Cancer-associated fibroblast subtypes modulate the tumor-immune microenvironment and are associated with skin cancer malignancy

Abstract

Cancer-associated fibroblasts (CAFs) play a key role in cancer progression and treatment outcome. This study dissects the intra-tumoral diversity of CAFs in basal cell carcinoma, squamous cell carcinoma, and melanoma using molecular and spatial single-cell analysis. We identify three distinct CAF subtypes: myofibroblast-like RGS5+ CAFs, matrix CAFs (mCAFs), and immunomodulatory CAFs (iCAFs). Large-cohort tissue analysis reveals significant shifts in CAF subtype patterns with increasing malignancy. Two CAF subtypes exhibit immunomodulatory properties via different mechanisms. mCAFs sythesize extracellular matrix and may restrict T cell invasion in low-grade tumors via ensheathing tumor nests, while iCAFs are enriched in late-stage tumors, and express high levels of cytokines and chemokines to aid immune cell recruitment and activation. This is supported by the induction of an iCAF-like phenotype with immunomodulatory functions in primary healthy fibroblasts exposed to skin cancer cell secretomes. Thus, targeting CAF variants holds promise to enhance immunotherapy efficacy in skin cancers.

Fig. 1. A single cell transcriptomic atlas of human BCC, SCC, melanoma, and healthy skin.

A Workflow of donor sample processing for Smart-seq2 scRNA-seq, data analysis and verification. B UMAP projection of first-level clustering of 4824 cells from 15 donors (left). Clusters are labeled by cell types, which were identified by commonly accepted marker genes (right). C Expression of top marker genes for the main cell types.

Fig. 2. Second-level clustering of non-mesenchymal cells and CNV analysis.

A, B UMAP projection of second-level clustering, violin plots of signature genes as well as bar plots showing donor sample (n = 15 donors) distribution per cluster are presented for healthy and neoplastic keratinocytes and melanocytes. C CNV analysis (based on inferCNV package) of tumor samples using stromal cells as reference controls. UMAPs for healthy and neoplastic keratinocytes and melanocytes: Malignant cells with a predicted CNV alteration are highlighted in red and PTCH1/PTCH2 overexpressing cells without CNVs are highlighted in orange or yellow, respectively. D UMAP projection of second-level clustering, violin plots of signature genes as well as bar plots showing donor sample distribution per cluster are presented for immune cells. The distribution of cytotoxic, helper, and regulatory T cells are depicted in separate cut-outs. E Heatmap of genes that reflect the resting, activation, cytotoxic, co-stimulatory or co-inhibitors status of T cell subsets from healthy and tumor samples. KC-keratinocyte, MC-melanocyte, MEL-melanoma cells, hTcells-healthy T cells, tCD8-Cytotoxic T cells, tCD4-Helper T cells, Tregs-Regulatory T cells.

Fig. 3. Second-level clustering of fibroblasts and vascular smooth muscle cells (vSMCs) results in two heathy fibroblasts populations, four CAF subsets and one vSMC cluster.

A UMAP of second-level clustered fibroblasts and vSMCs. Violin plots of signature genes and bar plots showing donor sample (n = 15 donors) distribution per cluster. B Heatmap of top ten differentially expressed genes per cluster. C Differentially expressed transcription factors between mCAFs and iCAFs. B, C Source data of all significantly differentially expressed genes and transcription factors including exact p values are provided in the Source Data file. D Trajectoy analysis using Monocle2. Cells were highlighted according to clusters, category or pseudotime. E UMAP colored in pseudotime showing trajectory results from Monocle3.

Fig. 4. The RGS5⁺ cells are an inhomogeneous population of CAFs and pericytes.

A Feature and violin plots showing the expression of fibroblast and pericyte marker genes in the RGS5⁺ cluster. Representative immunohistochemistry of TAGLN, DES and CD31 in different regions of the tumor (intratumoral, peritumoral) (n = 10 tumor samples from biologically independent donors). B Representative images of COL1A1 (green), RGS5 (red), and PDGFRA (blue) RNAScope fluorescence stainings in four different regions of FFPE tissue sections from donor sample SCC IV, representative for the n = 10 independent tumor samples. DAPI nuclear stain is shown in gray. Scale bar represents 20 μm. C Myofibroblasts in a HNSCC dataset from Puram et al. exhibit a very similar expression pattern in comparison to the RGS5⁺ cluster in our dataset.

Fig. 5. mCAFs and iCAFs are characterized by the expression of ECM and immunomodulatory genes, respectively.

Representative images from (A) COL1A1 (green), COL11A1 (red), and PTGDS (blue) and (B) COL1A1 (green) and MMP1 (red) RNAScope fluorescence stainings to identifiy mCAFs and iCAFs respectively in FFPE tissue sections from n = 52 biologically independent tumor samples. DAPI nuclear stain is shown in gray. Scale bar represents 20 μm. C Spatial plots highlighting the spatial distribution of total CAFs (COL1A1), iCAFs (COL1A1⁺MMP1⁺) and mCAFs (COL1A1⁺COL11A1⁺) and respective H&E stainings on consecutive sections. Dashed-lined boxes show approximate area of spatial plot in H&E stainings. D Quantification of total CAFs (COL1A1⁺), iCAFs (COL1A1⁺MMP1⁺), and mCAFs (COL1A1⁺COL11A1⁺MMP1⁻) in cells per mm² in 52 independent tumor samples of nodular (n = 8) and infiltrative BCC (n = 9), well (n = 8) and poorly (n = 10) differentiated SCC as well as low- (n = 8) and high-grade (n = 9) melanoma. Fibroblast numbers of at least five representative ROIs from each tumor were summed-up and normalized to the tissue area to capture the whole tumor tissue. Data are presented as box plots with median as center and whiskers ranging from minimum to maximum, bounds of boxes extend from the 25th to 75th percentiles. Statistical analysis by two-sided Mann-Whitney test. *p < 0.05, **p < 0.01. E Representative images from COL11A1 immunohistochemistry stainings (n = 10 biologically independent tumor samples). Scale bar represents 100 μm. F Image analysis of CD3⁺cells/mm² in tumor nests and total CAFs/mm² (high-low cutoff 140 cells/mm²), mCAFs/mm² (high-low cutoff 40 cells/mm²), or iCAFs/mm² (high-low cutoff 40 cells/mm²) in 97 ROIs from nodular and infiltrative BCCs (n = 15 biologically independent tumor samples). Linear regression analysis of log(CD3⁺cells/mm²) in tumor nests and log(CAFs/mm²). Representative images of CD3 immunohistochemistry and COL1A1 (green) COL11A1 (red) RNAScope fluorescence stainings. Data are presented as mean values ± SEM. Statistical analysis by unpaired two-sided t-test; *p < 0.05. D, F Source data and exact p values are provided in the Source Data file.

Fig. 6. Fibroblasts are an important source of chemokines in the tumor.

A Expression of immunomodulatory genes in iCAFs compared to healthy and neoplastic keratinocytes and melanocytes and interrogation for respective receptors in healthy and neoplastic keratinocytes and melanocytes as well as immune cells (n = 15). B Circular plots of selected receptor-ligand pairs from CellChat analysis, showing mCAF/iCAF as source cells (n = 15 donors). C Representative images of RNA ISH staining with probes against CXCL2, CXCL8, IL24, and COL1A1 of BCC, SCC and melanoma samples (n = 43 biologically independent tumor samples). D In vitro cytokine expression of NHDF after exposure to conditioned medium from NHDFs, VM08, VM15, VM26, VM19, VM25, and SCC13 cells for 72 h in comparison to the cytokine expression of the cancer cell lines VM08, VM15, VM26, VM19, VM25, and SCC13, and to primary melanoma-derived CAFs (pMel CAFs). Data from four independent experiments are presented as bar graphs showing mean values ± SD, overlayed with individual data points of the independent experiments. Statistical analysis by One-way-ANOVA and Tukey’s post hoc test for multiple comparison on log-transformed data. Significant comparisons to NHDFs are shown; Source data and exact p values are provided in the Source Data file. *p < 0.05, **p < 0.01, ***p < 0.001.

Fig. 7. Fibroblasts activate CD4⁺ and CD8⁺ T cells.

A Experimental setup of T cell assays shown in B and C. B Proliferation assessed by flow cytometry of CD4 or CD8 T cells upon co-culture with NHDFs pre-treated with conditioned medium from cancer cells, primary melanoma-derived CAFs (pMel CAFs) or cancer cells. C Upregulation of the early activation marker CD69 on CD4 or CD8 T cells after 24 h of co-culture with NHDFs pre-treated with conditioned medium from cancer cells, pMel CAFs or cancer cells. Data represented as fold change of percentages of cells positive for the indicated markers normalized to NHDFs. B, C n = 6 biologically independent samples for T cells only and untreated NHDFs, n = 5 biologically independent samples for pMel CAFs, and n = 3 biologically independent samples for T cells in co-culture with VM15, VM26, VM19, VM25. Data are presented as mean values ± SD. Statistical analysis in comparison to NHDFs or to T cells only by unpaired two-sided Student’s t test and Welch’s correction; B, C Source data and exact p values are provided in the Source Data file. *p < 0.05, **p < 0.01, ***p < 0.001. D Schematic summary of spatial distribution of distinct CAF subsets in human skin cancer.