pone-0110111-t003){ref-type="table"}). Addition of ECG variables to models containing similar echocardiographic variables did not significantly improve C statistic or NRI (Tables S5 and S6 in [File S1](#pone.0110111.s001){ref-type="supplementary-material"}). 10.1371/journal.pone.0110111.t003 ###### C-Statistic and net reclassification index (NRI) for the association of echocardiographic parameters (as continuous variables) with time to incident atrial fibrillation, ARIC Jackson Cohort, 1993--2009. {#pone-0110111-t003-3} ECHO variable C-statistic (95% CI) without ECHO variable C-statistic (95% CI) with ECHO variable NRI (0.05,0.10) P-value ------------------------------------------------------------- -------------------------------------------- ----------------------------------------- ----------------- --------- Left Atrial Diameter (cm) 0.718 (0.671--0.766) 0.741 (0.691--0.791) 0.12 0.01 Left Ventricular Diameter (Diastole) (cm) 0.740 (0.693--0.788) 0.750 (0.700--0.799) 0.03 0.41 Left Ventricular Mass Index (g/m^2^) 0.740 (0.693--0.787) 0.752 (0.706--0.799) 0.04 0.31 Depressed Left Ventricular Ejection Fraction (LVEF \<50%) 0.726 (0.681--0.772) 0.739 (0.694--0.785) 0.03 0.36 Mitral Early-to-Late (E/A) Diastolic Filling Velocity Ratio 0.733 (0.683--0.783) 0.735 (0.685--0.785) 0.003 0.90 \% Fractional Shortening of Left Ventricular Diameter 0.740 (0.692--0.787) 0.750 (0.705--0.795) 0.03 0.44 All Echo variables: n  =  1773; AF cases  =  72 0.744 (0.691--0.797) 0.767 (0.714--0.819) When we examined all the echocardiographic variables in a single model, C-statistic improved from 0.744 (0.691--0.797) to 0.767 (0.714--0.819). Addition of both echocardiographic and ECG variables in the same model produced a similar result: C-statistic 0.768 (0.715--0.821). Sample size for these analyses were smaller (1773 individuals with 72 AF cases) as these were the only individuals who had all the echocardiographic and ECG measures available. In a sensitivity analysis excluding AF cases identified during the first 2 years of follow-up, results were unchanged (data not shown). Finally, we examined four RWT-LV mass index groups to assess their relationship to AF. After adjustment for age, sex and the CHARGE-AF risk score, concentric and eccentric hypertrophy were associated with increased AF risk, compared with normal geometry, HR 3.17 (1.35--7.48) and 4.87 (1.37--17.35) respectively. There was no increased risk of AF with concentric remodeling, HR 1.72 (0.75--3.93). Discussion {#s4} ========== This population-based prospective study of blacks aged 51--70 years, found that echocardiographic measures of cardiac structure and function are associated with an increased risk of AF, independent of sex, lifestyles, clinical measures and chronic diseases. We found that the association of some echocardiographic measures (LA diameter, LV diameter, and %FS) with AF risk is complex and non-linear. Specifically, LA diameter showed a J-shaped association, with AF risk markedly increasing in the top quintile (\>3.78--5.18 cm). Several studies have shown that LA diameter is a risk factor for AF [@pone.0110111-Marcus2], [@pone.0110111-Psaty1], [@pone.0110111-Vaziri1]. It has also been shown that LA diameter is smaller in blacks than whites, and that this difference could contribute to lower AF risk in blacks [@pone.0110111-Marcus2], [@pone.0110111-Manolio1]. Marcus et al. [@pone.0110111-Marcus2] found that mean LA diameter was 4.26 cm in blacks vs. 4.33 cm in whites (p = 0.0026). Two prior reports using data from the Framingham and Cardiovascular Health studies have demonstrated a link between LA diameter and incident AF mostly in whites [@pone.0110111-Psaty1]--[@pone.0110111-Vaziri1]; however, non-linear associations were not evaluated in these reports. Rosenberg et al. [@pone.0110111-Rosenberg1] evaluated non-linear associations between diastolic parameters and AF also in whites, but found only a linear association between LA diameter and AF risk. Only one prior study has compared echocardiographic parameters and risk of AF in blacks and whites [@pone.0110111-Marcus2], however they only included 476 blacks with 6 AF cases. Since our study is the first to report this J-shaped relationship between LA diameter and AF risk, it remains to be seen if this finding is particular to blacks. In a similar way, in the present study, LV diameter and %FS both showed a U-shaped relationship with incident AF, with risk of AF lowest at the midpoint of LV diameter and %FS 32.41--36.55. The increased risk with smaller LV diameter may be explained by the effect of hypertensive heart disease leading to a thicker left ventricle and hence smaller internal diameter. In our analyses, there was however no effect modification of LVH (Cornell voltage criteria) on LV diameter. Both concentric hypertrophy (which can be caused by pressure overload from hypertension), and eccentric hypertrophy (caused by volume overload) were associated with increased AF risk. %FS is a surrogate of cardiac contractility and although it is dependent on preload, it is a long established method of estimating LV function [@pone.0110111-deSimone2]. It only fails in individuals with regional wall motion abnormalities which is not expected to be present in this population-derived cohort. That there was a lower risk of AF in patients with %FS 32.41--36.55, compared to those with lower %FS is not surprising, as low LVEF (another marker of reduction in myocardial function) was associated with lower AF risk. Increased catecholamine release is associated with AF [@pone.0110111-Linsay1]. In high catecholamine states, there may be hyperdynamic cardiac function, and hence %FS may be increased. This might explain the link between increased %FS and AF; hence the U shaped relationship between %FS and AF. Also, LV mass index (a marker of LVH), diastolic dysfunction measured by E/A ratio \<0.7 and \>1.5, and reduced LV systolic function were associated with increased risk of AF [@pone.0110111-Rosenberg1], [@pone.0110111-Kannel1], [@pone.0110111-Benjamin1]. These measures (except LV diameter), both individually and together in a single model improve risk stratification and discriminative ability over traditional risk factors for AF in blacks, although this improvement was not statistically significant. Addition of ECG measures did not significantly add to risk stratification. Although these measures are associated with risk of AF in blacks, the associations between these measures and AF in blacks are similar to those in whites [@pone.0110111-Marcus2], [@pone.0110111-Rosenberg1], [@pone.0110111-Psaty1], [@pone.0110111-Vaziri1]; hence these measures alone are unlikely to explain the racial differences found in AF. The complex mechanisms underlying development of AF remain incompletely understood. AF is thought to result from presence of multiple wave fronts conducted to the ventricles from the atria and requires triggers [@pone.0110111-Wyse1]. The regular impulses produced by the sinus node to provide rhythmic contraction of the heart are overwhelmed by these rapid randomly generated multiple wave fronts [@pone.0110111-Goette1]--[@pone.0110111-Chang1]. There have been a few mechanisms suggested by which echocardiographic markers of cardiac structure and function may increase risk of developing AF. Increased LA size has been thought to increase AF risk as a result of stretch of the atrial appendage which leads to remodelling of the anatomy and physiology of the left atrium and increases dispersion of atrial refractoriness [@pone.0110111-Vaziri1], [@pone.0110111-Tsang2], [@pone.0110111-Wang1]. Impairment of LV end-diastolic filling pressure and LA emptying caused by LV size, systolic and diastolic dysfunction can also lead to development of AF by worsening atrial structural remodeling. AF itself reduces atrial contractility, contributing to atrial dilatation and more remodeling, thus contributing to the progressive and perpetuating nature of this arrhythmia [@pone.0110111-Allessie1]. HTN, CHD, age, inflammation and other risk factors act as triggers and propagating factors directly influencing these measures of cardiac structure and function increasing AF risk. For example, age- and inflammation-related fibrosis can cause increased atrial size, and HTN and CHD lead to increased LV mass and cardiomyopathy. However, a positive association of echocardiographic measures with AF persisted even after we accounted for these traditional risk factors. Our study has some limitations. First, we did not have volumetric measures of atrial size. The limitations of assessment of atrial size in M-mode or 2-D are widely recognized, and it is accepted that volumetric measures are better. Second, markers of HF (NT-proBNP), thyroid dysfunction (TSH) and renal failure (eGFR), which are recognized risk factors for AF, were not available. Hence, we could not adjust for them. However, we do not expect that adjusting for these factors would have changed our overall findings, since accounting for HF and other traditional risk factors for AF did not completely explain the associations we found. Third, most of our AF cases are identified due to hospitalizations, which may not account for AF managed in outpatient settings. Also, as AF is often transient, asymptomatic patients may have been missed. Therefore, there is likely an under ascertainment of incident AF cases. However, AF cases through hospitalizations in ARIC have been validated [@pone.0110111-Alonso1], [@pone.0110111-Psaty1]. Fourth, since echocardiograms were performed between 1993 and 1995, measures needed to properly classify diastolic function by current ASE guidelines were not available. Prior studies, however, have shown that E/A ratio of \<0.7 and \>1.5 compared to 0.7-1.5 is a marker for diastolic dysfunction and risk factor for CVD [@pone.0110111-Fox1]--[@pone.0110111-Gardin1]. Also, while individuals with E/A ratio\>1.5 could potentially have normal filling, most in this category have demonstrated increased CVD risk [@pone.0110111-Fox1]--[@pone.0110111-Gardin1]. Since not all echocardiographic measures were available for all individuals in the analysis, we were unable to examine the overall relationship with AF and all measures in a single model. Lastly, whites did not have echocardiograms performed at visit 3, so we do not have direct racial comparisons. Despite its limitations, this study is novel and has important strengths, including the relatively large sample size, extended follow-up, and availability of information on potential confounders. Conclusions {#s4a} ----------- We have shown that echocardiographic measures of cardiac structure and function in blacks are associated with an increased risk of AF independent of traditional risk factors. Although these measures individually and in a single model modestly improve risk stratification and discriminative ability of AF in this population, the only statistically significant improvement was for LA diameter. Addition of ECG measures did not significantly add to risk stratification. Further research is needed to determine if blacks at increased risk based on echocardiographic measures can be targeted with therapies to attempt to reduce incidence of AF. Lastly, large prospective studies with direct racial comparisons of associations of imaging modalities with AF are warranted. Supporting Information {#s5} ====================== ###### **Supporting information.** Tables, Figures, and description of terms in methods section of paper. (DOCX) ###### Click here for additional data file. The authors thank the staff and participants of the ARIC study for their important contributions. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: WB AA JRM SK SDS EZS LRL FLL ERF THM. Performed the experiments: WB AA JRM. Analyzed the data: WB AA JRM. Contributed reagents/materials/analysis tools: WB AA JRM SK SDS EZS LRL FLL ERF THM. Wrote the paper: WB AA JRM SK SDS EZS LRL FLL ERF THM.

1. Introduction {#sec1-ijerph-16-02499} =============== Biodegradation of wood and polymeric materials involves a series of biological and chemical processes, resulting in the destruction of polymer chains and wood structures when exposed to fungal hyphae. During this process, different types of fungi produce a unique grouping of microbial volatile organic compounds (MVOCs), which can also be used for species identification. This phenomenon indicates the possibility of using MVOCs as detection markers for wood decay fungus \[[@B1-ijerph-16-02499]\]. There are two common methods for the collection of MVOCs. The first entails a collection of a whole air sample in a stainless-steel sampling vessel, such as a canister or mini-can, using passive diffusion devices. The second method involves passing whole air through an air sampler on a solid sorbent inside a glass tube. The absorbent tube method is markedly more sensitive than the solvent extraction technique, while analysis of desorbed compounds using TD-GC/MS is the most commonly used method for detecting, identifying, and quantifying MVOCs. Moreover, this method is very sensitive in complex environmental matrices where MVOC levels are low \[[@B2-ijerph-16-02499],[@B3-ijerph-16-02499]\]. The *Basidiomycetes* class is one of the most potent degraders of cellulose. These indoor, wood-destroying fungi are commonly found in European buildings \[[@B4-ijerph-16-02499]\]. The development of fungi causes changes in the chemical, physical and mechanical properties of wood or wood-based materials, which has a negative impact on their technical properties, especially durability. Resistance of wood-polymer composite (WPC) materials to an attack by *Basidiomycetes* determines the ENV 12038:2002 standard \[[@B5-ijerph-16-02499]\]. Degradation of the cell-wall structure in wood occurs under the influence of enzymes secreted by the hyphae of mycelium: cellulase and ligninase. The metabolism and MVOCs produced by wood-rotting *Basidiomycetes* \[[@B6-ijerph-16-02499],[@B7-ijerph-16-02499],[@B8-ijerph-16-02499]\] have been investigated extensively. However, there is a lack of research focused on profiling the MVOCs produced by *Basidiomycetes* colonization of WPC. Composite materials are a combination of two or more basic constituent materials, which have substantially different chemical or physical properties \[[@B9-ijerph-16-02499]\]. In WPC, polymers form a continuously binding matrix that surrounds the wood components. Matrix polymers are materials with a high molecular mass, such as polyethylene (PE), polypropylene (PP), polyvinyl chloride (PVC), polystyrene (PS) and polylactide (PLA), which possess desirable physical properties, such as viscoelasticity or thermoplastic properties. Wood fillers perform a reinforcing role thanks to their strength and hardness \[[@B10-ijerph-16-02499],[@B11-ijerph-16-02499]\]. Kim and Pal \[[@B12-ijerph-16-02499]\] indicate two main reasons for adding wood to polymers: to lower the price of the final product and to reduce dependency on mineral oil-based products. In addition, modification of wood fibers significantly reduces water absorption capacity, which is important for resistance to biological agents. The most widespread indoor applications of WPCs are railings, molding, trim, window and door frames, sheathings, elements of roof and floor construction, and indoor furniture. The most common outdoor uses are deck floors, landscaping timbers, cladding and siding, fences and park benches \[[@B13-ijerph-16-02499],[@B14-ijerph-16-02499],[@B15-ijerph-16-02499],[@B16-ijerph-16-02499]\]. Several hundred volatile organic compounds have been detected in residential indoor air. When a residential area is infested with microorganisms, their metabolic processes produce microbial volatile organic compounds (MVOCs). These microbial metabolites are released during growth of fungi, mold and other microbes. The diversity of MVOCs is large, estimated at over a thousand different compounds, which are a complex mixture of organic compounds, including fatty acids and their derivatives, such as hydrocarbons, aliphatic alcohols, aldehydes and ketones, aromatic compounds or terpenes and terpene derivatives \[[@B17-ijerph-16-02499]\]. Garcia et al. \[[@B2-ijerph-16-02499]\] summarized the most common MVOCs reported for indoor and outdoor environments. The data presented in this work show that MVOC concentrations are higher in indoor environments because they remain closed and ventilation rates are lower compared to outdoors. In turn, Korpi et al. \[[@B18-ijerph-16-02499]\] have found that single MVOC levels in indoor environments range from a few ng/m^3^ up to 1 mg/m^3^. MVOCs are products of the primary and secondary metabolism of organisms, such as fungi and bacteria. In primary metabolism, organisms create MVOCs as by-products, breaking down food in the environment to synthesize DNA, and amino and fatty acids used to maintain cell structures. The number of primary metabolites is relatively small, and their biosynthesis is well known. Secondary metabolism covers chemical compounds that may be necessary in order to survive in the environment and cope with various hazards. Secondary metabolites perform adaptive functions, playing a key role in communication between fungi and the environment and they are unique to individual species. They do not perform basic biological functions and many of them have a complex and branched biosynthetic pathway \[[@B18-ijerph-16-02499]\]. In other words, secondary metabolism consists of subsequent downstream reactions, while the production of secondary metabolites is often species-specific or restricted to a phylogenetic group \[[@B19-ijerph-16-02499],[@B20-ijerph-16-02499]\]. A World Health Organization (WHO) report on biological agents \[[@B21-ijerph-16-02499]\] and other sources \[[@B22-ijerph-16-02499],[@B23-ijerph-16-02499],[@B24-ijerph-16-02499],[@B25-ijerph-16-02499],[@B26-ijerph-16-02499]\] describes the results of research designed to identify which MVOC arise from the fungal and bacterial growth on the surfaces of various building materials and microbiological media. The growth of fungi on building materials generates emissions of both reactive and nonreactive volatile organic compounds. These emissions may contribute to "sick building syndrome" (SBS) \[[@B27-ijerph-16-02499],[@B28-ijerph-16-02499]\]. SBS is a situation in which building occupants experience acute health and comfort effects, such as irritation of the eyes, nose and throat. It has been shown that some compounds associated with SBS are related to those produced in the presence of *Basiciomecytes* growing on WPC. In this study, MVOCs have been measured in order to identify microbial growth in building construction in the case of WPC, as described in the following sections. 2. Materials and Methods {#sec2-ijerph-16-02499} ======================== Two separate composite boards per fungi species were placed on a sterile malt-extract agar medium in each Kolle flask (experimental chamber), and two sterilized samples in other Kolle flasks served as controls (control chambers). The experiment was carried out simultaneously using two experimental chambers for each fungal species. Only the MVOCs emitted in both experimental emission chambers for the same mold strain are reported in this study. 2.1. Experimental Design {#sec2dot1-ijerph-16-02499} ------------------------ To obtain emission profiles for the MVOCs, an experimental set-up was constructed ([Figure 1](#ijerph-16-02499-f001){ref-type="fig"}). Analyses were performed using the aspiration air sampling method. MVOCs were collected on adsorbent glass tubes filled with Tenax TA resin by air suction using an aspirator. The sample loop contained a rotameter to measure airflow. This was then transferred back into the growth chambers, thus forming a closed sample system. 2.2. Choice of Fungi Species {#sec2dot2-ijerph-16-02499} ---------------------------- Fungi from the *Basidiomycetes* class are the most dangerous factor in the destruction of wooden elements in objects. The emission profiles of MVOCs at various stages of WPC decay have been demonstrated in this study for *Coniophora puteana* and *Poria placenta*, which the European industrial standards for wood-decay fungi are based upon. Moreover, there is lack of literature relating to MVOC profiles produced from *Basidiomycetes* colonization of WPC. 2.3. Fungal Strains, Growth Conditions and Culture {#sec2dot3-ijerph-16-02499} -------------------------------------------------- The wood-decay fungi strains used in this study were brown-rot fungus (Coniophora puteana (Schumacher ex Fries) Karsten (BAM Ebw. 15)) and white-rot fungus (Poria placenta (Fries) Cooke sensu J. Eriksson (FPRL 280)). They were separately cultured in Kolle flasks on malt extract agar (MEA) which consisting of 25 g/L agar and 40 g/L malt extract (Zakład Enzymów i Peptonów BTL). Each bottle had a volume of 250 mL. Pilot studies were performed on different WPC types; however, it was difficult to detect differences between these, because emissions were conditioned mainly by fungal metabolites. Finally, WPCs with high-density polypropylene (HDPP) matrices were chosen as the cultivation substrate. The composite materials consisted of 51% wood fibers and 49% HDPP. WPCs were cut into 80 mm × 40 mm × 6 mm cubes and put into Kolle flasks ([Figure 2](#ijerph-16-02499-f002){ref-type="fig"}). MVOC emissions from the uncontaminated fungal composites on the agar medium blanks (in the control chambers) were used as background mass spectra. The Kolle flasks were incubated for four weeks in a dark incubator at a temperature of 22 ± 2 °C and humidity of 70 ± 5%. The Kolle flasks were sealed tightly with cotton wool to enable exchange of inside and outside air. During incubation, MVOCs were taken from the flasks (adsorption tubes were inserted next to the cotton wool) after one day, two weeks and one month. Experimental conditions were maintained at a constant level throughout the whole incubation period. 2.4. Sampling and Analysis {#sec2dot4-ijerph-16-02499} -------------------------- Thermal desorption tubes are the most common sampling technique used for environmental MVOCs; these tubes allow detection of directly desorbed compounds into the GC/MS and avoids sample preparation. Air samples from the incubation chambers were adsorbed in Tenax TA, which is preferred because of its efficiency in trapping low concentrations of organics from air samples and because it is chemically inert, thermally stable and has good storage stability \[[@B3-ijerph-16-02499],[@B29-ijerph-16-02499]\]. The Tenax desorption tube technique requires a short sampling time and low sample volume, thereby obtaining high sensitivity \[[@B2-ijerph-16-02499]\]. Moreover, Tenax TA has a low affinity for water, which makes it ideal for use in humid environments \[[@B7-ijerph-16-02499]\]. The air was collected over a period of 8 min at an air flow rate of 2 L/h (volume of the Kolle flask). 2.5. Thermal Desorption-Gas Chromatography/Mass Spectrometry (TD-GC/MS) {#sec2dot5-ijerph-16-02499} ----------------------------------------------------------------------- The separation process and these analyses of volatile compounds were achieved using a gas chromatograph equipped with a gas chromatography mass spectrometer (model GCMS-QP2010, Shimadzu, Tokyo Japan). The following GC oven temperature methodology was applied: an initial temperature of 40 °C for 5 min, then 10 °C increases per minute up to 260 °C, with a final temperature of 260 °C for 1 min. The transfer line temperature was held at 250 °C, whereas the source temperature was kept at 220 °C. Chromatographic separation was performed on a ZB-5MSi capillary column (30 m, 0.25 mm ID, 1 µm df), which favors identification of nonpolar compounds. The 1:10 split- ratio injection mode was applied. Volatile organic compounds were thermally desorbed using a thermal desorption (TD 20, Shimadzu) cold-trap injector (cryo-focused at −18 °C). The tubes were heated at a temperature of 280 °C. Adsorbed compounds were then released into the flow of helium carrier gas to the GC/MS, which was used for MVOC identification and quantification. The GC/MS analysis system includes an odor database (Smart Database; Shimadzu Corporation) with parameters and sensory information (such as types of odors and odor threshold values) for the primary compounds that cause odors. 2.6. Calculation Method for MVOC Concentrations {#sec2dot6-ijerph-16-02499} ----------------------------------------------- GC/MS software has the ability to subtract spectra. If the background spectra (from the control chamber) are subtracted from the spectra under investigation (experimental chamber), clearer spectra can be obtained. This can be done automatically or manually. For the data presented in this study, the automatic method was applied. Ions present in the background spectrum have been reviewed for possible background contamination or coeluting peaks. Taking into account the assumptions of the experiment, there are some circumstances in which this gives the best approximation. 2.7. Reference Compounds {#sec2dot7-ijerph-16-02499} ------------------------ Volatile compounds were identified by comparing the retention times of chromatographic peaks with those of reference compounds and by searching the NIST 2014 mass spectral database. All volatile compounds with mass spectra match factors *p* ≥ 90% were considered identified. The reference standards were analyzed in the same conditions as the experimental tubes. The reference compounds used to confirm the identity of the metabolites were as follows: alcohols (propanol, 2-hexanol and 2-heptanol); aldehydes (propionaldehyde, hexanal and 1-heptanal); carboxylic acid (propionic acid, hexanoic acid and heptanoic acid); ketones (acetone, 2-hexanone, 2-heptanone) and toluene. A solution of 200 µg/mL of each compound was produced by gravimetric measurement and dissolving the individual substances in methanol (CPAchem Ltd., Stara Zagora, Bulgaria). Standard solutions were prepared by injecting the compounds into the inert gas flow in the tube; the spiked adsorbent tubes were then thermally desorbed in the same conditions as the samples \[[@B30-ijerph-16-02499],[@B31-ijerph-16-02499]\]. Compound-specific response factors were used to estimate the MVOC concentrations, determined by a certificated standard solution of 200 µg/mL foreach compound (CPAchem Ltd.). Identified compounds were quantified using their individual response factors when the reference compound was available. Concentrations of MVOCs other than the reference compounds were estimated by referring to compounds belonging to the same homologous series. 3. Results {#sec3-ijerph-16-02499} ========== 3.1. Identification of MVOCs {#sec3dot1-ijerph-16-02499} ---------------------------- Mass spectra of the MVOCs were obtained using spectrum subtraction. The analysed mass spectra were obtained by subtracting background mass spectra (control chamber) from the spectra under investigation (experimental chamber) ([Figure 3](#ijerph-16-02499-f003){ref-type="fig"}). [Table 1](#ijerph-16-02499-t001){ref-type="table"} contains the indicative concentration of MVOCs released from the culture of Coniophora puteana and Poria placenta growing on the WPC and agar-based medium. MVOCs are divided due to their chemical nature (alcohols, aldehydes, carboxylic acids, terpenes, ketones and sulphur compound). The concentrations measured in this study are equilibrium values obtained at samplingpoints, attributed by the exchange of outside and inside air in the Kolle flasks. 3.2. Identified Odor Compounds {#sec3dot2-ijerph-16-02499} ------------------------------ In accordance with the WHO report \[[@B21-ijerph-16-02499]\], there is a strong probability that the intensity of the odor emitted by the MVOCs (e.g., an earthy, musty, fruity or mushroom-like smell) is responsible for the general smell of the building. Many worldwide committees are working on unifying and adopting the level of odor intensity emitted by indoor MVOCs as an indicator of air quality perceptible by humans \[[@B21-ijerph-16-02499]\]. The musty-type mold odor of experimental cultures cannot be thus based on the presence of a single substance but rather on the coexistence of several relevant compounds. Many MVOCs have distinctive odors, perceptible by the human sense of smell. Excessive humidity supports their growth and, consequently, smell. Odors derived from the compounds identified in this study are tabularized in [Table 2](#ijerph-16-02499-t002){ref-type="table"}. 4. Discussion {#sec4-ijerph-16-02499} ============= 4.1. Analysis of MVOC Emission Profiles {#sec4dot1-ijerph-16-02499} --------------------------------------- The emissions of MVOCs change in accordance with the fungi type and their growth phase. Emissions also depend on existent environmental conditions, such as moisture, temperature and substrate composition \[[@B26-ijerph-16-02499]\]. In a nutrient-rich environment, the microbe produces a particular MVOC profile, which corresponds to stress-free growth, but in stressed environmental conditions (with competition or limited nutrients), a different MVOC profile may be produced, which enables adaptation to changing resource availability \[[@B19-ijerph-16-02499]\]. Fungi produce a number of different MVOCs including alcohols, aldehydes, carboxylic acids, terpenes, ketones and sulphur compounds. These compounds have been reported as fungal second metabolites in numerous former studies \[[@B32-ijerph-16-02499],[@B33-ijerph-16-02499],[@B34-ijerph-16-02499]\]. In each case, after one day of incubation, 1-octen-3-ol was detected in the range from 50 to 100 µg/m^3^. This unsaturated secondary alcohol gives a distinctive earthy mushroom odor (see [Table 2](#ijerph-16-02499-t002){ref-type="table"}) and is one of the most common compounds detected in damp houses suffering from fungal growth \[[@B35-ijerph-16-02499],[@B36-ijerph-16-02499]\]. It is a precursor to the formation of 1,3-octadiene. Korpi et al. \[[@B18-ijerph-16-02499]\] indicated that microbial air contamination of an indoor environment can be identified when 1-octen-3-ol reaches or exceeds the threshold value, which is equal to 10 µg/m^3^. Some alcohols such as 3-methyl-1-butanol, 4-methyl-1-pentanol and 1-octen-3-ol, can be emitted when fungi grow on media containing carbohydrates. Higher carbon alcohols (1-decanol or 5-hexadecanol) are produced by fatty acid degradation and reduction of amino acids. After only a month of incubation, 4-methyl-1-pentanol and 1-hexanol were identified. Ketones are produced by fatty acid degradation of, e.g., linoleic acid \[[@B37-ijerph-16-02499]\]. Lower carbon ketone (e.g., acetone) is a product of fundamental biochemical processes such, as the Krebs cycle and glycolysis \[[@B38-ijerph-16-02499]\]. In turn, unsaturated fatty acids may be transformed into aldehydes, such as octanal, nonanal and decanal, which were found in the largest quantities after two weeks of incubation. Glutaraldehyde was only observed as a species-specific compound for *Poria placenta*. Phytochemicals (like oxime-methoxy-phenyl or epoxy-linalooloxide) are chemical compounds formed during the metabolic process. These chemicals are often referred to as secondary metabolites. They are generally biologically active in the fungi hosts and play a role in their growth or defense against competitors or pathogens. Carboxylic acids are both substrates and products in the metabolism of MVOCs. Nonanoic acid, n-decanoic acid and tetradecanoic acid are the most abundant acids identified, especially for *Poria placenta* growth. Dimethyl disulphide is produced from the degradation of the sulphur-containing amino acids derived from the malt-extract agar medium \[[@B38-ijerph-16-02499]\]. Many other metabolites and metabolic pathways exist, but a detailed description of these is not the main objective of this thesis. The highest amount of ΣMVOC was observed two weeks after incubation for both fungi species: 1045 μg/m^3^ in the case *Coniophora puteana* and 1686 μg/m^3^ for *Poria placenta*. The sum total of MVOCs emitted from the Kolle flask in which *Poria placenta* was located was equal to 3461 μg/m^3^ ([Figure 4](#ijerph-16-02499-f004){ref-type="fig"}). In turn, the sum total of MVOCs emitted from *Coniophora puteana* was equal to 2139 μg/m^3^. Analysis of the data shows that of all the constituents, aldehydes (up to 50%) and carboxylic acid (up to 41%) were the predominantly identified groups of MVOCs ([Figure 4](#ijerph-16-02499-f004){ref-type="fig"}). Nonanal and decanoic acids are compounds present at higher concentrations. Moreover, 0.005% of dimethyl disulphide was identified for *Poria placenta* (not marked on the respective diagram). 4.2. Species-Specific MVOCs {#sec4dot2-ijerph-16-02499} --------------------------- Production of MVOCs mainly depends on the growth medium, although many compounds are produced by different species (species-specific compounds) and types of fungi. Significant differences were observed between the compounds produced during growth on paper, building materials and microbiological media \[[@B23-ijerph-16-02499],[@B24-ijerph-16-02499],[@B25-ijerph-16-02499],[@B26-ijerph-16-02499]\]. They can develop on organic materials (wood, wallpaper and cardboard), as well as polymeric (expanded polystyrene, mineral wool, plastic film and artificial leather) and mixed materials (carpeting and WPC). As a result of the described research, 2-ethyl-4-methyl-1-pentanol, 5-hexadecanol, 3-isopropyl-2-phenyl-pent-4-en-2-ol, p-cymene, 2-ethylbutanal, undecanal, 2-ethylhexanoic acid, octadecanoic acid and 1-methyl-2-pyrrolidinone were found to be *Coniophora puteana* species-specific MVOCs. In turn, glutaraldehyde, 1-decanol, 3-hydroxydodecanoic acid, oxime-methoxy-phenyl, 5-methyl-3-hexanol, 3-isopropyl-2-phenyl-pent-4-en-2-ol, 6-methyl-5-hepten-2-one, benzophenone and dimethyl disulphide were found to be *Poria placenta* species-specific MVOCs. 4.3. MVOCs Associated with SBS {#sec4dot3-ijerph-16-02499} ------------------------------ Damp in residential buildings is a risk factor for the onset of general symptoms and mucosal symptoms. Associations between MVOC and health concerns have been investigated in studies of several countries in Europe and Asia \[[@B39-ijerph-16-02499],[@B40-ijerph-16-02499],[@B41-ijerph-16-02499],[@B42-ijerph-16-02499]\]. In the prevalent study, any SBS symptom was associated with some individual volatile organic compounds of possible microbial origin (MVOC) e.g., 2-pentanol, 2-hexanon, 2-pentylfuran, 1-octen-3-ol and 3-methylofuran \[[@B43-ijerph-16-02499]\]. The compounds associated with SBS are the same as those found to be associated with the presence of *Basiciomecytes* growing on WPC. The main risk factor for SBS seems to be 1-octen-3-ol. Secondary indicators are 1-hexanol, 2-octen-1-ol and/or dimethyl disulphide, but this finding needs to be collaborated by further investigations. 5. Conclusions {#sec5-ijerph-16-02499} ============== The mass spectra of fungal metabolites were obtained by subtracting background (control chamber) spectra from the spectra under investigation (experimental chamber). The authors have demonstrated that the experimental set-up and data analysis method (using subtracted spectra) are capable of measuring MVOCs. The study of metabolite production resulting from cultures growing on different building materials is of great importance in order to provide information on which metabolites may be expected from WPCs affected by microbial growth. Despite the results obtained, we cannot definitively state the extent to which wood-decay fungi colonize WPC because not all the necessary data were available (e.g., the mass loss rate of the samples and strength parameters). The emissions of MVOCs change with the growth phase. The largest amount and concentrations of secondary metabolites were identified by both fungi species after two weeks of incubation. The concentration of MVOCs in building materials usually falls within the range 1--50 µg/m^3^. For *Poria placenta*, the most predominant metabolic volatile group identified was carboxylic acids, while aldehydes were the dominant group for *Coniophora puteana*. Terpenes, alcohols and ketones shared similar proportions, and α-pinene, 1-octen-3-ol and acetophenone are the compounds found at higher levels, respectively. Moreover, many compounds were found to overlap for both species, suggesting dependence on the malt-extract agar substrate. The main indicator of fungal exposure in indoor environments seems to be 1-octen-3-ol, which has also been used as a marker for active growth or damp building materials. Some MVOCs, such as 1-hexanol, 2-octen-1-ol and/or dimethyl disulphide, seem to be positively associated with any SBS symptoms, but epidemiological studies on exposure to MVOC are needed. Conceptualization, M.K.; methodology, M.K., A.W., A.A.; investigation, M.K., A.W., A.A.; writing---original draft preparation, M.K.; writing---review and editing, M.P.; supervision, M.P.; project administration, M.K. This research received no external funding. The authors declare no conflict of interest. {#ijerph-16-02499-f001} {#ijerph-16-02499-f002} {#ijerph-16-02499-f003} {#ijerph-16-02499-f004} ijerph-16-02499-t001_Table 1 ###### Overview showing the range of MVOC levels emitted by *Coniophora puteana* and *Poria placenta* after one day, two weeks and one month of incubation. *Coniophora puteana* *Poria placenta* ------------------------------------- ---------------------- ------------------ ----- ----- ----- ---- Alcohols 4-Methyl-1-pentanol \+ \+ 1-Hexanol ++ ++ 2-Ethyl-4-methyl-1-pentanol \+ 3-Methyl-1-butanol \+ \+ \+ \+ 1-Octen-3-ol ++ ++ 2-Octen-1-ol \+ \+ \+ \+ 5-Methyl-3-hexanol \+ 3-Isopropyl-2-phenyl-pent-4-en-2-ol \+ \+ 1-Decanol ++ 2-Phenoxy-ethanol \+ ++ 5-Hexadecanol \+ Aldehydes Hexanal \+ \+ \+ \+ \+ 2-Ethyl-butanal \+ Heptanal \+ \+ \+ ++ \+ Octanal \+ ++ \+ \+ ++ Glutaraldehyde \+ \+ \+ Nonanal \+ +++ \+ \+ +++ Decanal ++ +++ +++ Undecanal \+ \+ Dodecanal \+ \+ 13-Methyltetradecanal \+ Carboxylic acids Pentanoic acid \+ \+ \+ Hexanoic acid \+ \+ 3-Hydroxydodecanoic acid \+ 2-Ethylhexanoic acid \+ Nonanoic acid \+ \+ \+ ++ ++ 3-Hydroxydodecanoic acid \+ n-Decanoic acid \+ \+ ++ +++ +++ Dodecanoic acid ++ ++ Tridecanoic acid \+ \+ Tetradecanoic acid \+ +++ +++ \+ Octadecanoic acid ++ Terpenes alpha-Pinene \+ +++ +++ \+ ++ ++ Vanillin \+ \+ 3-Carene \+ ++ \+ \+ \+ \+ p-Cymene \+ M-Pyrol ++ Epoxy-linalooloxide \+ \+ Oxime-methoxy-phenyl \+ Ketones Acetone \+ \+ \+ \+ \+ \+ 3-Methyl-2-cyclopenten-1-one \+ \+ 6-Methyl-5-hepten-2-one \+ \+ 1-Methyl-2-pyrrolidinone \+ 6,10-Dimethyl-5,9-undecadien-2-one ++ ++ Benzophenone \+ Acetophenone \+ +++ ++ \+ Sulphur compound Dimethyl disulphide \+ Indicative concentrations of MVOCs (µg/m^3^) are shown as (+) for concentrations below 50 µg/m^3^ (++) for concentrations between 50--100 µg/m^3^ and (+++) for those above 100 µg/m^3^. ijerph-16-02499-t002_Table 2 ###### List of MVOCs with types of odors for some identified compounds \[[@B8-ijerph-16-02499],[@B22-ijerph-16-02499]\]. CAS No. Compound Name Odor ------------ --------------------- --------------------------- 66-25-1 Hexanal fat, tallow, grass 124-13-0 Octanal green, fat, soap, lemon 112-31-2 Decanal soap, tallow, orange peel 112-54-9 Dodecanal fat, citrus, lily 109-52-4 Pentanoic acid sweet 142-62-1 Hexanoic acid sweat, fatty, cheesy 3391-86-4 1-Octen-3-ol earthy, "mushroomy" 22104-78-5 2-Octen-1-ol green lemon, melon 123-51-3 3-Methyl-1-butanol truffle 112-05-0 Nonanoic acid green, fat 334-48-5 Decanoid acid fat, rancid 143-07-7 Dodecanoid acid metal 80-56-8 Alpha-Pinene solvent 121-33-5 Vanillin vanilla 119-61-9 Benzophenone almond, burnt sugar 98-86-2 Acetophenone flower, musty, almond 624-92-0 Dimethyl disulphide cabbage, onion, putrid

INTRODUCTION {#rjy135s5} ============ Thrombophilia is a hypercoagulable state that predisposes to thrombosis \[[@rjy135C1]\]. In addition to the well-established acquired risk factors for venous thromboembolic events, several genetic risk factors such as factor V Leiden (FV Leiden) (G1691A), prothrombin gene (factor II) (PTH) (G20210A) and MethyleneTetrahydrofolate Reductase (*MTHFR*) (C677T) mutations have been shown to predispose to thromboembolic events \[[@rjy135C1]\]. Patients with one or more of these mutations are at risk of developing venous thromboembolism (VTE), including deep vein thrombosis (DVT) and pulmonary embolism (PE), or developing adverse pregnancy outcomes including recurrent abortions, severe pre-eclampsia, intrauterine growth restriction (IUGR) and premature delivery \[[@rjy135C2]\]. Bariatric surgery patients are at elevated risk for thromboembolic events because of underlying venous stasis, immobility, increased intra-abdominal pressure and the risk of surgery itself \[[@rjy135C2]\]. Studies have shown that PE accounts for as many as half of all deaths after Roun-en-Y Gastric bypass surgery \[[@rjy135C3]\]. We report a case of a 56-year-old male patient presenting with simultaneous multiple thromboembolic events and was found to have three major thrombophilia gene co-mutations in FV Leiden, PTH and methylenetetrahydrofolate reductase genes. CASE PRESENTATION {#rjy135s6} ================= A 56-year-old male patient underwent sleeve gastrectomy one year prior to presentation, presented to our institution with left upper abdominal pain of few days duration. There was no history of chest pain or dyspnea, there were no focal neurological symptoms. On physical examination, the patient had left upper quadrant abdominal tenderness, no rebound tenderness or guarding. The patient had no known previous history of thromboembolic events. Laboratory values on presentation were within normal limits, coagulation profile: INR = 1.4, aPTT = 35.5 s, PT = 16.6 s, Anithrombin III = 96%, protein S = 104%, potein C = 65%. A computed tomography scan of the chest, abdomen and pelvis revealed splenic vein thrombosis (Fig. [1](#rjy135F1){ref-type="fig"}); and a filling defect in the lower lobar segmental branch of the right pulmonary artery consistent with acute segmental pulmonary embolism. A venous duplex scan of the extremities showed thrombosis of the left axillary and basilic veins (Figs [2](#rjy135F2){ref-type="fig"} and [3](#rjy135F3){ref-type="fig"}). Echocardiogram was normal. {#rjy135F1} {#rjy135F2} {#rjy135F3} The presence of multiple simultaneous thrombotic events in this patient triggered molecular work-up: heterozygous for Factor V Leiden (G1691A) mutation, Factor II (G20210A) mutation and MTHFR (C677T) mutation. The patient was started on therapeutic dose of subcutaneous low molecular weight heparin (LMWH), hospitalized for 4 days, discharged home on LMWH bridged to oral acenocoumarol, on which he was maintained for 6 months. DISCUSSION {#rjy135s7} ========== Genetic profiling for patients with thromboembolic events is frequently performed \[[@rjy135C2]\]. Factor V Leiden is the most common predisposing factor, accounting for 40--50% of the cases. The prothrombin (Factor II) gene mutation, deficiencies in protein S, protein C and antithrombin account for most of the remaining cases. In Lebanon, the prevalence of FV Leiden is 14% among healthy individuals and 40% among patients with DVT \[[@rjy135C4]\]. Double heterozygosity for FV Leiden and factor II mutation (G20210A) is the most common genetic combination with an increased risk of thrombosis \[[@rjy135C2]\]. This increased risk of thrombosis could be explained by the synergistic effect between the two mutations. Triple mutations involving these three genes remains a rare entity, as is the case with our patient herein reported. A study conducted on Omani population showed no FV Leiden and prothrombin gene mutations in any subject. However, MTHFR and CBS 844ins68 along with protein C deficiency were important risk factors. More importantly, 56.4% of these cases had either one or a combination of these three mutations \[[@rjy135C5]\]. Multiple studies have evaluated the role of FV Leiden in the risk of VTE. Approximately 10--26% of patients with VTE are carriers of the factor V Leiden mutation \[[@rjy135C5]\]. Heterozygosity for FV Leiden mutation increases the risk of venous thrombosis 5--10 times, while with homozygosity for the mutation there is an 80--100 times increased risk for venous thrombosis \[[@rjy135C6]\]. Prothrombin gene mutation G20210A causes elevated levels of Prothrombin (factor II) and thus increased risk for venous thrombosis \[[@rjy135C7]\]. Patients with *MTHFR* gene mutation may develop more than one thrombophilic gene mutation resulting in a higher risk of thrombosis than those with a single gene mutation \[[@rjy135C8]\]. In addition, carriers of the G20210A prothrombin gene mutation have an increased risk of deep vein, hepatic, portal, mesenteric, cerebral vein thrombosis and an increased risk to pulmonary emboli \[[@rjy135C9]\]. Bariatric surgery candidates have high rates of acquired as well as inherited thrombophilias \[[@rjy135C3]\]. Data are still not clear on whether to proceed with preoperative VTE screening through laboratory studies and imaging, in the setting of improved perioperative prophylaxis for thromboembolism in all bariatric surgery patients \[[@rjy135C3]\]. Data are still not clear on whether to proceed with preoperative VTE screening through laboratory studies and imaging, in the setting of improved perioperative prophylaxis for thromboembolism in all bariatric surgery patients. For instance, several studies have employed profiling of VTE risk through laboratory testing for thrombophilias in all patients, through studying D-Dimer, Fibrinogen, Factor VIII, Factor IX and Factor XI, and other factors, in addition to preoperative imaging with Dupplex \[[@rjy135C2], [@rjy135C9]\]. Anticoagulation options for acute thromboembolic events include unfractionated heparin, LMWH, fondaparinux and the direct oral anticoagulants (DOACs) with equivalent efficacy \[[@rjy135C10]\]. To note, the assay done at our institution detects only the FV Leiden G1691A mutation, Factor II (Prothrombin) G20210A mutation and the MTHFR C677T mutation. Heterozygosity for MTHFR by itself has not been found to contribute to thrombosis, however, it is important to note that other mutations may be associated to this gene and may lead to an increased homocysteine level. In the case of our patient, homocysteine level was mildly elevated (15 μmol/L). A variety of other genetic and non-genetic factors may contribute to increased risk of thrombosis. CONCLUSION {#rjy135s8} ========== Here we report a rare case of triple thrombophilic mutations associated with multiple thrombophilic episodes. We suggest a role for screening for thrombophilic mutations in the Eastern Mediterranean patients undergoing bariatric surgeries for morbid obesity due to the increased risk of thrombosis in this group of patients. CONFLICT OF INTEREST STATEMENT {#rjy135s9} ============================== The authors declare that they have no conflict of interest. ETHICAL APPROVAL STATEMENT {#rjy135s10} ========================== All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. INFORMED CONSENT STATEMENT {#rjy135s11} ========================== Informed consent was obtained from all individual participants (one patient) included in the study.

Based on the reflection on these three misunderstandings for the design of negative pressure isolation ward, as well as a series of experimental studies, both the concept and the effective technology for effective isolation of airborne transmission were formed under the dynamic condition during the common operation of negative pressure isolation ward. The dynamic isolation concept proposed by author in 2002 \[[@CR1]\] can be established completely \[[@CR2]\]. The dynamic isolation is relative to the static isolation which means the application of the barrier (such as closing the air-proof door) and the static pressure difference to prevent leakage through gap. This concept has been adopted by Beijing local standard DB11/663-2009 "*Essential construction requirements of negative pressure isolation wards*". It has also been applied in the design and the construction of negative pressure isolation ward in different places. Proper Pressure Difference for Isolation {#Sec1} ======================================== According to the principle of air cleaning technology, the purpose of isolation is to prevent infection and cross infection. It is especially applied to prevent the transmission of infectious pathogen between indoors and outdoors through air movement. It is an effective measure to realize the purpose of infection control. For example, the following aspects are considered as the main reasons for the prolonged period of infection by pulmonary tuberculosis, including the delayed treatment on the tuberculosis patients, shortage of isolation period, and deficiency of ventilation in isolation ward. Except for isolation with barrier (physical isolation) such as isolation room, the terminology of isolation mentioned here mainly means the isolation with pressure difference. This has been illustrated in the previous chapter. It will be unsuccessful to rely on the pressure difference to realized dynamic isolation. The pressure difference is aimed to maintain static isolation. Therefore, in the concept of dynamic isolation, isolation with proper pressure difference is needed. Physical Significance of Pressure Difference {#Sec2} -------------------------------------------- When all the doors and windows indoors are closed, the pressure difference is the resistance of airflow through the gap of the closed door or window, which is from the high pressure towards low pressure. From Fig. [3.1](#Fig1){ref-type="fig"}, when the pressures at both sides of the gap are assumed *P* ~1~ and *P* ~2~, the pressure difference can be expressed as:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta P = P_{ 1} - P_{ 2} = \left( {\xi_{ 1} + \xi_{ 2} } \right)\frac{{v^{ 2} \rho }}{ 2} + h_{w} ({\text{Pa}})$$\end{document}$$whereFig. 3.1Schematic diagram of gap $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\left( {\xi_{ 1} + \xi_{ 2} } \right)\frac{{v^{ 2} \rho }}{ 2}}$$\end{document}$is the local resistance at the gap where air flows through;*h*~w~is the frictional resistance. Since the depth of the gaps on door, window and panel are in the magnitude of 10^−2^ m, *h* ~w~ can be ignored completely;*ξ*~1~is the local resistance at the sudden contraction position. Since the cross sectional area of the gap is extremely small, *ξ* ~1~ ≈ 0.5;ξ2is the local resistance at the sudden expansion position. Since the cross sectional area of the gap is extremely small, ξ1 ≈ 1;ρis the air density, which is about 1.2 kg/m^3^ From Eq. ([3.1](#Equ1){ref-type=""}), we can obtain$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$v = \frac{ 1}{{\sqrt {\xi_{ 1} + \xi_{ 2} } }}\sqrt {\frac{ 2\Delta P}{\rho }} = \varphi \sqrt {\frac{ 2\Delta P}{\rho }\,} ({\text{m}}/{\text{s}})$$\end{document}$$where *φ* is the velocity coefficient, $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varphi = \frac{1}{{\sqrt {\xi_{1} + \xi_{2} } }} = 0.82}$$\end{document}$. Because the resistances of the gaps are different, the value of *φ* could be large or small. The air velocity of leakage air at different parts of the gap will be different. Because the geometry of the gap is relatively complex, its resistance will be increased. The value of the velocity coefficient *φ* will be reduced. With the given Δ*P*, the air velocity through the gap will be decreased, which will be introduced later. Determination of Pressure Difference {#Sec3} ------------------------------------ How to determine the pressure difference for the isolation ward under the condition of door and window closing? It has already been pointed out that it should depend on the enough flow rate of outdoor air to be sucked in through the gap of the door, so that the leakage pollutant airflow through the door gap can be prevented \[[@CR3]\]. This belongs to the static isolation. According to the field test by author, the maximum air velocity induced by occupant movement is 0.34 m/s when the walking velocity is 1 m/s. The air velocity indoors created by air supply from air conditioner is usually not larger than 0.3 m/s. The air velocity in normal room with natural ventilation is not larger than 0.2 m/s, which means that the air velocity induced through the gap will not be larger than 0.5 m/s. From Eq. ([3.2](#Equ2){ref-type=""}), when the air velocity through the gap is 0.5 m/s, the theoretical pressure difference can be calculated with:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta P = \frac{{\rho v^{2} }}{{2{\varphi}^{2} }} = \frac{{1.2 \times 0.5^{2} }}{{2 \times 0.82^{2} }} = 0.22\,{\text{Pa}}$$\end{document}$$ This means that in theory when the door is closed and the pressure difference reaches 0.22 Pa, the requirement for common leakage prevention can be satisfied. It has been pointed out early that when Δ*P* = 1 Pa the air velocity through the gap in theory could reach 1.06 m/s \[[@CR4]\], which could counteract the leakage completely. After the epidemic of SARS, it has been pointed out by "*Guidelines for Preventing the Transmission of Mycobacterium tuberculosis in HealthCare Facilities*" published in 1994 by CDC that for maintenance of the negative pressure and prevention of air flow into the ward, the minimum pressure difference is extremely small (0.001 in H~2~O). It was considered that with the pressure difference 0.001 in H~2~O, the leakage flow rate could reach 50 ft^3^/min (85 m^3^/h). In this case, the minimum air velocity of leakage air sucked inwardly is 100 ft/min, which corresponds to 0.51 m/s. The static pressure 0.001 in H~2~O corresponds to 0.25 Pa. With Eq. ([3.2](#Equ2){ref-type=""}), the corresponding ideal air velocity through the gap is 0.53 m/s. Table [3.1](#Tab1){ref-type="table"} shows the requirements of pressure difference in isolation ward from standards abroad. In U.S.A., the pressure difference increases from the initial value 0.25 Pa to the current value 2.5 Pa. The reason why the pressure difference should be increased by ten time is not explained.Table 3.1Requirements of pressure difference in isolation ward from standards abroadStandard or guidelineControl objectNegative pressure difference between the ward and the corridor (buffer room), PaCDC guideline in U.S.A. (1994)Mycobacterium tuberculosis0.25ASHRAE handbook (*Health care facilities*) in U.S.A. (2003)Not specified0.25UK "*guidance on the prevention and control of transmission of multiple drug*-*resistant tuberculosis*" \[[@CR27]\]Mycobacterium tuberculosis0.25CDC in U.S.A. "*guidelines for environmental infection control in health care facilities*" \[[@CR27]\]Not specified2.5DHHS in U.S.A. "*guidelines for construction and equipment of hospital and medical facilities*" \[[@CR27]\]Not specified2.5AIA in U.S.A. "*guidelines for design and construction of hospital and health care facilities*" \[[@CR27]\]Not specified2.5Australia "*guidelines for the classification and design of isolation rooms in health care facilities*" \[[@CR27]\]Aerosol15ASHRAE 170 "*ventilation of health care facilities*" (2013)Isolation ward with airborne infection2.5Russian standard GOST R 52539-2006 "*air cleanliness in hospitals. general requirements*"Isolation ward with airborne infection10--15 Are the real air velocity through the gap and the pressure difference consistent with the above-mentioned condition? The necessary minimum pressure difference will be analyzed further. The case with *φ* = 0.82 belongs to the ideal condition for the gap. In fact, the resistance on the gap is much larger. Table [3.2](#Tab2){ref-type="table"} shows the air velocity through the door gap in real situation. Of course, it is difficult to perform measurement. Therefore, there is error. But from these data, the credibility of the theoretical equation can be found.Table 3.2Measured air velocities through door gapsNo.LocationDoor gap, mmPressure difference, PaAir velocity through gap, m/sTheoretical air velocity though gap for *φ* = 0.82Actual velocity coefficient *φ*YearDepthLengthMeasured valueAverage1No. 1 operating room at Binzhou people's hospital201500+222.803.03.03.403.903.224.970.5320042No. 1 operating room at Zhengzhou people's hospital201500+101.000.650.450.650.750.73.340.17220043No. 1 Asepsis room at inner Mongolia biological & pharmaceutical Co., Ltd8960+70.550.602.501.222.80.3620044No. 3 Asepsis room at inner Mongolia biological & pharmaceutical Co., Ltd.8960+10.30.280.20.261.060.2520045Bacteria Room at inner Mongolia biological & pharmaceutical Co., Ltd.5960−70.850.800.700.782.80.2320046Jiangxi Keda animal pharmaceutical Co., Ltd.//+40.480.520.560.490.510.480.512.110.22004Average velocity coefficient(0.29)7Chemical analysis lab at inner Mongolia biological & pharmaceutical Co., Ltd.596000.170.200.180.180(0.29)2004 In case 7 from Table [3.2](#Tab2){ref-type="table"}, the actual measured pressure difference is 0. The air velocity through the door gap should be zero, but the measured value is 0.18 m/s. This means there is measurement error for the pressure difference 0 Pa, which could not be used to calculate the actual velocity coefficient. In the previous 6 cases from Table [3.2](#Tab2){ref-type="table"}, the average value is $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\overline{\varphi } = 0. 2 9}$$\end{document}$. When it is used as the velocity coefficient for case 7, we obtain:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${0.18 = 0.29\sqrt {\frac{{2\Delta P}}{1.2}} }$$\end{document}$$ The actual pressure difference for case 7 is Δ*P* = 0.23 Pa. The pressure difference value 0.23 Pa is much smaller than the half of the resolution of the current liquid column manometer, i.e., 1 Pa. It is immeasurable, not alone to say the needle manometer. Table [3.2](#Tab2){ref-type="table"} shows the measured pressure difference was 0, which is natural. This measured result 0 Pa does not represent the actual pressure difference. From Table [3.2](#Tab2){ref-type="table"}, the actual value *φ* is between 0.2 and 0.5. Suppose it was 0.5 (when the air-tightness level of doors is less than that of 0.2), when *v* = 0.5 m/s, Δ*P* = 2.6 Pa. This means that both the theoretical calculation result and the pressure difference 0.23--0.25 Pa provided by CDC from U.S.A. are not practical. It is not only unsafe, but also difficult to measure and control automatically. In fact, the more air-tightness the geometry is, the larger the resistance of the gap is. In this case, the necessary Δ*P* is much larger. When the air velocity through the door gap is required to be larger than 0.5 m/s, the minimum pressure difference in theory should be larger than 3 Pa. It is shown that when the door and the window are closed, the pressure difference to prevent the leakage through the gap could be as small as 3 Pa. The opinion is unnecessary that the larger the pressure difference is, the better it is, which will be explained with the experimental data later. But when the pressure difference is as small as 1 Pa, the requirement cannot be satisfied. Therefore, the following two concepts are provided:The pressure difference of the room needed is not large. It is feasible to adopt the common value 5 Pa. The reason why the pressure difference +5 Pa is adopted in common cleanroom will be introduced. One is that it can meet the requirement. The other is that 5 Pa corresponds to 0.5 mm H~2~O. It is the half of the smallest scale of the manometer in the metric system, which means the resolution is 0.5 mm. Therefore, in imperial unit system the half of the smallest scale is adopted as the minimum pressure difference, which is 0.05 in H~2~O or 1.27 mm H~2~O or 12.5 Pa. For the purpose of automatic control, in order to prevent too much instantaneous fluctuation of the pressure difference, it is impossible to maintain 5 Pa, so it is possible to use 10 Pa. When the relative pressure difference is too much, such that larger than 30 or 50 Pa, the people or the little creature will feel uncomfortable. Therefore, in the "*Requirements for ultra*-*clean ventilation* (*UCV*)*, Systems for operating departments*" published by National Health Service in U.K. and National Institute for Health Research in U.K., as well as "*Architectural technical code for hospital clean operating department*" (GB 50333-2002) in China, the pressure difference is specified not to exceed the limit of 30 Pa. In the later revision of GB 50333 issued in 2013, the pressure difference reduces from original 30 to 20 Pa based on the requirement of ISO 14644.(2)Trouble maybe occur when the envelope of the room is extremely air-proof. This has also been discovered by literature \[[@CR5]\]. It has been pointed out in this paper that when the automatic air valve at the exhaust air pipeline changes the position slightly because of the control error, although the variation of air volume induced is very small, the influence on the fluctuation of indoor pressure will be very large. During the adjusting process on negative pressure isolation room, when the variation of air volume for the room with the sealing strip near the edges of the door is only several m^3^/h, which corresponds to one thousandth of the exhaust air volume from the room, the change of the pressure difference could reach 1 Pa. For example, when the air volume in a lab (23.9 m^2^ × 3 m) changed by 5--10 m^3^/h, the pressure difference varied by 1 Pa \[[@CR6]\]. In order to stabilize the pressure difference, the sealing strip was forced to be removed. It is common that the variation of the air volume reaches several m^3^/h. Buffer Room for Isolation {#Sec4} ========================= From the aforementioned analysis, no matter whether the door is air-proof, with the influence of door opening, occupant movement and the temperature difference, pollutant will release outwardly with the same magnitude during the opening process of door. However, when door is closed and the pressure difference is as small as 5 Pa, the velocity of the entrained air at the gap could reach more than 2 m/s, which could prevent the outward leakage of pollutant. The leakage rate reduced by 40 and 60% when the pressure difference is −6 and −30 Pa, respectively. Therefore, the concept of buffer room to prevent the outward leakage of pollutant efficiently will be proposed. Mode of Buffer Room {#Sec5} ------------------- Basic mode Buffer room is the air lock room where clean air is supplied through HEPA filter. Figure [3.2](#Fig2){ref-type="fig"} shows the schematic diagram of an air lock room. It acts as an auxiliary part of the cleanroom. It was firstly proposed by "*Contamination Control of Aerospace Facilities*" (TO 00-25-203) issued by U.S. Air Force in 1961. Air lock room is a small room near the entrance of the cleanroom. For several doors in the air lock room, only one door could be open at the same time. It is aimed to prevent the contaminated air in the outside area from flowing into the cleanroom, so that the "sealing" function works.Fig. 3.2Schematic diagram of an air lock room Of course, the air lock room could also be used to prevent the contaminated air inside the room from flowing into the environment. In "*Good Manufacturing Practice*" (GMP) by WHO, "Airlock is an enclosed space with two or more doors, which is interposed between two or more rooms, e.g. of differing classes of cleanliness, for the purpose of controlling the airflow between those rooms when they need to be entered. An airlock is designed for use either by people or for goods and/or equipment". It is shown that air lock room is only a room with interlock doors. It is the same as the delivery window. When its volume is not large, the maximum quantity of polluted air for the other side is equivalent to one fourth of its volume. But this kind of polluted air is different from that enters into the buffer room, which has not been diluted by clean air. Air lock room can only control airflow, but cannot dilute airflow. Gradient pressure difference is established between two adjacent connected areas, which reduces the pressure value from the pollutant prevention side to the polluted side. In this way, pollution through the gap between two regions (rooms) by induction of some factor can be prevented, which moved from the polluted side to the pollution prevention side. In general, the area with high pressure for pollution prevention and the area with low pressure for isolation should be place at the end or at the center of the plane, which is shown in Fig. [3.3](#Fig3){ref-type="fig"}. The pressure difference of the isolation ward relative to the atmosphere could be positive or negative. In this book, it is mainly aimed for negative pressure.Fig. 3.3Gradient pressure difference on the plane Buffer room is placed outside of the isolation ward. Positive pressure is maintained in the buffer room relative the isolation ward, while negative pressure or zero pressure is kept in buffer room relative to the outside of the buffer room. This kind is called Three-Room-One-Buffer, or Two-Area-One-Buffer, which is shown in Fig. [3.4](#Fig4){ref-type="fig"}. Three rooms mean the isolation ward, the buffer room and the corridor. Two areas mean the polluted isolation ward and the corridor with potential pollution.Fig. 3.4Schematic diagram of three-room-one-buffer Buffer room is placed outside of the isolation ward. Inner corridor is set outside of the buffer room. The second buffer room is set outside of the inner corridor. Preparatory area for medical personnel is set outside of the buffer room. Positive pressure or zero pressure is maintained in the preparatory area. Negative pressures are kept inwardly. The magnitude of negative pressure increases gradually. This type is called Five-Room-Two-Buffer, or Three-Area-Two-Buffer, which is shown in Fig. [3.5](#Fig5){ref-type="fig"}. Five rooms mean the isolation ward, the buffer room 1, the inner corridor, the buffer room 2, and the clean area. Three areas mean the polluted area, the area with potential pollution, and the clean area.Fig. 3.5Schematic diagram of five-room-two-buffer The isolation ward belongs to the polluted area. The inner corridor belongs to the semi-polluted area. The preparatory area belongs to the clean area.2.Analysis on pollutant flux After theoretical analysis on the effect of the buffer room on the negative pressure isolation ward, quantitative assessment of the pollution flux and the effect of the buffer room was obtained. Novel founding was provided for the effect of the buffer room \[[@CR7]--[@CR9]\]. Figure [3.6](#Fig6){ref-type="fig"} shows the phase diagram calculation is provided for analysis of pollutant flux. In the figure, No. 1--5 is the serial number of the room. *V* is the room volume. *N* ~1~ is the pollutant concentration in Room 1 or at the gate of Room 1, pc/m^3^. *Q* ~1~ is the flow rate from Room 1 to Room 2 because of the pressure difference which is not counteracted after door is open, m^3^.Fig. 3.6Phase diagram calculation of five-room-two-buffer Now the analysis on the pollutant flux is performed as follows.At the moment of door opening for Room 1, the pollutant flux at the gate is *N* ~1~ *Q* ~1~.The volume of the buffer room is very small (it is usually not larger than 5--6 m^3^) and the air change rate is very large (it is usually about dozens h^−1^). With the effect of dispersion, during the 2--3 s for opening and closing of door for Room 1, three conditions of pollutant distribution which enters into Room 2 can be assumed, which is shown in Fig. [3.7](#Fig7){ref-type="fig"}.Fig. 3.7Extent of mixtureIn Fig. [3.7](#Fig7){ref-type="fig"}a, pollutant is fully mixed in the whole room (this is the common case for the buffer room).In Fig. [3.7](#Fig7){ref-type="fig"}b, pollutant is distributed in part of the room, but it does not reach the exit of Room 2 (For Room 3 without buffer room belongs to this case).In Fig. [3.7](#Fig7){ref-type="fig"}c, pollutant is also distributed in part of the room, but it also reaches the exit of Room 2 (When the room is spacious with shallow depth, this situation may appear). How is the pollutant distributed, i.e., how is the performance of mixture? It can be expressed with the mixture coefficient *α*. For small room such as the buffer room, the performance of mixture is good, which is shown in Fig. [3.7](#Fig7){ref-type="fig"}a. In this case, *α* ~2~ is set 1. For the case with poor performance of mixture, *α* ~2~ may be 0.2. When there is no air supply in this case, the mixture coefficient could chosen with half at the maximum, i.e., 0.5. On average, it could set 0.35. Therefore, the resultant concentration can be expressed as follows:After the door of Room 1 is closed, the initial pollutant concentration in Room 2 is: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N_{20} = \frac{{N_{1} Q_{1} }}{{V_{2} \alpha_{2} }}\,({\text{pc}}/{\text{m}}^{3} )$$\end{document}$$ (2)With the self-purification effect of the air by HEPA filter installed in return air or supply air pipeline, when there is no pollutant particle source in the buffer room and in the atmosphere, the self-purification for this increased concentration can be expressed as follows based on the instantaneous concentration equation \[[@CR10]\]. $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\frac{{N_{2t} }}{{N_{20} }} \approx {\text{e}}^{{\frac{ - nt}{60}}} }$$\end{document}$$ Next the derivation of this equation will be introduced. For a room with air supply and air return (exhaust) system and HEPA filter installed in the supply air or return (exhaust) air pipelines, the instantaneous concentration *N* ~t~ inside the room can be expressed as:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N_{t} = \frac{{60G \times 10^{ - 3} + M_{n} (1 - S)(1 - \eta_{n} )}}{{n\left[ {1 - S(1 - \eta_{r} )} \right]}} \times \left\{ {1 - \left[ {1 - \frac{{N_{0} n\left[ {1 - S(1 - \eta_{r} )} \right]}}{{60G \times 10^{ - 3} + M_{n} (1 - S)(1 - \eta_{n} )}}} \right]{\text{e}}^{{\frac{{ - nt\left[ {1 - S(1 - \eta_{r} )} \right]}}{60}}} } \right\}$$\end{document}$$where *G* is the particle generation rate per volume from occupants and surfaces indoors, pc/(m^3^ · min); *M* ~n~ is the atmospheric particle concentration, pc/L; *S* is the ratio of the return air volume to the supply air volume; *η* ~n~ is the total efficiency of filters installed on fresh air pipeline; *η* ~r~ is the total efficiency of filters installed on return air pipeline; *N* ~0~ is the initial concentration indoors. Let $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{60G \times 10^{ - 3} + M_{n} \left( {1 - S} \right)\left( {1 - \eta_{n} } \right)}}{{n\left[ {1 - S\left( {1 - \eta_{r} } \right)} \right]}} = A$$\end{document}$, Eq. ([3.4](#Equ4){ref-type=""}) can be re-written as:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N_{t} = A \times \left\{ {1 - \left[ {1 - \frac{{N_{0} }}{A}} \right]{\text{e}}^{{\frac{{ - nt\left[ {1 - S\left( {1 - \eta_{r} } \right)} \right]}}{60}}} } \right\} = A - A{\text{e}}^{{\frac{{ - nt\left[ {1 - S\left( {1 - \eta_{r} } \right)} \right]}}{60}}} + N_{0} {\text{e}}^{{\frac{{ - nt\left[ {1 - S\left( {1 - \eta_{r} } \right)} \right]}}{60}}}$$\end{document}$$ For the special case when pathogenic bacteria is released from patients in the isolation ward, it will enter into the buffer room because of the door opening, then it will enter into other rooms. Since there is no bacteria release source in the buffer room and the next following rooms, *G* = 0. Since there is no such bacteria in the atmosphere, *M* ~n~ = 0. So *A* = 0. Therefore, we obtained$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N_{t} = 0 - 0 + N_{0} {\text{e}}^{{\frac{{ - nt\left[ {1 - S\left( {1 - \eta_{r} } \right)} \right]}}{60}}} = N_{0} {\text{e}}^{{\frac{{ - nt\left[ {1 - S\left( {1 - \eta_{r} } \right)} \right]}}{60}}}$$\end{document}$$ Because *η* ~r~ is the total efficiency of filters installed on return (exhaust) air pipeline, according to Chinese standard, HEPA filters with Type B or higher requirement will be used for this application. The filtration efficiency for bacteria reached more than 99.9999% (refer to Sect. [3.4](#Sec10){ref-type="sec"}). So 1 − *η* ~r~ ≈ 0. Therefore, we obtain:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} N_{t} & = N_{0} {\text{e}}^{{ - {\text{nt}}/60}} \\ \frac{{N_{t} }}{{N_{0} }} & = {\text{e}}^{{\left( { - nt/60} \right)}} \\ \end{aligned}$$\end{document}$$ Neither the common cleanroom where there is bacteria generation inside nor the common environment where particles or bacteria exist in the atmosphere is suitable to adopt this equation. As for Eq. ([3.5](#Equ5){ref-type=""}), when *t* → ∞, we obtain$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{N_{t} }}{{N_{0} }} = 0$$\end{document}$$ This means *N* ~t~ = 0. In physics, when a certain amount of bacteria enter into the buffer room during the door opening process, it will be self-purified continuously in the buffer room or diluted by the incoming flow and then exhausted. When *t* → ∞, all the bacteria will be captured by HEPA filter or exhausted outdoors, so that air cleanliness level will be recovered in the buffer room. When *t* = ∞, we obtain$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{N_{t} }}{{N_{0} }} = 1$$\end{document}$$ This means *N* ~t~ = *N* ~0~. At the moment when bacteria enter into the buffer room, the instantaneous concentration is *N* ~1~, which is the initial concentration of the bacteria in the buffer room. From the above analysis, before the pollutant enters into Room 3, the pollutant concentration in Room 2 (refer to Fig. [3.8](#Fig8){ref-type="fig"}) will be$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N_{2t} = \frac{{N_{1} Q_{1} }}{{V_{2} \alpha_{2} }}{\text{e}}^{{\left( { - nt/60} \right)}}$$\end{document}$$where *t* is the self-purification time, min. It starts when the door in Room 1 is closed. It equals to the period for the walking from the door in Room 1 to the door in Room 2 before which is open (the interlock time of the door is included). It is usually between 5 s (\~0.1 min) and 30 s (\~0.5 min). *n* is the air change rate, h^−1^.Fig. 3.8Distribution of pollutant after it enters into room 2 (3)When the door of Room 2 is open and occupant moves from Room 2 to Room 3, the pollutant flux at the gate of Room 3 is shown in Fig. [3.9](#Fig9){ref-type="fig"}, which isFig. 3.9Schematic of pollutant entering into room 3 $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{N_{1} Q_{1} Q_{2} }}{{V_{2} \alpha_{2} }}{\text{e}}^{{\left( { - nt/60} \right)}} ({\text{pc}})$$\end{document}$$ So when the door of Room 2 is closed, the initial concentration of Room 3 becomes$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N_{30} = \frac{{N_{1} Q_{1} Q_{2} }}{{V_{2} V_{3} \alpha_{2} \alpha_{3} }}{\text{e}}^{{\left( { - nt/60} \right)}}$$\end{document}$$ After Room 3 is self-purified, the concentration of the airflow entering into Room 4 becomes$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N_{3t} = \frac{{N_{1} Q_{1} Q_{2} }}{{V_{2} V_{3} \alpha_{2} \alpha_{3} }}\left[ {{\text{e}}^{{\left( { - nt/60} \right)}} } \right]^{2} ({\text{pc}}/{\text{m}}^{3} )$$\end{document}$$ For Room 2, *n* is big and *t* is small. While for Room 3, *n* is small and *t* is big. For the simplification of assumption in calculation, values of *nt* are assumed the same, which is shown in Fig. [3.10](#Fig10){ref-type="fig"}.Fig. 3.10Distribution of pollutant after it enters into room 3 (4)At the moment when the door of Room 3 is open, the pollutant flux at the gate of Room 4 (buffer room) is shown in Fig. [3.11](#Fig11){ref-type="fig"}, which isFig. 3.11Schematic of pollutant entering into room 4 $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{N_{1} Q_{1} Q_{2} Q_{3} }}{{V_{2} V_{3} V_{4} \alpha_{2} \alpha_{3} \alpha_{4} }}\left[ {{\text{e}}^{{\left( { - nt/60} \right)}} } \right]^{2}$$\end{document}$$ (5)With the same analysis method for (2), after the pollutant is fully mixed in Room 4, the concentration in Room 4 before the pollutant enters into Room 5, which is shown in Fig. [3.12](#Fig12){ref-type="fig"}, becomesFig. 3.12Distribution of pollutant after it enters into room 4 $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N_{4t} = \frac{{N_{1} Q_{1} Q_{2} Q_{3} }}{{V_{2} V_{3} V_{4} \alpha_{2} \alpha_{3} \alpha_{4} }}{\left[ {{\text{e}}^{{\left( { - nt/60} \right)}} } \right]^{3} }({\text{pc}}/{\text{m}}^{3} )$$\end{document}$$ (6)When the pollutant enters into Room 5 from Room 4, the pollutant flux at the gate of Room 5 becomes $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{N_{1} Q_{1} Q_{2} Q_{3} Q_{4} }}{{V_{2} V_{3} V_{4} \alpha_{2} \alpha_{3} \alpha_{4} }}\left[ {{\text{e}}^{{\left( { - nt/60} \right)}} } \right]^{3}$$\end{document}$$ After the door of Room 4 is closed, the initial concentration of Room 5 (refer to Fig. [3.13](#Fig13){ref-type="fig"}) becomes$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N_{50} = \frac{{N_{1} Q_{1} Q_{2} Q_{3} Q_{4} }}{{V_{2} V_{3} V_{4} V_{5} \alpha_{2} \alpha_{3} \alpha_{4} \alpha_{5} }}\left[ {{\text{e}}^{{\left( { - nt/60} \right)}} } \right]^{3} ({\text{pc}}/{\text{m}}^{3} )$$\end{document}$$ Fig. 3.13Distribution of pollutant after it enters into room 5 Isolation Coefficient of Buffer Room {#Sec6} ------------------------------------ Isolation coefficientIsolation coefficient is the ratio of the initial concentration in the isolation ward to the induced pollutant concentration in the third room by opening of two doors when the buffer room is set. The larger the total isolation coefficient is, the stronger the prevention ability is. It represents the enhancement ratio of the total prevention ability with buffer room to that without buffer room, which is labeled with *β*. Suppose *α* ~2~ = *α* ~3~ = *α* ~4~ = *α* ~5~ = *α* for Three-Room-One-Buffer scheme, we obtain$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta_{3 \cdot 1} = \frac{{N_{1} }}{{N_{30} }} = \frac{{V_{2} V_{3} \alpha^{2} }}{{Q_{1} Q_{2} \left[ {{\text{e}}^{{\left( { - nt/60} \right)}} } \right]}}}$$\end{document}$$ When it is the same assumption for Five-Room-Two-Buffer scheme, we obtain$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta_{5 \cdot 2} = \frac{{N_{1} }}{{N_{50} }} = \frac{{V_{2} V_{3} V_{4} V_{5} \alpha^{4} }}{{Q_{1} Q_{2} Q_{3} Q_{4} \left[ {{\text{e}}^{{\left( { - nt/60} \right)}} } \right]^{3} }}}$$\end{document}$$ With the same amount *Q* (m^3^) of incoming flow, suppose the number of the buffer rooms is *m* and the total number of the rooms is *k*. Volumes of two isolation wards are usually the same, i.e., *V* ~3~ = *V* ~5~ = *V*, m^3^. Volumes of two buffer rooms are also usually the same, i.e., *V* ~2~ = *V* ~4~, m^3^. Suppose the parameter *X* represents the ratio of the volumes between the isolation ward and the buffer room, we obtain the following general formula.$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta_{k \cdot m} = \frac{{V^{{\left( {k - 1} \right)}} \alpha^{{\left( {k - 1} \right)}} }}{{X^{m} Q^{{\left( {k - 1} \right)}} \left[ {{\text{e}}^{{\left( { - nt/60} \right)}} } \right]^{{\left( {k - 2} \right)}} }}}$$\end{document}$$ 2.Example Calculation of the entrainment airflow *Q* from one room to the other can be performed as follows. Since the flow rate induced by the pressure difference 5 Pa is less than 0.05 m^3^/s, the counteracting effect of the flow rate by the pressure difference less than 5 Pa can be ignored. From Table 10.1007/978-981-10-2923-3_2, the flow rate *Q* of the convection by door opening with Δ*t* = 1 °C within 2 s is 0.44 m^3^. From Sect. 10.1007/978-981-10-2923-3_2, the maximum flow rate *Q* of the entrainment flow during the door opening process is 0.9 m^3^. The flow rate of the entrainment flow by occupant movement within 2 s is 0.28 m^3^. Therefore, the total flow rate *Q* is Σ*Q* = 1.62 m^3^. (In literatures \[[@CR7]--[@CR9]\], it was 1.52 m^3^.) Suppose *nt* = 6. Since in the buffer room *n* is very large, we can assume *α* ~2~ = 1. In Room 3 the performance of the air change with air cleaning technique is good, but the value of *n* is smaller than that in the buffer room. We can assume *α* ~3~ = 0.8. On average *α* = 0.9. Suppose the volume of the isolation ward is 25 m^3^ and *X* = 5, we obtain$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{array}{*{20}l} {\beta_{3 \cdot 1} = \frac{{25^{2} \times 0.9^{2} }}{{5 \times 1.62^{2} \times 0.9}} = 42.9} \hfill \\ {\beta_{5 \cdot 2} = \frac{{25^{4} \times 0.9^{4} }}{{5^{2} \times 1.62^{4} \times 0.9^{3} }} = 2042} \hfill \\ \end{array}$$\end{document}$$ When Room 4 is the buffer room with positive pressure, we assume that the relative pressure difference between Room 4 and the exterior room is Δ*P* = +5 Pa. When the non-air-tight door is used, the pressurized outward flow rate *Q* ~4~ is$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Q_{4} = 1.62 + 0.08 = 1.7\,{\text{m}}^{3}$$\end{document}$$ $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta_{5 \cdot 2} = \frac{{25^{4} \times 0.9^{4} }}{{5^{2} \times 1.62^{3} \times 1.7 \times 0.9^{3} }} = 1943}$$\end{document}$$ It is shown that the influence of Room 4, whether it is the negative or the positive pressure buffer room, on the result is not large. The aforementioned isolation coefficients are related to some prescribed parameters such as the size of the door opening, the period of entrainment by occupant, and the induced flow rate. Therefore only the relationship of the order of the magnitude is reflected in the result. Please do not focus on the specific value. Influencing Factors for Performance of Buffer Room {#Sec7} -------------------------------------------------- Air change rate The size of the buffer room is very small. Even when the air change rate reaches 60 h^−1^, the corresponding flow rate is only 300 m^3^/h, which is inconsiderable. Therefore air must be supplied into the buffer room. For the common range of the air change rate in the buffer room, the influence of the air change rate on the isolation performance is not large. Figure [3.14](#Fig14){ref-type="fig"} shows the influence of the air change rate on the self-purification of the particle concentration (the background concentration is not included) entering into the buffer room, when the air change rate is within the common range \[[@CR9]\].Fig. 3.14Influence of the air change rate on the self-purification of the particle concentration After the door is closed, it usually takes 0.1 min to walk to the door of the other side. In Eq. ([3.5](#Equ5){ref-type=""}), e^−*nt*/60^ = *N* ~t~/*N* ~0~. It is shown from Fig. [3.14](#Fig14){ref-type="fig"} that the value with 120 h^−1^ is less than that with 60 h^−1^ by about 10%, and the influence on *β* ~3·1~ is slightly more than 10%. Table [3.3](#Tab3){ref-type="table"} illustrates the influence of the air change rate on the isolation coefficient. It is shown that when the air change rate increases by one time, the isolation performance only increases by more than 10%.Table 3.3Influence of the air change rate on the isolation coefficientAir change rate, h^−1^*t*, min*β* ~3·1~600.142.9800.143.91000.145.41200.147.7 Of course, when *n* reaches 1200 h^−1^, the performance will be improved obviously. However, this is impractical and unrealistic.2.Volume The larger the volume of the buffer room is, the better the isolation performance is. From the general formula mentioned above, with the given air change rate, when the volume of the buffer room is larger and when *x* is smaller, the isolation performance will be better. But when the volume is large, the mixture performance within a certain period will be poor, which means *α* will be smaller. When the reduction rate of *α* is not proportional to the increase rate of *x*, the total isolation performance will improve more. Considering the feasibility for the layout, the volume of the buffer room should not be smaller than 2--3 m^2^.3.Self-purification time The more the time it takes for occupant to walk through the buffer room, or the more the time it is set for self-lock, the better the isolation performance is. It is much economic and simpler to increase the self-purification time than to increase the air change rate. When *t* increases from 6 to 12 s, it is relative easy. But it is equivalent to the increase of the air change rate by one time. Of course, the time for self-lock should not be too long, which is usually within 30 s. Therefore, when serious condition of pollution occurs, the time to open the other door should be delayed, or the time of self-lock should be prolonged to 30 s. Compare with the condition of 6 s, for the air change rate 60 h^−1^ as shown in Fig. [3.28](#Fig28){ref-type="fig"}, the value of e^−*nt*/60^ decreases from 0.9 for the self-lock time 6 s to 0.6 for the self-lock time 30 s. The value of *β* ~3·1~ will increase by 1.5 times, and the value of *β* ~5·2~ will increase by (1.5)^3^ times. The total isolation coefficient will be increased as follows:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{array}{*{20}l} {\beta_{ 3\cdot 1} {\text{will}}\,{\text{increase}}\,{\text{from}}\, 4 2. 9\,{\text{to}}\, 6 4. 4} \hfill \\ {\beta_{ 5\cdot 2} {\text{will}}\,{\text{increase}}\,{\text{from}}\, 20 4 2\,{\text{to}}\, 6 8 9 2} \hfill \\ \end{array}$$\end{document}$$ 4.Conclusion From the above-mentioned points 2 and 3, we could not obtain the conclusion that the smaller the buffer room is, the better the isolation performance will be. With the given air change rate, the volume of the buffer room exerts no influence on the air change rate. But when the buffer room becomes small, the value of *x* in Eq. ([3.7](#Equ7){ref-type=""}) will become large, and the isolation performance will be poor. When the air supply rate is fixed, with the decrease of the volume of the buffer room, the air change rate will become large, and the isolation performance depends on the relative influences on *x* and *n*. Even when the isolation performance increases, in essence it is caused by the increase of the air change rate. It should be noted that if the buffer room is as small as half step between two doors, the value of *t* may be smaller than 6 s by 50%. In this case, the loss outweighs the gain for improving the isolation performance. Experimental Validation {#Sec8} ----------------------- Experimental schemeMicrobial experiment In the project entitle with "*Research on isolation performance of the isolation ward*", studies on performance of the buffer room was carried out in Institute of Building Environment and Energy, China Academy of Building Research, Beijing, China. Microbial experiment was performed in a simulated isolation ward, where the aforementioned experiment on the pressure difference was performed. Figure [3.15](#Fig15){ref-type="fig"} shows the appearance of the isolation ward. Figure [3.16](#Fig16){ref-type="fig"} shows the preparatory work performed in the ward. Figure [3.17](#Fig17){ref-type="fig"} illustrates the layout of the ward. Parameters in the experiment are shown in Table [3.4](#Tab4){ref-type="table"}.Fig. 3.15Appearance of the isolation ward Fig. 3.16The preparatory work performed in the ward Fig. 3.17Layout of the simulated isolation ward (in the figure, the bathroom did not work. There is no supply air and exhaust air. If the bathroom were real, the door of the bathroom should be open towards the ward inside) Table 3.4Related parameters for determining the isolation coefficient in the buffer room with microbial experimentVolumeCondition of air exchangeScheme 1Scheme 2Temperature (°C)Pressure difference (Pa)Temperature (°C)Pressure difference (Pa)Isolation ward a27.6 m^3^All fresh air with air change rate 12 h^−1^20.1∆*P* ~a−b~ = −520.2∆*P* ~a−b~ = 0Buffer room b6.25 m^3^No air exhaust. The air supply rate is very small, which is equivalent to the natural ventilation18∆*P* ~b−c~ = −518∆*P* ~b−c~ = 0Exterior room cAbout 27.6 m^3^No air supply and exhaustNormal temperatureNormal pressureNormal temperatureNormal pressure Figure [3.18](#Fig18){ref-type="fig"} shows the bacteria solution atomization system. Figure [3.19](#Fig19){ref-type="fig"} shows the simulated bacteria generation condition at the mouth.Fig. 3.18Schematic of the bacteria solution atomization system Fig. 3.19Simulated bacteria generation condition at the mouth In order to check the function of the buffer room, occupant was required to move outwardly through the door by opening and closing the door for one time (about 2 s). The type of the bacteria generated is the Bacillus atrophaeus (formerly *Bacillus subtilis* var. niger) with the strain number ATCC: 15442; 1.3343. It was provided by Biological Resource Center, Institute of Microbiology Chinese Academy of Sciences (IMCAS-BRC). There are different opinions in literatures about the naked size of the bacteria, which includes 0.5, 0.8, 1 and 1.5 μm. According to the data in literature \[[@CR11]\] published in 2003, the linear length and the width of this kind of *Bacillus subtilis* are 1 and 0.5 μm, respectively. But according to the SEM figure of this paper (which are shown in Figs. [3.20](#Fig20){ref-type="fig"} and [3.21](#Fig21){ref-type="fig"}), the linear length for most of the *Bacillus subtilis* is about 1.2 μm, and a few are smaller than 0.5 μm. As for whether the size of the spores generated in our experiment was the same as that in the published literatures, validation was not performed.Fig. 3.20SEM figure of the *Bacillus atrophaeus* (amplification ratio 1700) Fig. 3.21Enlarged SEM figure of the *Bacillus atrophaeus* (amplification ratio 13500) After being cultured to be colony forming unit, the color of this kind of bacteria becomes light yellow to red, which is rare for the hybrid strain (including *Bacillus subtilis*) in atmosphere. So we could consider the background of this kind of bacteria was zero. Error could be avoided. They are easily discovered. The concentration of the bacteria solution, the quantity of the liquid atomized, and the period used for atomization of the liquid should be controlled, so that the bacteria concentration was not too high to be counted in the isolation ward. At the same time, those bacteria passing through the buffer room can be sampled in exterior room, so that the isolation coefficient could be calculated. Therefore, according to the trial experiment, the bacteria solution concentration was set 10^10^--10^11^ pc/mL. In practice the bacteria solution concentration reached 8 × 10^10^ pc/mL. The bacteria solution concentration was determined by the gradual dilution method with the dilution ratio 10. According to the microbiological experiment method specified in foreign standard, the quantity of the solution atomized should be between 5 and 10 mL. The quantity of air flow rate for generation of bacteria should be 17 L/min. The period for generation of bacteria should be 30 min. Based on experience, the generated liquid droplet by the atomizer was between 1 and 5 μm. The pressure for atomization was 1 kg/cm^2^. Bacteria were sampled with the sedimentation method. The petri dishes should be placed on the floor and near the gate. The measurement points are shown in Fig. [3.22](#Fig22){ref-type="fig"}.Fig. 3.22Layout of five sampling points on the floor (the direction of the door is illustrated in Fig. 10.1007/978-981-10-2923-3_7) (2)Experiment with atmospheric dust Table [3.5](#Tab5){ref-type="table"} shows the experiment parameters.Table 3.5Related parameters for experiment on isolation coefficient of buffer room with atmospheric dustVolumeCondition of air exchangeOne people open and then close the doors to walk from the ward to the buffer roomTemperature, °CRelative humidity, %Isolation ward27.6 m^3^All fresh air20.055Buffer room6.25 m^3^Air supply and exhaust with air change rate 112 h^−1^20.654Exterior roomAbout 27.6 m^3^No ventilation---- All fresh air was supplied into the isolation ward when the fresh air did not pass through HEPA filter. When the indoor concentration was close to that of the atmospheric dust outdoors, the indoor concentration was measured. The self-purification process was turned on in the buffer room. Measurement was performed in the buffer room to determine if the air cleanliness level ISO 6 has been arrived. Then the opening and closing of doors were completed. The concentration in the buffer room was measured immediately. The particles entering into the buffer room were considered as microbes. Measurement was performed every 1 min. Three to four readings were recorded. The concentration in the isolation ward was monitored until it reached stable.(3)Experiment with temperature difference Table [3.6](#Tab6){ref-type="table"} shows the experiment with the temperature difference. For detailed information about the method, please refer to literature \[[@CR12]\].Table 3.6Conditions in the experiment with the temperature differenceConditionRelative pressure difference between the isolation ward and the buffer room, PaRelative pressure difference between the buffer room and the exterior room, PaTemperature in the ward, °CTemperature in the buffer room, °CRelative temperature difference between the ward and the buffer room, °CBefore opening the door of the wardDuring opening the door of the wardBefore opening the door of the wardDuring opening the door of the ward1−100+1--2017.016.8+0.22−100+1--2020.817.8+33−100+1--2022.717.7+5 2.Experimental resultsMicrobiology method \[[@CR11], [@CR13]\] Figure [3.23](#Fig23){ref-type="fig"} shows the results on No. 5 microbial sampling point in the isolation ward.Fig. 3.23Figure of the sedimentation bacteria in the ward Figure [3.24](#Fig24){ref-type="fig"} shows the results on No. 5 microbial sampling point in the buffer room.Fig. 3.24Figure of the sedimentation bacteria in the buffer room Figure [3.25](#Fig25){ref-type="fig"} shows the results on No. 5 microbial sampling point in the exterior room.Fig. 3.25Figure of the sedimentation bacteria in the exterior room In these figures, the white colony was cultured with the hybrid strain in atmosphere. It should be noted that only clear figures are presented. Results on No. 5 microbial sampling point in various rooms are relative clear. They cannot be used to obtain the isolation performance exactly, but they can be applied to show the trend of the isolation performance. Tables [3.7](#Tab7){ref-type="table"} and [3.8](#Tab8){ref-type="table"} show the measured data with Scheme 1 and Scheme 2, respectively. In Scheme 1, there is pressure difference and small temperature difference. In Scheme 2, there is no pressure difference and small temperature difference.Table 3.7Experimental result on CFU of isolation performance for three-room-one-buffer with pressure difference (−5 Pa) and small temperature differenceBacteria solution concentration, 8 × 10^10^ pc/mL\ Quantity of the solution atomized, 6.12 mL\ Period for liquid spray, 30 minIsolation wardMeasurement point12345AverageData986872823688720817.8Exterior roomMeasurement point678910AverageData--186206160164179Buffer roomMeasurement point1112131415AverageData433040464540.8 Table 3.8Experimental result on CFU of isolation performance for three-room-one-buffer with no pressure difference (0 Pa) and small temperature differenceBacteria solution concentration, 8 × 10^10^ pc/mL\ Quantity of the solution atomized, 6.12 mL\ Period for liquid spray, 30 minIsolation wardMeasurement point12345AverageData800752672784--752Exterior roomMeasurement point678910AverageData190172194184--185Buffer roomMeasurement point1112131415AverageData4240--44--42 According to the above measured data, the isolation coefficients can be obtained as follows. For Scheme 1 with pressure difference and small temperature difference,$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta_{3 \cdot 1} = \frac{817.8}{40.8} = 20.04}$$\end{document}$$ For Scheme 2 with no pressure difference and small temperature difference,$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{array}{*{20}c} {\beta_{3 \cdot 1} = \frac{752}{42} = 17.9} \\ {\overline{{\beta_{3 \cdot 1} }} = 19} \\ \end{array}$$\end{document}$$ (2)Experimental method with atmospheric dust \[[@CR14]\] The isolation performance with atmospheric dust is obtained through proceeding of data in Table 10.1007/978-981-10-2923-3_4, which is shown in Table [3.9](#Tab9){ref-type="table"}.Table 3.9Experimental result on isolation coefficient for two-room-one-buffer (one isolation ward and one buffer room) with atmospheric dustRelative pressure difference between the ward and the buffer room, PaIsolation coefficientNoteOne people walks out for 2 s−3137.0Maximum concentration appears at the second minute−3040.0Maximum concentration appears at the first minute−638.5Maximum concentration appears at the second minute024.4Maximum concentration appears at the first minute018.9Maximum concentration appears at the first minuteAverage31.8 (3)Experimental method with temperature difference It has been shown in Chapter 10.1007/978-981-10-2923-3_2.(4)Experimental method with artificial dust From the data provided in literature \[[@CR15]\], the isolation performance with artificial dust for the scheme with one operating room and the other buffer corridor is obtained, which is shown in Table [3.10](#Tab10){ref-type="table"}.Table 3.10Experimental result on isolation coefficient for two-room-one-buffer (one operating room and one buffer corridor) with atmospheric dustType of doorRelative pressure difference between the operating room and the buffer corridor, PaAverage isolation coefficientOutwardly opening door0$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\frac{{2.6 \times 10^{8} }}{32.4} \div \frac{{4.2 \times 10^{6} }}{13} = 24.2}$$\end{document}$−30$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\frac{{2.6 \times 10^{8} }}{32.4} \div \frac{{1.7 \times 10^{6} }}{13} = 59.8}$$\end{document}$*Note* In the above equations, the values 32.4 and 13 represent the volumes of the operating room and the buffer corridor, respectively 3.Analysis Comparison between theoretical analysis and experimental result Now the isolation performance with the microbiology method will be analyzed. In the buffer room, there is no exhaust air. The air supply volume can only provide the extremely small amount to maintain the pressure drop across the gap. Therefore, the mixing performance in the buffer room is relatively poor. *α* ~2~ should be smaller than 1, which could be set 0.84. There is no ventilation in the exterior room. According to the analysis in previous section, *α* ~3~ could be set 0.35. For the scheme with no pressure difference and small temperature difference, the flow rate of air passing through one room to the other room should be as follows. The temperature difference between the buffer room and the exterior room was ∆*t* = 1 °C, so the total air exchange rate was *Q* = 1.62 m^3^. The temperature difference between the isolation ward and the buffer room was ∆*t* = 2 °C, so the total air exchange rate was *Q* = 1.8 m^3^. According to Table [3.4](#Tab4){ref-type="table"}, we know$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${X = \frac{27.6}{6.25} = 4.42}$$\end{document}$$ Since in fact there was only air supply with *n* \< 60 and the sedimentation samplings were performed simultaneously in the isolation ward and the buffer room, the influence of the time could be ignored. When *t* = 2 s, *nt* \< 6 and e^(−*nt*/60)^ ≈ 1. Therefore, in theory we obtained$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta_{3 \cdot 1} = \frac{{27.6^{2} \times 0.85 \times 0.35}}{4.42 \times 1.62 \times 1.8 \times 1} = 17.6}$$\end{document}$$ The theoretical value is quite close to the measured data which is *β* ~3·1~ = 17.9. For the scheme with pressure difference and small temperature difference, when the flow rate (in 2 s) 0.082 m^3^ by negative pressure is counteracted, in theory we obtain *β* ~3·1~ = 18.5. It is also close to the measured value 20.04.(2)By experiments on isolation coefficient with the above-mentioned microbiology method and the experimental method with temperature difference, the isolation coefficients with the microbiology method and the temperature difference method are much smaller than that with the atmospheric dust in the scheme with one isolation ward and one buffer room. The experimental value of the isolation coefficient with atmospheric dust is much larger than the theoretical value. The reason is the self-purification time, which will be analyzed as follows. The theoretical isolation coefficient is defined by Eq. ([3.9](#Equ9){ref-type=""}), which only considers the entrance of "exterior particles" from the isolation ward to the buffer room.$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta_{2 \cdot 1} = \frac{{N_{1} }}{{N_{2t} }} = \frac{{N_{1} }}{{\frac{{N_{1} Q_{1} }}{{V_{2} \alpha_{2} }}{\text{e}}^{ - nt/60} }} = \frac{{V_{2} \alpha_{2} }}{{Q_{1} {\text{e}}^{ - nt/60} }}}$$\end{document}$$ For the data in Table [3.4](#Tab4){ref-type="table"}, the related parameters are as follows. *V* ~2~ = 6.25 m^3^ *α* ~2~ = 1 (with exhaust air) *Q* ~2~ = 1.62 m^3^/s *n* = 112.5 h^−1^ The maximum value for the pressure difference 0 Pa appears at 1 min (which is shown in Table [3.7](#Tab7){ref-type="table"}), so *t* = 1 and we obtain e^−*nt*/60^ = 0.153 Therefore, for the pressure difference 0 Pa, the isolation coefficient is$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta_{2 \cdot 1} = \frac{6.25 \times 1}{1.62 \times 0.153} = 25.2}$$\end{document}$$ The theoretical value 25.2 under the real situation is very close to the average measured data 21.7 shown in Table [3.7](#Tab7){ref-type="table"}. The maximum value for the pressure difference -30 Pa also appears at 1 min, so *t* = 1. But for the case with large pressure difference, the flow rate by the negative pressure should be counteracted. From Table 10.1007/978-981-10-2923-3_2, *Q* ≠ 1.62. Since the air velocity through the door gate is 0.112 m/s, with the area of the door we obtain the flow rate 0.22 m^3^/s. So *Q* = 1.4. Therefore, for the pressure difference −30 Pa, we obtain the isolation coefficient$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta_{2 \cdot 1} = \frac{6.25 \times 1}{1.4 \times 0.153} = 29.2}$$\end{document}$$ It is very close to the experimental data 40. The maximum value for the pressure difference −31 Pa also appears at 2 min, so *t* = 2 and we obtaine^−*nt*/60^ = 0.025 Therefore, for the pressure difference −31 Pa, the isolation coefficient is$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta_{2 \cdot 1} = \frac{6.25 \times 1}{1.4 \times 0.025} = 178.6}$$\end{document}$$ The theoretical value 178.6 under the real situation is far from the average measured data 37. The reason may be that for the timing of the optical particle counter, when the end of the reading was 2 min, which may be in fact just passing through 1 min. But it was recorded in the region of 2 min. When 1 min was used for calculation, the theoretical value *β* ~2·1~ became 29.2, which is close to the measured data 37. In literature \[[@CR16]\], the experimental data was also calculated with Eq. ([3.9](#Equ9){ref-type=""}) by the Japanese scholar, i.e.,$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta_{2 \cdot 1} = \frac{{V_{2} \alpha_{2} }}{{Q_{1} {\text{e}}^{ - nt/60} }}}$$\end{document}$$where *V* ~2~ is the volume of the isolation corridor, which could set 13 m^3^. *α* ~2~ is the coefficient, which could set 1 for the buffer room where air is supplied and exhausted. *Q* ~1~ is the flow rate. The size of the door is the same as that of the previous case. The time for the previous case is 2 s, but in this case it is 1.6 s. When people arrive at the door, they began to open and close the door. The average time of this process for ten people was obtained. The average flow rate obtained was 1.3 m^3^ (not 1.62 m^3^). *n* is the air change rate, which is 260/13 = 20 h^−1^. *t* is the first minute when the maximum concentration on the reading of the optical particle counter. Therefore, for the pressure difference 0 Pa, the isolation coefficient is$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta_{2 \cdot 1} = \frac{13 \times 1}{1.3 \times 0.72} = 14}$$\end{document}$$ For the case with the pressure difference -30 Pa, the flow rate by the negative pressure should be counteracted. *Q* = 1.1 m^3^. The theoretical value *β* ~2·1~ became 16.4.(3)During the experimental method with temperature difference, the air supply rate in the buffer room is 379 m^3^/h and *n* = 60.64 h^−1^. Since there is air supply and air exhaust with large air change rate, *α* = 1. Based on past calculation data, the flow rates *Q* under different temperature difference conditions is: 1.38 m^3^ for 0.2 °C, 1.92 m^3^ for 3 °C, and 2.14 m^3^ for 5 °C. In the previous 2 min, e^−*nt*/60^ = 0.134. Theoretical and experimental data for the isolation coefficient are shown in Table [3.11](#Tab11){ref-type="table"}.Table 3.11Theoretical and experimental data for the isolation coefficient of the outwardly opening door during the experiment with the temperature differenceΔ*t*, °C0.235*β* ~2·1~Calculation33.924.321.9Experiment28.623.215.2 It is shown that the calculation result in theory could be used to estimate the actual isolation performance on average. Both the theoretical and measured data are summarized in Table [3.12](#Tab12){ref-type="table"}. The difference between the theoretical and measured data is large only for the method with atmospheric dust. This is related to the explanation of the original data. But for other methods, the difference is very small. When some parameter is not accurate enough, the influence on the result will be very large. For example, when the error for determining the value of *t* is several seconds, this influence will appear. Therefore, there is still relative reference value in the theoretical equation.Table 3.12Comparison of the theoretical and measured isolation coefficientsType of experimentCountryConstitution of roomsTheoretical valueMeasured dataMicrobiology methodChinaThree-room-one-buffer17.6 for 0 Pa18.5 for 5 Pa17.920.04Method with atmospheric dustChinaTwo-room-one-buffer25.2 for 0 Pa29.2 for 30 Pa178.6 or 29.2 for 31 Pa21.740.037.0Method with artificial dustJapanTwo-room-one-buffer14 for 0 Pa16.4 for 30 Pa24.259.8Method with temperature differenceChinaTwo-room-one-buffer16.9 (average value with three data under different temperature difference values)22.3 (average value with three data under different temperature difference values) (4)Influence of the pressure difference on the isolation performance Based on the aforementioned experimental data, it is shown that the influence of the pressure difference on the isolation performance is not large. This is because the counteracting flow rate by the negative pressure difference on the outward leakage flow rate is very small. Results are shown in Table [3.13](#Tab13){ref-type="table"}.Table 3.13Influence of the pressure difference on the isolation performanceType of experimentAmplification ratio of the isolation performance in experimentRatio for −5 (or −6) Pa to 0 PaRatio for −30 to 0 PaMicrobiology experimentMethod with atmospheric dustMethod with artificial dust0.120.67−0.7--0.671.47 It takes too much effort to create the pressure difference −30 Pa, compared with the situation with the pressure difference 0 Pa. However, the amplification magnitude for the increase of the isolation coefficient is not too much. Besides under the condition with the pressure difference −30 Pa, when there is temperature difference, the pollutant cannot be prevented from leakage outwardly during the opening process of the door. From Table 10.1007/978-981-10-2923-3_2, the quantity of the leakage flow rate still reached more than one fifth of the original concentration. Therefore, the cost-effective performance to adopt the pressure difference −30 Pa is almost the same as that of the situation with −5 Pa. In short, it is not reasonable and safe to pursue the isolation performance by increasing the pressure difference. When the pressure difference increases from 0 to −30 Pa, the isolation performance is increased by more than one time. However, when one buffer room is added, the isolation coefficient *β* ~3·1~ will be increased by a dozen times. Door of Buffer Room {#Sec9} ------------------- There are two doors for the buffer room. One is the door adjacent to the isolation ward, which can also be called the door of the ward. The other is the door for entrance into the interior corridor. For the purpose of convenience, the former door is called the inner door of the buffer room, and the latter is called the outer door of the buffer room. As mentioned before, the extent of the air-proof for the inner door has no effect on the outward leakage of air flow. However, there are still differences for different kinds of the doors. From Fig. 10.1007/978-981-10-2923-3_2 and Table 10.1007/978-981-10-2923-3_2, the performance of the outwardly opening door is poorer than that of the inwardly opening door which has worse performance than the sliding door. The air velocity of the counter current has been introduced before. Figure [3.26](#Fig26){ref-type="fig"} illustrates the air velocities of the counter current during the opening and closing of three different door under different pressure difference conditions. In the figure, the component of −*x* means the direction from indoors towards outdoors, and the component of +*y* means the direction from the heel post of the door towards the door handle, which are shown in Fig. [3.27](#Fig27){ref-type="fig"}. Since the air velocity of the *y* component during the closing of the door is larger than that during the opening of the door, the air velocity of the +*y* component during the closing of the door is presented in the figure \[[@CR15]\].Fig. 3.26Relationship between the air velocities of the counter current during the opening and closing of the door and the pressure difference Fig. 3.27Schematic diagram of the air velocity components for the counter current In this experiment, the air velocity of the counter current reached more than 1 m/s, which is obviously larger than the measured values by American scholar and our study. In this experiment, the air velocity is the one during the pushing process of the door, while in our study the air velocity was measured after the door was open. This may be the reason for the difference. From the figure, it is shown that:The magnitude of the air velocity for the counter current is: outwardly opening door \> inwardly opening door \> sliding door. This can be used to explain the sequence for the relationship between the particle number transmitted during the opening of the door and the type of the door shown in Fig. 10.1007/978-981-10-2923-3_2.The air velocity of the counter current is not related to the pressure difference. This means the air velocity of the counter current for Δ*P* = 0 Pa is close to that for Δ*P* = 30 Pa. It even beyonds the imagination that for the inwardly opening door, the air velocity of Δ*P* = 0 Pa is the maximum while that for Δ*P* = 30 Pa is the minimum. From the aforementioned analysis, with the suitable size of the buffer room, the sliding door could be adopted as the inner door, and the outwardly opening door could be utilized as the outer door. For the ward with negative pressure, the outwardly opening door can be used only. The problem of increasing the air-proof performance cannot be considered. The door with common air-proof performance is fine, except that the wooden material is not used. Airflow Isolation in Mainstream Area {#Sec10} ==================================== Concept of Mainstream Area {#Sec11} -------------------------- Medical personnel for ward round or operation near the sickbed will face the pollutant source directly. During the talking, coughing and sneezing processes of patients, it will pose a threat to the medical personnel. The range hood introduced in Chapter One has been denied by practice. There is a certain effect of prevention for wearing common masks. As for the designer of the isolation ward, how can we reduce the infectious risk of the medical personnel under the dynamic condition of the operation with the current air cleaning system? For protecting the medical personnel, CDC manual in U.S.A. proposed that clean air should be passing through the working area of the medical personnel. No only CDC emphasized this point, but also other related literatures has mentioned it. But one fact has been ignored, which is shown in Fig. [3.28](#Fig28){ref-type="fig"}. When clean air is supplied from the back of the people, negative pressure area will be formed in the front breathing zone of the people, which has no protective effect and is harmful.Fig. 3.28Negative pressure area formed in the front zone of the people Therefore, in the concept of dynamic isolation theory, the measures to utilize the mainstream area are proposed, which has been validated effectively. In 1979, the concept of the mainstream area was proposed by author \[[@CR3]\], which is shown in Fig. [3.29](#Fig29){ref-type="fig"}. The region of the downward supply air below the air supply outlet can be termed as the mainstream area. In this area, the air cleanliness level is the best and the ability to exhaust pollution is the best. At the upper region of the mainstream area, some of the surrounding air will be sucked in, which will be diluted and then exhausted together from the lower part of the room.Fig. 3.29Schematic diagram of the mainstream area When air is supplied and then exhausted as shown in Fig. [3.29](#Fig29){ref-type="fig"}, the average indoor concentration is:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${N_{v} = N_{s} + \psi \frac{{60G \times 10^{ - 3} }}{n}}$$\end{document}$$where *N* ~v~ is the average indoor concentration; *N* ~s~ is the concentration of the supplied air; *G* is the bacteria generation rate per unit volume; *n* is the air change rate; *ψ* is the non-uniform distribution coefficient (please refer to Table [3.14](#Tab14){ref-type="table"}).Table 3.14Non-uniform distribution coefficient *ψ*Air change rate, h^−1^1020406080*ψ*1.51.221.161.060.99 The parameter *ψ* can be calculated as follows.$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\psi = \left( {\frac{1}{\varphi } - \frac{\beta }{\varphi } + \frac{\beta }{1 + \varphi }} \right) \times \left( {\varphi + \frac{{V_{b} }}{V}} \right)}$$\end{document}$$where *φ* is the carrier ratio of the airflow. The value *φ* becomes 1.5, 1.4, 1.3, 0.65, 0.3 and 0.2 when the area corresponding to each air supply unit with air filter inside on the ceiling is ≥7, ≥5, ≥3, ≥2.5, ≥2 and ≥1 m^2^, respectively. *β* is the ratio of the particle generation rate in the mainstream area to that in the whole room; *V* is the room volume; *V* ~b~ is the volume of the vortex area. For the isolation ward with single patient and with the air change rate *n* ≈ 10^−1^, the value of *ψ* is 1.5 based on Table [3.14](#Tab14){ref-type="table"}. In this case, the average indoor concentration is:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${N_{v} = 1.5 \times \left( {N_{s} + \frac{{60G \times 10^{ - 3} }}{n}} \right)}$$\end{document}$$ Function of Mainstream Area {#Sec12} --------------------------- The mainstream area was proposed to protect the medical personnel. Air supply outlet is placed on top of the positions where medical personnel performed the ward round and operation near the sickbed of the patient, which is shown in Fig. [3.30](#Fig30){ref-type="fig"}.Fig. 3.30Schematic of the medical personnel near the sickbed under the protection of the mainstream area If there is no particle generation source in the mainstream area as shown in Fig. [3.30](#Fig30){ref-type="fig"}, *β* = 0. The area of the room is larger than 10 m^2^. If there is only one air supply outlet, the parameter is *φ* = 1.5. Therefore, the non-uniform distribution coefficient in the mainstream area becomes$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\psi = 1 - \frac{\beta }{1 + \varphi } = 1 - \frac{0}{1 + 1.5} = 1$$\end{document}$$ The average concentration in the mainstream area is$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${N_{a} = N_{s} + \frac{{60G \times 10^{ - 3} }}{n}}$$\end{document}$$ Therefore, based on the expressions of the concentration in the mainstream area and the average indoor concentration *N* ~v~, we obtain$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$N_{\text{a}} = 0.67N_{\text{v}}$$\end{document}$$ This means that the pollutant concentration in the mainstream area where the medical personnel stay and face as the breathing zone is about 2/3 of the average indoor concentration. This is much less than that in the vortex area. In other words, when the medical personnel stays in any place indoors (except the breathing area in front of the patient), the pollutant concentration could be considered as the average indoor concentration *N* ~v~. But when the medical personnel stay inside the mainstream area, the pollutant concentration can be reduced to 2/3 of *N* ~v~. Compared with the concentration in the breathing zone of the patient, the concentration in the mainstream area is much smaller. This is because the pollutant entering into the mainstream area only accounts for a small proportion, which is further diluted by the supplied air in the mainstream area. Therefore, the concentration of the mainstream area is much less than 2/3 of the average indoor concentration. Because in the ward, patient stays inside alone, or stays with other patients who carry the same kind of disease. They are already adapted to this habitat environment. The average indoor concentration is not too important for them. However, for the medical personnel, although they walk around in the ward, they will usually stay for a long period besides the sickbed. Because they are healthy people, they are sensitive to the indoor pollutant. Therefore, the pollutant concentration in this region should be as small as possible. This is the reason why the air supply outlet should be placed above the side of the sickbed where the medical personnel usually stay. It is aimed to protect the medical personnel through the mainstream area. Because of such performance of the mainstream area, the mainstream area should be enlarged as much as possible. With the given area of the air filters, the area of the air supply outlet should be increased, as long as the air velocity at the air supply outlet is not small than 0.13 m/s \[[@CR4]\]. Otherwise, the performance will be opposite. Application of Self-circulation Air Through HEPA Filter {#Sec13} ======================================================= Application Principle of Circulation Air {#Sec14} ---------------------------------------- From Chap. 10.1007/978-981-10-2923-3_2, we know that it is a misunderstanding that circulation air is not allowed for usage. But for the application of circulation air in the concept of the dynamic isolation, there are several principles:The circulation air does not mean the circulation of the return air in the central system for different rooms. It is the self-circulation for the air in each room.There must be a proportion of circulation air to be exhausted outdoors, so that negative pressure can be kept indoors. On the exhaust air outlet, HEPA filter must be installed.The exhaust air outlet and the return air outlet can be combined to be one apparatus. HEPA filter could be shared for both the exhaust air and the return air.The exhaust air outlet and the return air outlet are not required to be free of leakage. Function of HEPA Filter {#Sec15} ----------------------- Common disinfection methodsFeatures of common disinfection methodIn order to deal with pollution and infection in hospitals, pipeline cleaning and spraying sanitization were usually adopted in the past. Chemical and physical disinfection methods were utilized indoors. This kind of treatment methods with pollution at first and then control is termed as the passive treatment. It will yield twice the result with half the effort. Besides, it is a method with sequela. It is not a method to control pollution during the whole process. Instead, it only pays attention to pollution control at the beginning and the end.In hospitals the objects where disinfection is needed include: occupant, surfaces of object, and air. For occupant, hands need to be disinfected. People needs to take a bath and wear aseptic clothes.For surfaces of object, it should be wiped clean. The following methods may be used, which include the disinfection with chemical agent (more than one type must be used), the moist heat sterilization, the dry heat sterilization, the radiation sterilization, the gas sterilization, sterilization with air filtration (when sterilization cannot be performed eventually in the container).Examples of air disinfection methodTable [3.15](#Tab15){ref-type="table"} illustrates various methods for air disinfection, where the air disinfection efficiencies were shown in the product catalog, literatures, claim from manufacturer or the test report.Table [3.16](#Tab16){ref-type="table"} illustrates the measured result on electrostatic sterilization equipment performed by experts from Institute of Microbiology and Epidemiology at the Academy of Military Medical Sciences \[[@CR17]\].Table [3.17](#Tab17){ref-type="table"} shows the report from Chinese Journal of Nosocomiology \[[@CR18]\]. Table [3.18](#Tab18){ref-type="table"} shows the data from Tianjin University \[[@CR19]\] and National Center for Quality Supervision and Test of Building Engineering \[[@CR20]\]. Table [3.19](#Tab19){ref-type="table"} presents the report data by National Center for Quality Supervision and Test of Building Engineering. Table [3.20](#Tab20){ref-type="table"} provides the data from the thesis at China Academy of Building Research \[[@CR20]\]. Table [3.21](#Tab21){ref-type="table"} shows the data from literature \[[@CR21]\]. Table [3.22](#Tab22){ref-type="table"} presents data from the Academy of Military Medical Sciences \[[@CR17]\] and the dissertation at Tongji University \[[@CR22]\]. Table 3.15Various methods for air disinfectionDisinfection methodPrincipleEfficiencySingle-stage electrostatic precipitatorHigh voltage electric field will form corona, and generate free electron and ion. The dust and bacteria will contact them to become charged. Then they will deposit on the dust collecting electrode to be removed. For relative larger particles and fibers, the efficiency is poorer, because discharge will occur. The advantage is that it's able to remove dust and bacteria with small pressure drop. The shortcoming is that it is difficult and time-consuming to wash, and pre-filter must be installed. It may generate ozone and nitrogen oxides, which forms secondary-contamination problem50% (test on some products show that the efficiency only reaches about 20%)PlasmaUnder the condition of heating or strong electromagnetic field, gas will generate the electron cloud. Active free radical and rays have broad-spectrum germicidal effect. But it cannot remove dust66.7%Atractylodes lanceaTraditional Chinese medicine68.2%AnionWith the electric field, UV irradiation field, ray and impact of water, air is ionized to generate anion. It can adsorb dust particle, so that dust becomes heavy ion to settle down. The shortcoming is that secondary airborne dust can be formed. It is of little use in HVAC system73.4%Nano-photocatalysisWith the irradiation of sunlight and UV, oxidative decomposition of the volatile organic gas or the bacteria occurs on the surface of catalytic active material. They are converted to CO~2~ and water. The disinfected air must contact the catalytic material for a certain period. With the dust loading process, the performance becomes poor. So pre-filter must be installed. UV irradiation will generate ozone. In experiment, the efficiency may become negative75% (test on some products show that the efficiency only reaches only about 30%, and some of the efficiency even became negative)Formaldehyde fumigationIt is the chemical agent, which has been confirmed carcinogenic.77.42%Ultraviolet irradiationThe air velocity in HVAC system is very large. When it is applied in HVAC system, the irradiation dose on bacteria is very small, so the performance is poor. Only bacteria can be removed, but dust cannot. Ozone will be generated. WHO and GMP from EU claim that this method is not acceptable. It can not be used as the final disinfection method82.90%Electron sterilization lampPhysical method85%Double-stage electrostatic precipitatorThe ionization electrode is separate from the dust collecting electrode90% (test on some products show that the efficiency only reaches about 60%)OzoneIt is a light blue gas, its oxidation performance is strong. The oxygen atom generated during the oxidation process can oxidize and penetrate through the cell wall of the bacteria. It has broad-spectrum effect of germicide, but it cannot remove dust. During usage, occupant should not stay indoors. Various goods may be destroyed. Little effect will be exerted on surface microorganism. It is detrimental to the respiratory tract of people. This method is not suggested for usage91.82%High and medium efficiency air filter with ultra-low resistanceIt is a method with physical barrier. When it is used for common air supply outlet, the pressure difference is only about 10 Pa, which is one third of the value for coarse filter. But the efficiency is high or medium (for particles with diameter ≥0.5 μm, the efficiency reaches more than 70--80%). It is light and easy for installation, and there is no secondary-contamination92--98%HEPA filterIt is a method with physical barrier. There is no side effect. It is disposable. In the "*Technical Standard For disinfection*" issued by Ministry of Health of the People's Republic of China, only air filtration is proposed for air disinfection in cleanroom. The resistance is large99.9999--99.99999% (test with *Bacillus subtilis* in 2004) Table 3.16Bacteria removal efficiency by electrostatic sterilization equipmentSampling time after start-up, hSterilization efficiencyOperating roomICU0.561.2055.201.050.4073.201.557.0065.102.050.8086.503.078.8078.904.066.7078.60*Note* Bacterial concentration before start-up was 504--900 CFU/m^3^ Table 3.17Comparison on sterilization performance of different disinfection methodsDisinfection methodBacterial concentration before disinfection, CFU/m^3^Bacterial concentration after disinfection, CFU/m^3^Disinfection rate, %Note*Atractylodes lancea*64220468.22Experiment was performed by Wu Chun-lan, which was published in Chinese Journal of NosocomiologyOzone disinfection machine6735591.82Ultraviolet ray60110382.86Formaldehyde fumigation59813577.42 Table 3.18Data from Tianjin University \[[@CR19]\]UV lamp installed in ductV = 3 m/sOne-pass sterilization efficiency 80%Experiment was performed by Wang Ying, which was published in Journal of Tianjin University Table 3.19Report data by National Center for Quality Supervision and Test of Building EngineeringSterilization equipment with circulation-air and UV lamp installation in a room with area 11.6 m^2^Coarse, medium and sub-high efficiency air filters installed in the pipeline of the outdoor airThe sterilization rate indoors after start-up reaches 93%Experiment was performed by National Center for Quality Supervision and Test of Building Engineering Table 3.20The data from the thesis at China Academy of Building Research(1) One-pass sterilization efficiency of high and medium efficiency air filter with ultra-low resistanceAt the half-th minuteAt the eighth minuteAt the thirteenth minuteAt the nineteenth and a half minuteAt the twenty-first minuteAt the twenty-fifth minuteAverageExperiment was performed by National Center for Quality Supervision and Test of Building Engineeringf building engineering. Authors include Pan Hong-Hong, Cao Guo-Qing, et al.99.34%99.34%99.39%99.47%99.26%99.43%99.37%(2) Pressure drop of high and medium efficiency air filter with ultra-low resistanceAir velocity, m/s0.310.410.520.610.710.790.890.99Experiment was performed by National Center for Quality Supervision and Test of Building Engineering. Authors include Pan Hong-Hong, Cao Guo-Qing, et al.Pressure drop, Pa810121315171921(3) Efficiency of HEPA filterType B HEPA filter based on national standardType C HEPA filter based on national standardResearch report from China Academy of Building Research. Authors include Zhang Yi-Zhao, Yu Xi-Hua, et al.One-pass efficiency is 99.99997% for *B. Subtilis* sporesOne-pass efficiency is 99.999997% for *B. Subtilis* spores Table 3.21Efficiency of electrostatic air cleanerElectrostatic air cleanerThe first ionization (U.K.)One-pass dust removal efficiency is 80%\<*Fundamentals of air cleaning technology*\>Electrostatic air cleanerThe first ionization (Japan)One-pass dust removal efficiency is 72.8%Electrostatic air cleanerThe secondary ionization (China)One-pass dust removal efficiency is 99.3% Table 3.22Sterilization efficiency of electrostatic air cleanerElectrostatic air cleaner4 h after start-upSterilization efficiency in operating room is 66.7%Report by Yang Ming-Hua by Institute of Microbiology and Epidemiology at the Academy of Military Medical SciencesElectrostatic air cleaner0.5 h after start-upSterilization efficiency in ICU is 61.2%Thesis by Mao Hua-Xiong from Tong UniversityElectrostatic air cleaner1 h after start-upSterilization efficiency with atmospheric bacteria is 79.9%2 h after start-upSterilization efficiency with atmospheric bacteria is 91.1% 2.Features of disinfection method with HEPA filterWhole-process controlWhen air cleaning system with HEPA filter is used during the whole operation process indoors, the indoor air environment can be controlled. This is different from "sterilization at the beginning" or "sterilization at the end".Both dust and bacteria are removedBy using the principle of physical barrier, air filter can remove dust, as well as microbes (including bacteria and virus). Since microbes are carried by particles, bacteria can be removed during the removing process of dust.No generation of other gradientsSome sterilization method may generate toxic gases such as nitrogen oxides or ozone. However, air filtration is a pure physical method, which will not generate other gradients.No side effect of generating toxic materialSome sterilization method may generate radiation, which is harmful to occupant's health. Some generated electric field or magnetic field has influence on instrument and equipment. Some may promote the mutation of bacteria. For example, this will occur during the UV irradiation. The drug resistance of the bacteria may become very strong.High and complete sterilization efficiencyFor HEPA filter, the sterilization efficiency could reach more than 99.99999%. The sterilization performance is complete. There are no semi-lethal bacteria left, which may revive afterwards. For bacteria suffering UV irradiation, they may revive under the light if they are not killed by irradiation. Bacteria corpse or secretion will not left if HEPA filter is used.Broad-spectrum of sterilization efficiency and easy for selectionThe range of sterilization efficiency for different products is from 10 to 99.999999%, while that of other sterilization methods is very narrow, which is from 70 to 90%.It is not selective for dust and bacteria removalThe performance of some sterilization method such as the electrostatic method is influenced by the conductivity property of the dust particles. Some sterilization method is selective for the type of bacteria, which means the sensitivity level is different.For example, the UV-C ray with wavelength 253.7 nm has different sensitivity levels for various microbes, which is shown in Table [3.23](#Tab23){ref-type="table"} \[[@CR23]\].Table 3.23Sensitivity of various typical microbes for UV raySensitivity to UV rayGroup of microbesTypical microorganismThe most sensible             ↓             ↓             ↓             ↓             ↓             ↓             ↓             ↓             ↓             ↓             ↓             ↓             ↓             ↓             ↓The least sensibleEndophytic bacterium*Staphylococcus arueusStreptococcus pyogenesEscherichia coliPseudomonas aeruginosaSerratia marcescens*Mycobacteria*Mycobacterium tuberculosisMycobacterium bovisMycobacterium leprae*Spore bacteria*Bacillus anthracisBacillus cereusBacillus subtilis*Fungal spore*Aspergillus versicolorPenicillium chrysogenumStachybotrys chartarum*It is a method to prevent aerosol and airborne microbes from entering into the air ducts. It is equivalent to the method of preventing the enemy from invading into our country. It is the highest state of war to subdue the enemies without fight. This method will not cause corpses of the bacteria cover the plain. Therefore, it is an active pollution control method, and not the passive method to perform disinfection after bacteria come indoors. It is a green method to build the green hospitals. No new pollution will be generated. Of course, the main shortcoming of air filter is its relative large pressure drop. But pre-filter or filter for air supply outlet should also be installed for some methods, including the nano-photocatalysis and UV equipment. Air filters with lower pressure drop should be utilized. Related technique and products have been available.Classification of air filtersTable [3.24](#Tab24){ref-type="table"} shows the classification of air filters according to national standard GB/T 14295-2008.Table 3.24Performance classification of air filtersTypeSymbolFace velocity, m/sEfficiency *E* under rated flow rate, %Initial pressure drop ∆*P* ~i~ under rated flow rate, PaFinal pressure drop ∆*P* ~f~ under rated flow rate, PaSub-high efficiencyYG1.0Particle diameter ≥0.5 μm99.9 \> *E* ≥ 95≤120240High and medium efficiencyGZ1.595 \> *E* ≥ 70≤100200Medium efficiency 1Z12.070 \> *E* ≥ 60≤80160Medium efficiency 2Z260 \> *E* ≥ 40Medium efficiency 3Z340 \> *E* ≥ 20Coarse efficiency 1C12.5Particle diameter ≥2.0 μm*E* ≥ 50≤50100Coarse efficiency 2C250 \> *E* ≥ 20Coarse efficiency 3C3Arrestance with standard artificial dust*E* ≥ 50Coarse efficiency 4C450 \> *E* ≥ 10*Note* When measured efficiency value satisfies two types in this table, the higher classification level can be used for assessmentClassification for HEPA and ULPA filterClassifications for HEPA and ULPA filters according to national standard GB/T13554-2008 are shown in Tables [3.25](#Tab25){ref-type="table"} and [3.26](#Tab26){ref-type="table"}, respectively.Table 3.25Classifications for HEPA and ULPA filtersClassification for HEPA filterTypeEfficiency *E* with sodium flame method under rated flow rate, %Efficiency *E* with sodium flame method under 20% of rated flow rate, %Initial pressure drop ∆*P* ~i~ under rated flow rate, PaA99.99 \> *E* ≥ 99.9No requirement≤190B99.999 \> *E* ≥ 99.9999.99≤220C*E* ≥ 99.99999.999≤250Classification for ULPA filterTypeParticle counting efficiency *E* with 0.1--0.3 μm particles under rated flow rate, %Initial pressure drop ∆*P* ~i~ under rated flow rate, PaNoteD99.999≤250Leakage scanningE99.9999≤250Leakage scanningF99.99999≤250Leakage scanning Table 3.26Approximated comparison of several standards for air filters from China and abroadChinese standardEU standard Euvovent 4/9ASHRAE standard with arrestance, %ASHRAE standard with dust spot method, %U.S. DOP method (0.3 μm), %EU standard EN779German standard DIN 24185U.S. standard MERVCoarse filter 4EU1G1A1Coarse filter 3EU1\<65G1A2--4Coarse filter 2EU265--80G2B15--6Coarse filter 1EU380--90G3B27--8Fine filter 3EU4≥90G4B29--10Fine filter 2EU540--6020--55G5C111Fine filter 1EU660--80F6C1/C212High and medium efficiency filterEU780--9055--60F7C213High and medium efficiency filterEU890--9565--70F8C314High and medium efficiency filterEU9≥9575--80F9--14Sub-high efficiency filterEU10\>85F10Q15Sub-high efficiency filterEU11\>98F11R16HEPA filter AEU12\>99.9F12R/S17HEPA filter AEU13\>99.97F13S17HEPA filter BEU14\>99.997F14S/T18-19HEPA filter CEU15\>99.9997U15T19ULPA filter DEU16\>99.99997U16U--ULPA filter E--FEU17\>99.999997U17V--Comparison of classification standards between China and abroadThe test conditions for air filters in standards from China and abroad are quite different, including the particle source and its diameter. Only approximated instead of quantitative comparison can be performed, which is shown in Fig. [3.26](#Fig26){ref-type="fig"}. Experimental Validation for Application of HEPA Filter with Circulation Air {#Sec16} --------------------------------------------------------------------------- Experimental methodExperiment was performed in the aforementioned simulated isolation ward as shown in Fig. [3.17](#Fig17){ref-type="fig"}. Bacteria were released outside of the air filter for exhaust air. The bacterial concentration for spray was 8 × 10^8^ pc/mL. Leakage may occur on the frame of the air filter as shown in Fig. [3.31](#Fig31){ref-type="fig"}, which may not be detected beforehand and then sealed (a leakage-free exhaust air apparatus was invented, which will be introduced in detail later). In order to prevent the spread of aerosol into the room from the bacterial release position at the exhaust (return) air outlet, casing pipe must be used for bacteria generation inside.Fig. 3.31Leakage on the frame of air filter for exhaust (return) air If the frame of air filter is very air-tight, the casing pipe should contact air filter directly, which is shown in Fig. [3.32](#Fig32){ref-type="fig"} \[[@CR24]\]. In the figure, *Q* is the flow rate of the circulation air in the room, m^3^/h; *q* ~1~ is the air flowrate during the spray, m^3^/h; *q* ~2~ is the air flowrate sucked into the casing pipe, m^3^/h; *k* is the penetration of HEPA filter at the return air outlet, %; *C* is the bacterial generation rate, pc/h.Fig. 3.32Direct contact of casing pipe with air filter It is obvious that the aerosol concentration in the supply air is *kC*/*Q*. When there is leakage air on the frame of air filter or when the casing pipe does not closely contact the air filter as shown in Fig. [3.33](#Fig33){ref-type="fig"} \[[@CR24]\], is there any influence on the measured result?Fig. 3.33Leakage air on the frame of air filter and casing pipe In Fig. [3.33](#Fig33){ref-type="fig"}, *q* ~1~ is the air flowrate during the spray, m^3^/h; *q* ~2~ is the air flowrate sucked into the casing pipe, m^3^/h; *q* ~3~ is the flowrate of the leakage air, m^3^/h; *Q* is the flow rate of exhaust air through air filter, m^3^/h; *k* is the penetration of air filter,%; *C* is the bacterial generation rate, pc/h; *Q*/*X* is the flowrate inside the casing pipe (*Q*/*X* = *q* ~1~ + *q* ~2~, *Q* = *q* ~2~ + *Q*/*X*), where *X* is the ratio which is larger than 1. In this case, the concentration of entering air from return air opening is *C*/(*Q*/*X*), pc/m^3^. The penetrated aerosol in air is *kC*/(*Q*/*X*). The total quantity of aerosol penetrated is:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\frac{kC}{{\frac{Q}{X}}} \cdot \frac{Q}{X}/Q = \frac{kC}{Q}}$$\end{document}$$ This means that the concentration is the same as that from the supply air for the case when the casing pipe is required to contact air filter directly and when there is no leakage on frame of air filter. This proves that it is feasible to use casing pipe for bacterial generation. With the same method, bacterial concentration can be measured below the air supply outlet with the sampling method for planktonic bacteria, which is shown in Fig. [3.34](#Fig34){ref-type="fig"}.Fig. 3.34Measurement of planktonic bacteria at the air supply outlet 2.Experimental results The measurement condition is shown in Table [3.27](#Tab27){ref-type="table"} \[[@CR24]\]. The measurement results are presented in Table [3.28](#Tab28){ref-type="table"} \[[@CR14]\]. In the table, the measured concentration for the supply air at the air supply outlet is expressed with CFU/petri dish.Table 3.27Experimental conditionNo. of experiment123Type of HEPA filterBCBParticle counting efficiency of HEPA filter before delivery from factory (≥0.5 μm)99.99999.9999499.999Bacterial solution concentration (pc/mL)8 × 10^10^4.5 × 10^10^4.5 × 10^10^Spray volume of bacterial solution (mL/min)0.204 0.1590.153Air flowrate during bacterial generation (L/min)171717Number of vessel in Andersen samplerOne vessel in each layer, 2 layersOne vessel in each layer, 2 layersOne vessel in each layer, 2 layersFlow rate through Andersen sampler (L/min)28.328.328.3Number of vessel in centrifugal sampler (Type WL 1)One vessel in each layerOne vessel in each layerOne vessel in each layerFlow rate through centrifugal sampler (Type WL 1) (L/min)28.328.328.3Period of bacterial generation (min)306336Volume of bacterial solution (mL)6.12105.5Sampling position15 cm from the center of the air supply outlet15 cm from the center of the air supply outlet15 cm from the center of the air supply outletFlow rate of return air (m^3^/h)380223233 Table 3.28Measurement result for filtration efficiency with bacteriaType of air filterSampling methodSampling time, minConcentration at air supply outlet, CFU/dish or CFU/m^3^Bacterial generation rate at return air opening, CFU/hBacterial concentration released at return air opening, CFU/m^3^Filtration efficiency, %Average filtration efficiency, %No. of experimentBAndersen sampler141 (41 × 1000/28.3 = 1448)9.79 × 10^11^2.58 × 10^9^99.99994499.99998999.99994999.999971391 (91 × 1000/28.3 × 3 = 1072)99.9999585200 (200 × 1000/28.3 × 5 = 1414)99.999945BAndersen sampler15.17 (5.17 × 1000/28.3 = 182)4.14 × 10^11^1.8 × 10^9^99.9999999.99999199.999991334.5(4.5 × 1000/28.3 × 3 = 53)99.999992Andersen sampler14.7(4.7 × 1000/28.3 = 166)4.29 × 10^11^1.9 × 10^9^99.99999199.99999317(17 × 1000/28.3 × 3 = 200.2)99.999989CAndersen sampler34.8(4.8 × 1000/28.3 × 3 = 56.5)4.29 × 10^11^1.9 × 10^9^99.99999799.99999799.99999799.9999972Centrifugal sampler38(8 × 1000/28.3 × 3 = 94.2)99.99999599.999996Centrifugal sampler511.3(11.3 × 1000/28.3 × 5 = 80.1)4.29 × 10^11^1.9 × 10^9^99.999996 Hourly bacterial generation rate at the return air opening = Bacterial solution concentration per minute × Spray quantity of bacterial solution × 30 min × 2 Bacterial generation concentration at the return air opening = Bacterial generation rate at the return air opening ÷ flow rate of return air It is shown from Table [3.27](#Tab27){ref-type="table"} that for HEPA filter B, the filtration efficiency for the spray strain reached 99.99997%. For HEPA filter C, the filtration efficiency for the spray strain reached 99.999997%. Both efficiency of these two kinds of filters are larger than the measured efficiency with atmospheric dust. This is consistent with the aforementioned factor of equivalent diameter of microorganism. Filtration efficiency of HEPA filter C for bacteria is higher than that of HEPA filter B by one order of magnitude. This is also consistent with the relationship of the filtration efficiency with atmospheric dust between two products. Next quantitative analysis will be performed. The relationship between different particle diameters is given by empirical equation \[[@CR25]\].$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k_{2} =\frac{{k_{1} }}{{{\text{e}}^{{\left( {d/d_{0.3} } \right)^{2} }} }}}$$\end{document}$$where *k* ~1~ and *k* ~2~ are the penetrations for particle size 0.3 μm and particle size larger than 0.3 μm, respectively; *d* ~0.3~ and *d* are the particle size 0.3 μm and the particle size larger than 0.3 μm, respectively; From the above experiment shown in Table [3.27](#Tab27){ref-type="table"} \[[@CR14]\], the measured efficiency of HEPA filter B before delivery from factory for atmospheric dust with diameter ≥0.5 μm is 99.999%. From literature \[[@CR25]\], the efficiency with atmospheric dust for particle size 0.3 μm can be calculated to be 99.93%, which corresponds to the penetration *k* = 0.07%. The measured efficiency of HEPA filter C before delivery from factory for atmospheric dust with diameter ≥0.5 μm is 99.99994%. The efficiency with atmospheric dust for particle size 0.3 μm can be calculated to be 99.998%, which corresponds to the penetration *k* = 0.002%. From Chap. 10.1007/978-981-10-2923-3_5, water component in the sprayed droplet will evaporate quickly during the spray process, but residual solute will be left. Since the size of the sprayed droplet is usually 1--5 μm, the average size is 3 μm. From Chap. 10.1007/978-981-10-2923-3_5, the final size of the solute is 0.16 × 3 = 0.48 μm, so it should be added into the size of naked bacteria. Since the size of the spore is about 1 μm, the increase of the linear thickness on the surface can be ignored. Table [3.29](#Tab29){ref-type="table"} illustrates the efficiency for spores with different diameters.Table 3.29Efficiency for spores with different diametersAir filterCalculated efficiency, %Measured efficiency, %0.5 μm0.8 μm0.9 μm1.0 μmB99.995799.99994699.9999916≈10099.99997C99.9998899.999995299.9999993≈10099.999997 Based on the measurement result from Table [3.28](#Tab28){ref-type="table"}, the measured efficiency is within the range of the calculated efficiency for spores with diameter 0.8 and 0.9 μm. From Table [3.24](#Tab24){ref-type="table"}, the concentration at the air supply outlet, when HEPA filter C was applied, was in the order of the natural number magnitude. This means for such high efficiency in experiment, when the bacteria concentration was increased by an order of magnitude, the measured efficiency may become higher. The above measured result is equivalent with the foreign experimental data in Table [3.30](#Tab30){ref-type="table"} \[[@CR26]\]. The corresponding filtration velocity is only less than a half of the filtration velocity in Table [3.30](#Tab30){ref-type="table"}, so the efficiency value is slightly higher.Table 3.30Filtration efficiency of various air filters for serratia marcescens (the bacterial solution concentration for spray is 1.1 × 10^7^ pc/L)Type of air filterTimes of experimentsEfficiency, %Filtration velocity, m/sNoteDOP 99.97DOP 99.97201999.999999.9994 ± 0.00070.0250.025Equivalent to HEPA filter B in ChinaDOP 99.97DOP 95201799.996 ± 0.002499.989 ± 0.00240.0250.025Equivalent to sub-high efficiency air filter in ChinaDOP 75DOP 60DOP 4020202099.88 ± 0.017997.2 ± 0.29183.8 ± 1.0060.050.050.05Equivalent to high-medium efficiency air filter in ChinaDOP 20--301854.5 ± 4.9030.20Equivalent to or slightly better than fine filter in China The above efficiency values with the sodium flame method, DOP method and particle counting method are different. For particle counting method, the efficiency for particle size 0.3 μm is different from that for particle size ≥0.5 μm, which could be referred to literature \[[@CR25]\]. In short, the theoretical efficiency with actual particle size should be larger than the experimental data. So it is much safe to use experimental data. It is known that the number of aerosol during coughing could reach more than 3 × 10^5^. When ten times of this concentration, i.e., 3 million bacterial particles, pass through HEPA filter B, only one particle could penetrate. When one hundred times of this concentration, i.e., 30 million bacterial particles, pass through HEPA filter C, only one particle could penetrate. Therefore, the bacterial concentration in the circulation air through HEPA filtration unit is much smaller than indoor bacterial concentration. For patients staying in a room with such low bacterial concentration for a long time, it is not doubt that the influence of circulation air is little. This means it is feasible to apply HEPA filtration unit in isolation ward.

Introduction ============ When we open a newspaper, we often encounter headlines such as "Hometown driver now a local hero" or "A lot of heroes around here." We may also come across a headline that reads "While at war, female soldiers fight to belong" (New York Times, 25 May 2015). What images of men and women do such newspaper articles create? Does the language they contain influence these images? In our research, we studied the effects of using either only masculine or both masculine and feminine role nouns in newspaper articles. A large body of research documents that women are less visible in the media in general: only 13% of all news stories are about women ([@B15]). Furthermore, the media often depict women and men in a stereotyped manner, with 46% of news stories reinforcing gender stereotypes, and only 6% challenging such stereotypes ([@B15]). Gender stereotypes that prevail in a society are reflected in the media, but the media also influence how women and men are perceived in the respective society. Also, the way research findings are reported in the popular press may affect readers' beliefs and attitudes and may reinforce stereotyping. For example, a series of studies showed that readers of an article that stressed biological explanations of gender differences endorsed gender stereotypes more strongly than readers of a similar article that focused more on sociocultural explanations for gender differences ([@B4]). The image of women is not only influenced by what is said or not said, but also by how it is said. In grammatical gender languages, such as German, French, or Russian, nouns and pronouns have masculine and feminine forms and thus differentiate for gender, for instance, "he" vs. "she" or "hero" vs. "heroine." However, when referring to mixed-gender groups, to persons with unknown gender or persons whose gender is irrelevant, "masculine generics" are used, i.e., grammatically masculine nouns and pronouns ([@B12]). In contrast, gender-inclusive language makes explicit reference to women and men (word pairs, e.g., "he or she," "firemen and firewomen") or uses gender-neutral forms (e.g., "they," "firefighters," [@B19]). Past research has revealed that gender-inclusive language makes women more visible than masculine generics do ([@B19]). When gender-inclusive forms are used, people assume percentages of women in a profession to be higher than when masculine generics are used ([@B3]) and more female exemplars of a category are named ([@B21]; [@B16]) or drawn ([@B2]). Most research of this kind investigated common social roles (e.g., student, professor, physician, [@B2]) and only some studies tested more prominent roles (e.g., musician, politician, [@B21]). In one investigation, German participants listed more women when asked to name their favorite "heroine or hero in a novel"^[1](#fn01){ref-type="fn"}^ (word pair form; German: *Romanheldin oder Romanheld*) than those who were asked to name their favorite "hero in a novel" (*Romanheld*; [@B20]). In addition to the impact of language on mental representations of women and men, recent research has shown that after reading a text with gender-inclusive wording, speakers used more gender-inclusive forms themselves in a fill-in-the-blank task ([@B14]). While most research focused on the effects of gender-inclusive language on perceivers (readers or listeners), there is much less research on the effects of this usage on producing such language in writing or speaking. One study that addressed effects of gender-inclusive language on users showed that participants who were made to employ forms such as *he or she* in a sentence-completion task imagined more female characters as protagonists of the sentences they were completing than participants who were made to use generic *he* ([@B10]). The author does not explain the mechanisms underlying this effect, but it is conceivable that speakers' own use of gender-inclusive forms enhances the cognitive availability ([@B23]) and causes a more intensive processing of these forms and of their referents. We were interested in the impact of gender-inclusive language in media texts on mental representations of women and men and in the mechanism underlying this impact. To this end, we investigated reactions to newsworthy, exceptional social roles that are often dealt with in the media: hero and murderer. Both social roles attract much attention and have similarly low percentages of women (ca. 10--20%). In the US, only 9% of the recipients of the Carnegie Hero Medal for saving others are women, and in Germany only about 20% of similar medals are awarded to women. This may be because there are fewer women in professions such as firefighters, soldiers, or police officers---jobs involving dangerous situations where jobholders can act heroically. As for murderers, in 2014 women committed 11% of all homicides in the US ([@B6]); in Germany, where the present study was conducted, it was 9% ([@B22]). As mentioned above, reading a gender-inclusive text provoked more use of such language in a fill-in-the-blank task ([@B14]). To replicate and extend this finding, we examined whether the effect on language use also held for more natural forms of language production, namely for texts written entirely by the participants themselves. We presented participants with "a short science-based press release," and asked them to summarize the text in their own words and to indicate the percentage of women in the described social role. We expected participants summarizing a gender-inclusive text to use more word pairs and other gender-fair forms than those summarizing a text with masculine generics (Hypothesis 1). In line with past research ([@B21]), we expected that reading a text with gender-inclusive forms would result in higher estimates of the percentage of women in the respective roles than reading a text with masculine generics (Hypothesis 2). In addition, and more importantly, we examined whether participants' own use of gender-inclusive language would lead to a higher perceived percentage of women in a given social role. In the present study, we aimed at eliciting gender-inclusive forms by having participants read a text containing such forms and by asking them to summarize its content in their own words. Thus, we directly manipulated only the text, but not participants' language use *per se*. We expected that receiving a message in gender-inclusive wording would enhance speakers' inclination to use such forms themselves ([@B14]). This use, in turn, was expected to result in higher estimates of the percentage of women in the respective roles than reading a text with masculine generics ([@B10]). In other words, we expected speakers' own use of gender-inclusive forms to mediate the relationship between the forms appearing in the text and mental representations of women in the role described in the text (Hypothesis 3). The present investigation integrates hitherto separate lines of research on gender-inclusive language by studying three effects in the framework of one experiment: (1) the effect of reading gender-inclusive forms on speakers' own language use, (2) the effect of reading gender-inclusive forms on mental representations of women, and (3) the effect of own language use on mental representations of women. By integrating these aspects in one experiment, the present study enables identifying the process underlying effects of gender-inclusive language. In this way, it contributes to a more comprehensive understanding of how gender-inclusive language leads to a reduction of gender stereotyping and discrimination ([@B18]). Materials and Methods {#s1} ===================== Participants and Design ----------------------- We recruited participants for an online study via a local newspaper, university mailing lists, and during the *Long Night of Scientists* in Jena, Germany (an event popularizing science among the general public). Participants were 267 native speakers of German. After deleting answers of one participant who took 17 h, the average time for completing the survey was 20 min (*SD* = 17 min). We deleted six participants who completed the survey in less than 3 min (-1 *SD*) and five who took more than 54 min (+2 *SD*). The final sample consisted of 256 persons (170 women, 86 men). Participants were of different ages (range: 14--82 years, *M* = 32.30, *SD* = 14.32) and most of them (68%) were living in the German state of Thuringia. About half (55%) of them were university students from different departments. The experiment had a 2 (social role: heroes vs. murderers) × 2 (linguistic form: masculine generics vs. word pairs) design, with use of gender-inclusive language and estimated percentage of female heroes/murderers as dependent variables. The use of gender-inclusive language in participants' own writing served as a potential mediator of the relationship between linguistic forms used in the text and the estimated percentage of female heroes/murderers. Procedure and Measures ---------------------- The study was carried out in accordance with the recommendations of Swiss and German Human Research Acts. It was presented as "a study on the perception of science-based press releases." After providing their informed consent and indicating their age and gender, participants read one (randomly chosen) of four versions of a 250-word media text about the socialization of (1) heroes, (2) heroines and heroes, (3) murderers, or (4) murderesses and murderers. The texts were identical except for the role nouns, which varied in the heading and in three other places in the text. Each text was about a study that attempted to identify factors explaining why some people become heroes/murderers (see Supplementary Materials). After reading the text, participants were asked to sum it up in their own words in 3--4 sentences. To rule out that the text quality differed depending on linguistic forms used, we asked participants whether the information in the text was credible (accurate, credible, reliable, reflecting reality, α = 0.88) and whether the text was informative (convincing, informative, should be published, relevant, α = 0.87; scale: 1 = *totally disagree*, 7 = *fully agree*). The respective analyses showed that the texts were perceived as similarly credible and informative in both linguistic versions (*F*s \< 2.60, *p*s \> 0.10, $\eta_{p}^{2}$ \< 0.01). The texts about heroism were perceived as more credible than the texts about murderers, *F*(1,251) = 8.23, *p* = 0.004, $\eta_{p}^{2}$ = 0.03, but the texts about murders were perceived as more informative, *F*(1,251) = 8.23, *p* = 0.004, $\eta_{p}^{2}$ = 0.03. No interactions were observed, *F*s \< 1. Then we asked five "memory questions," including the dependent variable. The questions concerned facts and numbers that appeared in the text (How many participants took part in the study? What hobby did most of them have? How well did most of them do at school?), but also "What proportion of heroic deeds/murders are committed by women?" Finally, participants answered some additional questions, specified demographic data^[2](#fn02){ref-type="fn"}^, were debriefed, and given the possibility to make comments. Those who left their e-mail addresses had the chance to win three prizes of €20 each. Results ======= Use of Gender-Inclusive Language -------------------------------- After reading about heroism or murders, participants were asked to summarize the text in their own words. We coded the linguistic forms they used to refer to the roles of hero and murderer. The three linguistic forms most often used to describe the protagonists of the texts were: masculine forms (singular or plural, e.g., German *Helden* "heroes"), word pairs (e.g., German *Heldinnen und Helden* "heroines and heroes"), and neutralization (e.g., German *Menschen* "humans," *Leute* "people," *Personen* "persons"). For each participant we built an index: we computed the proportion of gender-inclusive language by dividing all gender-inclusive forms (word pairs and neutralization) by all forms used for person reference (both gender-inclusive and masculine). The resulting index ranged from 0 to 1; the higher the index, the more gender-inclusive forms were used by the participant. A 2 (social role: heroes vs. murderers) × 2 (linguistic form: masculine generics vs. word pairs) ANOVA showed that, in accord with Hypothesis 1, participants used more gender-inclusive language after reading texts with word pairs (*M* = 0.72, *SD* = 0.38) than after reading texts with masculine generics (*M* = 0.18, *SD* = 0.26), *F*(1,259) = 177.04, *p* \< 0.001, $\eta_{p}^{2}$ = 0.42 (see **Figure [1](#F1){ref-type="fig"}**). Furthermore, participants used more gender-inclusive language when writing about heroism (*M* = 0.51, *SD* = 0.42) than about murders (*M* = 0.39, *SD* = 0.42), *F*(1,259) = 9.53, *p* = 0.002, $\eta_{p}^{2}$ = 0.04. The interaction of social role and linguistic form was not significant, *F* \< 1. {#F1} Given the large age range of the sample (14--82 years), and the possibility of participant gender playing a role, we examined whether age and gender modulated the results. An ANCOVA showed no significant influence of gender (*F* \< 1) and a small effect of age, *F*(1,246) = 3.30, *p* = 0.07, $\eta_{p}^{2}$ = 0.01: younger participants used slightly more gender-inclusive forms. The remaining results were virtually the same with or without both covariates. Estimated Percentage of Female Heroes and Murderers --------------------------------------------------- A 2 (social role: heroes vs. murderers) × 2 (linguistic form: masculine generics vs. word pairs) ANOVA revealed that, in accord with Hypothesis 2, participants who read the texts with word pairs estimated a higher percentage of women in both roles (*M* = 35.93; *SD* = 16.98) than participants who read the texts with masculine generics (*M* = 31.72; *SD* = 16.49), *F*(1,231) = 5.53, *p* = 0.02, $\eta_{p}^{2}$ = 0.02 (see **Figure [2](#F2){ref-type="fig"}**). Furthermore, participants estimated a higher percentage of women among heroes (*M* = 42.48; *SD* = 14.57) than among murderers (*M* = 24.37; *SD* = 13.78), *F*(1,231) = 98.87, *p* \< 0.001, $\eta_{p}^{2}$ = 0.30. An interaction effect, *F*(1,231) = 5.00, *p* = 0.03, $\eta_{p}^{2}$ = 0.02, showed that those who read about "heroines and heroes" estimated a higher percentage of female heroes than those who read about "heroes," *p* = 0.001. There was no difference in the estimated percentage of female murderers between the two language conditions, *p* = 0.94. Also, the more gender-inclusive forms participants used in their earlier summaries, the more women they perceived in the respective social role, *r*(231) = 0.26, *p* \< 0.001. {#F2} An ANCOVA including participants' gender and age again failed to show a significant influence of gender, *F*(1,246) = 2.67, *p* = 0.10, $\eta_{p}^{2}$ = 0.01, but revealed an influence of age, *F*(1,246) = 12.88, *p* \< 0.001, $\eta_{p}^{2}$ = 0.05: younger participants perceived more women in the social roles described in the texts. However, the age effect was not strong and the results were the same with or without covariates. The Mediating Role of Gender-Inclusive Language Use --------------------------------------------------- Our hypothesis predicted that gender-inclusive language would make participants use more of such language themselves, think in more gender-inclusive ways, and, in turn, estimate a higher percentage of women in the respective roles (Hypothesis 3). Therefore, we examined the potential mediating effect of own language use in the relationship between linguistic form used in the media text and perceived percentage of women in a role. As the results above showed, gender-inclusive language had a similar effect on speakers' own use of such forms for both social roles (heroes and murderers), but affected the perceived percentage of women only among heroes. In view of this finding, we conducted a moderated mediation model ([@B11], p. 369), with social role as a moderator. The analysis showed that after reading the media text containing word pairs participants used more gender-inclusive forms themselves (see **Figure [3](#F3){ref-type="fig"}**), in line with our previous findings. The more gender-inclusive language they used, the more women they perceived in this role. As shown above, the social role described in the text did not affect participants' own use of gender-inclusive language (no interaction of linguistic form and social role), but affected the perceived percentage of women (higher percentage of women for heroes, but not for murderers). Beyond previous findings and in accord with Hypothesis 3, the analysis revealed that when own use of gender-inclusive language was included, the direct effect of linguistic forms in the text disappeared for the category heroes and remained insignificant for murderers. Indirect (i.e., mediational) effects occurred for both roles. Thus, having read a text with word pairs, participants used more such forms themselves, and this, in turn, promoted more gender-balanced representations in both roles. {#F3} Post-test --------- The main effects for estimated percentage of women in the two social roles resembled the results obtained for own language use: participants used more gender-inclusive language and estimated a higher percentage of women after reading the texts with word pairs and after reading the texts about heroism. However, gender-inclusive language increased only the estimated percentage of female heroes, but not the percentage of female murderers. It has been suggested that the number of female exemplars available in a person category may play a role for the changes in gender perceptions that are evoked by gender-inclusive language ([@B3]). Thus, the observed asymmetry in our results may be due to a different availability of female exemplars among heroes and murderers. To address this possibility, we conducted a post-test. The post-test assessed how often people thought of women when asked about heroism and murders, without variation in language forms. We asked passers-by (*N* = 35, 21 women, 14 men, *M*~age~ = 35.17, *SD* = 14.82, age range: 19--74) in Jena, Germany, to fill out a short questionnaire in exchange for a chocolate bar. The instruction given in a booklet asked participants to name "people who behaved heroically" and, on the next page, "people who murdered someone" (order counterbalanced). Next, we explicitly asked participants to write down "women who behaved heroically" and "women who killed someone." The results showed that participants spontaneously named a total of 82 (*M* = 2.34) male and 12 (*M* = 0.34) female murderers, as well as 45 (*M* = 1.29) male and 22 (*M* = 0.63) female heroes. Thus, they mentioned more men than women for both roles, but the proportion of women was higher for heroes (33%) than for murderers (13%). Furthermore, participants named more female heroes than female murderers, while the opposite was true for male heroes and murderers. When asked explicitly about women, they again named more female heroes (37) than female murderers (23). In all, the post-test revealed that more exemplars of female heroes were available to the participants than of female murderers. Discussion ========== The present study shows that the language used in the media to describe social roles affects readers' own language use and that this, in turn, can influence the social perception of groups. In line with and extending earlier research, which relied on fill-in-the-gap sentences ([@B14]), participants used more gender-inclusive forms after reading a text with word pairs when summarizing the text in their own words. Moreover, participants who read about "heroines and heroes" estimated a higher percentage of female heroes than those who read about "heroes." Most importantly, we found that gender-inclusive language triggered a more gender-balanced use of language, and this use resulted in a more gender-balanced perception of social roles. In this process, participants' own language use functioned as a mediator between the language of the media texts and the mental representation of social roles. Thus, our results not only replicate and link previous findings ([@B10]; [@B21]; [@B14]), but also illuminate the mechanism behind the effects of gender-inclusive language. Participants' own use of gender-inclusive language made them think in more gender-inclusive ways. While readers of a text about "heroines and heroes" perceived more female heroes than readers of a text about "heroes," the estimated percentage of women committing murder did not differ between readers of a text about "murderesses and murderers" and about "murderers." Our post-test showed that people were aware of more female heroes than murderers. It seems that when people know some women in a given role ([@B1]; [@B17]), gender-inclusive language can trigger the female exemplars that are known and can make the mental representation of a role more gender-balanced ([@B3]). In other words, language can impact speakers' perceptions to some extent, but only within the boundaries of social reality. Other explanations for the observed effect are also conceivable. It is possible, for instance, that norm congruence plays a role here. Speaking of women as murderers might be too negative and less reconcilable with the female stereotype of a caring, nurturing, and selfless mother, than speaking of women as heroes. Perhaps gender-inclusive language makes women more visible in positive roles, but not in negative ones. However, our participants generally overrated the percentage of murderesses, which suggests that they were not trying to protect women's image. Furthermore, earlier studies have shown that gender-inclusive language has an effect also when talking about disliked personalities (e.g., the least liked politician, [@B9]; [@B8]). Still another explanation of the observed effect could be the different flexibility of the definition of both roles. Perhaps category boundaries are clearer and more straightforward for "murderers" than for "heroes." It might be easier to extend the definition of a hero to include more women than to extend the definition of a murderer ([@B1]; [@B7]). Although the effect of gender-inclusive language on the estimated percentage of women was significant only for heroes but not for murderers, the mediation effect occurred for both roles. The mediation findings suggest that triggering people to use gender-inclusive forms in their writing can make them process information more intensively and enhance cognitive availability of these forms and their referents ([@B23]). In addition, participants may have inferred that because they used gender-inclusive forms in their summaries themselves, there must be a considerable percentage of women in these roles. This self-perception mechanism ([@B5]) could be operating in parallel to a more intensive processing due to own use of gender-inclusive language. Future research should aim at replicating the obtained results and at further contributing to an explanation of the mechanisms underlying them. It could be also studied whether similar effects are present while listening to gender-inclusive language and while using it in one's own speech. Potential mediating variables could be assessed and/or manipulated. The accessibility of exemplars, for instance, could be manipulated by carefully choosing roles that differ in this respect, but not in valence, or by experimentally making some exemplars more accessible. The influence of the roles' valence could be studied by investigating roles with comparable numbers of known female exemplars, but with a different valence. To generalize the effects of valence, it would be advisable to include several positive and negative roles. Future research could also test whether the effect of texts in gender-inclusive language on the own use of such language and on the mental representations "spills over" to other social roles than the ones provided. Gender-inclusive language can be used to reduce gender-stereotypic images of certain male-typed social roles in a community with a grammatical gender language ([@B13]). Our results show that both perceiving and producing gender-inclusive language can evoke more gender-balanced mental representations of social roles. Such language can be effective not only when read or heard, but it can expand and resonate later when reproduced. The present study also suggests that reality will not be distorted if the media use more gender-inclusive language, but that this type of language may help to present women and men more equally in various social roles. Author Contributions ==================== All authors listed, have made substantial, direct and intellectual contribution to the work, and approved it for publication. Conflict of Interest Statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer SM and handling Editor declared their shared affiliation, and the handling Editor states that the process nevertheless met the standards of a fair and objective review. **Funding.** This research was supported by the European Commission's Seventh Framework Programme (FP7/2007-2013, grant agree ment 237907) and by public funds for scientific research in Poland (BST 174437/2015). When writing this article, KH was supported by the Polish National Science Centre (NCN) grant (DEC-2013/08/S/HS6/00573) and the Foundation for Polish Science (FNP) scholarship (START 030.2015-W). We thank Conni Winkler for advertising the study, and Friederike Braun, Lara Benteler, Agata Ostrowska, and the reviewers for their comments on an earlier version of this manuscript. We use the expressions "heroines" along with "heroes" as well as "murderesses" along with "murderers" to render the German wording. However, we are aware that in English these expressions are not as gender-fair as in German. We also included a few measures that are not the focus of the current article. The description and basic results on these measures are available in the Supplementary Materials. Supplementary Material ====================== The Supplementary Material for this article can be found online at: <http://journal.frontiersin.org/article/10.3389/fpsyg.2016.00369> ###### Click here for additional data file. [^1]: Edited by: *Manuel Carreiras, Basque Center on Cognition, Brain and Language, Spain* [^2]: Reviewed by: *Simona Mancini, Basque Center on Cognition, Brain and Language, Spain; Margo J. Monteith, Purdue University, USA* [^3]: ^†^Present address: *Cindy Littwitz, University of Hagen, Hagen, Germany* [^4]: This article was submitted to Language Sciences, a section of the journal Frontiers in Psychology

Introduction {#s1} ============ Influenza A virus (IAV) infection is a global health threat to livestock and humans, causing substantial mortality and morbidity. Pandemic and avian-like H1N1 and H3N2 IAV are endemic in pigs and humans in addition to H1N2 in pigs. The human pandemic 2009 H1N1 strain is also found in pigs, and human viruses or human-origin gene segments frequently adapt to transmit efficiently in pigs. Thus, pigs play a critical role in the emergence and epidemiology of novel IAV. Apart from the economic loss to farmers attributable to IAV infections of pigs, the generation of new reassortant strains of IAV poses a zoonotic threat to humans. Immunization is a cost-effective control measure to combat influenza infection, but the rapid evolution of the virus is a major obstacle. Conventional inactivated IAV vaccines are strain specific and protection correlates with the presence of neutralizing Ab. However, IAV infection and live attenuated viral vaccines have been shown to induce a degree of cross-protection in several species ([@r1]), suggesting that it may be possible to develop an effective "universal vaccine" for use in pigs and humans. Influenza virus in which the hemagglutinin (HA) signal sequence is suppressed (S-FLU) is a candidate universal vaccine. Suppression of the signal sequence limits expression of the viral encoded HA protein to the cytosol. The HA protein coating the S-FLU virus is provided from a transfected complementing cell line by pseudotyping ([@r2]), and is expressed from a codon-optimized cDNA lacking 5′ and 3′ untranslated regions of the viral RNA. This design allows infection by the vaccine virus to occur once only, resulting in attenuation. Replication of vaccine in the lung or nose is prevented, but all of the conserved viral proteins are expressed in the cytosol of S-FLU--infected cells for optimal Ag presentation ([@r3]). Immunization with S-FLU initiates broadly reactive cell-mediated immune responses to conserved viral Ags shared by all influenza viruses, which limit viral replication in the lungs of mice and ferrets even when heterologous challenge IAV are used ([@r2], [@r4]). Neutralizing Ab is not induced. In this study, we tested the effect of S-FLU--expressing pandemic origin H1 or H5 HA on viral load and pathology after challenge with IAV-expressing homologous or heterologous HA in pigs. Materials and Methods {#s2} ===================== S-FLU vaccine and influenza challenge virus {#s3} ------------------------------------------- The design and production of S-FLU, that is, \[S-eGFP/N1(Eng)\].H1(Eng), have been described previously ([@r2]). We used two vaccines, H1 and H5 S-FLU. H1 S-FLU expresses the HA and neuraminidase (NA) of the A/England/195/2009(pandemic \[pdm\]H1N1) (N1 <http://www.ncbi.nlm.nih.gov/nuccore/GQ166659> and surface H1 HA <http://www.ncbi.nlm.nih.gov/nuccore/237689564>) and internal protein genes of PR8. H5 S-FLU expresses the HA of the highly pathogenic avian influenza virus (HPAIV) A/Vietnam/1203/2004 (H5N1 clade 1, VN/04) with the NA and internal protein genes of PR8 ([@r4]). The H5 HA cDNA was codon optimized and the polybasic site was replaced with a trypsin cleavage site. The pig isolate of A/Swine/England/1353/09(pdmH1N1) (A/Sw/Eng/1353/09) was provided by Dr. Sharon Brookes (Animal and Plant Health Agency, U.K.). Virus stocks were propagated in the allantoic cavities of 9- to 11-d-old embryonated specific pathogen-free hens' eggs and used for viral challenge. For all serological and immunological assays the virus was propagated in Madin--Darby canine kidney (MDCK) cells. Animals and immunization {#s4} ------------------------ All experiments were approved by the ethical review processes at the Pirbright Institute and the University of Bristol, according to the U.K. Animal (Scientific Procedures) Act 1986. Twenty-four twelve-week-old Landrace cross female pigs were obtained from a commercial high health status herd. Pigs were screened for absence of influenza A infection by matrix gene real-time RT-PCR and for Ab-free status hemagglutination inhibition using four swine influenza virus Ags. Pigs were randomly divided into four groups of six and immunized as follows: 1) control received 4 ml DMEM, 0.1% (w/v) BSA, and 1× penicillin and streptomycin intratracheally (i.t.), 2) H1 S-FLU was administered i.t. at 2 × 10^7^ tissue culture--infective dose (TCID)~50~ in 4 ml (H1 i.t.), 3) H1 S-FLU was administered by aerosol (aer) by delivering ∼6.85 × 10^6^ TCID~50~ in 1 ml (H1 aer), and 4) H5 S-FLU was administered i.t. at 8 × 10^7^ TCID~50~ in 4 ml (H5 i.t.). The animals received an identical booster immunization 28 d later. For i.t. immunization, following sedation, a 20-gauge needle was inserted through the skin and wall of the trachea cranial to the sternum and for aerosol immunization an InnoSpire Deluxe Philips Respironics nebulizer was fixed to a small-sized anesthetic mask held over the animal's nose and mouth. The aerosol delivery process was designed to deliver 6.85 × 10^6^ TCID~50,~ although the actual retained dose in the lungs cannot be measured directly. However, we have determined that the nebulization process results in between 20 and 50% loss of viability of either S-FLU or the challenge virus, which means that the actual dose of viable S-FLU administered by aerosol is lower. For logistical reasons, two influenza virus challenges were performed, with half of the animals challenged at 31 d postboost (dpb) and the remainder at 45 dpb. Pigs were rehoused so that each challenge pen contained one animal from each immunization group. Animals were challenged intranasally with 6 × 10^6^ 50% egg infectious dose, equivalent to 1.5 × 10^5^ PFU/pig of A/Sw/Eng/1353/09. Two milliliters was administered to each nostril using a mucosal atomization device (MAD300; Wolfe Tory Medical). Pathological and histopathological examination of lungs {#s5} ------------------------------------------------------- Animals were humanely killed 3 d postchallenge (dpc) with an overdose of pentobarbital sodium anesthetic. The lungs were removed and digital images taken of the dorsal and ventral aspects. Macroscopic pathology scoring was performed blind at the University of Bristol using ImageJ v1.46r software (National Institutes of Health) to determine the percentage of the total surface area of the lung (dorsal and ventral aspects) affected by consolidation. Lung tissue samples were collected into 10% neutral buffered formalin for routine histological processing (University of Bristol). Formalin-fixed tissues were paraffin wax embedded and 5-μm sections were cut and then stained with H&E. Microscopic changes in the sections of lung were scored by a veterinary pathologist blinded to the treatment group. Five parameters were each scored on a 5-point scale of 0--4 and then summed to give a total slide score ranging from 0 to 20. Scoring criteria ([Supplemental Table I](#DC1){ref-type="supplementary-material"}) were based on a previously published method ([@r5]) with modifications. In cases where more than one lung sample was taken from an individual pig (e.g., lesion and nonlesion sites), the section with the highest composite score was used for subsequent statistical analysis. In these cases, this always corresponded to a sample containing a lung lesion that was visible on gross examination. Tissue sample processing {#s6} ------------------------ Two nasal swabs (one per nostril) were taken at 0, 1, 2, and 3 dpc. The swabs were placed into 2 ml virus transport medium comprising tissue culture medium 199 (Sigma-Aldrich) supplemented with 25 mM HEPES, 0.035% sodium bicarbonate, 0.5% BSA, penicillin, streptomycin, and nystatin, vortexed, centrifuged to remove debris, and stored at −80°C for subsequent virus titration. Serum and citrate anticoagulated blood samples were collected at the start of the study (prior to immunization), at days 0, 7, 13, and 28 dpb and 0 and 3 dpc. Citrate blood samples were diluted 1:1 in PBS before density gradient centrifugation at 1200 × *g* for 30 min over Histopaque 1.083g/ml (Sigma-Aldrich). PBMCs were harvested from the interface, washed, and RBCs were lysed with ammonium chloride lysis buffer, washed again, and cryopreserved in FCS (Life Technologies) with 10% DMSO (Sigma-Aldrich). Bronchoalveolar lavage (BAL) was collected from the entire right lung lobe with 150 ml virus transport medium (described above). BAL cells were isolated by centrifugation of the lavage fluid at 800 × *g* for 15 min. The supernatant was removed, aliquoted, and frozen for Ab analysis, whereas the cell pellet was washed in PBS and filtered through a 70-μm cell strainer and cryopreserved. Tracheobronchial lymph nodes (TBLN) were dissected out. TBLN cells were dissociated into a single-cell suspension with the gentleMACS Octo dissociator (Miltenyi Biotec, Woking, U.K.), using C tubes (Miltenyi Biotec) with 3 ml RPMI 1640. The single-cell suspension was filtered twice using a 70-μm cell strainer, washed, and RBCs were lysed. Cells were washed again and cryopreserved. The accessory lung lobes were dissected out and frozen at −80°C for subsequent virus titration. The whole lobe was homogenized using the gentleMACS Octo dissociator, the homogenate was clarified by centrifugation, and supernatant was used for virus titration. Virus titration {#s7} --------------- Viral titers in nasal swabs and the lung accessory lobe were determined by plaque assay on MDCK cells. Samples were 10-fold serially diluted and 100 μl was overlaid on confluent MDCK cells in 12-well tissue culture plates. After 1 h, the plates were washed and 2 ml 2% agarose/medium (1:3) was overlaid. Plates were incubated at 37°C for 48 h and plaques visualized using 0.1% crystal violet. Ab ELISA, HA inhibition, and microneutralization assay {#s8} ------------------------------------------------------ Influenza-specific Ab titers in serum and BAL fluid were determined by HA inhibition (HAI) using standard protocols ([@r6]). Briefly, H1 HAI Ab titers were determined using 0.5% chicken RBCs and A/Sw/Eng/1353/09 at a concentration of 4 HA units/ml. Neutralizaing Ab titers were determined in serum and BAL fluid using a microneutralization assay as described in the *WHO Manual on Animal and Influenza Diagnosis and Surveillance*. Briefly, samples were diluted 10-fold and heat inactivated at 56°C for 30 min, and further serially diluted 2-fold in PBS. Samples were then incubated with 4 × 10^4^ PFU/ml A/swine/England/1353/09(pdmH1N1) in serum-free DMEM for 1 h at 37°C in 5% CO~2~. The virus/sample mix was then added onto confluent MDCK cells in a 96-well plate and incubated at 37°C in 5% CO~2~. After 1 h, the plates were washed and overlaid with serum-free DMEM containing 2 μg/ml TPCK trypsin. Plates were incubated at 37°C for 48 h and cytopathic effect was visualized using 0.1% crystal violet. Titers were determined as the final dilution of serum that prevents cytopathic effect on MDCK cells following incubation with virus. The IgG and IgA ELISAs were performed in 96-well ELISA plates (BD Biosciences) coated with a preoptimized concentration of A/swine/England/1353/09(pdmH1N1). Two-fold dilutions of BAL fluid samples were added starting from 1:2 dilution. Binding of Abs was detected using preoptimized concentrations of polyclonal goat-anti pig IgG HRP or IgA-HRP (AbD Serotec) and tetramethylbenzidine substrate (BioLegend). Ab values were expressed as endpoint titers defined as the highest dilution at which the OD was higher than twice the background OD. IFN-γ ELISPOT {#s9} ------------- Frequencies of IFN-γ secreting in PBMCs, BAL, and TBLN cells were determined by ELISPOT using either fresh or cryopreserved cells. MultiScreen-HA ELISPOT plates (Merck Millipore) were coated with 0.5 μg/ml anti-pig IFN-γ (clone P2G10; BD Pharmingen) in carbonate buffer and incubated at 4°C overnight. The plates were washed five times in PBS and blocked using 4% (w/v) milk powder in PBS for 2 h. After five washes in PBS, 5 × 10^5^ cells were seeded in triplicate wells and stimulated with either live MDCK cell--grown A/Sw/Eng/1353/09 (multiplicity of infection \[MOI\] of 6), medium control, or 10 μg/ml Con A (Sigma-Aldrich). Plates were incubated overnight or for 40 h at 37°C in a 5% CO~2~ incubator, depending on whether fresh or frozen cells were used, followed by washes with PBS, 0.05% Tween 20, and addition of 0.25 μg/ml anti-pig biotinylated IFN-γ detection Ab (clone P2C11; BD Pharmingen). Plates were incubated for 2 h at room temperature, washed five times, and streptavidin--alkaline phosphatase (Invitrogen) was added for a further 1 h at room temperature. Spots were visualized using an alkaline phosphatase substrate kit (Bio-Rad) and the reaction was stopped using tap water. Immunospots were counted using the AID ELISPOT reader (AID Autoimmun Diagnostika). Results were expressed as number of IFN-γ--producing cells per 10^6^ cells after subtraction of the average number of IFN-γ^+^ cells in medium control wells. Flow cytometry {#s10} -------------- Cryopreserved mononuclear cells from blood, TBLN, and BAL were thawed and stimulated for 12 h at 37°C with live MDCK cell--grown strain A/Sw/Eng/1353/09 (MOI of 6 for BAL and PBMCs, MOI of 0.6 for TBLN) or MDCK cell mock supernatant as control. GolgiPlug (BD Biosciences) was added according to the manufacturer's instructions for a further 5 h before intracellular cytokine staining. Cells were stained for surface markers with CD3ε-PeCy5 PPT3 (AbCam), biotinylated CD4 clone MIL17 (in-house), with secondary streptavidin-allophycocyanin (SouthernBiotech), CD8α-FITC MIL12 (AbD Serotec), and near-IR fixable Live/Dead stain (Invitrogen). Cells were permeabilized using Cytofix/Cytoperm (BD Biosciences) as per the manufacturer's instructions before intracellular staining with IFN-γ--PE P2G10 (AbD Serotec) and cross-reactive anti-human TNF-α--Pacific Blue Mab11 (BioLegend). Samples were fixed in 1% paraformaldehyde before analysis using an LSRFortessa instrument (BD Biosciences). Data were analyzed using FlowJo v10 (Tree Star), and fluorescence minus one primary Ab controls were used to set gates. Samples were batch gated ([Supplemental Fig. 1](#DC1){ref-type="supplementary-material"}) on lymphocytes based on side scatter area/forward scatter area, followed by single cells on forward scatter height/forward scatter area. Live CD3^+^ cells were analyzed for expression of CD4 and CD8α. Boolean gating was used to determine the levels of IFN-γ and TNF-α expression in CD8α^hi^, CD4CD8α double-positive, and CD4 T cell subsets. Postcytometric multivariate data analysis was performed using SPICE version 5.3 (<http://exon.niaid.nih.gov>.). Statistical analysis {#s11} -------------------- One-way or two-way ANOVA with a Dunnett posttest for multiple comparisons were used to compare immunized groups to the control group, and analysis was performed using GraphPad Prism 6. A general linear model (GLM), together with Tukey honestly significant difference and least significant difference post hoc tests, was used to identify any significant differences between groups for individual variables. For pen and batch, when *p* \> 0.2, the variable was excluded from the GLM in a stepwise fashion. Samples taken during multiple days were analyzed using a repeated measures GLM. A nonparametric Freidman rank test was used when it was not possible to transform data to normality. A principal component analysis was performed using 17 factors (including virus titer, ELISPOT, pathology, and flow cytometry results) to explain overall variation between groups. Statistical analysis was performed using The R Project for Statistical Computing (version 3.0.1) (<http://www.r-project.org/>), IBM SPSS statistics software (version 21.0), and the GGobi data visualization system (version 2.1.10a) (<http://www.ggobi.org/>). Results {#s12} ======= Pulmonary immunization with S-FLU reduces viral load in nasal swabs and lungs {#s13} ----------------------------------------------------------------------------- S-FLU has previously been shown to protect mice and ferrets against homologous and heterologous IAV challenge, and it has been demonstrated that to confer protection, the vaccine needs to be delivered to the lung ([@r2], [@r4]). We therefore administered S-FLU either i.t. or by aerosol with a nebulizer to reach the lower respiratory tract (LRT) of the pigs. We used two S-FLU vaccines: H1 S-FLU expressing the HA and NA of A/England/195/2009(pdmH1N1), and H5 S-FLU expressing the HA of the highly pathogenic avian influenza virus A/Vietnam/1203/2004(H5) and N1 from A/PR/8/34. Groups of six pigs were immunized with two doses of H1 S-FLU i.t. (H1 i.t.) or by aerosol (H1 aer) or with H5 S-FLU i.t. (H5 i.t.); the control group received medium i.t. (control). Because of the different viral titers of the H1 and H5 S-FLU vaccine stocks and different delivery routes, different doses of the viruses were administered. H5 i.t. animals received 8 × 10^7^ TCID~50~, and the H1 aer group received a lower dose of ∼6.85 × 10^6^ TCID~50~. Most likely the dose actually received was even lower owing to losses during the aerosolization and administration procedure. Animals were challenged with the swine pdm09 strain A/Sw/Eng/1353/09 at either 31 or 45 d after the second immunization and euthanized 3 dpc to determine the effect of immunization at the peak of virus replication. Despite immunization with \>1 log less H1 S-FLU, the pigs immunized by the aerosol route showed the most efficient reduction of challenge virus in the nasal swabs at 1, 2, and 3 dpc ([Fig. 1A](#fig01){ref-type="fig"}). H1 S-FLU administered i.t. significantly reduced viral shedding at 2 and 3 dpc (*p* = 0.001), whereas H5 S-FLU reduced viral shedding at 2 dpc (*p* = 0.016) and completely prevented shedding in two individual animals at 3 dpc ([Fig. 1A](#fig01){ref-type="fig"}). Aerosol administration of H1 S-FLU also significantly reduced the viral titer in the accessory lobe of the lung (*p* = 0.006) ([Fig. 1B](#fig01){ref-type="fig"}). {#fig01} No ill effects were observed in any of the pigs immunized with H1 or H5 S-FLU by either method of administration. Gross pathology after viral challenge was minimal in most immunized animals; however, the control group contained two individuals with severe gross pathology ([Fig. 1C](#fig01){ref-type="fig"}). Most lung sections had total scores of ≤4 of a maximum score of 20, showing that there were only very subtle histopathological changes. The pigs with the highest scores belonged to the control or to the H5 IT treatment groups ([Fig. 1D](#fig01){ref-type="fig"}), but no significant difference between groups was detected. In addition to the parameters scored, several lung sections showed moderate levels of hemorrhage (data not shown) that was considered most likely to be a postmortem artifact. Furthermore, collapse of alveolar spaces frequently made assessment of alveolar septal thickness difficult. However, these issues simply highlight amendments that can be considered for sample collection for future studies. These results indicated that administration of S-FLU in pigs reduced the viral load in nasal swabs and lungs after closely matched viral challenge and that this was most efficiently induced by aerosol administration. H5 S-FLU reduced the viral load in nasal swabs and lung following heterologous viral challenge, but this was significant in nasal swabs only at 2 dpc. Influenza-specific immune responses in PBMCs {#s14} -------------------------------------------- As the analysis of samples from pigs challenged at days 31 and 45 did not reveal any significant differences ([Fig. 1](#fig01){ref-type="fig"}), for simplicity in presentation the results of the immunological assays carried out on pigs challenged on both days have been amalgamated. Because most commercial vaccines used to control influenza virus elicit a strong Ab response in the host, we first determined whether the reduced viral replication after S-FLU immunization resulted from a humoral immune response. HAI and microneutralization titers in serum and BAL fluid of all immunized groups were comparable with unimmunized controls at 3 dpc ([Table I](#tI){ref-type="table"}), consistent with previous studies showing lack of neutralizing Ab responses following S-FLU immunization of mice and ferrets ([@r2], [@r4]). Influenza-specific IgG and IgA were also quantified by ELISA in BAL fluid samples at 3 dpc and similarly very low endpoint titers for IgG and IgA were detected ([Table I](#tI){ref-type="table"}). ###### Serum and BAL fluid A/swine/1353/09(pdmH1N1)--specific HAI, MN, IgG, and IgA titers Group Serum BAL Fluid BAL Fluid 3 dpc --------- ------------ ------------ ----------------- ------ ------------ ------ ----------- ----------- Control 8.0 ± 0.0 7.3 ± 1.5 \<10 \<10 4.0 ± 0.0 \<10 0.8 ± 1.1 0.0 ± 0.0 H1 i.t. 9.6 ± 3.2 14.7 ± 8.5 \<10 \<10 5.3 ± 1.9 \<10 3.6 ± 7.0 2.4 ± 2.2 H1 aer 11.2 ± 4.0 14.7 ± 3.0 \<10 \<10 5.6 ± 2.0 \<10 3.0 ± 7.0 2.0 ± 2.0 H5 i.t. 8.0 ± 0.0 10.0 ± 4.5 \<10 \<10 12.0 ± 4.0 \<10 8.0 ± 8.8 2.0 ± 1.3 HAI titers were determined in serum at 0 and 3 dpc and BALF after sacrifice at 3 dpc using 4 HA units of A/Sw/Eng/1353/09 virus. Results shown are the mean of six animals ± SD. As a positive control, serum from pigs immunized with commercial A(H1N1)pdm09 vaccine Pandemrix (GlasxoSmithKline) and challenged with A/England/195/09(pdmH1N1) virus as described ([@r6]) was used and gave a titer of 2048. MN titers were determined in serum at 13 dpb and 3 dpc and BALF after sacrifice at 3 dpc using 100 PFU A/Sw/Eng/1353/09 virus. Results shown are the mean of six animals. Serum and BALF collected from A/Sw/Eng/1353/09 virus--infected pigs at 14 dpc were used as controls with both giving a titer of 80. IgG and IgA titers in BALF after sacrifice at 3 dpc. Results shown are the mean of six pigs ± SD. As a positive control, BALF collected from A/Sw/Eng/1353/09 virus--infected pigs at 14 dpc was used and gave an endpoint titer of 1024 for IgA and 128 for IgG. We next determined average influenza-specific T cell responses in PBMCs by IFN-γ ELISPOT at 0 and 3 dpc. All S-FLU--immunized groups showed a virus-specific response to the challenge pdmH1N1 virus of ∼30--60 spot-forming cells (SFC) per 10^6^ cells just before the challenge (0 dpc), indicating that all animals had been immunized successfully. H5 i.t. immunized animals, which showed the strongest T cell response to A/Sw/Eng/1353/09 (*p* = 0.001), responded equally well to stimulation in vitro with H1 S-FLU or H5 S-FLU vaccine constructs, whereas H1 i.t. and H1 aer animals made minimal responses to H1 S-FLU or H5 S-FLU ([Fig. 2A](#fig02){ref-type="fig"}). At 3 dpc the number of SFC following pdmH1N1 stimulation declined in all immunized groups, most likely due to the migration of influenza-specific PBMCs to the lung. Intracellular cytokine staining of PBMCs at 3 dpc shows the highest IFN-γ and TNF-α responses in the H5 i.t. group ([Fig. 2B](#fig02){ref-type="fig"}). {#fig02} Taken together, these data show that the magnitude of the response to IAV in peripheral blood, measured either as number of IFN-γ□ or TNF-secreting cells, does not correlate with reduction of viral load in nasal swabs or the accessory lung lobe. It is not clear why PBMCs from the H1 i.t. and H1 aer--immunized animals do not respond in vitro to stimulation with H1 S-FLU or H5 S-FLU, whereas H5 i.t. immunized animals do, but this may be related to the higher dose of H5 S-FLU administered. BAL and TBLN influenza-specific responses {#s15} ----------------------------------------- We next measured the local influenza-specific immune responses in cells from BAL and TBLN, the main draining lymph nodes of the lungs. H1 aer immunization induced the highest numbers of IFN-γ--secreting cells in the BAL (∼200 SFC per 10^6^ cells, *p* = 0.001), whereas a much lower response occurred in the H5 i.t. immunized animals and none in the H1 i.t. and control groups ([Fig. 3A](#fig03){ref-type="fig"}). {#fig03} Because it has been suggested that cells simultaneously producing IFN-γ, IL-2, and TNF-α may provide optimal effector function and protection against viral infections ([@r7]), we defined the proportions of bifunctional cells in our immunized animals. We performed intracellular staining for IFN-γ and TNF-α on BAL cells stimulated with A/Sw/Eng/1353/09. H1 aer immunization induced the strongest, single IFN-γ responses by CD8 and CD4CD8 cells (0.41% in CD8 and 0.32% in CD4CD8 cells), followed by the H5 i.t. regimen ([Fig. 3B](#fig03){ref-type="fig"}, [3C](#fig03){ref-type="fig"}). The proportion of cells producing TNF-α was very low in all cell subsets, and neither TNF-α single-producing nor IFN-γTNF-α double-producing cells differed between the groups ([Fig. 3B--D](#fig03){ref-type="fig"}). The pie charts summarize the proportions of single and double cytokine-producing cells in the BAL. The H1 aer and H5 i.t. regimes appear to induce different proportions of single or double cytokine-producing influenza-specific cells compared with H1 i.t. and controls. However, H1 i.t. and controls induced very few Ag-specific cells so that interpretation of these data requires caution. Nevertheless, H1 aer immunization induced greater numbers of IFN-γ single producers in all cell subsets. TBLN responses differ from those in the BAL. H5 i.t. immunization induced the strongest IFN-γ ELISPOT response followed by the H1 aer regimen ([Fig. 4A](#fig04){ref-type="fig"}). Similarly, the H5 i.t. regimen induced the significantly highest CD4CD8 IFN-γ--producing cells as determined by intracellular cytokine staining staining (*p* = 0.009) followed by the H1 aer regimen (*p* = 0.047) ([Fig. 4C](#fig04){ref-type="fig"}). TNF-α production was higher compared with the BAL, although no differences between the groups were detected in the proportion of single TNF-α-- or double TNF-αIFN-γ--producing cells. As for BAL, the pie charts show that the best protected H1 aer group had the highest proportion of IFN-γ single-producing cells ([Figs. 4B--D](#fig04){ref-type="fig"}). {#fig04} In summary, the most protective H1 aer immunization induces a strong BAL and weaker TBLN response to IAV. These responses consist of cells with high amounts of intracellular IFN-γ, which correlate with and may account for their protective efficacy. S-FLU may have the potential to be a candidate universal vaccine {#s16} ---------------------------------------------------------------- Principal component analysis (PCA) was used to reduce a large number of original variables to a smaller number of undefined, underlying variables that are responsible for the variation in the data. The first principal component, PC1, could explain 23.1% of the total variation between pigs ([Fig. 5A](#fig05){ref-type="fig"}). A general linear model showed that PC1 was significantly associated with immunization status (*p* = 0.0014). In particular, the post hoc test revealed significant differences between the control and most of the immunization groups (H1 aer, *p* = 0.0009**;** H5 i.t., *p* = 0.0012). A second component, PC2, explained a further 14.9% of the variation and could be theorized to be associated with route of administration, although there was no significant difference between the aerosol and i.t. delivery routes (data not shown). Because of missing data points, it was not possible to obtain a full set of principal components for four pigs (one in each treatment group). {#fig05} The variables used for the PCA were assigned to one of four groups relating to gross pathology, viral titer, histopathology, or immune response. [Fig. 5B](#fig05){ref-type="fig"} shows the PCA loadings for each of these variables. The loadings explain how much variation within a particular variable can be assigned to a principal component. Three distinct clusters can be identified; viral titer and histopathology cluster together in the same quadrant, suggesting that there is an association with viral load and histopathological signs. Interestingly, IFN-γ production by CD4^+^ T cells in BAL also appears in this quadrant. Although gross pathology and histopathology appear in different quadrants, they do not directly oppose each other, suggesting that there is still some level of association between the two variables. Considering [Fig. 5A](#fig05){ref-type="fig"} and [5B](#fig05){ref-type="fig"} together, the PCA analysis shows that virus titer and histopathology scores were highest in the control group and lowest in the H1 aer group. Note that the H5 i.t. group also had a reduction in viral load and histopathology. Additionally, the figures show that the strongest immune response was seen in the H5 i.t. group, closely followed by the H1 aer group. This may be explained by the heterologous challenge virus used in the H5 i.t. group, as opposed to homologous challenge. The H1 i.t. and H1 aer groups also had lower gross pathology scores than did the control and H5 i.t. groups. Consistent with the previous analyses on individual variables, it appears that the H1 aer vaccination was most effective at reducing viral load and histopathological and gross pathological signs, as well as inducing an effective immune response to challenge with a homologous virus. Discussion {#s17} ========== Abundant evidence in humans and animals shows that natural influenza virus infection or immunization with live attenuated influenza viruses can induce a degree of cross protection ([@r1], [@r8], [@r9]). Cross-reactive CD4 and CD8 T cells or anti-HA stem Abs have been postulated as correlates of this cross protection in humans, but vaccines currently in use induce these responses poorly. In the present study, we have used a novel candidate universal vaccine, S-FLU, based on the suppression of HA signal sequence. S-FLU has the advantage of inducing an immune response to all viral components, but without replication of virus or danger of reassortment of HA viral RNA into seasonal influenza, making immunization directly to the site of infection safe. Our data show that immunization with S-FLU--expressing H1 HA (H1 S-FLU) reduces the viral load in nasal swabs and lungs after challenge with the closely matched pdmH1N1 virus strain. The reduction of viral load was greatest after aerosol administration of S-FLU. Detectable neutralizing Ab was not produced in S-FLU--immunized pigs. Rather, the reduction of viral load in the H1 aer group correlated with the presence of IFN-γ--producing CD8 or CD4CD8 double-positive cells in the BAL. CD4CD8 double-positive cells in pigs are activated memory CD4 cells ([@r10]), meaning S-FLU appears capable of inducing both CD4 and CD8 memory, both of which have been shown to be important in protective immunity to influenza ([@r11]--[@r13]). H5 i.t. immunization induces a much weaker BAL response, even though a higher dose of virus was administered, but stronger TBLN and PBMC responses. These results suggest that the effects of priming by aerosol and i.t. administered vaccines differ. The higher TBLN response and lower BAL response after i.t. immunization suggest either that priming by this route may have slower kinetics than aerosol immunization so that responding cells are still in the draining nodes at the time of euthanasia or that cells primed by this route do not acquire homing molecules allowing them to enter the alveolar spaces. It is also possible that the HA and NA molecules expressed on different S-FLUs may influence which cells within the respiratory tract can be infected. This in turn might influence the magnitude and nature of the immune response. Further experiments are required to define more clearly distribution of S-FLU delivered to pigs by aerosol or to the upper trachea and the properties of the Ag-specific cells induced. Our preliminary data (not shown) indicate that S-FLU administered intranasally did not induce lung immune responses or reduce viral load. Finally, we have recently shown that S-FLU can induce inhibitory Abs to the encoded NA in mice (T. Powell, P. Rijal, K.-Y. Huang, R. McEwen-Smith, H. Byun, M. Hardwick, L. Schimanski, R. Daniels, and A. Townsend, submitted for publication), so it is possible that the homologous immunization with H1N1 S-FLU induced a protective anti-N1(Eng) response. This is less likely with the H5N1 S-FLU, which encoded a distantly related N1 sequence from A/PR/8/34. Recent overwhelming evidence shows that immunization or infection via the respiratory tract results in generation of lung-resident memory T cells, which are much more numerous than previously thought and are critical for protection ([@r14]--[@r16]). These lung-resident T cells are phenotypically different from their peripheral blood counterparts and respond more vigorously to respiratory viruses, thus providing additional protection against infection. Further investigation will be needed to reveal the distribution within the lungs of memory cells induced by different methods of immunization and determine whether the BAL cells we detect in H1 aer--immunized pigs are part of the lung-resident memory population. The influenza-specific responses in our S-FLU--immunized animals were detectable up to 45 dpb but neither the frequency of IFN-γ--secreting cells nor the presence of multifunctional cells in the blood correlated with reduction of viral burden in nasal swabs or lungs. In humans, specific subsets of blood memory T cells with specificity for conserved internal proteins of influenza virus have been associated with protection against infection ([@r11]--[@r13]). Additional work may indicate whether such memory subsets exist in pigs and are induced by immunization with S-FLU. Influenza-specific cross-reactive CD8 T cells are also detectable in the lungs and persist for at least a year in mice and humans ([@r14], [@r17]). Longer studies in pigs will be required to establish the persistence of memory in this species and whether lung immune responses correlate with protective immunity during an extended period. Our experiments in pigs demonstrate that targeting the LRT by aerosol is an efficient way to induce an immune response and to reduce viral load. This is in agreement with earlier studies showing that for optimum induction of protective T cell immune response in mice and ferrets, virus infection of the lung is required, as opposed to infection of the upper respiratory tract or other peripheral sites ([@r18]--[@r20]). Aerosol delivery of H1 S-FLU (at least 4-fold less than H1 S-FLU i.t. and ∼1 log less than H5 S-FLU i.t. administered dose) induced a potent immune response in the BAL compared with the i.t. route of administration and was the most efficient in reducing viral load. A higher dose H5 i.t. immunization provided some degree of reduction of viral load after heterologous pdmH1N1 challenge. However, our method of i.t. delivery to the upper trachea clearly fails to immunize the lung efficiently and therefore further studies using aerosol administration will be required to establish whether S-FLU provides a high degree of heterologous as well as homologous protection in pigs. Similar comparisons between i.t. and aerosol delivery have been performed with adenoviral-vectored tuberculosis or Ebola vaccines in nonhuman primates, and in both cases aerosol delivery offered superior protection compared with other mucosal routes ([@r21], [@r22]). Although aerosol delivery of measles vaccine has been successfully deployed in humans ([@r23], [@r24]), to our knowledge this is the first report of delivery of an influenza vaccine to pigs by aerosol. Feng et al. ([@r25]) delivered a *Mycoplasma hyopneumoniae* vaccine by aerosol to pigs, and although they showed that the vaccine reached the LRT and induced local IgA, they did not assess protection or T cell responses induced by the immunization. Further studies will be required to develop practical devices for aerosol immunization in the field, but our data provide proof of principle that S-FLU can be efficiently delivered by aerosol to a large animal, strongly suggesting that this would be a highly efficient method of immunizing both pigs and humans against influenza viruses. Various live attenuated influenza vaccines (LAIV) have been developed and shown to be effective when delivered to pigs by the intranasal route ([@r26]--[@r29]). Loving et al. ([@r29]) used LAIV temperature-sensitive polymerase mutants to show prevention of viral shedding and transmission, but this did not correlate with Ab titers in serum or mucosal secretions. At first sight this result appears to contradict our finding that delivery of the vaccine to the LRT is the most efficient means of inducing a protective immune response. There are also other differences between the present study and that of Loving et al. ([@r29]). They used much younger (5 wk) and smaller animals and a different challenge virus (H3N2), and the details of the method of intranasal administration are not clear. Furthermore, no assays of T cell immunity were carried out so that it is not possible to exclude the possibility that a lung immune response was induced. Only head-to-head studies of different vaccines, together with assays of local and systemic immunity, will reveal the factors underlying these apparent differences. Vaccine-associated enhanced respiratory disease (VAERD) has been reported after immunization with inactivated swine H1N2 vaccine followed by infection with pdmH1N1 and correlated with the presence of anti-HA stem Ab ([@r30]). However, it has been demonstrated that VAERD could not be induced after administration of LAIV ([@r5]). This and our data showing lack of VAERD after S-FLU immunization support the idea that mucosal administration of vaccines avoids this complication. The licensed cold-adapted LAIV (FluMist) can induce cross-protective immunity when delivered to the lung ([@r2], [@r4], [@r31]), but not the upper respiratory tract of mice ([@r20]). However, there are two barriers to delivering existing LAIV to the LRT. The first is that delivery to the LRT of humans, although it may be the most effective means to induce protective immunity, may be dangerous, as LAIV retains some potential to replicate. Second, the viral RNA encoding pandemic HA in the vaccine virus is capable of reassortment with circulating seasonal viruses, with the potential to render the latter highly virulent ([@r32]). In contrast, S-FLU should be safe for aerosol administration to the lung because it cannot replicate, nor can it donate, a viable HA gene to seasonal influenza. Although understanding of immunity to influenza virus remains incomplete, our experiments suggest an important role for local lung immunity. Heterotypic immunity of the type induced by S-FLU, which is predominantly T cell mediated, does not prevent viral entry, but limits viral replication. Additional studies will be required to test whether the reduction in viral load shown in these studies is sufficient to block transmission and prevent disease. Nevertheless, the best protected lungs showed the highest numbers of Ag-specific cells in the BAL, indicating that local immunity in the lung is an important mechanism by which universal vaccines can confer protection. These data therefore have important implications for the design of pig and human universal vaccines against influenza. Challenges for the future will include the design of immunization regimes and devices that can safely induce long-lasting lung-resident memory in humans and livestock. Supplementary Material ====================== We thank the animal staff at Compton, The Pirbright Institute, and Bristol University for excellent animal care. This work was supported by Biotechnology and Biological Sciences Research Council Strategic Longer and Larger Grant BB/L001330/1. The challenge influenza virus strain was characterized under Department for Environment, Food and Rural Affairs Project SV3041, Monitoring of Influenza A Viruses in the UK Pig. The online version of this article contains [supplemental material](http://www.jimmunol.org/lookup/suppl/doi:10.4049/jimmunol.1502632/-/DCSupplemental). Abbreviations used in this article:aeraerosolA/Sw/Eng/1353/09A/Swine/England/1353/09(pdmH1N1)BALbronchoalveolar lavagedpbday postboostdpcday postchallengeGLMgeneral linear modelHAhemagglutininHAIHA inhibitionIAVinfluenza A virusi.t.intratracheal(ly)LAIVlive attenuated influenza vaccineLRTlower respiratory tractMDCKMadin--Darby canine kidneyMOImultiplicity of infectionNAneuraminidasePCAprincipal component analysispdmpandemicSFCspot-forming cellS-FLUinfluenza virus in which the hemagglutinin signal sequence has been suppressedTBLNtracheobronchial lymph nodeTCIDtissue culture--infective doseVAERDvaccine-associated enhanced respiratory disease. Disclosures {#s18} =========== A.M.T. is named on a European patent (publication no. EP2758525 A2, published July 30, 2014) concerning the use of S-FLU as a vaccine. The other authors have no financial conflicts of interest. [^1]: S.B.M. and J.D.H. contributed equally to this work.

Introduction {#s1} ============ Juvenile hormones (JHs) are acyclic sesquiterpenoids which contain an α,β-unsaturated methyl ester and a terpenoid backbone with an epoxide distal to the ester. Both the ester and epoxide groups are required for hormone regulatory functions. JH regulates diverse processes including the growth, development, metamorphosis and reproduction of insects [@pone.0056261-Gilbert1], [@pone.0056261-Truman1]. The JH levels control the full cycle of insect development from the immature larval stage to the adult form. Hence, insect growth regulators (IGRs) with JH-agonistic or -antagonistic activity have become attractive candidates for insect pest control since the identification of the structure of JH in 1967 [@pone.0056261-Rller1]. Several JH analogues (JHAs) such as methoprene [@pone.0056261-Henrick1], fenoxycarb [@pone.0056261-Grenier1] and pyriproxyfen [@pone.0056261-Miyamoto1] have been proven to be effective IGRs and useful as insecticides against household and agricultural pests that cause damage worth billions of pounds to global agriculture each year. The diversity of JH-mediated physiological effects suggests that target cells may respond to the hormone directly by gene expression and/or via a second messenger [@pone.0056261-Sevala1], [@pone.0056261-Wyatt1]. Since JH actions for the individual processes occur in tissue- and stage-specific manner, it is generally assumed that multiple numbers of JH receptors exist at the membranes and in cytosols and nuclei. Upon release from the *corpora allata* where JH is synthesized, the hormone is dispersed to the hemolymph to act at distant peripheral sites. In Lepidoptera almost every molecule of JH in the insect hemolymph appears in a complex with a specific 30 kDa JH-binding protein (JHBP) which serves as a carrier supplying the hormone to target cells [@pone.0056261-Kramer1]--[@pone.0056261-Touhara1]. Complex formation provides protection of the chemically labile JH against nonspecific enzymatic degradation and/or sequestration, and is crucial for effective signaling by the low amounts of the hormone. In this sense, JHBP is one of the most important proteins that regulate the JH signaling. In previous reports [@pone.0056261-Suzuki1], [@pone.0056261-Suzuki2], we have reported a gate-latch mechanism of JH delivery in hemolymph by JHBP based upon the crystal and solution structures of apo- and JH-bound forms of the recombinant JHBP from the silkworm, *Bombyx mori*. JHBP adopts a unique elongated β-barrel fold, consisting of a long spine helix (α3) wrapped in a highly curved β-sheet, which is suitable for a transport of the hydrophobic JH ([Figure 1](#pone-0056261-g001){ref-type="fig"}). Nearly the same folds are shared by several other lipid binding proteins: Takeout, a potential ubiquinone binding protein [@pone.0056261-Hamiaux1]; a bactericidal permeability-increasing protein [@pone.0056261-Beamer1]; and a cholesteryl ester transfer protein [@pone.0056261-Qin1]. JH binds to a hydrophobic JH-binding pocket located at the one end of the elongated structure near the C-terminus of α3. The uptake and release of JH are regulated by the opening and closing of the α1-helix over the JH-binding pocket that functions as a gate sensing the ligand entry. In the crystal structure of the apo-JHBP (gate-open conformation), the location of the α1-helix generates a wide open conformation which permits access of JH to the hormone-binding site ([Figure 1A](#pone-0056261-g001){ref-type="fig"}). Binding of the JH molecule induces a dramatic change in the orientation of α1, which swings towards the pocket about 70° from the position of the corresponding helix in the apo-structure. In the resulting fully gate-closed JHBP-JH complex structure ([Figure 1B](#pone-0056261-g001){ref-type="fig"}), the bound JH is completely buried inside the protein and is thus protected from unfavorable nonspecific absorption and enzymatic degradation during its transport in the hemolymph. In solution, the apo-JHBP assumes multiple conformations of the gate α1-helix ranging from the fully closed to open forms while the protein core is well maintained in all of these structures. JH binding silences conformational dynamics of the α1 gate and leads to the formation of the unique conformation seen in the JHBP-JH complex [@pone.0056261-Suzuki1]. Our dynamic structural model visualizes the JH-induced conformational change of JHBP previously reported by a range of different techniques such as proteolysis and ultracentrifugation [@pone.0056261-Wieczorek1], UV-difference and CD spectroscopy [@pone.0056261-Touhara1], [@pone.0056261-Krzyzanowska1], and electrochemical impedance spectroscopy [@pone.0056261-Stobiecka1]. {#pone-0056261-g001} Structural studies revealed that JHBP has an additional hydrophobic cavity (second cavity) on the side opposite from the JH-binding pocket [@pone.0056261-Suzuki1], [@pone.0056261-Kolodziejczyk1]. In both the apo- and JH-bound structures of the recombinant *B. mori* JHBP, this second cavity remains empty [@pone.0056261-Suzuki1]. In contrast, ligand binding to the second cavity is suggested by the crystal structure of the native JHBP from the wax moth *Galleria mellonella* in which the second cavity accommodates an undefined small molecule while the JH-binding pocket remains empty [@pone.0056261-Kolodziejczyk1]. Although the nature of the second ligand is not known, it was speculated that the second ligand could be bound to the cell membrane to allow for hormone delivery site recognition by JHBP [@pone.0056261-Kolodziejczyk1]. Further studies are necessary to validate the accurate functional roles of the second ligand-binding site. Lepidopteran insects are major agricultural pests. For example, the tobacco hornworm *Manduca sexta* and the tobacco budworm *Heliothis virescens* are the most destructive pests of tobacco [@pone.0056261-Metcalf1], while the wax moth *G. mellonella* causes serious damage to beehives [@pone.0056261-Shimanuki1]. Since the hemolymph JHBP mediates the first step in the JH signal transduction cascade, and is highly specific to Lepidopteran insects, JHBP is expected to be an excellent target for the design of novel specific IGRs and insecticides. The first three-dimensional structures of the JHBP-JH II complex in the crystalline state and the JHBP-JH III complex in solution, together with subsequent biochemical assays, enabled us to elucidate the molecular mechanism of JH recognition by JHBP that clearly explains the ligand specificity and enantioselectivity [@pone.0056261-Suzuki1], [@pone.0056261-Goodman1]--[@pone.0056261-Schooley1]. The structural information derived from the JHBP-JH complexes opens the way to the structure-based design of IGRs which inhibit complex formation between JH and JHBP, and thus disrupt the JH signaling in Lepidoptera. The development of such IGRs should be further accelerated by enhancement of our understanding of interactions of JHBP with any ligands structurally related and unrelated to the JH molecule. Here we report the crystal structure of the recombinant *B. mori* JHBP in complex with 2-methyl-2,4-pentanediol (MPD). This is the first structure of JHBP in complex with a ligand which is structurally unrelated to JH. In this structure, we found two bound MPD molecules, one (MPD1) in the JH-binding pocket and the other (MPD2) in a second cavity. MPD1 is anchored in the same hydrophobic cage as the epoxide of the JHBP-bound JH. Furthermore, interactions of MPD1 with JHBP are nearly the same as those observed for the epoxide moiety of JH in the JHBP-JH complex. We confirm that MPD and methoprene containing a MPD-like structural element, but not the unepoxydated JH III (methyl farnesoate), inhibit the binding of JH to JHBP using the ligand competitive binding assay. These findings strongly demonstrate that the epoxy group and its mimic structures are critically important for the ligand binding to the JH-binding pocket of JHBP, and should provide useful information for IGR design targeted for JHBP. We also found that MPD2 binding induces a significant conformational change in the second cavity. The MPD2-JHBP interactions reported here should provide important guidance in the search for the natural ligand of the second cavity. Results and Discussion {#s2} ====================== Overall Structure of the JHBP-MPD Complex {#s2a} ----------------------------------------- The crystal structure of *B. mori* JHBP in complex with MPD was solved at a resolution of 2.6 Å by a single-wavelength anomalous dispersion (SAD) method using selenomethionine (SeMet)-substituted crystals ([Figure 2](#pone-0056261-g002){ref-type="fig"} and [Table 1](#pone-0056261-t001){ref-type="table"}). As shown in [Figure 2A](#pone-0056261-g002){ref-type="fig"}, the MPD-bound JHBP adopts a gate-closed conformation which is similar to the structure of the JH-bound JHBP [@pone.0056261-Suzuki1]. The refined model contains four JHBP monomers (A, B, C and D) in the crystal asymmetric unit, which are related by 222 point group symmetry ([Figure 2B](#pone-0056261-g002){ref-type="fig"}). No stable dimerization could be seen in any two pairs of the four NCS-related molecules. Consistent with this, gel filtration and ultracentrifugation experiments as well as NMR spectroscopy have shown unambiguously that JHBP is monomeric in solution [@pone.0056261-Suzuki1], [@pone.0056261-Suzuki2], [@pone.0056261-Wieczorek1], [@pone.0056261-Oyhar1], [@pone.0056261-Dobryszycki1]. {#pone-0056261-g002} 10.1371/journal.pone.0056261.t001 ###### Summary of data collection and refinement statistics of the crystal structures of the JHBP-MPD complex. {#pone-0056261-t001-1} MPD --------------------------------------------------- ------------------------ **Data collection** Space group *P*2~1~2~1~2~1~ Unit cell parameters * a* (Å) 54.9 * b* (Å) 114.7 * c* (Å) 192.9 Beam Line PF 17A Wavelength (Å) 0.97000 Resolution (Å) 50--2.60(2.69--2.60) Total reflections 473812 Unique reflections 38636 (3701) *R*-merge 0.099(0.489) Completeness (%) 99.7(97.5) Average *I/σ(I)* 22.2(3.1) Average redundancy 12.3(9.2) **Structure refinement** Resolution (Å) 49.5--2.59(2.66--2.59) *R*-factor 0.222(0.310) *R* ~free~-factor[a](#nt101){ref-type="table-fn"} 0.290(0.382) RMSD from ideal Bond lengths (Å) 0.016 Bond angles (°) 1.75 Ramachandran plot (%) Favoured regions 92.2 Allowed regions 5.6 Outlier regions 2.2 *R* ~free~-factors were calculated using 5% of the unique reflections. We could trace the electron density of residues 7--174 and 183--222 for molecule A, of 7--175 and 180--222 for molecule B, of 5--175 and 183--223 for molecule C, and of 7--174 and 182--222 for molecule D. The four molecules overlay well with one another, giving root mean square difference (RMSD) values of less than 0.70 Å for C^α^ atoms between any two pairs of them. For all the four JHBP molecules, we revealed two bound MPD molecules, one (MPD1) in the JH-binding pocket and the other (MPD2) in the second cavity at each end of the elongated structure ([Figures 2A and C](#pone-0056261-g002){ref-type="fig"}). These MPD molecules were most likely incorporated from the precipitant solution during crystallization of the apoprotein. Structural Plasticity of the JH Binding Pocket in JHBP {#s2b} ------------------------------------------------------ Comparison of the MPD-bound and JH II-bound JHBP structures reveals that the core of the protein assumes nearly the same structure with RMSD values ranging from 0.69 to 1.05 Å for C^α^ atoms of residues 32--172 and 190--210 depending on the pair of chains superposed ([Figure 3A](#pone-0056261-g003){ref-type="fig"}). However, pronounced displacements are observed at both ends of the elongated structure. The largest displacement involves the gate α1 helix which covers the JH-binding pocket ([Figure 3B](#pone-0056261-g003){ref-type="fig"}). The structural rearrangement of α1 is correlated with the difference in the molecular size between JH and MPD. Comparing with the JH II-bound structure, the α1 helix of the MPD-bound structure is folded back into the hormone binding pocket about 11° and generates contacts of the Leu17 and Thr21 side chains in helix α1 with MPD, which is smaller in size than JH. Concomitantly, the N-terminal arm and the C-terminal tail that serve as a latch for the JH-binding pocket are pushed out toward solvent. When compared with the unliganded apo-JHBP structure [@pone.0056261-Suzuki1], the movement of the α1 helix is 81° ([Figure 3B](#pone-0056261-g003){ref-type="fig"}). {#pone-0056261-g003} Reflecting differences in the orientation of the gate α1 helix, the sizes of the JH-binding pockets are different between the MPD-bound and JH II-bound forms. According to the Swiss-PbdViewer version 4.1 (<http://www.expasy.org/spdbv>) [@pone.0056261-Guex1], the JH-binding pocket of the JH II-bound form is a huge continuous cavity. The volume was calculated to be 874 and 903 Å^3^ for the two JH-complexes A and B in the crystal asymmetric unit, respectively. The cavity is completely closed, and matches almost accurately the van der Waals surface of the bound JH II molecule as shown in [Figure 4A](#pone-0056261-g004){ref-type="fig"} (drawn for JH-complex A). The cavity is sealed by two hydrogen bonds, Cys9 NH-Phe220 CO and Glu222 NH-Cys9 CO, between the N-terminal arm and the C-terminal tail ([Figure 4B](#pone-0056261-g004){ref-type="fig"}). In contrast, the JH-binding pocket in the MPD-bound form is separated into two smaller cavities, a large one with a volume of 507 Å^3^ at the upper part which accommodates the MPD1 molecule and a tiny one with a volume of 56 Å^3^ at the bottom wall as shown in [Figure 4C](#pone-0056261-g004){ref-type="fig"} (drawn for MPD-complex C). The latter tiny cavity corresponds to the open space below the methyl ester of the bound JH in the JHBP-JH complex. The former large cavity is slightly expanded upward compared with that of the JH-bound form, and has an open hole due to the structural rearrangements of the α1 helix, N-terminal arm, and C-terminal tail as mentioned above. It is noteworthy, however, that the Cys9 NH-Phe220 CO and Glu222 NH-Cys9 CO hydrogen bonds are still maintained in the MPD-bound structure ([Figure 4D](#pone-0056261-g004){ref-type="fig"}). These observations suggest a high adaptability of the JH binding pocket which could change its size and shape in a ligand-dependent manner due to flexibility of the gate α1 helix. {#pone-0056261-g004} MPD Mimics Epoxide of JH {#s2c} ------------------------ The MPD molecule (MPD1) bound to the JH-binding pocket is confined in a hydrophobic cage formed by Phe78 and Met80 in β2b, Val87 and Leu89 in β3, Tyr128 and Tyr130 in β4, Phe142 and Val144 in β5a, and Phe220 and Phe221 in the C-terminal tail as shown in [Figure 5A](#pone-0056261-g005){ref-type="fig"} (drawn for MPD-complex C). The same cage accommodates the epoxide group of JH in the JHBP-JH II complex [@pone.0056261-Suzuki1] as shown in [Figure 5B](#pone-0056261-g005){ref-type="fig"} (drawn for JH-complex A). In fact, MPD1 overlaps the epoxide group of JH II in the complex with JHBP as shown in [Figure 5C](#pone-0056261-g005){ref-type="fig"}. Furthermore, the recognition mode of MPD1 is essentially the same as that of the epoxide group of JH ([Figures 5A and B](#pone-0056261-g005){ref-type="fig"}). MPD1 is anchored in the cage by a direct hydrogen bond of the O^2^H group of MPD with the hydroxyl group of Tyr128 which forms a direct hydrogen bond with the epoxy oxygen of JH in the JHBP-JH II complex. In addition, MPD1 is further stabilized by CH-π stacking interactions between two methyl groups attached to C^2^ of MPD and the aromatic side chains of Tyr130 and Phe142. These two aromatic residues make similar CH-π stacking interactions with the methyl and ethyl groups attached to the epoxide of JH II in the JHBP-JH II complex. {#pone-0056261-g005} Similarity in the binding mode between MPD1 and the epoxide moiety of JH raises the question if MPD competes with the binding of JH to JHBP. To answer this question, we performed a ligand competition binding assay by using a described method [@pone.0056261-Vermunt1]. Binding of ^3^H-JH III to JHBP was measured in the absence and presence of MPD. For comparison, we also tested methyl farnesoate (MF) which is the unepoxidated form of JH III and JH-agonist methoprene which can be used as an insecticide. [Figure 5D](#pone-0056261-g005){ref-type="fig"} shows chemical structures of the tested ligands as well as JH II. Since the local structure of methoprene around the methoxy group resembles the MPD structure, the methoxy oxygen is expected to act as the hydrogen-bond acceptor from the Tyr128 of JHBP. It has been reported by electrochemical impedance spectroscopy that methoprene binds to the recombinant *G. mellonella* JHBP with the dissociation constant (*K* ~d~) of 33.3 nM which is comparable to the *K* ~d~ values for JH I (28.6 nM) and JH III (43.5 nM) [@pone.0056261-Stobiecka1]. As shown in [Figure 5E](#pone-0056261-g005){ref-type="fig"}, MPD inhibits the binding of ^3^H-labeled JH III to JHBP in a dose-dependent manner. The IC~50~ value was determined to be 160 µM by least-square-fitting to the experimental data using Eq. 1 described in Materials & Methods. Methoprene displayed 5-fold stronger inhibitory activity (IC~50~ = 35 µM) than MPD likely due to additional interactions with JHBP. The curve-fitting also provided the maximum reduction in the radioactivity, ΔF~max~ in Eq. 1, 66% for MPD and 76% for methoprene, which are comparable to the value for JH III (68%) estimated by use of the same binding assay [@pone.0056261-Vermunt1]. In contrast with MPD and methoprene, MF inhibited only 30% of the ^3^H-JH III binding to JHBP even at the concentration of 2 mM ([Figure 5E](#pone-0056261-g005){ref-type="fig"}). The weak inhibitory activity of MF could be attributed to the lack of the epoxy group because the structure of the remaining part is the same as JH. Hence the results of our binding assay strongly demonstrate that the epoxy group and its mimic structures, particularly the existence of the hydrogen-bond acceptor from the Tyr128 O^η^H, are critically important for the ligand binding to the JH-binding pocket of JHBP. The dissociation constant (*K* ~d~) for the tested ligand can be theoretically calculated from the IC~50~ value if the amount of the unliganded JHBP is negligible. For the ligand competition assay, we employed the condition that JHBP exists much more than JH as in the insect hemolymph (800 nM JHBP and 10 nM ^3^H-JH III in our case). It is, therefore, difficult to estimate the precise *K* ~d~ value from the IC~50~ value obtained from our assay. Applying the reported *K* ~d~ values of JH III (43.5 nM) and methoprene (33.3 nM) [@pone.0056261-Stobiecka1] to the equation of *K* ~d~ = IC~50~/(1+ \[JH\]/*K* ~d,JH~), the IC~50~ of methoprene is theoretically calculated to be 41 nM which is three-order of magnitude smaller than the IC~50~ value (35 µM) obtained by our assay. Assuming that this three-order difference between the theoretical and our observed IC~50~ values is held in the case of MPD, the *K* ~d~ value is estimated to be 152 nM. This value might be the maximum because MPD can also bind to the second cavity of JHBP. Ligand-induced Conformational Change of the Second Cavity in JHBP {#s2d} ----------------------------------------------------------------- JHBP has a second hydrophobic cavity located on the side opposite from the JH-binding pocket. The second cavity is formed by the α2-helix, the N-terminal portion of the α3-helix, and the inner side of the highly curved β-sheet. The entrance of the cavity is formed by the β5c strand, the α2-helix, the N-terminal portion of the α3-helix, and presumably the α2-α3 loop, which is not observable for the MPD-bound JHBP because of the flexibility of the loop or conformational heterogeneity. The end of the β5c strand and the whole α2-helix forms one side-wall while the N-terminal portion of the α3-helix forms the other side-wall. Seven residues, Val50, Phe62, Tyr116, Ile159, Arg189, Ala192 and Ile193, form the inside-wall which closes the cavity at the middle of the protein structure. In contrast, the outside-part at the end of the second cavity remains open. Comparison of the MPD-bound JHBP structure with the crystal structures of the apo- and JH II-bound JHBP [@pone.0056261-Suzuki1] suggests a ligand-induced structural rearrangement of the second cavity. This cavity accommodates one MPD molecule (MPD2) in the MPD-bound form ([Figure 6A](#pone-0056261-g006){ref-type="fig"}) but remains empty in the apo- and JH-bound forms [@pone.0056261-Suzuki1] ([Figure 6B](#pone-0056261-g006){ref-type="fig"}). [Figures 6A and 6B](#pone-0056261-g006){ref-type="fig"} are drawn for the MPD-complex A and the JH-complex A, respectively. There is essentially no difference in the structure of the unliganded second cavity between the apo- and JH II-bound forms. MPD2 is partially solvent-exposed, and is held in the cavity by a water-mediated hydrogen-bond network with Tyr116 O^η^H and Arg189 CO, and hydrophobic interactions with Ile60 and Phe62 in β2a, Ile101 in β3, Tyr116 in β4, Ile159 and Leu161 in β5c, and Leu185, Leu189, Arg189 and Ala192 at the N-terminus of helix α3 ([Figure 6A](#pone-0056261-g006){ref-type="fig"}). The α3-helix has no kink in its conformation in the MPD-bound second cavity ([Figure 6A](#pone-0056261-g006){ref-type="fig"}), but is kinked at position 190 near the N-terminus in the unliganded second cavity ([Figure 6B](#pone-0056261-g006){ref-type="fig"}), where the side chain of Leu188 occupies the same space as MPD2 in the MPD-bound structure. In addition, a loop connecting helices α2 and α3 is visible in the apo- and JH II-bound forms. As a result, the unliganded second cavity is shallow when compared to the liganded one. {#pone-0056261-g006} Ligand binding to the second cavity is also suggested by the crystal structure of the native *G. mellonella* JHBP [@pone.0056261-Kolodziejczyk1]. In this structure, the second cavity accommodates an undefined small molecule for which a 7 Å long patch of electron density starts from a point about 3 Å from the functional groups of Lys51, Tyr62 and Thr193 toward the main entrance. The equivalent residues in *B. mori* JHBP are hydrophobic: Val50, Phe62 and Ala192. It is worth mentioning that the starting point of this undefined molecule seems to match the location of MPD2 in our JHBP-MPD complex structure. Furthermore, in the *G. mellonella* JHBP structure long spine α4 helix, which corresponds to the α3 helix in the *B. mori* JHBP, has no kink in its conformation as the MPD-bound structure of *B. mori* JHBP reported here. Unlike the liganded and unliganded *B. mori* JHBP, the α2 and α4 helices of *G. mellonella* JHBP are connected by a clearly observable extra helix α3. The structural change of the α3 helix of the *B. mori* JHBP from the kinked conformation to the straight one after ligand binding to the second cavity results in a cavity expansion. We used the GHECOM program (<http://strcomp.protein.osaka-u.ac.jp/ghecom/>) [@pone.0056261-Kawabata1] to evaluate the volume of the second cavity instead of the Swiss-PbdViewer program [@pone.0056261-Guex1] used for the JH-binding pocket because the former program provides more reasonable results for shallow pockets like the second cavity. The GHECOM analysis revealed that the volume of the MPD-bound second cavity of the *B. mori* JHBP is in a range of 741--929 Å^3^ ([Figure 6C](#pone-0056261-g006){ref-type="fig"}) while the unliganded second cavity of the same protein is separated into two smaller cavities with volumes of 86 (site 2a) and 233 (site 2b) Å^3^ by the Leu185 side chain in the kinked α3 helix ([Figure 6D](#pone-0056261-g006){ref-type="fig"}). It has been reported that the volume of the liganded second cavity of the *G. mellonella* JHBP is estimated to be 668 Å^3^ by the CASTp analysis [@pone.0056261-Kolodziejczyk1], [@pone.0056261-Dundas1]. Our GHECOM [@pone.0056261-Kawabata1] analysis of the same structure provided the volume of 706 Å^3^ for the second cavity of the *G. mellonella* JHBP. Hence, we suggest that binding of a small molecule to the second cavity of JHBP straightens the long spine helix, α3 in *B. mori* JHBP or α4 in *G. mellonella* JHBP, and creates a proper space for the ligand. This finding, together with the MPD2-JHBP interaction mechanism, should assist in identification of the natural ligand(s) for the second cavity of JHBP. Conclusion {#s2e} ---------- In this paper, we present the crystal structure of the recombinant *B. mori* JHBP in complex with two molecules of MPD (MPD1 and MPD2), where MPD1 is bound in the JH-binding pocket and MPD2 in a second hydrophobic cavity. They are at different ends of the elongated protein structure. This is the first structure of JHBP in complex with a ligand which is structurally unrelated to JH. Detailed comparison with the apo-JHBP and JHBP-JH II complex structures previously reported by us [@pone.0056261-Suzuki1] led to a number of intriguing findings. First, the JH-binding pocket changes its size and shape in a ligand-dependent manner due to flexibility of the gate α1 helix. Second, MPD1 mimics interactions of the epoxide moiety of JH previously observed in the JHBP-JH complex, and MPD can compete with JH in binding to the JH-binding pocket. We also confirmed that methoprene, with an MPD-like structural element, inhibits the complex formation between JH and JHBP. The existence of the hydrogen-bond acceptor from Tyr128 is critical for ligand binding to the JH-binding pocket, because methyl farnesoate, which is structurally similar to JH but lacks such acceptor, showed significantly weaker binding to JHBP. These findings could open the door to the structure-based design of novel Lepidoptera-specific IGRs which inhibit complex formation between JH and JHBP, and thus disrupt JH signaling. Third, binding of MPD to the second cavity of JHBP induces significant conformational changes accompanied with cavity expansion. This finding, together with the MPD2-JHBP interaction mechanism identified in the JHBP-MPD complex, should provide important guidance in the search for the natural ligand of the second cavity, which is presumably important for JH delivery site recognition by JHBP. Materials and Methods {#s3} ===================== Crystallization, Data Collection and Structure Determination {#s3a} ------------------------------------------------------------ Mature JHBP from *Bombyx mori* was expressed as a GST-fusion protein in *Escherichia coli* and purified as described previously [@pone.0056261-Suzuki2], [@pone.0056261-Vermunt1]. SeMet-substituted JHBP was expressed in *E. coli* strain B834 (DE3). Crystals of JHBP in complex with MPD and SeMet derivative crystals were obtained by the hanging-drop vapor-diffusion method at 20°C from solution containing 3 µL of apoprotein at 20 mg mL^−1^ in 2 mM Tris buffer, pH 8.0, and 1.5 µL of crystallant: 25% (±)-MPD (Hampton Research, Aliso Viejo, CA, USA), 0.05 M zinc acetate (Wako Pure Chemical Industry, Osaka, Japan) and 0.1 M sodium cacodylate buffer, pH 7.0 (Hampton Research). After a week, thin plate crystal clusters appeared and a typical crystal grew to the size of dimensions 500 × 20 × 10 µm. Diffraction data were collected at beamlines of the Photon Factory (PF), High Energy Accelerator Research Organization, Tsukuba, Japan. A single crystal was scooped up in a nylon loop and directly flash-frozen in a nitrogen stream at 95 K. Native data were collected by CCD detectors (Area Detector Systems Corp., Poway, CA, USA). Data were integrated and scaled using the program DENZO and *Scalepack* in the HKL2000 program suite [@pone.0056261-Otwinowski1]. MPD complex and SeMet crystals belonged to the space group *P*2~1~2~1~2~1~ and four molecules were present in the asymmetric unit with a Vm value of 3.0 Å^3^ Da^−1^ and a solvent content of 59.6% [@pone.0056261-Matthews1]. Anomalous dispersion data sets for the SeMet crystal were collected near the selenium absorption edge. Structure solution of JHBP was performed using the SAD method. The heavy atom search and initial phase calculation were conducted using the program SOLVE/RESOLVE [@pone.0056261-Terwilliger1], [@pone.0056261-Terwilliger2]. A total of 16 heavy atom positions were determined, four of which were identified as bound zinc ions. An initial model produced by RESOLVE indicated that the crystal contains four JHBP molecules in the asymmetric unit. Successively, the structural model was refined using the MPD complex data of 2.6 Å resolution. Manual model rebuilding, introduction of water molecules, and molecular refinement were conducted using Coot [@pone.0056261-Emsley1] and Refmac5 [@pone.0056261-Murshews1]. A total of 20 zinc ions, eight MPD molecules and 121 water molecules were added into the final model. The stereochemistry of the models was analyzed with the program RAMPAGE [@pone.0056261-Lovell1]. Data collection and structure refinement statistics are summarized in [Table 1](#pone-0056261-t001){ref-type="table"}. The atomic coordinates and structure factors have been deposited in the Protein Data Bank (3A1Z). Figures of the proteins were generated by a combination of PyMOL (version 1.5: Schrödinger, LLC.) and PovRay (Version 3.7: Persistence of Vision Pty. Ltd.). Cavities on the protein surface were detected either by the Swiss-PbdViewer version 4.1 [@pone.0056261-Guex1] or the program GHECOM [@pone.0056261-Kawabata1]. The shapes of the cavities detected by GHECOM were visualized by use of OOSAWA (<http://www.cfca.nao.ac.jp/~takedatk/COMPUTER/OOSAWA/oosawa.html>) and PovRay. Competitive Ligand Binding Assay {#s3b} -------------------------------- Competitive ligand binding assays were conducted based on the method used for the JH binding assay [@pone.0056261-Vermunt1] with slight modifications. For the assay, the recombinant *B. mori* JHBP (800 nM) was dissolved in 20 mM Tri-HCl buffer, pH 7.9, supplemented with 5 mM magnesium acetate, 1 mM EDTA, 1 mM DTT, 1 mM PMSF (phenylmethylsulfonyl fluoride), 2 mg mL^−1^ Leupeptides, 1 mg mL^−1^ pepstatin A, 0.1 mM diisopropyl fluorophosphate, and 10 µM 3-octylthio-1,1,1-trifluoropropanone. The protein samples were incubated for 30 min at 4°C in a volume of 100 µL containing 10 nM ^3^H-JH III (62.90 GBq mmol^−1^; New England Nuclear Chemicals) in the absence or the presence of the competitive ligand delivered in ethanol (1% v/v). Unbound JH was removed by the addition of dextran-coated charcoal solution (100 µL) and centrifugation for 2 min at 10,000 *g*. The radioactivity of a 50 µL aliquot of the supernatant was measured using Perkin-Elmer Liquid Scintillation Counter (Tri-Carb 2900TR) for quantitative determination of JHBP-bound ^3^H-JH III. Data were fitted to a hyperbolic equation describing binding to a single site [@pone.0056261-Mogensen1],where F~obs~ is the observed radioactivity at any given competitive ligand concentration, F~0~ is the radioactivity of the JHBP-bound ^3^H-JH III in the absence of competitors, and ΔF~max~ is the maximum reduction in the radioactivity. IC~50~ and ΔF~max~ are fitted as free parameters by non-linear squares regression analysis. We would like to thank the beamline researchers and staff at the Photon Factory for X-ray diffraction data collection. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: TS TY. Performed the experiments: ZF RS TS WT AT MM TY. Analyzed the data: ZF RS TS TY. Contributed reagents/materials/analysis tools: WT AT MM. Wrote the paper: TY.

Background {#Sec1} ========== Necrotizing enterocolitis (NEC) is the most common surgical gastrointestinal emergency in neonates with high morbidity and mortality. The exact etiology of NEC is still unknown and prematurity is considered as the most important risk factor Beeby & Jeffery ([@CR2]). More than 90% of the infants with NEC are born before 36 weeks gestation Henry & Moss ([@CR17]). Immaturity and ischemia of the gut, infection and introduction of hyperosmolar feeds are the key factors postulated in the development of NEC Henry & Moss ([@CR17]). Regardless of the underlying risk factors, the endogenous inflammatory mediators such as interleukins, platelet activating factor (PAF) and nitric oxide (NO) have been implicated in the final common pathogenesis of NEC Edelson et al. ([@CR10]; Ewer [@CR11]; Muguruma et al. [@CR25]; Caplan et al. [@CR9]). PAF (1-O-alkyl-2-acetyl-sn-glycero-3-phosphocholine) is a potent pro-inflammatory mediator, involved in the signaling and activation of proinflammatory cells such as platelets, neutrophils, and macrophages Prescott et al. ([@CR28]). PAF causes a variety of biological functions such as epithelial apoptosis, increased vascular permeability, bronchoconstriction and smooth muscle contraction Hsueh et al. ([@CR19]; Caplan et al. [@CR6]; Sun et al. [@CR32]). The proinflammatory effects of PAF are terminated by the enzyme PAF acetylhydrolase (PAFAH). It exists in two forms - intracellular and secretory. The secretory form is found in breast milk and in other body fluids Moya et al. ([@CR24]). Studies in human infants with NEC have shown increased luminal and systemic PAF and reduced expression of PAFAH Caplan et al. ([@CR5]). In animal models with NEC, increased levels of serum PAF and decreased serum levels of the PAFAH have also been clearly documented Lu et al. ([@CR22]). The role of PAF and PAFAH has been extensively documented in various disease states involving inflammation Tjoelker & Stafforini ([@CR34]). Asthma and atopy are the prototypic diseases where the role of PAF mediating inflammation been extensively studied. Increased PAF and reduced PAFAH activity have been proven in asthma patients Tsukioka et al. ([@CR35]; Miwa et al. [@CR23]). Ile198Thr (in exon 7, position 593;T \> C) and Ala379Val (in exon 11, position 1136;T \> C) variants in PAFAH gene have been associated with an increased risk for the development of atopy and asthma in Caucasian population Kruse et al. ([@CR20]). Even though the role of PAF in NEC is well documented, no prior study has been reported about the prevalence of PAFAH variants in NEC infants and the role of potential pathogenesis. The rationale for the present study is based on the research by Kruse et al. ([@CR20]). In neonates, the polymorphisms of coding genes of various enzymes involved in the inflammatory pathways can influence various conditions such as respiratory distress syndrome, bronchopulmonary dysplasia, intraventricular hemorrhage and retinopathy of prematurity. For example, in our previous studies, we have reported associations between genotypes of the nitric oxide synthase gene and two neonatal conditions namely intraventricular hemorrhage and retinopathy of prematurity Yanamandra et al. ([@CR39]; Vannemreddy et al. [@CR37]). In this prototype study, we attempted to find the prevalence of the PAFAH variants (Ile198Thr and Ala379Val) in Northwest Louisiana infants both with and without NEC and any causal relationship between these variants and the NEC. Materials and Methods {#Sec2} ===================== Subjects {#Sec3} -------- Institutional review board (IRB) approval was obtained. We screened peripheral EDTA blood samples from 570 preterm infants admitted to the neonatal intensive care unit (NICU), Louisiana State University Health Sciences Center, Shreveport, Louisiana for the presence of Ile198Thr and Ala379Val polymorphisms. The diagnosis of NEC was based on diagnostic coding, the International Classification of Diseases, 9th revision, Clinical Modification, [@CR33]; ICD-9-CM code for the perinatal disorders of the digestive system -- 777.5 is NEC in newborn; 777.50 is NEC in newborn unspecified; 777.51 is stage I NEC; 777.52 is stage II NEC (NEC with pneumatosis, without perforation) and 777.53 is stage III (NEC with perforation or NEC with pneumatosis and perforation) (The International Classification of Diseases, 9th revision, Clinical Modification, [@CR33]). Only limited clinical information was available to us such as ethnicity of the infants, gestational age and birth weight. The important clinical information pertinent to NEC such as feeding history, sepsis, presence of patent ductus arteriosus, use of inotropes, administration of indomethacin, radiological findings and surgical management were not available for analysis. Out of 570 infants, 36 infants (20 Caucasian infants and 16 African-American infants) had either stage I or II NEC. The number of infants with stage I and II were equal in both ethnicities. In view of small numbers, stages I and II were combined into one category. The 36 infants in the NEC population did not have the ICD-9-CM diagnostic codes 777.5, 777.50 or 777.53. Also in this sample of 570 infants, no infants died. The remaining 534 infants (570--36) did not have any diagnostic coding for NEC, and they were recruited as the control population. The control population included 192 Caucasian and 342 African-American infants. The demographic and the gestational characteristics such as birth weight and gestational age for the NEC and the control groups were summarized in Table [1](#Tab1){ref-type="table"}. In the Caucasian group, the birth weight and the gestational age in the NEC group and the control group were comparable (p value \>0.05 by two sided z test). In the African-American group, the NEC population was significantly younger and more premature when compared to the control group (p value \< 0.05 by two sided z test).Table 1**Summary statistics on demographic and gestational characteristics of the study subjects**NECEthnicityNBirth weight (gms)Gestational age (weeks)RangeMeanRangeMeanYesC201100-2837214929.7 - 36.834.2NoC192720 - 2609227827.6 - 36.833.9YesAA16810 - 1510122424.7 - 32.129.1NoAA342540 - 2480196523.4 - 36.832.8NEC -- necrotizing enterocolitis; C- Caucasian; AA -- African-American. Methods {#Sec4} ------- DNA was isolated from patients and controls using Qiagen blood mini DNA isolation kits. Microplate PCR method was carried out for each polymorphism. The genotype assay used was modified from the assay used by Kruse et al. ([@CR20]). Ala379Val polymorphism {#Sec5} ---------------------- Briefly 10 microlitre (μl) of the PCR mix contained the following components in the final concentration of dH20 5.35 μl, 25 mM MgCl2 1.3 μl, 10× Applied Biosystems buffer 1 μl, 3.75 pmoles of each of 4 dNTPs, forward and reverse PAFAH primers 0.5 μl, Taq polymerase (gold, Perkin-Elmer) 0.5 units, and 1--2 ng of patient DNA. The primer sequences used were: GGG AGA CAT AGA TTC AAC TG (forward) and CGT TTT GTA AGA ATG CTA ATG A (reverse). The PCR conditions used were : Initial denaturation for 10 min at 94°C, 35 cycles of the following program: 94°C for 45 seconds, 53°C for 90 seconds, 72°C for 90 seconds with the final extension at 72°C for 10 minutes. Enzyme digestion was carried out using nuclear enzyme buffer (0.5 μl), bovine serum albumin (0.01 μl) and restriction endonuclease (Pst I). To 5 μl of the PCR product, 1 μl of the enzyme mix was added and incubated at 37°2C for 2 hours. The digested PCR fragments were separated by electrophoresis on 3% agarose gel for 2 hours. Ethidium bromide staining was used to reveal the fragments under an ultraviolet light transillumination. Ile198Thr polymorphism {#Sec6} ---------------------- Assay parameters and conditions were similar as in the Ala379Val except that the restriction enzyme was Pvu II instead of Pst I, and the primer sequences were as follows: ATA AAT AAT TTT TGC TTG TAT (forward) and AGA GCC AAG ACT TGT CAG CT (reverse). Results {#Sec7} ======= Distribution of PAFAH (Ile198Thr & Ala379Val) variants {#Sec8} ------------------------------------------------------ In view of small sample size, both variants Ile198Thr and Ala379Val were combined for analysis. In Table [2](#Tab2){ref-type="table"}, the study population was cross classified by three variables (ethnicity, presence or absence of NEC and the study genotypes). NEC group: Among the NEC group, the homozygous abnormal genotype (198ThrThr + 379ValVal) was not seen in any infants. Eight infants (1 Caucasian and 7 African-American infants) had heterozygous abnormal genotype (198IleThr + 379AlaVal). The normal homozygous (198IleIle + 379AlaAla) genotype was found in 28 infants (19 Caucasian and 9 African-American infants). Control group: Among the control population of healthy infants without NEC, homozygous abnormal genotype (198ThrThr + 379ValVal) was seen in 29 infants (7 Caucasian + 22 African-American infants). Heterozygous abnormal genotype (198IleThr + 379AlaVal) was identified in 159 infants (36 Caucasian + 123 African-American infants). Normal homozygous (198IleIle + 379AlaAla) genotype was found in 346 infants (149 Caucasian + 197 African-American infants).Table 2**Study population classified based on three variables (ethnicity, presence or absence of NEC and the study genotypes)**PAFAH GenotypesEthnicityNEC(1)(2)(3)TotalCaucasianYes011920CaucasianNo736149192African-AmericanYes07916African-AmericanNo22123197342Total29167374570PAFAH Genotypes -- (1) is 198 ThrThr + 379 ValVal (homozygous abnormal genotype), (2) is 198 IleThr + 379AlaVal (heterozygous abnormal genotype) and (3) is 198IleIle + 379AlaAla (homozygous normal genotype). Statistical analysis {#Sec9} -------------------- Statistical analysis was performed using SAS Version 9.2 (SAS Institute Inc., Cary, NC, USA). A total of 570 infants were classified into three PAFAH genotypes-198ThrThr + 379ValVal or homozygous abnormal genotype (represented as genotype 1), 198 IleThr + 379AlaVal or heterozygous abnormal genotype (represented as genotype 2) and 198IleIle + 379AlaAla or homozygous normal genotype (represented as genotype 3). Table [2](#Tab2){ref-type="table"} shows the frequency of the genotypes stratified by ethnicity (Caucasian or African-American) and disease status (NEC present or absent). Given the relatively small sample size for genotype 1, which has zero infants with occurrence of NEC, it was combined with genotype 2 and the combined group was considered abnormal genotype. This combined cross classified table is illustrated in Table [3](#Tab3){ref-type="table"}.Table 3**Study population cross classified by three variables (ethnicity, presence or absence of NEC and the study genotypes)**EthnicityNECAbnormal genotypeNormal genotypeTotalCaucasianYes11920CaucasianNo43149192African-AmericanYes7916African-AmericanNo145197342Total196374570In view of small numbers, the homozygous abnormal genotype (genotype 1 in Table [2](#Tab2){ref-type="table"}) and heterozygous abnormal genotype (genotype 2 in Table [2](#Tab2){ref-type="table"}) were combined into one category (abnormal genotype). Multiple logistic regression analysis was used to determine the effects of genotype and race on NEC occurrence adjusted for the effects of each other with the following results (Table [4](#Tab4){ref-type="table"}) Adjusted for the effect of genotype, odds for NEC among Caucasian infants were 2.03 times the odds for NEC among African-American infants in the study population. The adjusted odds ratio (OR) was significant with p-value \<0.05 and a 95% confidence interval (CI), which did not contain 1 (value of OR under the null hypothesis of no association between race and NEC).Table 4**Results of multiple logistic regression analysis on the study population for the occurrence of NEC**FactorOdds ratio (OR)95% CI for ORp-valueRace (C vs. AA)2.031.01 -- 4.070.046\*GenotypeAbnormal vs. Normal0.620.27 -- 1.420.26^NS^C- Caucasian; AA -- African-American. \*Significant at 5% level (0.01 \< p-value \< 0.05) ^NS^ Not significant at 5% level (p-value \> 0.05). Adjusted for effect of race, genotype had no significant effect on NEC; that is, adjusted for race, odds for NEC among infants with abnormal genotype was similar to those among infants with normal genotype. \[Note: Although the adjusted OR for genotype was not significant, the calculated OR (\< 1) seemed to suggest that infants with abnormal genotype had smaller odds for NEC than the infants with normal genotype\]. Alternatively, since the focus of the study was on the role of polymorphisms in the etiology of NEC, the 3-way cross classified table was analyzed with the 2 races as strata in determining the association between PAFAH genotype and NEC using the Cochran-Mantel-Haenszel (CMH) test. The CMH statistic, its 95% CI and its p-value were calculated as CMH statistic = 1.65, 95% CI = (0.71 to 3.84), p-value = 0.26. The CMH statistic was not significant as its 95% CI contained 1 and its p-value \> 0.05. Thus, when the infants were stratified by race (Caucasian or African-American) there was no association between genotype and NEC. The p-value for the CMH statistic (p = 0.26) was the same as the p-value for the adjusted OR for abnormal vs. normal genotype (p = 0.26) using multiple logistic regression analysis. Also, the CMH did not test for effect of race; rather it was used to stratify the infants. Discussion {#Sec10} ========== Necrotizing enterocolitis (NEC) remains one of the leading causes of morbidity and mortality in neonatal intensive care units, with an incidence of 10% among very low birth weight infants (\<1500 grams) Uauy et al. ([@CR36]) and a mortality of approximately 26% Hack et al. ([@CR16]). The lower the gestational age and birth weight, higher is the risk of developing NEC. Despite improvements in overall neonatal survival, this mortality rate has not changed significantly over the last three decades Henry & Moss ([@CR17]; Ewer [@CR11]). In addition to the mortality, NEC is associated with high morbidity resulting from multiple surgeries, need for parenteral nutrition, cholestatic liver disease and development of short gut syndrome. Although several predisposing factors such as prematurity, bacterial infection, enteral feeding and gut ischemia have been identified, the exact pathogenesis of NEC remains unknown Beeby & Jeffery ([@CR2]). Regardless of the initial triggering factor, the final pathology is similar which strengthens the possibility of the role of inflammatory mediators such as platelet activating factor (PAF) Ewer ([@CR11]; Muguruma et al. [@CR25]; Caplan et al. [@CR9]). PAF is an endogenous phospholipid with potent proinflammatory actions. It is synthesized from the plasma membrane precursors by cells such as intestinal cells and neutrophils. The half-life is very short and it is degraded by the enzyme PAF acetylhydrolase (PAFAH). PAFAH can be therefore be considered as the functional antagonist of PAF. The PAF receptor is a G protein-coupled seven transmembrane domain receptor family (GPCR). PAF combines with its receptor which is present on most cells resulting in activation of the inflammatory cascade Caplan et al. ([@CR9]; Prescott et al. [@CR28]). The exact mechanism of pathogenesis of PAF in NEC is not known, but exposure to a high local concentration of PAF results in increased mucosal permeability, apoptosis, activation of inflammatory cascade and finally bowel necrosis Frost et al. ([@CR13]). Numerous clinical and experimental studies have substantially strengthened our understanding of the role of PAF and PAFAH in the pathogenesis of NEC. The PAFAH activity is generally lower in newborns compared to adults, which enhances the susceptibility of neonates to the NEC Caplan et al. ([@CR5]). NEC induced by intravenous administration of PAF has been documented in neonatal experimental piglet models Ewer et al. ([@CR12]) and also in rats Gonzalez-Crussi & Hsueh ([@CR14]). Both in human and animal subjects, it has been proven that in NEC, the local as well as systemic concentrations of PAF were significantly high with concurrent decreased expression of PAFAH Caplan et al. ([@CR5]; Lu et al. [@CR22]). Increased concentration of stool PAF in NEC has also been documented Amer et al. ([@CR1]). In contrast, reduced activity and concentration of PAFAH has been found in the neonatal population particularly in infants with NEC Caplan et al. ([@CR5]). Additionally, enteral administration of recombinant PAFAH or PAF receptor antagonist to neonatal rats resulted in a significant decrease in NEC Caplan et al. ([@CR7];[@CR8]) strengthening the concept of functional antagonism of PAFAH in the NEC medicated by PAF. The human PAF acetylhydrolase gene is located in chromosome 6p12.p21.1. Stafforini et al. ([@CR30]). Reduced PAFAH has been documented in many disease states where PAF plays a role in mediating the inflammation such as in asthma Kruse et al. ([@CR20]), sepsis Graham et al. ([@CR15]), stroke Satoh ([@CR29]; Hiramoto et al. [@CR18]), multiple organ failure Partrick et al. ([@CR27]) and myocardial infarction Yamada et al. ([@CR38]). A difference in PAFAH activity between individuals/ethnicities has been extensively studied and genetic variations in the candidate genes has been proposed as the etiology for this difference in activity Tjoelker & Stafforini ([@CR34]; Stafforini et al. [@CR30]). The single nucleotide polymorphisms (SNPs) commonly described in the human PAFAH gene were Arg92His (exon 4), Ile198Thr (exon 7), Val279Phe (exon 9), Gln281Arg (exon 9) and Ala379Val (exon 11) Li et al. ([@CR21]). The influence of these PAFAH polymorphisms is well studied in many disease states where PAF plays a significant role in the pathogenesis. The role of SNPs in the pathogenesis of diseases is variable in different ethnicities and controversial Kruse et al. ([@CR20]; Li et al. [@CR21]). For example, in the Japanese population, two loss of function mutations on PAFAH gene have been described in asthmatic children and similar mutations were not seen in the Caucasian population with asthma Kruse et al. ([@CR20]; Stafforini et al. [@CR31]). The two PAFAH variants (Ile198Thr and Ala379Val) were also studied in conditions such as asthma Kruse et al. ([@CR20]), aging Campo et al. ([@CR4]), acute respiratory distress syndrome Li et al. ([@CR21]) and in schizophrenia Bell et al. ([@CR3]; Ohtsuki et al. [@CR26]). Even though the role of PAF in NEC is well known, similar SNP studies in the PAFAH gene have not been reported in the literature. We selected the two PAFAH variants (Ile198Thr and Ala379Val) based on the study by Kruse et al. ([@CR20]). Kruse et al. ([@CR20]) demonstrated that in Caucasian population, Thr198 and Val379, the two common variants of the PAFAH gene were associated with lower substrate affinities (higher *Km*) of PAFAH. They also demonstrated that decreased affinity of PAFAH for PAF results in reduced catalytic activity of PAFAH, which is believed to further prolong the activity of PAF and increase the risk of development of asthma and atopy. We applied this rationale to our study and attempted to find out the relationship between the two PAFAH variants and NEC. In our study population, adjusted for the effect of race, genotypic variants had no significant effect on NEC; that is, adjusted for race, odds for NEC among infants with an abnormal genotype was similar to those among infants with a normal genotype. Because of the controversial relationship of variant genotypes to the PAFAH enzyme levels in different ethnicities Kruse et al. ([@CR20]; Stafforini et al. [@CR31]) and our study being the first study of its kind in the literature, the exact significance of our observations is unknown at present. There are several limitations in this study that warrant discussion. First, the sample size was small. The concurrent measurements of PAF and PAFAH activity would be more meaningful. As our study was retrospective, concurrent measurements of PAF and PAFAH levels were not feasible. Another limitation is that the definition and the inclusion criteria for NEC was strictly based on diagnostic coding, the International Classification of Diseases, 9th revision, Clinical Modification, [@CR33]; ICD-9-CM. Even though very desirable, pertinent clinical information such as feeding history, sepsis, presence of patent ductus arteriosus, use of inotropes, indomethacin, etc. which play a vital role in the etio-pathogenesis of NEC were not taken for analysis due to unavailability. Our study sample did not have advanced stages of NEC such as stage 3 NEC (NEC with intestinal perforation or NEC with pneumatosis with perforation) or infants who died from NEC which is also a limitation. Despite the limitations, our study is a prototype pilot study in providing the prevalence of PAFAH gene polymorphisms in the Northwest Louisiana population. Also we attempted to find a relationship between PAFAH variants and NEC. A Medline search using the terms "necrotizing enterocolitis" and "platelet activating factor acetylhydrolase" revealed 20 articles. To the best of our knowledge, a detailed review of these articles did not reveal any previous study showing the role of PAFAH polymorphisms in the pathogenesis of NEC. Based on our results, we conclude that PAFAH genotype variants (Ile198Thr and Ala379Val) have no significant effect on NEC. A prospective study involving a larger population with concurrent measurements of PAF and PAFAH along with detailed clinical information is recommended to confirm or refute our initial findings. **Competing interest** The authors declare that they have no competing interests. **Authors' contribution** SS -- Conceptualized the study, analyzed the data, drafted the initial manuscript and revised. KY -- Conceptualized the study, designed the study, critically reviewed and revised the manuscript. DN -- Carried out the molecular genetic studies. GC -- Performed the statistical analysis. RD -- Critically reviewed and provided subject matter expertise, revised the manuscript. All authors read and approved the final manuscript. We sincerely thank Mr. John Cyrus for providing assistance in writing the manuscript.

Sporozoites of the human malaria parasite *Plasmodium falciparum* (Pf) express a surface protein, circumsporozoite protein (PfCSP), with an immunodominant central NANP (Asn-Ala-Asn-Pro) repeat region ([@cit0001]*--*[@cit0003]). Antibodies against the repeat can mediate protection from *Plasmodium* infection in animal models ([@cit0004]*--*[@cit0006]). However, anti-NANP antibody-- mediated protection is not readily achieved through vaccination. Thus, the induction of protective PfCSP NANP antibodies is a major goal in pre-erythrocytic vaccine development ([@cit0007]). We recently showed that the anti-NANP PfCSPmemory B cell response in Pf-naïve volunteers after immunizationwith live Pf sporozoites under chloroquine prophylaxis (PfSPZ-CVac)matured predominantly through the clonal selection and expansion of potent Pf inhibitory *IGHV3-33*-- and *IGKV1-5*-- encoded germline antibodies with an 8--amino acid--long immunoglobulin (Ig) light chain k complementarity-determining region 3 (CDR3) (this 8--amino acid CDR3 is hereafter designated KCDR3:8) ([@cit0008]*,* [@cit0009]). We analyzed five representative germline or low-mutated antibodies with reported affinities for a NANP 5-mer peptide (NANP~5~) between 10^−6^ and 10^−9^ M ([Fig. 1A](#f0001){ref-type="fig"} and table S1) ([@cit0009]). Antigen binding was abrogated when the original Ig Vκ1-5 light chain was replaced by Vk2-28 or when the native Ig heavy chains were paired with a Vκ1-5 light chain with 9--amino acid--long KCDR3 ([Fig. 1B](#f0001){ref-type="fig"}), demonstrating the importance of these specific Ig features in antigen recognition. ![**Affinity maturation of high-affinity human PfCSP NANP antibodies. (A)** Surface plasmon resonance (SPR) affinity and SHM of selected (labeled) V~H~3-33--Vκ1-5--KCDR3:8 (green) and non--V~H~3-33--Vκ1-5--KCDR3:8 (gray) anti-PfCSP antibodies ([@cit0009]). (**B** to **D**) Original and mutated antibodies. \[(B) and (C)\] PfCSP enzymelinked immunosorbent assay reactivity. Data in (A), (B), and (C) are from one experiment representative of at least two independent experiments. OD, optical density; Ab, antibody. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. (D) Pf liver cell traversal inhibition. Bars represent means from two to four independent experiments (symbols represent results from individual experiments). *\*\*P* = 0.01 (significant) for two-tailed Student's t test. (**E**) Silent (gray) and replacement (red) SHM (bars) in V~H~3-33--Vκ1-5 antibodies (n = 63). FWR, framework region; aa, amino acid. (**F**) Observed (obs) amino acid usage compared with a baseline (base) model ([@cit0022]*,* [@cit0023]). (**G** and **H**) Independent NANP~3~ SPR affinity measurements (dots) and means (gray lines). *\*\*P* = 0.01 (significant) for Bonferroni multiple-comparisons test; ns, not significant. KD, equilibrium dissociation constant.](Science-360-1358-g001){#f0001} All V~H~3-33--Vκ1-5--KCDR3:8 antibodies were encoded by the IGHV3-33^\*^01 allele ([@cit0009]). *IGHV3- 33^\*^01* differs from three otherwise highly similar gene segments (*IGHV3-30, IGHV3-30-3*, and *IGHV3-30-5*) at position 52 of heavy-chain CDR2 (HCDR2), which encodes strictly a tryptophan residue and not serine or arginine ([Table 1](#t0001){ref-type="table"} and table S2).HCDR2W^52^→S (H.W52_S)andH.W52_R mutants of the selected antibodies, as well as an H.W52_A mutant of antibody 2140 and a double mutant (H.V50_F\_W52_R) to mimic the *IGHV3- 30\*02* and *IGHV3-30-5^\*^02* alleles, all showed reduced PfCSP repeat reactivity associated with reduced in vitro parasite inhibitory activity ([Fig. 1, C and D](#f0001){ref-type="fig"}; single-letter amino acid abbreviations are defined in the legend to [Fig. 1](#f0001){ref-type="fig"}). ###### HCDR2 residues encoded by different IGHV3-33, IGHV3-30, IGHV3-30-3, and IGHV3-30-5 alleles. Gene and allele data are from [www.imgt.org/genedb/](http://www.imgt.org/genedb/). Gene Allele(s) Residue at position ------------------ ------------------------------------------------------------------------ --------------------- --- --- --- ***IGHV3-33*** 01, 02, 03, 04, 06 V I W Y ***IGHV3-33*** 05 V I S Y ***IGHV3-30*** 01, 03, 04, 05, 06, 07, 08, 09, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 V I S Y ***IGHV3-30-3*** 01, 02, 03 V I S Y ***IGHV3-30-5*** 01 V I S Y The majority of NANP-reactive V~H~3-33--Vκ1-5-- KCDR3:8 B cells belonged to clonally expanded and somatic hypermutation (SHM)--diversified cell clusters with strong selection for replacement mutations in HCDR1 (H.S31) and HCDR2 (H.V50 and H.N56), as well as KCDR3 \[KCDR3 S^93^ (K.S93)\], likely as a result of affinity maturation ([Fig. 1, E and F](#f0001){ref-type="fig"}) ([@cit0009]). The introduction of missing somatic mutations (mut) or reversions (rev) at H.V50 and, to a lesser extent, H.S31 revealed a role in binding to a minimal NANP~3~ peptide ([@cit0010]*,* [@cit0011]), as demonstrated for the germline antibody 2163 and the low-mutated antibody 1210 ([Fig. 1, G and H](#f0001){ref-type="fig"}, and table S3). In contrast, exchanges at H.N56 and K.S93, either alone (in antibodies 1210_H.K56_N^rev^, 1210_K. N93_S^rev^, and 2163_H.N56_K^mut^) or in combination (in 1210_NS and 2163_KN), showed no significant effect ([Fig. 1, G and H](#f0001){ref-type="fig"}, and table S3). Thus, affinitymaturation to the repeat explained the strong selection for only two of the four characteristic replacementmutations in V~H~3-33-- Vκ1-5--KCDR3:8 anti-NANP antibodies. We next determined the cocrystal structure of the 1210 antigen-binding fragment (Fab) with NANP~5~ ([Fig. 2](#f0002){ref-type="fig"}, fig. S1A, and tables S4 to S6). The NANP core epitope contained a type I β turn and an elongated conformation ([Fig. 2, A and C](#f0002){ref-type="fig"}, and fig. S1B), similar to NANP bound to a chimeric 2140 Ig heavy chain--1210 Ig k antibody and in line with previous observations (fig. S1C and tables S4 and S7) ([@cit0010]*--*[@cit0014]).Main-chain atoms in KCDR3 were optimally positioned to mediate H bonds with the repeat, likely contributing to the strong selection of KCDR3:8 ([Fig. 2, B and C](#f0002){ref-type="fig"}, and tables S2, S5, and S10). V~H~3-33 germline residues, notably H.V50 and H.W52 (the residue encoded only by *IGHV3-33* alleles), as well as H.Y52A and H.Y58 in HCDR2, mediated the majority of antigen contacts (table S5 and fig. S2) ([@cit0015]). Affinity maturation at H.V50 and H.S31 may be explained by strengthened van der Waals interactions with the repeat ([Fig. 2C](#f0002){ref-type="fig"}). ![**Affinity maturation drives homotypic repeat binding.** (**A** to **H**) 1210 Fab-NANP~5~ cocrystal structure. (A) Superposition of the four NANP-bound Fabs. (**B**) Surface representation of the antigen-antibody interaction. (**C**) Details of core epitope recognition by 1210. Black dashes indicate H bonds. (**D**) Two 1210 Fabs in complex with NANP~5~. \[(**E**) and (**F**)\] Surface representations of Fab-B (E) and Fab-A (F). Residues involved in homotypic interactions are dark gray. \[(**G**) and (**H**)\] Details of homotypic interactions. Affinitymatured residues are labeled in red. (**I**) Mean ± SEM K~D~ determined by isothermal titration calorimetry (ITC). Dots represent independent measurements. One-tailed Mann-Whitney test: *\*P* \< 0.05, *\*\*P* \< 0.01. (**J**) Results from size exclusion chromatography coupled with multiangle light scattering (SEC-MALS) for the 1210 Fab--PfCSP complex. The red line indicates mean ± SD molar mass from two measurements. RIU, refractive index units. (**K**) Two-dimensional class averages for the 1210 Fab--PfCSP complex. Red arrows indicate individual Fabs, and red lines indicate the binding angle observed in the crystal structure (D). NF54, Pf strain. Scale bar, 10 nm.](Science-360-1358-g002){#f0002} Notably, our crystal structure also revealed that two 1210 Fabs (designated 1210 Fab-A and Fab-B) bound to one NANP~5~ peptide in a headto- head configuration at a 133° angle ([Fig. 2D](#f0002){ref-type="fig"} and fig. S3). This binding mode led to six homotypic antibody-antibody H bonds providing 263 Å^2^ of buried surface area (BSA) between the two Fabs and an additional \~120 Å^2^ of BSA between the Fabs and the repeat ([Fig. 2, E and F](#f0002){ref-type="fig"}, and tables S5, S6, and S10). Two highly selected mutations, H.N56_K and K.S93_N ([Fig. 1, E and F](#f0001){ref-type="fig"}), formed H bonds with H.Y52A and H.S99 in the opposing Fab, thereby stabilizing the head-tohead configuration ([Fig. 2, G and H](#f0002){ref-type="fig"}). KCDR3:8 optimally contacted HCDR3 of the opposite 1210 molecule, providing another explanation for the length restriction in KCDR3. To investigate homotypic interactions, we next measured the Fab affinities for NANP~5~ and NANP~3~ for 1210, 1210_NS (which lacks the selected mutations involved in homotypic binding), a 1210H.D100_Y^mut^ K.N92_Y^mut^ mutant (1210_YY, designed to disrupt head-to-head binding through steric clashes), and a 1210 germline antibody (1210_GL) ([Fig. 2I](#f0002){ref-type="fig"} and fig. S4). Compared with 1210, 1210_YY and 1210_NS showed significantly weakened affinity for NANP~5~ but not for NANP~3~, whereas 1210_GL was significantly worse than 1210 at binding both peptides ([Fig. 2I](#f0002){ref-type="fig"} and fig. S4) ([@cit0016]). These data suggest that only 1210 efficiently recognized the repeat in a high-affinity homotypic head-to-head binding configuration. An analysis of full-length PfCSP with 38 NANP repeats confirmed this hypothesis. Approximately twelve 1210 Fabs bound PfCSP and recognized the NANP repeat in a head-to-head binding configuration similar to the 1210 Fab--NANP~5~ crystal structure ([Fig. 2, J and K](#f0002){ref-type="fig"}, and fig. S3D) ([@cit0011]*,* [@cit0017]). Furthermore, 1210_YY IgG, with its restricted ability to engage in homotypic antibody interactions, showed a lower binding avidity to fulllength PfCSP than 1210 (fig. S5). Thus, affinity maturation selects for mutations that improve homotypic antibody interactions, thereby indirectly increasing PfCSP NANP binding. To better understand the selection of SHM at the cellular level, we measured the degree of B cell activation in response to NANP~5~ of transgenic B cell lines expressing 1210 or variant B cell receptors (BCRs) ([Fig. 3, A to D](#f0003){ref-type="fig"}). BCR signaling was delayed in cells expressing 1210_GL compared with that in cells expressing 1210. This effect was even more pronounced in 1210_YY mutant cells. As expected, 1210_H.V50_I^mut^ (1210 with HCDR2 V^50^→I),withhigh repeataffinity,mediated stronger signals than 1210, especially with low antigen concentrations, whereas 1210_NS showed no significant differences ([Fig. 3D](#f0003){ref-type="fig"}). Thus, B cell activation is promoted by both direct NANP binding and homotypic antibody interactions. Despite a 2-log difference in NANP~3~ affinities ([Fig. 1, G and H](#f0001){ref-type="fig"}) and the varied potential of these antibodies to engage in homotypic interactions, all showed similar capacities to inhibit Pf sporozoites in vitro ([Fig. 3E](#f0003){ref-type="fig"} and fig. S6). Likewise, all antibodies conferred similar levels of dose-dependent protection from the development of blood-stage parasites after passive immunization in mice, presumably because of strong avidity effects ([Fig. 3F](#f0003){ref-type="fig"}). These data provide a mechanistic explanation for the strong in vivo selection of antihomotypic antibody mutants by affinity maturation, independently of their protective efficacy as soluble antibodies. ![**B cell activation and parasite inhibition.** (**A** to **D**) NANP~5~-induced calcium signaling of 1210 and variants. \[(A) and (B)\] Reaction kinetics and percentages of activated cells (A) and overlay of median signal intensities (B) with 1 mg/ml NANP5 for one of at least six representative experiments. Indo, calcium indicator. \[(C) and (D)\] Percentages of activated cells and median activation time after the addition of 1 mg/ml (C) (*n* = 6 or 7 experiments) and 0.1 mg/ml (D) (*n* = 3 experiments) NANP~5~. Symbols indicate results from independent experiments, and lines and error bars indicate means ± SD. *\*\*P* = 0.01 (significant) for Bonferroni multiple-comparisons test. (**E** and **F**) Parasite inhibition. (E) Mean ± SD median inhibitory concentration (IC~50~) values from at least three independent experiments for 1210 and 2163 antibodies with indicated NANP~3~ affinities.We detected no significant differences between IC~50~ values because of extensively overlapping confidence intervals. (F) Percentages of parasite-free mice after passive immunization with 30 or 100 mg of 1210 or variants 24 hours before subcutaneous injection with *Plasmodium berghei* sporozoites expressing PfCSP (Pb-PfCSP). Data are from one (100 mg) or two (30 mg) independent experiments with five mice per group.We detected no significant differences in survival for 1210 variants (Mantel-Cox test).](Science-360-1358-g003){#f0003} V~H~3 antibodies dominate the anti-PfCSP memory response ([@cit0009]*,* [@cit0011]*,* [@cit0014]). In addition to V~H~3-33-- Vκ1-5--KCDR3:8, we observed a cluster of highly mutated, affinity-matured V~H~3-23--Vκ1-5 NANPreactive memory B cell antibodies in our selection ([Fig. 4, A andB](#f0004){ref-type="fig"}) ([@cit0009]). Although the NANP~5~-binding mode of a representative V~H~3-23--Vκ1-5 antibody, 1450, was different from that of 1210, it also recognized NANP~5~ in a head-to-head configuration, with HCDR3s in direct juxtaposition and the affinity-matured K.N30 residues forming an H bond between Fab-A and Fab-B ([Fig. 4, C to E](#f0004){ref-type="fig"}; fig. S7; and tables S4, S8, and S9). Sequence analysis of the V~H~3-23--Vκ1-5 antibody cluster confirmed enrichment for amino acid exchanges that participate directly in antibody-antigen interactions or antibody-antibody contacts or favor a 1450 paratope conformation optimal for NANP epitope recognition ([Fig. 4B](#f0004){ref-type="fig"}). ![**Antihomotypic affinity maturation in *IGHV3-23*-encoded PfCSP NANP antibodies. (A)** SPR affinity and SHM of 1450 out of all V~H~3-23--Vκ1-5 (green) and non--V~H~3-23--Vκ1-5 (gray) anti-PfCSP antibodies ([@cit0009]). (**B**) Silent and replacement SHM (bars) in V~H~3-23--Vκ1-5 antibodies (*n* = 100). (**C** to **E**) Fab 1450--NANP~5~ cocrystal structure. Head-to-head binding mode (C), Fab-Fab (**D**), and Fab-NANP~5~ (E) interactions. Black dashes indicate H bonds. Affinity-matured residues are colored according to the SHM amino acid usage scheme and are labeled in red. Observed amino acid usage is compared with a baseline model ([@cit0022]*,* [@cit0023]). (**F**) V~H~3-33-- Vκ1-5--KCDR3:8 or V~H~3-23--Vκ1-5 antibodies in total memory B cells ([@cit0018]), CD19^+^CD27^hi^CD38^hi^ plasmablasts (PB), and CD19^+^CD27^+^ PfCSP-reactive memory B cells (CSPmem) ([@cit0008]*,* [@cit0009]). Dots represent subsamples of 1500 sequences. Box plots show the median, SD, maximum, and minimum of the distribution. *\*\*\*P* = 0.001 (significant) for two-tailed Student's t test. (**G**) Frequency of V~H~3-33--Vκ1-5-- KCDR3:8 and V~H~3-23--Vκ1-5 antibodies among clonally expanded versus singlet pooled PB and CSPmem ([@cit0009]).](Science-360-1358-g004){#f0004} After the immunization of malaria-naïve individuals with PfSPZ-CVac, \~15% of PfCSP-reactive memory B cells showed V~H~3-33--Vκ1-5--KCDR3:8 or V~H~3-23--Vκ1-5 sequence characteristics ([Fig. 4F](#f0004){ref-type="fig"}). Furthermore, these cells were strongly enriched in the expanded anti-PfCSP memory B cell pool compared with the nonexpanded population ([Fig. 4G](#f0004){ref-type="fig"}). Thus, antihomotypic affinity maturation is observed after repeated Pf sporozoite immunization ([@cit0008]*,* [@cit0009]) in both low-mutated high-affinity V~H~3-33 antibodies and lower-affinity antibodies utilizing other gene combinations. This phenomenon also likely takes place in B cell responses elicited by RTS,Smalaria vaccination (fig. S8) ([@cit0011]). Thus, antihomotypic affinity maturation, in addition to traditional antibody-antigen affinity maturation, promotes the strong clonal expansion and competitive selection of PfCSP-reactive B cells in humans. Even in the absence of affinity maturation, V~H~3-33--Vκ1-5--KCDR3:8 antibodies are moderate to strong NANP binders and potent Pf inhibitors. This critically depends on H.W52 in HCDR2. Because *IGHV3-33* is located in a region of structural polymorphism of the *IGH* locus, haplotype frequencies, especially in areas where Pf is endemic, may determine the efficient induction of protective humoral anti- PfCSP repeat responses upon vaccination ([@cit0019]). Indeed, one donor in our study was *IGHV3-33* negative (fig. S9). We propose that antihomotypic affinity maturation may be a generalizable property of B cell responses if a repetitive antigen (malarial or other) brings two antibodies into close proximity to optimize binding and promote clustering of surface Igmolecules through homotypic interactions ([@cit0020]*,* [@cit0021]). Supplementary Material ====================== ###### Antihomotypic affinity maturation improves human B cell responses against a repetitive epitope ###### Click here for additional data file. We thank C. Busse (German Cancer Research Center) for general discussions; R. Übelhart (NCT, Heidelberg) for experimental advice; and D. Foster and C. Winter (German Cancer Research Center) and C. Kreschel, H. Krüger, D. Tschierske, L. Spohr, D. Eyermann, and M. Andres (Max Planck Institute for Infection Biology) for technical assistance. We acknowledge S. M. Khan and C. J. Janse (Leiden University Medical Center, Leiden, The Netherlands) for providing transgenic Pb-PfCSP parasites. HC-04 human hepatocytes (MRA-975), contributed by J. S. Prachumsri, were obtained from BEI Resources. We are grateful to the Genomics & Proteomics and Chemical Biology Core Facilities (German Cancer Research Center) for gene sequencing services and assistance with SPR measurements, respectively, and to the Structural & Biophysical Core Facility (Hospital for Sick Children) for access to the ITC, biolayer interferometry, and SEC-MALS instruments. This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. To view a copy of this license, visit <http://creativecommons.org/licenses/by/4.0/>. This license does not apply to figures/ photos/artwork or other content included in the article that is credited to a third party; obtain authorization from the rights holder before using such material. Funding ======= This work was funded by the Bill & Melinda Gates Foundation (OPP1179906; J.-P.J. and H.W.), as well as the German Research Foundation (CRC 1279, B03; IRTG-TRR130, P01; H.J.). This research was undertaken, in part, thanks to funding from the Canada Research Chairs program (J.-P.J.). S.W.S. was supported by a Hospital for Sick Children Lap- Chee Tsui postdoctoral fellowship and a Canadian Institutes of Health Research (CIHR) fellowship (FRN-396691). X-ray diffraction experiments were performed by using beamline 08ID-1 at the Canadian Light Source, which is supported by the Canada Foundation for Innovation, the Natural Sciences and Engineering Research Council of Canada, the University of Saskatchewan, the Government of Saskatchewan, Western Economic Diversification Canada, the National Research Council Canada, and the CIHR. X-ray diffraction experiments were also performed at GM/CA\@APS, which has been funded in whole or in part with federal funds from the National Cancer Institute (ACB-12002) and the National Institute of General Medical Sciences (AGM-12006). The Eiger 16M detector was funded by an NIH--Office of Research Infrastructure Programs High-End Instrumentation grant (1S10OD012289-01A1). This research used resources of the Advanced Photon Source, a U.S. Department of Energy (DOE) Office of Science user facility operated for the DOE Office of Science by Argonne National Laboratory under contract DE-AC02-06CH11357. Author contributions ==================== K.I., S.W.S., H.J., E.A.L., J.-P.J., and H.W. designed experiments; P.G.K., B.K.L.S., S.L.H., and B.M. provided clinical samples; K.I., S.W.S., G.C., and G.P.M. performed experiments; A.B., G.T., R.M., and V.R. contributed to the experimental work; K.I., S.W.S., G.C., G.P.M., E.A.L., J.-P.J., and H.W. analyzed the data; K.I., S.W.S., J.-P.J., and H.W. wrote the manuscript; and J.-P.J. and H.W. conceived the study. Competing interests =================== B.K.L.S. and S.L.H. are salaried employees of Sanaria, the owner of the PfSPZ-CVac and the sponsor of the clinical trial. B.K.L.S. and S.L.H. have financial interest in Sanaria. All other authors declare no conflicts of interest. Data and materials availability =============================== Structural data are deposited under Protein Data Bank (PDB) IDs 6D01, 6D0X, and 6D11. All other data needed to evaluate the conclusions in this paper are present either in the main text or in the supplementary materials. Materials from the German Cancer Research Center and the Max Planck Institute for Infection Biology will be available upon reasonable request under material transfer agreements (MTAs). Sharing of the NF54 *P. falciparum* parasite is limited by an MTA with the Radboud University Medical Center; sharing of the P. berghei strain Pb-PfCSP is limited by an MTA with the Leiden University Medical Center. SUPPLEMENTARY MATERIALS ======================= [www.sciencemag.org/content/360/6395/1358/suppl/DC1](http://www.sciencemag.org/content/360/6395/1358/suppl/DC1) Materials and Methods Figs. S1 to S9 Tables S1 to S10 References (*24--35*) [^1]: These authors contributed equally to this work.

{#F6} Julie Theriot graduated from the Massachusetts Institute of Technology as a double major in biology and physics, and her career as a biologist ever since has been notable for the quantitative rigor of her approach to the messy world of biology. As a graduate student at the University of California San Francisco, she began studying the subversion of actin polymerization by pathogenic bacteria in animal cells, and more general issues of bacterial and eukaryotic motility remain the focus of her group's research at Stanford University. With colleagues Rob Phillips, Jane Kondev, and Hernan Garcia, she has published a textbook, Physical Biology of the Cell, exploring the applications of mathematical and physical modeling in cell biology. In the interview here, she applies a breathtaking breadth of scholarship and a fearless imagination to the fundamental question of the difference between bacterial cells and ours. Aren't more and more similarities being found between bacterial cells and eukaryotic ones? How different are they in fact? ========================================================================================================================== It is true that over the past 15 or 20 years we have identified a surprisingly large number of molecular similarities between bacterial cells and eukaryotic cells. Of course we have known about the profound similarities across the entire phylogenetic tree of life in many of the machines of the central dogma (ribosomes, polymerases, and so on) and the enzymes of central metabolism, but now we've also found homologs of the major eukaryotic cytoskeletal proteins in bacteria and many other surprises. But it is still a fundamental observable fact that the vast majority of bacterial cells are physically small and morphologically simple compared with the vast majority of eukaryotic cells. There are certainly exceptions to this - there are bacteria that are large and complicated and there are eukaryotes that are small and simple - but if you just look at any random bacterium versus a random eukaryote, it is clear that there is a fundamental quantitative and qualitative difference in size and complexity. Archaea, which make up the third major domain of life, have some molecular signatures that seem quite similar to those in eukaryotes \[[@B1]\], but morphologically they look very much like bacteria. Indeed this is the reason that we didn't recognize them as a distinct domain until very recently \[[@B2]\]. The overall argument about the origins of morphological complexity that I want to make here applies equally to bacteria and archaea, but I'm going to focus on bacteria for specific examples just because we know so much more about them. The most obvious difference between eukaryotes and bacteria is that there is a membrane-bounded nucleus in eukaryotes and not in bacteria - again, for the most part: there is a bacterium with the wonderful name *Gemmata obscuriglobus* that is described as having a double membrane enclosing the DNA in a nucleus-like structure \[[@B3]\], although the structure is apparently contiguous with the plasma membrane \[[@B4]\], so in that sense it is very different from a eukaryotic nuclear membrane and this is certainly a special case. But leaving that example aside, the main consequence biologically of having a membrane-enclosed nucleus is that transcription and translation are uncoupled. So there is a fundamental kinetic and organizational difference between eukaryotes and bacteria in the way that genetic information is expressed in the form of protein and is therefore allowed to be converted into cellular structure, function and organization. So how does that affect the function of bacterial and eukaryotic cells? ======================================================================= Well, let's now think a little bit about what other cellular features go along with a membrane-enclosed nucleus. Another major difference between eukaryotes and bacteria is the proliferation of other membrane-bounded organelles, of which you see many different kinds within single eukaryotic cells - for example, the Golgi apparatus, the endoplasmic reticulum, and so on. Again, there are a few bacteria that have internal membranes, although in most cases those membrane-enclosed organelles in bacteria are contiguous with the plasma membrane, like the pseudo-nuclear membrane of *Gemmata*. One example is the magnetosomes of the bacterium *Magnetospirillum magneticum*; these are little crystals of magnetite wrapped inside of membrane invaginations that the cells use to orient themselves along the earth's magnetic field lines \[[@B5]\]. Because these structures are continguous with the plasma membrane, they don't really act as topologicaly separate compartments. A critically important exception is the cyanobacteria, which carry out photosynthesis in the elaborate thylakoid endomembrane system. The thylakoids do appear to be truly separate from the plasma membrane and can be topologically quite complicated \[[@B6]\]. But although we know quite a lot about the mechanisms of photosynthesis in the thylakoids, we know relatively little about membrane traffic in these organisms, so I can't really comment on how similar their organizational mechanisms are to eukaryotic endomembranes. Going along with the proliferation of membrane-enclosed organelles in eukaryotes is usually a higher degree of subcellular compartmentalization, of assigning different kinds of functions to different regions of the cell. And of course, eukaryotes have endosymbionts, the mitochondria and chloroplasts that used to be bacteria that the eukaryotes have taken into themselves and tamed for their own purposes \[[@B7]\]. Another major observable difference is that eukaryotic cells are able to make very big, fancy, multicellular organisms like redwood trees and elephants. Among the three major groups of macro-organisms (those visible to the naked eye), animals and plants are the better studied, but the largest fungi are also remarkable for their vast size and lifespan \[[@B8]\]. Bacteria can also form multicellular structures, such as biofilms, that require complex intercellular signaling and developmental programs, as well as deposition of extracellular matrix \[[@B9]\], but they do not approach the structural complexity of eukaryotic multicellular organisms. The largest of the bacterial communities are formed by cyanobacteria and are called stromatolites; these are made up of beautiful layered structures that form through cycles of bacterial growth, matrix deposition, and accretion of mineral particles \[[@B10],[@B11]\]. Stromatolite structures, though, have remained fundamentally unchanged for over three billion years, as stromatolites make up the oldest recognizable fossils of living organisms. They flourished until the Cambrian explosion, when they became much more rare as, presumably, the newly evolved animals began to crawl around and nibble on them. In support of this idea, stromatolites became more abundant in the fossil record after the major extinction events that wiped out most of the animals, and then receded again when the animals bounced back \[[@B12]\]. Today the only living stromatolites are found in extremely salty bays that are hostile to animal life. So I would say qualitatively in terms of complexity as well as direct competition, true and highly evolvable (and apparently hungry) multicellularity is a feature of the eukaryotes, not of the bacteria. Finally, and I think not coincidentally, eukaryotes typically have genomes that are greatly expanded in length by as much as several orders of magnitude beyond those of bacteria, and those genomes usually contain a lot more noncoding DNA whose function we don't understand. Can you explain why eukaryotes have such an expanded genome, given that we don't think most of it is doing much or we don't know what it's doing? Sadly I don't have an answer to that question, and as you know the possible function of noncoding DNA is an intensely controversial area right now \[[@B13],[@B14]\]. I will point out that it has been known for quite a while that genome size in a wide variety of organisms seems to correlate better with cell size than with number of protein-coding genes or apparent complexity \[[@B15]\], so if cell size itself is a selectable trait that might be part of the answer. But what I am going to try to explain is why eukaryotes do not seem to worry about how much extra DNA they are carrying around. In principle that opens an opportunity for picking up more genes and more chromosomes, more bits of DNA whose function may not yet be obvious to us, but may well be important to the cells that are carrying it. Or might evolve =============== Yes, or might evolve. Having the capacity to carry around and segregate lots and lots of DNA also just gives the eukaryotic cells more options and more flexibility. The much larger cell size for eukaryotic cells, which seems to be connected with all of the other differences between eukaryotes and bacteria, brings up the issue of the diffusion limit, which Kevin Young wrote about in his contribution to the Forum you recently published on cell size \[[@B16]\]. That was a terrific article, and I agree with everything he said, but I think he didn't take the argument quite far enough, and that's what I'm going to do next. His essential point was that bacterial size and structure are constrained by the need to import nutrients efficiently and divide accurately through mechanisms that depend only on diffusion. Even some of the largest bacterial cells we know are still effectively diffusion-limited; for example, *Thiomargarita namibiensis* appears as a sphere up to 750 μm across, easily visible to the naked eye, but is organized as a very thin shell of cytoplasm, less than 2 μm thick, surrounding a gigantic vacuole \[[@B17]\]. But as soon as you can set up an intracellular molecular transport machinery such as a filamentous cytoskeleton and associated molecular motors, then having the genome be readily accessible to diffusive transport becomes less of an issue, freeing up eukaroytic cells to become physically large. What we'd really like is some simple, cogent explanation that ties all of these eukaryotic features together: the membrane-enclosed nucleus, the elaboration of other topologically separate membrane-bound compartments, the ability to capture endosymbionts, the ability to make fancy multicellular organisms, the greatly expanded genome, and the large cell size. When I was in graduate school, the explanation was known and it was very straightforward. It was that eukaryotes have a cytoskeleton and bacteria do not. If you go down the list of all the things that are special about eukaryotic cells, you can ascribe virtually all of them to functions of the cytoskeleton. For example, you need structural elements, including microtubules, to organize the membrane-enclosed nucleus and the extensive internal membrane system. And coming back to the expanded genome, we can see that it is simple to divide if you have a mitotic spindle, because adding another chromosome, or even doubling or quadrupling the size of your genome, is no big deal; the mitotic spindle can take care of segregating extra chromosomes using the same mechanism that it uses to segregate just a few. This is because eukaryotic spindles use essentially the same microtubule-kinetochore interface structure repeated for every chromosome, and the collective decisions such as when to enter anaphase are carried out by checkpoint machineries that enforce the rule that all of the kinetochores must be attached before the next step can proceed \[[@B18]\]. In contrast, bacteria that have multiple chromosomes seem to segregate them by using independent, orthogonal machineries specific for each chromosome \[[@B19]\], and don't appear to have anything as general or as scalable as a mitotic spindle. Turning to the actin cytoskeleton, this is also vital for many of the eukaryotic-specific features we have discussed. Dynamic actin assembly and disassembly are necessary for phagocytosis, to separate a large membraneous organelle from the plasma membrane compartment, and to also capture an endosymbiont \[[@B20]\]. And then to make a multicellular organism, you need two kinds of interactions between cells. First, you need the ability to lay down an extracellular matrix, which bacteria are also perfectly capable of doing. But then you need some kind of structural elements within cells that can connect to the extracellular matrix and to one another in such a way that forces can be continuously transmitted from the cells to the matrix and from one cell to another. This is the property that is necessary for cells to make simple tissues such as epithelia, where sheets and ensembles of cells can get bigger and bigger and perform coherent behaviors. In animal cells, these processes rely on the actin cytoskeleton \[[@B21]\], and there is evidence that similar cytoskeleton-based processes are also necessary for simpler kinds of multicellularity in non-metazoan eukaryotes such as *Dictyostelium*\[[@B22]\] and *Volvox*\[[@B23]\]. The problem with this argument about the basis of the difference between eukaryotes and bacteria is that it all depends on bacteria not having a cytoskeleton, which is what we believed in the early 1990s. But then it was discovered by several very convincing converging lines of evidence, spearheaded by Joe Lutkenhaus, that the bacterial protein FtsZ, which forms a ring around the middle of the bacterial cell and has an essential role in cell division \[[@B24]\], is a homolog of tubulin \[[@B25],[@B26]\]. And when the atomic structures for both tubulin and FtsZ were solved at the same time, it was absolutely clear that they were nearly superimposable and almost certainly true homologs in the sense of being derived from a common ancestor \[[@B27],[@B28]\]. So there went the assumption that bacteria do not have a cytoskeleton. My research up until that point had focused on the actin cytoskeleton, so for a little while I could maintain my eukaryotic-centric world view by saying to myself that bacteria have tubulin but they don't have actin, and so that must be the most important difference between us and them. But then a few years later, in a series of quite spectacular papers where the cell biological evidence for the shape-determining role of a certain class of bacterial actin-like proteins including MreB \[[@B29]\], was staggeringly confirmed by the undeniable structural similarity between MreB and actin \[[@B30]\], it was quite clearly demonstrated that bacteria do in fact have actin homologs. In the 10 years or so since that discovery, a lot of people have been searching for more different examples of actin and tubulin homologs in bacteria, and indeed we can find a tremendous number of such homologs, a vast proliferation with different biological functions, with various actin homologs like ParM involved in plasmid segregation \[[@B31]\] and MamK necessary for magnetosome alignment \[[@B5]\]. I'm particularly fond of the work of Joe Pogliano, who has gone searching for actins and tubulins carried by plasmids and bacteriophages, and has found an outrageously big zoo of both actins and tubulins \[[@B32],[@B33]\]. And in a few bacteria, there is even some evidence that they have homologs (or at least functional analogs) of intermediate filament proteins \[[@B34]\]. So we must absolutely acknowledge that the major eukaryotic cytoskeletal proteins are also present in our bacterial comrades, indeed there are many copies of them with distinct biological functions. So I would like to rephrase the question about what the difference is between eukaryotes and bacteria. We now know that everyone has a cytoskeleton, but still there are fundamental and easily observable morphological differences between these two domains of life, where eukaryotes have used their cytoskeletons to get larger and more morphologically complex and even truly multicellular, while bacteria basically have not done so. So the question I'd really like to ask is, if bacteria have a cytoskeleton, why don't they do anything more interesting with it? And are you going to explain why bacteria don't do what we do with our cytoskeletons? ===================================================================================== I am. At least, I have a hypothesis. It is an untested hypothesis, but I've been thinking about this now for a few years, and there is a lot of supporting evidence. I think it is at least a unifying concept that I hope will be provocative, and perhaps lead to experiments and analysis that might really test this idea. The starting point for my hypothesis is that the central feature of the cytoskeletal elements that are universally shared among organisms, and are necessary for cellular life, is the ability to form protein polymers that can give rise to large-scale cell organization and cell division via the dynamic assembly and disassembly of helical protein filaments. That is found everywhere. Besides the actin- and tubulin-related cytoskeletal proteins in bacteria, there are structures like bacterial flagella and bacterial pili, which are also fundamentally helical homopolymers of proteins. Bacteria are perfectly good at making those kinds of structures. They are perfectly good at governing the dynamics of those structures. So why don't they do anything more interesting with them? Here is my hypothesis: eukaryotes enhance the intrinsic assembly features of the helical filament protein systems with two particular kinds of cytoskeleton-associated factors, which have not yet been found in bacteria. And those two are regulated nucleators - centrioles for example - and linear stepping molecular motor proteins - the eukaryotic myosin and kinesin molecules. For actin, the best-characterized of the regulated nucleators is the Arp2/3 complex, which has two actin-related proteins as part of the complex and then five other proteins that hold them together \[[@B35]\] (Figure [1](#F1){ref-type="fig"}a). In its isolated form, the two actin-related proteins of the Arp2/3 point off in slightly different directions \[[@B36]\], but when the complex is activated for its nucleation activity they swing around to imitate the starting point of the two protofilaments of the actin filament structure, and this structural mimicry of the growing tip of an actin filament is probably the basis of the nucleating activity for the Arp2/3 complex \[[@B37]\]. For microtubules, the best characterized nucleator is the γ-tubulin ring complex, which has 13 copies of the protein γ-tubulin (a paralog of α- and β-tubulin) and then some other proteins that hold them in a slightly distorted ring that can template the growth of a microtubule with 13 protofilaments \[[@B38],[@B39]\] (Figure [1](#F1){ref-type="fig"}b). There are other actin nucleators and there are other microtubule nucleators that operate by different mechanisms. But it seems from those two examples that a very reasonable way to regulate the initiation and assembly of helical cytoskeletal polymers is to just make another copy of the gene for the subunit and then allow it to specialize a little bit so that it becomes a regulatable nucleator. Certainly that is the sort of thing that bacteria could do if they wanted. They would have no problem duplicating and modifying the genes for the cytoskeletal proteins, as they have demonstrated with the proliferation of the different flavors of actin and tubulin homologs that are used in such a wide variety of contexts. For example, *Bacillus subtilis* has three different chromosomally encoded paralogs, each of which is homologous to actin, MreB, Mbl, and MreBH, that appear to have somewhat overlapping functions \[[@B40]\]. But so far, we do not know of any specialized actin- or tubulin-related proteins in bacteria that are used specifically as regulated nucleators for their main self-assembling subunits MreB and FtsZ. ![**Cytoskeletal filament nucleation by modified subunits. (a)** Nucleation of actin filaments by the Arp2/3 complex. Left: diagram of Arp2/3 complex before and after activation, showing rearrangement of actin-like subunits leading to templated filament growth (Copyright 2008 from *Molecular Biology of the Cell,* 5th edition by Alberts *et al*. Reproduced by permission of Garland Science/Taylor & Francis LLC \[[@B41]\]). Right: electron micrograph showing the appearance of an actin filament nucleated by Arp2/3 (at the bottom) (from *Proc Natl Acad Sci U S A*\[[@B35]\]). **(b)** Nucleation of microtubules by the γ-tubulin ring complex. Left: diagram of microtubule templated from a ring complex (Copyright 2008 from *Molecular Biology of the Cell,* 5th edition by Alberts *et al*. Reproduced by permission of Garland Science/Taylor & Francis LLC \[[@B41]\]). Middle, structure of the ring complex by cryo-electron microscopy, showing how the γ-tubulins are held in the proper configuration to imitate a microtubule plus end (reprinted by permission from Macmillan Publishers Ltd: *Nat Rev Mol Cell Biol***12:**709**--**721, copyright 2011 \[[@B38]\]). Right, electron micrograph of the end of a microtubule nucleated from a ring complex (reprinted by permission from Macmillan Publishers Ltd: *Nat Cell Biol***2:**365**--**370, copyright 2000 \[[@B42]\]).](1741-7007-11-119-1){#F1} So why don't bacteria want regulated nucleation? ================================================ This is the corollary to my argument. If my hypothesis that bacteria do not have regulated cytoskeletal nucleation proteins is true - and I will go through the cell biological evidence that makes me think this is true - then the question is whether they really do not want to have them or whether they just never had the opportunity to develop them. I think, at least as far as nucleators go, the opportunity to develop them is not a very high barrier. So I think it must be that bacteria simply have a fundamentally different strategy for cytoplasmic organization as compared to eukaryotes. What makes you say it's not a high barrier? Do we have evidence that it's happened more than once in eukaryotes? ================================================================================================================ I don't have good evidence that forming nucleating factors by duplication of the subunits has happened more than once for each of the two major cytoskeletal structures because both the Arp2/3 complex \[[@B43]\] and the γ-tubulin ring complex \[[@B44]\] are very well conserved across all eukaryotes, so it is most likely that the relevant duplications happened fairly early in the eukaryotic lineage and have been maintained ever since. However, at least in the case of actin, there are many different, distinct molecular families of nucleators that can operate by different but equally simple mechanisms. For example, the actin nucleators Spire \[[@B45]\] and Cordon-bleu \[[@B46]\] both appear to nucleate actin by having a series of three or four domains that bind directly or indirectly to actin monomers; these domains can bring the actin subunits into close enough proximity and appropriate enough orientation to get over the kinetic barrier to actin nucleation and start the growth of a filament. In the particular case of this category of nucleators, I am quite confident that bacteria would be able to develop them if they wanted to, as indeed two bacterial pathogens are known to express secreted virulence factors that act as host cell actin nucleating factors by exactly this mechanism \[[@B47],[@B48]\]. For these virulence factors, it is not clear whether the pathogens picked up their actin nucleators by horizontal gene transfer or by convergent evolution, but in either case it is still striking that bacteria are easily able to nucleate eukaryotic actin filaments but do not seem to have any regulated protein nucleators for their own cytoskeletal filaments. But the thing that I think is really interesting about cytoskeletal filament nucleation in this context is that classically when we were taught the theory of protein polymerization from Fumio Oosawa \[[@B49],[@B50]\] and Terrell Hill \[[@B51],[@B52]\] and all those giants in the field, their argument was that it is important, kinetically, that nucleation be the rate-limiting step for polymer formation. And that is indeed observably true for actin and for microtubules and for the bacterial flagellum, the classical examples of helical protein self-assembly that they were trying to describe with their comprehensive theoretical treatments. But when people started doing very careful kinetic studies on the bacterial cytoskeletal proteins - and this I think has been done best for FtsZ \[[@B53]\] and for ParM \[[@B54]\] - it became clear that nucleation for the bacterial cytoskeletal proteins is actually very, very fast. It's spontaneous. The way bacterial cells regulate where they have their filaments is not by regulating the site of nucleation, but rather by regulating the sites of stabilization and destabilization of spontaneously nucleating filaments. For those of us who have been raised on the thermodynamic theory of protein polymerization in the context of cell biology, this is deeply shocking. Spatial localization of cytoskeletal components in bacteria simply appears to use a fundamentally different mode of organization from the one we see for all of the organized cytoskeletal assemblies in eukaryotes, and frankly we as cell biologists are justified in being a little bit freaked out. These are mechanisms that regulate fundamental processes, aren't they? This is bacterial cell division? ======================================================================================================= Yes, that's right. The dynamic cytoskeletal polymers found in bacteria seem to be just as important to the bacterial cells as they are to us eukaryotes, and they are involved in similarly crucial cell biological processes. Also the bacterial cytoskeletal proteins are very widely distributed among bacteria and even archaea \[[@B55],[@B56]\]. So again, my premise is that since we must now accept that bacteria do have a dynamic cytoskeleton, we must now try to understand why they don't do something more interesting with it, and when I say \'interesting' I mean in my eukaryotic-centric view becoming larger, more morphologically complex, or multicellular. I absolutely do not mean to disparage the many very interesting things that bacteria do and have done in their evolutionary history. The cyanobacteria invented oxygenic photosynthesis for which I am very grateful, and in general bacteria have much more interesting twists on metabolism than do us chemically unimaginative eukaryotes. But I do realistically claim organismal size, morphological complexity, and true multicellularity as eukaryote-specific features that deserve explaining. As we delve into the details of my argument I will delineate a few of the many biological examples of well-understood systems that have convinced me that bacteria simply do not have cytoskeletal nucleators or cytoskeletal motor proteins as we understand them in eukaryotes. At present, I hope you'll bear with this assertion for just a bit, so that I can more fully explain my hypothesis. If you'll accept for the moment my premise that the real difference between bacterial cells and eukaryotic cells lies in the eukaryotic proliferation of cytoskeletal nucleators and molecular motor proteins, then a relevant question becomes, what kinds of cellular structures can you make if you have nucleators and motors versus the structures that you can make if you don't? The diagram in Figure [2](#F2){ref-type="fig"} shows - given some reasonable assumptions about the universality and fundamental nature of helical protein filament assembly - what larger-scale structures you can get with and without nucleators and motors. In particular these drawings show structures that can be formed by polarized cytoskeletal filaments, where the subunits assemble in a head-to-tail fashion so that the two ends of the filaments are structurally distinct. According to the basic theories of protein polymerization, this is expected to give a polymer where the kinetics of subunit addition and loss at the two ends are also distinct, where one end grows and shrinks more quickly than the other \[[@B51]\]. In microtubules, the fast-growing end is called the plus end and the slow-growing end is called the minus end. In actin filaments, the fast-growing end is called the barbed end and the slow-growing end is called the pointed end. (Incidentally, both the Arp2/3 complex and the γ-tubulin ring complex nucleate their cognate filaments from the slow-growing end.) {#F2} The simple structures that can be made from polarized filaments I will call type A structures. In the absence of nucleators you can obviously make a single filament of essentially any length and that single filament can have many protofilaments. A microtubule is a single filament with 13 protofilaments that can be arbitrarily long. A bacterial flagellum is also a single filament that happens to have 11 protofilaments, and flagella can also be very long - 10 microns long *in vivo*. Both of these structures self-assemble quite nicely from solutions of purified protein monomers; indeed these were the examples that have formed much of the basis of our understanding of the fundamental thermodynamics of protein polymerization \[[@B50]\]. So those kinds of structures you can make regardless of whether you are a bacterium or a eukaryote and regardless of the presence of nucleators or motors. The other kind of structure that is very easy to make is a mixed polarity bundle. In crowded solutions, such as in the cytoplasm of a living cell, colloidal rods will tend to align with one another simply because of entropy and excluded volume effects \[[@B57]\]. When the rods happen to be cytoskeletal filaments, they can easily form bundles either by interacting with one another laterally, or else by having cross-linking proteins that help pull them together. For the bacterial cytoskeleton, the clearest example of a mixed polarity bundle is the plasmid-segregating actin homolog ParM, which can assemble into mixed polarity bundles on its own \[[@B58]\]. It is also very likely that the FtsZ ring in bacterial cytokinesis is essentially a mixed polarity bundle, formed with the help of cross-linking proteins \[[@B59]\]. The kinds of structures for which I think, theoretically, you need to have either localized nucleation or motor activity, or both, the type B structures, are structures like asters, where many cytoskeletal filaments with the same polarity emanate from a single location, or parallel bundles of filaments, where all of the filaments are pointing in the same direction. If filaments form spontaneously and then come together through purely entropic effects, there is no intrinsic reason for them to assemble in a particular orientation. So if you want to have a parallel bundle, such as in a muscle sarcomere, you have to control the assembly or orientation of the filaments, for example by having them all nucleated from the same site. And of course a great example of all of these properties is the mitotic spindle, where you have parallel bundling and anti-parallel bundling of microtubules, and also their nucleation from particular sites at the spindle poles. My assertion, and I've really scoured the literature here, is that no type B structures - asters and parallel bundles and spindles - have been observed in the cytoplasm of bacteria (with one very interesting exception which is I think the exception that proves the rule - and I'll come back to that a bit later). There are plenty of examples of single polarized filaments in bacteria. There are plenty of examples of mixed polarity filament bundles in bacteria. But the type B structures are critical I think to making eukaryotes what we are today, by allowing the elaboration of the microtubule cytoskeleton to give complex organelle dynamics and fabulously flexible DNA segregation capacity, and elaboration of the actin cytoskeleton to give us the possibility of amoeboid motion and phagocytosis, which allow us to run around and eat all those pesky bacterial biofilms and tame endosymbionts. And then once we have those kinds of structures and mechanisms, we are able to overcome the diffusion barrier and the increase in size and complexity of eukaryotic cells follows naturally from that. That's the hypothesis. The supporting details can be discussed from three different perspectives. The first focuses on self-assembly dynamics, and the rules about the kinetics and thermodynamics of self-assembly that come from the intrinsic properties of proteins - can these really be different between bacteria and eukaryotes? The second perspective focuses on the nucleators - is it true that bacteria don't have them? And if not, why not? And then the third perspective is all about the motors - is it true that bacteria don't have them? And if not, why not? And beyond that, there are also other possible explanations besides the cytoskeletal hypothesis for why eukaryotes and bacteria are different; this is a fourth level, even more general and more speculative, but one that I think helps tie this whole story together. I think it would be good to know all four supporting arguments for your hypothesis. Can we start with number one? ================================================================================================================= The first thing to think about is the question of protein self-assembly, because classically, when we think about the cytoskeleton, we imagine lots of little subunits that are able to assemble in an oriented fashion, to make larger structures. The ability of proteins to form homo-oligomers is very prevalent and, in fact, I would say it is almost the default thing for proteins to be able to do. There have been some genome-wide studies showing, for example, that in *Escherichia coli*, if you look at the known protein oligomers (and of course there may be some we don't know), something like 80% of them are homo-oligomers, where proteins assemble with other copies of themselves \[[@B60]\]. Structural biologists have done a very nice job of breaking down the kinds of symmetries you can get in these homo-oligomers into different kinds of classifications. Really making a helix is just one particular phylogenetic group, if you will, of the kinds of structures that proteins can make by self-assembly. Now there are two really nice things about helices. One is that a helix enables you to make structures of variable length, while most other oligomer types make a closed structure with a defined size, such as a viral capsid. But a helix that grows by addition of subunits onto the end can in principle be tuned over a very wide size (or length) range. The second thing that's nice about the helix as a mode for protein self-assembly was pointed out originally by HR Crane in 1950 \[[@B61]\] and then followed up by Linus Pauling in 1953 \[[@B62]\]. They used protein structural arguments to explain that when you allow many copies of the same protein to aggregate together you can hardly help but make a helix (Figure [3](#F3){ref-type="fig"}a). If you allow a protein to self-assemble, a helix of some kind is going to be the default. It is actually going to take more effort, in an evolutionary sense, to try and make something that's not a helix. ![**Helical protein filaments formed by self-assembly. (a)** General scheme for protein self-assembly into helices. For any globular protein of arbitrary shape, as shown at the top, considered as interacting with a second copy of itself in all possible orientations, there will be some pair of surface patches that result in optimal binding energy. It is highly unlikely that those two interface patches will happen to reflect any specific geometrical symmetry. When many copies of the same subunit self-associate by binding to one another through these surface interactions, a one-start helix with a single protofilament is the default structure formed, as shown in the middle. At bottom, if weaker interactions can also form laterally between subunits, multi-start helices may be stabilized (adapted with permission from the Royal Society of Chemistry \[[@B62]\]). **(b)** Electron micrograph showing a single filament of sickle-cell hemoglobin (HbS) (reprinted by permission from Macmillan Publishers Ltd: *Nature***272:**506**--**510, copyright 1978 \[[@B63]\]).](1741-7007-11-119-3){#F3} And in fact, mutant hemoglobin makes helical fibers, doesn't it? ================================================================ Yes, hemoglobin is a terrific example. Hemoglobin, of course, has been selected through evolution to be extremely soluble, so that within a red blood cell you can have 300 mg/ml of this one protein, which is an outrageously high concentration. In sickle-cell disease, a single point mutation in hemoglobin changes one charged residue on the surface to a neutral residue \[[@B64]\], and now in this dense cellular bag of the erythrocyte, filled almost entirely with one protein, you have a condition where the oxygen-depleted form of hemoglobin is able to self-assemble into a spectacularly beautiful helical structure with 14 protofilaments that looks absolutely classically like a microtubule or some other cytoskeletal filament \[[@B63]\] (Figure [3](#F3){ref-type="fig"}b). Sickle-cell hemoglobin is, of course, a very famous example of many principles of protein structure and function, but in this particular case it clearly shows that when you take a very soluble protein and create a condition in which it is not quite soluble, a helix is what you get. That's the default. If any old protein will assemble into a helix, then what is special about the cytoskeletal proteins? There are several possible answers, but one that I find compelling is that the common feature of the universally conserved cytoskeletal proteins - the actin superfamily, the tubulin superfamily - is that both of them are nucleotide hydrolases. They use the energy of nucleotide hydrolysis to switch between at least two distinct conformations. One of those conformations has a lower energy barrier to forming a filament than the other one. What this means is that if you can couple nucleotide hydrolysis kinetics to the interactions that the protein can form when it is in a helix, you can use the energy of nucleotide hydrolysis to regulate stability \[[@B65]\]. You can have the filaments assemble when the subunits have the ATP or GTP bound, and then after hydrolysis takes place, the energy released by hydrolysis is stored in the lattice in such a way that now disassembly becomes favorable. And this means that within a cytoplasm, where you have a good supply of ATP and GTP, you could have constantly dynamic filaments without having to change the concentration of anything. So are you going to suggest that bacteria don't have the energy to regulate filament assembly? ============================================================================================== Absolutely not. Bacteria have a ton of energy; I don't know of any cases where ATP availability is limiting for any normal biological process. And in fact bacteria use the cycle of nucleotide hydrolysis to modulate the assembly of their cytoskeletal filaments quite nicely. This is not the difference between bacteria and eukaryotes. If you look at the dynamics of, for example, FtsZ, it turns over very fast, even in the cytokinetic ring. You can see a beautiful ring that persists stably for some minutes before cytokinesis and before the cells separate \[[@B66]\], and yet there are very convincing photobleaching studies showing that the filaments within that ring are continuously turning over just like the microtubules in a mitotic spindle, or the actin filaments in a lamellipodium. Indeed it has been shown that mutants in FtsZ that have slowed GTP hydrolysis kinetics also have a slower turnover rate inside the living cell \[[@B67]\]. ParM, which is the very well characterized actin homolog that is used to segregate plasmids in bacteria \[[@B31]\], even shows dynamic instability \[[@B54]\], which is one of the classic outcomes of the coupling of assembly to nucleotide hydrolysis for eukaryotic cytoskeletal filaments \[[@B65],[@B68]-[@B70]\]. I think it is very clear that those intrinsic, dynamic properties of the self-assembling filaments - the coupling to nucleotide hydrolysis, the rapid turnover, kinetic properties like dynamic instability - those things are universal in cellular cytoskeletons (Figure [4](#F4){ref-type="fig"}). That is not a problem for bacteria, and that is not the difference between bacteria and eukaryotes. ![**Dynamic instability of cytoskeletal filaments from eukaryotes and bacteria. (a)** Dynamic instability of eukaryotic microtubules. Left: direct observation using dark-field microscopy of a microtubule undergoing dynamic instability. Middle: graph showing position of plus ends (top) and minus ends (bottom) for two dynamically unstable microtubules, with repeated cycles of growth and shrinkage. Numbered points correspond to individual video frames as labeled on the left (reprinted by permission from Macmillan Publishers Ltd: *Nature***321:**605**--**607, copyright 1986 \[[@B69]\]). Right, schematic diagram showing the connection between nucleotide hydrolysis and dynamic instability (Copyright 2008 from *Molecular Biology of the Cell,* 5th edition by Alberts *et al*. Reproduced by permission of Garland Science/Taylor & Francis LLC \[[@B41]\]). **(b)** Dynamic instability of bacterial ParM filaments. Left: fluorescence time-lapse images of a single ParM filament over time. Blue arrowhead shows position of initial filament appearance; red arrowheads mark the most extreme positions of the two tips. Video frames are separated in time by 5 s; scale bar is 2 μm. Right, traces of filament length over time for six different ParM filaments, showing a phase of growth followed by catastrophic shrinking. (From Garner EC, Campbell CS, Mullins RD: Dynamic instability in a DNA-segregating prokaryotic actin homolog. *Science* 2004, **306:**1021**--**1025. Reprinted with permission from AAAS \[[@B54]\].)](1741-7007-11-119-4){#F4} Moving on to the second perspective for my argument, if helical protein self-assembly regulated by nucleotide hydrolysis is universal, then what can we say about the role of regulated nucleation of cytoskeletal filaments in determining the difference between bacterial and eukaryotic cell organizational strategies? Here I think we are digging into much richer soil. I briefly mentioned this earlier, but now I'd really like to emphasize the striking observation that both FtsZ (bacterial tubulin) and ParM (bacterial actin) nucleate like mad \[[@B53],[@B54]\]. As a cell, you would really have to put a lot of effort into not nucleating them. For ParM, the filaments undergo very rapid dynamic instability and shrink back to nothingness unless they are stabilized by encountering cognate segments of DNA bound by the correct protein partner, both of which are normally found on the plasmid that is using ParM for segregation \[[@B71]\]. This mechanism rather neatly ensures that ParM filaments forming in a cell will be stabilized to push the plasmids apart only when there are two copies of the plasmid present, one to stabilize each end of the normally unstable filament. For FtsZ, its major regulator is a destabilizing factor, MinC \[[@B72]\], which undergoes its own very fascinating form of spatial regulation, but the short version is that the FtsZ ring that initiates bacterial cell division can form only where MinC is not; that is, FtsZ nucleation is spontaneous, but filament stability is regulated. If we had much more time to talk, I'd also tell you the whole beautiful story about the spatial regulation of MinC \[[@B73]\]. In *E. coli*, MinC is carried around by MinD, which arguably is yet another spontaneously nucleating self-assembled polymer that doesn't happen to be homologous to any of the known eukaryotic cytoskeletal proteins, so it is not really part of my central story here, but I can't stop myself from mentioning it anyway, and its kinetic regulation is highly relevant. MinD self-assembles on the bacterial membrane, and the MinD filaments are then destabilized by another protein factor, MinE. The kinetic interaction between MinD assembly and MinE destabilization results in spectacular oscillatory positioning of the MinC inhibitor inside of cells \[[@B74]\] and self-propagating waves when reconstituted *in vitro*\[[@B75]\]. In brief, this impressively dynamic and very precise system that the bacterial cell uses to choose the site of division depends on the spontaneous nucleation of one filamentous structure (MinD) that is destabilized by a regulator (MinE). The biological purpose of MinD and MinE is to regulate the localization of MinC, which acts to destabilize the spontaneously nucleating tubulin homolog FtsZ. Over and over for bacterial cytoskeletal and cytoskeletal-like elements, we are seeing spontaneous nucleation followed by spatially localized stabilization or destabilization as the general organizing principle. Again the really surprising thing here is that, for the cases that we understand well, nucleation plays no obvious part in the spatial regulation of cytoskeletal assembly for bacteria; everything where we understand the molecular details of spatial regulation regards filament stabilization and destabilization. My examples here are the best-characterized systems that we know in bacteria. For most of the other examples of bacterial cytoskeletal filaments, too little is known about their dynamics to enable us to guess how the nucleation versus stabilization equation will play out. I think it will be very, very interesting in the next few years to see if this is really a universal, decisive difference between the eukaryotes and the bacteria, or just an intriguing feature of the first few well understood systems. As we've already discussed, there are several simple strategies for developing regulatable nucleators for cytoskeletal filaments, either through specialization of a copy of the gene encoding the structural subunit, or just by recruiting another protein that has multiple binding sites for the structural subunits. Honestly, I really think bacteria could do that if they wanted to. But so far we do not know of any bacterial proteins that are specifically dedicated to nucleation of bacterial cytoskeletal filaments. So if nucleation can evolve easily, the question, again, is why didn't it in bacteria? ====================================================================================== I think you could argue that once you commit to a certain kind of dynamic strategy for your cytoskeletal filaments, back in the ancient past - maybe 3 billion years ago, when the modern version of FtsZ first came into being - then it's not worth changing it. That's possible. But there may be something else that we're missing, that makes the domain-based choice of cellular organizational strategy more likely to be universal. In the fourth part of this argument, the wild speculation, I'll get to what I think that might be. But as far as the nucleators go, it's not so much that I think that bacteria can't have them, it's just that there's no positive evidence yet that they do. There are many cases where having localized nucleators has been shown to be sufficient to give you really very interesting kinds of self-organized systems. A famous example I really like comes from experiments on dropping centrosomes or beads covered with microtubule nucleators into little microfabricated wells - you can grow up asters of microtubules and these will push the bead or the centrosome into the center of that well \[[@B76]\] (Figure [5](#F5){ref-type="fig"}a). Each growing microtubule end pushes against the wall of the well, generating a few picoNewtons of force \[[@B77]\], and the forces are equally balanced when the nucleating bead is near the middle. Because the microtubules are dynamic, and specifically because they are undergoing dynamic instability and occasionally shrinking back to their origin, the system does not get stuck and the centering can be maintained. This mechanism of self-centering by having centrally nucleated microtubules nudging at walls appears to be the way that the fission yeast *Schizosaccharomyces pombe* maintains the mid-cell location of its nucleus \[[@B78]\]. In the example of the nucleating bead in the well, we can see that just by localizing nucleation, you can set up a coordinate system that will tell you within the microchamber or within the cell where you are and which direction is inside and which is outside. If you imagine some cargo attached to a molecular motor encountering this assembly at any point in the space, the cargo attached to a minus-end directed motor such as dynein will end up in the middle, and the cargo attached to a plus-end directed motor such as kinesin-1 will go to the periphery. Going from that to being able to make something like the mitotic spindle is a relatively straightforward couple of steps, adding a second nucleating center and a protein that preferentially cross-links overlapping antiparallel microtubules, but you can't do it at all if you don't have the nucleator. ![**Self-centering activity of dynamic microtubule arrays. (a)** Self-centering by nucleation and dynamic instability, for a microtubule-nucleating bead. Images are separated by 3 minutes, scale bar is 10 μm (From *Proc Natl Acad Sci U S A*\[[@B76]\]). **(b)** Self-centering by motors. Left: diagram of crosslinked motors reorienting microtubules. Right: fluorescence image of an aster formed in a microwell by this mechanism (reprinted by permission from Macmillan Publishers Ltd: *Nature***389:**305**--**308, copyright 1997 \[[@B79]\]).](1741-7007-11-119-5){#F5} Now this brings me to the exception I mentioned earlier where bacterial cytoskeletal proteins can actually form a type B structure, specifically a self-centering aster. This has been seen for at least two of the eukaryotic cytoskeletal homologs associated with independent DNA elements in bacteria, an actin homolog that is encoded by a plasmid \[[@B80]\] and a tubulin homolog that is encoded by a bacteriophage \[[@B81]\]. In both cases, it appears that the self-centering activity of the associated cytoskeletal filament structures is useful to promote replication or segregation of the associated DNA element. In these cases, the plasmid or bacteriophage DNA itself is acting as the nucleating center. Other filament-forming proteins encoded by plasmids in bacteria, such as ParA, appear to help regulate the positioning of their plasmids in much the same way, even though these are not obviously homologous to one of the eukaryotic cytoskeletal proteins \[[@B82]\]. So it is clear that the basic mechanics for self-centering by localizing nucleation of self-assembled filaments do work just fine with the bacterial cytoskeletal and cytoskeletal-like proteins. But, and I think this is an important distinction, these structures are self-centered in more than just one way; the oriented cytoskeletal filaments do not appear to serve as tracks to provide spatial information for other cellular elements. Unlike the microtubule asters that set up a global coordinate system used by molecular motors and membrane-enclosed organelles to generate large-scale organization in eukaryotes, the plasmid and bacteriophage systems seem to operate with every man for himself. That is, they spatially localize only the very DNA element that encodes them. This observation points out a really interesting and probably important difference between bacteria and eukaryotes that I think is fundamental. When people first started discovering all of these tubulin and actin homologs in bacteria, many of us were initially amazed at how many there seem to be, with each one apparently tuned for a single specific purpose. But maybe what we should really be amazed about is how few tubulins and actins seem to be present in eukaryotic cells. For the major filament-forming cytoskeletal subunits in eukaryotes, there may be multiple genes encoding them in any given organism, but the subunits are typically able to assemble together into a single all-purpose cytoskeleton that is used for an outrageous variety of biological processes. In eukaryotes, functional variety appears to be largely carried by the large numbers of different kinds of actin-binding and tubulin-binding proteins that are present \[[@B83],[@B84]\]. Frankly it is rather extraordinary that the same kind of microtubule structure can be used to make mitotic spindles and beating cilia. As far as I can tell, this kind of creative multi-purposing of cytoskeletal filaments just does not happen in bacteria, where the rule seems to be one filament for one function. With this in mind - the idea that eukaryotes have to deal with just one kind of actin filament and just one kind of microtubule, while bacteria juggle many kinds of each along with other cytoskeletal-like filaments such as MinD and ParA - let's move on now to discussing the molecular motor proteins. This is the second major group of cytoskeletal regulators, after the nucleating proteins, that I suspect might simply be missing in bacteria. Like regulated nucleators, cytoskeletal motor proteins can cooperate with their filaments to generate very large-scale structures. For example, clusters of motor proteins can generate very nice organized asters *in vitro*, much as the nucleating beads do, even if their associated filaments are stabilized and non-dynamic \[[@B79]\] (Figure [5](#F5){ref-type="fig"}b). The motors, because they move toward only one end of the polarized filament substrate, are essentially able to sort out a disorganized clump of mixed-polarity filaments into something nice and orderly with uniform polarity. So the cytoskeletal molecular motors, together with localized nucleators, can make the type B cytoskeletal structures that I am arguing are so important for eukaryotic cell organization. Obviously bacteria do have some kinds of molecular motors, if we define molecular motors very generally as just being engines that convert chemical energy into mechanical energy, which I think is a fair definition. You mean bacterial motors such as flagella and pili and so forth? ================================================================= Yes. And the bacterial flagellar motor is just spectacular. It is an extraordinarily energy-efficient and complicated and beautiful object \[[@B85]\]. In fact, it is so beautiful that in the United States, the anti-evolutionary creationists seized upon it as being something so fantastic that it could not possibly have evolved \[[@B86]\]. Happily there is actually very nice structural evidence that evolution of the flagellar rotor has indeed occurred \[[@B87]\]. There are other several kinds of biological motors that can convert chemical energy into mechanical energy, and it is convenient to classify all of the biological motors we know about into five classes, which are not really mutually exclusive. The rotary motors such as the flagellar rotor would be one. Linear stepper motors, like kinesin, myosin and dynein, would be another \[[@B88]\]. Assemby and disassembly motors - using the forces that you get from polymerization of and depolymerization of microtubules or actin - make up another class \[[@B70]\]. Or there can be pre-stressed springs that are built in such a way that they store mechanical energy that can be released all at once, as, for example, in the acrosomal reaction in the horseshoe crab sperm \[[@B89]\]. And then there are also extrusion nozzles, where a cell will squirt out very hygroscopic polysaccharide that can allow it to jet along. *Myxococcus xanthus* does that \[[@B90]\]. Bacteria have some examples of all of those classes of biological motors. And they have linear stepper motors that work on DNA, or work on RNA, as substrates. What they don't have, or at least what has not yet been found, is any linear stepper motors that work on the cytoskeletal filaments. There is nothing known that does linear stepping on FtsZ. There's nothing known that does linear stepping on MreB or ParM or any of the other actin homologs. Why should bacteria not have evolved linear stepper motors? =========================================================== Why should it be so difficult? Let's take a look at the eukaryotes and see where they got their motors from. Looking just at the linear stepper motors for microtubules and actin, there are three major classes \[[@B88]\]. There are the myosins for actin, and the kinesins and dynein for microtubules. It has been shown structurally - and this was a real surprise for me and I think for most people - that kinesin and myosin have very similar central folds around the region where they couple nucleotide hydrolysis to piston-like motion, and are almost certainly derived from a common ancestor \[[@B91],[@B92]\]. Dynein is definitely the odd man out. It is a very different kind of motor, related to a completely different class of ATPases. So I hope you'll forgive me, for purposes of my speculative argument here, if I leave dynein aside and focus just on myosin and kinesin, and where did they come from, and why don't bacteria have them? It has been speculated that there was some kind of motor precursor that was the common ancestor of myosin and kinesin \[[@B93]\]. I don't think that we can make any reasonable argument about which kind of cytoskeletal filament it was more likely to walk on. But the heart of both of those motors is the nucleotide switch that converts hydrolysis into a large-scale protein conformational change resulting in stepping movement. Other aspects of motor function, such as the binding to the filament, are quite different among different motors, and if you look even just within the families - the myosin family, the kinesin family - the way they couple that nucleotide switch to motion is actually very wildly, dramatically different among different individuals \[[@B94]\]. For example, most myosins walk toward the barbed end of the polarized actin filament, but one particular subfamily, myosin VI, walks in the opposite direction toward the pointed end \[[@B95],[@B96]\]. Focusing on the nucleotide switch at the heart of the motor, these cytoskeletal molecular motors are members of what is called the P-loop NTPase family. This includes lots and lots of different ATPases and GTPases that are found in all domains of life. There has been a heroic attempt made by Eugene Koonin and colleagues to classify all of these many very divergent proteins into a reasonable phylogenetic tree based on sequence and structural similarities \[[@B97]\]. Given that this is such a diverse protein family spanning essentially the whole history of cellular evolution, there is some uncertainty here, but one thing about their reconstructed phylogeny really leapt out at me. According to their analysis, there is a entire branch of the P-loop NTPases that is found only in eukaryotes, and not in bacteria or archaea. This branch includes not only myosin and kinesin, but also many other critical proteins that we associate with eukaryotic cellular complexity. These include the Rho GTPase superfamily, which act as master regulators for actin cytoskeletal assembly \[[@B98]\], the Rab GTPases that govern many aspects of membraneous organelle identity \[[@B99]\], the Arf GTPases that are also associated with membrane traffic \[[@B100]\], the Ran GTPase that governs the directionality of nuclear import and export \[[@B101]\], and the heterotrimeric G proteins that influence so many aspects of eukaryotic cell-to-cell signaling \[[@B102]\]. So, wow. This looks very much like the list of eukaryotic-specific cellular features that we started off with. It seems historically as if a branch of the P-loop NTPase family might have arisen in eukaryotes at some point when they had presumably already been evolutionarily separated from the bacteria and the archaea, and this novel protein family gave rise not just to the myosins and kinesins, but also to many of the regulatory and signaling proteins that we most closely associate with the eukaryotic way of life. Bacteria, of course, have very good signalling proteins, such as the large family of two-component signal transduction systems involving histidine kinases and response regulators \[[@B103]\]. But, bacteria just don't seem to have the GTPases that we associate with eukaryotic signaling and large-scale cellular organization, and (particularly in animals) with complicated kinds of multicellular life. Who knows why that happened - maybe it was just good luck, maybe the innovation that led to those branches of the P-loop NTPase superfamily is something that happened in eukaryotes so that they were able to seize advantage of it and then combine it with their other properties and develop the ability to make these very large and elaborate, well organized and polarized cytoskeletal structures that would enable them to do things like build a mitotic spindle. So you're arguing that there might have been a couple of relatively low-probability changes that helped eukaryotic development but weren't important enough for bacteria to be forced to evolve that way because they could survive without it? Bacteria already had a perfectly good strategy going without these kinds of systems. Arguably in many ways the prokaryotic side of the tree, the bacteria and archaea, are much more diverse and more successful than eukaryotes - certainly there are many more of them than there are of us. They are particularly good at diversifying their metabolisms. All of the really exciting inventions in biological chemistry, I would say, have been generated in the prokaryotic branches of the tree. Photosynthesis, for example, is simply an awesome idea, and it was cyanobacteria that came up with that. Eukaryotes never could come up with that whole crazy business about using a cubic manganese cluster to strip the electrons off of water \[[@B104]\]. The best that eukaryotes could do was to tame the cyanobacteria and get them to come and live inside and become chloroplasts. I think the bacterial strategy is terrific, it is just different from our eukaryotic strategy. Our strategy has much more to do with morphological diversification, including getting very large both as cells and as organisms, and developing hunting strategies of various different kinds. Could we come back from this prokaryotic chauvinism for a moment to the crucial differences between them and us? ================================================================================================================ OK, finally I'm going to bring this whole argument back full circle and say that really the crucial difference between them and us is the membrane-enclosed nucleus. I think this is probably both a consequence and a cause in a feedback loop mechanism of the diversification of cytoplasmic cytoskeletal structures that then gave rise to larger-scale morphological diversity in eukaryotes. This fourth part of my argument is now much more speculative than even the most speculative parts of what I have said before. Let us stipulate that it is observable that all cells are organized in some way. What is their central organizing principle? Where is the information that is used by various different components of the cell to know where they are in relationship to everyone else? Well, if you're a bacterium and your chromosome is in the cytoplasm, the chromosome is a spectacular source of spatial information. In most bacteria there are only one or a few chromosomes. They tend to be oriented in a very reproducible way as you go from one individual to the next \[[@B105],[@B106]\] and because of the coupled transcription and translation, the physical site where you have a bit of DNA is also connected to the physical site where you make the RNA and the physical site where you make the protein from that bit of information \[[@B107]\]. If it is important to a bacterial cell to be able to target something to a specific location, it already has all the information it could ever hope for about which location in the cytoplasm is which because it has a well-defined, oriented chromosome present there. Now, once you wrap that beautifully organized chromosome up in a nucleus, all of a sudden you've lost all that spatial information. It is a very difficult chicken-and-egg problem as to what came first. Was it the wrapping of the nucleus that caused the actin and tubulin cytoskeletons to expand their capacities, or was it the explosion of the capacity of the cytoskeleton that wrapped up the nucleus in membrane? I like to imagine that at some point the nucleus got sequestered away somehow by some sort of prototypical membrane, maybe like what we see now in *Gemmata*, and then the poor little cytoskeletal elements were left out there in the cytoplasm on their own. They had no way of knowing where they were or of measuring space or position. So they had to figure out how to do it by themselves, without the chromosome there to help. Our eukaryotic cytoskeletons figured out how to do this by setting up large-scale arrays that can be oriented by virtue of having nucleators and molecular motor proteins to make those type B structures that are so useful for spatial organization over vast distances of many tens of micrometers. I think that this is a very elegant solution. The other benefit that the eukaryotes may have gotten from this strategic decision is extra morphological evolvability. In one of your other interviews, Marc Kirschner made some very interesting points about how certain kinds of preexisting conditions may make it relatively easy for some animal lineages to generate highly variable morphology \[[@B108]\]. I think the eukaryotic cytoskeleton may well be an example of this at the cellular level, an idea that Marc also certainly shares \[[@B109]\]. Once the lonely but inventive eukaryotic cytoskeletal proteins committed to the strategy of using a very small number of filament types to perform a large number of different functions, the addition of a new kind of organizational function to the underlying cytoskeletal framework may have been as simple as coming up with a few new modulators of cytoskeletal filament dynamics, or another kind of slightly modified motor protein. This diversification may have happened very quickly on an evolutionary scale. Sequence analysis of the myosin and kinesin motor families seems to suggest that the most recent common ancestor for all the currently living eukaryotes already had several different kinds of each motor \[[@B110],[@B111]\]. Indeed this most recent common ancestor may even have been capable of both amoeboid crawling motion and flagellar swimming \[[@B112]\]. It may be that the bacteria just never had to face this particular problem because, again, almost universally they have kept their chromosome right there in the cytoplasmic compartment where they could use it for spatial information. So typically, when a particular bacterium needs to make a filamentous structure for a novel purpose, such as orienting the magnetosomes in *Magnetospirillum*\[[@B5]\], it duplicates the gene for a cytoskeletal filament and adapts it for that one new purpose. This works fine for the purpose at hand, but forgoes the opportunity for flexibility and truly large-scale cellular organization that are intrinsic features of both the eukaryotic actin and microtubule cytoskeletons. Does that take us back to what the original eukaryotic cell might have looked like? =================================================================================== We're certainly never going to know what the original eukaryote looked like. One major reason we're never going to know is that all existing eukaryotes are very similar in many ways that must have come much, much later than that original separation of the eukaryotic lineage from the bacterial and archaeal lineages, suggesting that our most recent eukaryotic common ancestor was already quite a bit different from the original eukaryote and probably much more morphologically complex. So many of the most deeply rooted eukaryotic branches are just gone from the earth now, and we're never going to see them. Knowing eukaryotes, I would guess that the ones that figured out how to do phagocytosis first just ate everybody else. At some point initially, the earliest eukaryote must have looked much like its contemporary bacterial and archaeal counterparts, but it had secrets inside it that enabled it to become different. I think the fact that you see that both the diversification of the important NTPase families and the elaboration of cytoskeletal functions seem to be universal among eukaryotes means that probably those things happened relatively quickly. So when the lineage branched off, and maybe somehow the DNA got trapped in a nucleus and/or somehow membranes started being messed around with, that then generated a positive feedback loop that pretty quickly in evolutionary time caused it to turn into something with internal membrane-enclosed organelles and a mitotic spindle, and everything else we associate with eukaryotes came downstream of that. So I suspect the original eukaryote was small. I suspect it was pretty simple-looking compared with *Stentor* or one of the really fabulous single-celled eukaryotes. Also possibly simpler than the most complicated bacterium? ========================================================== Certainly simpler than the most complicated bacterium. But with potential.

Introduction {#sec1-1} ============ The word dolichoectasia is derived from Greek words "dolicos" and "ectasis" which means abnormally long and dilate, respectively.\[[@ref1]\] Many terms have been used to describe this arteriopathy (e.g., mega artery, mega dolichol artery, fusiform aneurysm, cirsoid aneurysm, and serpentine aneurysm). The prevalence of intracranial dolichoectasia was reported 0.1%--6.5% in the general population\[[@ref2]\] and about 12% among stroke patients.\[[@ref3]\] The risk factors associated with dolichoectasia are old age, hypertension, and male gender.\[[@ref4][@ref5]\] The hereditary conditions including Marfan syndrome, Ehlers--Danlos syndrome Type IV, pseudoxanthoma elasticum, Fabry disease, Pompe diseases, neurofibromatosis Type I, tuberous sclerosis, moyamoya diseases, autosomal dominant polycystic kidney diseases, fibromuscular dysplasia, and acquired immune deficiency syndrome.\[[@ref6][@ref7]\] The dolichoectasia can cause a transient ischemic attack, ischemic or hemorrhagic stroke, and compressive symptoms of the surrounding structure.\[[@ref5][@ref8]\] An aortic aneurysm, saccular aneurysm, and coronary artery disease\[[@ref9]\] may coexist with intracranial dolichoectasia. The dolichoectatic artery lack definite neck and this make it difficult to treat surgically. Till date, little is known about its natural course and prognosis. In this article, we have been tried to present the radiological course, treatment options, and outcome of one dolichoectatic patient with a review of the literature. Case Report {#sec1-2} =========== A 40-year-old male patient came to our hospital in 2014 with multiple intracranial artery dolichoectasia. It was diagnosed as an incidental finding on radiological examination of the patient. The patient had a history of coiling of the left cavernous internal carotid artery (ICA) aneurysm, which was done outside of our hospital. The patient had a history of systemic hypertension, and he was a chronic smoker. The computed tomography (CT) angiography findings in the year 2014 were, dilated and tortuous right supraclinoid ICA, fusiform dilated M1 segment of right and left middle cerebral artery (MCA) \[[Figure 1a](#F1){ref-type="fig"}\], fusiform dilated segment of the right posterior communicating artery (PCOM), P2 segment of right posterior cerebral artery (PCA) and left vertebral artery \[[Figure 2a](#F2){ref-type="fig"}\]. On the left vertebral artery, two fusiform aneurysm was present, out of which one was located at the level of posterior inferior cerebellar artery origin and the second one just distal to its origin \[[Figure 2a](#F2){ref-type="fig"}\]. The largest size of fusiform aneurysm segment was present in the right M1 MCA segment (size -- 7.3 mm), so clipping of an aneurysm was done to prevent future risk of rupture on the year 2014. Then, on the next stage procedure, trapping of left vertebral artery fusiform aneurysm segment was done, as its morphology was not suitable for the endovascular procedure \[[Figure 2b](#F2){ref-type="fig"}\]. The patient was advised to stop smoking and regular antihypertensive medicine. {#F1} {#F2} On follow-up CT angiography of the patient in the year 2015, a new fusiform aneurysmal dilated segment was observed on the distal A1 segment of right anterior cerebral artery (ACA) \[[Figure 1b](#F1){ref-type="fig"}\]. The maximum diameter of the fusiform segment of the right ICA, M1 segment of the right MCA, PCOM, and P2 segment of Posterior cerebral artery (PCA) had been increased from previous follow-up image \[[Table 1](#T1){ref-type="table"}\]. In the year 2016, the left cavernous ICA aneurysm refilling was observed, so again repacking of the coil was done. On CT head, no infarction and hemorrhage were observed and patient advice to continue follow-up. ###### Measurement of right intracranial artery Right intracranial artery 2014 2015 2017 2018 --------------------------- ------ ------ ------ ------ ICA (mm) 5.7 6 6.6 7.5 M1 MCA (mm) 7.3 10 11.3 8.5 A1 ACA (mm) 2.6 5.4 5.4 5.7 PCOM (mm) 3.5 6.7 7 7.2 P1 PCA (mm) 0.7 1.7 2.1 2.2 P2 PCA (mm) 2.6 4.3 4.5 5.3 BA (mm) 2.2 2.3 2.3 2.4 ICA -- Internal carotid artery; MCA -- Middle cerebral artery; ACA -- Anterior cerebral artery; PCOM -- Posterior communicating artery; BA -- Basilar artery; PCA -- Posterior cerebral artery In the year 2017, magnetic resonance imaging (MRI) angiography was done, which showed further increase in size of the fusiform dilated segment of right M1 MCA and right PCOM \[[Figure 1c](#F1){ref-type="fig"}\] as compared to previous cerebral angiography in the year 2015 \[[Table 1](#T1){ref-type="table"}\]. The contrast enhancement was noted in the wall of fusiform aneurysm segment of the right A1 ACA on MRI examination, which may be due to increased dilated vasa vasorum on the wall of an aneurysm \[[Figure 3](#F3){ref-type="fig"}\]. {#F3} On the next follow-up CT angiography in the year 2018, size of the fusiform segment of right ICA, PCOM and A1 ACA was further increased slightly from previous cerebral angiography in the year 2017 \[[Table 1](#T1){ref-type="table"}, Figures [1d](#F1){ref-type="fig"} and [2c](#F2){ref-type="fig"}\]. However, the right MCA fusiform aneurysmal segment size was decreased slightly on CT angiography, which may be due to thrombosis in the aneurysmal sac \[[Figure 1d](#F1){ref-type="fig"}\]. On sequential follow-up cerebral angiography, size of the fusiform segment of left M1 MCA was increased from 3.3 mm in the year 2014 to 5.4 mm in the year 2018 \[[Table 2](#T2){ref-type="table"}\]. The computational fluid dynamic (CFD) study was done in the year 2018 on left M1 MCA fusiform aneurysm segment, which showed wall pressure high and wall shear stress low \[Figure [4a](#F4){ref-type="fig"} and [b](#F4){ref-type="fig"}\]. The streamline was showing slow flow on the dome \[[Figure 4c](#F4){ref-type="fig"}\], and the vector was convergent in proximal fusiform aneurysm segment of left M1 MCA \[[Figure 4d](#F4){ref-type="fig"}\]. As CFD analysis was showing a risk of impending rupture on fusiform aneurysm segment of left M1 MCA, so clipping of an aneurysm was done in the year 2018 with preserving parent artery \[[Figure 4f](#F4){ref-type="fig"}\]. Intraoperatively numerous dilated vasa vasorum was observed in the fusiform segment of left M1 MCA aneurysm \[[Figure 4e](#F4){ref-type="fig"}\]. Till date, no hemorrhage and infarction were observed. ###### Measurement of left intracranial artery Left intracranial artery 2014 2015 2017 2018 -------------------------- ------ ------ ------ ------ ICA (mm) 4.8 5.7 5.7 6 M1 MCA (mm) 3.3 3.6 4.7 5.4 A1 ACA (mm) 1.9 2.3 2.3 2.3 PCOM (mm) 1.3 1.3 2.4 2.4 P1 PCA (mm) 0.9 0.9 1 1 P2 PCA (mm) 1.5 2.4 2.4 2.5 ICA -- Internal carotid artery; MCA -- Middle cerebral artery; ACA -- Anterior cerebral artery; PCOM -- Posterior communicating artery; PCA -- Posterior cerebral artery {#F4} Discussion {#sec1-3} ========== Definition and diagnostic criteria {#sec2-1} ---------------------------------- Intracranial arterial dolichoectasia describes the presence of at least one ectatic or enlarged artery in the cerebral vasculature.\[[@ref10]\] The diffuse intracranial arterial dolichoectasia defined by the involvement of two or more cerebral vessel and had been suggested with poorer natural history.\[[@ref11]\] A fusiform aneurysm is a form of nonsaccular arterial dilatation for short segment of the wall as compared to dolichoectasia which involve a long segment of the vessel wall.\[[@ref12]\] The irregular arterial course in the supraclinoid segment of internal carotid artery, anterior cerebral artery, MCA and PCA is determined by a visual assessment based on compression of surrounding structure and tortuosity of the vessel when compared to the contralateral side.\[[@ref13]\] The smoker and colleagues \[[Table 3](#T3){ref-type="table"}\] recommended a cutoff of 4.5 mm diameter at the level of mid pons to define basilar artery ectasia.\[[@ref14]\] The Passero and Rossi have suggested cutoffs diameter for internal carotid (≥7 mm), MCA (≥4 mm), and vertebral artery (≥4 mm) to indicate ectasia.\[[@ref5]\] ###### Diagnostic criteria and scores for basilar artery dolichoectasia based on computed tomography scan Basilar artery diameter at mid pons Laterality Height of bifurcation ------------------------------------- ---------------------------------------------------------- -------------------------------------------------------- 0-1.9--4.5 mm (normal range) 0 - Midline throughout 0 - At or below dorsum sellae 1 - \>4.5 mm (ectasia) 1 - Medial to lateral margin of clivus or dorsum sellae 1 - In suprasellar cistern 2 - Lateral to lateral margin of clivus or dorsum sellae 2 - At third ventricle floor 3 - At cerebellopontine angle 3 - Indentation and elevation of third ventricle floor Value of 1-deemed abnormal. Value of 2 or more suggests an abnormality Epidemiology {#sec2-2} ------------ The prevalence of intracranial arterial dolichoectasia in stroke-free patients ranges from 0.8% to 18.8%.\[[@ref4]\] The prevalence in series with stroke patients ranges from 3.1% to 17.1%.\[[@ref15][@ref16]\] The criteria used for the definition of dolichoectasia varied in the above series and diagnostic method used were CT angiography, MRI angiography, and autopsy findings. To date, no difference in prevalence has been found in a different race or ethnicity. The dolichoectasia is found to associated with hypertension, myocardial infarction,\[[@ref3]\] cerebral small vessel diseases,\[[@ref17]\] and smoking.\[[@ref5]\] The series have shown dolichoectasia to associated with polycystic kidney disease,\[[@ref18]\] moyamoya disease,\[[@ref6]\] ectrodactyly, ectodermal dysplasia and cleft lip-palate syndrome,\[[@ref19]\] Marfan syndrome,\[[@ref20]\] Ehlers--Danlos syndrome,\[[@ref21]\] Pompe disease,\[[@ref22]\] PHACES syndrome,\[[@ref23]\] tuberous sclerosis,\[[@ref24]\] cavernous angioma,\[[@ref25]\] craniocervical malformation,\[[@ref26]\] head trauma,\[[@ref27]\] pseudoxanthoma elasticum, Fabry disease, acquired immune deficiency syndrome.\[[@ref6]\] Pathophysiology {#sec2-3} --------------- In animal models acute increase in blood flow that causes dilatation of intracerebral artery has been shown to cause disruption of internal elastic lamina.\[[@ref28]\] The anatomical, hemodynamic and biological factors probably trigger dolichoectasia. Studies have shown that intracranial arterial diameter should be adjusted according to the circle of Willis, a reduction in a connection between anterior and posterior circulation lead to large increase in basilar artery diameter.\[[@ref29]\] Hemodynamics may have an important role in dolichoectasia, similar to distal abdominal ectasia, the basilar artery has an obtuse angle at apex mimicking bifurcation of the abdominal aorta, which is known to generate reflection wave, with maximum shear stress in distal abdominal aorta.\[[@ref30]\] Formation of CAs initiates in response to excessive hemodynamic stress to intracranial arterial wall of vascular bifurcation. The hemodynamic stress may lead to the endothelial dysfunction/injury, infiltration of inflammatory cells, phenotypic modulation and degeneration of smooth muscles, remodeling of extracellular matrix, and subsequent cell death and vessel wall degeneration. In this process, hemodynamic stress, an inflammatory reaction by activated macrophages, and vascular smooth muscle cell death are presumably crucial for the formation of a cerebral aneurysm.\[[@ref31]\] The disruption of the internal elastic membrane is common in patients of dolichoectasia.\[[@ref32]\] Marked increase in matrix metalloproteinase enzyme \[[Figure 5](#F5){ref-type="fig"}\] has been associated with many arteriopathies such as an aortic aneurysm, vascular dementia, and cerebral microangiopathy. The association of dolichoectasia with an aortic aneurysm, multiple lacunar infarct, perivascular atrophy, and leukoaraiosis support probably common pathological process. The role of atherosclerosis in dolichoectasia is unclear. Recent experimental studies using animal models, several pro-inflammatory cascades seem to be activated during aneurysmal progression including NF-κB, tumor necrosis factor-α, prostaglandin, myeloperoxidase, and reactive oxygen species.\[[@ref31]\] The other trophic factor such sympathetic innervation may have a role, such as in posterior circulation has less sympathetic innervation make susceptible to deformation with increase blood flow.\[[@ref33]\] {#F5} Clinical course {#sec2-4} --------------- The dolichoectasia may be symptomatic or with subclinical infarction to have diagnosed incidentally in radiology.\[[@ref34]\] The patients may present with compressive symptoms or vascular event. The dolichoectasia of posterior circulation may present with cranial neuropathy; the eight nerve is primarily involved in posterior circulation dolichoectasia.\[[@ref35]\] The posterior circulation dolichoectasia may manifest as ophthalmoplegia, hemifacial spasm, nystagmus, facial palsy, dizziness, tinnitus, hearing loss, dysarthria, trigeminal neuralgia, and diplopia.\[[@ref36]\] The dolichoectasia of posterior and anterior circulation\[[@ref37]\] may present with hydrocephalus. The mechanism may be from direct compression of the ventricle to water hammer effect from pulsation of the dolichoectatic artery. The infrequent presentations of dolichoectasia may be central apnea, cerebellar ataxia, normal-pressure hydrocephalus, and compressive symptoms contralateral to the affected site due to deformation of the brain stem.\[[@ref38]\] The dolichoectasia of anterior circulation may present with seizure, visual field defect, retinal ischemia, Horner\'s syndrome, and pyramidal sign \[[Table 4](#T4){ref-type="table"}\].\[[@ref38]\] ###### Clinical presentation Asymptomatic Compressive symptoms Vascular events -------------- ---------------------- --------------------------- Cranial neuropathy Brain infarction Hydrocephalus Transient ischemic attack Cerebellar ataxia Hemorrhagic stroke Central apnea Subarachnoid hemorrhage Seizure Pyramidal signs etc. Visual defects The dolichoectasia may present with brain infarction, transient ischemic stroke, hemorrhagic stroke, and subarachnoid hemorrhage. A systemic review of 375 patients with vertebrobasilar dolichoectasia the 5 years risk of brain infarction (17.6%), brainstem compression (10.3%), transient ischemic attack (10.1%), hemorrhagic stroke (4.7%), and subarachnoid hemorrhage (2.3%).\[[@ref39]\] The etiologies of stroke are generally due to an artery to artery embolism, traction, and occlusion of a small perforating branch,\[[@ref40]\] *in situ* thrombosis less commonly due to arterial dissection\[[@ref41]\] and vasospasm following subarachnoid hemorrhage. The hemorrhage in dolichoectasia is quite rare with prevalence reported in different case series range from 0.0% to 6.6%. Treatment {#sec2-5} --------- At present, no specific treatment exists to prevent arterial dilatation or tortuosity. The antiplatelet is used to prevent cerebral ischemia. In a case series of 40 patients, good surgical outcome (Glasgow Outcome Scale scores 1--2) was observed in 78% of patients. The various surgical techniques have done including direct clipping, trapping with bypass, proximal occlusion, resection with re-anastomosis, transposition, aneurysmorrhaphy with thrombectomy, and wrapping. There was no surgical mortality in this series.\[[@ref38]\] The endovascular procedure like coils at the top and bottom of the widen artery has been proposed.\[[@ref42]\] A combination of balloon and stent or use of several overlapping stents and coiling has been proposed.\[[@ref43]\] The randomized controlled trial is lacking for management of dolichoectatic artery. Several studies continue on the development of the drug for the prevention of progression of an aneurysm is in primitive stage at present. The experiments support the use of hydroxymethylglutaryl-CoA reductase inhibitor (statins) presumably through their potent inhibitory action on NF-κB.\[[@ref44]\] Conclusion {#sec1-4} ========== Intracranial arterial dolichoectasia may be associated with certain cardiovascular risk factors and acquired collagen vascular disorders. The exact pathogenesis in progression and risk of rupture of the dolichoectatic segment is not yet confirmed. It can present as asymptomatic incidental finding to local compressive symptoms. In our study, even with the best possible surgical modality, the disease was progressed in size and shape. The randomized controlled trial will be needed in future regarding the best possible management of such intracranial arterial disease. The new modality of treatment such as drug therapy to stop its progression and rupture risk will be required. Declaration of patient consent {#sec2-6} ------------------------------ The authors certify that they have obtained all appropriate patient consent forms. In the form the patient(s) has/have given his/her/their consent for his/her/their images and other clinical information to be reported in the journal. The patients understand that their names and initials will not be published and due efforts will be made to conceal their identity, but anonymity cannot be guaranteed. Financial support and sponsorship {#sec2-7} --------------------------------- Nil. Conflicts of interest {#sec2-8} --------------------- There are no conflicts of interest.

1. Introduction {#sec1-molecules-21-01714} =============== For the last several decades, fused pyrimidine derivatives have become a significant attraction in the field of medicinal chemistry research. This is attributed to the fact that pyrimidine is the basic unit of DNA and RNA structure. This fact explains the wide range of pharmacological activities of pyrimidine derivatives. Pyrazolo\[3,4-*d*\]pyrimidine derivatives are a class of fused pyrimidines possessing significant biological activities \[[@B1-molecules-21-01714],[@B2-molecules-21-01714]\]. They act as purine analogs \[[@B3-molecules-21-01714]\] and many of their derivatives act with antimicrobial \[[@B2-molecules-21-01714],[@B4-molecules-21-01714],[@B5-molecules-21-01714]\], antiviral \[[@B1-molecules-21-01714],[@B6-molecules-21-01714]\], antimetabolites \[[@B7-molecules-21-01714]\], anticancer \[[@B8-molecules-21-01714],[@B9-molecules-21-01714]\], anti-inflammatory \[[@B10-molecules-21-01714],[@B11-molecules-21-01714],[@B12-molecules-21-01714]\], and xanthine oxidase inhibitor activities \[[@B13-molecules-21-01714],[@B14-molecules-21-01714]\]. Furthermore, pyrimidopyridazine derivatives have a significant interest owing to the fact that they have a potent pharmacological effect as therapeutic agents \[[@B15-molecules-21-01714],[@B16-molecules-21-01714],[@B17-molecules-21-01714]\]. They have monoamine oxidase (MAO) inhibitory effect and subsequent modification on the diazine ring results in different inhibitory activities \[[@B18-molecules-21-01714]\]. MAO inhibitory drugs play an important role in clinical management of depression, as well as Alzheimer's disease \[[@B19-molecules-21-01714]\]. It has been established that cancer is spread worldwide and responsible for about 15% of all deaths \[[@B20-molecules-21-01714]\]. Many drugs with anticancer and antiviral activities have been developed \[[@B21-molecules-21-01714]\], such as zidovudine (AZT) \[[@B22-molecules-21-01714]\], zalcitabine (DDC) \[[@B23-molecules-21-01714]\], brivudine (BVDU) \[[@B24-molecules-21-01714]\], and methotrexate (MTX) \[[@B25-molecules-21-01714]\]. 4-Deazatoxaflavin (1,6-dimethyl-1,5,6,7-tetrahydropyrimido\[4,5-*c*\]pyridazine-5,7-dione) binds to herring sperm DNA and inhibits growth of *Pseudomonas* 568 \[[@B21-molecules-21-01714],[@B26-molecules-21-01714]\]. On account of these facts, a new series of substituted pyrimidopyridazines and pyrazolopyrimidines have been synthesized starting from 6-hydrazinyluracil derivatives and their antimicrobial, as well as antitumor activity has been evaluated and reported. 2. Results and Discussion {#sec2-molecules-21-01714} ========================= 2.1. Chemistry {#sec2dot1-molecules-21-01714} -------------- Extending our work in the synthesis of non-nucleosidic compounds of fused uracils \[[@B27-molecules-21-01714],[@B28-molecules-21-01714],[@B29-molecules-21-01714]\], we tried to synthesize pyrimidopyridazines and pyrazolopyrimidines from 6-hydrazinyluracils. The regioselective alkylation of 6-chlorouracil **1** \[[@B30-molecules-21-01714]\] with methyl---and/or propyl iodide in dimethyl sulfoxide (DMSO) in the presence of K~2~CO~3~ as basic medium afforded about 60%--70% yield of alkylated uracils **2a**--**c** \[[@B31-molecules-21-01714],[@B32-molecules-21-01714],[@B33-molecules-21-01714],[@B34-molecules-21-01714]\]. The nucleophilic substitution of 6-chlorouracil **2a**--**c** using hydrazine hydrate afforded 6-hydrazinyluracils **3a**--**c** \[[@B34-molecules-21-01714],[@B35-molecules-21-01714]\]. Heating of **3a**--**c** under reflux with ninhydrin for 5--10 min in the presence of AcOH resulted the desired compounds **4a**--**c** in good yield, as shown in [Scheme 1](#molecules-21-01714-sch001){ref-type="scheme"}. The structure of which was confirmed on the basis of analytical and spectral data. Thus, the ^1^H-NMR (DMSO-*d*~6~) spectrum of compounds **4a**,**b** showed a singlet around δ 12.24--12.23 ppm exchangeable characteristic for NH, while in **4c** a triplet splitting signal at δ 4.36 ppm characteristic for -NCH~2~ of propyl group. Additionally, a characteristic signal of the phenyl group for compounds **4a**--**c** appears around δ 9.26--7.75 ppm and the characteristic signal disappeared at 5--6 ppm of CH (5) in compounds **4a**--**c**. ^13^C-NMR showed 14 signals for compound **4a** and 16 signals for **4b** characteristic for carbon atoms. While, refluxing of **3b**,**c** with isatin for 1--2 h in AcOH gave the open form **5a**,**b** as shown in [Scheme 1](#molecules-21-01714-sch001){ref-type="scheme"}. Compounds **5a**,**b** were proved by the ^1^H-NMR spectrum, which showed three singlets at δ 13.01, 11.35, 10.99 ppm characterized for three NH groups, a singlet at δ 5.61 characterized for CH(5) of compound **5a** and two singlets at δ 13.03, 11.37 ppm characteristic for two NH groups, a singlet at δ 5.74, characterized for CH(5) of compound **5b**. ^13^C-NMR showed 15 signals for compound **5a** and 18 signals for **5b** characteristic for carbon atoms. The mechanism formation of **4a**--**c** is shown in [Scheme 2](#molecules-21-01714-sch002){ref-type="scheme"}. On the other hand, the reaction of **3b**,**c** with even benzylidene malononitrile or benzylidene ethyl cyanoacetate derivative via Michael addition reaction by heating under reflux for 6--8 h in dimethylformamide (DMF) in the presence of triethylamine as basic medium furnished the same products of pyrazolopyrimidines **6a**--**f**, as shown in [Scheme 3](#molecules-21-01714-sch003){ref-type="scheme"} by the elimination of malononitrile and ethyl cyanoacetate moieties, respectively, as shown in [Scheme 4](#molecules-21-01714-sch004){ref-type="scheme"}. Compounds **6a**--**f** were confirmed on the basis of analytical and spectral data. The ^1^H-NMR spectrum showed a characteristic singlet at δ 10.95, 11.19 ppm for NH(5) of compounds **6a** and **6b**, respectively, a singlet around δ 7.95--7.50 ppm for NH(1) and characteristic signals for the phenyl group around δ 7.73--6.72 ppm for compounds **6a**--**f**. ^13^C-NMR for compounds **6b** and **6e** showed 12 and 17 signals characteristic for carbon atoms respectively. Heating of **3b**,**c** under reflux conditions for 4--5 h with benzil in dimethylformamide in the presence of triethylamine furnished **7a**,**b** in moderate yield, as shown in ([Scheme 3](#molecules-21-01714-sch003){ref-type="scheme"}). On the other hand, compound **7a** was also obtained via heating of **3b** with α-phenyl phenacyl bromide under reflux conditions for 5 h. Compounds **7a**,**b** were confirmed on the basis of analytical and spectral data. ^1^H-NMR spectra of **7a** showed a singlet at δ 11.72 characteristic for NH(6) and signals of phenyl groups for **7a**,**b** around δ 7.26--7.10 ppm. ^13^C-NMR showed 16 signals characteristic for carbon atoms of compound **7a**. Finally, reaction of **3b** with different phenacyl bromides such as phenacyl-, *p*-methoxyphenacyl-, and *p*-nitrophenacyl bromide by heating under reflux for 4--6 h in dimethylformamide in the presence of triethylamine afforded pyrimidopyridazines **8a**--**c** in moderate yields ([Scheme 3](#molecules-21-01714-sch003){ref-type="scheme"}). Compounds **8a**--**c** were identified on the basis of analytical and spectral data. The ^1^H-NMR spectra showed a singlet around δ 12.08--11.96 ppm characteristic for NH(6), a characteristic singlet aromatic proton at CH(4) around δ 8.69--8.43 ppm of pyridazine ring, and signals of phenyl groups around δ 8.69--7.00 ppm. ^13^C-NMR showed 14 signals for compound **8b** characteristic for carbon atoms. The expected mechanism for the reaction of 6-hydrazinyl uracil with benzylidene malononitrile and/or benzylidene ethyl cyanoacetate ([Scheme 4](#molecules-21-01714-sch004){ref-type="scheme"}). 2.2. Antimicrobial Screening {#sec2dot2-molecules-21-01714} ---------------------------- As shown in [Table 1](#molecules-21-01714-t001){ref-type="table"}, the newly-synthesized compounds tested displayed variable in vitro antibacterial and antifungal activities. From the screening results, it can be seen that compound **4b** showed the highest activity against Gram-positive bacteria *Bacillus subtilis* compared with the standard drug, followed by compounds **4c**, **5a**, **6b**, **6d**, **4a**, **5b**, **6c** and **6f**, respectively. Similarly, compound **4b** showed the highest activity against Gram-positive bacteria *Streptococcus pneumonia* in comparison to the standard drug, followed by compounds **5a**, **4c**, **6b**, **6d**, **6c**, **4a**, and **5b**, respectively. On the other hand, compound **4b** showed the highest activity against Gram-negative bacteria *Escherichia coli* compared with the standard drug, followed by compounds **4c**, **6b**, **5a**, **6d**, **4a**, and **6c**, respectively. However, the order of activity against *Pseudomona aeruginosa* was **4b**, followed by compounds **5a**, **4c**, **6b**, **6d**, **6c**, **4a**, and **5b**, respectively. Regarding the activity of the tested compounds against the tested filamentous fungus *Aspergillus fumigatus*, the order of activity being **4b**, **4c**, **6b**, **5a**, **6d**, **4a**, **6c**, **6f**, respectively. Compound **6f** showed a weak antimicrobial effect on Gram-positive bacteria *Bacillus subtilis* as well as the tested filamentous fungus *Aspergillus fumigatus*. No antimicrobial activities were detected for compounds **7a** and **7b**. None of the tested compounds exert any activity against the pathogenic yeast species (*Candida albicans*) under these screening conditions. The minimum inhibitory concentration of the six most active synthesized compounds were detected, as shown in [Table 2](#molecules-21-01714-t002){ref-type="table"}. It was shown that **4b** showed the highest potential where its minimum inhibitory concentration (MIC) was comparable with that of the standard compounds, whereas **4a** showed the lowest potential and a very high MICs in comparison to the standard. ### Anticancer Activity {#sec2dot2dot1-molecules-21-01714} The in vitro growth inhibitory activity of the synthesized compounds was investigated in comparison with the well-known anticancer standard drug 5-flourouracil under the same conditions using colorimetric viability assay. Data generated were used to plot a dose response curve of which the concentration of test compounds required to kill 50% of cell population (IC~50~) was determined. The results revealed that all the tested compounds showed inhibitory activity to the tumor cell lines in a concentration dependent manner. Cytotoxic activity was expressed as the mean IC~50~ of three independent experiments. The results are represented in [Table 3](#molecules-21-01714-t003){ref-type="table"} and [Figure 1](#molecules-21-01714-f001){ref-type="fig"}a,b showed that compound **4a** was the most active against the breast carcinoma cell line (MCF-7), compared with the reference drug with IC~50~ values of 3.6 and 4.1 μg/mL, respectively. Interestingly, compounds **4a**, **4c**, and **8a** exhibited potent antitumor activity against breast cancer, respectively, and were the most active among their analogues. Moreover, the other compounds were less active. 3. Experimental Section {#sec3-molecules-21-01714} ======================= 3.1. General {#sec3dot1-molecules-21-01714} ------------ All melting points were determined with an Electrothermal Mel.-Temp. II (Registered trademark of Barnstead, Barnstead, NH, USA) apparatus and were uncorrected. Element analyses were performed at Regional Center for Mycology and Biotechnology at Al-Azhar University. The infrared (IR) spectra were recorded using a potassium bromide disc technique on a Nikolet IR 200 FT IR spectrometer (Thermo Electron Scientific Instruments LLC, Madison, WI, USA) and carried out in Taif University, Taif, KSA. Mass spectra were recorded on DI-50 unit of Shimadzu GC/MS-QP 5050A mass spectrometer (Shimadzu Corporation, Tokyo, Japan) at the Regional Center for Mycology and Biotechnology at Al-Azhar University. ^1^H-NMR and ^13^C-NMR spectra were recorded in DMSO-*d*~6~ as a solvent using a Varian Mercury spectrometer at 400 MHz and 125 MHz, respectively, Applied Nucleic Acid Research Center, Zagazig University, Egypt. Chemical shifts (δ) are given in ppm and coupling constants (*J*) are given in Hz. All reactions were monitored by TLC using pre-coated plastic sheet silica gel (0.25 mm, 20 × 20 cm, 60F~254~, E. Merck KGaA, Konstanz, Germany) and spots were visualized by irradiation with UV light (254 nm). The used solvent system was chloroform:methanol (9:1) and ethyl acetate:toluene (1:1). 6-Chlorouracil (**1**) was prepared according to the reported method \[[@B30-molecules-21-01714]\]. *6-Chloro-1-alkyl- and/or 1,3-Dialkyluracils* **2a**--**c** \[[@B31-molecules-21-01714],[@B32-molecules-21-01714],[@B33-molecules-21-01714],[@B34-molecules-21-01714]\] *6-Chloro-1-propyluracil* (**2b**) and *6-chloro-1,3-dipropyluracil* (**2c**): A solution of 6-chlorouracil (**1**) (40 mmol) in dimethyl sulfoxide (25 mL) was heated gently until 6-chlorouracil dissolved, and then potassium carbonate (20 mmol) was added with stirring. Propyl iodide (40 mmol) was added one time and the mixture was stirred at room temperature for 6 h. Water (40 mL) was added, and cooled in an ice box for several hours. The formed precipitate was collected by filtration, washed with water, dried in the oven at 80 °C, and crystallized from methanol to give 3.8 g of a white crystalline precipitate (51% yield) **2b** with m.p. = 165 °C. The mother liquor was evaporated in vacuo until dryness, then water (30 mL) was added, followed by extraction with chloroform (40 mL × 3). The chloroformic layer was evaporated and the obtained colorless crystals was dried in desiccator to give 2.3g (25%) of **2c** with m.p. = 58 °C; ^1^H-NMR (DMSO-*d*~6~) δ ppm: 6.03 (s, 1H, CH-5), 4.26 (t, 2H, NCH~2~), 3.89 (t, 2H, NCH~2~), 1.52--1.62 (m, 4H, 2CH~2~), 0.84--0.88 (m, 6H, 2CH~3~). *6-Hydrazinyl-1-methyl-, 1-Propyl- and/or 1,3-Dipropyluracils* (**3a**--**c**) \[[@B34-molecules-21-01714],[@B35-molecules-21-01714]\] **3a**: Yield 95%; m.p. 254 °C, lit. \[[@B34-molecules-21-01714]\] = 255 °C; **3b**: Yield 84%; m.p. 238--240 °C; **3c**: Yield 91%; m.p. 120 °C. *2,4-Disubstituted-1H-indeno\[2,1-c\]pyrimido\[5,4-e\]pyridazine-1,3,7(2H,4H)-triones* (**4a**--**c**) A mixture of 6-hydrazinyl-1-substituted and/or -1,3-disubstituteduracils (**3a**--**c**) (1.9 mmol) and ninhydrin (1.9 mmol) in acetic acid (5 mL) was heated under reflux for 5--10 min. The formed precipitate after cooling was filtered, washed with ethanol and crystallized from DMF/ethanol (1:3). *4-Methyl-1H-indeno\[2,1-c\]pyrimido\[5,4-e\]pyridazine-1,3,7(2H,4H)-trione* (**4a**): Yield: 48%; m.p. \>300 °C; IR (KBr) ν~max~ (cm^−1^): 3178 (NH), 3062 (CH arom), 2885 (CH aliph), 1687, 1631, 1597 (C=O), 1447 (C=C); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 12.24 (s, 1H, NH),9.20 (d, 1H, *J* = 7.6 Hz), 7.89--7.76 (m, 3H, arom), 3.66 (s, 3H, CH~3~); ^13^C-NMR (DMSO-*d*~6~): δ = 188.2, 160.9, 153.5, 151.2, 149.8, 140.8, 137.5, 136.5, 134.8, 133.9, 131.4, 129.2, 109.9, 29.9; MS: *m*/*z* (%) = M^+^, 280 (52), 251 (10), 182 (59), 155 (52), 154 (28), 153 (22), 138 (100), 126 (32), 111 (23), 99 (20), 76 (37). Anal. Calcd for C~14~H~8~N~4~O~3~: C, 60.00; H, 2.88; N, 19.99. Found: C, 60.16; H, 2.85; N, 20.14. *4-Propyl-1H-indeno\[2,1-c\]pyrimido\[5,4-e\]pyridazine-1,3,7(2H,4H)-trione* (**4b**): Yield: 51%; m.p. 281--283 °C; IR (KBr) ν~max~ (cm^−1^):3174 (NH), 3047 (CH arom), 2974, 2838 (CH aliph), 1720, 1692, 1555 (C=O), 1458 (C=C); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 12.23 (s, 1H, NH, exchangeable), 9.23 (d, 1H, *J* = 7.6 Hz), 7.90--7.75 (m, 3H, arom), 4.31 (t, 2H, *J* = 7.6 Hz, CH~2~), 1.77--1.71 (m, 2H, *J* = 7.6 Hz, CH~2~), 0.96 (t, 3H, 7.6 Hz, CH~3~). ^13^C-NMR (DMSO-*d*~6~): δ =188.1, 160.8, 153.5, 151.1, 149.6, 141.0, 137.6, 136.5, 134.7, 133.9, 131.4, 129.2, 110.0, 44.0, 20.4, 11.1; MS: *m*/*z* (%) = M^+^, 308 (100), 267 (94), 266 (80), 265 (36), 238 (50), 223 (77), 210 (32), 196 (48), 195 (34), 181 (49), 167 (36), 155 (33), 154 (37), 153 (15), 152 (24), 139 (49), 138 (39), 127 (36), 126 (53), 125 (46), 112 (24), 99 (28). Anal. Calcd for C~16~H~12~N~4~O~3~: C, 62.33; H, 3.92; N, 18.17. Found: C, 62.51; H, 3.95; N, 18.25. *2,4-Dipropyl-1H-indeno\[2,1-c\]pyrimido\[5,4-e\]pyridazine-1,3,7(2H,4H)-trione* (**4c**): Yield: 71%, m.p. 260--262 °C; IR (KBr) ν~max~ (cm^−1^): 3050 (CH arom), 2961, 2873 (CH aliph), 1710, 1667, 1566 (C=O), 1432 (C=C); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 9.26 (d, 1H, *J* = 7.6 Hz, arom), 8.11--7.61 (m, 3H, arom), 4.37 (t, 2H, *J* = 6.8 Hz, CH~2~), 3.95 (t, 2H, *J* = 6.8 Hz, CH~2~), 1.76--1.73 (m, 2H, CH~2~), 1.66--1.65 (m, 2H, CH~2~), 0.96 (t, 3H, *J* = 6.8 Hz, CH~3~), 0.93 (t, 3H, *J* = 6.8 Hz, CH~3~); MS: *m*/*z* (%) = M^+^, 350 (62), 323 (33), 290 (27), 268 (28), 262 (50), 240 (20), 238 (16), 236 (42), 223 (26), 209 (17), 196 (15), 195 (19), 192 (47), 180 (63), 177 (41), 169 (64), 154 (21), 140 (22), 138 (83), 123 (46), 112 (45), 99 (17), 180 (63), 98 (45), 97 (47), 94 (53), 74 (81), 73 (100); Anal. Calcd for C~19~H~18~N~4~O~3~: C, 65.13; H, 5.18; N, 15.99. Found: C, 65.45; H, 5.24; N, 16.17. *6-(2-(2-Oxoindolin-3-ylidene)hydrazinyl)-1-propyl- and/or 1,3-Dipropylpyrimidine-2,4(1H,3H)-diones* **5a**,**b** A mixture of 6-hydrazinyl-1-propyl- and/or 1,3-dipropyluracils (**3b**,**c**) (1.6 mmol) and isatin (1.6 mmol) in acetic acid (5 mL) was heated under reflux for 1--2 h. The formed precipitate after cooling was filtered, washed with ethanol, and crystallized from DMF/ethanol (1:3). *6-(2-(2-Oxoindolin-3-ylidene)hydrazinyl)-1-propylpyrimidine-2,4(1H,3H)-dione* (**5a**): Yield: 53%; m.p. \>300 °C; IR (KBr) ν~max~ (cm^−1^): 3195 (br., NH), 3095 (CH arom), 2956, 2815 (CH aliph), 1702, 1594, 1515 (C=O), 1458 (C=C); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 13.01 (s, 1H, NH), 11.35 (s, 1H, NH), 10.99 (s,1H, NH), 7.63 (d, 1H, *J* = 7.6 Hz, arom), 7.39--7.36 (m, 1H, arom), 7.12--7.09 (m, 1H, arom), 6.97 (d, 1H, *J* = 7.6 Hz, arom), 5.61 (s, 1H, CH-5), 3.81 (t, 2H, *J* = 7.6 Hz, CH~2~), 1.69--1.64 (m, 2H, *J* = 7.6 Hz, CH~2~), 0.94 (t, 3H, *J* = 7.6 Hz, CH~3~); ^13^C-NMR (DMSO-*d*~6~): δ= 163.5, 162.3, 162.3, 151.0, 141.9, 136.1, 131.4, 122.7, 120.7, 119.5, 111.3, 78.5, 42.8, 20.9, 10.7; MS: *m*/*z* (%) = M^+^, 313 (21), 285 (34), 253 (30), 243 (16), 226 (11), 213 (12), 200 (14), 158 (10), 147 (14), 145 (19), 132 (13), 118 (39), 117 (35), 104 (34), 103 (22), 101 (19), 90 (32), 77 (47), 76 (29), 68 (100); Anal. Calcd for C~15~H~15~N~5~O~3~: C, 57.50; H, 4.83; N, 22.35. Found: C, 57.78; H, 4.90; N, 22.52. *6-(2-(2-Oxoindolin-3-ylidene)hydrazinyl)-1,3-dipropylpyrimidine-2,4(1H,3H)-dione* (**5b**): Yield: 65%, m.p. 283--285 °C; IR (KBr) ν~max~ (cm^−1^):3137 (br., NH), 3084 (CH arom), 2966, 2877 (CH aliph), 1691, 1600, 1542 (C=O), 1458 (C=C); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 13.03 (s, 1H, NH), 11.37 (s, 1H, NH), 7.95--7.87 (m, 1H, arom), 7.65 (d, 1H, *J* = 6.8 Hz, arom), 7.40--7.36 (m, 1H, arom), 6.98 (d, 1H, *J* = 6.8 Hz, arom), 5.74 (s, 1H, CH-5), 4.25 (t, 2H, *J* = 7.4 Hz, NCH~2~), 3.87 (t, 2H, *J* = 7.4 Hz, NCH~2~), 1.76--1.69 (m, 2H, *J* = 7.4 Hz, CH~2~), 1.54--1.53 (m, 2H, *J* = 7.4 Hz, CH~2~), 0.96--0.89 (m, 6H, 2CH~3~); ^13^C-NMR (DMSO-*d*~6~): δ = 163.5, 161.2, 158.0, 149.6, 141.9, 136.2, 131.4, 122.7, 120.7, 119.4, 111.3, 78.1, 43.2, 41.8, 20.1, 19.6, 11.1, 10.7; MS: *m*/*z* (%) = M^+^, 355 (86), 338 (20), 327 (47), 313 (24), 285 (34), 243 (59), 227 (20), 226 (12), 213 (27), 200 (52), 187 (36), 186 (54), 172 (12), 166 (41), 161 (100), 158 (32), 153 (31), 148 (50), 147 (42), 146 (17), 145 (69), 125 (17), 118 (42), 117 (44), 111 (66); Anal. Calcd for C~18~H~21~N~5~O~3~: C, 60.83; H, 5.96; N, 19.71. Found: C, 61.07; H, 6.03; N, 19.94. *3-Substituted-7-propyl- and/or 5,7-Dipropyl-1H-pyrazolo\[3,4-d\]pyrimidine-4,6(5H,7H)-diones* **6a**--**f** Method A: A mixture of 6-hydrazinyl-1-propyl- and/or 1,3-dipropyluracils (**3b**,**c**) (1.7 mmol) and appropriate benzylidene malononitriles (1.7 mmol) in DMF (5 mL) in the presence of TEA (1mL) was heated under reflux for 6--8 h. The reaction mixture was evaporated under reduced pressure. The residue was treated with ethanol (10 mL), the formed precipitate was filtered, washed with ethanol, and crystallized from DMF/ethanol (2:1) to afford **6a**--**f**. Method B: A mixture of 6-hydrazinyl-1,3-dipropyluracil (**3c**) (1.7 mmol) and 4-chlorobenzylidene ethyl cyanoacetate (1.7 mmol) in DMF (5 mL) in the presence of TEA (1 mL) was heated under reflux for 8 h. The reaction mixture was evaporated under reduced pressure. The residue was treated with ethanol (10 mL), the formed precipitate was filtered, washed with ethanol, and crystallized from DMF/ethanol (2:1) to afford **6f**. *3-Phenyl-7-propyl-1H-pyrazolo\[3,4-d\]pyrimidine-4,6(5H,7H)-dione* (**6a**): method A: Yield: 72%; m.p. \>300 °C; IR (KBr) ν~max~ (cm^−1^): 3170 (br., NH), 3058 (CH arom), 2965 (CH aliph), 1679, 1595 (C=O), 1452 (C=C); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 10.95 (s, 1H, NH), 7.95 (s, 1H, NH), 7.41--7.25 (m, 5H, arom), 3.84 (t, 2H, *J* = 7.2 Hz, NCH~2~), 1.71--1.69 (m, 2H, *J* = 7.2 Hz, CH~2~), 0.99 (t, 3H, *J* = 7.2 Hz, CH~3~); MS: *m*/*z* (%) = M^+^, 270 (4), 231 (8), 184 (25), 176 (10), 165 (19), 139 (10), 130 (11), 111 (23), 109 (14), 107 (11), 98 (17), 96 (16), 95 (17), 83 (30), 81 (23), 71 (30), 69 (99), 67 (26), 55 (100), 44 (20), 43 (86), 41 (75); Anal. Calcd for C~14~H~14~N~4~O~2~: C, 62.21; H, 5.22; N, 20.73. Found: C, 62.48; H, 5.24; N, 21.04. *3-(4-Chlorophenyl)-7-propyl-1H-pyrazolo\[3,4-d\]pyrimidine-4,6(5H,7H)-dione* (**6b**): Method A: Yield: 69%; m.p. \>300 °C; IR (KBr) ν~max~ (cm^−1^): 3215 (br., NH), 3050 (CH arom), 2965, 2869(CH aliph), 1684, 1614 (C=O), 1435 (C=C), 810 (*p*-substituted); ^1^H-NMR (DMSO-*d*~6~) δ: 11.19 (s, 1H, NH), 7.82 (s, 1H, NH), 7.47 (d, 2H, *J* = 8.6 Hz, arom), 7.29 (d, 2H, *J* = 8.6 Hz, arom), 4.05 (t, 2H, *J* = 7.4 Hz, NCH~2~), 1.65--1.63 (m, 2H, *J* = 7.4 Hz, CH~2~), 0.91 (t, 3H, *J* = 7.4 Hz, CH~3~); ^13^C-NMR (DMSO-*d*~6~) δppm: 160.4, 154.6, 150.3, 135.9, 133.0, 129.4, 128.7, 115.3, 99.2, 42.7, 20.6, 11.1; MS: *m*/*z* (%) = 306 (M+2, 9), M^+^, 304 (25), 271 (30), 265 (42), 211 (20), 176 (27), 145 (21), 138 (21), 131 (38), 125 (34), 116 (34), 114 (23), 110 (100), 87 (64), 84 (86), 82 (34), 43 (84), 42 (48); Anal. Calcd for C~14~H~13~ClN~4~O~2~: C, 55.18; H, 4.30; N, 18.39. Found: C, 55.37; H, 4.36; N, 18.57. *3-Phenyl-5,7-dipropyl-1H-pyrazolo\[3,4-d\]pyrimidine-4,6(5H,7H)-dione* (**6c**): Method A: Yield: 77%; m.p. \>300 °C; IR (KBr) ν~max~ (cm^−1^): 3195 (br., NH), 3040 (CH arom), 2964 (CH aliph), 1605, 1598 (C=O), 1449 (C=C); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 7.95 (s, 1H, NH), 7.56--7.15 (m, 5H, arom), 3.89 (t, 2H, NCH~2~), 3.84 (t, 2H, *J* = 8.8 Hz, NCH~2~), 1.70--1.68 (m, 2H, CH~2~), 1.60--1.58 (m, 2H, CH~2~), 0.91--0.89 (m, 6H, 2CH~3~); MS: *m*/*z* (%) = M^+^, 312 (21), 282(18), 255 (20), 247 (14), 194 (31), 180 (16), 163 (46), 126 (17), 125 (60), 121 (33), 105 (22), 97 (77), 83 (42), 81 (19), 80 (38), 69 (58), 57 (31), 56 (100), 43 (71); Anal. Calcd for C~17~H~20~N~4~O~2~: C, 65.37; H, 6.45; N, 17.94. Found: C, 65.48; H, 6.53; N, 18.09. *3-(4-Bromophenyl)-5,7-dipropyl-1H-pyrazolo\[3,4-d\]pyrimidine-4,6(5H,7H)-dione* (**6d**): Method A: Yield: 78%; m.p. \>300 °C; IR (KBr) ν~max~ (cm^−1^): 3135 (NH), 3023 (CH arom), 2966, 2837 (CH aliph), 1678, 1676 (C=O), 1496 (C=C); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 7.84 (s, 1H, NH, exchangeable), 7.73--7.21 (m, 4H, arom), 4.0--3.64 (m, 4H, 2NCH~2~), 1.68--1.61 (m, 4H, 2CH~2~), 0.91--0.80 (m, 6H, 2CH~3~), MS: *m*/*z* (%) = M^+^ + 2, 393 (1), M^+^, 391 (3), 355 (10), 327 (19), 283 (13), 262 (17), 220 (17), 207 (14), 160 (15), 159 (13), 157 (16), 141 (13), 129 (12), 119 (16), 115 (12), 109 (34), 97 (46), 95 (34), 87 (17), 85 (24), 84 (70), 81 (100), 71 (33); Anal. Calcd for C~17~H~19~BrN~4~O~2~: C, 52.19; H, 4.89; N, 14.32. Found: C, 52.53; H, 4.91; N, 14.39. *3-(2-Hydroxyphenyl)-5,7-dipropyl-1H-pyrazolo\[3,4-d\]pyrimidine-4,6(5H,7H)-dione* (**6e**): Method A: Yield: 72%; m.p. \>300 °C; IR (KBr) ν~max~ (cm^−1^): 3435 (OH), 3180 (NH), 3055 (CH arom), 2965 (CH aliph), 1605, 1545 (C=O), 1485 (C=C), 750 (*o*-substituted phenyl); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 9.41 (s, 1H, OH), 7.59--6.72 (m, 5H, 1NH & 4H arom), 3.80--3.66 (m, 4H, 2NCH~2~), 1.67--1.65 (m, 2H, CH~2~), 1.52--1.50 (m, 2H, CH~2~), 1.16 (t, 3H, *J* = 7.6 Hz, CH~3~), 0.91 (t, 3H, *J* = 7.6 Hz, CH~3~); ^13^C-NMR (DMSO-*d*~6~): δ = 162.5, 158.0, 152.0, 151.9, 134.4, 125.3, 124.9, 118.6, 116.4, 115.9, 97.3, 43.1, 42.8, 20.2, 20.0, 11.1, 11.0; MS: *m*/*z* (%) = M^+^, 328 (57), 296 (23), 278 (48), 266 (15), 223 (14), 179 (14), 165 (15), 140 (10), 136 (14), 129 (16), 127 (17), 125 (20), 116 (31), 115 (25), 113 (20), 111 (31), 109 (23), 107 (27), 97 (33), 81 (19), 77 (40), 69 (77), 67 (48), 59 (20), 56 (94), 43 (100); Anal. Calcd for C~17~H~20~N~4~O~3~: C, 62.18; H, 6.14; N, 17.06. Found: C, 62.44; H, 6.21; N, 17.23. *3-(4-Chlorophenyl)-5,7-dipropyl-1H-pyrazolo\[3,4-d\]pyrimidine-4,6(5H,7H)-dione* (**6f**): Method A: Yield: 69%, method B: Yield: 61%; m.p. \>300 °C; IR (KBr) ν~max~ (cm^−1^ ): 3187 (NH), 3053 (CH arom), 2965, 2870 (CH aliph), 1685, 1613 (C=O), 1497 (C=C), 814 (*p*-substituted); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 7.95 (s, 1H, NH), 7.52 (d, 2H, *J* = 8.8 Hz, arom), 7.32 (d, 2H, *J* = 8.8 Hz, arom), 4.99--3.81 (m, 4H, 2NCH~2~), 1.72--1.55 (m, 4H, 2CH~2~), 0.88--0.86 (m, 6H, 2CH~3~); MS: *m*/*z* (%) = M^+^ + 2, 348 (3), M^+^, 346 (8), 341(21), 314(25), 311 (16), 280 (29), 269 (17), 266 (23), 247 (22), 245 (17), 238 (25), 226 (20), 207 (18), 206 (19), 203 (30), 184 (22), 154 (33), 146 (33), 145 (40), 127 (56), 125 (26), 123 (25), 119 (43), 89 (82), 82 (100), 73 (51), 67 (63), 66 (45), 40 (87); Anal. Calcd for C~17~H~19~ClN~4~O~2~: C, 58.87; H, 5.52; N, 16.15. Found: C, 59.05; H, 5.61; N, 16.23. *3,4-Diphenyl-8-propyl- and/or 6,8-Dipropylpyrimido\[4,5-c\]pyridazine-5,7(6H,8H)-diones* (**7a**,**b**) Method A: A mixture of 6-hydrazinyl-1-propyl and/or 1,3-dipropyluracils (**3b**,**c**) (1.6 mmol) and benzil (1.6 mmol) in DMF (5 mL) in the presence of TEA (1 mL) was heated under reflux for 4--5 h. The reaction mixture was evaporated under reduced pressure. The residue was treated with ethanol (10 mL), the formed precipitate was filtered, washed with ethanol, and crystallized from DMF/ethanol (2:1) to afford compounds **7a**,**b**. Method B: A mixture of 6-hydrazinyl-1-propyluracil (**3b**) (1.6 mmol) and α-phenylphenacyl bromide (1.6 mmol) in DMF (5 mL) in the presence of TEA (1 mL) was heated under reflux for 5 h. The reaction mixture was evaporated under reduced pressure. The residue was treated with ethanol (10 mL), the formed precipitate was filtered, washed with ethanol, and crystallized from DMF/ethanol (2:1) to afford **7a**. *3,4-Diphenyl-8-propylpyrimido\[4,5-c\]pyridazine-5,7(6H,8H)-dione* (**7a**): Yield: method A: 76%, method B: 68%; m.p. 226--228 °C; IR (KBr) ν~max~ (cm^−1^): 3159 (NH), 3003 (CH arom), 2966, 2835 (CH aliph), 1674, 1538 (C=O), 1496 (C=C); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 11.72 (s, 1H, NH), 7.25--7.10 (m, 10H, arom), 4.34 (t, 2H, *J* = 7.6 Hz, CH~2~), 1.82--1.77 (m, 2H, *J* = 7.6 Hz, CH~2~),0.99 (t, 3H, *J* = 7.6 Hz, CH~3~); ^13^C-NMR (DMSO-*d*~6~): δ = 159.8, 157.5, 151.7, 149.8, 139.3, 136.5, 134.6, 129.6, 128.9, 128.0, 127.6, 127.3, 111.7, 43.1, 20.5, 11.1; MS: *m*/*z* (%) = M^+^, 358 (16), 317 (10), 316 (46), 315 (100), 255 (10), 189 (9), 171 (9), 128 (4), 77 (5); Anal. Calcd for C~21~H~18~N~4~O~2~: C, 70.38; H, 5.06; N, 15.63. Found: C, 70.49; H, 5.10; N, 15.84. *3,4-Diphenyl-6,8-dipropylpyrimido\[4,5-c\]pyridazine-5,7(6H,8H)-dione* (**7b**): Yield: method A: 64%; m.p. 220--222 °C; IR (KBr) ν~max~ (cm^−1^): 3056 (CH arom), 2963, 2873 (CH aliph), 1718, 1670 (C=O), 1495 (C=C); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 7.26--7.10 (m, 10H, arom), 4.42 (t, 2H, *J* = 7.4 Hz, NCH~2~), 3.75 (t, 2H, *J* = 7.4 Hz, NCH~2~), 1.85--1.79 (m, 2H, *J* = 7.4 Hz, CH~2~), 1.53--1.47 (m, 2H, *J* = 7.4 Hz, CH~2~), 1.00 (t, 3H, *J* = 7.4 Hz, CH~3~), 0.83 (t, 3H, *J* = 7.4 Hz, CH~3~); MS: *m*/*z* (%) = M^+^, 400 (6), 383 (10), 369 (7), 267 (22), 223 (17), 135 (12), 133 (26), 127 (11), 125 (11), 112 (20), 110 (17), 101 (18), 95 (26), 90 (23), 86 (24), 83 (25), 80 (24), 76 (41), 72 (31), 70 (31), 69 (29), 59 (43), 55 (42), 44 (100), 42 (81), 40 (62); Anal. Calcd. for C~24~H~24~N~4~O~2~: C, 71.98; H, 6.04; N, 13.93. Found: C, 72.21; H, 6.12; N, 14.12. *3-Substituted-8-propylpyrimido\[4,5-c\]pyridazine-5,7(6H,8H)-diones* **8a**--**c** A mixture of 6-hydrazinyl-1-propyluracil (**3b**) (1.6 mmol) and appropriate phenacyl bromides (1.6 mmol) in DMF (5 mL) in the presence of TEA (1 mL) was heated under reflux for 4--6 h. The reaction mixture was evaporated under reduced pressure. The residue was treated with ethanol (10 mL), the formed precipitate was filtered, washed with ethanol, and crystallized from DMF/ethanol (2:1) to afford **8a**--**c**. *3-Phenyl-8-propylpyrimido\[4,5-c\]pyridazine-5,7(6H,8H)-dione* (**8a**): Yield: 56%; m.p. 262--264 °C; IR (KBr) ν~max~ (cm^−1^): 3178 (NH), 3039 (CH arom), 2965, 2866 (CH aliph), 1668, 1592 (C=O), 1496 (C=C); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 11.99 (s, 1H, NH), 8.50 (s, 1H, arom), 8.22--8.19 (m, 2H, arom), 7.57--7.46 (m, 3H, arom), 4.29 (t, 2H, *J* = 7.4 Hz, NCH~2~), 1.79--1.73 (m, 2H, *J* = 7.4 Hz, CH~2~), 0.97 (t, 3H, *J* = 7.4 Hz, CH~3~); MS: *m*/*z* (%) = M^+^, 282 (13), 254 (9), 241 (23), 240 (89), 239 (19), 197 (36), 77 (14), 44 (30), 40 (100); Anal. Calcd. for C~15~H~14~N~4~O~2~: C, 63.82; H, 5.00; N, 19.85. Found: C, 63.97; H, 5.08; N, 20.02. *3-(4-Methoxyphenyl)-8-propylpyrimido\[4,5-c\]pyridazine-5,7(6H,8H)-dione* (**8b**): Yield; 61%; m.p. 219--221 °C; IR (KBr) ν~max~ (cm^−1^): 3162 (NH), 3039 (CH arom), 2967, 2826 (CH aliph), 1670, 1600 (C=O), 1449 (C=C), 838 (*p*-substituted); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 11.96 (s, 1H, NH), 8.43 (s, 1H, arom), 8.17 (d, 1H, *J* = 7.0 Hz, arom), 7.46 (d, 1H, *J* = 8.4 Hz, arom), 7.10 (d, 1H, *J* = 7.0 Hz, arom), 7.01 (d, 1H, *J* = 8.4 Hz, arom), 4.28 (t, 2H, *J* = 7.4 Hz, NCH~2~), 3.84 (s, 3H, CH~3~), 1.78--1.72 (m, 2H, *J* = 7.4 Hz, CH~2~), 0.96 (t, 3H, *J* = 7.4 Hz, CH~3~); ^13^C-NMR (DMSO-*d*~6~): δ = 160.8, 160.1, 160.0, 150.9, 149.8, 131.1, 127.4, 126.3, 120.7, 114.5, 55.3, 42.8, 20.5, 11.1; MS: *m*/*z* (%) = M^+^, 312 (19), 271 (12), 270 (50), 269 (100), 255 (21), 239 (24), 135 (20), 40 (23); Anal. Calcd for C~16~H~16~N~4~O~3~: C, 61.53; H, 5.16; N, 17.94. Found: C, 61.71; H, 5.23; N, 18.13. *3-(4-Nitrophenyl)-8-propylpyrimido\[4,5-c\]pyridazine-5,7(6H,8H)-dione* (**8c**): Yield: 54%; m.p. 190--192 °C; IR (KBr) ν~max~ (cm^−1^): 3176 (NH), 3063 (CH arom), 2965, 2873 (CH aliph), 1684, 1595 (C=O), 1449 (C=C), 1510, 1332 (NO~2~), 851 (*p*-substituted); ^1^H-NMR (DMSO-*d*~6~) δ ppm: 12.08 (s, 1H, NH), 8.69 (s, 1H, arom), 8.52 (d, 2H, *J* = 8.8 Hz, arom), 8.38 (d, 2H, *J* = 8.8 Hz, arom), 4.31 (t, 2H, *J* = 7.4 Hz,NCH~2~), 1.77--1.74 (m, 2H, *J* = 7.4 Hz, CH~2~), 0.97 (t, 3H, *J* = 7.4 Hz, CH~3~); MS: *m*/*z* (%) = M^+^, 327 (13), 286 (44), 285 (100), 243 (14), 242 (90), 40 (24); Anal. Calcd for C~15~H~13~N~5~O~4~: C, 55.05; H, 4.00; N, 21.40. Found: C, 55.12; H, 4.09; N, 21.57. 3.2. Biological Evaluation {#sec3dot2-molecules-21-01714} -------------------------- ### 3.2.1. Antimicrobial Bioassay by Using the Agar Diffusion Cylinder Method \[[@B36-molecules-21-01714]\] {#sec3dot2dot1-molecules-21-01714} All microbial strains were provided from the culture collection of the Regional Center for Mycology and Biotechnology (RCMB), Al-Azhar University, Cairo, Egypt. The newly-synthesized target compounds were tested in vitro against different types of bacteria, *Streptococcus pneumoniae* and *Bacillus subtilis* as examples of Gram-positive bacteria, and *Pseudomonas aeruginosa* and *Escherichia coli* as examples of Gram-negative bacteria. Fungi, as well as bacteria, were used for testing the antifungal activity of the synthesized compounds. *Aspergillus fumigates* and *Candida albicans* were used as example of fungi and yeast, respectively. The stock solution of concentrations (1 mg/mL) of the synthesized compounds were used. The plates were incubated at 37°C for 24 h for bacteria and yeast, and for 48--72 h for fungi. Tetracycline was used as the standard antibacterial drug while amphotericin B was used as the standard antifungal drug. The diameters of the inhibition zones (mm) were measured and used as criterion for the antimicrobial activity. ### 3.2.2. Determination of the Minimum Inhibitory Concentration (MIC) {#sec3dot2dot2-molecules-21-01714} Serial dilutions of the promising compounds were subjected to MIC determination. The different concentrations of each compound were tested with the modified agar diffusion cylinder method as was described before. ### 3.2.3. Evaluation of the Antitumor Activity Using Viability Assay {#sec3dot2dot3-molecules-21-01714} All human anticancer cell lines were obtained from the American Type Culture Collection. The cells were grown on RPMI-1640 medium supplemented with 10% inactivated fetal calf serum and 50 µg/mL gentamycin. The cells were maintained at 37 °C in a humidified atmosphere with 5% CO~2~ and were subcultured two to three times a week. For antitumor assays, the tumor cell lines were suspended in medium at concentrations of 5 × 10^4^ cell/well in Corning^®^ 96-well tissue culture plates, then incubated for 24 h. The tested compounds were then added into 96-well plates (three replicates) to achieve eight concentrations for each compound. Six vehicle controls with media or 0.5% DMSO were run for each 96-well plate as a control. After incubating for 24 h, the numbers of viable cells were determined by staining the cells with crystal violet \[[@B37-molecules-21-01714],[@B38-molecules-21-01714]\], followed by cell lysing using 33% glacial acetic acid and read the absorbance at 590 nm using microplate reader (Sunrise, TECAN, Inc., Morrisville, NC, USA) after well mixing. The percentage of viability was calculated as \[1−(ODt/ODc)\] ×100%, where ODt is the mean optical density of wells treated with the tested sample and ODc is the mean optical density of untreated cells. The relation between surviving cells and drug concentration is plotted to obtain the survival curve of each tumor cell line after treatment with the specified compound. The 50% inhibitory concentration (IC~50~), the concentration required to cause toxic effects in 50% of intact cells, was estimated from graphic plots \[[@B37-molecules-21-01714]\]. 4. Conclusions {#sec4-molecules-21-01714} ============== The newly synthesized compounds of indeno\[2,1-*c*\]pyrimido\[5,4-*e*\]pyridazines, oxoindolinylidene hydrazinyl pyrimidines, pyrazolo\[3,4-*d*\]pyrimidines and pyrimido\[4,5-*c*\]pyridazines were prepared by a simple method. The novel compounds were screened for both antimicrobial and anticancer activities. Compound **4b** showed a very high MICs in comparison to the standard drug tetracycline. Compounds **4a**, **4c** and **8a** exhibited potent antitumor activity against breast cancer in comparison to the standard drug 5-flourouracil. The Author wishes to thanks Mahmoud Elaasser to carry out the biological activity of this work at Regional Center for Mycology and Biotechnology at Al-Azhar University, Cairo, Egypt. **Sample Availability:** Samples of the compounds are available upon request. Samar El-Kalyoubi formulated the research idea, conceived and prepared the manuscript; Samar El-Kalyoubi and Fatimah Agili performed the experiments and analyzed the data; Samar El-Kalyoubi wrote the paper. The authors have read and approved the final manuscript. The authors declare no conflict of interest. Figure, Schemes and Tables ========================== {#molecules-21-01714-sch001} {#molecules-21-01714-sch002} {#molecules-21-01714-sch003} {#molecules-21-01714-sch004} {#molecules-21-01714-f001} molecules-21-01714-t001_Table 1 ###### In vitro antimicrobial activity of the tested compounds by well diffusion agar assay expressed as inhibition zone diameter (mm) in the form of mean ± SD \*. Tested Compounds Gram-Positive Bacteria Gram-Negative Bacteria Fungi ------------------ ------------------------ ------------------------ ------------- ------------- ------------- ------------- **4a** 20.6 ± 0.63 18.3 ± 0.72 20.6 ± 1.2 15.2 ± 0.58 NA 18.3 ± 1.5 **4b** 25.3 ± 1.2 22.6 ± 0.72 28.3 ± 0.72 24.2 ± 0.58 NA 23.4 ± 1.5 **4c** 23.6 ± 0.63 21.1 ± 1.5 23.4 ± 1.2 19.2 ± 1.5 NA 21.5 ± 1.2 **5a** 21.3 ± 0.72 22.4 ± 0.63 21.3 ± 0.37 20.1 ± 0.63 NA 19.3 ± 0.63 **5b** 18.1 ± 0.63 17.3 ± 0.63 NA 17.3 ± 0.63 NA NA **6b** 22.3 ± 1.5 20.1 ± 0.58 22.4 ± 0.58 18.6 ± 1.2 NA 21.3 ± 1.2 **6c** 17.3 ± 0.63 19.2 ± 0.72 16.3 ± 0.46 17.3 ± 0.63 NA 17.3 ± 0.63 **6d** 20.9 ± 1.5 19.2 ± 1.2 21.3 ± 0.37 17.3 ± 0.63 NA 18.9 ± 1.2 **6f** 15.2 ± 0.63 NA NA NA NA 13.6 ± 0.63 **7a** NA NA NA NA NA NA **7b** NA NA NA NA NA NA Tetracycline 28.7 ± 0.5 26.4 ± 0.7 30.2 ± 0.6 27.4 ± 0.8 Amphotericin B \- \- \- \- 25.4 ± 0.63 23.7 ± 0.72 \* NA: No activity under the screening conditions; -: Not tested. molecules-21-01714-t002_Table 2 ###### The MIC of the synthesized compounds. Sample Tested Microorganisms 4a 4b 4c 5a 6d Standard ------------------------------------------ ------ ---------------- ------ ------ ------- ---------- Fungi Amphotericin B *Aspergillus fumigatus* (RCMB 02568) 7.81 1.95 3.9 3.9 3.9 1.95 *Candida albicans* (RCMB 05036) NA NA NA NA NA 0.98 Gram Positive Bacteria: *Ampicillin* *Streptococcuspneumonia* (RCMB 010010) 7.81 1.95 3.9 3.9 3.9 1.95 *Bacillis subtilis* (RCMB 010067) 3.9 0.98 0.98 1.95 3.9 0.49 Gram negative Bacteria: *Gentamicin* *Pseudomonas aeruginosa* (RCMB 010043-5) 62.5 0.98 3.9 3.9 15.63 0.98 *Escherichia coli* (RCMB 010052-6) 3.9 0.49 1.95 1.95 3.9 0.49 \* NA: No activity. molecules-21-01714-t003_Table 3 ###### The in vitro inhibitory activity of tested compounds against breast carcinoma cell line (MCF-7) expressed as IC~50~ values (μg/mL) ± standard deviation from three replicates. Tested Compounds IC~50~ Values (μg/mL) ±Standard Deviation ------------------ ----------------------- --------------------- **4a** 3.6 0.4 **4b** 47.6 2.8 **4c** 4.6 0.3 **5a** \>200 \>8 **5b** 95.1 2.6 **6a** 106.7 2.5 **6b** 42.2 1.9 **6c** 160.4 5.8 **6d** 189.9 7.6 **6e** 49.8 2.4 **6f** 34.8 3.2 **7a** 104.5 4.9 **7b** 68.1 1.7 **8a** 20.4 0.8 **8c** 86.1 1.7 5-Flurouracil 4.1 0.6

[^1]: Academic Editor: John O\'Donnell

GENOME ANNOUNCEMENT {#s0} =================== *Streptococcus pneumoniae* is a major human pathogen causing a diverse array of respiratory and invasive infections. Because antibiotic resistance has increased among strains of *S. pneumoniae*, lytic phages and endolysins are currently being reconsidered as alternatives to antibiotics. Despite the isolation of several pneumophages in the past ([@B1]), only three virulent pneumococcus phages are readily available through phage collections, Dp-1 (*Siphoviridae* family) as well as Cp-1 and SOCP (*Podoviridae* family) ([@B2][@B3][@B5]). Here, a new virulent pneumococcal phage was isolated from a swab sample collected at the Centre Hospitalier de l'Université Laval (Quebec City). Briefly, the swab sample was eluted, filtered (0.45 µm), and propagated on *S. pneumoniae* strain R6 using brain heart infusion (BHI) broth supplemented with 5 ng/mL choline chloride (BHI-fc) and incubated overnight at 37°C ([@B2]). Several phage plaques were obtained and based on the DNA restriction profile of their genome, only one distinct phage was isolated, purified, and designated MS1. Genomic DNA of phage MS1 was isolated using a Lambda maxi kit (Qiagen) from a lysate. Genome sequencing was performed on a 454 FLX instrument (IBIS, Université Laval). The genomic sequence was completed by primer walking and the sequencing of several PCR products. The 17,814 raw reads (total of 6,821,949 bases) were assembled into one contig using the GS De Novo Assembler (Roche), with an average coverage of 122-fold. Genomic termini were determined using PhageTerm ([@B6]). Phage MS1 has a circularly permuted genome and uses headful (*pac*-type) packaging system. The genome of the virulent pneumococcal phage MS1 has a low G+C content (42.3%), and is composed of 56,075 bp and 77 genes. No tRNA genes were found. Phages were washed with ammonium acetate (0.1 M, pH 7.5), stained with uranyl acetate (2%), and observed under an electron microscope. Observations revealed an icosahedral capsid of 67 ± 2 nm in diameter with a noncontractile tail of 179 ± 12 nm in length, indicating that phage MS1 belongs to the *Siphoviridae* family. Gene prediction and annotation were performed using customized RASTtk ([@B7]) workflow with GeneMark ([@B8]). Automated annotations were verified with Glimmer's predictions ([@B9]) and manually curated. MS1 genome is syntenic to *Streptococcus* Dp-1 phage ([@B2]), with an average nucleotide identity (ANIb) 73.3% on 62.3% of aligned nucleotides calculated using JSpeciesWS ([@B10]). Putative functions of gene products were assigned based on results found by BLASTP, HHpred, pfam (v29.0), and TMHMM (v2.0) searches ([@B11][@B12][@B14]). Promoters and terminators were predicted using BPROM and ARNold, respectively ([@B15], [@B16]). No known toxins or remnants of the lysogeny module were identified in the MS1 genome. Interestingly, phage MS1 also possesses queuosine biosynthesis genes, which have been found in a few other phage genomes ([@B2], [@B17][@B18][@B19]). Phage MS1 was deposited at the Félix d'Hérelle Reference Center for Bacterial Viruses (<http://www.phage.ulaval.ca>) under the number HER 532. Accession number(s). {#s1} -------------------- The complete genome sequence of phage MS1 is available in GenBank under the accession number [KY629621](https://www.ncbi.nlm.nih.gov/nuccore/KY629621). **Citation** Kot W, Sabri M, Gingras H, Ouellette M, Tremblay DM, Moineau S. 2017. Complete genome sequence of *Streptococcus pneumoniae* virulent phage MS1. Genome Announc 5:e00333-17. <https://doi.org/10.1128/genomeA.00333-17>. We acknowledge funding from the Natural Sciences and Engineering Research Council of Canada and the Canadian Institutes of Health Research. W.K. is a recipient of a grant from the Danish Research Council for Technology and Production (4093-00198). M.O. is the holder of a Canada Research Chair in Antimicrobial Resistance. S.M. holds a tier 1 Canada Research Chair in Bacteriophages.

Introduction {#Sec1} ============ Natural organisms not only have to fit their environment, but also have to adapt to changes in their environment which means that they have to be plastic. While plasticity occurs in many forms, here we focus on *neural plasticity* which we define as an organisms ability to use "experiences" to improve later decisions and behavior. Being able to solve a T-maze repetitively, remembering where food is located, avoiding places where predators have been spotted, and even learning another language are all cognitive abilities that require an organism to have neural plasticity. We will show how this neural plasticity can evolve in a computational model system. Neural plasticity allows natural organisms to learn due to reinforcement of their behavior^[@CR1]^. However, learning is tied to specific neural mechanisms- working memory (WM), short-term memory (STM) and long-term memory (LTM). While learning was initially perceived as a new "factor" in evolution^[@CR2]^, potentially even independent, it has since been well integrated into the *Modern Synthesis of Evolution* ^[@CR3]^. Evolution and learning can have a positive effect on each other^[@CR4],[@CR5]^, however, this is not necessarily always the case^[@CR6]^. This has several implications: Evolution begin with organisms that could not adapt during their lifetime, which means that they had no neural plasticity. The only feedback that the evolutionary process receives is differential birth and death. As a consequence, learning will only evolve if it can increase the number of viable offspring, and it can only do so if there is a signal that predictably indicates a fitness advantage^[@CR7]^. Organisms receive many signals from their environment which have to be filtered and interpreted. Irrelevant signals should be ignored while others require adaptive responses. This can be done through instincts or reflexes in cases where fixed responses are necessary. In other cases, information has to be stored and integrated in order to inform later decisions, which requires memory and learning. To distinguish between actions that lead to advantageous results and those that are disadvantageous organisms need positive or negative feedback. However, none of the signals organisms receive are inherently "good" or "bad"; even a signal as simple as food requires interpretation. The consumption of food has to trigger positive feedback within the organism in order to function as a reward. The machinery that triggers the feedback is an evolved mechanism and is often adaptive to the environment. If food were a global positive feedback signal, it would reinforce indiscriminate food consumption. Organisms would not be able to avoid food or store it for later, but instead eat constantly. Another important detail we have to consider is the difference between learning and memory. While memory is information about the past, learning is the process that takes a sensorial percept and, typically by reinforcement, retains that information for later use. Specifically, sensory information is stored in working memory (WM)^[@CR8],[@CR9]^. Imagine this as the flurry of action potentials that go through the brain defining its current state. Information that a living organism needs to store for a moment is believed to reside in STM^[@CR9],[@CR10]^, but how information transforms from WM to STM is not fully understood^[@CR8],[@CR10],[@CR11]^. Natural systems use their LTM if they want to keep information for longer. Presumably, information from STM becomes reinforced and thus forms LTM, this is sometimes referred to as consolidation^[@CR9],[@CR10],[@CR12]^. The reinforcement process takes time and therefore is less immediate than STM. In addition, memories can be episodic or semantic^[@CR9],[@CR10],[@CR13]^ and can later be retrieved to influence current decisions. While information in the working memory can be used to influence decisions, it does not change the cognitive substrate. However, long term potentiation (or other neural processes) use this information to change the neural substrate by presumably forming or modifying connections. In summary, if we want to model the evolution of learning in natural organisms properly we need to take the following statements seriously:evolution happens over generations while learning happens during the lifetime of an organismevolution is based on differential birth and death (selection) and learning evolved to increase the number of viable offspring and/or to avoid deathorganisms do not receive an objective "positive" or "negative" signal, but instead evolved mechanisms to sense and interpret the world so that they can tell what actions were positive and which ones were notmemory is information about the past and can be transientinformation in the WM does not change the cognitive machinery, while learning changes the substrate to retain information for longer, which turns transient into permanent information Machine Learning {#Sec2} ================ Computer science and engineering are typically not concerned with biological accuracy but more with scalability, speed, and required resources. Therefore, the field of machine learning is much more of a conglomerate of different methods which straddle the distinct concepts we laid out above. Machine learning includes methods such as data mining, clustering, classification, and evolutionary computation^[@CR14]^. Typically, these methods try to find a solution to a specific problem. If we provide an explicit reference or example class we refer to it as supervised learning since the answer is known and the fitness function quantifies the difference between the provided solution and the ones the machine generates. For unsupervised learning we provide a fitness function that measures how well a machine or algorithm performs without the need to know the solution in advance. Genetic algorithms (GAs), which are a form of evolutionary search, work in supervised or unsupervised contexts, whereas learning algorithms are typically supervised. A special class are learning to learn algorithms, which improve their learning ability while adapting to a problem^[@CR15]^ but do not necessarily apply evolutionary principles. Genetic algorithms clearly optimize from one generation to the other, while learning algorithms on the other hand could be understood as lifetime learning. *Q*-learning^[@CR16]^ optimizes a Markov Decision Processes by changing probabilities when a reward is applied. Typically, delayed rewards are a problem, that deep-*Q* learning and memory replay try to overcome^[@CR17]^. Artificial neural networks can be trained by using back propagation^[@CR18]--[@CR20]^, the Baum-Welch algorithm^[@CR21],[@CR22]^, or gradient decent^[@CR23],[@CR24]^ which happens episodically, but on an individual level and not to a population that experiences generations. Multiplicative weights algorithm strengthens or weakens connections in a neural network-based on the consensus of a pool of experts^[@CR25],[@CR26]^, again on an individual level. At the same time, memory and learning are often treated interchangeably. Recurrent artificial neural networks can store information in their recurrent nodes (or layer) without changing their weights, which would be analogous to WM. Similarly, the system we use, Markov Brains (MB), can form representations about their environment and store this information in hidden states (WM), again transiently without changing its computational structure^[@CR27]^ (For a general explanation of Markov Brains see <https://arxiv.org/abs/1709.05601> ^[@CR28]^). Changes to the weights of an ANN, or the probabilities of a Markov process, or the probabilities of a  POMDP^[@CR29]^ reflect better learning, since those changes are not transient and change all future computations executed by the system. We also find a wide range of evolvable neural network systems^[@CR30]--[@CR32]^ (among many others) which change from generation to generation and allow for memory to form by using recurrent connections. Alternatively, other forms of evolving systems interact and use additional forms of memory^[@CR33],[@CR34]^. In order to evolve and learn, other systems allow the topology and/or weights of the neural network to change during evolution while also allowing weight changes during their lifetime^[@CR34]--[@CR44]^. Presenting objective feedback to adapt these systems during their lifetime allowed their performance to improve. As a consequence, the machinery that interprets the environment to create feedback was of no concern, but, as stated above, natural organisms also need to evolve that machinery to learn. We think it is quite possible to change these systems to not rely only on external feedback. Instead, they themselves could create the feedback signal as part of their output. However, none of the systems mentioned above is an evolvable MB (for a comparison see Fig. [1A vs. B](#Fig1){ref-type="fig"}).Figure 1Comparison of two different approaches to feedback generation. Traditional methods have the environment evaluate the actions of a neural network (or similar system) and then generate feedback which is provided as an external signal, similar to adding another input (panel A). Our approach (panel B) integrates both the feedback generation and the feedback gates into the Markov Brain. This way, the world only provides consistent information about itself which can be evaluated so that feedback generation does not need to be offloaded to the environment, but becomes part of the entire machine. The entire connectivity as well as feedback gates are integrated as part of what needs to evolve. In our approach we use MBs^[@CR45]^, which are networks of deterministic and probabilistic logic gates, encoded in such a way that Darwinian evolution can easily improve them. MBs have been proven to be a useful tool to study animal behavior^[@CR46],[@CR47]^, neural correlates^[@CR27],[@CR48],[@CR49]^, evolutionary dynamics^[@CR50]^, decision making^[@CR51],[@CR52]^, and can even be used as a machine learning tool^[@CR53],[@CR54]^. One can think of these MBs as artificial neural networks (ANN)^[@CR55]^ with an arbitrary topology that uses Boolean logic instead of logistic functions. Through sensors these networks receive information about their environment as zeros or ones, perform computations and typically act upon their environment through their outputs. We commonly refer to MBs that are embodied and through that embodiment^[@CR56]^ interact with their environment as agents (others use the term animat which is synonymous). MBs use hidden states to store information about the past similar to recurrent nodes in an ANN. The state of these hidden nodes has to be actively maintained which makes the information volatile. The information in the hidden states can be used to perform computations and functions as memory. This form of memory resembles WM or STM more than LTM due its volatile nature. In the past, the entire structure of a MB would be encoded by the genome and would not change over the lifetime of the agent. Here we introduce what we call feedback gates, which allow MBs to use internal feedback to store information by changing their probabilistic logic gates (see Methods for a detailed description of feedback gates). Like other systems these updates do not change the topological structure of the node network but rather the probabilities within the gates; similar to how learning in ANN is achieved through weight changes. However, feedback is not an objective signal coming from the environment but must be generated as part of the evolved controller. The feedback gates only receive internally generated feedback to change their behavior. This linkage between inputs, evaluation of the environment to generate feedback, how feedback gates receive this information, and how everything controls the actions of the agent evolves over time (see Fig. [1B](#Fig1){ref-type="fig"}). Here we introduce feedback gates that allow MBs to change their internal structure which is akin to learning and forming long-term memories during the lifetime of an organism. The closest similarity can be found in *Q*-learning or deep-*Q* learning, which changes the probabilities within a Markov Decision process even if rewards are delayed^[@CR14]^ (for a detailed explanation of the feedback gate function see Materials and Methods as well as Supplementary Fig. [1](#MOESM1){ref-type="media"}). We will show that feedback gates function as expected and allow agents to evolve the ability to decipher sensor inputs so that they can autonomously learn to navigate a complex environment. Results {#Sec3} ======= Natural organisms have to learn many things over their lifetime including how to control their body. Here the environment used to evolve agents resembles this problem. Agents have to learn to use their body in order to navigate properly. The environment is a 2D lattice (64 × 64 tile wide) where a single tile is randomly selected as the goal an agent must reach. The lattice is surrounded by a wall so agents can not escape the boundary, and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{1}{7}$$\end{document}$ of the lattice is filled with additional walls to make navigation harder. From the goal the Dijkstra's path is computed so that each tile in the lattice can now indicate which of its neighboring tiles is the next closest to the goal. In cases where two neighbor tiles might have the same distance, one of these tiles is randomly selected as the next closest. For ease of illustration we can now say that a tile has an arrow pointing towards the tile that should be visited next to reach the goal in the shortest number of tiles. The agent, controlled by a MB, is randomly placed on a tile that is 32 tiles away from the goal and facing in a random direction (north, west, south, or east). Agents can see the arrow of the tile they are standing on. The direction indicated by the tile is relative to the that of the agent, so that a tile indicating north will only be perceived as a forward facing arrow if the agent also faces north. The agent has four binary sensors that are used to indicate which relative direction the agent should go to reach the goal. The agent can move over the lattice by either turning 90 degrees to the left or right, or by moving forward. So far, in order to navigate perfectly, the agent would simply need to move forward when seeing a forward facing arrow, or turn accordingly. Instead of allowing the agent to directly pick a movement, it can choose one of four intermediate options (A, B, C, or D) at any given update. At the birth of an agent, these four possible options are mapped to the four possible actions: move forward, turn left, turn right, do nothing. As a result, the complexity of the task increases when the agent has to learn which of the 24 possible option-to-action maps currently applies to navigate the environment properly. The agent is not given any direct feedback about its actions; a mechanism must evolve to discern the current mapping and which is rather difficult. In prior experiments^[@CR27],[@CR45]--[@CR51],[@CR53]^, MBs were made from deterministic or probabilistic logic gates that use a logic table to determine the output given a particular input. Deterministic gates have one possible output for each input, while probabilistic gates use a linear vector of probabilities to determine the likelihood for any of the possible outputs to occur. To enable agents to form LTM and learn during their lifetime we introduce a new type of gate: feedback gate. These gates are different from other probabilistic gates, in that they can change their probability distribution during their lifetime based on feedback (for a detailed description, see below). This allows for permanent changes which are akin to LTM. While MBs could already retain information by using hidden states, now they can also change "physically". MBs must evolve to integrate these new gates into their network of other gates and find a way to supply feedback appropriately. Feedback Gate Usage {#Sec4} ------------------- To test if the newly introduced feedback gates help evolution and increase performance, we compare three different evolutionary experimental conditions. Agents were evolved over 500,000 generations that could use only deterministic logic gates, only probabilistic logic gates, or all three types of gates--deterministic, probabilistic, and feedback gates to solve the task. When analyzing the line of descent (LOD; see materials and methods), we find a strong difference in performance across the three evolutionary conditions (see Supplementary Information Fig. [2](#MOESM1){ref-type="media"}). None of the 300 agents that were evolved using only probabilistic logic gates were capable of reaching the goal in any of the 24 mappings. The agents that were allowed to use only deterministic logic gates failed to reach the goal in 225 of the 300 experiments, but the remaining 75 agents never reached the goal more than five times. Agents allowed to use all gates including feedback gates only failed to reach the goal in 75 experiments and in the remaining 225 experiments they reached the goal on average 5 times; with the best performer reached the goal 9 times on average. A Wilcoxon rank sum test^[@CR57]^ comparing the final distributions of performances for each experimental condition showed that we can reject the hypothesis that they were drawn from the same distribution (with the lowest p-value smaller than 10^−20^). This shows that agents that were allowed to use feedback gates outperform all other conditions by far. It is not entirely surprising to us that agents using only probabilistic gates struggle in this task because agents using probabilistic gates generally evolve slower, which might explain the effect. However, to our surprise, we found a couple of agents who were only allowed to use deterministic gates evolved to solve the task at least a couple of times. In the group that could use all three types of gates, we found 3 agents that could reach the goal on average 5 times using only probabilistic gates, as opposed to the 9 times the top agent with feedback gates could reach the goal on average. This shows two things: the task can be solved using only the gate inputs (i.e. WM) and providing agents with feedback gates during evolution allows them to reach the goal more often. This is an important control because from a computational point of view there is no qualitative difference between WM and LTM as both methods allow for recall of the past. Populations allowed to use feedback gates quickly evolve the ability to reach the goal in any of the 24 possible environments. The variance of their performance supports the same idea, that agents do not become better by just performing well in one environment, but instead evolve the general ability to learn the mapping each environment presents (See Fig. [2](#Fig2){ref-type="fig"} panel B).Figure 2Performance over evolutionary time. Panel A solid line shows how often the goal was reached (*W*) by all 300 replicate experiments across all 24 possible environments for agents on the line of descent. The dotted line is the same average performance for the same agents when their feedback mechanism was disabled. The underlying gray-shade indicates the standard error. Panel B shows how often on average the 300 replicate agents on the line of descent could not reach the goal a single time in any of the 24 possible environments as a black line. In red is the variance in performance on the line of descent as an average over all 300 replicate experiments. Now that we have shown that agents with feedback gates are capable of evolving a solution to navigate in this environment, we have to ask if they actually utilize the feedback gates. For that, all agents on the LOD were tested again but their feedback gates were kept from changing their probability tables. Comparing these results with the agents performance when using the feedback gates regularly reveals that the agents rely heavily on their feedback gates (see Fig. [2](#Fig2){ref-type="fig"} panel A). This implies that the evolved agents indeed store information about the environment for longer periods using their feedback mechanism. If they would have used hidden states as the only means to save information for later, blocking the feedback signal would not have caused a loss of function. Therefore feedback gates indeed allow for the formation of LTM. Feedback gates change over time {#Sec5} ------------------------------- We find that the probability tables modified by feedback become specifically adapted to each of the 24 possible mappings the agents get tested in. See Fig. [3](#Fig3){ref-type="fig"} as an example of the best performing agent using only one feedback gate. Some rows in the probability tables converge to having a single high value that is specific to the environment the agent experienced (for more details see Supplementary Information Figs [3](#MOESM1){ref-type="media"}--[4](#MOESM1){ref-type="media"}). This shows, that indeed feedback gates become specifically adapted to the environment the agent experiences. It also indicates, that agents change their computational machinery according to their environment and do not rely solely on WM to perform their task.Figure 3Probability tables of a feedback gate after it learned each of the 24 possible mappings. Each of the 24 gray scale images corresponds to a feedback table adapted during the lifetime of an agent to a different mapping. The darker the color, the lower the probability, observe that rows have to sum to 1.0. Some rows, apparently those of input combinations that were never experienced, remain unadapted. Rows one, two, and seven of each table converge to a single high value surrounded by low probabilities. The change to the feedback gates' probability tables can be quantified by measuring the mutual information each table conveys at birth and after the agent completes the task. We find that the mutual information is generally lower at birth (\~0.25) and higher at the end of the task (\~0.8). This  signifies that the agents have more information about the environment at death than they did at birth, as expected. We then compute the difference between both measurements, ($\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{{\rm{\Delta }}}$$\end{document}$), which quantifies the change of mutual information over the lifetime of the agent. When discretizing these values for different levels of performance, we find a strong correlation (*r* = 0.922) between performance and increase in mutual information (see Fig. [4](#Fig4){ref-type="fig"}). This shows that agents who perform better increase information stored in their feedback gates over their lifetime.Figure 4Change in mutual information of feedback gates for different levels of performance. The change in performance ($\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{{\rm{\Delta }}}$$\end{document}$) is shown on the y-axis for different performances (W). The performance for all 300 final organisms using all types of gates was measured and put into ten bins, ranging from the lowest performance 2 to the highest performance of 12, with a bin size of 1. Error-bars indicate the variance. Differences between agents using feedback gates and those who do not {#Sec6} -------------------------------------------------------------------- We wanted to test if feedback gates improve the agents ability to learn. Those agents that evolved to use feedback gates generally perform very well in the environment, however, we also find other agents that only use deterministic gates and still have an acceptable performance. This either suggests that feedback gates are either not necessary or that they do not provide enough of an selective advantage to be used every time. It is also possible that there is more than one algorithmic solution to perform well in this navigation task. All of these points suggest that a more thorough analysis and comparison of the independently evolved agents is necessary. We find that agents that do not use feedback gates require a much greater number of logic gates than those who do (see Fig. [5](#Fig5){ref-type="fig"}). This seems intuitive, since feedback gates can store information in their probability tables whereas agents that do not use them need to store all information in their WM. This suggests that there might be a difference in the evolved strategy between those agents that use feedback gates and those who do not. When observing the behavior of the differently evolved agents we find that there are two types of strategies (see Supplementary Information Figs [5](#MOESM1){ref-type="media"}--[6](#MOESM1){ref-type="media"} for details). Agents that evolved brains without feedback gates use a simple heuristic that makes them repeat their last action when the arrow they stand on points into the direction they are facing. Otherwise, they start rotating until they move off a tile, which often results in them standing again on a tile that points into the direction they are facing, which makes them repeat the last action (see Supplementary Information Fig. [7](#MOESM1){ref-type="media"}).Figure 5Correlation of number of feedback gates to deterministic gates. Each dot represents the number of feedback gates the last agent on the line of descent has versus the number of deterministic gates it has in 300 replicates. The more feedback gates an agent has the less deterministic gates it evolves; the feedback gates allow agents to decrease brain size while still solving the task. The Pearson correlation coefficient between the number of feedback gates and deterministic gates is −0.52 with p-value of 8.64*e* ^−06.^ Agents that evolved to use feedback gates appear to actually behave as if they learn how to turn properly. They make several mistakes in the beginning, but after a learning period perform flawlessly. Of the 300 replicate evolutionary experiments where agents were allowed to use all types of logic gates 56 did not evolve using feedback gates. Comparing the actions those 56 agents take with the remaining 244 which do use feedback gates we first find that as expected the group using feedback gates reached the goal more often on average (5.33 times, versus 4.89 times for those agents not using feedback gates) which suggests a difference in behavior. The usage of actions is also drastically different during evolution (see Fig. [6](#Fig6){ref-type="fig"}). Agents using feedback gates reduce the instances where they do nothing, minimize turns and maximize moving forward. Agents not using feedback gates are less efficient because they rely on forward movements while minimizing times where they do nothing and turns. In conjunction with the observations made before we conclude that indeed agents not using feedback gates use some form of heuristic with a minimal amount of memory while agents using feedback gates learn to navigate the environment properly and quickly (see Supplementary Information Fig. [8](#MOESM1){ref-type="media"} and Section [8](#MOESM1){ref-type="media"} for a more details regarding the evolved solutions).Figure 6Different actions performed while navigating over evolutionary time. Both panels show the average use of forward (green), do nothing (red), and turn (black) commands over generations. Gray background indicates the standard error. Panel A shows those 244 agents that do use feedback gates, Panel B shows the remaining 56 independently evolved agents that do not use feedback gates. Agents on the LOD were analyzed. Discussion {#Sec7} ========== In prior experiments,  Markov Brains could evolve to solve various complex tasks and even form memories that represent their environment^[@CR27]^. However, these MBs use binary hidden states to store information about the environment similar to WM or STM, but lacked a feedback mechanism that allowed them to change their internal structure, very much like long term potentiation would lead to LTM. Other systems like artificial neural networks already allowed for a similar process, where weights could be adapted due to feedback. We augmented MBs with feedback gates that use the multiplicative weight update algorithm to change their probabilities given positive or negative feedback. It is important to note that this feedback is not directly provided by the environment to the agents but must be internally generated by the MBs. In addition, MBs need to integrate these feedback gates into their cognitive machinery during evolution. We showed that we can successfully evolve MBs to use feedback gates. These gates generated internal feedback is and they change their internal probabilities as expected. However, we also find competing strategies, which instead of evolving to learn deploy some form of heuristic. In the future it might be interesting to study under which evolutionary conditions either heuristics or learning strategies evolve (similar to Kvam 2017). We used a biological inspired task and we see the future application of this technology in the domain of computational modeling to study how intelligence evolved in natural systems and to eventually use neuroevolution as the means to bring about general artificial intelligence. ANNs that are specifically optimized by machine learning will solve specific classification tasks much faster than the model introduced here. The idea is not to present a new paradigm for machine learning, but to implement a tool to study the evolution of learning. While using MBs augmented with feedback gates will probably not be competitive with other supervised learning techniques, it remains an interesting question: How would typical machine learning tools perform if challenged with the task presented here? (see Supplementary Information Fig. [9](#MOESM1){ref-type="media"} for a comparison with Q learning on a single environment). One problem encountered by systems that receive external feedback is "catastrophic forgetting"^[@CR58],[@CR59]^. When the task changes, feedback will cause the system to adapt to their new objective but they forget or unlearn former capabilities. Requiring adaptive systems to interpret the environment in order to generate internal feedback might be a way to overcome this problem; assuming that the fitness function or evolutionary environment not only changes but also actively rewards organisms who do not suffer from "catastrophic forgetting". We now have the ability to evolve machines inspired by natural organisms that can learn without an objective external signal. This gives us a tool to study the evolution of learning using a computational model system instead of having to use natural organisms. It will also allow us to study the interplay between evolution and learning (aka Baldwin effect) and explore under which circumstances evolution benefits from learning and when it does not; we propose to use this model to study these questions in the future. Another dimension we will investigate in the future is the ability of MBs to change their connections due to feedback, not just the probabilities within their gates. Conclusion {#Sec8} ========== As stated before, combining evolution with learning is not a new idea. We think that it is in principle very easy for other systems to internalize the feedback. For example, it should be easy to evolve an artificial neural network to first interpret the environment, and then use this information to apply feedback on itself. However, we need to ask under which circumstances this is necessary. By providing training classes for supervised learning situations we can already create (or deep learn) machines that can learn to classify these classes. In addition, we often find these classifiers exceed human performance^[@CR60],[@CR61]^. If we are incapable of providing examples of correctly classified data we use unsupervised learning methods and only need to provide a fitness function that quantifies performance. But we know that fitness functions can be deceptive and designing them is sometimes more of an art than a science. When interacting with future AI systems we should find a different way to specify what we need them to do. Ideally they should autonomously explore the environment and learn everything there is to know without human intervention - nobody tells us humans what to research and explore, evolution primed us to pursue this autonomously. The work presented here is one step into this direction and will allow us to study evolution of learning in a biological context as well as explore how we can evolve machines to autonomously learn. Methods {#Sec9} ======= Markov Brains are networks of probabilistic and deterministic logic gates encoded by a genome. The genome contains genes and each gene specifies one logic gate, the logic it performs and how it is connected to sensors, motors, and to other gates^[@CR27],[@CR45],[@CR50]^. A new type of gate, the feedback gate, has been added to the Markov Brain framework (<https://github.com/LSheneman/autonomous-learning>),and this framework has been used to run all the evolutionary experiments. The Markov Brain framework has since been updated to MABE^[@CR62]^. See below for a detailed description of each component: Environment {#Sec10} ----------- The environment the agents had to navigate was a 2D spatial grid of 64 × 64 tiles. Tiles were either empty or contained a solid block that could not be traversed. The environment was surrounded by those solid blocks to prevent the navigating agent from leaving that space. At the beginning of each agent evaluation a new environment was generated and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{1}{7}$$\end{document}$ of the tiles were randomly filled with a solid block. The randomness of the environments maintains a complex maze-like structure across environments, but no two agents saw the exact same environment. A target was randomly placed in the environment, and Dijkstra's algorithm was used to compute the distance from all empty tiles to the target tile. These distances were used to label each empty block so that it had an arrow facing to the next closest tile to the target. When there was ambiguity (two adjacent tiles had the same distance) a random tile of the set of closest tiles was chosen. At birth agents were randomly placed in a tile that had a Dijkstra's number of 32 and face a random direction (up, right, down, or left). Due to the random placement of blocks it was possible that the goal was blocked so that there was no tile that is 32 tiles away, in which case a new environment was created, which happened only very rarely. Agents were then allowed to move around the environment for 512 updates. If they were able to reach the target, a new random start orientation and location with a Dijkstra's number of 32 was selected. Agents used two binary outputs from the MB to indicate their actions-- 00, 01, 10, or 11. Each output was translated using a mapping function to one of four possible actions- move forward, do nothing, turn left, or turn right. This resulted in 24 different ways to map the four possible outputs of the MB to the four possible actions that moved the agent. The input sensors gave information about the label of the tile the agent was standing on. Observe that the agent itself had an orientation and the label was interpreted relative to the direction the agent faced. There were four possible arrows the agent could see--forward, right, backward, or left--and were encoded as four binary inputs, one for each possible direction. Beyond the four input and two outputs nodes, agents could use 10 hidden nodes to connect their logic gates. Performance (or fitness) was calculated by exposing the agent to all 24 mappings and testing how often it reached the goal within the 512 updates it was allowed to explore the world. At every update agents were rewarded proportional to their distance to the goal (*d*), and received a bonus (*b*) every time they reached the goal, thus the fitness function becomes:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$W=\prod _{n\mathrm{=0}}^{n\mathrm{ < 24}}((\sum _{t\mathrm{=0}}^{t\mathrm{ < 512}}\frac{1}{1+d})+b)$$\end{document}$$ Selection {#Sec11} --------- After all agents in a population were tested on all 24 option-to-action mappings at each generation, the next generation was selected using tournament selection where individuals are randomly selected and the one with the best fitness transmits offspring into the next generation^[@CR63]^. The tournament size was set to five. Mutation {#Sec12} -------- Genomes for organisms in the first generation were generated randomly with a length of 5000 and 12 start codons were inserted that coded for deterministic, probabilistic, and feedback gates. Each organism propagated into the next generation inherited the genome of its ancestor. The genome had at least 1,000 and at most 20,000 sites. Each site had a 0.003 chance to be mutated. If the genome had not reached its maximum size stretches of a randomly selected length between 128 and 512 nucleotides got copied and inserted at random locations with a 0.02 likelihood. This allowed for gene duplications. If the genome was above 1000 nucleotides, there was a 0.02 chance for a stretch of a randomly selected length between 128 and 255 nucleotides to be deleted at a random location. Feedback Gates {#Sec13} -------------- At every update of a probabilistic gate, an input *i* resulted in a specific output *o*. To encode the mapping between all possible inputs and outputs of a gate we used a probability matrix *P*. Each element of this matrix *P* ~*io*~ defined the probability that given the input *i* the output *o* occurred. Observe that for each *i* the sum over all *o* must be 1.0 to define a probability:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1.0=\sum _{o\mathrm{=0}}^{O}{P}_{io}$$\end{document}$$where *O* defines the maximum number of possible outputs of each gate. A feedback gate uses this mechanism to determine its output at every given update. However, at each update we consider the probability *P* ~*io*~ that resulted in the gate\'s output to be causally responsible. If that input-output mapping for that update was appropriate then in future updates that probability should be higher. If the response of the gate at that update had negative consequences, then the probability should be lower. As explained above, the sum over all probabilities for a given input must sum to one. Therefore, a single probability can not change independently. If a probability is changed, the other probabilities are normalized so that Eq. [2](#Equ2){ref-type=""} remains true. But where is the feedback coming from that defines whether or not the action of that gate was negative or positive? Feedback gates posses two more inputs, one for a positive signal, and one for a negative signal. These inputs can come from any of the nodes the Markov Brain has at its disposal, and are genetically encoded. Therefore, the feedback can be a sensor or an output of another gate. Receiving a 0at either of the two positive or negative feedback inputs has no effect, whereas reading a 1 triggers the feedback. In the simplest case of feedback a random number in the range \[0, *δ*\] is applied to the probability *P* ~*io*~ that was used in the last update of the gate. In case of positive feedback the value is increased; in the case of negative feedback the value is decreased. The probabilities cannot exceed 0.99 or drop below 0.01. The rest of the probabilities are then normalized. The effects of the feedback gate are immediately accessible to the MB. However, because MBs are networks the signal that a feedback gate generates might need time to be relayed to the outputs via other gates. It is also possible that there is a delay between an agent's actions and the time it takes to receive new sensorial inputs that give a clue about the situation being improved or not. Thus, allowing feedback to occur only on the last action is not sufficient. Therefore, feedback gates can evolve the depth of a buffer that stores prior *P* ~*io*~ values up to a depth of 4. When feedback is applied all the probabilities identified by the elements in the queue are altered. The *δ* is determined by evolution and can be different for each element in the queue. Line of descent {#Sec14} --------------- An agent was selected at random from the final generation to determine the line of descent (LOD) by tracing the ancestors to the first generation^[@CR64]^. During this process, the most recent common ancestor (MRCA) is quickly found. Observe that all mutations that swept the population can be found on the LOD, and the LOD contains all evolutionary changes that mattered. Electronic supplementary material ================================= {#Sec15} Supplementary Material **Electronic supplementary material** **Supplementary information** accompanies this paper at 10.1038/s41598-017-16548-2. **Publisher\'s note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This work is in part supported by the NSF BEACON-center for the study of evolution in action under Cooperative Agreement DBI-0939454. We would like to thank Christoph Adami, Charles C. Ofria, and Fred C. Dyer for useful discussions. L.S. and A.H. conceived of the experiments and wrote the manuscript. L.S. performed all computational experiments and data analysis. Competing Interests {#FPar1} =================== The authors declare that they have no competing interests.