Advertisement
Journal of Investigative Dermatology Home

A GWAS Finds Variants at 2p21 Associated with Self-Reported Sensitive Skin in the Han Chinese Population

  • Author Footnotes
    7 These authors contributed equally to this work
    Bingjie Li
    Footnotes
    7 These authors contributed equally to this work
    Affiliations
    Department of Skin and Cosmetic Research, Shanghai Skin Disease Hospital, Tongji University School of Medicine, Tongji University, Shanghai, China

    CAS Key Laboratory of Computational Biology, Shanghai Institute of Nutrition and Health, University of Chinese Academy of Sciences, Chinese Academy of Sciences, Shanghai, China
    Search for articles by this author
  • Author Footnotes
    7 These authors contributed equally to this work
    Xiyang Cai
    Footnotes
    7 These authors contributed equally to this work
    Affiliations
    CAS Key Laboratory of Computational Biology, Shanghai Institute of Nutrition and Health, University of Chinese Academy of Sciences, Chinese Academy of Sciences, Shanghai, China
    Search for articles by this author
  • Author Footnotes
    7 These authors contributed equally to this work
    Lizhong Wang
    Footnotes
    7 These authors contributed equally to this work
    Affiliations
    Hunan Provincial Key Lab on Bioinformatics, School of Computer Science and Engineering, Central South University, Changsha, China

    WeGene, Shenzhen Zaozhidao Technology, Shenzhen, China
    Search for articles by this author
  • Jiarui Li
    Affiliations
    CAS Key Laboratory of Computational Biology, Shanghai Institute of Nutrition and Health, University of Chinese Academy of Sciences, Chinese Academy of Sciences, Shanghai, China
    Search for articles by this author
  • Ying Zou
    Affiliations
    Department of Skin and Cosmetic Research, Shanghai Skin Disease Hospital, Tongji University School of Medicine, Tongji University, Shanghai, China
    Search for articles by this author
  • Gang Chen
    Affiliations
    Hunan Provincial Key Lab on Bioinformatics, School of Computer Science and Engineering, Central South University, Changsha, China

    WeGene, Shenzhen Zaozhidao Technology, Shenzhen, China

    Shenzhen WeGene Clinical Laboratory, Shenzhen, China
    Search for articles by this author
  • Sijia Wang
    Correspondence
    Corresponding author
    Affiliations
    CAS Key Laboratory of Computational Biology, Shanghai Institute of Nutrition and Health, University of Chinese Academy of Sciences, Chinese Academy of Sciences, Shanghai, China

    Center for Excellence in Animal Evolution and Genetics, Chinese Academy of Sciences, Kunming, China
    Search for articles by this author
  • Author Footnotes
    7 These authors contributed equally to this work
Open AccessPublished:May 19, 2021DOI:https://doi.org/10.1016/j.jid.2021.04.021
      To the Editor
      Sensitive skin is defined by the occurrence of unpleasant sensations (e.g., stinging, burning, pain, pruritus, tingling) in response to various stimuli (e.g., cosmetics, temperature variation, emotion) that generally should not provoke such sensations (
      • Brenaut E.
      • Barnetche T.
      • Le Gall-Ianotto C.
      • Roudot A.C.
      • Misery L.
      • Ficheux A.S.
      Triggering factors in sensitive skin from the worldwide patients’ point of view: a systematic literature review and meta-analysis.
      ;
      • Misery L.
      • Ständer S.
      • Szepietowski J.C.
      • Reich A.
      • Wallengren J.
      • Evers A.W.
      • et al.
      Definition of sensitive skin: an expert position paper from the Special Interest Group on Sensitive Skin of the International Forum for the Study of Itch.
      ). The face tends to be the most common site of skin sensitivity owing to its rich innervation and its exposure to multiple types of irritants. Facial sensitive skin affects around 55‒57% of the Asian population but varies substantially among countries and studies (
      • Kamide R.
      • Misery L.
      • Perez-Cullell N.
      • Sibaud V.
      • Taïeb C.
      Sensitive skin evaluation in the Japanese population.
      ;
      • Kim Y.R.
      • Cheon H.I.
      • Misery L.
      • Taieb C.
      • Lee Y.W.
      Sensitive skin in Korean population: an epidemiological approach.
      ). Genetic factors probably play a role, supported by the significant ethnic differences in skin sensitivity (
      • Jourdain R.
      • de Lacharrière O.
      • Bastien P.
      • Maibach H.I.
      Ethnic variations in self-perceived sensitive skin: epidemiological survey.
      ) and relevant skin properties (e.g., permeability [
      • Kompaore F.
      • Tsuruta H.
      In vivo differences between Asian, black and white in the stratum corneum barrier function.
      ] and transepidermal water loss [
      • Machado M.
      • Hadgraft J.
      • Lane M.E.
      Assessment of the variation of skin barrier function with anatomic site, age, gender and ethnicity.
      ]). However, the genetics of sensitive skin is largely unknown. In a clinical study, sensitive skin is usually measured by stinging test methods (
      • Christensen M.
      • Kligman A.M.
      An improved procedure for conducting lactic acid stinging tests on facial skin.
      ), but this is not quite feasible in population-based studies. Instead, self-reported questionnaires have been the validated tools for identifying individuals with sensitive skin (
      • Brenaut E.
      • Barnetche T.
      • Le Gall-Ianotto C.
      • Roudot A.C.
      • Misery L.
      • Ficheux A.S.
      Triggering factors in sensitive skin from the worldwide patients’ point of view: a systematic literature review and meta-analysis.
      ). In this study, we used self-reported sensitive skin as the phenotype for our GWAS. A direct-to-consumer platform (WeGene) was used to collect questionnaires and DNA data, a strategy that has been successfully applied in previous GWASs of various phenotypes (
      • Kang K.
      • Sun X.
      • Wang L.
      • Yao X.
      • Tang S.
      • Deng J.
      • et al.
      Direct-to-consumer genetic testing in China and its role in GWAS discovery and replication.
      ;
      • Shaffer J.R.
      • Li J.
      • Lee M.K.
      • Roosenboom J.
      • Orlova E.
      • Adhikari K.
      • et al.
      Multiethnic GWAS reveals polygenic architecture of earlobe attachment.
      ). The discovery cohort was collected through an online platform in February and March 2018 (n = 1,872, age = 29.0 ± 7.2 years, 63.68% females) (Supplementary Table S1). An independent replication cohort was collected online in the next 10 months after the discovery set (n = 817, age = 28.6 ± 8.1 years, 62.30% females). Sensitive skin was defined as a self-reported case/control trait. Saliva samples of subjects were collected and then genotyped on Affymetrix WeGene V1 Arrays (Santa Clara, CA). Imputation was performed by Eagle and Minimac4 using 1000 Genomes phase 3 data as the reference. After quality control, a total of 596,744 genotyped SNPs and 5,370,088 imputed SNPs were included in the GWAS (Supplementary Materials and Methods). We found no outlier in our samples from a principal component analysis (Supplementary Figure S1). This research was conducted with official approval from the Ethical Committee of WeGene, and all subjects provided electronic informed consent.
      When asked, “Would you consider your facial skin sensitive, and prone to a sensation of redness, itching, tension, sting, or burning when exposed to external stimuli?,” 55.56% of the subjects in the discovery cohort and 59.73% in the replication cohort reported “very or rather sensitive” (Supplementary Tables S1 and S2), similar to the prevalence (55‒57%) reported in East Asian populations (
      • Kamide R.
      • Misery L.
      • Perez-Cullell N.
      • Sibaud V.
      • Taïeb C.
      Sensitive skin evaluation in the Japanese population.
      ;
      • Kim Y.R.
      • Cheon H.I.
      • Misery L.
      • Taieb C.
      • Lee Y.W.
      Sensitive skin in Korean population: an epidemiological approach.
      ). The prevalence of self-reported sensitive skin was significantly higher in females than in males (63.30% vs. 45.67%, P = 9.98 × 10–19), also consistent with findings from previous studies (
      • Chen W.
      • Dai R.
      • Li L.
      The prevalence of self-declared sensitive skin: a systematic review and meta-analysis.
      ). Overall, the age of the individuals with self-reported sensitive skin was slightly younger than that of the control group (P = 0.04); however, there was no significant association after sex stratification (Pfemale = 0.17, Pmale = 0.15) (Supplementary Figure S2).
      We then performed a GWAS using logistic regression with an additive genetic model in the discovery cohort, including sex, age, and the first five genetic principal components as covariates. We identified a signal at 2p21 significantly associated with self-reported sensitive skin (rs72803822 as the lead SNP; OR = 1.47, confidence interval = 1.28–1.68; P = 2.28 × 10–8) (Table 1 and Supplementary Figure S3a). This locus was successfully replicated (P = 0.03) in the replication cohort (Table 1). In the meta-analysis of the discovery and replication cohorts, four SNPs in strong linkage disequilibrium (r2 = 0.95~0.99) at 2p21 reached the genome-wide significance level (OR = 1.42, confidence interval = 1.36–1.48; P = 3.78 × 10–9) (Figure 1a and b and Supplementary Figure S3b). Analysis after sex stratification showed that the effect size of the associated SNPs did not differ between males and females (Supplementary Table S3). In a GWAS further incorporating dermatologic conditions associated with self-reported sensitive skin as covariates (e.g., eczema/dermatitis, eczema/dermatitis since childhood, and asthma/hay fever) (Supplementary Table S4), the locus remained genome-wide significant in a meta-analysis (OR = 1.42, confidence interval = 1.30–1.53; P = 5.43 × 10–9) (Supplementary Figure S3c), indicating that these findings were not affected by the skin disease history. None of the above SNPs has ever been reported to be associated with any dermatologic conditions in the GWAS catalog.
      Table 1Summary of the Four Signal SNPs at 2p21 in GWAS


      CHR
      SNPBP
      Positions are according to human reference hg19.
      REFALTEAEAFDiscoveryReplicationMeta-Analysis
      OR95% CIP-ValueOR95% CIP-ValueOR95% CIP-Value
      2rs1703020343155982TGG0.5031.45(1.27–1.66)4.17 × 10−81.25(1.02–1.53)0.0341.39(1.24–1.55)9.04 × 10−9
      2rs1703020643156974TCC0.5051.45(1.27–1.66)3.82 × 10−81.23(1.00–1.51)0.0491.38(1.23–1.54)1.35 × 10−8
      2rs5790898143157164CAA0.5041.46(1.28–1.67)3.11 × 10−81.23(1.00–1.51)0.0441.39(1.24–1.55)9.90 × 10−9
      2rs7280382243158175GAA0.5021.47(1.28–1.68)2.28 × 10−81.26(1.03–1.56)0.0261.42(1.25–1.56)3.78 × 10−9
      Abbreviations: ALT, alternative allele; BP, base-pair; CHR, chromosome; CI, confidence interval; EA, effect allele; EAF, effect allele frequency; REF, reference allele.
      1 Positions are according to human reference hg19.
      Figure thumbnail gr1ac
      Figure 1GWAS of self-reported sensitive skin identified a significant signal at 2p21. (a) Manhattan plot and quantile‒quantile plot showing the results of the meta-analysis for GWASs of self-reported sensitive skin on discovery and replication sets. The red line corresponds to the genome-wide threshold (P ≤ 5 × 10–8). SNPs within 5 kb to the signals were plotted in red. (b) Regional association plot for the significant region at 2p21. Increasing color intensities indicated increasing linkage disequilibrium (r2) with rs72803822. (c) Epigenetic annotation at a region within 1 kb to rs17030203, 160 kb upstream of MTA3, and outlined in yellow. (d) Functionality scores from 3DSNP annotated the significant SNPs as TFBSs and enhancers. (e) The effect allele G on rs17030203 showed an increasing effect on more sensitive skin. (f) Geography of allele frequency on rs17030203. Chr, chromosome; Ctrl, control; Mb, megabase; SS, sensitive skin; TFBS, transcriptionfactor–binding site.
      Figure thumbnail gr1df
      Figure 1GWAS of self-reported sensitive skin identified a significant signal at 2p21. (a) Manhattan plot and quantile‒quantile plot showing the results of the meta-analysis for GWASs of self-reported sensitive skin on discovery and replication sets. The red line corresponds to the genome-wide threshold (P ≤ 5 × 10–8). SNPs within 5 kb to the signals were plotted in red. (b) Regional association plot for the significant region at 2p21. Increasing color intensities indicated increasing linkage disequilibrium (r2) with rs72803822. (c) Epigenetic annotation at a region within 1 kb to rs17030203, 160 kb upstream of MTA3, and outlined in yellow. (d) Functionality scores from 3DSNP annotated the significant SNPs as TFBSs and enhancers. (e) The effect allele G on rs17030203 showed an increasing effect on more sensitive skin. (f) Geography of allele frequency on rs17030203. Chr, chromosome; Ctrl, control; Mb, megabase; SS, sensitive skin; TFBS, transcriptionfactor–binding site.
      All four SNPs are located in the intergenic region between ZFP36L2 and MTA3 genes. Interestingly, rs17030203, one of the significant SNPs, is an expression quantitative trait locus of MTA3 in the blood (z-score = 3.07, P = 0.002 [
      • Westra H.J.
      • Peters M.J.
      • Esko T.
      • Yaghootkar H.
      • Schurmann C.
      • Kettunen J.
      • et al.
      Systematic identification of trans eQTLs as putative drivers of known disease associations.
      ]; data from the other tissues are not significant in Genotype-Tissue Expression). Reportedly, MTA3 plays a role in the maintenance of the normal epithelial architecture by regulating E-cadherin levels (
      • Fujita N.
      • Jaye D.L.
      • Kajita M.
      • Geigerman C.
      • Moreno C.S.
      • Wade P.A.
      MTA3, a Mi-2/NuRD complex subunit, regulates an invasive growth pathway in breast cancer.
      ). E-cadherin and its encoding gene CDH1, essential in the maintenance of epithelium integrity and keratinocyte differentiation, was recently found to be upregulated in sensitive skin samples (
      • Kim E.J.
      • Lee D.H.
      • Kim Y.K.
      • Kim M.K.
      • Kim J.Y.
      • Lee M.J.
      • et al.
      Decreased ATP synthesis and lower pH may lead to abnormal muscle contraction and skin sensitivity in human skin.
      ). Querying the signal locus against the 3DSNP database, we found high functionality scores on transcription factor‒binding sites and enhancers (53.51~100) (Figure 1d). In addition, on the basis of the encyclopedia of DNA elements and Roadmap database, the locus exhibited distinct signatures of active enhancer epigenetic markers such as H3K4me1 histone modifications in fibroblasts and keratinocytes (Figure 1c). This evidence thus revealed that the signal locus has a potential regulatory function. Taken together, we speculate that SNP rs17030203 regulates the expression of the MTA3 gene in an enhancer/transcription factor binding sites‒dependent manner, and thus regulating E-cadherin level to affect skin sensitivity.
      Individuals with the derived allele (G) of rs17030203 have a higher frequency of self-reported sensitive skin, with an increased prevalence of approximately 7.6% per copy (Figure 1e). The derived allele of rs17030203 is of high frequencies in East Asian populations but mostly rare in other populations (Figure 1f). The other three significant SNPs in linkage disequilibrium also show similar patterns (Supplementary Figure S4 and Supplementary Table S5). These results were consistent with the report of comparatively higher skin sensitivity in East Asians than in Europeans (
      • Aramaki J.
      • Kawana S.
      • Effendy I.
      • Happle R.
      • Löffler H.
      Differences of skin irritation between Japanese and European women.
      ). However, we found no signal for positive natural selection in the 2p21 region in East Asians, Europeans, or Africans (Supplementary Figure S5).
      In conclusion, this GWAS on self-reported sensitive skin in the Han Chinese population, to our knowledge previously unreported, identified a risk locus at 2p21, and one of the lead SNPs, rs17030203, is an expression quantitative trait locus of MTA3. A limitation of the study is that the self-reported phenotype is likely composed of several endophenotypes, and the questionnaire-based approach makes the phenotype even rougher. Further investigations on the role of subjective perception on sensitive skin will be beneficial. On the other hand, genetic findings of this GWAS are likely robust. Our findings provide insights into a mechanism involving the disrupting of the maintaining epithelial architecture in the development of sensitive skin.

       Data availability statement

      The GWAS summary statistics were deposited in the publicly available National Omics Data Encyclopedia (http://www.biosino.org/node/) and are available under the project identification document OEP001547. Data usage shall be in full compliance with the Regulations on Management of Human Genetic Resources in China.

      ORCIDs

      Conflict of Interest

      LW and GC are employees of WeGene. The remaining authors state no other conflict of interest.

      Acknowledgments

      This work was supported by the National Key Research and Development Project (grant number 2018YFC0910403 to SW), the Strategic Priority Research Program of the Chinese Academy of Sciences (grant number XDB38020400 to SW), and the Shanghai Municipal Science and Technology Major Project (grant number 2017SHZDZX01 to SW). We thank all WeGene users who consented to participate in the research. For data-related information, please correspond with GC ( [email protected] ).

      Author Contributions

      Conceptualization: BL, SW; Data Curation: BL, XC; Formal Analysis: BL, XC, LW; Funding Acquisition: SW; Investigation: BL, XC; Methodology: BL, XC, LW; Project Administration: BL, GC, SW; Resources: LW, GC, SW; Software: BL, XC, LW; Supervision: GC, SW; Visualization: BL, XC, LW, GC, SW; Writing - Original Draft Preparation: BL, XC; Writing - Review and Editing: JL, YZ, GC, SW

      Supplementary Materials and Methods

       Population and samples

      All participants were recruited from consenting WeGene customers from Shenzhen Zaozhidao Technology, a direct-to-consumer genetic testing service provider. The discovery set was collected in February and March 2018 and included 1,872 volunteers (1,192 females and 680 males). The replication set was collected in the following 10 months and included 817 volunteers (509 females and 308 males). The sample summary is provided in Supplementary Table S1.

       Phenotyping

      Information on personal data and phenotypes for each participant was gathered through a self-reported questionnaire. Participants replied to the question “Would you consider your facial skin sensitive, and prone to a sensation of redness, itching, tension, sting, or burning when exposed to external stimuli?” In the following analysis, we defined individuals who answered very sensitive and rather sensitive as a sensitive group and those who answered slightly sensitive and not sensitive as a nonsensitive group (
      • Kamide R.
      • Misery L.
      • Perez-Cullell N.
      • Sibaud V.
      • Taïeb C.
      Sensitive skin evaluation in the Japanese population.
      ). Potentially relevant dermatologic conditions (e.g., eczema/dermatitis, eczema/dermatitis since childhood, and asthma/hay fever) were also queried in the survey. The initial survey questions were worded in Chinese and are shown in Supplementary Table S2.

       Genotyping

      Saliva samples for DNA extraction and genotyping were collected and then genotyped on the Affymetrix WeGene V1 Arrays (Santa Clara, CA), which covered 596,744 SNPs at the WeGene genotyping center, Shenzhen. To control for genotype quality, we used PLINK, version 1.9 (
      • Purcell S.
      • Neale B.
      • Todd-Brown K.
      • Thomas L.
      • Ferreira M.A.
      • Bender D.
      • et al.
      PLINK: a tool set for whole-genome association and population-based linkage analyses.
      ), to exclude individuals with >5% missing data, outlying heterozygous rate, or discordant sex information. Unrelated was filtered by pairwisely checking for all the samples, and those whose identities were identified by descent scores >0.125 were removed. Ancestry was assigned from self-reported surveys and further examined with principal component analysis. We also discarded SNPs with unbalanced call rates in case and controls, those with >2% missing data, those with a minor allele frequency <1%, or any SNPs that failed the Hardy Weinberg deviation test (P < 1 × 10–5). After pruning, 373,988 SNPs were retained for further analysis. We used Eagle (
      • Loh P.R.
      • Danecek P.
      • Palamara P.F.
      • Fuchsberger C.
      • Reshef Y.A.
      • Finucane H.K.
      • et al.
      Reference-based phasing using the Haplotype Reference Consortium panel.
      ) and Minimac4 (
      • Fuchsberger C.
      • Abecasis G.R.
      • Hinds D.A.
      minimac2: faster genotype imputation.
      ) to impute the genotypes at nongenotyped SNPs with a 10-Mb chunk size, a 3-Mb step size, and 1000 Genomes phase 3 data (
      • Auton A.
      • Brooks L.D.
      • Durbin R.M.
      • Garrison E.P.
      • Kang H.M.
      • et al.
      1000 Genomes Project Consortium
      A global reference for human genetic variation.
      ) as an imputation reference panel. X chromosome was also phased and imputed by Eagle and Minimac4 following the guidelines of Minimac4. The final imputed dataset contained genotypes for 5,370,088 autosomal SNPs and 183,199 X-chromosomal SNPs with an imputation quality score (Minimac Rsq) > 0.3, mean allele frequency > 1%, and missing rate < 2%.

       Statistical analyses

       Population stratification analysis

      We first used PLINK, version 1.9 (
      • Purcell S.
      • Neale B.
      • Todd-Brown K.
      • Thomas L.
      • Ferreira M.A.
      • Bender D.
      • et al.
      PLINK: a tool set for whole-genome association and population-based linkage analyses.
      ), to carried out principal component analysis to correct for possible population stratification. We combined our datasets with samples of 102 Han Chinese in Beijing, China, 98 Utah residents with Northern and Western European ancestry from the CEPH collection, and 107 Yoruba in Ibadan, Nigeria from 1000 Genomes phase 3 data and selected 56,723 SNPs with linkage equilibrium r2 < 0.1 for analysis to account for potential biases introduced by linkage disequilibrium structure. Han Chinese in Beijing, China, and both our discovery and replication set were clustered, and there were no outliers in our dataset (Supplementary Figure S1). The top five genetic principal components were later included in the GWAS model. The number of principal components used was based on the proportion of variance explained and the degree of genomic inflation (lambda).

       Association test

      Correlation analysis was performed using R software (version 4.0.1) between sensitive skin and sex by Pearson chi-square test and between sensitive skin and chronological age by Welch two-sample t-test. We found that reported case of self-reported sensitive skin was significantly higher in females than in males (63.30% vs. 45.67%, P = 9.98 × 10–19). Overall, the age of the individuals with self-reported sensitive skin was slightly younger than that of the control group (P = 0.04); however, there was no significant association between age and self-reported sensitive skin after stratified by sex (Pfemale = 0.17, Pmale = 0.15). The distribution of the trait stratified by sex and age is shown in Supplementary Figure S2. Age and sex are included as covariates in the GWAS model as a convention. We also performed Pearson chi-square test between potentially relevant dermatologic conditions with self-reported sensitive skin and found significant associations (Supplementary Table S4).
      We then performed a genome-wide association analysis for self-reported sensitive skin with logistic regression and an additive genetic model incorporating age, sex, and the first five genetic principal components. Associations were significant if the P-values passed the genome-wide significant threshold (Supplementary Figure S3a). We performed a meta-analysis with PLINK, including the discovery set and the replication set, applying the standard error‒based method (
      • Purcell S.
      • Neale B.
      • Todd-Brown K.
      • Thomas L.
      • Ferreira M.A.
      • Bender D.
      • et al.
      PLINK: a tool set for whole-genome association and population-based linkage analyses.
      ). X chromosomes were also included in the GWAS and meta-analysis (Supplementary Figure S3d). For loci of interest, we performed linkage disequilibrium analysis using an online tool LocusZoom (
      • Pruim R.J.
      • Welch R.P.
      • Sanna S.
      • Teslovich T.M.
      • Chines P.S.
      • Gliedt T.P.
      • et al.
      LocusZoom: regional visualization of genome-wide association scan results.
      ). We also performed a conditional GWAS, controlling the top-associated SNP to identify any independent signals (Supplementary Figure S3b). We further performed another GWAS, including potentially relevant dermatologic conditions as covariances to control for skin health (Supplementary Figure S3c). In addition, GWAS and meta-analysis were repeated after sex stratification (Supplementary Table S3) or on SNP‒sex interaction (Supplementary Figure S3e).

       Tests for natural selection

      To test for signals of natural selection, we used the Composite of Multiple Signals method (
      • Grossman S.R.
      • Shlyakhter I.
      • Karlsson E.K.
      • Byrne E.H.
      • Morales S.
      • Frieden G.
      • et al.
      A composite of multiple signals distinguishes causal variants in regions of positive selection.
      ), which incorporates the scores on the basis of three distinct signatures of selection: long-range haplotypes, differentiated alleles, and high frequency‒derived alleles. We tested for signals of positive natural selection with Composite of Multiple Signals scores at 2p21 region extracted from the genome-wide scores of Yoruba in Ibadan, Nigeria, Utah residents with Northern and Western European ancestry from the CEPH collection, and East Asia (Japanese in Tokyo, Japan + Han Chinese in Beijing, China) provided by the Broad Institute (https://www.broadinstitute.org/cms/cms-composite-multiple-signals). Results are presented in Supplementary Figure S5.

       Functional annotation

      We queried the signal locus against the 3DSNP database (
      • Lu Y.
      • Quan C.
      • Chen H.
      • Bo X.
      • Zhang C.
      3DSNP: a database for linking human noncoding SNPs to their three-dimensional interacting genes.
      ) for functionality scores, and we used HaploReg, version 4.1 (
      • Ward L.D.
      • Kellis M.
      HaploReg: a resource for exploring chromatin states, conservation, and regulatory motif alterations within sets of genetically linked variants.
      ), to annotate the identified variants for their functional relevance and the University of California, Santa Cruz Genome Browser (
      • Kent W.J.
      • Sugnet C.W.
      • Furey T.S.
      • Roskin K.M.
      • Pringle T.H.
      • Zahler A.M.
      • et al.
      The human genome browser at UCSC.
      ) on Human Feb. 2009 (GRCh37/hg19) Assembly for visualization (http://epigenomegateway.wustl.edu/legacy/). DNase hypersensitive site and Histone mark annotation (H3K27ac, H3K4me1) peaks for the human skin cells were obtained from the encyclopedia of DNA elements and Roadmap database (
      • Birney E.
      • Stamatoyannopoulos J.A.
      • Dutta A.
      • Guigó R.
      • Gingeras T.R.
      • et al.
      ENCODE Project Consortium
      Identification and analysis of functional elements in 1% of the human genome by the ENCODE pilot project.
      ). Results are presented in Figure 1c and d.
      Figure thumbnail fx1
      Supplementary Figure S1Structural analysis of the samples from discovery and replication sets, combined with representative populations from the 1000 Genomes Project. (a) Principal component analysis for the discovery and replication set showed that the two populations cluster together. (b) Joint principal component analysis with three populations from 1000 Genomes phase 3 data. The zoomed-in lot again showed that WeGene cohort samples and Han Chinese samples are mostly in one cluster, which indicated that our samples are well overlapped with Han Chinese samples from 1000 Genomes phase 3 data and that there were no outliers in our samples. CEU, Utah residents with Northern and Western European ancestry from the CEPH collection; CHB, Han Chinese in Beijing, China; Dis, discovery; PC, principal component; Rep, replication; YRI, Yoruba in Ibadan, Nigeria.
      Figure thumbnail fx2
      Supplementary Figure S2Distribution of self-reported sensitive skin, stratified by sex and chronological age. The case of the trait was counted in every 10 years of age, and the count numbers were marked on the corresponding bars.
      Figure thumbnail fx3ac
      Supplementary Figure S3Manhattan plot and quantile‒quantile plot showing the results of the genome-wide scan for self-reported sensitive skin in the discovery set, the conditional model, and a GWAS-controlling health factors. The panels showed the Manhattan plot illustrating the results of (a) the GWAS in 1,872 samples adjusted for age, sex, and the first five genetic PCs; (b) the conditional GWAS adjusted for age, sex, the first five genetic PCs, and the top-associated SNP rs72803822; (c) the meta-analysis of discovery and replication set, adjusted for age, sex, the first five genetic PCs, and potentially relevant health factors, eczema/dermatitis, atopic dermatitis in childhood, and asthma/hay fever; and (d) the meta-analysis of the discovery and replication set, adjusted for age, sex, the first five genetic PCs on the X chromosome, and (e) on SNP‒sex interaction. The red line corresponds to the threshold for genome-wide statistical significance (P ≤ 5 × 10–8). SNPs that were close (<5 kb) to the genome-wide significant signals were plotted in red. The quantile‒quantile plot (embedded) compared the empirical values on the vertical axis with simulated values on the horizontal axis. The x-axis and y-axis showed the expected P-values under null distribution and the observed P-values, respectively. The overall inflation of the observed versus expected distribution of association test statistics was reflected by lambda (λ). Here, λ = 0.998~1.006, showing no sign of confounding effects. PC, principal component.
      Figure thumbnail fx3de
      Supplementary Figure S3Manhattan plot and quantile‒quantile plot showing the results of the genome-wide scan for self-reported sensitive skin in the discovery set, the conditional model, and a GWAS-controlling health factors. The panels showed the Manhattan plot illustrating the results of (a) the GWAS in 1,872 samples adjusted for age, sex, and the first five genetic PCs; (b) the conditional GWAS adjusted for age, sex, the first five genetic PCs, and the top-associated SNP rs72803822; (c) the meta-analysis of discovery and replication set, adjusted for age, sex, the first five genetic PCs, and potentially relevant health factors, eczema/dermatitis, atopic dermatitis in childhood, and asthma/hay fever; and (d) the meta-analysis of the discovery and replication set, adjusted for age, sex, the first five genetic PCs on the X chromosome, and (e) on SNP‒sex interaction. The red line corresponds to the threshold for genome-wide statistical significance (P ≤ 5 × 10–8). SNPs that were close (<5 kb) to the genome-wide significant signals were plotted in red. The quantile‒quantile plot (embedded) compared the empirical values on the vertical axis with simulated values on the horizontal axis. The x-axis and y-axis showed the expected P-values under null distribution and the observed P-values, respectively. The overall inflation of the observed versus expected distribution of association test statistics was reflected by lambda (λ). Here, λ = 0.998~1.006, showing no sign of confounding effects. PC, principal component.
      Figure thumbnail fx4
      Supplementary Figure S4Effect and frequency of derived alleles on chromosome 2p21. The left panel showed the proportion of the trait against the genotypes of the three loci on region 2p21. The right panel showed the geography of allele frequency on the corresponding SNPs. The effect alleles were marked by blue. Allele frequencies were obtained from 1000 Genome data and visualized by the Geography of Genetic Variants Browser (
      • Marcus J.H.
      • Novembre J.
      Visualizing the geography of genetic variants.
      ). Ctrl, control; SS, sensitive skin.
      Figure thumbnail fx5
      Supplementary Figure S5Results of the test for natural selection on chromosome 2p21. CMS scores were plotted against physical distance for region 2p21 in CEU, CHB + JPT, and YRI populations, shown in green, orange, and blue, respectively. The LD block containing the four significantly associated SNPs was denoted by a yellow rectangle. No enrichment for high CMS scores (Top 0.1%) was found, indicating that the region was not subject to natural selection. CEU, Utah residents with Northern and Western European ancestry from the CEPH collection; CHB, Han Chinese in Beijing, China; CMS, Composite of Multiple Signal; LD, linkage disequilibrium; POS, position; JPT, Japanese in Tokyo, Japan; YRI, Yoruba in Ibadan, Nigeria.
      Supplementary Table S1Sample Summary of the Two Genome-Wide Studies in WeGene Cohort
      CharacteristicsDiscoveryReplication
      Subjects, n1,872817
      Sex, n (%)
       Female1,192 (63.68)509 (62.30)
       Male680 (36.32)308 (37.70)
      Age, years, mean (SD)29.03 (7.22)28.55 (8.14)
      Self-reported sensitive skin case, n (%)
       Very sensitive270 (14.42)152 (18.60)
       Rather sensitive770 (41.13)336 (41.13)
       Slightly sensitive591 (31.57)227 (27.78)
       Not sensitive241 (12.87)102 (12.48)
      Supplementary Table S2Original Survey Questions in Chinese
      请问您认为自己面部皮肤是否容易对外界刺激产生不适,如发红,瘙痒,紧绷,刺痛,灼烧感等等不舒服的感觉?:非常敏感/较为敏感/不太敏感/不敏感
      请问您目前是否患有湿疹或皮炎:是/否
      请问您是否从小就患有湿疹或皮炎(非后天传染所得):是/否
      请问您是否患有哮喘或者过敏性鼻炎:是/否
      Supplementary Table S3Summary of Locus at 2p21 in Meta-Analysis after Sex Stratification


      CHR
      SNPBP
      Base-pair positions are according to human reference hg19.
      EAFemale (n = 1,701)Male (n = 988)
      OR95% CIP-ValueOR95% CIP-Value
      243155982rs17030203G1.40(1.21–1.62)3.68 × 10‒61.36(1.14–1.63)0.0008
      243156974rs17030206C1.40(1.21–1.62)4.49 × 10‒61.35(1.12–1.62)0.0012
      243157164rs57908981A1.40(1.21–1.62)3.74 × 10‒61.35(1.13–1.61)0.0010
      243158175rs72803822A1.41(1.22–1.63)2.64 × 10‒61.39(1.16–1.67)0.0005
      Abbreviations: BP, base-pair; CHR, chromosome; CI, confidence interval; EA, effect allele.
      1 Base-pair positions are according to human reference hg19.
      Supplementary Table S4Breakdown of Dermatologic Conditions and Self-Reported Sensitive Skin


      Dermatologic Conditions
      CatalogSelf-Reported Sensitive Skin
      DiscoveryReplication
      CaseControlCaseControl
      Eczema/dermatitisCase40917321079
      (present)Control631659278250
      PChi-squared test1.14 × 10173.75 × 10–8
      Atopic dermatitis in childhoodCase189629129
      Control851770397300
      PChi-squared test2.14 × 10–111.49 × 10–4
      Asthma/hay feverCase44823918195
      Control592593307234
      PChi-squared test2.11 × 10–100.018
      Supplementary Table S5Allele Frequencies of Locus at 2p21 across Populations


      CHR
      BP
      Base-pair positions are according to human reference hg19.
      SNPEAFrequencies
      Allele frequencies were obtained from HaploReg, version 4.1.
      AFRAMRASNEUR
      242928842rs17030203G0.010.120.510.06
      242929834rs17030206C0.060.140.510.07
      242930024rs57908981A0.010.120.510.06
      242931035rs72803822A0.010.120.510.06
      Abbreviations: AFR, African; AMR, Ad Mixed American; ASN, Asian; BP, base-pair; CHB, Han Chinese in Beijing, China; EA, effect allele; EUR, European; JPT, Japanese in Tokyo, Japan.
      1 Base-pair positions are according to human reference hg19.
      2 Allele frequencies were obtained from HaploReg, version 4.1.

      References

        • Aramaki J.
        • Kawana S.
        • Effendy I.
        • Happle R.
        • Löffler H.
        Differences of skin irritation between Japanese and European women.
        Br J Dermatol. 2002; 146: 1052-1056
        • Brenaut E.
        • Barnetche T.
        • Le Gall-Ianotto C.
        • Roudot A.C.
        • Misery L.
        • Ficheux A.S.
        Triggering factors in sensitive skin from the worldwide patients’ point of view: a systematic literature review and meta-analysis.
        J Eur Acad Dermatol Venereol. 2020; 34: 230-238
        • Chen W.
        • Dai R.
        • Li L.
        The prevalence of self-declared sensitive skin: a systematic review and meta-analysis.
        J Eur Acad Dermatol Venereol. 2020; 34: 1779-1788
        • Christensen M.
        • Kligman A.M.
        An improved procedure for conducting lactic acid stinging tests on facial skin.
        J Soc Cosmet Chem. 1996; 47: 1-11
        • Fujita N.
        • Jaye D.L.
        • Kajita M.
        • Geigerman C.
        • Moreno C.S.
        • Wade P.A.
        MTA3, a Mi-2/NuRD complex subunit, regulates an invasive growth pathway in breast cancer.
        Cell. 2003; 113: 207-219
        • Jourdain R.
        • de Lacharrière O.
        • Bastien P.
        • Maibach H.I.
        Ethnic variations in self-perceived sensitive skin: epidemiological survey.
        Contact Dermatitis. 2002; 46: 162-169
        • Kamide R.
        • Misery L.
        • Perez-Cullell N.
        • Sibaud V.
        • Taïeb C.
        Sensitive skin evaluation in the Japanese population.
        J Dermatol. 2013; 40: 177-181
        • Kang K.
        • Sun X.
        • Wang L.
        • Yao X.
        • Tang S.
        • Deng J.
        • et al.
        Direct-to-consumer genetic testing in China and its role in GWAS discovery and replication.
        Quant Biol. 2020; : 1-15
        • Kim E.J.
        • Lee D.H.
        • Kim Y.K.
        • Kim M.K.
        • Kim J.Y.
        • Lee M.J.
        • et al.
        Decreased ATP synthesis and lower pH may lead to abnormal muscle contraction and skin sensitivity in human skin.
        J Dermatol Sci. 2014; 76: 214-221
        • Kim Y.R.
        • Cheon H.I.
        • Misery L.
        • Taieb C.
        • Lee Y.W.
        Sensitive skin in Korean population: an epidemiological approach.
        Skin Res Technol. 2018; 24: 229-234
        • Kompaore F.
        • Tsuruta H.
        In vivo differences between Asian, black and white in the stratum corneum barrier function.
        Int Arch Occup Environ Health. 1993; 65: S223-S225
        • Machado M.
        • Hadgraft J.
        • Lane M.E.
        Assessment of the variation of skin barrier function with anatomic site, age, gender and ethnicity.
        Int J Cosmet Sci. 2010; 32: 397-409
        • Misery L.
        • Ständer S.
        • Szepietowski J.C.
        • Reich A.
        • Wallengren J.
        • Evers A.W.
        • et al.
        Definition of sensitive skin: an expert position paper from the Special Interest Group on Sensitive Skin of the International Forum for the Study of Itch.
        Acta Derm Venereol. 2017; 97: 4-6
        • Shaffer J.R.
        • Li J.
        • Lee M.K.
        • Roosenboom J.
        • Orlova E.
        • Adhikari K.
        • et al.
        Multiethnic GWAS reveals polygenic architecture of earlobe attachment.
        Am J Hum Genet. 2017; 101: 913-924
        • Westra H.J.
        • Peters M.J.
        • Esko T.
        • Yaghootkar H.
        • Schurmann C.
        • Kettunen J.
        • et al.
        Systematic identification of trans eQTLs as putative drivers of known disease associations.
        Nat Genet. 2013; 45: 1238-1243