Rare missense variants in the SH3 domain of PSTPIP1 are associated with hidradenitis suppurativa

Abstract

Hidradenitis suppurativa (HS) is a chronic, debilitating skin disease for which few treatment options are available. While most HS is sporadic, some rare kindred show a high-penetrance, autosomal-dominant inheritance. We wanted to identify rare variants that could contribute to HS risk in sporadic cases using candidate gene sequencing. We ultimately identified 21 genes for our capture panel. We included genes of the γ-secretase complex (n = 6) because rare variants in these genes sometimes cause familial HS. We added Notch receptor and ligand genes (n = 13) because γ-secretase is critical for processing Notch receptor signaling. Clinically, some people with PAPA (pyogenic arthritis, pyoderma gangrenosum, and acne) syndrome, a rare inflammatory disease, have concurrent HS. Rare variants in PSTPIP1 are known to cause PAPA syndrome, so we included PSTPIP1 and PSTPIP2 in the capture panel. We screened 117 individuals with HS for rare variations and calculated the expected burden using Genome Aggregation Database (gnomAD) allele frequencies. We discovered two pathogenic loss-of-function variants in NCSTN. This class of NCSTN variant can cause familial HS. There was no increased burden of rare variations in any γ-secretase complex gene. We did find that individuals with HS had a significantly increased number of rare missense variants in the SH3 domain of PSTPIP1. This finding, therefore, implicates PSTPIP1 variation in sporadic HS and further supports dysregulated immunity in HS. Our data also suggests that population-scale HS genetic research will yield valuable insights into disease pathology.

Keywords: PSTPIP1; gamma-secretase; genetics; hidradenitis suppurativa; nicastrin; rare variants.

Figure 1

Ancestry categorization of the patient cohort Data for simulated controls (1,000 individuals/population) and our cohort (n = 117). (A) The PC1/PC2 control plot shows the separation between African and non-African ancestries. (B) The other ancestries separate more from each other on PC1 and PC3, though there is still substantial overlap. (C) PCA-informed UMAP projections of controls have better spatial separation between ancestries than the PCA alone. (D) The PC1/PC2 projection with our cohort overlaid (black dots). Our samples correspond well to the gnomAD populations. (E) The same projection with our samples alone, colored by assigned ancestry. (F) PCA-informed UMAP of controls when our samples are included. (G) The same projection with our cohort alone, colored by ancestry assignment. Ancestries: African/African American, purple; admixed American/Latino, orange; East Asian, blue; non-Finnish European, red; South Asian, green. PC, principal component; UMAP, uniform manifold approximation and projection.

Figure 2

Experimental design for control simulation Shown is a diagram of the experimental design for simulating controls. We first had to filter ExAC and gnomAD sites to only regions included in our targeted capture. Once the sites were filtered, we simulated training data to predict population ancestries for our cohort. We then simulated control rare variants for a control population with ancestries matched to our cohort, counting the total number of observed alternate alleles per gene. The process was repeated a total of 200 times to estimate the number of variants to expect on average for each gene.

Figure 3

Variant enrichment burden test visualization (A) The simulation results for the expected number of rare variants per gene that alter amino acids. Each black dot represents the number of total minor alleles in a single simulation. The actual number of rare variants observed in our cohort is overlaid as a red dot. (B) This panel has the same layout but is for synonymous variant burden, and our cohort observations are in green. (C and D) Panels (C) and (D) show the burden results for protein-affecting variants and synonymous variants, respectively. Each dot is the result for one gene. The x-axis is the log2 fold change of our observation versus the mean expected number of variants from simulation, and the y-axis shows the -log10 of the Bonferroni corrected p value for the burden. Dashed vertical lines are a 1.5-fold change. The horizontal dashed line indicates a 0.05 cutoff. In both plots, most of our observations closely match the simulation results and are not significant. Only the PSTPIP1 SH3 domain shows a significantly increased number of rare variants.

Figure 4

Localization of PSTPIP1 rare variants (A) A lollipops plot of the PSTPIP1 protein. While two variants are near the mid-point of the protein, the majority of identified variants are in the SH3 domain. (B) Pairwise alignment of the amino acid sequences of PSTPIP1 in mice and humans shows the terminal portion of the protein shares a high identity. All four missense variants are in amino acids that are conserved in mice. (C) The SH3 domain crystal structure shows the position of the missense variants in three dimensions. The folding of the protein puts the variants in closer physical proximity than one might expect from the linear sequence. They also appear to be surface exposed and could alter the binding of this domain to different targets.

Figure 5

Western blots of PSTPIP1 wild-type and variant interactions with known binding partners Western blot of binding partners immunoprecipitated by FLAG-tagged PSTPIP1. The variants tested included wild-type (WT) PSTPIP1 and a construct missing the entire SH3 domain (no SH3) as a negative control for SH3 domain interactions. The other 4 SH3 variants (T371I, A382T, G403E, and R405C) were detected in our HS cohort and generated with site-directed mutagenesis. The anti-FLAG blot (top panel) shows that the PSTPIP1 protein was expressed and that the SH3 domain deletion migrated faster since it was smaller. The anti-WASp blot (middle panel) showed intact binding between PSTPIP1 and WASp in WT, T371I, A382T, and G403E variants. The A382T and G403E variants may have slightly enhanced binding. The SH3 domain deletion and R405C failed to bind to WASp. The anti-PTP-PEST blot (bottom panel) confirmed that the interaction between PTP-PEST and PSTPIP1 is SH3-domain independent. All PSTPIP1 variants, including the SH3 domain deletion, are bound normally to PTP-PEST.