Genomic loci and molecular genetic mechanisms for hidradenitis suppurativa

Abstract

Background: Hidradenitis suppurativa (HS) is a common, chronic, and debilitating inflammatory disease that most commonly affects intertriginous skin. Despite its high heritability, the genetic underpinnings of HS remain poorly understood.

Objective: To identify genetic signals associated with HS, determine genetic relationships with other diseases, and investigate potential molecular genetic mechanisms.

Methods: We performed a genome-wide association study meta-analysis of six studies, totaling 4,540 cases and over 1 million controls and identified genetic correlations with other common diseases. We integrated the HS data with expression quantitative trait loci from 10 trait-relevant tissues, epigenomic and transcriptomic data from human scalp, differential expression data from HS lesions versus adjacent skin, and mesenchymal Hi-C chromatin looping data. To identify functional noncoding variants, we performed transcriptional reporter assays for signals near KLF5 and SOX9.

Results: We identified eleven significant HS signals across seven loci: four corresponded to previously reported associations, four represented novel signals within known loci, and three were signals in newly implicated loci. We identified significant genetic correlation between HS and other inflammatory conditions, particularly inflammatory bowel disease, rheumatoid arthritis, type 2 diabetes, and asthma. We prioritized candidate genes for the 11 signals. The risk allele at KLF5 exhibited 10-fold greater transcriptional activity than the non-risk allele, while risk alleles at SOX9 showed significantly reduced transcriptional activity.

Conclusions: Our results provide insights into potential genetic mechanisms underlying HS and suggest potential therapeutic targets for this challenging condition.

Figure 1:

HS genetic association signals. A: Manhattan plot of the associations between variants and HS in all populations. Variants within 1 Mb of a previously known HS locus are colored in blue; variants within 1 Mb of a new HS locus (P-value ≤ 5×10⁻⁸) are colored in red. Loci with asterisks represent multi-signal loci. B: Variant associations for new HS loci. Left: a signal near SUNCR1 was significantly associated with HS in the all-population analysis; Middle: a signal near NIN was significantly associated with HS in the EUR+AMR stratified analysis; Right: a signal near MXRA7 was significantly associated with HS in males.

Figure 2:. HS locus detected in male-only GWAS.

Left: Variants associated with HS in men (top) versus women (bottom). Right: Forest plot of male-only and female-only GWAS results. The MXRA7 locus is strong in men and shows consistent direction and strength of effect across all contributing studies. An effect is not detected in any female-only or sex-combined GWAS results. Studies noted with an asterisk (*) were not included in the meta-analysis due to small sample size.

Figure 3:. Genetic correlation between HS and other complex traits.

Green indicates Bonferroni-significant correlations. IBD: inflammatory bowel disease; T2D: type 2 diabetes; PCOS: polycystic ovarian syndrome; RA: rheumatoid arthritis.

Figure 4:. Differential expression of prioritized candidate genes in lesion vs non-lesion tissue and in scalp cell types.

We prioritized 20 genes that were within 100 kb of a lead variant, or the nearest protein-coding gene if none were within 100 kb, and genes with external evidence of relevance to HS. Top: the log₂ fold change of gene expression in HS lesion versus matched non-lesion tissue. Bottom: the log₂ fold change of gene expression in scalp cell types. Only increased expression in a cell type is shown. To be included, the cell type must be more highly expressed in at least one prioritized gene. Circles show the size of the fold change. Orange: keratinocytes; yellow: melanocytes; green: fibroblasts; purple: immune cells; brown: muscle or endothelial cells.

Figure 5:. HS loci KLF5 (left) and SOX9 (right) may influence expression levels of their target genes in keratinocytes via a regulatory element that appears to be only accessible in keratinocytes.

Top row: Variant associations with HS for the two signals. Middle left: Genome browser plots showing scalp chromatin accessibility by cell type. The lead variant of KLF5 (rs981625) is located within a keratinocyte-accessible regulatory element that was not observed in other examined cell types (chr13_73465564_73466065) and is strongly conserved across vertebrates. No other LD proxies of the lead variant are located within accessible chromatin. KLF5 was much more strongly expressed in keratinocytes than any other cell type (log₂fold change = 0.56 – 1.93 for keratinocyte cell clusters, P-values ≤ 4×10⁻¹⁶⁶). Bottom left: the rs981625-C allele, which was associated with higher HS risk, showed a 10-fold greater transcriptional activity than the rs981625-G reference allele (P-value = 2.9 × 10⁻⁷). Middle right: the SOX9 lead variant rs17226067, and its LD proxies rs17825774 and rs17825799 (r² = 1, shown in gold) are located within a region of open chromatin detected in keratinocytes but not in other surveyed cell types (chr17_71519287_71519788); no other LD proxies are located within accessible chromatin. SOX9 was also more highly expressed in hair-follicle (HF) keratinocyte clusters than other scalp cell types (log₂ fold-change = 0.28 – 5.19, P-values ≤ 5×10⁻⁸⁶). Bottom right: the 3-variant risk haplotype of rs17226067-G, rs17825774-A, and rs17825799-A alleles showed >2-fold reduced transcriptional activity versus the reference haplotype AGG haplotype (P-value = 0.001).

Figure 6:. Both conditionally distinct signals on chromosome 9 exhibit a significant Hi-C chromatin interaction peak in mesenchymal data.

A: Variants associated with HS at two conditionally distinct signals. B: Mesenchymal cell Hi-C data anchored around rs1994799, the lead variant for the primary HS signal. C: Mesenchymal cell Hi-C data anchored around rs1535620, the lead variant for the secondary HS signal. The target regions immediately flanking the anchor regions and target regions with FitHiC Q-values ≥ 0.01 are not shown. Arc colors are scaled by strength of association measured by −log₁₀(P-value), with the darkest lines showing the link with the strongest P-value. Hi-C data in the red box, centered around KLF4, showed significant interactions; no significant interactions were observed around the more proximal gene ACTL7B (gray box).