Inflammatory skin disease atlas allows disease stratification based on shared and distinctive pathogenic mechanisms

Summary: Abstract Body: Inflammatory skin diseases (ISD) represent a large proportion of dermatologic conditions. Prior studies identified dysregulated pathways compared to healthy controls (HC). However, diagnosis of ISD in dermatology is often challenging as clinicians must differentiate between multiple different conditions that may have similar appearances or distribution. Here, we generated bulk-RNA sequencing data from skin biopsies of 753 different patients across 40 different ISD to build a transcriptome atlas. Importantly, we also included five ISD from the scalp, which have not been thoroughly studied. Unsupervised clustering including cytokine and chemokine signatures identified 4 major disease clusters that divided papulosquamous (e.g. psoriasis) from fibrotic (e.g. scleroderma), autoimmune (e.g. lupus) and scalp (e.g. folliculitis decalvans). Interestingly, some samples harbored molecular signatures shared between clusters, highlighting heterogeneity within each disease. We identified shared and distinctive pathways enriched in each cluster and individual disease. Cytokine gene expressions were able to split cases versus HC in most diseases. Importantly, using specific response signatures in keratinocytes, to IL-4, IL-13, type I and II interferons, IL-36, IL-17A, TNFα, and IL-17A+TNF, we stratified ISD across their cytokine signature enrichment. Furthermore, these cytokine signatures were used to train a machine learning prediction model to classify every disease against all the other conditions. Despite unique and shared cytokine responses across many diseases, the majority showed high areas under the receiver operating characteristic curve (AUC >89%), indicating that multidimensional cytokine response signatures can distinguish ISD from each other. In summary, our transcriptome atlas of ISD represents a new resource for dermatologists that can be used for clinical diagnosis, further research, and decision making for targeted treatment based on cytokine signatures. Benjamin Klein¹, Kelsey Van Straalen¹, Joseph Kirma¹, Michelle Kahlenberg¹, Killian Eyerich², Lam C. Tsoi¹, Johann E. Gudjonsson¹ 1. University of Michigan, Ann Arbor, MI, United States. 2. Albert-Ludwigs-Universitat Freiburg, Freiburg, BW, Germany. Innate Immunity, Microbiology, and Microbiome