LGR5 expressing skin fibroblasts define a major cellular hub perturbed in scleroderma

Summary

Systemic sclerosis (scleroderma, SSc) is an incurable autoimmune disease with high morbidity and mortality rates. Here, we conducted a population-scale single-cell genomic analysis of skin and blood samples of 56 healthy controls and 97 SSc patients at different stages of the disease. We found immune compartment dysfunction only in a specific subtype of diffuse SSc patients but global dysregulation of the stromal compartment, particularly in a previously undefined subset of LGR5⁺-scleroderma-associated fibroblasts (ScAFs). ScAFs are perturbed morphologically and molecularly in SSc patients. Single-cell multiome profiling of stromal cells revealed ScAF-specific markers, pathways, regulatory elements, and transcription factors underlining disease development. Systematic analysis of these molecular features with clinical metadata associates specific ScAF targets with disease pathogenesis and SSc clinical traits. Our high-resolution atlas of the sclerodermatous skin spectrum will enable a paradigm shift in the understanding of SSc disease and facilitate the development of biomarkers and therapeutic strategies.

Graphic abstract.

Introduction

Systemic sclerosis (SSc) is a rare chronic autoimmune disease that primarily affects women between 30 and 50 years old with high morbidity and mortality rates greater than those of any other rheumatic disease (Bhattacharyya et al., 2011; Allanore et al., 2015). SSc involves a broad spectrum of clinical features that encompass vascular, immune, and fibrotic manifestations affecting the skin and internal organs (Trojanowska, 2010; Denton and Khanna, 2017). A striking feature of SSc is the heterogeneity in patients' clinical manifestations, autoantibody profiles, rate of disease progression, response to treatment, and survival (Allanore et al., 2015). Skin fibrosis, the distinguishing hallmark of SSc, allows the classification of patients into two major subsets: limited cutaneous SSc (lSSc), confined distally to the elbows, knees, or face and diffuse cutaneous SSc (dSSc), also involving the proximal portion of the extremities as well as the trunk and face. It remains uncertain whether these two subsets represent distinct molecular diseases or merely reflect two extremes on the disease spectrum. Both the SSc subsets may involve visceral organs, including the lungs, gastrointestinal tract, kidneys, and heart (Denton and Khanna, 2017), and the degree of skin thickness score correlates with organ involvements (Matsuda et al., 2019).

Immunological abnormalities of the innate and adaptive immune system have long been recognized in SSc. These include chronic mononuclear cell infiltration into affected tissues, increased peripheral blood monocyte count, dysregulation of chemokines, cytokines, and diverse growth factors (York, 2011; Pattanaik et al., 2015; Kania et al., 2019). A recent single-cell RNA-seq (scRNA-seq) study (Gaydosik et al., 2021) identified perturbations in CXCL13⁺ T cells in the SSc skin. Autoimmunity in SSc is best exemplified by the presence of multiple specific autoantibodies, such as the anti-topoisomerase1 (Scl-70), anticentromere, and anti-RNA polymerase III antibodies, enriched in distinct clinical subsets (Kayser and Fritzler, 2015). However, given the limited impact of canonical immunomodulatory therapies on SSc progression, the contribution of the immune compartment to disease initiation and/or disease maintenance remains an open question.

Despite the fibrotic nature of SSc, the precise molecular profiles of the skin stromal cell compartment, which consists of fibroblasts, myofibroblasts, blood vessels, and pericytes, are not fully elucidated. Skin fibrosis can occur as part of the natural wound-healing process upon injury (Eming et al., 2017) or because of diseases, such as SSc or graft versus host disease (GVHD) (Strong Rodrigues et al., 2018). Stromal cells' contribution to SSc was studied as early as the 1970s, focusing specifically on myofibroblasts (Gabbiani et al., 1972; Hinz and Lagares, 2020). Recently, scRNA-seq studies described heterogeneity within subsets of fibroblasts in normal and aging skin, following wound healing, and in the skin and lungs of SSc patients (Philippeos et al., 2018; Tabib et al., 2018, 2021; Guerrero-Juarez et al., 2019; Valenzi et al., 2019; Solé-Boldo et al., 2020; Buechler etal., 2021).

However, the comprehensive scRNA-seq profiling of skin samples from large SSc patient cohorts encompassing the diverse clinical spectrum of SSc has never been done. Furthermore, the major drivers and fundamental mechanisms regulating fibrosis in SSc remain largely unknown.

Here, we report the first population-scale scRNA-seq profiling of skin biopsies derived from the dorsal forearm of 56 healthy controls and 97 lSSc and dSSc patients at different stages of disease progression and from an additional 10 patients with other fibrotic skin diseases (e.g., the sclerodermatous type of GVHD and localized scleroderma [morphea]). Although SSc is considered to be an immune-mediated disease, we did not find consistent immune perturbations shared across patients; rather, we found that changes in immune cell composition were confined to a specific subtype of dSSc patients with early disease or a high skin score. In striking contrast, most dSSc patients shared global and systemic perturbations in the stromal cell compartment. We defined 10 skin fibroblast cell states from healthy and SSc patients based on the different expressions of gene modules and signaling pathways, pointing toward diverse functions including skin maintenance, ECM remodeling, fibrosis, and immune modulation. A fibroblast subset defined by the expression of the LGR5 receptor, which we term scleroderma-associated fibroblasts (ScAFs), was significantly diminished in the dSSc patients compared with that in the lSSc patients and healthy donors. Importantly, the remaining ScAFs of the SSc patients express both known and unknown SSc driver genes and pathways associated with excessive ECM deposition and reduced ECM degradation, activation of complement and coagulation systems, the modulation and proliferation of blood vessels, and the activation of cellular senescence and profibrotic signaling pathways. Combining single-cell sequencing assay for transposase-accessible chromatin (scATAC-seq) with scRNA-seq from the same individual cells, we revealed the major coordinators of the transition of healthy to pathological ScAFs. We found that the open chromatin regions in SSc patients are enriched with AP-1 motifs (including ATF3, JUN, FOS, FOSL, and JDP2) and other transcription factors (TF, e.g., BACH1, MAFK, and TEAD1/3). Using 3D whole-mount single-molecule fluorescence in situ hybridization (smFISH) analysis, we discovered that ScAFs in the healthy skin are scattered mainly in the deep reticular dermis and are largely perturbed in their number and morphology in SSc. We further showed that ScAFs display a large number of ligand-receptor interactions with the skin immune and stromal compartment. In summary, coupling population-scale scRNA-seq with detailed clinical metadata identifies a distinct LGR5-expressing fibroblast subset (ScAFs) as the central axis of perturbation in SSc patients. Targeted therapies directed against these ScAF pathways might begin a new era for SSc therapies.

Results

Atlas of skin and blood immune cells of SSc patients and healthy subjects

To better understand the SSc pathogenesis and characterize the immune and stromal cell function and diversity in SSc patients, we designed a robust protocol for massively parallel single-cell RNA-sequencing (MARS-seq) (Jaitin et al., 2014; Keren-Shaul et al., 2019) to characterize whole skin tissue biopsies derived from the mid-forearm of SSc patients and age-matched healthy controls with no medical morbidities (Figure 1A; STAR Methods). Our design focused on maintaining the in situ cellular and molecular composition of each sample by instantly cooling the skin and blood samples, enzymatically and mechanically dissociating the skin, and immediately sorting the blood and skin immune cells (CD45⁺) as well as skin stromal cells (Thy-1/CD90⁺) for MARS-seq analysis (Figures 1A and S1; STAR Methods). In order to comprehensively map the full spectrum of SSc disease, we recruited a heterogeneous cohort of SSc patients with diverse demographic characterizations, disease stages, and clinical manifestations (Table S1; STAR Methods).

Figure 1. Single-cell atlas of blood and skin immune cells from SSc patients and healthy subjects.

(A) Overview of the experimental setting. Skin and blood samples of healthy donors and SSc patients were dissociated into single cells. CD45⁺ and CD90⁺ cells were sorted for transcriptional profiling.

(B) 2D projection of the subclustered blood and skin immune cells with subpopulations marked by a color code.

(C) 2D projection of the single cells by the source of the samples (blood or skin).

(D) 2D projection of a selected set of marker genes over the metacell model.

(E) Normalized expression of selected genes across the metacell model. Each bar represents one metacell, colored as in (B).

(F) Bar plot showing the sample source contribution of individual immune cell populations.

To better define the role of the immune system in SSc, we first analyzed the blood and skin immune cells. We profiled a total of 67,156 quality control (QC)-positive (STAR Methods) blood and skin immune cells derived from 86 SSc patients (38 lSSc and 48 dSSc) with diverse clinical manifestations, and 35 healthy donors (Figures S2A and S2B; Table S2). We used the MetaCell algorithm (STAR Methods) to identify the homogeneous and robust groups of cells based on the cluster-specific expression patterns of the 2,202 most variable genes, resulting in a detailed map of 385 meta cells organized into 8 broad immune cell lineages (Figures 1B, 1C, and S2C). Atwo-dimensional projection of all immune single cells revealed separation into diverse cell types and states, based on lineage-specific marker genes (Figures 1D and 1E). This immune atlas consisted of T cells, including blood and skin effector T cells (T_Effector and sT_Effector, respectively), regulatory T cells (Treg, sTreg, and sTreg_CXCR4), γδ T cells (T_GD), and blood and skin naive T cells (T_naive and sT), all characterized by a high expression of CD3. The immune atlas also contained subsets of NK cells (NK, NK_XCL1, and NK_XCL1_CXCR4), plasmacytoid dendritic cells (pDCs and pDC_CXCR4), conventional dendritic cells (DCs, DC_CCL22, DC_CXCL10, and DC_XCR1), macrophages (Mf and MF_TREM2), monocytes (Mo, Mo_CD16, M_IL1B, and M_CD16_IL1B), Langerhans cells (LCs), mast cells (Mast and Mast_CLC), B cells (B and B_CXCR4), and plasma cells. Many myeloid and lymphoid subsets were predominantly found in the skin, including the DC subsets, LCs, skin effector and regulatory T cells, Mf, mast cells, and subsets of NK cells, while the other immune subsets were predominantly found in the blood (Figures 1C and 1F). Further comparison of the skin and blood immune cells showed molecular differences between the blood and skin immune subsets and provided insights into immune cell adaptations to the skin tissue and its intrinsic activation properties (Figures S2D−S2H).

Local changes define the immune cell composition of SSc patients compared with healthy controls

To estimate alterations in the immune compartment between SSc patients and healthy subjects, we performed differential composition analysis between SSc patients and healthy donor blood and skin immune cells. We visualized the results by computing the 2D density of the projected cells from each paired group, controlling for the number of cells and patients (STAR Methods). To our surprise given that SSc is considered to be an immune-mediated disease, we did not find major global and consistent changes between SSc patient groups and healthy subjects in blood (Figures 2A−2C, S3A, and S3B) or skin (Figures 2A, 2D, 2E, S3C, and S3D) immune subsets.

Figure 2. Blood and skin immune cells composition of healthy subjects and SSc patients.

(A) 2D density plots showing immune cell population enrichment over the metacell model in blood and skin from healthy donors and SSc patients’ groups (STAR Methods). (Left) Reference map of skin and blood immune populations as in Figure 1B; (Right) down-sampled blood (upper) and skin (lower) immune cells of healthy donors and SSc patients are shown with contour lines indicating the density of projected cells.

(B and D) Bar plots showing the blood (B) and skin (D) immune cell-type composition within the healthy donors and the SSc patients in each group (left) or in the individual patients (right). Cell types are colored the same as in (A).

(C and E) Dot plots showing percentages of selected blood (C) and skin (E) immune cell populations within the healthy donors and SSc patients’ groups. *p < 0.1, **p < 0.05, ***p < 0.01, Mann-Whitney U test, two-sided. Error bars indicate mean ± SEM.

Since SSc is a heterogeneous disease with several patient subtypes and progression phases that are potentially associated with the molecular patterns of disease, we proceeded with a sub-analysis of the blood and skin immune cells based on the patient's metadata, including disease subtype (lSSc versus dSSc), duration (early ≤3 years versus long ≥ 10 years), activity (high skin score ≥20 versus low ≤10), and autoantibody status. We found significant (p < 0.05, Wilcoxon rank-sum test, two-sided) changes in immune compositions in several subsets of SSc patients, primarily in the skin of dSSc patients with early or active disease (Figures 2B−2E and S3A−S3D; STAR Methods). In the blood, we observed a decrease in naive T cells in both the SSc groups and an increase in classical and non-classical (CD16⁺) monocytes mainly in the dSSc patients, consistent with previous studies (Lescoat et al., 2017; Kania et al., 2019). We also found increased DCs in dSSc patients with long disease duration or with a low skin score, increased pDCs in patients with a high skin score, and increased plasma cells in early dSSc disease compared with those in the other study groups (Figure S3A). In the skin, we observed an increase in IFN-γ-producing T effector cells mainly in the dSSc patients (Figures 2D and 2E), an increase in skin cytotoxic NK cells specifically in the dSSc patients with an early disease or with a high skin score, increased pDC, non-classical monocytes, and plasma cells in the dSSc patients with a high skin score, and decreased DC in the patients with early dSSc (Figure S3C). Interestingly, we found strong correlations between specific autoantibody levels and immune cell subsets, including associations between anti-RNA polymerase III and relatively high levels of skin DCs, and between anti-topoisomerase1 (Scl-70) and relatively high levels of blood mono-cytes and skin T effectors (Figures S3B and S3D). Finally, differential gene expression analysis between the immune compartments of SSc patients and healthy subjects revealed molecular perturbations associated with different immune subsets (Figures S3E−S3H; STAR Methods). In summary, we identified immune cell composition changes associated with specific disease subtype, duration, severity, and autoantibody status, suggesting a complex and heterogenous immune involvement across the SSc disease stages.

The human skin stroma is defined by a diverse functional landscape of fibroblast subsets

The precise molecular definition and functions of the human skin fibroblasts and other stromal cell populations remain only partially characterized (Lemos and Duffield, 2018; Philippeos et al., 2018; Tabib et al., 2018, 2021). To provide a detailed and high-resolution atlas of the skin stromal cells in healthy and SSc patients, we used a broad flow cytometry gating strategy, using the cell surface marker Thy-1/CD90 to label and enrich for stromal cells in the skin. Single-cell analysis using CD45⁻CD90⁻ gating to investigate potentially relevant CD90⁻ stromal cells confirms that CD45⁻CD90⁻ skin cells are mainly composed of epithelial cells, pericytes, and less than 1% of fibroblasts (Figures S4A and S4B). Therefore, we concluded that our Thy-1 gating strategy is optimal for sampling the non-epithelial stromal cells from skin biopsies.

We profiled a total of 64,152 QC-positive CD90⁺ skin stromal cells derived from 94 SSc patients (42 lSSc and 52 dSSc) and 50 age-matched healthy donors (Figure S4C; Table S2; STAR Methods). Metacell analysis divided the skin stromal compartment into 292 metacells, organized into 8 broad lineages including fibroblasts, pericytes, vascular endothelial cells, and cells of different epidermal and neuronal origins (e.g., melanocytes, keratinocytes, and nerves; Figures 3A−3C and S4D−S4F). The fibroblasts formed a highly diverse group of transcriptional states consisting of 228 metacells. We further annotated the fibroblast metacells by examining the 559 most variable genes, resulting in the de novo characterization of 10 distinct fibroblasts subsets (Figures 3D, 3E, and S4G).

Figure 3. Single-cell atlas of skin stromal cells from SSc patients and healthy subjects.

(A) 2D projection of subclustered skin stromal cells with subpopulations marked by a color code.

(B) 2D projection of a selected set of marker genes over the metacell model.

(C) Normalized expression of selected genes across the metacell model.

(D) Heatmap showing the 559 top variable genes within the 228-skin fibroblast metacells which are grouped into 10 identified skin fibroblast subsets (top) by 10 gene modules (left).

(E) Heatmap showing selected gene ontology (GO) functional enrichment (FDR adjusted p value <0.001) of the gene modules defined in (D) with the most associated fibroblast subsets colored on the top.

Our deep characterization of skin fibroblasts highlighted COCH-expressing fibroblasts (Fibro_COCH), recently discovered as one of the major fibroblast subsets in the skin (Ascensión et al., 2021), which express specific ECM components and assembly genes (e.g., COCH, ASPN, and MMP16), genes of endocrine components (e.g., DIO₂, PTH1R, IRS2, and PCSK1N), and genes associated with neural development (e.g., GAP43, MAB21L2, and NCAM1). Another fibroblast subset was annotated by its high expression of PTGDS (Fibro_PTGDS), a gene important in prostaglandins (PGs) synthesis. Interestingly, these fibroblasts also highly express genes associated with antigen presentation, such as the major histocompatibility complex (MHC) class I (B2M, HLA-A, HLA-B, HLA-C, HLA-F, and HLA-H) and MHC class II (CD74), and important immune cytokine and chemokine genes, such as IL32, IL33, CCL19, and CXCL12.

Two additional fibroblast populations expressed the marker gene MYOC (Fibro_MYOC1 and Fibro_MYOC2) encoding for proteins associated with assembly and disassembly of actin filaments and smooth muscle relaxation (e.g., MYOC and HSPB6). The Fibro_MYOC2 subset expressed the C7 gene, which encodes a subunit of the membrane attack complex (MAC) and other complement component genes (e.g., CFD, CFH, and C3). Both the populations express genes important for adipogenesis, lipid metabolism, and protection against oxidative stress (e.g., APOC1, APOD/E, GPX3, and MGST1). Myofibroblasts (Fibro_ACTA2), highly expressing gene modules of both fibroblast (e.g., COL1A1, COL3A1, COL5A1/2, COL8A2, and COL11A1) and smooth muscle cells (e.g., ACTA2 and MYL4), are considered key effector cells in SSc (Gabbiani et al., 1972; Denton and Khanna, 2017) but in our analysis comprised only a small fraction (on average 1.05%) of skin fibroblasts.

Interestingly, we identified a highly abundant, uncharacterized fibroblast subset (29.1% ± 4.5% of all fibroblasts in healthy subjects, 95% confidence interval) expressing LGR5. LGR5 is a leucine-rich repeat-containing receptor for R-spondins that activates the canonical Wnt signaling pathway and acts as a stem cell marker of the intestinal epithelium and skin hair follicle (Barker et al., 2007; Jaks et al., 2008). Similar to myofibroblasts and MYOC expressing fibroblasts, the LGR5⁺ fibroblasts express a large gene module involved in the contractile system of smooth muscle and genes important during wound healing (e.g., MYL6, TPM1, ACTB, and ACTG1) in addition to genes encoding for unique ECM components (e.g., COL12A1, MFAP5, PRG4, CILP, SPARC, and LUM) (Figures 3D and 3E; Table S3). The LGR5⁺ fibroblast subset overexpresses many genes associated with ECM degradation and skin remodeling, including metalloproteinase MMP2 and its inhibitors TIMP2, the dipeptidyl peptidase IV (DPP4/CD26), and the secretory leukocyte peptidase inhibitor (SLPI), along with other ECM remodeling/assembly genes (Table S3). Additionally, this subset also highly expresses complement regulatory genes (e.g., CD55, CD59, and CLU), which prevent complement activation and generation of the MAC, genes associated with vascular formation and remodeling, platelet aggregation, and coagulation (e.g., ANGPTL1/2, CTHRC1, PTGIS, THBS2, CD248, PDGFRB/L, and ANXA5) as well as genes important for iron storage and delivery (e.g., FTH1 and SCARA5). The LGR5⁺ fibroblasts express high levels of the Annexin gene family and VAT1 gene, which both play a key role in the trafficking, organization, and exocytosis of vesicles critical for cell-cell communication. Finally, the LGR5⁺ fibroblasts express unique atypical chemokine receptors, including ACKR3, which encodes a receptor for chemokines CXCL11, CXCL12, and opioids (Burns et al., 2006; Meyrath et al., 2020).

Overall, our comprehensive analysis revealed a remarkable degree of functional heterogeneity within skin fibroblasts, including an LGR5 expressing fibroblast subset. Skin fibroblasts express gene modules of diverse functions far exceeding their role as matrix-producing reparative cells, reflecting a wide range of potential functions in fibrosis, inflammation, immune responses, metabolism, angiogenesis, vascular and blood pressure modulation, coagulation, and neurogenesis (Figure 3E).

LGR5 expression marks a unique fibroblast type associated with systemic sclerosis disease

Following the generation of a population-level high-resolution cell atlas of healthy and SSc patients' skin, we searched for significant changes in stromal cell populations and signaling pathways during SSc disease progression. Comparing the stromal cell compartment of SSc patients (42 lSSc and 52 dSSc) with that of healthy subjects (n = 50), we found significant perturbations in fibroblasts, vascular, and pericyte lineages at both the individual and group levels (Figures 4A−4C, and S5A; STAR Methods). Strikingly, the most significant changes in the fibroblast lineage were observed in the LGR5⁺ fibroblast subpopulation, which are substantially reduced in the dSSc patients compared with those in healthy subjects and lSSc patients (Figures 4A−4C and S5A); hence, we define this cell population as ScAFs. Progressive reduction in ScAFs was observed from the early to late stages of the dSSc disease, with the strongest reduction in patients with an early disease or high skin score and in part of the lSSc patients (Figures S5A and S5B). To evaluate if ScAF perturbation is a specific aspect of the SSc fibrotic process or potentially a more general mechanism of loss of tissue structure, we conducted a single-cell analysis of 6 patients with a related fibrotic skin disease, GVHD (5 with sclerodermatous type and 1 with lichenoid type) and 4 patients with localized scleroderma (morphea), and identified a similar perturbation of the ScAF in patients with the sclerodermatous type of GVHD (Figures S5C and S5D). We further validated the perturbation of ScAF in SSc by profiling the stromal cells in the skin of families containing both healthy subjects and SSc patients (Figure S5E). In addition to the reduction of ScAF, we also revealed significant perturbations in other fibroblast populations, including decreased MYOC subsets 1 and increased MYOC subsets 2 in SSc patients compared with healthy subjects (Figures 4A−4C and S5A). Finally, we found a significant increase in the fraction of vascular endothelial cells (expressing CD74 and ACKR1/RBP7) and pericytes (expressing RGS5 and TGFBI), mainly in the dSSc patients with a high skin score compared with other groups (p < 0.05, Figures 4A−4C and S5A).

Figure 4. Perturbation of stromal cell-type composition and transcriptional program in SSc patients.

(A) 2D projection with density plots showing skin stromal cell population enrichment over the metacell model in the various groups. (Left) Reference map of the skin stromal populations as in Figure 3A; (Right) down-sampled skin stromal cells are shown with contour lines indicating the 2D density of projected cells.

(B Bar plots showing the stromal cell-type composition within the healthy donors and SSc patients in each group (left) or in individual patients (right). Cell types are colored the same as in (A).

(C) Dot plots showing percentages of selected stromal cell populations within the healthy and SSc patients’ groups. **p < 0.05, ***p < 0.01, Mann-Whitney U test, two-sided. Error bars indicate mean ± SEM.

(D) Bar plot showing the numbers of upregulated and downregulated genes for individual fibroblast subsets in the SSc groups compared with the healthy donors.

(E) Scatter plots showing gene expression of the ScAFs from the SSc groups (y axis) compared with that from the healthy donors (x axis). Differentially expressed genes (log₂ fold change >1.5) are colored in red with selected genes highlighted.

(F) Dot plots showing LGR5 gene expression in the ScAFs of each donor or patient in the different groups. ***p < 0.001, Mann-Whitney U test, two-sided. Error bars indicate mean ± SEM.

(G) Weighted nearest neighbor analysis (Wnn) Uniform Manifold Approximation and Projection (UMAP) of the CD90⁺ skin cells from 9 healthy donors and 9 dSSc patients with each subpopulation colored the same as in Figure 3A (STAR Methods). The top left corner shows the sources of individual cells.

(H) Clustering dendrogram of cell populations based on ATAC signals (top) and mRNA signals (bottom).

(I) Snapshots showing normalized ATAC-seq signals at selected loci for selected fibroblast subsets from the skin of healthy donors and dSSc patients. Dashed boxes highlight ATAC-seq peaks with differential signals.

(J) Heatmap showing DNA motifs with differential enrichment in skin stromal cell populations from the dSSc patients compared with those from the healthy donors. Selected associated genes are highlighted on the right.

To define molecular perturbations within cell subpopulations in lSSc and dSSc patients, we computed differentially expressed genes between the different SSc groups and healthy subjects for every fibroblast subset (STAR Methods). The most substantial transcriptional changes in the fibroblast lineage were observed in the ScAF subset of the dSSc patients (Figures 4D−4F). In addition to significant downregulation of the LGR5 gene itself in the remaining LGR5⁺ cells sampled from dSSc patients, we identified (mainly in the earlier disease stages) many perturbations in genes and signaling pathways implicated in key pathophysiological processes that drive SSc: fibrosis, vasculopathy, and inflammation (Figures 4E, 4F, and S5F; Table S3). The substantial differences in gene expression of the ScAFs of dSSc patients included upregulation of genes associated with excessive production and deposition of specific ECM components (e.g., COL1A1, COL3A1, COL5A1,2, EFEMP1, BGN, POSTN, and COMP) in parallel with the inhibition of ECM breakdown and uncontrolled tissue proteases activity (upregulation of the SERPINH1, SERPINE2, and PRSS23 genes and the downregulation of the SLPI and TIMP2 genes). Additionally, the ScAFs derived from the dSSc and lSSc patients shared activated profi-brotic signaling programs, including genes associated with the type I IFN signature, TGF-β pathway, and IL1 pathway. Differential regulation of the Wnt signaling associated genes (upregulation: SFRP4 and DKK3; downregulation: WISP2), IGF1 signaling associated genes (IGF1 and IGFBP4), and genes encoding for the CCN family proteins (an acronym for CYR61, CTGF, and NOV), which constitute a signaling hub associated with fibrosis (Jun and Lau, 2011), was distinct to dSSc patient ScAFs. Other gene perturbations in the ScAFs of dSSc patients are associated with increased and dysregulated angiogenesis, generation of abnormal leaky vessels, activation of coagulation and platelet aggregation (i.e., THBS1, PLPP3, and SERPINH1/SERPINE2, respectively), and with abnormal inter-cellular communication and vesicle and granule targeting (i.e., GJA1 and RAB31). Compared with the ScAFs of the lSSc patients, those of the dSSc patients downregulated genes associated with protection against oxidative damage (e.g., GPX3, MGST1, and SCARA5) and adipogenesis (e.g., ADIRF), which correspond to the widespread tissue damage and loss of subcutaneous adipose tissue, respectively, in dSSc patients (Marangoni and Lu, 2017).

Interestingly, the ScAFs in the lSSc and early disease dSSc patients significantly overexpressed genes connected with cellular senescence, including cyclin-dependent kinase inhibitor p16 (CDKN2A) and CDKN1A (p21) (Figures 4E, S5F, and S7E−S7G), consistent with previous studies that show evidence for cellular senescence in SSc (Dumit et al., 2014; Lakota and Varga, 2021).

In addition to the ScAFs, we observed signaling perturbations in other fibroblast subsets in the SSc patients, mainly in the MYOC and PTGDS subpopulations (Figures 4D and S5G). These two subsets overexpressed unique genes (e.g., SVEP1, TIMP1, METAP2, TPP1, CCL19, and IL33) but also share several gene modules with the ScAF subset (e.g., type I IFN and IL1-related genes, CCNs, and ECM-related modules). This resemblance suggests that all of these subsets are affected by the same pathological microenvironment; however, it is also possible that the MYOC and PTGDS subsets transdifferentiate from the ScAFs or compensate for the loss of the LGR5 expressing fibroblasts. Taken together, our analysis defined a distinctive fibroblast subset, ScAFs, which is drastically perturbed both in its abundance and signaling activity in the SSc patients compared with healthy subjects and within the SSc subgroups. The ScAF subset, mainly in the dSSc patients, expresses many genes critical for fibrosis and other essential aspects of SSc pathogenesis and may define a potential molecular diagnostic and therapeutic tool for SSc patients.

ScAF cis- and trans-regulatory circuits in physiology and SSc pathology

Since ScAFs are strongly associated molecularly with SSc pathogenesis, we proceeded with in vivo characterization of the ScAF accessible chromatin regions and TF regulatory circuits and their perturbation in SSc. For this purpose, we performed single-cell genome-wide multiome profiling of fresh fibroblasts including ScAFs and other stromal cells derived from the affected skin of 9 dSSc patients (7/9 with skin score ≥20 and 3/9 with disease duration <1 year) and 9 healthy subjects, simultaneously recovering the mRNA and accessible chromatin regions (by ATAC-seq) from the same individual cells (Figures 4G, S6A, and S6B). We obtained 6,710 and 2,417 QC-positive cells containing both RNA and ATAC signals from the healthy and dSSc samples, respectively. Using label transferring from our RNA cellular reference map, we assigned these cells to predefined clusters, overall retrieving 2,083 ScAFs. The ATAC signal revealed that the same stromal cell subsets identified by RNA expression signatures are also defined by unique chromatin accessible regions (Figure 4H).

To assess the gene regulation programs in skin stromal cells, we computed differential ATAC peaks across individual skin stromal cell populations from healthy subjects. DNA motif enrichment analysis on these peaks revealed that these programs are orchestrated by dedicated TF circuits uniquely expressed in the relevant cell subset (Figure S6C). We then compared the ATAC peaks of individual skin stromal cell populations from the dSSc patients with those of healthy subjects. We detected limited changes at promoter regions (287 regions) but substantial differences at distal chromatin open regions (5,016 regions). These differential regulatory regions are enriched for genes perturbed in the different fibroblast subsets, e.g., LGR5, CDKN2A, and SLPI/MATN4 in the ScAFs (Figure 4I). DNA motif scanning on the differential regions revealed the 63 most significantly overrepresented motifs contributing to the skin stromal cell gene perturbation in dSSc disease (Figure 4J). We found that the DNA motifs for ATF3, JUN, and CREB1 are highly enriched in pathogenic ScAFs (Figure 4J), while regions containing TWIST1, HAND2, MXI1, NEUROD1, PRRX1, and PRRX2 motifs were specifically and significantly repressed in ScAFs from the dSSc patients compared with those from healthy and other skin stromal cells (Figures 4J and S6C−S6E). Some enriched motifs in the pathogenic ScAFs are common to other fibroblast subsets in SSc, suggesting global TF circuitry perturbation (e.g., FOS, FOSL1/2, BACH1, and JDP2) (Figure 4J). In summary, our data uncovered 827 dSSc ScAF-specific regulatory regions that likely play a role in the SSc path-ogenicity of ScAFs. These regions are enriched in the AP-1 motifs, highlighting this TF circuit as a major candidate regulator of the transition from healthy to pathogenic ScAFs. Future studies will focus on the exact mechanism that drives healthy ScAFs into the SSc disease state.

ScAFs present unique morphological and signaling characteristics and are localized in the deep reticular dermis

The complex communication networks between heterogeneous cell types in tissue are mediated in part through ligand and receptor (L-R) signaling connecting signal-transmitting and -receiving cells (Zhou et al., 2018). Examining L-R pairs in single-cell maps can potentially reveal central cellular components shaping tissue fate (Camp et al., 2017; Cohen et al., 2018; Vento-Tormo et al., 2018; Zhou et al., 2018). In order to systematically map cellular interactions and reveal potential communication factors linked with SSc pathogenesis, we characterized L-R pairs between ScAFs and other skin cell types in normal and in disease state as described previously (Vento-Tormo et al., 2018; Figures 5A and 5B; Table S4). We focused on L-R pairs whose ligand or receptor expression was significantly perturbed in SSc patients (Mann-Whitney test, FDR adjusted p < 0.05). Grouping the L-R pairs by functional annotations revealed systemic perturbations in ScAF signaling related to pathways implicated in SSc pathogenesis, including angiogenesis, vascular and blood pressure modulation, coagulation and fibrinolysis, immune dysregulation, complement activation, lipid metabolism, and ECM remodeling (Figures 5A and 5B; Table S4).

Figure 5. ScAF cell interactions and their spatial localization in the dermis.

(A and B) Selected ligand-receptor (L-R) pairs between the ScAFs and other skin cells in healthy donors and SSc patients. L-R pairs are grouped and colored by functional annotations.

(A) L-R interactions in which the ligand (left) expression is perturbed in the ScAFs of SSc patients (red, upregulated; blue, downregulated), with the normalized expression of corresponding receptors in other skin cell types (right). Error bars indicate mean ± SEM.

(B) The same as (A) but the receptor expression (left) is perturbed with corresponding ligands on the right.

(C) Graphical model of the skin layers and ScAFs.

(D) Skin sections from the healthy donors and SSc patients stained with DAPI (4’,6-diamidino-2-phenylindole, blue), anti-hCD55 mAb (green), and smFISH probes for COL1A1 (red) and LGR5 (white) mRNA. Scale bars, 20 μm. Inset scale bars, 10 μm.

(E) Boxplot showing quantification of LGR5⁺ cells in the different dermal layers. *p < 0.1, Mann-Whitney U test, two-sided.

(F and G) Boxplot showing quantification of LGR⁵⁺ cells (F) and LGR5 signals in the LGR5⁺ cells (G) in the reticular dermis of the healthy controls (n = 3), lSSc (n = 4), and dSSc (n = 4) patients. *p < 0.1, ***p < 0.01, Mann-Whitney U test, two-sided.

Within the ScAFs of the SSc patients, we observed the upregulation of different ligands recognized by skin vasculature receptors. The over-expression of these ligands is implicated in the increase of endothelial proliferation and migration, blood vessel formation, and tone modulation (via the VEGFB-NRP1/FLT1, THBS1-CD36, and ADM-CALCRL L-R interactions, respectively). Moreover, we detected perturbed signaling resulting in the dysregulation of coagulation and complement activation (e.g., THBS1-SCARB1, F3-TFPI/IL6, F2R-GNAI2, SERPING1-SELP/SELE, and C3-CD46), in line with the vascular aberrations and hypercoagulative state associated with SSc. Importantly, we identified perturbed signaling between the ScAFs of SSc patients and various skin immune cells. For example, the ScAFs of dSSc patients exhibited downregulation of the non-classical MHC class I antigen HLA-E, involved in immune self-non-self-discrimination through inhibition of activation and killing of the NK and effector T cells via its cognate receptors, encoded by KLRD1 and KLRC1. In addition to their role in the perturbed L-R interactions associated with vascular dysfunction, coagulation, and immune modulation, the ScAFs in SSc patients are key players in other aberrant L-R interactions connected with lipid metabolism (e.g., APOD-LEPR, APOE-SCARB1), matrix deposition (e.g., CTGF-LRP1), and activation of profibrotic signaling, including Wnt and TGFβ signaling (e.g., LGR5-RSPO1/3/4 and TGFBR2-TGFB1, respectively) (Figures 5A and 5B; Table S4). Perturbations of L-R interactions were also observed in the vascular endothelial cells and pericytes of SSc patients. (Figure S6F). Next, we explored the spatial localization of ScAFs in the dermis of healthy subjects and in SSc patients. To visualize ScAFs and spatially localize and quantify them to specific skin niches in both healthy subjects and SSc patients (Figures 5C and S7A), we combined smFISH with immuno-fluorescent staining in whole-mount skin tissue derived from healthy subjects, lSSc, and dSSc patients (STAR Methods). Following the confirmation of a high expression of LGR5 and CD55 proteins in ScAFs by FACS sorting of CD45⁻/CD907⁺/LGR5⁺/CD55⁺ cells (Figure S7B), we stained fixed skin tissue with these ScAF-specific markers (LGR5 mRNAand the CD55 protein) together with the fibroblast signature gene COL1A1 mRNA (Figure 5D; STAR Methods). Surprisingly, we found that in the healthy skin, the ScAFs are scattered mainly in the deep reticular dermis (Figures 5C−5E) and featured long and thin extensions, represented by both CD55 and phalloidin staining of actin filaments (Figures S7C and S7D). In the dSSc patients, ScAF abundance and LGR5 expression on ScAFs were significantly reduced, providing orthogonal validation of our scRNA-seq analysis (Figures 5D, 5F, and 5G). In addition, we found that ScAFs from the dSSc patients are also morphologically perturbed, losing their elongated extensions (Figures 5D and S7D). Additionally consistent with our scRNA-seq data, we detected upregulation of COL1A1 in dSSc ScAFs as wel l as in other fibroblasts in the dermis (Figure 5D). The lSSc patients displayed intermediate ScAF levels and COL1A1 expression in their deep dermis (Figure 5D). We also validated that the ScAFs derived mainly from lSSc patients show significantly elevated levels of the p16 protein (Figures S7F and S7G). Taken together, we found that ScAFs participate in many ligand-receptor interactions with the immune and stromal compartments and that perturbation of these interactions is associated with the major SSc pathophysiological processes, suggesting that ScAFs may constitute a signaling hub coordinating cell-cell communication in the healthy skin. The finding that ScAFs localize specifically in the deep reticular dermis of healthy skin and display a protruding structure suggests that they may normally serve as a guidance scaffold to define the correct tissue organization.

ScAFs are significantly associated with multiple SSc clinical manifestations

SSc is a heterogeneous disease, with variable clinical manifestations encompassing skin fibrosis and other fibrotic traits such as interstitial lung disease (ILD), gastrointestinal (GI) dysmotility and malabsorption, impaired cardiac function, flexion contractures of joints, myopathy, and vascular traits like pulmonary arterial hypertension (PAH), scleroderma renal crisis (SRC), and digital ulcers (DU) (Denton and Khanna, 2017; Royle et al., 2018). There are currently no standard molecular biomarkers for the prediction of these diverse SSc manifestations. To test whether specific disease manifestations (Table S1) are associated with specific molecular features, we analyzed patients' cellular and molecular profiles jointly with their clinical features (STAR Methods). We found very few associations between immune cell subset gene expression and SSc clinical manifestations but high numbers of clinical disease-related genes linked to skin stromal cells, particularly fibroblasts (Figure S7H). Within the fibroblast populations, ScAFs and the MYOC subset 1 population are highly associated with both skin and extra-skin systemic manifestations (Figures 6A and 6B). For example, a high skin score is strongly correlated with low ScAF percentage and high RGS5⁺ pericyte and skin NK cell percentage (Figures 6A and 6B).

Figure 6. Association of cell-type composition and gene expression with clinical manifestations.

(A) Heatmap showing significant associations between cell population fractions and clinical manifestations (STAR Methods).

(B) Scatter plot showing selected association pairs in (A) with each circle representing a donor or patient colored by the group.

(C) Significant associations between gene expression in the ScAFs and clinical manifestations (STAR Methods). Genes associated with at least one significant clinical manifestation are shown on the left (q value < 0.01; STAR Methods) with their normalized expression in the individual donors or SSc patients on the right.

(D) Associations of a selected set of differentially expressed genes in ScAFs with various clinical manifestations.

Focusing on the ScAF subset, we analyzed the genes associated with at least one clinical disease trait and identified 473 upregulated and 280 downregulated genes significantly associated with SSc pathology (Figure 6C). These genes make up four major modules: one represents ScAF genes highly expressed in healthy subjects, two include either upregulated or downregulated genes associated with dSSc patients and a high skin score (Figure 6C), and one contains upregulated genes shared between lSSc and dSSc patients but with no significant skin score associations (Figure 6C). Within these ScAF gene modules, we identified potential biomarkers and therapeutic targets implicated in the key pathophysiological processes of SSc (Figures 6C and 6D; Table S5). For example, EFEMP1 is highly correlated with dSSc and a higher skin score but also associated with GI manifestations and joint contractures (JCS) in both the SSc groups (Figure 6D). Likewise, SMOC2, which encodes a protein-promoting matrix assembly and endothelial cell proliferation (Jang et al., 2020), is correlated with dSSc, a higher skin score, DU, and JCS (Figure 6D). Finally, IL1R1, highly expressed in the ScAFs of SSc patients, is not correlated with a disease subset or a skin score but is positively associated with ILD severity, consistent with previous studies that note the importance of the IL-1 pathway in lung fibrosis (Birnhuber et al., 2019). Interestingly, we further found a positive correlation between IL-1β-producing skin monocytes and ILD (Figure 6A). Together, our data illustrate that diverse dermal pathways and cells, specifically the ScAFs, are significantly associated with SSc clinical traits, and these may serve as targets for more effective biomarkers and therapeutic strategies for SSc.

Discussion

In this study, we report a population-scale single-cell genomic analysis of immune and stromal cells derived from blood and skin samples of the entire clinical spectrum of SSc patients and healthy subjects. We observed immune cell compartment changes mainly in subsets of dSSc patients, corresponding with different SSc skin pathology phases (early inflammatory phase with immune activation versus fibrotic phase). Specifically, we found a significant increase in skin NK cells in the dSSc patients with an early disease or with a high skin score, implicating NK cells in disease initiation. In parallel, we observed significant downregulation of the non-classical MHC class I antigen HLA-E on the pathogenic ScAF, which is important for immune self-non-self-discrimination by NK cells. Consistently, a previous study (Benyamine et al., 2018) highlights the role of NK cells purified from SSc patients as a potent immune inducer of endothelial activation at the early stage of SSc. We further found increased skin pDCs in the dSSc patients with a high skin score. The high IFN I signature present in the blood and affected skin of SSc patients is indicative of the presence of aberrant pDCs. In fact, several studies have confirmed that pDCs largely infiltrate the skin of SSc patients, release high quantities of CXCL4 and IFN-α, and play a major role in determining the severity of the disease (Carvalheiro et al., 2020).

Our study reveals highly diverse fibroblasts with gene expression modules reflecting their wide range of critical functions in the healthy and sclerodermatous skin, including skin remodeling, metabolism, immune modulation, angiogenesis, coagulation, and neurogenesis. While other stromal subtypes have been previously implicated as key players in SSc pathogenicity, mainly αSMA⁺ myofibroblast (Gabbiani et al., 1972; Denton and Khanna, 2017), we could not confirm these phenomena in our SSc cohort.

We discovered a discrete fibroblast subset, ScAFs, that displays elongated extensions, potentially important for coordinating the correct tissue organization and homeostasis, similar to other LGR5-expressing mesenchymal cells (Lee et al., 2017; Halpern et al., 2020). Re-analyzing the published scRNA-seq dataset from scleroderma patients (Tabib et al., 2021) revealed the same ScAF subset although in lower abundance. In the sclerodermatous skin, we found global and significant perturbations in ScAFs, reflected by alterations in their numbers, morphology, interactions with other cells, and their massive transcriptional and regulatory reprogramming of genes related to SSc fibrotic and vascular aberrations. Examining the skin of dSSc patients with short disease duration highlighted the importance of ScAF perturbation in the early disease stage. We found not only a shared gene module in ScAFs from the skin of limited and diffuse SSc patients compared with healthy controls but also a large number of unique molecular features representing each group of SSc patients. These findings support a model of SSc patho-genesis that involves two steps of reprogramming of the healthy skin ScAF from the critical skin homeostatic hubs to key pathogenic cells in SSc (Figure 7). The major differences between the limited (step 1) and the diffuse ScAF (step 2) cells are additional perturbations in the dSSc ScAF ECM deposition pathways in parallel with inhibition of ECM breakdown, activation of angio-genesis, and activation of profibrotic signaling pathways such as JAK-STAT and IGF-1 (Figure 7). Additionally, senescence hallmark genes were upregulated mainly in lSSc and early dSSc ScAF. The upregulation of antioxidant markers in ScAF of lSSc and their downregulation in dSSc compared with controls, in addition to significant downregulation of HLA-E and aberrant CD55 expression on dSSc ScAF, may also contribute to the extensive tissue damage found in the diffuse form.

Figure 7. ScAF cells are regulated through a two-step activation mechanism.

Schematic illustration showing ScAFs switching from homeostatic to stage 1 ScAFs (ISSc) and stage 2 ScAFs (dSSc). The key genes involved in each stage are shown below each condition. The transcription factors with altered DNA-binding landscapes involved in the transition to pathological ScAFs are shown below the key genes. Arrows indicate up (red) or down (green) regulation of the gene/transcription factor (TF) in the specific stage.

Multiome analysis of skin stromal cells revealed the involvement of epigenetic and TF circuits, highlighting a potential role for the AP-1 family members in regulating the transition from healthy to pathological ScAFs (Figure 7). We are approaching an exciting era of large single-cell genomic cohorts of hundreds of patients, combining deep single-cell molecular profiles with diverse phenotypic clinical metadata. Our SSc cohort is currently the largest published single-cell dataset connecting patient cellular, molecular, and epigenetic profiles with their clinical metadata. It enabled us to discover unique associations between diverse genes, pathways, and particular disease complications, which may serve as potential biomarkers for predicting disease trajectories in patients. Our detailed molecular characterization of ScAFs may also provide a foundation for the development of the next generation of drugs that leverage the unique signaling repertoire of these cells to restore their homeostatic phenotype (LGR5 dependence on R-spondin; Lin et al., 2017) or block specific checkpoints. These strategies can open new horizons in the field of fibrotic diseases in general and in SSc in particular.

Limitations of the study

To translate our finding to a mechanistic model of fibrosis and therapy intervention, it is of importance to design more relevant animal and human ex vivo models of SSc that will allow exploration of the initial insult leading to SSc and specifically to the perturbation of ScAFs and other stromal and immune cells. Additional research is needed to determine the role of LGR5 gain- or loss-of-function on ScAF physiology. Finally, given the huge heterogeneity in SSc, a larger cohort of SSc patients and longitudinal follow-up will enable further definition of additional immune subtypes with roles in disease manifestations and progression and additional molecular-clinical correlations and biomarkers.

STAR*METHODS

Detailed methods are provided in the online version of this paper and include the following:

• KEY RESOURCES TABLE

• RESOURCE AVAILABILITY

  • ○Lead contact
  • ○Materials availability
  • ○Data and code availability

• EXPERIMENTAL MODEL AND SUBJECT DETAILS

  • ○Collection of patients skin biopsies and blood samples

• METHOD DETAILS

  • ○Skin dissociation and peripheral blood mononuclear cells (PBMCs) separation
  • ○Flow cytometry single cell sorting for MARS-seq
  • ○Massively parallel single-cell RNA-seq library preparation (MARS-seq 2.0)
  • ○Single cell multiome analysis using Chromium 10x genomics platform
  • ○Whole mount smFISH and immunofluorescence staining

• QUANTIFICATION AND STATISTICAL ANALYSIS

  • ○Analysis of single-cell RNA-seq data
  • ○Metacell modeling for MARS-seq data
  • ○The 2D projection in the metacell modeling for MARS-seq data
  • ○Z-score of the gene expression in fibroblasts from healthy subjects and SSc patients
  • ○Normalized gene expression
  • ○Differential gene expression between healthy subjects and SSc patients
  • ○Processing and analysis of 10x Genomics Chromium Single Cell Multiome ATAC⁺ Gene Expression
  • ○Cell-cell Interaction maps
  • ○Quantification of LGR5 smFISH signal
  • ○Analysis of associations of patients cellular and molecular profiles with their clinical features
  • ○Statistical analysis

Star⋆Methods

Key Resources Table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
CD45 PE-Cy7	Biolegend	Cat #304016; AB_314404
CD45 PerCP Cy5.5	Biolegend	Cat # 304028; AB_893338
CD90 (Thy1)-BV650	Biolegend	Cat # 328144; AB_2734320
LGR5-PE human	Biolegend	Cat # 373804; AB_2686988
LGR5-PE human	Miltenyi biotec	Cat # 130-112-437
LGR5-PE mouse	Miltenyi biotec	Cat # 130-111-389
CD55-FITC	Biolegend	Cat #311306; AB_314863
CD55-UNLB	SouthernBiotech	Cat # 9661-01
P16 (F12)	Santa-crus Biotechnology	Cat # sc-1661
Phalloidin-Alexa Fluor™ 488	Invitrogen™	Cat # A12379
CD26-FITC	Biolegend	Cat # 302704; AB_314288
CD146-PE	eBioscience™	Cat # 12-1469-42; AB_11039529
CD31-F	Biolegend	Cat #303104; AB_314330
Chemicals, peptides, and recombinant Proteins
Ficoll-Paque™ Plus	Cytiva	Cat # 10308060
SmFISH probes (LGR5, Col1A1) coupled to Cy5, TMR or Alexa594 fluorescent dyes	Stellaris FISH Probe Designer Software Biosearch echnologies, Inc., Petaluma, CA
Critical commercial assays
Superscript III	Invitrogen	Cat # 18080-085
Exonuclease I	NEB	Cat # M0293
Second Strand Synthesis module	NEB	Cat # E6111
T7 High Yield RNA Synthesis Kit	NEB	Cat # E2040
Turbo DNase I	Ambion	Cat # AM2239
RNA fragmentation reagents	Ambion	Cat # AM8740
T4 RNA Ligase 1 (ssRNA Ligase)	NEB	Cat # M0204
AffinityScript Multi-Temp RT	Agilent	Cat # 600109
Kapa HiFi HotStart PCR ReadyMix	Kapa Biosystems	Cat # KK2601
Whole Skin Dissociation Kit, human	Miltenyi biotec	Cat # 130-101-540
Gentlemacs dissociator	Miltenyi biotec	Cat # 130-093-235
Gentlemacs tubes	Miltenyi biotec	Cat # 130-096-334
DAPI	Sigma-Aldrich	Cat # D9542
Proteinase K	Ambion	AM2546
SSC Buffer	Ambion	AM9765
Chromium Next GEM Single Cell multiome ATAC + gene expression kit	10x Genomics	PN-000285
Deposited data
scRNA-seq data of human blood and skin	This study	GSE195452
Oligonucleotides
NA	NA	NA
Software and algorithms
MetaCell R package	Baran et al., 2019	https://github.com/tanaylab/metacell/.
Cell Ranger ARC	10x Genomics	NA
Seurat 4.1.0	Hao et al., 2021	https://github.com/satijalab/seurat/
Signac 1.5.0	Stuart et al., 2021	https://github.com/timoast/signac/
FACSDiva 7	BD Biosciences	NA

Resource Availability

Lead contact

Further information and requests for reagents should be directed and will be fulfilled by the lead contact Ido Amit (ido.amit@ weizmann.ac.il).

Materials availability

We did not generate new unique reagents in this study.

Experimental Model and Subject Details

Collection of patients skin biopsies and blood samples

Patients who were diagnosed with SSc based on the 2013 American College of Rheumatology/European League against Rheumatism revised criteria for SSc (Van Den Hoogen et al., 2013), were recruited to the study from rheumatology departments in two medical centers in Israel.

The group of SSc patients in our cohort consisted of 80 female (82.5%). Disease subtype and internal organ involvement were assessed according to established criteria; 44 patients (45.4%) have limited cutaneous SSc (lSSc), and 53 (54.6%) have diffuse cutaneous SSc (dSSc). The mean disease duration from the first non-Raynaud's phenomenon scleroderma symptom is 7.9 years in the dSSc group (+/- 7.36 years, disease duration between 5 month to 26 years), and 10.2 (+/-8.02 years, disease duration between 3 month to 33 years) in the lSSc group. Early disease defined as disease duration of ≤3 years (n=28), 11 dSSc patients were within 1 year of first non-Raynaud's disease manifestation. In the dSSc patients 26 patients (49%) had high skin score (≥20) in the time of skin biopsy as measured by modified Rodnan Skin Score (mRSS). Active disease in our study refers to patients with high skin scores, and/or early in their disease process. Assessment of disease severity was done according to Medsger severity score (Medsger et al., 2003). Several of the patients were taking disease-modifying medications (DMARDs) as indicated in Table S1. SSc patients under biological Disease-Modifying antirheumatic Drugs (bDMARDs) were excluded. After informed consent (in accordance with Helsinki declaration), skin samples were obtained through punch biopsy (4 mm), 10 cm distal to the elbow and placed in saline containing tubes on ice and then transferred to cold FACS buffer (EDTA pH8.0, 2mM, BSA 0.5% in PBS). Whole blood samples were collected at the time of skin biopsy, were placed in EDTA-containing tubes (Beckton Dickenson) on ice and diluted 1:1 with ice cold FACS buffer. Skin and blood samples immediately transported to the lab. As a control, we also collected blood and skin samples from healthy patients following the same procedure.

Method Details

Skin dissociation and peripheral blood mononuclear cells (PBMCs) separation

In order to generate gentle and efficient single-cell suspensions from human skin tissue, we used Miltenyi human whole skin dissociation kit (catalog num. 130-101-540), which combines mechanical dissociation with enzymatic degradation of the extracellular adhesion proteins. In short, first we cut the skin biopsy into small pieces (< 1mm) and transferred them into gentleMACS C Tube containing 435μL of Buffer and 65 μL of enzyme mix. Next, samples were incubated in a water bath at 37 °C for 1h. After incubation we diluted the samples by adding 0.5 mL of cold cell culture medium, tightly closed C Tube and attached it upside down onto the sleeve of the gentleMACS dissociator, and run the gentleMACS Program h_skin_01. After a short centrifugation step to collect the sample material at the tube bottom, we applied the cell suspension to a 70 μm pre-separation filter, placed on a 15 mL tube, and washed with ice cold FACS buffer. Then, we centrifuged the cell suspension at 300xg for 10 minutes at 4°C and aspirated the supernatant completely. Finally, after red blood cell lysis (Sigma) for 5min at 4°C, centrifugation at 300xg for 10 minutes at 4°C, and washing, we resuspended the cells for FACS staining and sorting.

Blood mononuclear cell separation was performed by density centrifugation media (Ficol-paque, GE Life Sciences) in a 1:1 ratio with blood cells. Centrifugation (460g, 25min) was performed at 10°C, and the mononuclear cells were carefully aspirated and washed with ice cold FACS buffer. After red blood cell lysis (Sigma) for 5 minutes at 4°C, centrifugation at 300xg for 10 minutes at 4°C, and washing, we resuspended the cells for FACS staining and sorting.

Flow cytometry single cell sorting for MARS-seq

Skin cell suspension and blood mononuclear cells were stained with the following antibodies: CD45 (PerCp/Cyanine 5.5, 304028, Biolegend), CD90 (PE 328144, Biolegend), LGR5 (PE, 373803, Biolegend), CD55 (FITC, 311306, Biolegend). All FACS antibodies were used with a 1:100 dilution. Samples were filtered through a 40μm strainer before commencing sorting. Single cell sorting was performed using FACSAria Fusion cell sorter, and FACS Symphony S6, cell sorter (BD Biosciences, San Jose, CA). After doublets exclusion, isolated live cells were single-cell sorted into 384-well cell capture plates containing 2μL of lysis solution and bar-coded poly(T) reverse-transcription (RT) primers for single-cell RNA-seq. Four empty wells were kept in each 384-well plate as a no-cell control for data analysis. Immediately after sorting, each plate was spun down to ensure cell immersion into the lysis solution, snap frozen on dry ice, and stored at –80°C until processed. Cells were analyzed using BD FACSDiva software (BD Bioscience) and FlowJo software (FlowJo, LLC).

Massively parallel single-cell RNA-seq library preparation (MARS-seq 2.0)

Single-cell libraries were prepared as previously described (Jaitin et al., 2014; Keren-Shaul et al., 2019). Briefly, mRNA from cells sorted into barcoded cell capture plates and converted into cDNA, later pooled by using an automated pipeline. The pooled sample is then linearly amplified by T7 in vitro transcription, and the resulting RNA is fragmented and converted into a sequencing-ready library by tagging the samples with pool barcodes and Illumina adapters during ligation, RT, and PCR. Each pool of cells was tested for library quality and concentration was assessed as described earlier (Jaitin et al., 2014; Keren-Shaul et al., 2019). Overall, barcoding was done in three levels: Cell barcodes allow attribution of each sequence read to its cell of origin, thus enabling pooling; Unique Molecular Identifiers (UMIs) allow tagging each original molecule in order to avoid amplification bias; and plate barcodes allow elimination of the batch effect.

Single cell multiome analysis using Chromium 10x genomics platform

Cells were bulk sorted on FACS Symphony S6, cell sorter (BD Biosciences) using 100 nozzle and flow rate of 2 for 3 hours and pooled together. 75.000 and 40000 cells were obtained equally from 9 healthy and 9 dSSc patients, respectively, and processed according to 10x Genomics Chromium Single Cell Multiome ATAC+ Gene Expression low cell number protocol according to protocol number CG000365_DemonstratedProtocol_NucleiIsolation_ATAC_GEX_Sequencing_RevB used for cell lysis (0.1x lysis buffer and lysed for 6.5 min) to obtain intact nuclei. Single cell ATAC and RNA-seq libraries were prepared using the Chromium single cell multiome ATAC + gene expression platform (10x Genomics). Nuclei were prepared and counted to ensure quality and concentration. Nuclei were then transposed according to the manufacturer's protocol. Transposed nuclei suspension was loaded onto Next GEM Chip J targeting 5000 nuclei and then ran on a Chromium Controller instrument to generate GEM emulsion (10x Genomics). Single-cell gene expression libraries, as well as single cell ATAC-seq libraries, were generated according to the manufacturer's protocol using the Chromium Next GEM Single Cell multiome ATAC+ gene expression kit. Final libraries were quantified using NEBNext Library Quant Kit for Illumina (NEB) and high sensitivity D1000 TapeStation (Agilent). Each library was sequenced separately on a NovaSeq 6000 instrument using an SP100 cycles reagent kit (Illumina), targeting 25,000 reads/nuclei for ATAC-seq and a minimum of 20,000 reads/nuclei for gene expression.

Whole mount smFISH and immunofluorescence staining

Single-molecule fluorescence in situ hybridization (smFISH) was used to reveal mRNA expression levels in the intact tissue, enabling the zonation patterns of landmark genes (LM) to be measured with a high spatial resolution. Probe libraries were designed using the Stellaris FISH Probe Designer Software (Biosearch Technologies, Inc., Petaluma, CA). SmFISH probes were coupled to Cy5, TMR or Alexa594 fluorescent dyes. First, skin biopsies (4mm of skin punch extends through to the subcutaneous fat) were taken into cold 4% PFA and incubated for 3 hours at 4°C. Then, samples were moved to 30% sucrose solution in cold 4% PFAover-night at 4°C. Samples were then rinsed in PBS and embedded in cryo-molds with Optimal Cutting Temperature (OCT) before freezing. 40-60 μm (for 3D imaging, LGR5 staining) or 10 μm (for 2D imaging, p16 staining) thick sections of fixed skin were cryosection onto poly L-lysine coated coverslips or cell imaging dish (35mm, 170μm, TCT, Eppendorf, 0030740017), respectively, and used for smFISH staining. Next, 40-60 μm\10 μm sections were fixed with cold 4% PFA for 30 minutes\10 minutes, respectively, and underwent gentler fixation and per-meabilization with ETOH 70% for a minimum 2h at 4°C. Then, the skin sections were hybridized with smFISH probe sets (LGR5 conjugated to Cy5, Col1A1-conjugated toalexa594), according to a previously published protocol with some modifications (Halpern et al., 2020). Briefly, tissues were treated for 10 minutes with proteinase K(10 μg/ml Ambion, AM2546)and washed twice with 2xSSC (Ambion, AM9765). Immuno-stained tissues were further permeabilized with either triton 0.5% (for CD55/phalloidin protein staining) or tween 0.2% (for p16 protein staining). Following two washes with 2xSSC, tissues were incubated in wash buffer (20% Formamide Ambion AM9342, 2xSSC) for 90 minutes and mounted with the hybridization buffer (10% Dextran sulfate Sigma D8906, 20% Formamide, 1 mg/ml E. coli tRNA Sigma R1753, 2xSSC, 0.02% BSA Ambion AM2616, 2 mM Vanadyl-ribonucleoside complex NEB S1402S) mixed with the probes. Hybridization mix was incubated with tissues for overnight in a 30° C incubator. Afterthe hybridization, tissues were washed with wash buffer twice at 30 °C. For protein staining, coverslip or cell imaging dishes were incubated with CD55 (1:100)/ p16 (1:50) primary antibody that was added to GLOX buffer (0.4% glucose, 10 mM Tris buffer PH=8, 2xSSC). For immuno-fluorescence staining, unlabeled mouse anti human CD55-UNBL (9661-01, SouthernBiotech, 1:100), or mouse Anti human p16 (F12) (sc-1661, Santa-crus Biotechnology, 1:50) were used as primary antibodies following the post-hybridization washes. Staining for CD55 was followed by CY2 anti-mouse (715-225-151, Jackson ImmunoResearch Labs, 1:100) secondary antibody whereas as secondary antibody for p16 staining, we used biotinylated donkey anti-mouse (715-065-151, Jackson ImmunoResearch Labs, 1:100), followed by streptavidin Cy2 (016-220-084, Jackson ImmunoResearch Labs, 1:100). We further used Alexa Fluor™ 488 conjugated phalloidin (A12379, Invitrogen, 1:500) for immunofluorescence staining of F-actin. When we combined phalloidin with CD55 staining, the secondary Ab for CD55 was CY3 anti-mouse (CN-715-165-151, Jackson ImmunoResearch Labs 1:100,). Lastly, samples were counterstained with 50 ng/ml DAPI (D9542, Sigma-Aldrich) in GLOX buffer.

Imaging of skin punch biopsies of 40-60 μm slices was performed using a Andor Dragonfly Spinning disc confocal imaging system (Andor Technology PLC) connected to a Leica DMi8 microscope (Inverted) (LeicaGMBH). Imaging was performed with a 40m pinhole disk. Slide overview was performed with a20X/0.75 air objective tiles with 10% overlap and then stitched together. Following imaging conditions were used. Channels: DAPI staining excitation 405nm, emission 450/50nm, exposure 100ms; CD55 staining excitation 488nm, emission 495/20nm, exposure 200ms; COL1A1 mRNA staining excitation 561nm, emission 620/60nm, exposure 200ms; LGR5 mRNA staining excitation 637nm, emission 700/75nm, exposure 200ms. Laser power for all lasers was set for 20%. Signal was detected by an sCMOS Zyla (Andor) 2048x2048 camera in 16-bit depth, lens 20X/0.75 air imaging 665μ x 665μ pixel size 0.324μ. High magnification of specific loci were imaged using the same conditions with a 63 x 1.3 Glycerol objective 211μ x 211μ (XY) pixel size 0.103μ. Images were deconvolved with the internal Andor Fusion deconvolution application.

Quantification and Statistical Analysis

Analysis of single-cell RNA-seq data

MARS-seq libraries, pooled at equimolar concentrations, were sequenced using an Illumina NextSeq 500 or NovaSeq 6000 sequencer, at a sequencing depth of 20K-50K reads per cell as previously described (Jaitin et al., 2014; Keren-Shaul et al., 2019). Reads are condensed into original molecules by counting same unique molecular identifiers (UMI). We used statistics on empty-well spurious UMI detection to ensure that the batches we used for analysis showed a low level of cross single-cell contamination (less than 3%). MARS-seq reads were processed as previously described (Jaitin et al., 2014; Keren-Shaul et al., 2019). Reads were mapped to human reference genome hg38 using HISAT (version 0.1.6); reads with multiple mapping positions were excluded. Reads were associated with genes if they were mapped to an exon, using the UCSC genome browser for reference. Exons of different genes that shared genomic position on the same strand were considered a single gene with a concatenated gene symbol.

Metacell modeling for MARS-seq data

To analyze the MARS-seq data from all the samples, we used the MetaCell package (Baran et al., 2019) as previously described (Jaitin et al., 2014; Keren-Shaul et al., 2019; Li et al., 2019; Cohen et al., 2021). Gene features with high variance to mean were selected using the variation-to-mean parameter T_vm > 0.08 and minimal total UMI >100. From the gene features, we excluded high abundance lincRNA and genes linked with poorly supported transcriptional models (such as genes annotated with the prefix "AC [0-9]", "AL [0-9]", etc.). Annotation of the metacell model was done using the metacell confusion matrix and analysis of marker genes. Finally, in the initial clustering, we identified immune cells, stromal cells, and doublets contamination. We next removed doublets contamination and performed second round clustering on immune cells and stromal cells separately. We identified immune or stromal cells using straightforward analysis of known cell type markers (e.g., COL1A2 – fibroblast cells, TRAC − T cells, C1QA - macrophages, S100A8 - monocytes, and more), and cross-validation with the existing data from previous works. In the final clustering, the gene feature selection strategy described above retained a total of 2,202 genes from immune cells and 2,026 genes from stromal cells for the computation of the Metacell balanced similarity graph. We used K = 500 bootstrap iterations. Metacell splitting was performed by clustering the cells within each metacell and splitting it if distinct clusters are detected. We examined batch effect by comparing the multiple plates from the same sample processed in the same batch or across different batches and cells from the same population but different healthy subjects, and confirmed our method displayed no clear batch effect and our analysis is robust.

The 2D projection in the metacell modeling for MARS-seq data

The 2D projection is computed using a force-directed layout algorithm on a balanced similarity graph. First, a raw similarity matrix is generated by computing Pearson's correlations on the transformed UMI counts of features (genes). Next, the raw similarity matrix is used to generate a weighted adjacency matrix for a weighted, directed cell graph G, in which a heavy edge from cell i to cell j indicates strong attraction of the former to the latter. Then, the balanced similarity graph G is partitioned into metacell graph G^M using an adaptation of k-means to graphs with bootstrapping. Finally, the coordinates for each metacell in the 2D projection are computed by applying a standard force-directed layout algorithm to the metacell graph G^M, and the cells are positioned by averaging the metacell coordinates of their filtered neighbor cells in the original balanced graph G.

Z-score of the gene expression in fibroblasts from healthy subjects and SSc patients

In order to provide robust estimates of gene expression between fibroblast metacells, we calculated a z-score for each gene of each fibroblast metacell according to the following steps. First, for a given fibroblast metacell, we randomly selected 100 cells from the metacell and 100 cells from all fibroblast cells. Second, we computed a p-value using Mann-Whitney test for each gene comparing the normalized UMI counts in the 100 cells from the metacell and the 100 cells from all fibroblast cells. This p-value then was converted into a z-score. For each fibroblast metacell, we performed the previous two steps 100 times to obtain 100 z-scores for each gene, and we calculated the average value of the 100 z-scores to represent the relative expression of each gene in the given fibroblast metacell compared with the whole fibroblast population. We performed the same calculation for all the fibroblast metacells.

Normalized gene expression

To calculate the normalized gene expression per metacell or per cell population, we first down-sampled total UMI counts of individual cells to the same level (n = 250 for immune cells and n = 400 for stromal cells). To account for the variable cell numbers, for each metacell or a population in each patient, we randomly sampled a fixed number of cells and calculated the average UMI counts for each gene. Lastly, we performed the previous step 100 times to estimate the means and distributions of gene expression in each metacell or cell population from a given subject, and finally use them as normalized gene expression.

Differential gene expression between healthy subjects and SSc patients

We used the rich single cell gene expression information and performed differential gene expression analysis by comparing cells of the same subset from healthy control group and SSc patient groups. To account for the variable cell numbers between individual patients, the same number of cells were sampled from each patient for a given cell population. In order to obtain comparable results across different cell populations, we also controlled the total number of cells we sampled for a given population. A z-statistics was computed for each gene using Wilcoxon rank sum test on the down-sampled UMI counts from the cells of healthy control group and two SSc patient groups. Z-statistics Z_mean for each gene was calculated by randomly repeating 100 times the sampling procedure. Z_mean were converted into p values, and then we performed multiple testing correction on the p values with a standard Benjamini-Hochberg procedure and controlled for false discovery rate (FDR) q-value < 0.05.

Processing and analysis of 10x Genomics Chromium Single Cell Multiome ATAC⁺ Gene Expression

Low-level read processing and alignment was performed with the Cell ranger ARC/2.0.0 tool. Data integration of RNA-seq data from 10x multiome with our MARS-seq reference atlas was done using label transfer. In short, we used a Knn-classifier (k=50) using the Pearson's correlation as a similarity metric over the normalized gene features defined for the reference model. The distribution of cluster memberships over these k-neighbors was used to associate the new cell with a reference metacell (by majority voting) and to project the cell in two dimensions by weighted average of the linked reference clusters' mapped x and y coordinates. We then performed WNN analysis with our 10x multiome data from healthy controls and dSSc patients using Seurat 4.0.6 and validated the label transfer from our MARS-seq reference atlas (Hao et al., 2021; Stuart et al., 2021). Specific parameters include min.dist = 0.5 for “RunMAP” and resolution = 1.2 for “FindClusters”. ChromVAR (1.16.0) was used for the analysis of motif accessibility on the merged 10 multiome ATAC-seq peaks from the ARC tools. Presto (1.0.0) was used to identify differential peaks and differential expressed genes. We defined differential ATAC-seq peaks with at least 5% cells with non-zero signals in at least one of the cell groups, FDR < 0.05, and logFC > 0.2. We used JASPAR2020 motif modelsfor DNA motif scanning. We used gene models from EnsDb.Hsapiens.v86 and defined promoter regions as within 1Kb from transcription start site (TSS) distal chromatin regions as 50Kb away from TSS.

Cell-cell Interaction maps

To visualize all skin interactions, we used a published dataset of ligand and receptor pairs (Ramilowski et al., 2015). We applied a lenient filtering, including all LR with > 0.2 mean detected UMIs in at least one meta-cell (normalized to meta-cell size). We applied FDR-adjusted Mann-Whitney test to screen for ligands and receptors whose mean expression within ScAF or pericytes/vascular cells is significantly altered in dSSc or lSSc patients compared to healthy controls.

Quantification of LGR5 smFISH signal

After excluding images containing hair follicles, z-stacks used to result in 113 z-stacks in total (35 stacks from 3 control donors, 41 from 4 lSSc patients, and 37 from 4 dSSc 4 patients). The first 30 slices of a z-stack were used for quantification. For the LGR5 mRNA channel, the 30 slices were maximum intensity projected by every three non-overlapping and adjacent slices, resulting in 10 merged slices. The DAPI channel was maximum intensity projected to one image. Maximum intensity projected images were fed into a pixel classification and object detection pipeline in Ilastik (Berg et al., 2019) resulting in object classified images. The classification pipelines were trained for the LGR5 and DAPI channels separately. Multiple maximum intensity projected images from each patient group were used for manual training (DAPI: 3 control, 5 lSSc, 4 dSSc; LGR5: 8 control, 5 lSSc, 7 dSSc) to identify the LGR5 signal spots and DAPI nuclei. Next, based on the object-classified images, the Euclidian distance between the centroids of LGR5 spots and DAPI nuclei were calculated. Each LGR5 spot was assigned to the closest nucleus. LGR5 spots located further than 80 μm from any nucleus were excluded. Image pre-processing was done with ImageJ. The calculation of Euclidian distances, the quantification and grouping of FISH signal was done with a custom Matlab script. Inferential statistics was done with R.

Analysis of associations of patients cellular and molecular profiles with their clinical features

To test whether a specific disease manifestation is associated with specific molecular features, we performed linear regression with the lm function in R, using patients cell type abundances and gene expression profiles from the scRNA-seq data with their clinical features. For the cell type abundances, we removed immune cell populations that are detected in less than two patients in a given source, and we only considered fibroblasts, pericytes, and vascular endothelial cell populations in the linear regression analysis. For the gene expression, for each population, only the data from the patients with a fixed minimal number of cells (20 cells, we also tested with 15 cells and the results were similar) were used in the analysis. To account for the variable cell numbers between individual patients for the same population, normalized gene expression was used. We performed linear regression for every expressed gene in a given population or a pooled cell linage and performed multiple testing correction on the p values with a standard Benjamini-Hoch-berg procedure and controlled for false discovery rate (FDR) q-value < 0.05.

Statistical analysis

Statistical analysis Data was presented as mean (±SEM) of three independent experiments. Comparisons between two groups of samples were evaluated using the Mann–Whitney U test. All P-values reported were two tailed and statistical significance was defined as P less than 0.05. All statistical analyses were conducted using R software (R Foundation for Statistical Computing, Vienna, Austria).

Highlights.

• Population-scale scRNA-seq of skin and blood of scleroderma (SSc) disease spectrum

• Immune cell composition changes are confined to a specific subtype of dSSc patients

• SSc is defined by global perturbation of a novel scleroderma-associated fibroblast (ScAF)

• ScAF coordinates the signaling pathways implicated in key processes driving SSc

Data and code availability

Single cell RNA-seq data that support the findings of this study is publicly available for download in the NCBI Gene Expression Omnibus (GEO) with accession number GEO: GSE195452. Source code used for scRNA-seq analysis is available at https://bitbucket.org/amitlab/.

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.