Single-cell RNA sequencing of psoriatic skin identifies pathogenic Tc17 cell subsets and reveals distinctions between CD8+ T cells in autoimmunity and cancer

Abstract

Background: Psoriasis is an inflammatory, IL-17-driven skin disease in which autoantigen-induced CD8+ T cells have been identified as pathogenic drivers.

Objective: Our study focused on comprehensively characterizing the phenotypic variation of CD8+ T cells in psoriatic lesions.

Methods: We used single-cell RNA sequencing to compare CD8+ T-cell transcriptomic heterogeneity between psoriatic and healthy skin.

Results: We identified 11 transcriptionally diverse CD8+ T-cell subsets in psoriatic and healthy skin. Among several inflammatory subsets enriched in psoriatic skin, we observed 2 Tc17 cell subsets that were metabolically divergent, were developmentally related, and expressed CXCL13, which we found to be a biomarker of psoriasis severity and which achieved comparable or greater accuracy than IL17A in a support vector machine classifier of psoriasis and healthy transcriptomes. Despite high coinhibitory receptor expression in the Tc17 cell clusters, a comparison of these cells with melanoma-infiltrating CD8+ T cells revealed upregulated cytokine, cytolytic, and metabolic transcriptional activity in the psoriatic cells that differed from an exhaustion program.

Conclusion: Using high-resolution single-cell profiling in tissue, we have uncovered the diverse landscape of CD8+ T cells in psoriatic and healthy skin, including 2 nonexhausted Tc17 cell subsets associated with disease severity.

Keywords: CD8(+) T cell; Single-cell RNA sequencing; Smart-seq2; T-cell exhaustion; melanoma; psoriasis.

Conflict of interest statement

Conflict of Interest

T.B. is currently an investigator for Celgene, Janssen, Merck, and Regeneron. She has served as an advisor for Abbvie, Lilly, and Pfizer. M.D.R is a founder and consultant for TRex Bio., a founder of Sitryx Bio., and receives funding from Abbvie, LEO Pharma and TRex Bio.

Copyright © 2020 American Academy of Allergy, Asthma & Immunology. Published by Elsevier Inc. All rights reserved.

Figures

Figure 1.

Heterogeneity of CD8+ T cells in healthy and psoriatic skin. Clustering of 4,575 CD8+ T cells shown in UMAPs colored by (A) cluster and (B) clinical status. (C) Composition of CD8+ T cells in each cluster by different subjects. (D) Average percentage of CD8+ T cells in each cluster out of all CD8+ T cells within each subject. Bar heights indicate the means across healthy or psoriasis subjects, error bars represent standard deviations. Clusters ordered from left to right by increasing ratio of healthy mean percentage to psoriasis mean percentage. (E) Proportion of cluster 6 and 10 cells in healthy and psoriatic subjects. Bars and whiskers represent means and standard deviations.

Figure 2.

Specific inflammatory CD8+ T cell subtypes in psoriatic lesions. (A) Average expression in each cluster of marker genes for various T cell functions and characteristics. Each column represents the average mean-centered and scaled expression z-score of each gene (in rows) for the cells in a cluster. Color intensities outside of the z-score range [−1, 1] were set to the maximum and minimum of this range. (B) Cell cycle classification of cells in each cluster (C) Normalized log-transformed expression of IL17A and IL17F in each cluster.

Figure 3.

Tc17 subsets express CXCL13, a clinically relevant psoriasis biomarker. (A) CXCL13 expression and (B) co-expression with IL17A across all clusters. Points represent normalized, log-transformed count values for each cell. (C) RNAscope staining of CXCL13 and IL17A mRNA in psoriatic lesions. (D) CD8+ T cell-colocalized RNAscope fluorescence for (from left to right) CXCL13, IL17A, and both genes. Points represent proportion of pixels positive for CD8A, CD3E, and DAPI that are also positive for each gene(s) of interest in psoriatic skin (PSO, n=6), eczema skin (n = 4), healthy skin (n = 6), and control (n=2) slides. (E) Correlation of serum CXCL13 with PASI in VOYAGE 2 cohort. Trendline based on least squares regression. (F) Skin and lesional CXCL13 expression in VOYAGE 1 participants treated with guselkumab. Error bars = Mean ± SD. NL = non-lesional sites sampled at Week 0. (G) Average classification accuracy of 5-fold cross validated SVMs trained on Gudjonsson et al. 2010 (GSE13355) and Li et al. 2014 (GSE54456) datasets using different psoriasis-associated genes (n = 50 for each gene combination). P-values are shown for Wilcoxon tests between groups indicated by brackets.

Figure 4.

Dysfunction characteristics in psoriatic and melanoma-infiltrating CD8+ T cells. (A) Dysfunction scores of psoriasis and healthy skin CD8+ T cells in each cluster according to the core dysfunction signature from Tirosh et al. 2016. (B) Coinhibitory receptor expression across CD8+ T cell clusters. (C) Metabolic scoring of glycolysis and oxidative phosphorylation pathway gene expression, along with score difference, for each cell. (D) Common pathways represented by differentially-expressed genes between most exhausted vs least exhausted clusters in psoriasis (this study) and melanoma (Tirosh et al. 2016) CD8+ T cells.

Figure 5.

TCR sharing between Tc17 clusters. (A) TCR sharing network between clusters. Thickness and intensity of each curve represent the proportion of shared TCRs in one cluster (at arrow head) that can be found in another (at arrow tail). (B) Cluster distributions of shared TCRs. Each plot shows, for each cluster, the mean proportion of its ‘shared’ TCRs that occurs in each other cluster. Error bars represent mean ± standard error. Asterisks indicate a mean proportion that is significantly different from all others within the same cluster. n = number of shared TCRs detected for a cluster.