The polygenic architecture of hidradenitis suppurativa reveals signaling mechanisms that implicate epithelial remodeling

Abstract

We sought to identify clinically relevant regulators of hair follicle inflammation by conducting a human genetic study of hidradenitis suppurativa (HS), a prevalent, understudied, inflammatory disease with limited effective treatments. We performed a GWAS with 6,300 cases and identified 12 independent risk loci. Epigenetic and transcriptomic analyses of HS risk variants defined cell-specific gene regulatory programs. We experimentally validated a coherent gene module defined by upregulated SOX9, CXCR4, and CD74 co-expression that maps to aberrant epithelial structures in the skin. Pharmacological inhibition of CXCR4 implicates CD74 mediated regulation of PI3K/AKT and NF-κB signaling to calibrate inflammation, proliferation and apoptosis in keratinocytes. We next used genome-wide methods to interrogate shared polygenic architecture and identified new clinically and mechanistically relevant disease associations, including another condition that involves aberrant hair follicle remodeling, male pattern hair loss. Our results point towards CXCR4-CD74 signaling in HS and hair follicle homeostasis and suggest CXCR4 blockade as a new therapeutic strategy in HS.

Figure 1.. Mapping HS risk variants to cis regulatory elements and genes.

(A) Meta-analysis GWAS identified 12 independent associations including 4 previously reported loci (red) and 8 new loci (black). (B) Hierarchical clustering of HS risk variants predicted to have functional effects in Roadmap and Encode samples identifies a cluster of variants in regions that are predicted to be active in immune cells and tissues (blue) and a cluster of variants predicted to be active in epithelial (red) and mesenchymal derived or derived mesenchymal (green) cells. (C) HS risk variants fall within transcription factor (TF) bind sites that are shared across GWAS loci. (D) HS risk variants were mapped to eQTL data in GTEx. Skin is the most highly represented tissue. (E) HS risk variants were linked to 79 transcripts through eQTL or chromatin mapping data. Each data set implicates overlapping sets of genes. (F) Genes predicted to be influenced by HS risk variants were tested for differential gene expression in single cell transcriptomic data comparing HS lesional skin to normal healthy skin.

Figure 2.. CXCR4+ keratinocytes in HS skin.

(A). CXCR4 protein expression in HS skin. (B) We used single cell transcriptomic data to compare CXCR4+ keratinocytes to CXCR4- keratinocytes and found upregulated SOX9 and HLA class II genes (red). We also identified transcription factors with binding sites that contain HS risk variants (blue) and a strong signature of epithelial mesenchymal transition associated with epithelial wound responses (black). Strong co-expression of CD74 suggest that CXCR4 may be forming a heterodimer with CD74 to amplify PI3K/AKT, ERK/MAPK, and NF-KB signaling. (C) Activation of these pathways is confirmed in CXCR4+ keratinocytes. Furthermore, exposure to Plerixafor in ex vivo HS skin attenuated these pathways in CXCR4+ keratinocytes. (D) Proposed model illustrating how upregulated CXCR4 expression in HS keratinocytes disrupts the balance of inflammatory signaling pathways, and how Plerixafor treatment may restore this balance (purple). In this model, CXCR4 upregulation in HS keratinocytes promotes activation of MAPK/ERK, PI3K/Akt, and NF-κB pathways, contributing to sustained inflammation and impaired resolution. Pharmacologic inhibition of CXCR4 with Plerixafor attenuates these pathways and enhances apoptotic and gene programs associated with physiological remodeling in the hair follicle. Downstream pathway components that are modulated in CXCR4⁺ keratinocytes following Plerixafor treatment are highlighted in purple. Genes identified as GWAS-associated are marked with light blue indicators, and genes differentially expressed between HS and healthy keratinocytes are marked with green indicators.

Figure 3:. GPS testing results and PheWAS analysis in the AoU dataset. (a) GPS testing results in the AoU Release 2 dataset.

The Odds Ratio (OR) corresponds to a model adjusted for age, sex, and principal components of ancestry. The Standard Deviation (SD) represents the GPS distribution in controls. The Area Under the Curve (AUC) indicates the performance of the model adjusted for age, sex, and principal components of ancestry. The “Crude” AUC refers to GPS performance without adjusting for covariates. The variance explained is calculated for the GPS component alone, without covariates. CI: Confidence Interval. (b) PheWAS with HS GPS in the AoU dataset. The analysis includes data from 148,234 AoU participants and was performed under a dominant inheritance model using logistic regression, adjusted for age, sex, and ancestry. Effect estimates and two-sided P-values were derived from logistic regression. The red horizontal line indicates the phenome-wide significance threshold after adjusting for multiple testing (P = 2.8 × 10⁻). The Y-axis represents −log10(P-value), while the X-axis displays system-based phecode groupings. Upward-pointing triangles denote increased odds for a phecode, whereas downward-pointing triangles indicate reduced odds.

Figure 4:. Genome-wide genetic correlations of GWAS results for HS with infectious, autoimmune, cardiometabolic, and hair disorders.

(a) Genetic correlations including the HLA region. (b) Genetic correlations excluding the HLA region.

Figure 5.. Genetic evidence supports CD74-mediated regulation of inflammation in HS and male pattern hair loss (MPHL).

CD74 is critical for balancing inflammation to restore tissue homeostasis. Our genetic and experimental data support a model of aberrant sustained inflammation in HS keratinocytes driven by upregulated CXCR4 expression. CXCR4 blockade attenuated markers of aberrant inflammation and proliferation (PI3K/Akt, WNT signaling). γ-secretase processing of CD74 is upstream from genes that attenuate inflammation (e.g. TNFAIP3), instigate apoptosis (e.g., BAK1), and are implicated in epithelial activation and downgrowth in hair follicle development and cycling (CD74, TNFAIP3, SOX9). Our model suggests that loss of function variants in γ-secretase genes (NCSTN, PSENEN) would have the same effect as CXCR4 overexpression – escalating inflammation and preventing its resolution. The negative genetic correlation we identified between HS and MPHL further suggests that MPHL may result from an imbalance of this signaling axis in the opposite direction. Insufficient inflammation and increased apoptosis would result in attenuated epithelial downgrowth leading to hair follicle miniaturization.