In total, genotype and hair color data were available for 7091 subjects from the NTR. We performed five case-control GWAS with logistic regression for all SNPs including age, sex, the three Dutch PCs and genotype platform as covariates. Familial structure was taken into account in the analysis using Plink and selecting a single monozygotic twin. The resulting Q–Q and Manhattan plots for all colors and light versus dark hair color are shown in Figures S1–S10. As shown with GCTA and with the twin heritability modeling, the PC’s are significantly related to hair color in the Netherlands along the three Dutch major axes of genetic variation (Figures S11 and S12). Using the PCs, we corrected for this population stratification, with post correction GWAS λs ranging between 1.004 and 1.027. However, as a consequence, we have likely also reduced the significance of SNPs, which are truly associated with hair color.
3.4.1. Known Hair Color Variants in Relation to the GWAS Results
Table 5 and
Table 6 display the gene variants for hair color, which are known from previous studies and our most significant SNPs within these genes. The two genes
HERC2 (15q11.2-13), along with neighboring gene
OCA2, are known as the most essential genes for determining human pigmentation traits including eye, hair and skin color [
15,
32]. These genes also show strong signals on chromosome 15 in our study, with SNPs rs7495174 and rs79097182 for blond, brown and light hair color.
The solute carrier (SLC) gene family group is a large family gene group that consists of 458 genes in 52 families. Three loci have been found to be associated with human pigmentation:
SLC24A5,
SLC45A2 and
SLC24A4 gene. Interactions between
HERC2 and
SLC24A4 play a role in determining blue eye color, but also light hair color and less tanning ability [
32,
33].
SLC24A4 (14q32.12) encodes the sodium/potassium/calcium exchanger 4 protein (
NCKX4). Alternative splicing of this gene results in multiple transcript variants. Variants in
SLC24A4 have been previously associated with eye and hair color, skin sensitivity to sun and coetaneous malignant melanoma [
34]. We confirmed the associations for hair color. Within
SLC24A4, rs8014907 was significantly associated with all hair colors, except red. The
SLC45A2 gene, which is in the same family as
SLC24A4, encodes a transporter protein that mediates melanin synthesis. The protein of
SLC45A2 is expressed in a high percentage of melanoma cell lines. Mutations in this gene are a cause of oculocutaneous albinism type 4, and polymorphisms in this gene are associated with variations in skin and hair color [
35,
36]. Multiple transcript variants encoding different isoforms have been found for this gene. Our results do show a
p-value = 9.1 × 10
−5 for the gene.
SLC24A5 was found to be involved in skin pigmentation in European populations [
33]. A 4-bp insertion (c.569_572insATTA rs1426654) in the
SLC24A5 gene, causing a frame shift and premature termination, was identified in a man with Indian ancestry [
37]. Homozygosity of this insertion results in extreme hypopigmentation and pinkish-white skin, with dark brown hair and a brown iris [
35]. However, we have not found any significant hits in this locus. For
SLC24A5 gene the lowest
p-value is 0.7 and the question is whether we have people with this particular insertion present within our Dutch population.
MC1R (16q24.3) is an intron-less gene of the size less than 1 kb. Non-synonymous variants in
MC1R are present in approximately 50% in the European population [
38]. Its multiple variants were first found to be associated with human red hair color in 1995 [
39]. A subsequent study found these variants to have the same effect on pigmentation at increased frequency with increasing latitude in humans [
40,
41]. We replicate this association in our study, as
MC1R is associated with multiple hair colors, except blond.
KITLG (12q22) is known to regulate the number of melanocytes during development, melanin distribution and hyper/hypo pigmentation. Sequence variation is thought to affect expression of
KITLG (184,745), which results in the blond hair color. In European populations rs12821256 T/C SNP is found to be associated with blond hair color [
42,
43], and this SNP explains 3%–6% of the variance [
10]. Our study confirmed this result, rs12821256 showed a significant associations with blond, brown and light
versus dark hair color. Recently, a functional study showed that this SNP alters a transcription factor binding site for lymphoid enhancer-binding factor 1 (LEF1), reducing
LEF1 responsiveness and enhancer activity in cultured human keratinocytes [
44].
IRF4 (6p25.3) is associated with hair color and skin pigmentation [
42]. There is a strong association of the A allele of a single-nucleotide polymorphism (SNP) on chromosome 6p25.3, rs1540771, with the presence of freckles in Icelandic and Dutch population samples (discovery OR = 1.40,
p-value = 3.7 × 10
−18) [
45]. In our study, the most significant SNP rs62389424 is a bit further away (34 kb) and is associated with all hair colors except red (
p < 1.3 × 10
−5).
Table 3.
The explained genetic variance of the hair color liability scale for autosomal common SNPs in GCTA.
Table 3.
The explained genetic variance of the hair color liability scale for autosomal common SNPs in GCTA.
Phenotypes | N Cases | N controls | Proportion of Genetic Variance Explained by All Common SNPs (SE) | p | Top Chromosome | Proportion of Genetic Variance Explained by Top Chromosome (SE) | p | Proportion of Genetic Variance Explained by Known Associated SNPs (SE) | p |
---|
Blond | 1547 | 1793 | 0.165 (0.081) | 1.1E−02 | 15 | 0.014 (0.017) | 1.8E−01 | 0.058 (0.022) | 5.75E−40 |
Brown | 1946 | 1394 | 0.095 (0.079) | 9.3E−02 | 15 | 0.011 (0.016) | 2.5E−01 | 0.059 (0.022) | 7.88E−39 |
Red | 87 | 3253 | 0.246 (0.087) | 1.9E−03 | 16 | 0.163 (0.025) | 3.2E−14 | 0.069 (0.026) | 2.25E−55 |
Black | 66 | 3274 | <0.001 (0.083) | 5.0E−01 | 6 | 0.031 (0.228) | 7.7E−02 | 0.005 (0.003) | 1.2E−02 |
Light versus dark | 1890 | 1450 | 0.140 (0.080) | 2.7E−02 | 15 | 0.014 (0.017) | 2.0E−01 | 0.069 (0.026) | 6.50E−46 |
Table 4.
The explained genetic variance of the hair color liability scale for autosomal common SNPs in GCTA including the three Dutch PCs as covariates.
Table 4.
The explained genetic variance of the hair color liability scale for autosomal common SNPs in GCTA including the three Dutch PCs as covariates.
Phenotypes | N Cases | N Controls | Proportion of Genetic Variance Explained by All Common SNPs (SE) | p | Top Chromosome | Proportion of Genetic Variance Explained by Top Chromosome (SE) | p | Proportion of Genetic Variance Explained by Known Associated SNPs (SE) | p |
---|
Blond | 1547 | 1793 | <0.001 (0.084) | 0.5 | 15 | 0.001 (0.015) | 0.48 | 0.054 (0.024) | 2.81E−37 |
Brown | 1946 | 1394 | <0.001 (0.082) | 0.5 | 15 | <0.001 (0.016) | 0.5 | 0.054 (0.021) | 6.95E−39 |
Red | 87 | 3253 | 0.165 (0.025) | 1.39E−03 | 16 | 0.255 (0.083) | 2.36E−14 | 0.053 (0.020) | 1.65E−55 |
Black | 66 | 3274 | <0.001 (0.084) | 0.5 | 6 | 0.027 (0.228) | 0.10 | 0.005 (0.004) | 2.4E−02 |
Light versus dark | 1890 | 1450 | <0.001 (0.084) | 0.5 | 15 | <0.001 (0.016) | 0.5 | 0.065(0.024) | 1.53E−43 |
Table 5.
SNP associations within our study for blond and brown hair against known hair color loci.
Table 5.
SNP associations within our study for blond and brown hair against known hair color loci.
Locus | Chromosome Location | Most Significant SNP | MAF | OR Blond | SE Blond | p Blond | OR Brown | SE Brown | P Brown |
---|
SLC45A2 | 5p13.2 | rs16891982 * | 0.050 | 0.5036 | 0.1752 | 9.1E−05 | 1.4506 | 0.1656 | 0.02472 |
IRF4 | 6p25.3 | rs1540771 | 0.489 | 1.0253 | 0.1978 | 0.003 | 0.9027 | 0.0396 | 9.7E−03 |
IRF4 | 6p25.3 | rs62389424 * | 0.087 | 2.4846 | 0.1188 | 1.3E−13 | 0.5551 | 0.1081 | 5.2E−08 |
TYRP1 | 9p23 | rs1408799 * | 0.291 | 0.8522 | 0.0484 | 0.001 | 1.1846 | 0.0468 | 2.9E−03 |
TPCN2 | 11q13.3 | rs72930659 | 0.109 | 0.6073 | 0.0657 | 3.2E−14 | 1.5544 | 0.0649 | 1.1E−11 |
KITLG | 12q21.33 | rs12821256 | 0.128 | 0.6715 | 0.0616 | 9.8E−11 | 1.4366 | 0.0596 | 1.2E−09 |
SLC24A4 | 14q32 | rs8014907 | 0.178 | 1.4317 | 0.0554 | 3.0E−10 | 0.7262 | 0.0529 | 3.1E−09 |
SLC24A5 | 15q21.1 | rs1834640 | 0.110 | 1.0163 | 0.1323 | 0.903 | 0.9858 | 0.1294 | 0.9121 |
HERC2 | 15q13 | rs79097182 | 0.038 | 2.9822 | 0.1426 | 1.8E−14 | 0.4827 | 0.1235 | 3.7E−09 |
OCA2 | 15q13.1 | rs7495174 * | 0.014 | 3.5485 | 0.2577 | 8.9E−07 | 0.3544 | 0.2168 | 1.7E−06 |
MC1R | 16q24 | rs2353688 | 0.028 | 1.8309 | 0.1432 | 2.4E−05 | 0.6028 | 0.1316 | 1.2E−04 |
MC1R | 16q24 | rs146972365 | 0.053 | 0.9765 | 0.0945 | 0.802 | 1.7771 | 0.0922 | 4.5E−10 |
MC1R | 16q24 | rs8063160 | 0.065 | 0.9119 | 0.0872 | 0.290 | 1.7837 | 0.0847 | 8.6E−12 |
MC1R | 16q24 | rs117322171 | 0.014 | 0.9469 | 0.1795 | 0.761 | 0.9039 | 0.1789 | 0.572 |
ASIP | 20q11.22 | rs1015362 | 0.273 | 1.0448 | 0.0464 | 0.344 | 0.9559 | 0.0452 | 0.3181 |
ASIP | 20q11.22 | rs4911414 | 0.347 | 0.9922 | 0.0428 | 0.855 | 1.0148 | 0.0417 | 0.724 |
Table 6.
SNP associations within our study for light
versus dark, red and black hair against known hair color loci (same loci as
Table 5).
Table 6.
SNP associations within our study for light versus dark, red and black hair against known hair color loci (same loci as Table 5).
Most Significant SNP | OR Red | SE Red | p Red | OR Black | SE Black | p Black | OR Light versus Dark | SE Light versus Dark | p Light versus Dark |
---|
rs16891982 * | 0.8443 | 0.6410 | 0.792 | 4.9177 | 0.3258 | 1.0E−06 | 1.9867 | 0.1764 | 1.0E−04 |
rs1540771 | 0.6853 | 0.367 | 0.010 | 0.8300 | 0.1112 | 0.104 | 0.6822 | 0.307 | 0.002 |
rs62389424 * | 0.6897 | 0.3084 | 0.228 | 0.3985 | 0.2107 | 1.3E−05 | 0.4424 | 0.1171 | 3.4E−12 |
rs1408799 * | 0.9043 | 0.1467 | 0.493 | 0.9806 | 0.1355 | 0.885 | 1.1823 | 0.0478 | 4.6E−04 |
rs72930659 | 1.2327 | 0.2211 | 0.344 | 1.3688 | 0.1968 | 0.1106 | 1.6087 | 0.0653 | 3.4E−13 |
rs12821256 | 1.0088 | 0.1951 | 0.964 | 1.3023 | 0.1895 | 0.1634 | 1.4829 | 0.0612 | 1.2E−10 |
rs8014907 | 1.0271 | 0.1788 | 0.709 | 0.8070 | 0.1453 | 0.1400 | 0.7028 | 0.0547 | 2.6E−10 |
rs1834640 | 1.1650 | 0.4093 | 0.831 | 0.8708 | 0.3677 | 0.7067 | 0.9691 | 0.1315 | 0.811 |
rs79097182 | 0.5954 | 0.3135 | 0.098 | 0.4352 | 0.2254 | 2.2E−04 | 0.3900 | 0.1386 | 5.2E−14 |
rs7495174 * | 0.6939 | 0.5975 | 1.7E−06 | 0.9691 | 0.3839 | 0.9349 | 0.3216 | 0.2443 | 3.4E−06 |
rs2353688 | 0.5149 | 0.3862 | 0.086 | 1.4452 | 0.4024 | 0.3601 | 0.6148 | 0.1348 | 3.1E−04 |
rs146972365 | 0.0720 | 0.1674 | 1.1E−55 | 1.8403 | 0.3244 | 0.6007 | 1.8751 | 0.0924 | 1.0E−11 |
rs8063160 | 0.0770 | 0.1728 | 7.8E−50 | 1.9278 | 0.3041 | 0.3090 | 1.8891 | 0.0845 | 5.2E−14 |
rs117322171 | 1.7219 | 0.6122 | 0.375 | 452.3873 | 1.0414 | 4.3E−09 | 1.0119 | 0.1771 | 0.947 |
rs1015362 | 0.9583 | 0.1500 | 0.776 | 1.0513 | 0.1282 | 0.6967 | 0.9619 | 0.0464 | 0.402 |
rs4911414 | 0.7644 | 0.1371 | 0.725 | 1.1988 | 0.1200 | 0.1309 | 1.0366 | 0.0426 | 0.399 |
Genetic variants in 3-prime-untranslated region of the ASIP result in skin/hair/eye pigmentation variation. ‘The
ASIP haplotype’, rs1015362G and rs4911414T was shown to be associated with red hair color, freckling, and skin sensitivity to sun, in addition to burning and freckling that reached genome wide significance (max odds ratio = 1.60,
p-value = 3.9 × 10
−9) in the Icelandic and Dutch populations [
45]. In our study, however, we did not find any significant results for the
ASIP locus. Minimum
p-values for these genes were above 0.35 for the hair colors. Note that red hair color in our sample is not so frequent, so this could be related to detection power.
TPCN2 (11q13.3.) was found to be significantly associated with blond
versus brown hair color in Icelandic Europeans [
42]. In our study SNPs in this locus are significantly associated with blond, brown, black and light
versus dark hair colors (
p < 4.7 × 10
−10).
The variants of
TYRP1 gene are known skin/hair/eye pigmentation variation locus. A suggestive association for blond
versus brown hair was observed for rs1408799 in Iceland and Dutch populations, and functional data suggest that the
TYRP1 gene encodes a melanosomal enzyme with a role in the eumelanin pathway [
45]. The
p-values in our study are about 10
−4 implying a potential, but weak association with hair color.