Single-cell analysis of psoriasis resolution demonstrates an inflammatory fibroblast state targeted by IL-23 blockade

Abstract

Biologic therapies targeting the IL-23/IL-17 axis have transformed the treatment of psoriasis. However, the early mechanisms of action of these drugs remain poorly understood. Here, we perform longitudinal single-cell RNA-sequencing in affected individuals receiving IL-23 inhibitor therapy. By profiling skin at baseline, day 3 and day 14 of treatment, we demonstrate that IL-23 blockade causes marked gene expression shifts, with fibroblast and myeloid populations displaying the most extensive changes at day 3. We also identify a transient WNT5A+/IL24+ fibroblast state, which is only detectable in lesional skin. In-silico and in-vitro studies indicate that signals stemming from these WNT5A+/IL24+ fibroblasts upregulate multiple inflammatory genes in keratinocytes. Importantly, the abundance of WNT5A+/IL24+ fibroblasts is significantly reduced after treatment. This observation is validated in-silico, by deconvolution of multiple transcriptomic datasets, and experimentally, by RNA in-situ hybridization. These findings demonstrate that the evolution of inflammatory fibroblast states is a key feature of resolving psoriasis skin.

Fig. 1. Identification of cell populations showing an early response to IL-23 blockade.

A Experimental design of the study, showing timepoints for the sampling of non-lesional (day 0) and lesional (day 0, day 3, day 14) skin biopsies (n = 5 patients). Created with BioRender.com. B Left: Percentage reduction of disease severity (measured by the psoriasis area and severity index, PASI) after 3 and 14 days of risankizumab treatment. Every line represents a patient. Right: representative skin images showing clinical improvement during risankizumab treatment. C Unifold Manifold Approximation and Projection (UMAP) of 164,553 single cells (lesional and non-lesional skin), visualized as clusters (left) or skin cell types (right). D Dot plot showing the expression of marker genes used for the annotation of cell identities. E UMAP visualization of lesional skin showing the number of differentially expressed genes (DEG) observed in each cluster, after 3 (left) and 14 (right) days of treatment. F Heatmap depicting the number of ligand-receptor interactions between cell types. Each ligand-receptor interaction was assigned to the condition in which it had the strongest score. Interactions were then counted. The cell types expressing the ligand are labeled on the y-axis, and those expressing the receptor are labelled on the x-axis. KC, keratinocytes; FB, fibroblasts; PE, pericytes; LEC, lymphatic endothelial cells; VEC, vascular endothelial cells; EpC, epithelial cells; MEL, melanocytes; MAST, mast cells; MYEL, myeloid cells; TC, T cells.

Fig. 2. Effects of IL-23 inhibition in keratinocytes.

A UMAP of 68,695 keratinocytes forming five distinct subclusters (n = 5 patients). B Dot plot showing the expression of marker genes used for the annotation of cell identities. C Abundance of proliferating cells (expressed as keratinocyte percentage) at different time points after risankizumab treatment. Every line represents a patient. *P < 0.05, **P < 0.01 (repeated measures one-way ANOVA with Dunnett’s post-test). D UMAP visualization of keratinocytes from lesional skin, showing the number of differentially expressed genes (DEG) observed in each cluster, after 3 (left) and 14 (right) days of treatment. KC, keratinocytes; L, lesional skin; NL, non-lesional skin.

Fig. 3. Effects of IL-23 inhibition on lymphocytes and dendritic cells.

A UMAP of 6263 lymphocytes forming seven distinct clusters (n = 5 patients). B Dot plot showing the expression of marker genes used for the annotation of T cell identities. C UMAP visualization of lymphocytes from lesional skin, showing the number of differentially expressed genes (DEG) observed in each cluster, after 3 (left) and 14 (right) days of treatment. D Downregulation of cytotoxicity markers in TEMRA T cells, following treatment with risankizumab. E UMAP of 18,544 myeloid cells, forming 11 distinct clusters. F Dot plot showing the expression of marker genes used for the annotation of myeloid cell identities. G UMAP visualization of myeloid cells from lesional skin, showing the number of differentially expressed genes (DEG) observed in each cluster, after 3 (left) and 14 (right) days of treatment. Treg, regulatory T cell; TEMRA, terminally differentiated effector memory T cell; NK, natural killer cell; ILC, innate lymphoid cell; DC, dendritic cell; Mac, macrophage; L, lesional skin.

Fig. 4. Effects of IL-23 inhibition on vascular endothelial cells and pericytes.

A UMAP of 16,420 vascular endothelial cells forming five clusters (n = 5 patients). B Dot plot showing the expression of marker genes used for the annotation of vascular endothelial cell identities. C Risankizumab-induced downregulation of proinflammatory molecules and collagen genes, in PLVAP+/RGCC+ capillaries. D UMAP of 13,507 pericytes forming two clusters. E Dot plot showing the expression of marker genes used for the annotation of pericyte cell identities. F UMAP visualization of pericytes from lesional skin, showing the number of differentially expressed genes (DEG) observed in each cluster, after 3 (left) and 14 (right) days of treatment. G Genes promoting leukocyte extravasation are downregulated in pericyte 1, following risankizumab treatment. L, lesional skin; NL, non-lesional skin.

Fig. 5. Effects of IL-23 inhibition on fibroblasts.

A UMAP of 31,765 fibroblasts forming eleven clusters (n = 5 patients). B Dot plot showing the expression of marker genes used for the annotation of cell identities. C Plot showing the decline in WNT5A+/IL24+ cell abundance, following treatment with risankizumab. Every line represents a patient. *P < 0.05; **P < 0.01 (repeated measures ANOVA, with Dunnett’s post-test). D Real time PCR analysis showing that key markers for WNT5A+/IL24+ cells are upregulated in primary human fibroblasts stimulated with IL-17A and TNF. n = 5 biologically independent samples; data are mean (SD); *P < 0.05; **P < 0.01, ***P < 0.001 (Friedman test with Dunn’s post-test). E Real time PCR analysis of inflammatory markers in human primary keratinocytes cultured with unconditioned medium (no FB grown), supernatants from IL-17A/TNF stimulated fibroblasts (Stim), control medium (supernatant from unstimulated fibroblasts, Unstim) or control medium supplemented with IL-24. n = 4 biologically independent samples for all conditions, with the exception of n = 2 biologically independent samples for Unstim 2 ng IL-24; data are mean (SD); *P < 0.05; **P < 0.01 (Kruskal Wallis test with Dunn’s post-test). F Plot showing the SCENIC regulons detected in WNT5A+/IL24+ fibroblasts. The three regulons with the highest specificity scores are indicated by labels, with their size (number of regulated genes) in brackets. G Boxplots showing fibroblast clusters ordered by pseudotime; data are median (interquartile range, minimum-maximum). H Dot plot showing the expression of the top regulons in the various fibroblast clusters. L, lesional skin; NL, non-lesional skin; SD, standard deviation.

Fig. 6. The abundance of WNT5A+/IL24+ fibroblasts declines rapidly following psoriasis treatment.

Deconvolution of transcriptome datasets shows that the frequency of WNT5A+/IL24+ fibroblasts is significantly reduced by treatment with guselkumab (n = 22 patients for L Week 0, n = 9 for L Week 1, n = 8 for L Week 12, n = 20 for NL Week 0; data are mean (SD)) (A), ixekizumab (n = 6 for placebo group, n = 7 for ixekizumab group; data are mean (SD)) (B) or topical glucocorticoid (halomethasone monohydrate 0.05%), with the latter effect going into reverse upon drug withdrawal and clinical relapse (n = 7 patients per group; data are mean (SD)) (C). D The percentage of WNT5A+ fibroblasts (PDGFRA+ cells) detected by RNA in-situ hybridization decreases following treatment with risankizumab, (n = 3 patients, data are mean (SD)). E Representative confocal microscopy image showing the results of RNA in-situ hybridization (10x magnification, scale bar 100 µm). Arrows indicate cells triple-positive for PDGFRA (green), COMP (red) and WNT5A (magenta). F Diagram showing the early effects of therapeutic IL-23 inhibition. Left: in lesional psoriasis skin, IL-23 drives the differentiation of T17 cells. These cells produce IL-17, leading to the activation of keratinocytes and the evolution of COMP+ fibroblasts into a WNT5A+/IL24+ state. The ensuing IL-24 production further reinforces keratinocyte activation. Right: following IL-23 blockade, T17 cell differentiation and IL-17 production are inhibited, while WNT5A+/IL24+ fibroblasts revert to their COMP+ state. As a result of diminished IL-17 and IL-24 production, keratinocyte homeostasis is re-established. Created with BioRender.com. Ixe, ixekizumab; L, lesional skin; NL, non-lesional skin; SD, standard deviation; ns, not significant; *P < 0.05; **P < 0.01; ****P < 0.0001 (mixed effects ANOVA with Dunnett’s post-test (A), two-sided paired t test (B), or repeated measures ANOVA with Dunnet’s post-test (C–D)).