If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Internal Medicine, Erasmus MC University Medical Center Rotterdam, Rotterdam, The NetherlandsDepartment of Epidemiology, Erasmus MC University Medical Center Rotterdam, Rotterdam, The Netherlands
10 These authors, respectively, contributed equally to this work.
10 These authors, respectively, contributed equally to this work.
CAS Key Laboratory of Genomic and Precision Medicine, Beijing Institute of Genomic, University of Chinese Academy of Sciences, Chinese Academy of Sciences, Beijing, People’s Republic of ChinaDepartment of Genetic Identification, Erasmus MC University Medical Center Rotterdam, Rotterdam, The Netherlands
Eyebrow color shows a high degree of variation in Europeans. Although no heritability estimate has yet been reported, eyebrow color may share a large genetic component with scalp hair color, which has an estimated heritability of up to 90% (
). Although a phenotypic relationship between eyebrow and scalp hair color clearly exists, such a correlation is not perfect, suggesting the existence of overlapping and unique genetic components for both traits. Although previous genome-wide association studies (GWASs) on human eye (
) have identified multiple DNA variants, no GWAS for eyebrow color has been reported as of yet.
The cohort-related studies included in this study were approved by the medical ethics committee of the Erasmus University Medical Center, the St. Thomas’ Hospital local research ethics committee, the Queensland Institute of Medical Research (QIMR) Berghofer human research ethics committee, and the Indiana University internal review board. All participants provided written informed consent under protocols reviewed by the corresponding institutions.
The discovery stage meta-analysis of three GWASs for eyebrow color included a total of 6,513 European individuals from three cohorts, the Rotterdam Study (RS) (n = 3,114, mean age = 68.48 ± 9.34 years, 53.6% female), the TwinsUK study (n = 1,038, mean age = 59.47 ± 9.50 years, 100% female), and the QIMR study (n = 2,361, mean age = 16.43 ± 0.80 years, 54.0% female) (see Supplementary Table S1 online). Eyebrow color was graded into four broad ordinal categories (red, blond, brown, and black) by using photonumeric scales (see Supplementary Table S1). Detailed phenotype evaluation is provided in Supplementary Tables S2–S6 and the Supplementary Materials online.
The discovery stage meta-analysis of three GWASs identified a total of 355 single nucleotide polymorphisms (SNPs) at six distinct genetic loci showing genome-wide significant association with eyebrow color (P < 5 × 10–8) (Figure 1 and see Supplementary Figure S1 and Supplementary Table S7 online). Among these six loci, one locus (10q22.2: C10orf11) had not been previously associated with any other human pigmentation trait; the top-associated SNP (rs11001536; β = –0.21, P = 3.16 × 10–8) (see Supplementary Table S7) is an intronic DNA variant in C10orf11 (see Supplementary Figure S2 online). The remaining five loci have been repeatedly reported to have genome-wide significant association with human eye, scalp hair, and/or skin color. These include 15q13.1 (rs7494942; β = 0.22, P = 6.36 × 10–58 for HERC2 and rs4778237; β = 0.17, P = 7.01 × 10–31 for OCA2), 16q24.3 (MC1R rs75570604; β = –0.25, P = 9.88 × 10–52), 14q32.12 (SLC24A4 rs12883151; β = 0.10, P = 4.44 × 10–27), 20q11.22 (ASIP rs6059655, β = –0.11, P = 4.60 × 10–10), and 5p13.2 (SLC45A2 rs16891982, β = 0.18, Ps = 2.60 × 10–8) (see Supplementary Table S7).
The replication was conducted in 2,054 individuals of European origin from an additional US cohort (mean age = 25.75 ± 11.39 years, 68% female) (see Supplementary Table S1). Five loci highlighted in our discovery GWAS meta-analysis (SCL45A2, SCL24A4, HERC2 and OCA2, MC1R, and ASIP) have been successfully replicated in the US cohort (P < 0.05) (see Supplementary Table S7) and showed consistent allele effects in all four cohorts (see Supplementary Figure S3 online). The significant association for rs11001536 in C10orf11 highlighted in the discovery GWAS was not replicated in the US cohort. This SNP was nominally significant in the RS (β = –0.23, P = 2.16 × 10–6) and TwinsUK (β = –0.27, P = 4.97 × 10–3) cohorts but was nonsignificant in the QIMR (β = –0.09, P = 0.26) and the US (β = 0.07, P = 0.60) cohorts (see Supplementary Table S7).
Notably, both cohorts (RS and TwinsUK) that showed significant association consist of older individuals, whereas the two datasets not showing significant association (QIMR and US) consist of adolescents. This suggests that the eyebrow color effect of C10orf11 may be age dependent, which warrants further investigation in future studies. The light eyebrow color-associated G-allele had a relatively low frequency in Europeans (f = 0.02) but was more frequent in Asians with darker eyebrows (see Supplementary Figure S4 online), potentially explained by different linkage disequilibrium structures between these populations. Previous studies suggest that C10orf11 is an effective melanocyte differentiation gene, which is known to cause the oculocutaneous albinism (i.e., OCA) 7 phenotype via a rare nonsense mutation, c.580C>T (p.Arg194*, rs587776952) (
), seven SNPs showed significant association after the correction for multiple testing (adjusted P < 4.67 × 10–4), including 15q13.1 HERC2 rs12913832 (P = 1.12 × 10–48), 16q24.3 MC1R rs1805007 (P = 1.25 × 10–47), 14q32.12 SLC24A4 rs17184180 (P = 1.67 × 10–26), 20q11.22 ASIP rs6059655 (P = 4.60 × 10–10), 5p13.2 SLC45A2 rs16891982 (P = 2.60 × 10–8), 6p25.3 IRF4 rs12203592 (P = 3.47 × 10–6), and 1q32.1 DSTYK rs2369633 (P = 5.03 × 10–5) (see Supplementary Table S8 online). The first six SNPs all have known effects on human pigmentation traits. The last SNP was only recently identified in association with hair color (
In a subset of the RS cohort (n = 1,656), we compared the eight top associated SNPs at the seven genetic loci that were highlighted with significant eyebrow color association in our GWAS and candidate gene study (see Supplementary Table S9 online). In general, the contributions of the eight SNPs to scalp hair color variation were slightly larger than their impact on eyebrow color variation (see Supplementary Figure S6 online). MC1R rs1805007 showed a much larger contribution to scalp hair color than eyebrow color, likely explained by an aging effect on red eyebrow color, because the scalp hair color information in the RS cohort was ascertained from a questionnaire item on “hair color when young.” SLC45A2 rs16891982 was the only DNA variant of the eight tested that showed a slightly larger contribution to eyebrow color than to scalp hair color. These results suggest that the prediction accuracy for eyebrow color should be at a similar level to that for hair color in the same sample set under equal phenotype accuracy.
An eyebrow color prediction model was trained in 3,114 RS participants and validated in 779 independent RS participants not included in the GWAS. Red eyebrow color was excluded from the prediction analysis because none of these individuals had the phenotype. A model including 25 SNPs achieved prediction accuracies expressed as area under the curve (AUC) of 0.701 (95% confidence interval [CI] = 0.621–0.781) for blond, 0.620 (95% CI = 0.576–0.658) for brown, and 0.674 (95% CI = 0.633–0.709) for black eyebrows (Figure 2a and b, and see Supplementary Figure S7 and Supplementary Tables S10 and S11 online). The AUC values reported here for eyebrow color are lower than those previously reported for scalp hair color with the 22-SNP HIrisPlex model, which ranged between 0.75 and 0.92 for the four scalp hair color categories used (
). This discrepancy can be explained by our data set lacking four important rare MC1R SNPs (which are used in the HIrisPlex model) and an aging effect that decreases phenotype quality, particularly for red color. With the midrange accuracy level, our 25-SNP model provided highly confident prediction results for approximately 7% of the validation set, for example, those with high (>0.80) or low (<0.20) prediction probabilities of certain eyebrow color type (Figure 2c). Applying this model to 2,504 participants from the 1000 Genomes Project showed that prediction outcomes were generally consistent with knowledge about the global distribution of eyebrow color variation (Figure 2d, and see Supplementary Figure S8 online). More detailed prediction results are provided in the Supplementary Materials.
In conclusion, this eyebrow color GWAS in Europeans (the first, to our knowledge) highlighted six genome-wide significant genetic loci harboring six well-known pigmentation genes (ASIP, HERC2, MC1R, OCA2, SLC24A4, SLC45A2) and a gene to our knowledge previously unreported (C10orf11). The finding at C10orf11 warrants further investigations in European individuals with different age distributions. A candidate gene study suggested the involvement of two additional known pigmentation genes, DSTYK and IRF4, in human eyebrow color. This DNA-based eyebrow color prediction model is useful in future forensic applications.
Conflict of Interest
The authors state no conflict of interest.
This work has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement no. 740580 (VISAGE Project). FL is additionally supported by the National Key R&D Program of China (2017YFC0803501), National Natural Science Foundation of China (91651507), and The Thousand Talents Plan for Young Professionals. MK is additionally supported by Erasmus University Medical Center, Rotterdam.
The generation of the GWAS data sets of the Rotterdam Study was supported by the Netherlands Organisation of Scientific Research NOW Investments (nos. 175.010.2005.011, 911-03-012), the Genetic Laboratory of the Department of Internal Medicine, Erasmus University Medical Center, the Research Institute for Diseases in the Elderly (014-93-015; RIDE2), and the Netherlands Genomics Initiative/Netherlands Organisation for Scientific Research Netherlands Consortium for Healthy Aging (project no. 050-060-810). The Rotterdam Study is funded by Erasmus Medical Center and Erasmus University, Rotterdam; Netherlands Organization for the Health Research and Development; the Research Institute for Diseases in the Elderly; the Ministry of Education, Culture and Science; the Ministry for Health, Welfare and Sports; the European Commission (DG XII); and the Municipality of Rotterdam.
The TwinsUK study is funded by the Wellcome Trust, Medical Research Council, European Union (FP7/2007-2013), National Institute for Health Research-funded BioResource, Clinical Research Facility, and Biomedical Research Centre based at Guy’s and St Thomas’ NHS Foundation Trust, in partnership with King’s College London.
The Queensland Institute of Medical Research gratefully acknowledges the participants and their families. We would like to thank Kerrie McAloney for data collection and Scott Gordon for data management. We acknowledge funding by Australian National Health and Medical Research Council grants 241944, 339462, 389927, 389875, 389891, 389892, 389938, 442915, 442981, 496739, 552485, and 552498 and Australian Research Council grants A7960034, A79906588, A79801419, DP0770096, DP0212016, and DP0343921 for building and maintaining the adolescent twin family resource through which samples were collected. SEM is supported by National Health and Medical Research Council fellowship APP1103623.
All information and work pertaining to the US cohort was generated under funding by the US National Institute of Justice (grant no. 2014-DN-BX-K031). We thank the National Institute of Justice and all individuals who participated in this collection.