Single-cell analysis reveals key differences between early-stage and late-stage systemic sclerosis skin across autoantibody subgroups

Abstract

Objectives: The severity of skin involvement in diffuse cutaneous systemic sclerosis (dcSSc) depends on stage of disease and differs between anti-RNA-polymerase III (ARA) and anti-topoisomerase antibody (ATA) subsets. We have investigated cellular differences in well-characterised dcSSc patients compared with healthy controls (HCs).

Methods: We performed single-cell RNA sequencing on 4 mm skin biopsy samples from 12 patients with dcSSc and HCs (n=3) using droplet-based sequencing (10× genomics). Patients were well characterised by stage (>5 or <5 years disease duration) and autoantibody (ATA+ or ARA+). Analysis of whole skin cell subsets and fibroblast subpopulations across stage and ANA subgroup were used to interpret potential cellular differences anchored by these subgroups.

Results: Fifteen forearm skin biopsies were analysed. There was a clear separation of SSc samples, by disease, stage and antibody, for all cells and fibroblast subclusters. Further analysis revealed differing cell cluster gene expression profiles between ATA+ and ARA+ patients. Cell-to-cell interaction suggest differing interactions between early and late stages of disease and autoantibody. TGFβ response was mainly seen in fibroblasts and smooth muscle cells in early ATA+dcSSc skin samples, whereas in early ARA+dcSSc patient skin samples, the responding cells were endothelial, reflect broader differences between clinical phenotypes and distinct skin score trajectories across autoantibody subgroups of dcSSc.

Conclusions: We have identified cellular differences between the two main autoantibody subsets in dcSSc (ARA+ and ATA+). These differences reinforce the importance of considering autoantibody and stage of disease in management and trial design in SSc.

Keywords: Autoantibodies; Fibroblasts; Scleroderma, Systemic.

Figure 1

Overview of scRNAseq landscape. Markers to identify clusters, and differences between early-stage and late-stage SSc and HC (A) UMAP of all samples from skin samples. (B) UMAP with named clusters. (C) Feature plot with key genes used to identify clusters. (D) Split UMAP showing gross differences in abundance between early dcSSc, late dcSSc and HC. Most obvious differences apparent between keratinocyte clusters and fibroblast clusters. (E) PCA plot composed using pseudobulk analysis of all cells and all samples, with ellipses highlighting early dcSSc (red), late dcSSc (green) and HC (blue). (F) Bar plot of proportion abundance of each cluster by stage of disease. Notable differences can be seen with a contracted proportion of T cells and expanded keratinocytes in dcSSc. (G) Bar plots showing proportion each cell type by stage of disease red=early dcSSc, green=late dcSSc, blue=HC. dcSSc, diffuse cutaneous systemic sclerosis; HC, healthy control; PCA, principal component analysis; scRNAseq, single-cell RNA sequencing.

Figure 2

Re-clustered fibroblast landscape for early-stage and late-stage SSc and HC. Heatmap of differentiating genes, differences KEGG pathways and naming fibroblast subsets. (A) UMAP of fibroblast subset from all samples discriminating 10 distinct fibroblast populations. (B) Heatmap of the top 10 differential overexpressed genes by statistical significance for each cluster. (C) Split UMAP of fibroblast clusters by stage of disease. Visually clear differences in cluster 0, cluster 4 and cluster 5. (D) PCA plot from pseudobulk analysis of all fibroblast cells from all samples, with ellipses highlighting early dcSSc (red), late dcSSc (green) and HC (blue). (E) Key differentiating genes by each fibroblast cluster. (F) KEGG pathway analysis, showing clear different gene set enrichment in each fibroblast cluster. (G) Barplot by abundance per subset of fibroblast clusters. dcSSc, diffuse cutaneous systemic sclerosis; HC, healthy control; PCA, principal component analysis.

Figure 3

Differences by autoantibody across all cells and within the fibroblast cluster. (A) UMAP plot of whole skin split by stage and autoantibody. (B) bar plot showing mean frequency and SD of each cluster by SSc stage and ANA subset. red=early ARA+dcSSc, olive=late ARA+dcSSc, green=early ATA+dcSSc, blue=late ATA+dcSSC, purple=HC. (C) Stacked bar plot of proportion abundance by individual sample. (D) PCA plot from pseudobulk analysis from whole skin. This shows much clearer differentiation of sample groups when both stage and antibody are taken into consideration. red=early ARA+dcSSc, olive=late ARA+dcSSc, green=early ATA+dcSSc, blue=late ATA+dcSSc, purple=HC. (E) PCA plot from pseudobulk analysis of fibroblast subset. Once again, there is clearer differentiation between the subsets when both stage and antibody are taken into consideration compared with only stage. More marked differentiation is apparent between early and late ATA+, than for ARA+ patients. Red=early ARA+dcSSc, olive=late ARA+dcSSc, green=early ATA+dcSSc, blue=late ATA+dcSSC, purple=HC. (F) KEGG pathway analysis of all fibroblasts by antibody and stage. ARA, anti-RNA-polymerase III; ATA, antitopoisomerase antibody; dcSSc, diffuse cutaneous systemic sclerosis; HC, healthy control; PCA, principal component analysis.

Figure 4

Differences by autoantibody and stage in cluster 0, KEGG differences, key gene expression violin plots and trajectory differences by antibody. (A) Gene expression differences across whole skin of key genes previously identified as having differential protein concentrations in the serum by stage and antibody. (B) Differential expression of key genes within the 10 fibroblast clusters by stage and antibody. Some key profibrotic genes are clearly only expressed in early dcSSc, or in certain clusters. SFRP4 only expressed cluster 8, consistent with myofibroblasts. (C) KEGG pathway differential expression in cluster 0 fibroblasts (CCN5+PTX3+ FBs) between each antibody and stage. (D) Violin plots from cluster 0 fibroblasts (CCN5+, PTX3+). Comparison between ARA+ early and late stage, where most overexpressed genes are seen in early ARA+. ATA+ cluster 0 fibroblasts (CCN5+PTX3+) shows significant differential expression in both earl-stage and late-stage disease. A key set of differentially expressed genes separate FB cluster 0 between early ARA+ and ATA+. (E) Pseudotime analysis of fibroblast clusters. Originator FB cluster in both early antibody subsets was identified as being cluster 0, and myofibroblasts were identified as cluster 8. However, branch points differ between ARA+ and ATA+ early dcSSc, with some terminal fibroblasts being cluster three in ATA+ early dcSSc, whereas in ARA+ early dcSSc, FBs do not terminally differentiate at cluster 3, but include cluster 5 and cluster 4. ARA, anti-RNA-polymerase III; ATA, antitopoisomerase antibody; dcSSc, diffuse cutaneous systemic sclerosis; HC, healthy control.

Figure 5

Comparison of signalling strength and pathways between early-stage and late-stage dcSSc by autoantibody subgroup. Heatmap of cell-to-cell interaction differences between early-stage and late-stage disease in (A) ARA+ patients and (B) ATA+ patients. The y axis indicates sending cells, and x axis is receiver cells. Red indicates a stronger signal in late-stage disease, and blue is stronger signal in early disease. (C+D) Differential gene set pathway analysis between early-stage and late-stage disease in (C) ARA+dcSSc and (D) ATA+ patients. Red indicates relative expression in early dcSSc, whereas turquoise is expression from late dcSSc. ARA, anti-RNA-polymerase III; ATA, antitopoisomerase antibody; dcSSc, diffuse cutaneous systemic sclerosis.

Figure 6

Functional differences between whole skin ARA+ and ATA+ cell cluster interactions for candidate signalling pathways (TGFb, CCL and complement). (A) Relative pathway differential expression in ARA+ early dcSSc (red) and ATA+ early dcSSc (turquoise). (B) Outgoing and incoming signal between cell clusters in whole skin in ARA+ early dcSSc and ATA+ early dcSSc. Notable incoming signal differences seen between fibroblast clusters and lymphatic endothelial cells. (C) Heatmap highlighting differential cell interaction strength between ATA+ and ARA+ early dcSSc. Red indicates higher interaction strength in ATA+ early dcSSc, and blue indicates higher interaction strength in ARA+ early dcSSc. (D) concentrating on TGFβ ligand-receptor interactions, differences can be seen between ATA+ and ARA+ patients in both source of ligands, and more notably where receptors found. In gene expression data from ATA+ skin, receptors were expressed by three fibroblast clusters, whereas in early ARA+ patients, endothelial cells express receptors for the ligands. Plots showing cell to cell interaction and strength of that interaction for specific pathways including (E) TGFβ, (F) CCl signalling and (G) complement by different early antibody states. The strength of the signal is determined by the intercellular lines thickness; the thicker the line, the stronger the signal intensity. (H) Hierarchical pattern of similar expression and cells responding to each pattern of pathway response. In top panels, TGFβ pattern is grouped in pattern 1, and the cells responding to this pathway are predominantly the fibroblasts. In ARA+ early dcSSc, TGFβ expression is grouped in pattern 3, and the cells showing strongest response to pattern 3 are the endothelial and lymphatic endothelial cells. ARA, anti-RNA-polymerase III; ATA, antitopoisomerase antibody; dcSSc, diffuse cutaneous systemic sclerosis.