Abstract
Objective
Systemic sclerosis (SSc) is characterized by dysregulated immunity, microvascular damage, and progressive fibrosis. SSc-related interstitial lung disease (SSc-ILD) is the leading disease specific cause of mortality; it is therefore important to identify biomarkers associated with SSc-ILD. We hypothesized that the skin transcriptome would provide mechanistic clues that could further our understanding of lung impairment in SSc-ILD.
Methods
We conducted a multi-stage analysis of SSc transcriptomes using weighted gene correlation analysis, gene set enrichment analysis and single cell RNA sequencing.
Results
We identified seven modules correlated with skin and lung function involvement. Unsupervised clustering of these genes classified patients into two groups, one of which showed severe multiorgan involvement characterized by increased skin fibrosis and decreased lung function (p < 0.05). The KRAS pathway was enriched in the severe cluster. Deconvolution of the severe cutaneous signature in SSc skin and lungs showed increased expression in stromal cells and monocytes whereas the KRAS pathway was enriched in stromal cells only. Profiling KRAS and downstream signaling in lungs from SSc-ILD demonstrated activation of these pathways (p < 0.05).
Conclusion
We found a severe cutaneous transcriptomic signature of SSc associated with multiorgan involvement. By deconvoluting the severe cutaneous signature, we have identified pathways enriched in specific cell types including KRAS in stromal cells, and a combination of stromal and immune cells in the severe signature. The severe SSc signature can provide a framework for understanding shared mechanisms in skin and lung fibrosis and may help identify additional molecular pathways.
Introduction
Systemic sclerosis (SSc) is characterized by dysregulated immunity, microvascular damage, and progressive fibrosis of the skin and internal organs. Skin involvement is often the first manifestation of SSc(1). Over half of patients with SSc suffer from interstitial lung disease (SSc-ILD) and it is the leading cause of death(2). However, it is difficult to predict which patients will develop SSc-ILD and at what rate lung function will decline. Early lung involvement is often accompanied by non-specific symptoms, and pulmonary function tests and routine plain radiographs do not always detect it, resulting in treatment delays.
According to the extent of cutaneous involvement, SSc can be classified as diffuse cutaneous (dSSc) or limited cutaneous (lSSc)(3). Although dSSc has a higher risk of ILD than lSSc, neither subtype can predict disease severity or progression. Additionally, ILD progression is not consistently associated with age, sex, autoantibody positivity, or extent of lung parenchymal involvement(4). Repeated damage to epithelial cells, endothelial damage, activation of innate and adaptive immune systems, and recruitment of fibroblasts and myofibroblasts are among the most well-known mechanisms involved in SSc-ILD(5,6). It remains unclear, however, which molecular mechanisms are shared by skin and lung fibrosis in SSc.
Transcriptomic profiling of the skin from people with SSc has shown the association between specific signaling pathways and skin involvement(7–9). In SSc skin, increased expression of genes involved in extravasation and adhesion of inflammatory cells has been associated with SSc-ILD(10). Despite these findings, there is limited evidence that other specific cells or pathways are also involved in both cutaneous and lung fibrosis.
Given the significant morbidity and mortality of SSc-ILD and our incomplete understanding of the shared molecular determinants of skin and lung involvement in SSc, we need to uncover the mechanisms behind organ involvement in SSc. We hypothesized that the skin transcriptomes of SSc could reveal similar mechanisms affecting the lungs in SSc-ILD. To test this hypothesis, we leveraged bulk skin transcriptomes in two cohorts to identify genes associated with cutaneous involvement in SSc. Next, we used two independent SSc cohorts profiled with single-cell RNA sequencing (scRNAseq) to dissect the cellular components of the severe cutaneous signature. Lastly, we used protein expression in SSc-ILD lungs to validate a transcriptomic pathway activated in the cutaneous SSc signature.
Methods
Study Populations
We used gene expression data from the Gene Expression Omnibus (GEO) database. GSE47162 included bulk RNA gene expression from the skin of 59 patients with SSc along with information on sex, immunosuppression, race, antibody profiles, modified Rodnan Skin Score (mRSS), forced vital capacity (FVC) and diffusing capacity of carbon monoxide (DLCO)(10). Patients with SSc were enrolled in either the Genetic versus Environment in Scleroderma Outcome Study (GENISOS) or a Phase I/IIa open-label Imatinib study (baseline visit)(11,12). All subjects in this study fulfilled the American College of Rheumatology preliminary classification criteria for SSc or demonstrated at least 3 of 5 CREST symptoms (calcinosis, Raynaud’s phenomenon, esophageal dysmotility, sclerodactyly, telangiectasias), and the presence of sclerodactyly was mandatory(10). Moderate-to-severe ILD, defined as FVC < 70% and high-resolution computed tomography (HRCT) scan consistent with ILD was present in 23 (39%) subjects.
GSE106358 included gene expression from skin biopsies and mRSS scores of patients treated with placebo who participated in Roche’s faSScinate phase 2 study of tocilizumab in SSc(13).
GSE138669 contained single cell RNA sequencing (scRNAseq) data from the skin of 12 patients with diffuse SSc (dSSc) and 10 healthy controls along with information on sex, duration of disease, mRSS and immunosuppression(14,15).
GSE128169 contained scRNAseq data from 13 explanted lung tissues of 4 SSc-ILD patients and 4 healthy controls(16).
Determination of Participant Race and Ethnicity
Patients in the GENISOS study self-reported race and ethnicity information(11). This information was not available for the remaining studies.
Statistical Analysis
All statistical analyses were performed using R software version 4.2.2.
Weighted Gene Correlation Network Analysis
Weighted Gene Correlation Network Analysis (WGCNA)(17), a method to identify clusters of highly correlated genes (modules), was implemented using dataset GSE47162. We used skin transcriptomes to identify the gene modules and used the patient’s clinical characteristics as external features. Following normalization with NormExp background correction and normalization using control probes in limma, genes and samples with excessive missing values were removed and genes expressing a log2(intensity) > 5 in more than 10% of samples were kept for subsequent analyses. Genes with a SD of zero were removed. Gene modules derived from 19,210 remaining transcripts were correlated with phenotypic characteristics across 59 patient samples.
Metacore Enrichment analysis
Integrated pathway analysis of the gene modules was conducted using the MetaCore (Clarivate) version 03/01/2024. FDR < 0.05 was considered significant.
Unsupervised hierarchical clustering to determine severe SSc signature
Similarity Network Fusion (SNF) (K = 20, alpha = 0.5, T = 15) was used for unsupervised hierarchical clustering of genes in modules that were correlated with mRSS, FVC and/or DLCO(18). Samples were hierarchically clustered using average linkage on the distance matrix. Inspection of the dendrogram revealed a distinct separation at a cut height of 0.9845, beyond which merging branches led to a sharp increase in within-cluster dissimilarity. Cutting the tree at this threshold divided the cohort into two well-defined groups, capturing the major bifurcation in sample relationships while preserving intra-cluster homogeneity and biological interpretability. Comparisons of phenotypic data between the two clusters were performed using Wilcoxon rank sum test for continuous data, Pearson’s Chi-squared test for categorical data, and Fisher’s exact test for proportions. Transcriptome-wide differential gene expression analysis was performed between the two clusters using Wilcoxon Rank Sum score test. FDR < 0.05 was considered significant.
Elastic-net regression of validation dataset
Elastic-net regularized generalized linear model was fitted with the glmnet package v4.1–7 in R. Prior to modelling, all baseline skin transcriptomes in GSE106358 (n=39) were z-score standardized to place the coefficients on a common scale and mRSS was used as the outcome variable. Genes with non-zero coefficients from the final model were identified as transcriptomic predictors. We used K-means clustering with k=2 to cluster skin samples based on the transcriptomic predictors identified with the elastic-net model.
Gene Set Enrichment Analysis
Gene Set Enrichment Analysis (GSEA) version 1.70.3 was used to analyze the two clusters derived from unsupervised clustering(19). Pre-ranked genes were compared to the Molecular Signatures Database Human C2, a curated gene set composed of previously defined molecular pathways. Normalized enrichment scores were reported, and significance was defined as FDR < 0.05.
Single Cell Analyses of skin and lung datasets
Analyses of scRNAseq data from GSE138669 and GSE128169 were performed with Seurat version 5.0.3(20). A Seurat object was created for both datasets and data were normalized. Cell cycle and mitochondrial gene signals can obscure real biological differences if they reflect technical or general variation. Removing or controlling for them makes clustering, trajectory analysis, and interpretation more reliable. These genes were identified and adjusted for in the datasets.
Skin RNAseq annotation, GSE138669, was performed using previously identified skin cell types(21). This dataset contained punch biopsies from the dorsal mid-forearm skin of 12 patients with dSSc and 10 healthy controls.
ScRNAseq annotation of the lung, GSE128169, was completed with Azimuth(22). GSE128169 dataset was derived from 13 explanted lung tissues from 4 SSc-ILD patients and 4 healthy controls (56,196 cells; HC=21,768, SSc-ILD=34,428). Separate upper and lower lobe samples were obtained from all four SSc-ILD patients and from one healthy control(16).
Principal components analysis (PCA) was performed, and 70 significant principal components (PCs) were chosen using Jack Straw for downstream cell clustering and visualization using the uniform manifold approximation and projection (UMAP) technique.
Deconvolution of the bulk skin signature with Cibersortx
Cibersortx (cibersortx.stanford.edu) was used to deconvolute the previously identified severe skin transcriptomic signature in bulk RNA profiling using the average expression of genes in skin cells identified in the analysis of the scRNAseq skin dataset (GSE138669). Cibersortx analysis was performed in the relative mode after disabling batch correction and quantile normalization(23). Cluster values resulting from this deconvolution were compared using the Wilcoxon test.
AddModuleScore analysis of scRNAseq
To identify cell types expressing specific sets of genes in the severe cutaneous signature as well as other molecular pathways identified from our analysis of GSE47162, we employed the AddModuleScore function to explore the scRNAseq datasets. AddModuleScore is a Seurat function that calculates the gene expression level of a cluster of genes on a single cell level and then subtracts the aggregated expression of a randomly assigned control set. Upregulated and downregulated genes were analyzed separately. These were visualized using the UMAP technique as well as Violin plots with bootstrap for nonparametric confidence intervals. Expression of individual genes of the severe signature and other highlighted pathways by cell type were represented between SSc-ILD lungs and controls using dot plots. AddModuleScore values were compared between conditions using a t-test, with significance defined as p < 0.05.
Split dataset analysis
Because the combined lung scRNA-seq dataset contained multiple lung samples from the same patients, we assessed whether the observed findings were robust to patient-level sampling. We generated split datasets in which only a single lung sample per patient was included. Using a custom R script with a random number generator, we randomly selected one SSc-ILD sample from each patient and created two independent datasets, each of which was compared against the same set of control samples. GEO accession numbers Dataset 1: SSc-ILD (GSM3666106, GSM3666107, GSM3666101, and GSM3666104); control (GSM3666096, GSM3666097, GSM3666098, and GSM3666100). GEO accession numbers Dataset 2: SSc-ILD (GSM3666102, GSM3666103, GSM3666105, and GSM3666108); control (GSM3666096, GSM3666097, GSM3666098, and GSM3666100).
Immunofluorescence of SSc-ILD lungs and controls
De-identified lung tissues, a kind gift from Dr. Carol Feghali-Bostwick, were previously collected from healthy donors and patients with SSc using informed consent on protocols approved by the Institutional Review Board at the University of Pittsburgh (IRB 970946). Normal lung tissues were donor lungs that were deemed unsuitable for transplantation. SSc-ILD lung tissue were lung explants at the time of transplantation. Using xylene, paraffin-embedded lung tissue slices from SSc-ILD and healthy subjects were deparaffinized and rehydrated using decreasing strengths of ethyl alcohol (100%, 90%, 80%, 70% and 50%). Epitopes on the cell surfaces were unmasked by steaming slides with 10X antigen retrieval buffer (ThermoFisher) for 45 mins. After a PBS wash, the area around tissues was dried and encircled using a hydrophobic pen, followed by incubation with a blocking solution (Abcam) for an hour. Human primary antibodies were diluted in an antibody diluting solution to obtain 1:100 concentration and incubated overnight at 4 degrees Celsius. The following primary antibodies were used in this study; anti-KRAS (Invitrogen, Cat. #415700), anti-total AKT (Cell Signaling Technology, Cat. #4691), anti-phospho AKT (Cell Signaling Technology Cat. #4060), anti-total ERK1/2 (Cell Signaling Technology, Cat. #4695), anti-phospho ERK1/2 (Cell Signaling Technology, Cat. #4370). Following washing with PBS, tissue slides were incubated with the Alexa 594 anti-rabbit secondary antibody (1:1000 dilution, ThermoFisher) for 2 hours. Slides were washed with PBS, and a coverslip was applied on the slide after adding VECTASHIELD® Antifade Mounting Medium with DAPI (Vector laboratories). Images were obtained under 63X magnification on a Stellaris 5 confocal microscope.
Barplots reflect relative fluorescence units (RFUs) from immunofluorescence slides from each of the tissue slices. RFUs were collected from regions of interest (ROI) with standardized image acquisition and intensity to enable comparison between groups. Slides were imaged under identical exposure, gain, and laser settings to ensure signal comparability. RFUs were normalized either to DAPI signal to control for cell number and staining variability. The resulting normalized fluorescence values were aggregated across multiple fields to generate group-level means to assess differences in protein expression between experimental groups.
Results
Transcriptomic networks are associated with skin and lung involvement in SSc
Skin involvement is a major feature of SSc, therefore identifying clinical associations with skin transcriptomes may improve the chance of finding shared molecular pathways implicated in skin and lung involvement. To test this hypothesis, we conducted a multi-stage analysis of SSc transcriptomes (Supp. Fig. 1). First, we performed WGCNA of cutaneous bulk transcriptomes and its corresponding clinical features in GSE47162 to determine whether specific gene networks were associated with skin involvement, measured by mRSS, or lung function impairment, measured by decreased FVC and DLCO(10). Subjects in this cohort were predominantly female (76.3%) with a mean age of 52 years and a mean disease duration of 7.6 years at the time of skin biopsy. Anti-topoisomerase antibodies, anti-centromere antibodies, and anti-RNA polymerase III antibodies were present in 28.8%, 10.2%, and 27.1% of patients, respectively. FVC <70%, an indicator of lung function impairment, was present in 23 subjects (39%) (Supp. Table 1).
We identified 31 WGCNA modules, of which 12 correlated with mRSS (Fig. 1A, Supp. Fig. 2). Seven of the 12 WGCNA modules correlated with mRSS were also associated with lung function, DLCO (n=7) and FVC (n=1). Five of the seven modules (black, purple, midnight blue, dark orange, and dark turquoise) correlated with worse skin involvement, defined by higher mRSS, and worse lung function, defined by lower DLCO, whereas the remaining two (steel blue and light yellow) had the opposite correlations with skin involvement and lung function.
Figure 1.
Multiple gene networks are associated with skin and lung involvement.
(A) Weighted gene correlation network analysis of skin gene expression and phenotypic data in patients with systemic sclerosis (n = 59 samples).
Positive correlations are red, negative correlations are blue. The dark turquoise, purple, midnight blue, black and orange modules are associated with higher mRSS (more severe skin involvement) and lower DLCO % predicted (more severe lung impairment). The steel blue and light-yellow modules are associated with lower mRSS (less severe skin involvement) and higher DLCO % predicted (less severe lung impairment). Values are correlation coefficient (nominal P value).
(B) Metacore pathway enrichment of modules correlated with increased mRSS and decreased FVC or DLCO. Metacore pathway enrichment of modules associated with decreased mRSS and decreased FVC or DLCO). DLCO=diffusing capacity of the lungs for carbon monoxide; FVC=forced vital capacity; mRSS=modified Rodnan Skin Score; anti-RNP=anti-ribonucleoprotein; ATA=anti-topoisomerase I; anti-RNAP=anti-RNA polymerase III; ACA=anti-centromere antibody.
WGCNA hub genes were defined by absolute module membership (MM) >0.7. This analysis identified hub genes across multiple modules: 243 in black, 28 in darkturquoise, 40 in lightyellow, 82 in midnightblue, 27 in orange, 158 in purple, and 15 in steelblue. Gene ontology analysis of the hub genes revealed module-specific themes, with the black module primarily associated with regulation of cell migration, darkturquoise with the uropod, lightyellow with keratinization, midnightblue with the immunological synapse, orange with keratohyalin granules, purple with collagen-containing extracellular matrix, and steelblue with melanosome (Supp. Table 2).
To further identify pathways associated with skin and lung involvement in SSc, we performed pathway enrichment analysis of the seven WGCNA modules. The five modules correlated with worse skin and lung involvement were enriched for pathways involved in extracellular matrix processes, wound healing, collagen binding, regulation of the immune response and cytokine production (Fig. 1B). In contrast, the two remaining modules included pathways responsible for normal skin function, pigment and melanin biosynthesis and cell survival.
Two skin transcriptomic clusters are characterized by distinct skin and lung involvement in SSc
To define subgroups of patients with different levels of skin and lung involvement, we applied unsupervised hierarchical clustering to genes included in the seven WGCNA modules (n=1,219). This analysis led to the identification of two subgroups of patients with distinct clinical characteristics. (Table 1, Fig. 2A). Patients in cluster 1, n=30, had less severe clinical characteristics, including a lower mRSS (p<0.001), a higher DLCO (p=0.034), and a higher FVC (p=0.031) (henceforth, non-severe). In contrast, patients in cluster 2, n = 29, showed worse lung function (lower DLCO and FVC) and a higher mRSS (henceforth, severe) (Figs. 2B–D). Subjects in the severe cluster also exhibited a significantly higher frequency of anti-RNAP positivity (p = 0.003, Table 1), which is noteworthy due to its link with accelerated skin thickening and an increased risk of complications(24).
Table 1.
Unsupervised hierarchical clustering of dataset GSE47162 identifies two groups of patients with different degrees of severity.
| Characteristic | 1, N = 301 | 2, N = 291 | p-value2 |
|---|---|---|---|
|
| |||
| Race | 0.9 | ||
| African American | 5 (17%) | 4 (14%) | |
| Asian | 1 (3.3%) | 0 (0%) | |
| Hispanic | 4 (13%) | 6 (21%) | |
| White | 20 (67%) | 19 (66%) | |
| Age | 58 (40, 63) | 52 (46, 58) | 0.6 |
| FVC %predicted | 81 (70, 93) | 70 (51, 85) | 0.031 |
| DLCO %Predicted | 67 (60, 82) | 52 (42, 70) | 0.034 |
| Modified Rodnan Skin Score (mRSS) | 6 (4, 11) | 20 (15, 27) | <0.001 |
| Number of Positive Antibodies | 0.007 | ||
| 0 | 11 (37%) | 6 (21%) | |
| 1 | 3 (10%) | 0 (0%) | |
| 2 | 8 (27%) | 9 (31%) | |
| 3 | 3 (10%) | 13 (45%) | |
| 4 | 5 (17%) | 1 (3.4%) | |
| Anti-RNP | 3 (10%) | 0 (0%) | 0.2 |
| Anti-ATA | 8 (27%) | 9 (31%) | 0.7 |
| Anti-RNAP | 3 (10%) | 13 (45%) | 0.003 |
| Anti-ACA | 5 (17%) | 1 (3.4%) | 0.2 |
| FVC Abnormal | 14 (47%) | 18 (62%) | 0.2 |
| DLCO Abnormal | 19 (63%) | 23 (88%) | 0.030 |
| Immunosuppression | 6 (20%) | 8 (27%) | 0.49 |
n (%); Median (IQR)
Fisher's exact test; Wilcoxon rank sum test; Pearson's Chi-squared test
Immunosuppression: Patients treated with immunosuppressive agents or corticosteroids (exception prednisone equivalent dose ≤ 5 mg or hydroxychloroquine).
Figure 2.
Unsupervised hierarchical clustering based on cutaneous signature identifies two distinct groups of systemic sclerosis patients characterized by differences in mRSS, FVC% predicted, DLCO % predicted and pathway enrichment.
(A) Heat map of top 50 differentially expressed genes (DEGs) between these two groups. Group 2 demonstrated clinical features of more severe disease including: (B) higher mRSS; (C) lower DLCO % predicted; (D) FVC % predicted. (E) Gene set enrichment analysis (GSEA) shows upregulation of pathways involved in extracellular matrix organization, KRAS activation, and epithelial to mesenchymal transition in the more severe group. (F) Running enrichment plots of Sweet KRAS Targets Up pathway (NES: 1.73, FDR: 0.03). Red represents correlation with more severe disease, blue represents correlation with less severe disease.
We identified 974 differentially expressed genes (DEGs) between the non-severe and severe clusters, henceforth the severe SSc signature (Fig. 2A: top 50 DEGs, Supp. Table 3). Enrichment analysis with GSEA showed increased expression of extracellular matrix production and immune activation pathways in the severe cluster compared with the non-severe cluster (q-value <0.05 and NES score >0.5) (Fig. 2E). In addition, GSEA analysis of the severe cluster revealed a significant enrichment of the Kristen rat sarcoma viral oncogene homologue (KRAS) pathway (NES: 1.73, FDR: 0.03) (Figs. 2E–F) and downstream signaling pathways (Supp. Fig. 3). Furthermore, the KRAS pathway included multiple genes involved in extracellular matrix production and growth factor signaling (Supp. Table 4).
Independent validation of cutaneous transcriptomic predictors of mRSS
We applied elastic-net regression to an independent validation cohort of skin transcriptomics, GSE106358(13), to assess the utility of genes in the severe SSc signature to predict mRSS. We identified 127 genes (Supp. Table 5), which were then used for clustering of the baseline skin transcriptomes (Supp. Fig. 4A). Clustering partitioned the samples into two transcriptomic groups with distinctive mRSS (Supp. Fig. 4B), reproducing the cluster structure observed in the discovery cohort. To further reinforce this validation, we conducted gene ontology analysis on genes upregulated in the severe cluster (Supp. Table 6). This analysis revealed three predominant biological themes: (1) immune activation and inflammation, (2) cell motility and adhesion, and (3) signaling and tissue remodeling. These enriched pathways provide additional evidence that, beyond reproducing the cluster structure and predicting mRSS, the underlying molecular mechanisms are consistent between the discovery and validation cohorts.
The severe SSc signature is highly expressed in skin fibroblasts and monocytes
To determine which specific skin cells were enriched for the severe SSc signature we used a scRNAseq dataset of SSc skin, GSE138669(14,15). Of the 12 patients with SSc in this cohort, 58.3% were female, the average duration of disease was 2.42 years, the average mRSS was 26.2, and 75% were on immunosuppressant medications at the time of the skin biopsy (Supp. Table 7).
We identified 24 cell types. Among the most prominent cells, fibroblasts were clustered into six groups, including fibroblast, fibroblast CCL-19-APOE, fibroblast SFRP2-WIF-1, fibroblast POSTN-ASPN, fibroblast SFRP2-PCOLCE-2 and fibroblast ANGPTL-myofibroblast (Supp. Fig. 5A). The fibroblast groups of SSc subjects and healthy controls were separated (Supp. Fig. 5B). In addition, skin fibroblasts and CD14+ monocytes from SSc subjects expressed higher levels of genes upregulated in the severe SSc signature (Figs. 3A–B). Conversely, downregulated genes in the severe SSc signature were mainly expressed in epidermal cells and melanocytes (Supp. Figs. 5C–D). Cibersortx analysis confirmed this enrichment pattern (Fig. 3C). Together, these results show enhanced expression of the severe SSc signature in skin fibroblasts and CD14+ monocytes. Additionally, the severe cutaneous signature was elevated in specific cell populations of SSc skin compared with controls, most notably in CD14+ monocytes, CCL19-APOE fibroblasts, and SFRP2-WIF1 fibroblasts (Supp. Table 8).
Figure 3.
The severe cutaneous signature is enriched in stromal cells and monocytes of the skin.
(A) Feature expression plot shows the highest average expression of upregulated differentially expressed genes (DEGs) concentrated in fibroblasts and CD14+ monocytes of the skin
(B) Violin plot representing the highest average expression levels of upregulated DEGs by cell type.
(C) Box plots of cell type percentages show increased fraction of epidermal cells and melanocytes in the less severe group (Cluster 1) and increased fraction of monocytes and fibroblasts in the more severe group (Cluster 2).
SSc-ILD lung stromal cells and monocytes express the severe SSc signature
Given the association between the severe SSc signature and lung function impairment, we investigated whether specific lung cells in SSc-ILD lungs expressed these transcripts using a scRNAseq dataset, GSE128169(16). In SSc-ILD samples, usual interstitial pneumonia (UIP) was the most common histologic pattern (7 out of 8), followed by non-specific interstitial pneumonia (NSIP) (Supp. Table 9).
Altogether, we identified 42 cell types in lung tissues (Supp. Fig 6A). We found that the highest expression of upregulated DEGs in the severe SSc signature was concentrated in myofibroblasts, lipofibroblasts, CD16+ and CD14+ monocytes (Figs. 4A–B). Comparison between conditions revealed differences in the expression of upregulated DEGs in myofibroblasts, fibroblasts, and monocytes between SSc-ILD and control samples (Fig 4C). Because the combined lung scRNA-seq dataset contained multiple lung samples from the same patients, we assessed whether the observed findings were robust to patient-level sampling. This analysis demonstrated that, although CD16+ monocytes were differentially enriched in the full dataset, this enrichment was not consistently observed across the split datasets, whereas CD14+ monocyte enrichment was maintained in both (Supp Figs. 6B–C). Lastly, the highest expression of downregulated DEGs was confined to alveolar epithelial cells, ciliated, goblet, and basal cells (Supp. Figs. 6D–E).
Figure 4.
The severe cutaneous signature and genes involved in KRAS signaling are enriched in stromal cells of the lung.
(A) Feature expression plot shows the highest average expression of upregulated differentially expressed genes (DEGs) concentrated in myofibroblasts, lipofibroblasts and CD16+ monocytes of the lung
(B) Violin plot representing the highest average expression levels of upregulated DEGs by cell type.
(C) Deconvolution of the highest average expression of upregulated differentially expressed genes of the severe signature in select cells of the lung, separated by disease status (SSc-ILD vs control).
(D) Feature expression plot shows the highest expression of genes involved in KRAS signaling in myofibroblasts, pericytes, smooth muscle cells, and lipofibroblasts of the lung.
(E) Violin plot shows highest expression of genes involved in KRAS signaling in lung by cell type.
The KRAS pathway is enriched in skin and lung stromal cells in SSc-ILD
To determine whether the severe SSc signature can be applied to detect common pathways between skin and lung fibrosis, we validated KRAS, a pathway enriched in GSEA of the severe SSc cluster. First, we determined which specific skin and lung cell types showed high expression of genes in the KRAS pathway. We found that pericytes, myofibroblasts and fibroblasts expressed high levels of KRAS signaling genes in the skin (Supp. Figs. 7A–B). We found a very similar pattern in the lung characterized by enrichment in myofibroblasts and pericytes, and unique to the SSc-ILD lung smooth muscle cells, and lipofibroblasts (Figs. 4D–E). KRAS enrichment was limited to stromal cells and not present in immune cells including monocytes. Furthermore, expression of the KRAS pathway transcripts was increased in the SSc-ILD lung when compared to controls (Supp Figs. 7C–E). This distinction between SSc and controls was less pronounced in the skin (Supp. Figs. 8A–C). Together, these analyses demonstrate a link between the severe SSc signature and expression of KRAS pathways in specific stromal cells in the SSc skin and SSc-ILD lung.
SSc-ILD lungs express more KRAS protein and activate downstream signaling
Given the complexity of KRAS pathways and the known limitations of gene expression to reflect the dynamic regulation of protein activity, we performed phosphorylation mapping of the KRAS signaling pathway in the SSc-ILD lung. We quantified the expression of KRAS protein (Fig. 5). and found that, indeed, SSc-ILD has a higher expression of KRAS in the lung tissue of patients with SSc-ILD (n=3) (Fig. 5A) vs. controls (n=3) (Fig. 5B–C) from an independent cohort. We also explored downstream signaling pathways, including extracellular signal-regulated kinase (ERK) and serine/threonine kinase (AKT), assessing both their total and phosphorylated forms (n=1 each). Our analyses revealed that SSc-ILD samples exhibited higher KRAS expression accompanied by increased activation of these KRAS-associated downstream pathways (Supp. Figs. 9–12). Thus, mapping transcriptomic signatures shared between skin and lung and correlated with the extent of tissue involvement in SSc allowed us to identify KRAS activation and suggest engagement of its downstream signaling pathways, including ERK and AKT. These findings point to a previously underrecognized set of shared pathogenic mechanisms that may contribute to SSc-ILD.
Figure 5.
Immunofluorescence shows increased expression of kristen rat sarcoma viral oncogene homologue (KRAS) in the SSc lung.
(A). Immuofluorescence analysis of KRAS expression in lung tissue from three different SSc-ILD lung explants.
(B). Immuofluorescence analysis of KRAS expression in lung tissue from three different control lung explants.
(C). Relative fluorescence units of KRAS IF SSc-ILD vs. Control (Mann-Whitney Test).
Discussion
We identified a cutaneous transcriptomic signature of SSc that was associated with severe skin and lung involvement. This severe SSc signature was enriched in fibroblasts and monocytes in the skin and similarly in specific stromal cells and monocytes in the lung. Analysis of the severe SSc cluster revealed enrichment of KRAS expression in the skin. We validated the association between genes in the severe SSc signature and mRSS in an independent cohort of skin transcriptomes. Based on the shared transcriptomic features between SSc skin and SSc-ILD identified in these analyses, we examined whether SSc-ILD lung explants demonstrated increased expression and activation of KRAS through phosphorylation, a key post translational mechanism, and downstream signaling. Altogether, our analyses demonstrated shared transcriptional signatures between immune cells and stromal cells in SSc and their association with severity of multiorgan involvement.
The severe SSc signature was enriched for pathways involved in extracellular matrix production and immune regulation. Although KRAS pathway enrichment was confined to stromal cells based on scRNA-seq deconvolution, the severe SSc signature also incorporated immune cells. The coexistence of profibrotic and immune activation pathways underscores the interplay between immune dysregulation and abnormal tissue repair processes driving the pathogenesis of SSc-ILD. Our findings are consistent with previous studies that have shown enrichment of profibrotic and immune-related pathways in the skin of patients with SSc(25,26). The non-severe cluster was associated with enrichment of pathways related to normal skin function, including melanin synthesis, likely due to preserved normal skin in these patients. The lack of fibrotic and immune pathway enrichment in the non-severe cluster could also reflect biological heterogeneity rather than absence of disease activity, as subtle or patchy molecular changes fall below detection in bulk profiling.
In the skin, the severe SSc signature was primarily found in fibroblasts. These findings are consistent with previous studies by Tabib et al.(14,15) demonstrating a diverse fibroblast population in the skin of SSc patients, including a SFRP2 high progenitor population. Our results align with those of Gur et al., who identified a subset of fibroblasts implicated in SSc pathogenesis but found minimal myofibroblast involvement(27). In the SSc-ILD lung, however, expression of the severe SSc signature was less pronounced in fibroblasts than in the skin, while its expression was significantly increased in myofibroblasts, key drivers of abnormal fibrosis in SSc-ILD(16). These differences may reflect distinct disease stages, with skin biopsies capturing early fibrotic activation and lung samples, such as explanted UIP lungs, representing more advanced fibrosis, as well as the inherently different tissue architecture of the two organs.
The severe SSc signature also included tissue macrophages and monocytes. CD14+ monocytes were consistently enriched for the severe signature in the skin and lung of patients with SSc. It is widely accepted that macrophages and monocytes play a significant role in the development and progression of SSc-ILD given their potential to serve as an inflammatory or fibrotic signal(28). Peripheral blood mononuclear cells (PBMCs) from patients with SSc have been found to have both a proinflammatory and a profibrotic expression profile and the overexpression of IL8 in the blood and bronchoalveolar lavage fluid (BALF) of SSc-ILD patients suggests continued macrophage activation(29,30). The number of circulating macrophages in SSc patients is also associated with impaired lung function(31), and transcriptomes of circulating macrophages are correlated with disease severity(32). Monocytes programmed with the SSc signature create a distinct connection between tissues and disease severity in SSc.
The severe SSc cluster was enriched for the KRAS pathway and single cell deconvolution showed high gene expression of KRAS in stromal cells of SSc skin and SSc-ILD. KRAS is a member of the rat sarcoma viral oncogene (RAS) family. RAS proteins are involved in cell proliferation and survival by acting as binary switches that cycle between active and inactive states. RAS mutations have been widely implicated in many human cancers(33), and recent studies have implicated them in the pathogenesis of renal fibrosis(34) and metaplastic honeycomb lesions of interstitial pneumonia(35). In the SSc-ILD lung, KRAS expression was high along with activation of the RAF-MEK-ERK pathway and the PI3K-AKT-mTOR pathway, two downstream pathways. Together, the identification of higher mRNA, protein expression, and pathway activation of KRAS in SSc-ILD suggest a potential role for this pathway in tissue fibrosis. It is possible that further validation of our findings could result in targeting KRAS downstream pathways could modulate fibrosis in SSc.
Our study has some limitations. Whereas non-specific interstitial pneumonia (NSIP) is more prevalent in SSc-ILD, lung samples examined in this study contain mostly usual interstitial pneumonia (UIP). This may reflect the very late-stage disease and future studies involving earlier stage lung disease are warranted. The strongest signal of lung involvement in our study was DLCO impairment, while it is possible that pulmonary hypertension contributes to this abnormality, the presence of histologic changes of fibrosis in the single cell validation cohort support the association of the severe signature with SSc-ILD. The cross-sectional nature of our study does not allow the longitudinal assessment of the severe SSc signature. Additionally, future studies warrant analysis of matched skin and lung samples. Validation of the scRNAseq findings in larger SSc cohorts is also needed to better capture disease heterogeneity. Despite these limitations, our study’s main strength is the integration of diverse molecular markers from multiple cohorts of patients leading to the identification of a severe SSc signature and validation of KRAS phosphorylation and activation in the SSc-ILD lung.
In conclusion, we have found a skin transcriptomic signature in SSc associated with severe multi-organ involvement. The severe signature is found in stromal and immune cells, which play a role in the progression of SSc. Using this framework, we leveraged the transcriptomic enrichment in KRAS of the skin to study whether this component of the severe SSc signature was activated in the SSc-ILD lung. The severe signature may lead to the identification of additional molecular pathways of organ involvement in SSc.
Supplementary Material
Support Statement
This project has been funded by NIH/NHLBI R01 HL153604 and R03 HL154275 to JLG.
National Heart, Lung, and Blood Institute 2T32HL007778-26 to support JZ.
NIH/NLBI R01HL152677 and R01 HL163984 to EH
NIH/NIAMS R01 AR073270 and NIH/NHLBI R01 HL152677 to MH.
NIAMS K24AR060297 to CFB
K08 HL151970-02 to CR. Manuscript contents are solely the responsibility of the authors and do not necessarily represent the official view of NIH.
Footnotes
Declaration of Interests
EH has received consulting fees from Boehringer Ingelheim. MH has received consultancy fees from AbbVie and has received research grant support from Boehringer Ingelheim for an investigator-initiated research project. She is a Scientific Advisory Board Member for the National Scleroderma Foundation. She has participated in clinical trials with Boehringer Ingelheim, AbbVie, Kadmon, Mitsubishi, Horizon, Emerald Health. JZ, NL, YL, XY, ZW, MR, AF, AK, HY, CB, IS, HS, CFB, MH CR, and JLG reported no conflicts of interest.
References
- 1.Krieg T, Takehara K. Skin disease: a cardinal feature of systemic sclerosis. Rheumatology 2009;48 Suppl 3:iii14–8. [DOI] [PubMed] [Google Scholar]
- 2.Tyndall AJ, Bannert B, Vonk M, Airò P, Cozzi F, Carreira PE, et al. Causes and risk factors for death in systemic sclerosis: a study from the EULAR Scleroderma Trials and Research (EUSTAR) database. Ann Rheum Dis 2010;69:1809–1815. [DOI] [PubMed] [Google Scholar]
- 3.LeRoy EC, Black C, Fleischmajer R, Jablonska S, Krieg T, Medsger TA Jr, et al. Scleroderma (systemic sclerosis): classification, subsets and pathogenesis. J Rheumatol 1988;15:202–205. [PubMed] [Google Scholar]
- 4.Le Gouellec N, Duhamel A, Perez T, Hachulla A-L, Sobanski V, Faivre J-B, et al. Predictors of lung function test severity and outcome in systemic sclerosis-associated interstitial lung disease. PLoS One 2017;12:e0181692. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Cole A, Denton CP. Biomarkers in Systemic Sclerosis Associated Interstitial Lung Disease (SSc-ILD). Current Treatment Options in Rheumatology 2022;8:152–170. [Google Scholar]
- 6.Kania G, Rudnik M, Distler O. Involvement of the myeloid cell compartment in fibrogenesis and systemic sclerosis. Nat Rev Rheumatol 2019;15:288–302. [DOI] [PubMed] [Google Scholar]
- 7.Gardner H, Shearstone JR, Bandaru R, Crowell T, Lynes M, Trojanowska M, et al. Gene profiling of scleroderma skin reveals robust signatures of disease that are imperfectly reflected in the transcript profiles of explanted fibroblasts. Arthritis Rheum 2006;54:1961–1973. [DOI] [PubMed] [Google Scholar]
- 8.Sargent JL, Milano A, Bhattacharyya S, Varga J, Kari Connolly M, Chang HY, et al. A TGFβ-Responsive Gene Signature Is Associated with a Subset of Diffuse Scleroderma with Increased Disease Severity. Journal of Investigative Dermatology 2010;130:694–705. Available at: 10.1038/jid.2009.318. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Lofgren S, Hinchcliff M, Carns M, Wood T, Aren K, Arroyo E, et al. Integrated, multicohort analysis of systemic sclerosis identifies robust transcriptional signature of disease severity. JCI Insight 12 2016;1. Available at: 10.1172/jci.insight.89073. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Assassi S, Wu M, Tan FK, Chang J, Graham TA, Furst DE, et al. Skin gene expression correlates of severity of interstitial lung disease in systemic sclerosis. Arthritis Rheum 2013;65:2917–2927. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Assassi S, Sharif R, Lasky RE, McNearney TA, Estrada-Y-Martin RM, Draeger H, et al. Predictors of interstitial lung disease in early systemic sclerosis: a prospective longitudinal study of the GENISOS cohort. Arthritis Res Ther 2010;12:R166. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Khanna D, Saggar R, Mayes MD, Abtin F, Clements PJ, Maranian P, et al. A one-year, phase I/IIa, open-label pilot trial of imatinib mesylate in the treatment of systemic sclerosis-associated active interstitial lung disease. Arthritis Rheum 2011;63:3540–3546. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Stifano G, Sornasse T, Rice LM, Na L, Chen-Harris H, Khanna D, et al. Skin gene expression is prognostic for the trajectory of skin disease in patients with diffuse cutaneous Systemic Sclerosis. Arthritis Rheumatol 2018;70:912–919. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Xue D, Tabib T, Morse C, Yang Y, Domsic RT, Khanna D, et al. Expansion of Fcγ receptor IIIa-positive macrophages, ficolin 1-positive monocyte-derived dendritic cells, and plasmacytoid dendritic cells associated with severe skin disease in systemic sclerosis. Arthritis Rheumatol 2022;74:329–341. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Tabib T, Huang M, Morse N, Papazoglou A, Behera R, Jia M, et al. Myofibroblast transcriptome indicates SFRP2hi fibroblast progenitors in systemic sclerosis skin. Nat Commun 2021;12:4384. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Valenzi E, Bulik M, Tabib T, Morse C, Sembrat J, Trejo Bittar H, et al. Single-cell analysis reveals fibroblast heterogeneity and myofibroblasts in systemic sclerosis-associated interstitial lung disease. Ann Rheum Dis 2019;78:1379–1387. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 2008;9:559. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wang B, Mezlini AM, Demir F, Fiume M, Tu Z, Brudno M, et al. Similarity network fusion for aggregating data types on a genomic scale. Nat Methods 2014;11:333–337. [DOI] [PubMed] [Google Scholar]
- 19.Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 2005;102:15545–15550. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Butler A, Hoffman P, Smibert P, Papalexi E, Satija R. Integrating single-cell transcriptomic data across different conditions, technologies, and species. Nat Biotechnol 2018;36:411–420. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Wiedemann J, Billi AC, Bocci F, Kashgari G, Xing E, Tsoi LC, et al. Differential cell composition and split epidermal differentiation in human palm, sole, and hip skin. Cell Rep 2023;42:111994. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Travaglini KJ, Nabhan AN, Penland L, Sinha R, Gillich A, Sit RV, et al. A molecular cell atlas of the human lung from single-cell RNA sequencing. Nature 2020;587:619–625. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Newman AM, Steen CB, Liu CL, Gentles AJ, Chaudhuri AA, Scherer F, et al. Determining cell type abundance and expression from bulk tissues with digital cytometry. Nat Biotechnol 2019;37:773–782. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Elhannani A, Martel M-E, Collet A, Chepy A, Sanges S, Hachulla É, et al. Anti-RNA polymerase III antibodies in systemic sclerosis: prevalence and clinical associations from a systematic review and meta-analysis. Rheumatology (Oxford) 2025. Available at: 10.1093/rheumatology/keaf392. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Moon S-J, Bae JM, Park K-S, Tagkopoulos I, Kim K-J. Compendium of skin molecular signatures identifies key pathological features associated with fibrosis in systemic sclerosis. Ann Rheum Dis 2019;78:817–825. [DOI] [PubMed] [Google Scholar]
- 26.Skaug B, Lyons MA, Swindell WR, Salazar GA, Wu M, Tran TM, et al. Large-scale analysis of longitudinal skin gene expression in systemic sclerosis reveals relationships of immune cell and fibroblast activity with skin thickness and a trend towards normalisation over time. Ann Rheum Dis 2022;81:516–523. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Gur C, Wang S-Y, Sheban F, Zada M, Li B, Kharouf F, et al. LGR5 expressing skin fibroblasts define a major cellular hub perturbed in scleroderma. Cell 2022;185:1373–1388.e20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Al-Adwi Y, Westra J, Goor H van, Burgess JK, Denton CP, Mulder DJ. Macrophages as determinants and regulators of fibrosis in systemic sclerosis. Rheumatology 2023;62:535–545. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Mathai SK, Gulati M, Peng X, Russell TR, Shaw AC, Rubinowitz AN, et al. Circulating monocytes from systemic sclerosis patients with interstitial lung disease show an enhanced profibrotic phenotype. Lab Invest 2010;90:812–823. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Fields A, Potel KN, Cabuhal R, Aziri B, Stewart ID, Schock BC. Mediators of systemic sclerosis-associated interstitial lung disease (SSc-ILD): systematic review and meta-analyses. Thorax 2022. Available at: 10.1136/thorax-2022-219226. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Trombetta AC, Soldano S, Contini P, Tomatis V, Ruaro B, Paolino S, et al. A circulating cell population showing both M1 and M2 monocyte/macrophage surface markers characterizes systemic sclerosis patients with lung involvement. Respir Res 2018;19:186. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Makinde H-KM, Dunn JLM, Gadhvi G, Carns M, Aren K, Chung AH, et al. Three Distinct Transcriptional Profiles of Monocytes Associate with Disease Activity in Scleroderma Patients. Arthritis Rheumatol 2023;75:595–608. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Simanshu DK, Nissley DV, McCormick F. RAS Proteins and Their Regulators in Human Disease. Cell 2017;170:17–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Muñoz-Félix JM, Martínez-Salgado C. Dissecting the Involvement of Ras GTPases in Kidney Fibrosis. Genes 2021;12. Available at: 10.3390/genes12060800. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Kataoka T, Okudela K, Matsumura M, Baba T, Kitamura H, Arai H, et al. Significant accumulation of KRAS mutations in bronchiolar metaplasia-associated honeycomb lesions of interstitial pneumonia. Oncol Lett 2022;24:225. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.