A genome-wide association scan in admixed Latin Americans identifies loci influencing facial and scalp hair features

Abstract

We report a genome-wide association scan in over 6,000 Latin Americans for features of scalp hair (shape, colour, greying, balding) and facial hair (beard thickness, monobrow, eyebrow thickness). We found 18 signals of association reaching genome-wide significance (P values 5 × 10(-8) to 3 × 10(-119)), including 10 novel associations. These include novel loci for scalp hair shape and balding, and the first reported loci for hair greying, monobrow, eyebrow and beard thickness. A newly identified locus influencing hair shape includes a Q30R substitution in the Protease Serine S1 family member 53 (PRSS53). We demonstrate that this enzyme is highly expressed in the hair follicle, especially the inner root sheath, and that the Q30R substitution affects enzyme processing and secretion. The genome regions associated with hair features are enriched for signals of selection, consistent with proposals regarding the evolution of human hair.

Figure 1. GWAS results overview.

At the top are shown drawings illustrating the seven hair features examined in the CANDELA study sample. Thick lines connect these features with the candidate genes identified in regions with SNPs reaching genome-wide significant association (Table 1). At the bottom is shown a composite Manhattan plot displaying all significantly associated SNPs for the hair features examined. The rs number of the SNP with the smallest P value is shown at the top of each association peak (Table 1 index SNP). Composite panels in this and subsequent figures were made using Photoshop.

Figure 2. Effect sizes for the derived allele at index SNPs (Table 1) in ten genomic regions not previously associated with hair traits.

(a) 10p14 hair shape, (b) 16p11 hair shape, (c) 2q12 beard thickness, (d) 4q12 bear thickness, (e) 6q12 beard thickness (f) 7q31 beard thickness (g) 3q22 eyebrow thickness, (h) 2q36 monobrow, (i) 6p25 hair greying, (j) 10q22 balding. Blue boxes represent regression coefficients (x axis) estimated in each country. Red boxes represent effect sizes estimated in the combined meta-analysis. Blue box sizes are proportional to sample size. Horizontal bars indicate a 95% confidence interval of width equal to 2 × standard errors. Meta-analysis P values are shown in Supplementary Table 4. Similar plots for regions previously associated with hair traits that were replicated here are shown in Supplementary Fig. 4.

Figure 3. Association plots for six regions with SNPs showing genome-wide significant association to hair traits.

(a) 2q12 hair shape, (b) 2q12 beard thickness, (c) 16p11 hair shape, (d) 3q22 eyebrow thickness, (e) 2q36 monobrow, (f) 6q25 hair greying. The index SNP in each region (Table 1) is shown as a purple diamond. At the top of the figure are shown the association results (on a -log₁₀ P scale; left y axis) for all genotyped and imputed SNPs. The dot colour indicates the strength of LD (r²) between the index SNP and each SNP (based on the 1000genomes AMR data set). Recombination rate across the region, in the AMR data, is shown as a continuous blue line (scale on the right y axis). Genes in the region are shown in the middle. These plots were produced using LocusZoom. Below each LocusZoom plot we show an LD heatmap (using r², red indicating r²=1 and white indicating r²=0) produced using Haploview. Coordinates used are from human genome sequence build 37. Plots for regions not shown here are presented in Supplementary Fig. 8.

Figure 4. EDAR effects on mouse facial hair follicle density and expression of PRSS53 in anagen (growing) human hair follicles.

(a) Frontal photographs showing part of the lower facial region from 14.5-day-old mouse embryos stained by in-situ hybridization for detecting Sostdc1 to reveal hair placodes (primordia of hair follicles) as blue foci. We compared placode density in the lower jaw of wild-type (+/+) mice with an Edar transgenic having a high copy number of Edar (Edar^Tg951/Tg951). The black scale bar equals 0.5 mm. (b) Bar plot comparing placode density in mice with different Edar genotypes (n=4). Mean density in Edar^+/+ mice was 53 placodes per mm² (standard deviation=1.3) and 35 placodes per mm² (standard deviation=1.5) in Edar^Tg951/Tg951 mice, the difference between means being significant (exact P value of 0.028). Error bars represent ±3 standard deviations. (c–f) Anagen human hair follicle stained with anti-PRSS53 antibody (green) and with anti-melanocyte antibody (red), and counterstained with 4,6-diamidino-2-phenylindole (DAPI; blue, nuclei). (c) Hair follicle bulb showing PRSS53 expression in the developing IRS, pre-cortex and in some melanocytes (arrow and inset) as indicated by the yellow–orange staining. (d) Mid hair follicle showing expression of PRSS53 in maturing IRS keratinocytes (arrows). (e) Distal hair follicle showing high expression of PRSS53 in IRS cells around the level of DNA degradation in hair fibre (HF) keratinocytes (as indicated by a reduction in DAPI staining in this region HF). PRSS53 is also expressed in the IRS companion layer (CL; *). (f) Upper distal hair follicle at sebaceous gland level (Sg) showing PRSS53 expression in scattered peri-follicular cells just below the Sg and in the IRS at the point of its dissolution (arrows). (g–j) Anagen human hair follicle stained with anti-PRSS53 antibody (green) and with anti-TCHH antibody (red), and counter-stained with DAPI (blue, nuclei). (g) Hair follicle bulb showing co-localization (orange/yellow) of PRSS53 and TCHH in the developing IRS, especially in the most external IRS layer. (h) Supra-bulbar region of the hair follicle showing expression of PRSS53 in the developing companion layer (*) of the IRS (green) and some co-localization with TCHH in the inner IRS. (i) Mid hair follicle showing expression of PRSS53 in the IRS companion layer (*) and the central medulla (Md) of the developing HF. (j) Upper hair follicle showing PRSS53 expression in the companion layer of the IRS (*) and in the TCHH-positive IRS. Co, hair fibre cortex; FP, follicular papilla; IRS, inner root sheath; Md, medulla; pCo, pre-cortex; ORS, outer root sheath; Sg, sebaceous gland. Grey scale bars correspond to 40 μm in each figure.

Figure 5. Processing of PRSS53 and signals of selection in the PRSS53 gene region.

Comparison of PRSS53 and PRSS53 (Q30R) from cell extracts (a) and media (b), after expression in 293-EBNA cells cultured in the absence (−) or presence (+) of DECA (decanoyl-RVKR-CMK, a pro-protein convertase inhibitor). (a) The top two bands seen in lanes loaded with PRSS53 cell extracts, result from partial processing of the signal peptide. In the absence of DECA, there appears to be an accumulation of PRSS53 (Q30R) without signal peptide. PRSS53 (Q30R) also has a slightly faster mobility compared with PRSS53 (a difference of eight amino acids based on the location of the pro-protein convertase site Supplementary Fig. 9). In the presence of DECA, both forms of the enzyme appear identically processed and there is no difference in mobility of the proteins. (b) The medium from cell cultures grown in the absence of DECA shows a reduced amount of PRSS53 (Q30R), compared with PRSS53. In the presence of DECA, the amount of protein in the media is similar for the two forms of the enzyme. Recombinant proteins were detected using an anti-FLAG antibody after migration on 13% SDS–polyacrylamide gel electrophoresis gels. Molecular weight markers (kDa) are indicated on the left. Beta-actin was used as a loading control (C=cells transfected with an empty vector). Full immunoblots for a and b are shown in Supplementary Fig. 11. (c) Immunostaining of 293-EBNA cells expressing PRSS53 and PRSS53 (Q30R). The enzyme was detected using an anti-FLAG antibody and the endoplasmic reticulum (ER) stained with an anti-calreticulin antibody. There is more abundant co-localization of the enzyme with the ER (white arrows) for PRSS53 (Q30R) compared with PRSS53, consistent with the intracellular accumulation of PRSS53 (Q30R). Negative controls, lacking primary antibodies for FLAG and calreticulin, are also shown. The scale bar indicates 10 μm. Magnification was the same for all photographs. (d) The top panel shows CMS scores (red dots) in the 1000 genomes ASN (JPT+CHB) data for SNPs in the 16p11 region associated with hair shape. The dotted line indicates the empirical significance threshold (1%). The SNP with the highest CMS score is rs11150606, coding for the Q30R substitution in PRSS53 (highlighted in purple) associated with hair curliness (Table 1 and Fig. 3c). Genes in the region are shown in the panel below (introns in magenta, exons in green). Coordinates are from human genome sequence Build 37.