Immune modules to guide diagnosis and personalized treatment of inflammatory skin diseases

Abstract

Previous advances have identified immune pathways associated with inflammatory skin diseases, leading to the development of targeted therapies. However, there is a lack of molecular approaches that delineate these pathways at the individual patient level for personalized diagnostic and therapeutic guidance. Here, we conduct a cross-comparison of expression profiles from multiple inflammatory skin diseases to identify gene modules defining relevant immune pathways. Seven modules are identified, representing key immune pathways: Th17, Th2, Th1, Type I IFNs, neutrophilic, macrophagic, and eosinophilic. These modules allow the development of a molecular map with high diagnostic efficacy for inflammatory skin diseases and clinico-pathologically undetermined cases. Aligning dominant modules with treatment targets offers a rational framework for treatment selection, improving response rates in both treatment-naïve patients and non-responders to targeted therapies. Overall, our approach offers precision medicine for inflammatory skin diseases, utilizing transcriptional modules to support diagnosis and guide personalized treatment selection.

Fig. 1. Identification of immune modules through cross-comparison of inflammatory skin disease profiles.

a Sentinel biopsies from psoriasis (PsO), atopic dermatitis (AD), neutrophilic diseases (NeuD), lichen planus (LP), cutaneous lupus erythematosus (CLE), and Wells syndrome (Wells) patients were profiled. The expression levels for each disease were plotted against profiles of all other diseases to identify differentially expressed genes (DEG) that define modules. Volcano plots depict DEGs, with dashed lines indicating the significance threshold defined as log2 fold change > 2 and p value < 0.01 derived from two-sided t-tests. b UMAP projection of sentinel profiles based on module expression, demonstrating clustering according to disease. c Heatmap of sentinel profiles based on modules, showing hierarchical clustering along with disease-specific expression of modules. The color gradient reflects expression levels given as z-scores. d Hierarchical clustering of sentinel profiles based on expression of either module genes, the complete gene panel, or NS Forest minimal gene markers, shown by dendrograms. Clustering accuracy is indicated by the Fowlkes–Mallows (FM) index.

Fig. 2. Module-based profiles can classify inflammatory skin diseases.

a Ridgeline plots depict Th1, Th2, Th17, type I interferon (IFN), neutrophilic (neutro), macrophagic (macro) and eosinophilic (eosino) module scores in sentinel biopsies across model inflammatory skin diseases (PsO, AD, LP, CLE, NeuD, Wells). Solid lines indicate the thresholds of module activation. b Box plots showing the normalized modules scores in sentinel biopsies across all model diseases. Each dot represents one sentinel biopsy. LP (n = 12), AD (n = 16), PsO (n = 25), NeuD (n = 10), CLE (n = 12), Wells (n = 3), Healthy (n = 8). c Box plots illustrating the normalized modules scores in test biopsies. Each dot represents one test biopsy. LP (n = 6), AD (n = 15), PsO (n = 8), NeuD (n = 5), CLE (n = 7). b–c the central line of the box plot is the median. The box’s edges are the lower (25th percentile) and upper quartiles (75th percentile). Whiskers extend to data points within 1.5 times the interquartile range (IQR). d UMAP projection showing clustering of test biopsies with sentinel biopsies having concordant disease diagnosis. Black and grey rings represent test and sentinel biopsies, respectively.

Fig. 3. Immune modules can classify non-sentinel inflammatory skin diseases such as bullous pemphigoid and maculopapular drug-hypersensitivity reactions.

a Heatmap displaying module profiles of maculopapular drug-hypersensitivity reaction (DHR) (red) and bullous pemphigoid (BP) (turquoise) plotted against the established module-based cartography of sentinel probes. b UMAP projection of sentinel, DHR, and BP probes showing disease clustering based on module expression. c Hierarchical clustering of DHR and BP probes based on module genes, the complete gene panel, or the NS Forest minimal genes, shown as dendrogram. Clustering accuracy is indicated by the Fowlkes–Mallows (FM) index. d Box plots illustrating immune module scores in DHR (n = 10), and BP (n = 12) compared to AD. Each dot represents one probe. The central line of the box plot is the median. The box’s edges are the lower (25th percentile) and upper quartiles (75th percentile). Whiskers extend to data points within 1.5 times the interquartile range (IQR).

Fig. 4. Erythroderma and undetermined skin rashes display module profiles of AD, PsO, LP, and DHR sentinels.

a Module-based UMAP projection of erythroderma probes (grey circles) plotted against the established psoriasis (green circles), AD (blue circles), and DHR (red circles) cartography, showing disease-specific clustering. b Box plots showing immune module expression scores in erythroderma probes split by the sentinel disease (AD, n = 16; PsO, n = 10; or DHR, n = 4) they cluster with. Each dot represents one probe. c Module-based UMAP projection of undetermined rashes (grey circles) plotted against the established PsO (green circles), AD (blue circles), CLE (yellow circles), BP (turquoise circles), and DHR (red circles) cartography, showing disease-specific clustering. d Box plots showing immune module expression scores in erythroderma probes split by the sentinel disease (AD, n = 8; PsO, n = 10; LP, n = 2; or BP, n = 1) they cluster with. b and d, the central line of the box plot is the median. The box’s edges are the lower (25th percentile) and upper quartiles (75th percentile). Whiskers extend to data points within 1.5 times the interquartile range (IQR).

Fig. 5. Module matching to treatment targets to enhance therapeutic efficacy.

a Profiling of 80 pre-treatment biopsies from patients with AD, PsO, and LP undergoing targeted therapy, separated into responders or non-responders based on clinical scoring at week 16. The dominant expression module in skin probes (Th1, Th2 or Th17) was matched to the treatment target, defined as Th2 for anti-IL4R and IL-13, Th17 for anti-IL-23 and IL-17A/F, and Th1 for JAK1/2 inhibitors. Data are represented as percentages of responding and non-responding patients in the matched and non-matched groups. b Cumulative Th1, Th2, and Th17 module scores in pre- and post-treatment biopsies in 5 non-responding AD patients undergoing anti-Th2 treatment with anti-IL4R.