Multicytokine-based transcriptome-wide association study for 7 inflammatory skin disorders identifies candidate causal genes in keratinocytes

Background Transcriptome-wide association studies (TWAS) identify genetically regulated expression (GReX) components and can pinpoint causal genes in genome-wide association studies but are often limited by a single-cell context. Objective We hypothesized that modeling GReX across multiple conditions could enhance power to identify causal genes for complex inflammatory diseases. Methods We conducted TWAS on 400 transcriptomes under 8 proinflammatory cytokine stimulations in keratinocytes, modeling GReX for 18,599 genes against genome-wide association studies from 7 inflammatory skin diseases: atopic dermatitis, psoriasis, acne, alopecia areata, systemic sclerosis, systemic lupus erythematosus, and vitiligo. Results Our TWAS identified 274 loci from the 7 diseases that harbor a single significant TWAS gene association. We nominated causal genes and their associated proinflammatory cytokine stimuli, including ERAP2 for psoriasis with IL-17A + TNF-α stimulation; WNT10A for acne with IFN-γ stimulation; and RAET1L, MAP3K11, and ITGAM for alopecia areata, acne, and systemic lupus erythematosus, respectively, with TNF stimulation. Notably, our TWAS-identified genes showed overwhelming evidence of colocalization with genome-wide association study signals ( P = 1.03 × 10 −15 ), and our method successfully captured >85% of all genes with colocalizing expression quantitative trait loci. Single-cell–resolution spatial profiling further demonstrated the modulation of TWAS signals in keratinocytes by close proximity to TNF/IL-17–expressing cells in psoriatic skin. Conclusion Modeling gene expression across relevant cellular states substantially improves the power and resolution of TWAS.