Plasticity and Functional Heterogeneity of Cancer-Associated Fibroblasts

Abstract

Tumor heterogeneity and plasticity enable adaptation to metastatic microenvironments and resistance to therapies. Recent progress in single-cell analyses has permitted detailed characterization of the complexity and diversity of the different tumor components in multiple tumor types. Cancer-associated fibroblasts (CAFs) are a central component of the tumor microenvironment (TME) and play critical roles in cancer progression and therapeutic response. The identification of different CAF subtypes and elucidation of their functional plasticity is crucial to identify novel therapeutic approaches to target pro-tumorigenic CAFs and harness tumor suppressive CAFs to enhance the efficacy of cancer treatments. In this review, we discuss how intrinsic and extrinsic factors and the extensive crosstalk between cancer cells and the TME promote CAF heterogeneity and their contributions to cancer progression and therapeutic resistance. Understanding the roles of CAF plasticity and their intercellular interactions may drive the development of effective treatment strategies to improve patient prognosis.

Introduction

Cancers are malignant lesions that arise from mutations accruing within cells. Malignant cells co-exist with non-cancerous cells, which constitute the tumor microenvironment (TME), a complex and evolving entity constantly shaped by its interaction with tumor cells. The main components of the TME are blood vessels, fibroblasts, extracellular matrix (ECM) and immune cells of myeloid and lymphoid origins (1). The heterogeneous and plastic nature of the TME strongly influences cancer progression and response to therapy.

The advent of single-cell omics analyses revealed the complexity of cancers and the extent of heterogeneity within individual tumor compartments. The ability of cells to modify their epigenomic and transcriptomic profiles fosters the onset of different cell states, leading to tumor heterogeneity (2). The intrinsic genomic plasticity of cancer cells promotes dynamic and potentially reversible phenotypic switching in response to therapeutic insults or changes in the TME to promote tumor survival and progression (3, 4). While initial studies focused on the heterogeneity of cancer cells, it is increasingly clear that non-malignant cells of the TME display plasticity, as their transcriptomes and phenotypes are molded by interactions with cancer cells and other cells of the TME (5, 6). The continuum of crosstalk between cancer cells and the TME directs malignant and non-malignant cells into a range of phenotypic states that might coexist in varying proportions (5, 6), posing a challenge to the development of effective therapies.

Cancer-associated fibroblasts (CAFs), a crucial component of the TME in solid tumors, are key factors in shaping the TME and promoting tumor growth, metastasis, inflammation, and ECM remodeling (7, 8). CAFs have heterogenous phenotypes, origins, and functions (9), and characterization of their heterogeneity in many cancer types has led to the definition of numerous CAF subtypes, including myofibroblast-like (myCAFs), inflammatory (iCAFs), and antigen-presenting (apCAFs), among others (10–16). CAFs can originate from multiple precursors, including resident fibroblasts reprogrammed by cancer cells (17), bone marrow (BM)-derived mesenchymal cells recruited to the TME (18, 19), adipose-derived cells (20), endothelial cells (21), mesothelial cells (22, 23) or pericytes (24). The extensive heterogeneity in CAF subpopulations is evident in the vast array of tasks they perform in tumor progression, metastasis, and response to therapy. CAFs modulate the composition and stiffness of the ECM, which physically interferes with drug delivery (25, 26) and facilitates cancer cell survival and migration, leading to cancer progression (27, 28). CAF-mediated secretion of multiple cytokines, chemokines, and growth factors promotes tumor progression by acting directly on cancer cells and by altering the immune cell milieu in support of an immunosuppressive TME, allowing cancer cells to evade immune surveillance (29, 30).

Over the past decade, multiple approaches to targeting CAFs and their crosstalk with cancer cells have been attempted in preclinical models and clinical trials. However, clinical trials targeting CAFs have mostly failed and in some cases even accelerated cancer progression (31). A better understanding of CAF plasticity, origins and interactions is indispensable to identifying novel therapeutic strategies to target pro-tumorigenic CAFs and improving response to current therapies. Here, we discuss how patient demographics (e.g., age, obesity), intrinsic tumoral factors (e.g., cancer type, cancer cell state, extracellular vesicles (EV) and cancer cell mutations) and extrinsic stressors promote CAF plasticity and functional heterogeneity. Additionally, we analyze the extensive crosstalk between cancer cells and CAFs in promoting different CAF subtypes.

CAF heterogeneity in cancer

Strengths and weaknesses of single-cell omics analyses to identify CAF subpopulations:

Transcriptional profiling has been widely used to define CAF subpopulations. The advent of single-cell RNA sequencing (scRNA-seq) has dramatically improved the resolution of transcript analysis compared to classic bulk RNA-seq (15, 32) (Fig. 1). scRNA-seq enables the identification of distinct CAF subsets based on gene expression profiles, potentially uncovering novel biomarkers and signaling pathways associated with CAF activation and function. However, this approach is limited by the use of cell type annotations based on the cluster expression of established CAF markers, which might not necessarily apply in all cancers (e.g., glioblastoma) (33). When general markers are available and utilized, it is crucial to balance the sub-cluster resolution to identify rare but distinct CAF types and/or states that are biologically meaningful (15). Similarly, it is critical to avoid over-clustering, which might identify CAF subpopulations without biological meaning. scRNA-seq is also limited by its inability to capture spatial information, making it challenging to understand the distribution of CAFs within tumors and their interactions with neighboring cells.

Figure 1. Omics analysis for CAF characterization.

Graphical representation of available technologies to detail CAF identity, functional properties, and spatial interactions. The main features of the represented technologies are labeled in red. Created in BioRender. Flynn, J. (2025) https://BioRender.com/fu3j89k

Spatially resolved transcriptomics and spatial transcriptomics imaging overcome some limitations of scRNA-seq and allow the visualization of gene expression patterns within intact tissue sections. These techniques provide spatial context to gene expression data, facilitating the mapping of CAFs and their molecular interactions within the TME (34–36). However, spatial transcriptomics methods often have lower resolution and/or coverage than scRNA-seq, limiting the detection of rare or low-abundance CAF populations. In this landscape, analysis algorithms play a fundamental role, offering the possibility to retrieve extensive information but also requiring elevated computational resources, especially for the integration of multiple omics data (e.g., ATAC-seq).

Novel technologies have enhanced the options available for analyzing protein expression. A classic technique used for CAF identification is immunofluorescence (IF), which can be multiplexed to enable the characterization of CAFs and their spatial distribution in relation to other cell types (Fig. 1). While IF is constrained by the availability of specific antibodies, systems detecting up to 16 antibodies have been developed and utilized in the last decade (37–39). Spatial proteomics techniques, such as mass spectrometry imaging (MSI) and DNA barcoding-based IF, enable the spatially resolved analysis of many proteins within tissue sections, providing spatial information on protein expression and post-translational modifications in situ (40) (Fig. 1). Still, spatial proteomics methods have limited coverage of the proteome (generally 40–80 proteins) and require careful validation of protein markers for accurate characterization of CAFs. Depending on the available technology, image acquisition speed constraints, cell segmentation challenges, and general background noise may limit the automation and robustness of the technique.

Developments in flow cytometry (FC) techniques with spectral deconvolution allow simultaneous screening of up to ~50 parameters, enabling high-dimensional profiling of cellular phenotypes. FC can capture a broad range of intracellular and extracellular proteins, allowing for the characterization of CAFs and their functional states within the TME (11, 41) (Fig. 1). However, like other single-cell techniques, FC lacks spatial information, which is crucial for understanding the organization of CAFs within the tumor niche. In addition, disaggregating tissues to recover fibroblasts requires rather aggressive methods, which can limit the abundance and quality of the recovered CAFs (33, 42).

In conclusion, the characterization of CAFs in cancer requires a multi-faceted approach that integrates transcriptomic, proteomic, and spatial information. While each technique offers valuable insights into CAF biology, their limitations underscore the importance of combining multiple approaches to achieve a comprehensive understanding of CAF heterogeneity, spatial organization, and functional contributions to cancer progression.

CAF origin, subclassification and characteristics:

Previous studies elucidating CAF origin have described phenotypic transitions arising from the activation of normal resident fibroblasts by tumor-derived stimuli (43). Such transitions to a CAF phenotype are characterized by enhanced ECM remodeling and secretion of various growth factors and cytokines (10, 43, 44). An overview of CAF origins, subtypes and characteristics discussed herein is shown in Figure 2A. Despite CAFs often originating from fibroblasts, cell types of non-fibroblast lineage have also been shown to express established CAF genes, including endothelial cells, adipocytes, pericytes, monocytes, and BM-derived mesenchymal stem cells (9, 18, 45). Fibroblast-derived CAFs express high levels of platelet-derived growth factor receptor (PDGFR) family proteins, α-smooth muscle actin (α-SMA), fibroblast activation protein (FAP), fibroblast-specific protein 1 (FSP1) and key mediators of inflammation, such as interleukins IL-1 and IL-6, matrix metalloproteinases (MMPs), and secreted protein acidic and rich in cysteine (SPARC) (9, 17, 46, 47). In pancreatic cancer, the secretion of cytokines and growth factors, such as transforming growth factor receptor β1 (TGF-β1) by pancreatic cancer cells, activates pancreatic stellate cells (PSCs) into a myofibroblast-like phenotype. Activated PSCs can then secrete ECM proteins in excess, composing the fibrous tissue characterizing the CAF-enriched desmoplastic reactions in pancreatic cancers (48). In liver cancer, activated hepatic stellate cells (HSCs) are a major source of CAFs, producing various cytokines, chemokines, and ECM proteins (49). In mouse models of melanoma and spontaneous pancreatic carcinoma, exposure to TGF-β1 induces endothelial cells to acquire a CAF phenotype via endothelial-to-mesenchymal (EndMT) transition, characterized by the increased expression of FSP1 and decreased expression of CD31 (21). Taken together, these studies suggest that CAFs originate from both fibroblast and non-fibroblast lineages, which may contribute to their heterogeneity (Fig. 2A).

Figure 2. Overview of CAF subtypes.

A: Schematic showing cell(s) of origin, phenotype and/or function, markers and/or gene signature, and prognosis associated with each CAF subtype. B: Summary of CAF subtypes identified in different tumor types, markers and/or gene signature, and prognosis associated with each subtype. proCAF: progenitor CAF; myCAF: myofibroblast CAFs; senCAF: senescent CAF; mCAF: matrix-producing CAF; iCAF: inflammatory CAF; apCAF: antigen-presenting CAF; pan-dCAF: pan-desmoplastic CAF; pan-pCAF: pan-proliferating CAF ; CAFadi: adipogenic CAF; CAFpn: peripheral-nerve like CAF; ifnCAF: interferon-response CAF; rCAF: reticular-like CAF; tCAF: tumor-like CAF; csCAF: complement-secreting CAF; dCAF: developmental CAF; vCAF: vascular CAF; cCAF: cycling CAF; eCAF: EMT-like CAF. Created in BioRender. Flynn, J. (2025) https://BioRender.com/fu3j89k

The heterogeneous origins of CAFs and their plastic nature have led to the classification of multiple subtypes (9, 46). Studies focused on phenotyping CAF subtypes in human cancers classify populations based on α-SMA levels and functional properties, such as angiogenesis, invasion, immune modulation, and conferral of therapeutic resistance (46). Pan-cancer scRNA-seq identified progenitor CAFs (proCAFs) as the initial CAF population found in the early stages of lung, stomach, colon adenocarcinoma, and breast cancer samples (43). proCAFs are characterized by the expression of insulin-like growth factor (IGF1), osteoglycin (OGN), complement component 7 (C7) and high levels of nuclear factor IX (NFIX), associated with an enrichment in proliferative pathways (43). Additionally, proCAFs were shown to share functions across CAF subtypes, suggesting that this subpopulation might serve as a foundational phenotype for several CAF subpopulations (43). Below we report the main CAF subtypes and their functional characteristics.

myCAFs and iCAFs:

Initial studies in pancreatic ductal adenocarcinoma (PDAC) and PSCs established two major CAF subtypes, myCAFs and iCAFs (10, 16). In PDAC, myCAFs and iCAFs have been described as reversible cell states, governed by the reprogramming of α-SMA and IL-6 signaling, respectively (10, 16). In pancreatic cancer, myCAFs are found adjacent to tumor cells while iCAFs are located within the desmoplastic stroma, distant from the tumor parenchyma (10, 44, 50). Conversely, in colorectal cancer (CRC), the phenotypic switch between iCAFs and myCAFs is controlled by Wnt signaling. Low Wnt signaling promotes the iCAF phenotype, which induces cancer cell epithelial-to-mesenchymal transition (EMT) and invasion. Conversely, high Wnt signaling via β-catenin induces the myCAF phenotype, which supports tumor growth (51).

The myCAF phenotype is induced by the activation of the TGF-β/SMAD2/3 pathway and characterized by high expression of α-SMA (ACTA2) and other contractile proteins, including transgelin (TAGLN), myosin light chain 9 (MYL9), tropomyosin (TPM1/2), matrix metallopeptidase 11 (MMP11), periostin (POSTN), and homeodomain-only protein homeobox (HOPX) (10). A pan-cancer analysis identified myosin heavy chain 11 (MYH11) and regulator of G-protein signaling 5 (RGS5) as additional myCAF markers (43). scRNA-seq analysis of CAFs in both mouse and human pancreatic and breast cancers identified a myCAF population marked by expression of LRRC15, a leucine-rich-repeat protein (52, 53). Differentiation of LRRC15⁺ myCAFs was shown to be indispensably driven by TGF-βR2 signaling in healthy fibroblasts. LRRC15⁺ myCAFs are rich in ECM components and can restrain CD8⁺ T cell function (53). myCAFs have been shown to play both tumor-promoting and tumor-restraining roles depending on tumor stage and microenvironmental context (44). The main functions of myCAFs include high ECM production and responsiveness to TGF-β, which can further drive the recruitment and activation of CAFs (46). Conversely, selective depletion of myCAFs led to the discovery of their tumor-restraining functions (54, 55), which have been shown to occur at both early and late tumor stages (56). However, CAFs with tumor-restraining function are generally more abundant in early-stage cancer, protecting normal tissue against the developing tumor (56). Although the underlying mechanisms maintaining the balance between tumor permission or inhibition are largely unknown, the continuous crosstalk between CAFs and cancer cells educates CAFs to acquire pro-tumorigenic functions. One proposed mechanism in the context of breast cancer foresees the secretion of Hedgehog ligand (SHH) by cancer stem cells (CSCs), which activates paracrine Hedgehog signaling in CAFs and supports tumor growth (57). Targeted inhibition of the Hedgehog pathway in mouse mammary gland tumors led to a reduction in tumor growth and CAF depletion, indicating that CSC-CAF crosstalk via Hedgehog signaling supports the presence of CAFs within the TME and promotes tumor growth (57). Conversely, in a PDAC mouse model, while selective Shh depletion in epithelial cells decreased myCAF content, this reduction was associated with more aggressive features like increased vascularity and undifferentiated histology marked by significant increases in zinc finger E-box binding homeobox 1 (Zeb1) and SNAI2 (Slug) expression (58). These findings support a tumor-suppressive function of myCAFs in this model (58).

CAF subsets S1-S4 have been identified in human triple-negative breast cancer (TNBC) and high-grade serous ovarian cancer (HGSOC). In both cancer types, CAF-S1 and CAF-S4 have been classified as myCAFs since they express α-SMA (11, 59). In TNBC, CAF-S1 are further characterized by high expression of FSP1, PDGFR-β, CD29, FAP, and low levels of caveolin 1 (CAV1). Conversely, CAF-S4 are negative for FAP and express low levels of FSP1 and PDGFR-β (11). In HGSOC, CAF-S1 and CAF-S4 both express high levels of FSP1, PDGFR-β and CD29, and low levels of CAV1, but have differential expression of FAP (CAF-S1, FAP^hi; CAF-S4, FAP^low) (11, 59). Despite their classification as myCAFs, the authors indicate that CAF-S1 and CAF-S4 have distinct transcriptomic profiles, likely reflecting differences in their functions (11, 59). Further characterization of CAF-S1 in both TNBC and HGSOC showed upregulation of genes involved in in cell adhesion, ECM organization, and immune response, while the CAF-S4 subset was enriched in muscle contraction, actin cytoskeleton regulation, and oxidative metabolism genes (11, 59). Additionally, in both cancer types, the CAF-S3 subset expresses little to no α-SMA, FAP, or CAV1, and higher levels of FSP1 and PDGFR-β (11, 59). Peculiarly, CAF-S2 were negative for all the aforementioned CAF markers (59). In HGSOC, CAF-S2 and CAF-S3 were identified as “non-activated” CAFs due to the lack of α-SMA expression (59).

Additional studies in breast cancer led to the identification of the senescent CAF (senCAF) phenotype. The restricted expression of the senescence marker p16 to the myCAF subtype determined the classification of p16⁺ senCAFs as myCAFs (60). p16⁺ senCAFs also express a wide range of ECM and ECM-modifying proteins, such as collagens and enzymes that alter collagen fibers, respectively (60). Additional senCAF markers identified in human breast cancer are p15 and β-galactosidase (β-gal), which are well-established markers of senescence (60). Of note, the presence of senCAFs has been shown to increase with tumor progression (61). In PDAC, senCAFs have been shown to upregulate immunoregulatory genes including IL-6, IL-17, TGF-β, and AP-1, orchestrating an immunosuppressive TME (61, 62).

Finally, uniform manifold approximation and projection (UMAP) of pan-cancer scRNA-seq revealed that a subpopulation identified as matrix-producing CAFs (mCAFs) is projected in close proximity to myCAFs, suggesting similar transcriptomic profiles (43). However, despite the high degree of similarity, mCAFs differed enough from myCAFs to be classified as a separate subtype. mCAFs express low levels of the myCAF marker α-SMA but high expression of ECM genes, including collagen type X alpha 1 (COL10A1) and POSTN (43). A similar upregulation in collagens and other ECM components (e.g., COL5A1, COL5A2, COL6A3), POSTN (43), fibronectin (FN1), lumican (LUM), decorin (DCN), and versican (VCAN) was consistently observed in mCAFs from human intrahepatic cholangiocarcinoma (ICC) samples (63). mCAFs are predominantly present in mid-to-late tumor stages, suggesting a possible link between these subtypes in advanced tumors (43).

iCAFs are induced by the IL-1/JAK/STAT pathway and characterized by a secretory phenotype. This subtype expresses low levels of α-SMA and high levels of cytokines IL-6, IL-8, IL-11, and several chemokines (e.g., C-C motif chemokine ligand CCL2, C-X-C motif chemokine ligands CXCL1, CXCL2, CXCL12, and CXCL14) (10, 14, 16, 43). iCAFs are additionally characterized by high expression of insulin-like growth factor family genes (e.g., IGF1 and IGFBP6) and genes involved in the complement cascade (e.g., C3, C7, and complement factor D (CFD)) (64). The iCAF phenotype is induced by cancer cell-mediated secretion of IL-1α and TNF-α (44), which stimulates their interaction with immune and endothelial cells through CXCL12-CXCR4 and CCL2-ACKR1 signaling axes, respectively (43, 46). scRNA-seq datasets derived from lung, gastric, breast, and colorectal cancers revealed that, like proCAFs, iCAFs are predominant in early stages of cancer development (43). Tracing the distribution of CAF subtypes throughout tumor progression highlights the dynamic ability of CAFs to display plasticity. In the context of disease prognosis, iCAF biomarkers produce variable outcomes (43) and their abundance has been associated with both poor and favorable clinical outcomes depending on tumor type (64), suggesting a context-dependent role of iCAFs. In a pan-cancer analysis, iCAF biomarkers were identified in immune-inclusive, or “hot” tumors, and absent in “cold” tumors (43). Given the continuous interaction of iCAFs with T cells, B cells, and myeloid cells to orchestrate immune responses, the low presence of iCAFs in “cold” tumors could partially contribute to the reduced response of these tumors to immunotherapy (43). However, iCAFs can also contribute to chemoresistance due to their increased inflammatory cytokine expression (64), which modulates the tumor immune microenvironment to promote immune cell dysfunction. Specifically, direct interactions between iCAFs and tumor-infiltrating T cells via chemokine ligand-receptor binding have been shown to induce CD8⁺ T cell dysfunction and exhaustion (64).

apCAFs:

Antigen presenting CAFs (apCAFs) are characterized by surface expression of MHC class II but lack co-stimulatory molecules expressed on professional antigen presenting cells (APCs) needed to activate CD4⁺ T cell signaling (22, 46). First characterized in the KPC mouse model of pancreatic cancer, apCAFs express H2-Aa and H2-Ab1, which encode the alpha and beta chains of MHC class II, respectively, and Cd74, which encodes the invariant chain (10). In addition to antigen presentation and processing, other upregulated pathways in apCAFs include fatty-acid metabolism, MYC targets, and MTORC1 signaling (10). Studies in PDAC suggested that tumor paracrine signals, including IL-1 and TGF-β, can also trigger the transition of mesothelial cells to an apCAF phenotype (22).

Additional CAF subtypes:

Despite myCAFs, iCAFs and apCAFs being the most-studied CAF subtypes, additional subtypes have been described. pan-CAF gene signatures were generated using melanoma, head and neck squamous cell carcinoma (HNSCC), and lung cancer scRNA-seq datasets by comparing the transcriptomes of CAFs versus non-CAF cells (65). The following pan-CAF subtypes were newly described: i) pan-desmoplastic CAFs (pan-dCAFs), characterized by upregulation of collagens (e.g., COL1A1 and COL3A1) and genes associated with ECM remodeling, and ii) pan-proliferating CAFs (pan-pCAFs), which express genes involved in cell cycle and proliferation (65). Other CAF subtypes identified in additional studies include: i) adipogenic CAFs (CAFadi), which overexpress CFD across cancer types (66) and ATP binding cassette subfamily A member 8 (ABCA8) in gastric cancer (67). ABCA8 is also expressed in lipid-rich CAFs and it has been associated with poor prognosis in PDAC ; ii) peripheral nerve-like CAFs (CAFpn), which express peripheral nerve-related genes (e.g., myelin protein zero (MPZ), S100 calcium binding protein B (S100B), leucine rich repeat LGI4, and proteolipid protein 1 (PLP1)) and high levels of SRY-box transcription factors (SOX) SOX2 and SOX10 (66); iii) interferon-response CAFs (ifnCAFs), characterized by the upregulation of genes expressed in response to interferon (IFN) signaling (e.g., CXCL9, CXCL10, CXCL11, indoleamine 2,3-dioxygenase 1 (IDO1)) (15); iv) reticular-like CAFs (rCAFs), a rare subclass characterized by high expression of chemokine genes CCL19 and CCL21, which facilitate homing of naïve T cells in lymphoid tissues (15); and v) tumor-like CAFs (tCAFs), which are characterized by expression of CD10 and CD73 and enriched in the same pathways upregulated in the adjacent tumor cells (68). In studies of breast and lung cancer, tCAFs exhibited upregulation of glycolysis, stress response, and metabolic activity (68). Further characterization of tCAFs in human breast cancer, via scRNA-seq and gene set enrichment analysis (GSEA), demonstrated enrichment in EMT-associated pathways, TGF-β signaling, KRAS signaling, glycolysis, MTORC1, PI3K/Akt/mTOR signaling, p53 pathway, and hypoxia (15). Within the tumor, tCAFs are in close proximity to tumor cells and are associated with poor patient prognosis (68).

The intricate landscape of CAFs within the TME reflects a dynamic interplay of cellular origins and functional diversity. The characterization of CAFs based on marker expression and function, such as immune modulation and matrix deposition, has led to the classification of distinct CAF subsets. However, the continuous identification of new CAF subpopulations across cancer types raises concerns about over-clustering and -classification, which could ultimately reduce the biological relevance of the CAF cell states identified. Moving forward, it is crucial to compare the various CAF phenotypes and assess which CAF subtypes are biologically relevant and which markers will facilitate a clear and rigorous sub-classification.

Effects of intrinsic tumoral factors on CAF heterogeneity and plasticity

The heterogeneous nature of cancer molds disease progression, response to therapy, and the TME. Hence, it is crucial to dissect CAF heterogeneity and their diverse functions in a context-dependent manner. Here, we review how different intrinsic tumoral factors promote CAF heterogeneity.

Cancer type:

Although a previous study identified a universal fibroblast type present in all organs that can transition into various cell states (69), it is possible that additional imprinting by the tissue of residence might contribute to CAF intertumoral heterogeneity and functional plasticity. Here we review CAF subtypes that have been described in specific cancer types (Fig. 2B).

PDAC:

Complement-secreting CAFs (csCAFs) have been identified in PDAC patient samples. This subpopulation is enriched in early stages of PDAC, with its presence significantly higher in stage I compared to stage II and III disease. Despite the biological function of csCAFs remaining largely unknown, these findings suggest a tumor-suppressive role in PDAC and indicate that csCAFs might gradually decrease during tumor progression. csCAFs show high transcriptional levels of complement system components, including C3 and C7, and complement factors B, D, H, and I, which may suggest a role in the regulation of intratumoral immune and inflammatory responses (70). Although the authors classified csCAFs as a new CAF subpopulation, the enrichment in complement components (i.e., C3, CFD) is shared with the iCAF subtype (70).

Additional CAF subtypes identified in PDAC include FAP⁺ CAFs and α-SMA⁺ CAFs, which have distinct roles in tumor progression (71). These subpopulations were identified following an in-depth analysis of scRNA-seq data from mouse PDAC, which defined six unique CAF subpopulations. The results indicated minimal overlap between FAP and α-SMA expression in CAFs, and these findings were recapitulated by scRNA-seq analysis of human PDAC samples (71). Furthermore, FAP⁺ CAFs were associated with a significant reduction in overall survival, whereas α-SMA⁺ CAFs were linked to increased overall survival in patients. Consistently, depletion of FAP⁺ CAFs reduced PDAC tumor progression in mice, while elimination of α-SMA⁺ CAFs led to more aggressive PDAC tumor growth, further supporting their classification as distinct CAF subpopulations (71). Further analysis of mouse PDAC tumors unveiled differential effects of FAP⁺ and α-SMA⁺ CAFs on the tumor immune microenvironment. Specifically, FAP⁺ CAF depletion reduced CD11b⁺ myeloid cell infiltration in the tumor, whereas α-SMA⁺ CAF depletion decreased the effector T cell to regulatory T cell (T_reg) ratio. In the context of response to therapy, α-SMA⁺ CAF-derived IL-6 was found to contribute to gemcitabine resistance by enhancing cancer cell survival, while FAP⁺ CAF-derived IL-6 had no effect on response to gemcitabine (71). Consistently, total loss of IL-6 from the TME or specifically from αSMA⁺ CAFs, achieved using transgenic mice, did not affect PDAC progression but improved the overall survival of mice upon treatment with gemcitabine (71), suggesting that IL-6 secretion is associated with chemoresistance. In this study the authors speculate that cancer cells likely exploit paracrine IL-6 to promote their survival through activation of MAPK/STAT3 signaling pathways in the presence of gemcitabine.

Breast cancer:

Additional CAF subtypes identified in the MMTV-PyMT murine breast cancer model include developmental CAFs (dCAFs), vascular CAFs (vCAFs), and cycling CAFs (cCAFs) (32, 44, 72). dCAFs have been proposed to originate from tumor cells that have undergone EMT. Moreover, since dCAFs express genes related to different types of stem cells, including stimulator of chondrogenesis 1 (Scrg1), Sox9, and Sox10, they have been associated with cell differentiation, tissue development, and morphogenesis functions (32, 44, 72). In contrast, vCAFs have been proposed to originate from a pool of perivascular cells. This subpopulation is closely associated with blood vessels in early-stage tumors, but invades the tumor stroma during tumor progression. vCAFs are enriched in transcription factors and genes involved in cell junctions and express high levels of nuclear receptor subfamily 2 group F member 2 (Nr2f2)) and genes associated with angiogenesis and vascularization (e.g., Notch3, Col18a1, endothelial PAS domain protein 1 (Epas1)) (32, 72). Additionally, in vitro studies determined that vCAFs promote cancer cell invasion. Further characterizations of these different CAF subtypes indicated that cCAFs, which share a similar transcriptome profile with vCAFs, are classified as a proliferative vCAF subpopulation. Consistently, scRNA-seq analysis showed that vCAF and cCAF cluster together but differ in the expression of cell cycle-associated genes, which are enriched in the cCAF subtype. Finally, in support of the central role of CAFs in tumor progression, the vCAF signature was identified as an independent prognostic marker associated with the risk of developing metastatic disease (32, 44).

CD10⁺GPR77⁺ CAFs, identified in breast and lung cancer models, have been characterized as a unique CAF subpopulation that promotes tumor formation, reduces patient survival, and induces chemoresistance. CD10⁺GPR77⁺ CAFs mediate these effects by providing a survival niche for CSCs through the constant production of IL-6 and IL-8 (73). Increases in IL-6 and IL-8, which are well-established iCAF markers, are promoted by an intricate positive autocrine loop. Specifically, CAFs self-sustain GPR77 signaling, a receptor for complement C5a, by producing autocrine C5a, which in turn drives persistent NF-κB activation. This activation is specifically maintained by phosphorylation and acetylation of p65, which drives IL-6 and IL-8 secretion and paracrine sustenance of the CSC phenotype. Additionally, CD10⁺GPR77⁺ CAFs promote chemotherapy resistance in CSCs by enhancing ATP-binding cassette subfamily G member 2 (ABCG2) expression, which has been previously described as the main mechanism of chemotherapy resistance for CSCs (73, 74).

In mouse and human breast cancer, mutually exclusive expression of Podoplanin (PDPN) and S100A4 characterized two main CAF populations: i) PDPN⁺ CAFs (pCAFs), and ii) S100A4⁺ CAFs (sCAFs) (13), which were additionally classified into eight CAF subtypes. Consistent with the idea of a dynamic and evolving TME, these CAF subsets emerge progressively following a time-dependent trajectory initiated by an immunoregulatory transcriptional program and leading to antigen-presenting and wound-healing programs (13). pCAFs comprised six subtypes defined by enrichment in different transcriptomic programs: two “immunoregulatory” subtypes, which express genes related to immune regulation and cell migration (Cxcl12 and serum amyloid A3 (Saa3)); one “wound-healing” subtype, characterized by a wound-healing signature and high α-SMA expression; one “ECM-organizing” subtype, marked by an extracellular fiber organization signature and expression of fibrillin-1 (Fbn1); and two “inflammatory” subtypes, defined by inflammatory signatures and high levels of Cxcl1 and Il-6 (13). pCAFs are thought to originate from normal mammary fat pad fibroblasts and have distinct roles in tumor progression. In vitro co-culture experiments indicated that the “immunoregulatory” pCAFs suppress CD8⁺ T cell activation and proliferation, while “wound-healing” pCAFs secrete high levels of collagen, consistent with their wound healing function (13). Unlike pCAFs, the origin of sCAFs is less clear and they do not seem to converge into a specific cell subtype but rather exhibit a continuum of cell states (13). Despite this, the authors were able to identify two main subsets. While one population (Secreted phosphoprotein 1 (Spp1)^highS100A4^low) was enriched in an antigen presentation signature (e.g., MHC class II, class II antigen beta 1 (H2-Ab1)) and ECM remodeling, the second subset (Spp1^lowS100A4^high) expressed high levels of genes involved in protein-folding and metabolism (e.g., heat shock protein family D member 1 (Hspd1)) (13). Of note, the authors report that pCAFs share similarities with previously described myCAFs and iCAFs. Specifically, the pCAF inflammatory subpopulations share common genes with iCAFs (e.g., Cxcl1, Il-6) while the “wound-healing” pCAFs (expressing α-SMA) resemble myCAFs. On the other hand, the sCAF “antigen-presenting” subtype (expressing high levels of H2-Ab1, CD74 and secretory leukocyte peptidase inhibitor (Slpi)) shares common genes with apCAFs, indicating that these newly characterized CAF subpopulations might play a similar role in multiple cancer types.

The coexistence of multiple CAF subpopulations and their ability to dynamically reprogram their transcriptomic profiles has prognostic and potentially therapeutic implications. In breast cancer patient cohorts, individuals with a higher sCAF/pCAF ratio had improved survival. These data are consistent with the concept that immunoregulatory pCAFs inhibit T cell activation, whereas antigen-presenting sCAFs might activate the immune system, leading to improved clinical outcomes. These findings emphasize the need to characterize and target tumor-promoting CAF subpopulations, while enriching tumor-suppressing populations.

Lung cancer:

Studies in non-small cell lung cancer (NSCLC) identified five CAF subtypes (CAF S1-S5) (75). While the CAF S1-S4 subtypes had been previously described in TNBC and HGSOC (11, 59), CAF-S5 had not, and was defined by the expression of PDPN and FAP and lack of α-SMA expression (75). scRNA-seq data from NSCLC indicated that the CAF-S5 subpopulation shows increased expression of C3, C7, selenoprotein P (SEPP1), and clusterin (CLU) (75). The upregulated expression of C3 and C7 suggests that CAF-S5 might have an inflammatory phenotype (75).

Intrahepatic cholangiocarcinoma (ICC):

scRNA-seq analysis of ICC patient samples identified two additional populations besides the well-established CAF subtypes: i) EMT-like CAFs (eCAFs), characterized by epithelium-specific marker genes such as cytokeratin-19 (KRT19) and -8 (KRT8), and serum amyloid A1 (SAA1) and ii) vascular CAFs (vCAFs), characterized by microvasculature signature genes such as CD146 (MCAM) and high levels of IL-6, which were predominantly found in the tumor core and microvascular regions (63). CD146⁺ vCAFs were shown to promote tumor growth and accounted for most of the fibroblast populations identified in the patient samples analyzed. CD146⁺ vCAFs interact with ICC cells through the IL-6/IL-6R axis, inducing the upregulation of epigenetic modification factors, like the enhancer of zeste homolog 2 (EZH2) in ICC cells, which has been previously associated with more malignant features.

Melanoma:

scRNA-seq studies in murine and human melanoma identified three distinct stromal subsets: Stromal 1 (S1, “immune”), Stromal 2 (S2, “desmoplastic”), and Stromal 3 (S3, “contractile”) (76). S1 and S2 were mostly present in early-stage disease, while S3 was enriched in the later stages. Although all three stromal subsets express markers typical of fibroblasts, suggesting that the stroma analyzed comprises CAFs, individual markers were highly variable across the different clusters (76). These findings corroborate the unique functions and temporal characteristics of the different stroma subtypes. Further characterization determined that S1 and S2 shared high levels of Pdpn and Pdgfra but differed in Cd34 expression, with S1 expressing high levels and S2 low levels. S1 shows an inflammatory phenotype characterized by high expression of genes involved in the recruitment and regulation of immune cells, including cytokines Cxcl12, Csf1, Ccl8, cytokine receptors Il-6ra and Il-6st, as well as complement components C3, C2, and C4b (76). Expression of C3 was specific to the S1 subpopulation and mediated their interaction with macrophages and myeloid cells through the C3a-C3aR axis, modulating their recruitment. Disruption of this interaction reduced tumor growth, impaired macrophage infiltration, and increased CD8⁺ T cell density (76). Conversely, S2 was identified as an intermediate population and was enriched in genes encoding ECM components associated with a fibrotic matrix (e.g., collagen family members like Postn and Tnc). Fibrotic matrices are often associated with tumor progression and immune exclusion, indicating that S2 might drive desmoplastic reaction. Finally, S3, characterized by high α-SMA levels, represents a contractile phenotype, enriched in genes involved in actin cytoskeleton regulation and rearrangement, including rho associated coiled-coil containing protein kinase 1 (Rock1), myosin light chain 2 (Mlc2), and myosin light-chain kinase (Mlck) (76). Unlike S1 and S2, S3 expresses pericyte-associated markers such as neuronal-glial antigen 2 (Ng2), Mcam, and Rgs5 in addition to typical fibroblast markers, complicating the precise identification of their cellular origin (76).

Prostate cancer:

Studies in murine castration-resistant prostate cancer (CRPC) described five different CAF clusters (CAF c1-c5) (77). While c1 and c2 CAFs are characterized by Il-6, Cxcl12, and Pdgfra expression, which is shared with iCAFs, c3 and c4 CAFs exhibited an intermediate state between iCAFs and myCAFs. Of note, c4 also expressed MHC-II-related genes, suggesting its possible function in antigen presentation and anti-tumor immunity. Lastly, c5 CAFs, characterized by a significant increase in α-Sma, Tgfb1, and thymocyte antigen 1 (Thy1) expression, exhibited myCAF features (77). Interestingly, pseudotime analysis revealed a single lineage trajectory, suggesting that c4 and c5 CAFs likely evolve from c1–3 CAFs as the tumor progresses (77). Consistently, c4 and c5 CAFs were exclusively found in CRPC, whereas c1 and c2 CAFs were abundant in other hormone-sensitive tumors (77).

Overall, the identification of unique features in CAF subpopulations derived from different tissues suggests a possible role for the hosting organs in promoting CAF phenotypic switching. Understanding how specific organs promote CAF plasticity could be highly relevant to fully unveil the processes driving CAF heterogeneity.

Cancer cell state:

Cancer cells can directly or indirectly reprogram the epigenomic and transcriptomic profiles of CAFs to promote immune evasion, drug resistance and metastasis (78). Since high levels of heterogeneity have been detected among malignant cells within a tumor (79–82), it is important to identify the causative links between specific cancer cell states and features of the TME (Fig. 3).

Figure 3. Overview of major studies and findings from the “Cancer Cell State” section, focusing on cancer cell states and CAF heterogeneity.

Key mechanisms of crosstalk, including signaling pathways, molecular interactions, and spatial distribution influencing CAF function and plasticity in the tumor microenvironment. iCAF (inflammatory CAF), myCAF (myofibroblastic CAF), pan-pCAF (pan-proliferating CAF), pan-dCAF (pan-developmental CAF). Created in BioRender. Flynn, J. (2025) https://BioRender.com/fu3j89k

In mouse models of TNBC, the production of Hedgehog (Hh) by neoplastic cells reprograms CAFs to provide a supportive niche for the acquisition of a CSC phenotype, which in turn promotes resistance to therapy (83). This occurs via CAF-mediated fibroblast growth factor 5 (FGF5) expression and production of fibrillar collagen (83). In multiple cancer types, neoplastic cells that undergo EMT are positively associated with fibroblasts (78, 84) and negatively with other malignant cells (78). Similar findings were observed in HNSCC, where cells expressing a partial EMT (pEMT) program were enriched at the leading edge of primary tumors and interacted with CAFs (78, 81). Analysis of putative tumor-stroma interactions found that bidirectional TGFβ3-TGFβR2, FGF7-FGFR2, and CXCL12-CXCR7 signaling between CAFs and malignant cells may promote EMT programming (81). Additional studies in clear cell renal cell carcinoma (ccRCC) classified the CAF subpopulation in proximity to mesenchymal cancer cells as myCAFs (85).

Transcriptional heterogeneity has also been identified in CRC, leading to the identification of four consensus molecular subtypes (CMS1–4) (86, 87). Consistent with the aforementioned literature, a recent study corroborated that the mesenchymal-like CMS4 is associated with a high level of CAFs (87). The authors propose a positive regulatory loop mediated by the crosstalk between cancer cells and CAFs to operate and foster the stability of the CMS4 state. Specifically, CMS4-secreted TGF-β and PDGF can activate CAFs, which in turn release factors (e.g., TGF-β, PDGF, IL-11, and ECM molecules) to sustain YAP/TAZ activation in the cancer cells, likely reinforcing their mesenchymal traits (87). Both cancer cell- and CAF-derived factors are likely involved in promoting immunosuppression, endothelial cell infiltration, and angiogenesis (87). An additional cell state positively associated with the presence of CAFs has been identified in microsatellite-stable (MSS) CRC (88). Murine and patient-derived scRNA-seq studies indicated that the high-relapse cell (HRC) state, a residual cancer subpopulation detectable following surgical resection of the primary tumor, promotes the recruitment of α-SMA⁺ CAFs (88). Further characterization using metastatic human-like mouse models of MSS CRC indicated that while T cells infiltrate HRC-rich liver micrometastases, the progressive recruitment of CAFs (α-SMA⁺ and/or POSTN⁺) and macrophages (CD68⁺) within the TME re-localized T cells to the periphery during outgrowth. Consistently, T cell exclusion was shown to coincide with CAF and macrophage recruitment to the metastatic TME (88).

In small cell lung cancer (SCLC), spatial transcriptomic studies have linked the neuroendocrine (NE) and hybrid-NE SCLC cell states to the CAF S3 cell state, characterized by expression of ECM components, collagen formation, and collagen degradation pathways (89). Pairwise correlation of CAF S3 with established CAF signatures revealed a similarity with previously described pan-pCAFs and pan-dCAFs (65, 89). These subtypes share the expression of genes such as tumor endothelial marker 8 (TEM8), FN1, inhibin beta-A (INHBA), POSTN, and THY1, which are associated with cancer stemness and tumor aggressiveness (89).

The central role of different cancer cell states in reprogramming CAFs was further confirmed in preclinical studies in lung adenocarcinoma (LUAD), which demonstrated that cells at either end of the EMT spectrum govern CAF heterogeneity (90). The presence of mesenchymal cells in LUAD is associated with the formation of invasion structures. These structures are initiated by the migration of CAFs to the tips of invasive projections, followed by LUAD cells that maintained contact with the leading CAFs. The reversible EMT state in LUAD is characterized by high levels of ZEB1, which reprograms CAFs through the activation of a secretory program enriched in collagens, matricellular proteins, and inter-alpha-trypsin inhibitor heavy chain 2 (ITIH2). Consistently, phenotypic switching toward an epithelial state mediated by ZEB1 depletion in cancer cells reversed the ability of CAFs to enhance the metastatic potential of mesenchymal LUAD cells. Altogether, these findings suggest that EMT sensitizes LUAD cells to pro-metastatic signals from CAFs (90).

In melanoma, the undifferentiated/invasive/mesenchymal cancer cell state (82, 91) is associated with increases in the TGF-β and YAP/TAZ pathways (92–95), both of which drive the expression of numerous genes associated with CAF activation and the myCAF phenotype (9). In an immunocompetent mouse model of melanoma, single-cell and spatial transcriptomics confirmed the loss of macroH2A in tumors, as observed in advanced human melanoma (96). Loss of macroH2A in murine tumors promotes dedifferentiation along the neural crest lineage towards a state associated with tumor progression, poor prognosis, immune escape, and immunotherapy resistance (96). MacroH2A-deficient CAFs display increased myeloid chemoattractant activity following hyperinducible expression of inflammatory genes as a mechanism of tumor immune evasion (97). Since microenvironmental inflammatory signaling has been shown to induce melanoma dedifferentiation (98), the authors propose that macroH2A deficiency in CAFs may promote melanocyte dedifferentiation. Finally, this study suggests that the increase in inflammation is likely due to the polarization of dermal fibroblasts towards iCAFs at the expense of myCAFs (97).

In a recent publication, Gavish et al. (99) analyzed 77 studies covering 24 tumor types across 1,163 tumor samples, identifying 41 shared metaprograms among malignant cells. The analysis was extended to six common non-malignant subpopulations to identify interactions between cancer cell states and associated TME metaprograms. Five clusters were identified in which specific fibroblast metaprograms result from direct or indirect crosstalk between cancer cells and other subpopulations within the TME. Direct interactions between cancer cells and fibroblasts included the interferon and stress metaprograms (99), previously associated with tumor progression. This study suggests that the correlations between cancer cell states and TME metaprograms are promoted by a multicellular TME network, which could evolve through direct interactions between cell types, transfer of proteins through exosomes, or other mechanisms (99).

In conclusion, the crosstalk between cancer cells, CAFs, and other cell subtypes within the TME can promote tumor-interconnected heterogeneity, facilitated by phenotypic switching in multiple malignant and non-malignant cell populations, which affects tumor progression and response to therapy.

Cancer cell mutations:

Cancer mutations are a novel factor that have been shown to transcriptionally rewire CAFs. Oncogenic or tumor suppressor mutations in cancer cells can trigger extrinsic effects and influence the features of the surrounding CAFs to modulate invasion, metastasis, and response to chemotherapy. Here, we outline how specific tumor-intrinsic mutations affect CAF heterogeneity and promote tumor progression.

In CRC, the KRAS^G12D activating mutation promotes the phenotypic switch of α-SMA⁻ CAF populations into lipid-rich CAFs, characterized by high levels of secreted vascular endothelial growth factor A (VEGFA) (100). Specifically, KRAS^G12D mediates the activation of transcription factor cellular promoter 2 (TFCP2), likely through ERK1 phosphorylation in cancer cells, which promotes upregulation of Wnt family member 5B (WNT5B) and bone morphogenic protein 4 (BMP4) in CRC cells. Cancer cell-mediated secretion of WNT5B and BMP4 drives the transformation of α-SMA⁻ CAF subpopulations into lipid-rich CAFs (100), which promote tumor angiogenesis and disease progression by producing abundant VEGFA. Conversely, studies in pancreatic intraepithelial neoplasia (PanIN) indicated that KRAS^G12D-expressing epithelial cells secrete high levels of leukemia inhibitory factor (LIF), which increases IL-33 levels in CAFs via paracrine activation of JAK1/2-STAT3 (101). These effects were shown to be dependent on oncogenic KRAS^G12D, although the mechanism by which this occurs remains unknown. Secretion of IL-33 led to the reprogramming of different components of the TME, including CAFs, myeloid cells, and lymphocytes, which ultimately promoted an immunosuppressive TME and tumor growth. Despite the nuclear function of IL-33 being previously described (102, 103), ablation of stromal IL-33 or knockout of the IL-33 receptor ST2 in the host TME reverted the effects of IL-33 on tumor growth and increased CD8⁺ T cell recruitment and activation. These findings suggest that the effects of IL-33 are largely dependent on its secreted form rather than DNA binding or intracellular functionality. Interestingly, loss of stromal IL-33 promoted the shift of fibroblasts towards a myCAF phenotype, while CAFs in control tumors exhibited an immunosuppressive, secretory iCAF phenotype (101). A further mechanism proposed for KRAS^G12D PDAC cells to promote CAF heterogeneity foresees the secretion of paracrine factors activating the CXCR2 axis to promote a CAF phenotype with a more secretory function and less fibrogenic features. The authors propose a paracrine signaling loop in which the higher levels of CXCR2 ligands (CXCL1, CXCL5, and CXCL7) in KRAS-mutant cell conditioned media (CM) induce stronger activation of CXCR2 in CAFs compared to cancer cells with wild type (WT) KRAS. CXCR2 activation, in turn, upregulates CXCR2 ligands in CAFs and possibly contributes to sustaining their secretory phenotype. Additionally, CM derived from Kras-mutant cancer cells reduced mRNA expression of α-SMA and ECM proteins (Col1A1, Col4A1), while upregulating tumor-supporting cytokines (IL-4, IL-10, and IL-13) and chemokines (CXCL2 and CXCL7) in immortalized mouse PSCs compared to CM from Kras-WT cells (104). However, treatment of CAFs with CM from KRAS-mutant cells only induced a significant increase in the expression of CXCL7, with a modest effect on CXCL2 levels. Finally, blocking CXCR2 signaling in CAFs reduced NF-κB activity, leading the authors to propose the involvement of NF-κB activation in maintaining this secretory CAF phenotype (104).

As described for KRAS mutations, alterations in TP53 can also shape the pancreatic tumor stroma (105). PDAC cells with a gain-of-function (GOF) p53 mutation have been shown to educate CAFs to establish a pro-metastatic environment for cancer cells. Interestingly, CAFs educated by p53-null cancer cells can be reprogrammed into a pro-invasive subtype by either GOF mutant p53 cells or their CAFs via direct contact or long-range paracrine interactions (105). Mechanistically, crosstalk between GOF mutant p53 cancer cells and the surrounding fibroblasts is partially mediated by NF-κB signaling, which in turn can stimulate heparan sulfate proteoglycan 2 (HSPG2) expression and ECM deposition. Loss-of-function (LOF) studies indicated that HSPG2 is a key factor in promoting an environment that allows cancer cell invasion and metastasis. Moreover, CAF education mediated by GOF mutant p53 cells provide cancer cells with stromal cues to delay and reduce response to gemcitabine/Abraxane, which can be reverted by reducing HSPG2 stromal deposition. This study further supports the potential of anti-stroma therapy for pancreatic cancer (105).

Additional genetic alterations that have been shown to shape the stromal landscape in PDAC are linked to BRCA mutational status. BRCA-deficient cancer cells promote a stressful TME leading to the activation of heat shock factor 1 (HSF1) in a subset of PSCs. HSF1 activation, in turn, drives the transformation of PSCs from a myofibroblastic phenotype into immune-regulatory Clusterin (Clu)⁺ CAFs (106). Clu⁺ CAFs express genes involved in inflammation (IL-6, CXCL12, CXCL1, NFKBIA), ECM remodeling (LIF, COL14A1, HAS1), and angiogenesis regulation (C3, IL-6) (106). In line with the role of HSF1 in regulating CLU expression, higher HSF1 activation was observed in BRCA-deficient patient samples compared to those with WT BRCA, and lower Clu expression was detected in PSCs isolated from Hsf1-null mice compared to WT (106). The significant upregulation of Clu expression was also confirmed in CAFs derived from Brca2-deficient KPC tumors. These findings demonstrate that BRCA deficiency promotes a different stromal landscape in PDAC, however further studies are needed to unravel the specific clinical implication of these alterations within the TME (106).

Cancer cell epigenetic dysregulation can also reshape the TME, impacting tumorigenesis and CAF heterogeneity (107). Histone methyltransferase SET domain containing 2 (SETD2) is the primary histone H3K36 trimethyl-transferase mutated in various cancer types (107). Recent studies have linked SETD2 deficiency in pancreatic tumor cells to tumorigenesis and immune evasion (107). In a PDAC mouse model, loss of Setd2 in tumor cells lead to histone 3 lysine 36 trimethylation (H3K36me3) depletion, resulting in ectopic gain of histone H3 lysine 27 acetylation (H3K27Ac) at multiple loci, including the bone morphogenic protein 2 (BMP2) promoter region. Upregulated BMP2 expression interacted with neighboring CAFs expressing the BMP2 receptor (BMPR1a), inducing their transformation into lipid-rich CAFs (107). These lipid-rich CAFs utilized ABCA8a transporters for lipid transfer into PDAC cells, where they are likely used as energy source to sustain fatty acid oxidation (FAO) and oxidative phosphorylation (OXPHOS) in tumor cells, suggesting a role in tumor progression (107). Blocking the BMP2/BMPR1a axis, either by reducing BMP2 levels in tumor cells or ablating BMPR1a in CAFs, significantly suppressed the transformation of CAFs into lipid-rich CAFs and slowed PDAC progression in vivo (107).

Overall, accumulating evidence suggests that tumor gene mutations may play a crucial role in promoting CAF heterogeneity across cancer types and further contributing to tumor progression.

Extracellular vesicles:

EVs are membrane-enclosed vesicles secreted by a variety of cell types, including cancer and stromal cells (108, 109). EVs mediate the transfer of bioactive cargo, including proteins, lipids, and nucleic acids, that enables the reprogramming of recipient cells to support tumor progression. In the TME, cancer cell-derived EVs transfer factors that promote interactions between cancer and stromal cells, which reprogram the host tissue and promote CAF activation and heterogeneity (108, 110).

In PDAC, data obtained from preclinical models and patient samples indicated that lack of Bcl-2-associated-anthanogene 6 (BAG6), a regulator of EV biogenesis, is associated with an accumulation of iCAFs, characterized by the upregulation of genes such as Saa1/2, Cxcl1, Il-6, Lif, and α-Sma (111), despite the latter gene being a known marker of myCAFs. Furthermore, Bag6-deficiency in mice accelerated tumor growth through increased secretion of tumor-derived EVs and subsequent mast cell (MC) activation (111). This activation is driven by cancer cell-derived EVs containing IL-33, which can bind to its receptor Il-1rl1 on MCs, resulting in the release of a secretome enriched in PDGF and CD73, likely promoting the iCAF phenotype (111). In contrast, in vitro studies suggested that the secretome from MCs stimulated with BAG6-expressing EVs promoted the myCAF phenotype (111).

Multiple studies across different cancer types unveiled an active role for coding and non- coding RNA in EV-mediated CAF activation and heterogeneity. In melanoma, mRNA-containing EVs derived from metastatic cells drive lung fibroblasts to adopt a CAF-like phenotype, characterized by the activation of proinflammatory signaling pathways and the upregulation of cytokines and chemokines (112). These EVs contained high levels of RNAs encoding proinflammatory factors such as high-mobility group box 1 (Hmgb1), thymic stromal lymphopoietin (Tslp), and interferon regulator factor 1 (Irf1) and collectively enhanced the recruitment of immune cells that support melanoma metastasis. In a xenograft model of gastric cancer, EVs containing miRNAs, such as miR-155 and miR-210, have been linked to the transition of resident fibroblasts to an iCAF-like phenotype characterized by increased CXCL1 and CXCL8 secretion (113). Of note, CXCL1 and CXCL8 expression in CAFs have been associated with poor patient outcome (113, 114). Finally, studies in PDAC determined that EVs containing repeat RNAs, such as pericentromeric human satellite II (HSATII) RNA, led to the polarization of myCAFs to iCAFs and correlate with an immunosuppressive TME in several cancer types (115). The EV-mediated increase of HSATII RNA in CAFs activates viral-like responses, such as the interferon regulatory factor 3 (IRF3)-mediated innate immune response (115). Spatial imaging of PDAC primary tumors at single-cell resolution demonstrated that IRF3 activation leads to the downregulation of myCAF gene expression, promoting the transition to the iCAF phenotype (115).

Additional EVs that have been linked to CAF heterogeneity are small EVs (sEVs), like exosomes, which can facilitate intercellular communication through exchange of metabolic and other molecular cargo between cells (108, 109, 116). In CRC, exosomes modulate the metabolism of nearby CAFs through transfer of HSPC111 (117). HSPC111 phosphorylates adenosine triphosphate citrate lyase (ACLY), an enzyme that converts citrate into acetyl-CoA, thus promoting acetyl-CoA production in CAFs. This increase in acetyl-CoA facilitates H3K27 acetylation on the CXCL5 promoter, resulting in increased CXCL5 expression in CAFs (117). Elevated CXCL5 secretion by CAFs, in turn, promotes a positive feedback loop that further stimulates the release of HSPC111-containing exosomes from CRC cells (117). This CAF-driven feedback loop accelerates CRC liver metastasis in preclinical models, and blocking the CXCL5-CXCR2 axis impairs metastasis and tumor progression (117). Additionally, exosomes from primary and metastatic sites in CRC were shown to induce distinct CAFs subsets (118). Proteomic profiling revealed that exosomes derived from early-stage CRC cells expressed high levels of proteins associated with angiogenesis (IL-8, RAS oncogene 10a (Rab10a), and N-Myc downregulated gene 1 (NDRG1)), promoting the transition of fibroblasts into a pro-angiogenic CAF subset (118). In contrast, exosomes from late-stage CRC cells induced the activation of a pro-invasive CAF subset, with upregulation of proteins involved in proliferation (S100-A6 and farnesyl diophosphate synthase (FDPS)), invasion (PDZ and LIM domain 1 (PDLIM1), myosin IB (MYO1B)) and ECM remodeling (MMP11, ECM metalloproteinase inducer (EMMPRIN) (118). In bladder cancer, tumor cell-derived exosomes can influence resident fibroblasts to adopt an iCAF phenotype and secrete IL-6, activating STAT3 signaling and promoting EMT in cancer cells (119). Finally, in ICC, cancer cell-derived exosomes are enriched in miRNAs, including miR-9–5p, which plays a crucial role in modulating the TME (63). These exosomal miRNAs induced IL-6 expression in vCAFs (CD146⁺), which activates the IL-6/IL-6R axis in ICC cells, leading to an increase in the expression of EZH2, an epigenetic factor that regulates tumor progression (63).

These findings support the role of EVs and their cargos in promoting CAF heterogeneity and cancer progression.

Patient demographics, extrinsic stressors, and their effects on CAF heterogeneity

Intrinsic and extrinsic stressors enable CAFs to reorganize the TME, leading to metastasis and therapeutic resistance. Studying patient lifestyles and demographics could help unravel the intricate intertwining of factors implicated in CAF plasticity. Below, we summarize the effects of age, obesity, and extrinsic stressors in promoting CAF heterogeneity.

Age:

Cancer is considered a disease of aging (120), and CAFs undergo epigenetic changes during aging. Hence, it is sensible to think that aging might promote CAF heterogeneity and functional plasticity (121). Although recent studies observed age-related changes in fibroblasts in multiple cancer types (122–128), knowledge about how age affects CAF heterogeneity is lacking. Aging fibroblasts express less collagen but are associated with increased ECM stiffness, likely due to age-related increases in collagen cross-linking. Similarly, age-related alterations in the mammary gland ECM are significantly influenced by fibroblasts (129) and may lead to a stiffer and denser ECM (130). Dense breast tissue has been identified as a risk factor for developing breast cancer and can make tumor detection challenging (131). The importance of assessing breast density is underscored by a recent Food and Drug Administration (FDA) update to the mammography protocol, whereby breast density must be disclosed to patients (132). Given the crucial role of fibroblasts in shaping the ECM, it will be important to assess whether different fibroblast subtypes promote breast density and their involvement in breast cancer initiation. Furthermore, increased ECM stiffness promotes the differentiation of fibroblasts into myofibroblasts and could stimulate the development of an immunosuppressive microenvironment (133). Studies in many cancer types have investigated the interaction between myCAFs and immunosuppressive cells, corroborating the role of myofibroblasts in modulating the immune system. Consistently, in melanoma, aged fibroblasts promote structural changes and crosslinking of the ECM, leading to increased tumor invasion and alteration in immune cell infiltration (124).

Aging is also associated with chronic low-grade inflammation, increased immunosuppression, and a general decline in immune system function. Increased immunosuppression is likely in place to counteract chronic low-grade inflammation (133). Although these processes have not all been specifically described in the context of cancer and CAFs, these broad age-related effects likely contribute to CAF functional plasticity and heterogeneity.

Even though still very few, recent studies have started investigating CAF heterogeneity in the context of aging. In low grade gliomas, more than 400 genes associated with aging CAFs were identified, leading to the identification of two CAF clusters associated with distinct prognostic values and TMEs. Age-related genes were then used to generate an aging CAF scoring system to predict response to immune checkpoint inhibitors (ICI) (134). scRNA-seq analysis of a normal postmenopausal mastectomy sample identified two fibroblast subsets with differential enrichment in ECM genes. A comparison between the transcriptomes of postmenopausal fibroblasts and breast tumors in The Cancer Genome Atlas (TCGA) revealed significant overlap with luminal breast tumors. Such findings indicate that these fibroblast clusters, identified in a normal postmenopausal sample, could contribute to breast cancer initiation in the elderly (135). Overall, multiple studies indicate that aging widely affects fibroblast function and phenotype. Although the idea that aging might affect CAF heterogeneity is becoming more appreciated, additional studies are needed to elucidate the role of aging in this context.

Obesity:

Obesity is associated with an increased risk of aggressive breast cancer and decreased patient survival, potentially due to the altered TME and the mesenchymal-like characteristics of cancer cells in obese patients (136). Transcriptomic analysis of adipose-derived mesenchymal stromal/stem cells (bASCs) adjacent to breast tumors revealed that bASCs from lean patients exhibit upregulation of genes associated with an iCAF phenotype, such as colony stimulating factor 3 (CSF3), CXCL10, IL-1β, and hyaluronan synthase 1 (HAS1) (136). In contrast, bASCs from obese patients showed an increase of myCAF-associated genes, including ACTA2 (α-SMA), TAGLN, and connective tissue growth factor (CTGF). Based on these findings, the authors suggest that the increase in the myCAF subpopulation in obese patients likely contributes to tumor progression, therapy resistance, and poor prognosis (136). In a separate study, scRNA-seq analysis of tumor tissues from breast cancer patients revealed six subpopulations of CAFs, including iCAFs and myCAFs (137). In obese patients, the prevalence of iCAF2 (characterized by COL10A1, COL11A1, INHBA, and POSTN) and myCAF2 (characterized by CCL19, CCL21, cathepsin C (CTSC), fatty acid binding protein 4 (FABP4), and CD36) subtypes were significantly higher compared to lean patients (137). Notably, the myCAF2 subtype exhibited upregulation of lipid metabolism related genes such as FABP4, which were associated with an enrichment in pathways related to lipid and fatty acid transport (137). These findings suggest that myCAF2 may contribute to the altered metabolic state within the breast cancer TME, linking obesity to disease progression (137).

Altogether, these studies suggest an association between obesity and the prevalence of different CAF subtypes, underscoring the importance of characterizing the TME heterogeneity in the context of obesity-associated cancers.

Therapy:

Given the central role of CAF heterogeneity in regulating response to therapy, it is also critical to understand whether therapy itself promotes different CAF phenotypes. Therapy-induced CAF reprogramming has been linked to changes in the tumor immune compartment in multiple cancer types, leading to altered treatment responses in cancer patients. In PDAC, treatment with the TGF-β inhibitor galunisertib drives CAFs toward an iCAF phenotype, marked by upregulation of Ly6C and Il-1a, and reduction of myCAF markers such as Acta2 (α-SMA) and Tagln (138).This shift promotes fibrosis and adaptive resistance by activating NF-κB, increasing Cxcl1 expression and recruiting granulocytic myeloid-derived suppressor cells (MDSCs), which contribute to immune evasion and tumor progression (138). Consistently, preclinical studies in PDAC indicated that tumor-secreted TGF-β mediates the reduction of IL-1R1 expression in CAFs, inhibiting the iCAF phenotype by preventing JAK/STAT pathway activation (50). This results in a phenotypic switch of CAFs in proximity of tumor regions toward the myCAF phenotype, while distally located CAFs remain susceptible to IL-1-induced activation of JAK/STAT signaling (50). Thus, the authors propose that targeting the iCAF subpopulation with IL-1 receptor antagonists and JAK inhibitors could potentially shift the balance toward a tumor-restrictive myCAF phenotype and improve tumor control and therapeutic efficacy (50). In HGSOC, the shift of the CAF-S1 population from an ECM-myCAF (FAP⁺ and α-SMA⁺) subtype to a Detox-iCAF subtype in response to platinum-based therapies has been associated with better patient responses (139). Mechanistically, chemotherapy-induced reduction of ECM-myCAFs leads to an increase in CD8⁺ T cell cytotoxicity mediated by YAP1 pathway downregulation, which reduces immunosuppressive signaling in the TME (139). Studies in oral squamous cell carcinoma (OSCC) determined that following chemotherapy treatment, iCAFs were predominantly found in the tumor bed, whereas apCAFs and myCAFs were enriched in the residual tumor, suggesting that chemotherapy alters the spatial distribution of different CAF subtypes (140). Finally, in CRPC, preclinical studies demonstrated that anti-androgen treatment (e.g. enzalutamide) unleashed TGF-β signaling, which promoted the SPP⁺ myCAF phenotype, characterized by high levels of SOX4, chromatin remodeling and upregulation of pro-tumorigenic pathways, which was associated with resistance to androgen deprivation therapy (ADT) (77). In line with these findings, anti-TGF-β antibody or inhibition of downstream ERK signaling reduced CAF polarization and restored sensitivity to ADT (77). Altogether, these studies highlight the importance of investigating drug-induced mechanisms that regulate CAF heterogeneity in cancer to guide the rational development of combinatory therapies that can selectively target tumor-promoting CAFs.

Additional intrinsic and extrinsic factors:

Additional factors have been linked to alterations in fibroblast morphology and function, including sex and environmental influences (141–143). Phenotypic differences between fibroblasts in males and females translate into varying effects on cancer progression, changes in the TME (142, 144, 145), and response to therapy (146, 147), highlighting the importance of investigating CAF heterogeneity in the context of sex disparities. Similarly, smoking and UV radiation can give rise to pro-tumorigenic fibroblasts (141, 143). While the role of these factors in cancer incidence, progression and mortality is well documented, how they affect CAF subtypes is largely unknown. Further studies are needed to better characterize the effects of the aforementioned factors in promoting CAF heterogeneity.

Effects of CAF heterogeneity in cancer and clinical relevance

Effects of CAF heterogeneity on tumor progression:

CAF plasticity and functional heterogeneity actively contribute to tumor growth by remodeling the TME and ECM and by stimulating survival signaling in cancer cells. Below we summarize some of the main mechanisms linking CAF heterogeneity to tumor progression.

PDAC:

iCAFs are typically associated with increased tumor progression in PDAC (50). In hypoxic regions of PDAC tumors, upregulation of LIF in PSCs has been shown to activate cell-autonomous JAK/STAT signaling, which induces the iCAF phenotype (148). Hypoxia can also trigger HIF1α expression in PSCs, which promotes their transition to iCAFs and promotes tumor progression by secreting VEGF and remodeling the tumor vasculature (148). Additional pro-tumorigenic functions of CAFs have been linked to their ability to interact with and modulate the tumor immune compartment. In PDAC tumors, CAFs are a main source of IL-33, with iCAFs expressing the highest levels (101). Activation of IL-33/ST2 signaling in cells expressing interleukin 1 receptor-like 1 (Il1rl1 or ST2), including mast cells, T_regs, and group 2 innate lymphoid cells (ILC2s), has been linked to poor prognosis in solid tumors (101). In vivo studies using a conditional Il-33 knockout in the stromal compartment (Pdgfra-CreER^T2/+; Il33^f/f-eGFP (CreER;Il33^f/f) mouse model) suggested that ablation of stromal Il-33 led to an increase in the recruitment and activation of CD8⁺ T cells, increase in helper T cells and reduction in tumor growth in PDAC syngeneic models (101). Moreover, in transgenic mice lacking the Il-33 receptor Il1rl1^−/−, PDAC syngeneic tumors displayed a smaller volume compared to Il1rl1^+/+ control mice. Analysis of scRNA-seq of CreER;Il33^f/f tumors confirmed the presence of Il1rl1⁺ receptor in the tumor immune microenvironment, suggesting that the pro-tumorigenic effects of iCAFs in PDAC could be mediated by their ability to modulate the immune system via IL-33 secretion (101).

Similar to iCAFs, myCAFs, apCAFs, and senCAFs have also been shown to present immunomodulatory functions in PDAC (44, 149). Specifically, high levels of the glycosyltransferase ST3 β-galactoside α-2,3-sialyltransferanse 4 (ST3GAL4) in CAFs are associated with an enrichment of sialic acid-containing glycans. Despite ST3GAL4 expression being significantly higher in myCAFs compared to iCAFs, the authors disclose that they did not identify significant differences in α2–3 sialylation genes between transitional CAFs, iCAFs, or myCAFs, suggesting that sialylation might be an overall signature of all CAFs. Sialic acid-containing glycans have been widely described as immunosuppressive carbohydrates, acting as ligands for inhibitory Siglec receptors on immune cells. In vitro studies confirmed the ability of ST3GAL4-expressing CAFs to promote the differentiation of monocytes to immunosuppressive tumor-associated macrophages (TAMs) likely via sialic acid-siglec-9 interaction (150). Despite this knowledge, future studies are needed to evaluate the role of CAF-derived sialic acids on tumor progression and immune modulation in vivo. The immunomodulatory function of apCAFs has been associated with their ability to induce transition of naïve CD4⁺ T cells into T_regs in an antigen-specific manner, contributing to immunosuppression (151). On the other hand, senCAFs facilitate pro-tumor and immunomodulatory effects by regulating the infiltration of activated CD8⁺ T cells. In support of this concept, genetic or pharmacological depletion of senCAFs in mouse PDAC tumors enhanced activated CD8⁺ T cell infiltration (149, 152).

Lastly, previous studies described the secretion of CAF-derived growth factors as one of the mechanisms used by the stroma to promote tumor progression (44). In PDAC, a subclass of myCAFs characterized by TGF-β-mediated increase of amphiregulin and sustained autocrine epidermal growth factor receptor (EGFR) activation has been shown to induce EMT and promote metastasis (153).

Breast cancer:

CAF-S1-S4 subsets have been identified as prevalent subpopulations in breast cancer lymph node metastases, highlighting their pro-invasive properties (41). CAF-S1 can promote tumor cell migration and EMT initiation via CXCL12 and TGF-β signaling, whereas CAF-S4 trigger cancer cell invasion and motility through NOTCH signaling (41). Conversely, the functions of CAF-S2 and -S3 have not been fully characterized (41). Additional functions of CAF-S1 include their ability to promote an immunosuppressive microenvironment (154). Mechanistically, CAF-S1-derived CXCL12 recruits and retains CD4⁺CD25⁺ T lymphocytes in the tumor, likely through high cell-surface levels of OX40L, PD-L2, and JAM2. Furthermore, expression of B7H3, CD73, and DPP4 on CAF-S1 cell membranes enhances T lymphocyte survival while promoting their differentiation into CD25^highFOXP3^high, which inhibits effector T cell proliferation in the tumor (154). These findings were further corroborated in TNBC patient samples enriched in CAF-S1 in which high levels of FOXP3⁺ T lymphocyte accumulation were observed (154). Another CAF subpopulation associated with breast cancer progression is senCAFs. scRNA-seq/GSEA analysis of mouse breast cancer tumors confirmed that senCAFs, a subset of myCAFs, are enriched in genes related to ECM organization, TGF-β signaling and collagen fiber assembly. The generation of a collagen-rich ECM by senCAFs has been shown to directly contribute to breast cancer development through inhibition of natural killer (NK) cell maturation and cytotoxicity (60). Finally, in agreement with the immunosuppressive role of iCAFs described in other cancer types (44), an enrichment in the iCAF gene signature in TNBC patients was significantly associated with cytotoxic T lymphocyte (CTL) dysfunction (14).

Bladder cancer:

Studies conducted in bladder cancer patient samples found that elevated levels of CXCL12 in the TME, a chemokine primarily produced by iCAFs, are associated with poor prognosis (12). CXCL12 accumulation correlates with an enrichment in the TAM gene signature (12) in the TCGA bladder urothelial carcinoma (BLCA) cohort (12). Since the CXCL12 receptor, CXCR4, is highly expressed on immune cells, the authors hypothesized that the CXCL12-CXCR4 interaction drives TAM accumulation in the TME (12). Additional factors secreted by bladder cancer iCAFs include VEGF, FGF, and IGF1, which were predicted to contribute to tumor proliferation and progression (12, 155, 156).

NSCLC:

A CAF subtype that has been explored in NSCLC, yet remains a topic of debate, is apCAFs. On one side, apCAFs have been proposed to promote NSCLC bone metastasis by activating the SPP1-CD44 and SPP1-PTGER4 signaling pathways associated with cancer stemness (157). Conversely, other studies foresee an anti-tumorigenic role for apCAFs. Although apCAFs do not express common co-stimulatory molecules, they produce the pro-survival ligand complement component 1q (C1q), which acts on C1q binding protein (C1qbp) on CD4⁺ T cells to activate and rescue CD4⁺ T cells from apoptosis. Overall, this interaction was found to expand the pool of CD4⁺ T cells and reduced tumor burden (158).

Papillary thyroid carcinoma (PTC):

Studies in PTC revealed pro-tumorigenic functions for iCAFs and myCAFs. Computational analysis of cell-cell communications based on the expression of ligand-receptor pairs inferred that iCAFs can interact with CD8⁺ T cells, NK cells, and tumor cells through CCL5-ACKR4 signaling, corroborating the role for iCAFs in immunoregulation (159). On the other hand, this analysis revealed that myCAFs appear to be deficient in intercellular crosstalk in the TME due to the lack of enriched ligand-receptor pairs. As a result of this analysis, the authors suggested that myCAFs might use mechanical and chemical influence to impact tumor progression in PTC (159).

Other cancers:

In gastric cancer, CAF-derived CXCL12 has been shown to activate CXCL12-CXCR4 signaling in tumor cells, which enhances integrin β1 clustering and promotes an invasive phenotype (160). Consistently, elevated levels of CXCL12 in gastric tumors have been linked to increased tumor size, greater tumor depth, lymph node invasion, and poor prognosis (160). A similar increase in CAF-mediated CXCL12 secretion has been observed in OSCC. Mechanistically, cancer cell-derived lactate activates the PI3K-Akt-HIF1α axis in fibroblasts, promoting their transformation into iCAFs (161, 162). Activation of the PI3K-Akt-HIF1α pathway in iCAFs results in CXCL12 secretion, leading to an increase in T_regs infiltration, which has been shown to limit anti-tumor immunity (162, 163). The pro-tumorigenic/immunosuppressive role of iCAFs has been investigated in CRC as well. In this cancer type, a fibroblast subpopulation characterized by autocrine pro-tumorigenic IL1R1⁺/IL-1-high-signaling has been identified. This signaling pathway has been shown to drive the iCAF phenotype, promote tumor growth, reduce T cell proliferation, and increase polarization of macrophages towards an ‘M2’ or pro-tumorigenic phenotype (164). These findings demonstrate the ability of IL1R1⁺ iCAFs to suppress antitumor immune responses and promote tumor progression in CRC. In ICC, both myCAFs and iCAFs have been shown to promote tumor progression using different signaling pathways. While iCAFs promote tumor growth by secreting HGF, which activates the MET receptor on cancer cells (165), myCAF pro-tumorigenic effects have been linked to the expression of hyaluronan synthase 2 (HAS2), but its precise function needs to be further investigated (165). Finally, studies in HGSOC identified a tumor mesenchymal molecular subtype characterized by high levels of stroma and expression of stromal-related gene signatures, which was associated with poor patient survival (166). Similar to findings in TNBC, CAF-S1 found in the stroma of the mesenchymal HGSOC subtype demonstrate immunosuppressive functions by promoting attraction, survival, and differentiation of CD25⁺FOXP3⁺ T_regs (166). Expression of the chemokine CXCL12β by CAF-S1 is required for T cell attraction, and once in contact, CAF-S1 utilize CD73, B7H3, and IL-6 to amplify the survival and number of CD25⁺FOXP3⁺ T_regs (166).

In conclusion, different CAF subtypes can promote tumor progression through divergent or common mechanisms. Understanding the characteristics and functions of the different CAF subtypes is necessary to aim the design of effective therapeutic approaches that target pro-tumorigenic pathways activated in the different CAF subsets.

Effects of CAF heterogeneity on response to therapy:

CAF heterogeneity plays a key role in promoting therapy resistance (167, 168). Hence, the plasticity of the different CAF subtypes is a topic that needs further investigation in the context of cancer therapy. Here, we provide an overview of how CAF heterogeneity influences therapeutic outcomes.

myCAFs:

The role of myCAFs in modulating therapy response was first investigated in PDAC (169). myCAF-mediated deposition of ECM proteins like type I collagen and FN1 results in a stiffened ECM which can limit drug delivery to the tumor cells through diffusion (170–172). In breast and pancreatic cancers, ECM stiffening has been associated with resistance to paclitaxel and doxorubicin chemotherapies by promoting EMT, cancer cell survival, and proliferation through YAP/TAZ pathway activation (173, 174). Furthermore, TGF-β signatures associated with myCAF phenotypes correlate with poor responses to anti-PD1 in patients with melanoma, urothelial, and colorectal cancers (52, 175–178), likely via TGF-β-mediated changes in immune infiltration (175). Likewise, in PDAC, TGF-β-driven LRRC15⁺ myCAFs correlate with poor clinical response to anti-PD-L1 (52). Finally, in CRPC, myCAFs have been shown to secrete SPP1, which binds to integrin receptors and activates ERK signaling, promoting prostate cancer cell survival under ADT (77).

iCAFs:

As with the myCAF phenotype, the role of iCAFs in promoting resistance to therapy has been investigated across cancer types. In rectal cancer, iCAFs have been shown to confer chemoradiotherapy resistance in patients who did not exhibit complete tumor regression. Mechanistically, following irradiation, cancer cell-derived IL-1 polarizes CAFs towards the iCAF phenotype, triggering inducible nitric oxide synthase (iNOS) and subsequent nitrite production, which leads to oxidative DNA damage. Further DNA damage caused by irradiation or chemotherapy triggers a p53-dependent senescence program in CAFs, leading to the secretion of cytokines and ECM constituents that counteract the irradiation-induced tumor cell death (179). The central role of IL-1 in rectal cancer chemoresistance and CAF polarization was further validated with LOF studies. Blocking IL-1 signaling, either using the IL-1 receptor antagonist anakinra or via fibroblast-restricted Il-1r depletion in Col1a2CreER^T2;Il1r^F/F mice, reverted iCAF polarization and rendered tumors sensitive to irradiation (179). In bladder cancer, overexpression of LOXL2 in iCAFs promotes resistance to immunotherapy through IL-32 (180), a pro-inflammatory cytokine that contributes to the formation of an immunosuppressive TME (181). Additional studies in TNBC propose the interaction of iCAFs with lipid-associated macrophages via the CXCR4-CXCL12 signaling axis, as a mechanism to inhibit CD8⁺ T proliferation and effector functions (182). Precisely, CAFs derived from TNBC tissues induced the reprogramming of monocytes towards a STAB1⁺TREM2^high immunosuppressive phenotype, leading to the inhibition of T cell activation and proliferation (183). In line with these findings, this STAB1⁺TREM2^high phenotype was shown to be enriched in TNBC patients resistant to anti-PD1. Finally, computational studies in HNSCC patient samples indicated that iCAFs expressing CDC28 protein kinase regulatory subunit 2 (CKS2) correlated with poor prognosis, reduced response to immunotherapy, and decreased infiltration of cytotoxic CD8⁺ T cells and NK cells (184).

apCAFs:

apCAFs can influence tumor immune responses due to their ability to present antigens to CD4⁺ T cells (10). apCAFs have been associated with immunotherapy resistance in breast and pancreatic cancers (22, 44, 175). Given the lack of co-stimulatory molecules required to induce full CD4⁺ T cell activation on apCAFs, it has been proposed that apCAFs can lead to T cell anergy or induction of T_regs through antigen-specific MHC II-TCR (T cell receptor) interaction, contributing to immunosuppression and immunotherapy resistance (22).

In summary, CAF heterogeneity plays a pivotal role in driving therapy resistance and tumor progression through multiple mechanisms including immunosuppression, maintenance of cancer stemness, and ECM remodeling (167, 168).

Targeting stroma heterogeneity to improve response to therapy:

Given the extensive involvement of CAFs in ECM remodeling, suppression of anti-tumor immunity, and therapeutic resistance, strategies targeting CAFs have generated significant interest. Modulating key signaling molecules and pathways in CAFs could represent an effective approach to target the heterogeneous and plastic nature of the tumor-supporting microenvironment. Here, we provide an overview of preclinical studies exploring different strategies to target pro-tumorigenic CAFs.

The superior efficacy and safety profile of small targeting inhibitors, compared to traditional chemotherapy agents, has led to their emergence as a cornerstone of modern cancer therapy. Nintedanib (BIBF1120), a small molecule inhibitor of receptor tyrosine kinases (RTKs) including FGFR, PDGFR, and VEGFR, is approved by the FDA for the treatment of idiopathic pulmonary fibrosis. In vitro studies in NSCLC have shown that Nintedanib suppresses activation of patient-derived CAFs in response to TGF-β1 in lung adenocarcinoma (ADC). The selective antifibrotic effect was shown to be mediated by a reduction in the secretion of CAF-derived pro-tumorigenic factors and by a decrease in the expression of fibrillar collagens such as COL1A1 and COL3A1. (185). Additionally, in PDAC, Nintedanib increases the therapeutic efficacy of mesothelin-targeted chimeric antigen receptor (CAR)-NK cells by targeting PDGFR-β signaling in CAFs. Specifically, in vitro and in vivo studies, using co-culture models of PDAC cancer cells and patient-derived CAFs, demonstrated that Nintedanib-mediated inhibition of PDGFR-β signaling decreases CAF activation and growth, leading to a significant reduction in IL-6 secretion. This reduction, in turn, increases the infiltration, activation, and killing efficacy of CAR-NK cells. In line with the role of IL-6 in modulating NK function, high levels of CAF-secreted IL-6 have been previously associated with PDAC metastasis via STAT3 activation and induction of NK dysfunction (186, 187). Comparably, in mouse models of melanoma, Nintedanib enhances anti-tumor immunity and improved the efficacy of ICI by targeting CAFs and enhancing CD8⁺ T cell infiltration and activation (188).

Monoclonal antibodies are another class of targeted drugs that have emerged as potent therapeutic agents, revolutionizing the landscape of modern medicine. IL-6/IL-6R signaling is commonly upregulated in CAFs across tumor types (189, 190) and high levels of IL-6 characterize the iCAF phenotype. Preclinical studies in gastric cancer suggest that Tocilizumab, an anti-IL-6R monoclonal antibody, significantly enhanced the efficacy of 5-fluorouracil (5-FU) by preventing CAF-mediated chemoprotection. CAFs exert their chemoprotection action by secreting IL-6, which activates the JAK1/STAT3 pathway in tumor cells, leading to the upregulation of anti-apoptotic proteins like Bcl-2 and survivin while suppressing pro-apoptotic markers. Tocilizumab-induced inhibition of IL-6/JAK1/STAT3 signaling, is thought to promote apoptosis in cancer cells leading to a reduction in tumor growth (191). Targeting CAF monoclonal antibodies have been also investigated in other cancer types. In PDX models of breast cancer, GPR77 neutralizing monoclonal antibodies significantly decreased CD10⁺GPR77⁺ CAF infiltration, suppressed tumor growth, and increased the efficacy of docetaxel (192). Finally, given the central role of TGF-β signaling in promoting myCAF differentiation (50) and its association with immune exclusion (193) and lack of response to ICI, antibodies targeting TGF-β have been explored as a strategy to prevent CAF activation and immunosuppression. In a murine model of metastatic urothelial cancer, the co-administration of anti-TGF-β and anti-PD-L1 antibodies decreased TGF-βR signaling in stromal fibroblasts and increased CD8⁺ effector T cell infiltration into the tumor, leading to enhanced anti-tumor immunity and tumor regression (193).

CAR-T cell therapy is a promising novel treatment for cancer that uses genetically engineered T cells isolated from the patient to recognize and attack cancer cells. However, given the increasing evidence foreseeing the stroma as a central component in promoting tumor progression, CAR-T cells targeting the tumor-promoting stroma have been developed. In preclinical models of NSCLC and B-cell lymphoma, the administration of CAR-T cells targeting FAP⁺ CAFs reduced tumor growth and increased the antitumor activity of CAR-T cells targeting the tumor-associated antigen EphA2, without significant toxicity (194). Furthermore, in immunocompetent mouse models of PDAC, FAP-targeted CAR-T cells improve the efficacy of sequential claudin18.2-targeted CAR-T cells by remodeling the TME, enhancing T cell infiltration, increasing CAR-T cell accumulation, and suppressing MDSCs (195).

iCAFs have been shown to contribute to immunosuppression by restricting CD8⁺ T cell infiltration and fostering an immunosuppressive TME through the CXCL12-CXCR4 axis (196). In preclinical models of breast cancer, blocking CXCR4, a receptor for CXCL12, with the antagonist plerixafor (AMD3100) results in a reduction of α-SMA⁺ CAFs, leading to smaller metastatic nodules, reduction in CD4⁺FoxP3⁺ T_regs, increase in CD8⁺ T cell infiltration, and improved response to ICI (196).

Although not always specifically designed to target CAFs, many current and past clinical trials have been evaluating the effectiveness of the aforementioned agents targeting pathways associated with CAF heterogeneity. An overview of these clinical trials is included in Table 1.

Table 1. Overview of therapeutic agents targeting CAFs or associated stromal components tested in clinical trials.

Trials included were selected based on the following criteria: status marked as “ongoing,” “completed,” or “unknown”; funded by the NIH or non-industry sources (e.g., individuals, universities, organizations); initiated within the last 10 years; classified as interventional; and excluded if focused on hematologic malignancies or leukemia.

Target	Drug	Cancer type	Phase	Status/Outcome
FAP	Nectin4–7.19 and FAP-12 CAR-T cellsa	Nectin4+ advanced malignant solid tumors	I	Unknown NCT03932565
IL-6R	Tocilizumabb + immunotherapy or chemotherapy	Multiple cancer types	I and/or II	Completed NCT04871854 NCT03601611 NCT03135171 NCT02767557 Ongoing NCT06442709 NCT04940299 NCT04729959 NCT04554771 NCT04395222 NCT03999749 NCT03869190 NCT03821246
PDGFR-β signaling	Nintedanibc + immunotherapy or chemotherapy	Multiple cancer types	I and/or II	Completed NCT04046614 NCT03292250 NCT03062943 NCT02835833 NCT02730416 NCT02619162 NCT02568449 NCT02531737 NCT02496585 NCT02399215 NCT02393755 NCT02308553 NCT02225405 Ongoing NCT03377023 NCT02863055 NCT02299141

^a:Nectin4/FAP-targeted fourth-generation CAR-T cells (expressing IL-7 and CCL19, or IL-12).

^b:Anti-IL-6 receptor humanized monoclonal antibody;

^c:Small molecule inhibitor of receptor tyrosine kinases FGFR, PDGFR, and VEGFR.

Altogether, the preclinical studies reported herein highlight the therapeutic potential of targeting CAFs to improve patient response to therapy. However, significant obstacles remain, including the lack of definitive surface biomarkers and the heterogeneous and plastic nature of CAFs. This underlines the importance of fully characterizing CAF phenotypes and functions to guide the design of clinical trials that can improve the treatment of cancer patients.

Conclusions/Future directions

The effects of CAF heterogeneity on tumor progression are profound, influencing diverse aspects such as tumor progression, immune cell infiltration, and therapy resistance. Moreover, intrinsic and extrinsic factors and the bidirectional interactions between cancer cells and CAFs further contribute to tumor functional heterogeneity, shaping disease progression and treatment response. While CAF heterogeneity has been thoroughly investigated in PDAC and breast cancer, a similar approach is lacking in many other cancer types. It is crucial to characterize the multitude of interactions between cancer cells, tumor stroma, and vascular and immune compartments to better understand the dynamic and extensive crosstalk that strongly influences tumor biology. Despite studies linking different cell states to a defined microenvironment are limited, it is becoming increasingly evident that different cell states can mold and subjugate the TME, to induce a distinct multi-compartment metaprogram, to support tumor progression, evade immunity, and promote therapeutic tolerance and resistance. In this scenario it is important to understand the heterogeneity and functional diversity of each tumor compartment (tumor, stroma, vasculature and immune) and the mechanisms in place to promote this inter-compartmental plasticity. Future studies are needed to unveil the role of this multilevel heterogeneity and how each metaprogram affects tumor progression and response to therapy. Identifying factors that support this multi-cellular tumor-TME network will likely aid the design of better therapeutic approaches to prevent tumor recurrence. Moving forward it will be critical to analyze tumors and the associated TME as a single multicompartment entity with the “common scope” of advancing tumor progression.

Single-cell omics analyses offer an invaluable and novel approach to characterizing the heterogeneity and plasticity of the different tumor compartments. However, while these techniques offer unprecedented resolution, their limitations underscore the necessity of integrating spatially resolved transcriptomics and proteomics. In the future it will be critical to merge these complementary approaches, to reach a more complete understanding of CAF biology, elucidating their diverse subpopulations, spatial organization within the TME, and functional roles in cancer progression. Targeting CAFs holds promise as a strategy to improve clinical responses to existing therapies and prevent disease progression. While challenges remain, continued efforts to characterize CAF phenotypes and functions are essential to develop novel therapeutic approaches in cancer treatment.