The term CAF generally refers to stromal cells present in the TME of solid tumors characterized by a distinct phenotype and functional and spatial patterns.

Within the tumor tissue, fibroblast activation relies on factors released by tumor cells and by tumor-infiltrating immune cells. These factors include, but are not limited to, TGF-β, platelet-derived growth factor (PDGF), and fibroblast growth factor 2 (FGF2) ^[13][30][13,30]. Altogether, the factors released in the TME may contribute to the acquisition of a pro-inflammatory gene signature. Indeed, these factors are responsible for driving the differentiation of stromal cells either residing in the tumor area or recruited to acquire functional and phenotypical features attributable to the CAF component, thus inducing a tumor-promoting phenotype in these cells [31].

Several researchers have attempted to identify shared criteria in order to build the basis for studying CAF starting from some common points.

The recent consensus paper published by Sahai and colleagues describes CAF as: (i) negative for epithelial, endothelial and lymphocytic markers such as E-cadherin, CD31 and CD45, respectively, and (ii) positive for markers attributable to a state of fibroblast activation, such as α-smooth muscle actin (αSMA) and fibroblast activation protein (FAP) [14]. Indeed, CAF have been identified both in vitro and in vivo based on proteins related to their activation status [13]. Over the years, these features have been complemented by the analysis of the expression of particular proteins involved in extracellular matrix (ECM) synthesis or modification, such as collagen type I and II, tenascin C, metalloproteinases (MMPs) and tissue inhibitors of metalloproteinases (TIMPs) [13]. Furthermore, CAF modify the TME through the release of factors such as TGF-β, as well as PDGF and FGFs directly affecting the differentiation of immune and stromal progenitor cells recruited in the tumor area (Figure 1B).

However, one major obstacle in the identification of a pan-marker for activated fibroblasts is represented by the marked heterogeneity of fibroblast phenotypes across different types of tumors. Indeed, to date, there are no biomarkers exclusively expressed by CAF, but rather molecules that are indicative of a change of functional state [14]. The consensus paper by Sahai and colleagues defines CAF by the combined expression of TGF-β and αSMA ^[14][32][14,32].

In the last decade, innovative approaches such as single-cell technology have made it possible to bring greater clarity on the complexity of CAF. Indeed, the transcriptional expression profile of CAF at a single-cell resolution has highlighted the great heterogeneity present within CAF, and to also acknowledge the uniqueness of each tumor type in this context ^[33][34][35][33,34,35]. A comparison between breast and pancreatic cancer [36] highlighted the presence of activated fibroblasts with a distinct phenotypic profile, and partially overlapping with that of normal fibroblasts [37]. Evidence of stromal cell variability in cancer emerged from a comparison of the transcriptomic profile of CAF from breast cancer with that of healthy controls via single-cell RNA sequencing (scRNAseq), whereby the presence of four distinct populations of fibroblasts inside were identified in the lymph node of the patients characterized by different metastatization ability [38]. Another study reported the presence of two distinct populations of CAF in 3D cultures of pancreatic ductal adenocarcinoma, one with myofibroblast-like morphology named myCAF, and the other one characterized by the expression of inflammatory cytokines, named iCAF [39]. Specifically, the authors reported the presence of two subsets of CAF characterized by either a myofibroblastic or an inflammatory profile, whose functional characterization correlated with the patient’s clinical response. In both subsets, the differences observed were attributable to differential modulation of TGF-β signaling [39]. Interestingly, it has been reported the possible transition between myCAF and iCAF phenotypes in a cell contact fashion [39], highlighting the plasticity of CAF. Indeed, while the differentiation of myCAF is contact-dependent, that of iCAF may occur in absence of contact with cancer cells. Similarly, functional heterogeneity was also reported in non-small-cell lung cancer, where three CAF subpopulations that differentially expressed hepatocyte growth factor (HGF) and fibroblast growth factor 7 (FGF7), were reported. The first subset characterized by high HGF expression and variable FGF7 exerted a pro-tumor effect suppressing the immune response. The second subset highly expressed FGF7 and was moderately protective of cancers, by moderately suppressing the immune system, while the last subset, presenting low level of FGF7 and HGF, scarcely suppressed the immune system [40].

At the moment, three distinguishable fibroblast subpopulations inside the TME are reported which can be principally classified as: myCAF, iCAF and antigen presenting CAF (apCAF) (Figure 1D).

myCAF represent the major CAF subset present in many tumors, and these cells have been found close to the tumor mass where they perform opposite functions ^[41][42][41,42]. Cytoskeleton proteins such as αSMA and vimentin (Vim) are considered specific to determine myCAF in combination with TGF-β expression and ECM-related proteins. On one side, factors released by myCAF generate a stiff TME, thus favoring, for example, tumor metastasis and chemoresistance ^[43][44][45][43,44,45]. On the other hand, TGF-β can activate multiple pathways leading to CAF activation ^[46][47][46,47]; for example, it can trigger the activation of sonic hedgehog (SHH) signaling in CAF, thus increasing ECM deposition [48]. Indeed, activated fibroblasts release various types of collagens, laminins, fibronectins, proteoglycans, periostins, and tenascin C that favor tumor progression and metastasis [49]. Moreover, myCAF have been shown to produce proteases, such as different members of the MMPs ^[47][48][47,48] that digest the ECM, thus facilitating invasion of tumor cells into the nearby tissue. For example, stromelysin 1 is produced by activated fibroblasts and acts by cleaving E-cadherin and consequently promoting a premalignant phenotype in mammary epithelial cells, characterized by EMT and invasiveness ^[50][51][50,51]. Furthermore, myCAF are able to guide the direct invasion of cancer cells by means of a CAF-cancer cell co-migration mechanism [52], thus suggesting also a role for CAF in governing metastasis.

iCAF are inflammatory fibroblasts distinguished from other subtypes by their expression of the C-X-C motif chemokine 12 (CXCL-12) and interleukin-6 (IL-6) and of the inflammatory macrophage identification marker lymphocyte antigen 6 (Ly6C) ^[39][53][39,53]. Conversely to myCAF that are located close to the tumor mass, iCAF are spatially located at the periphery of the tumor, and their differentiation is induced by interleukin-1 (IL-1) released from nearby macrophage populations (Figure 1C) ^[39][53][39,53]. iCAF, in turn, can release leukemia inhibitory factor (LIF), which is responsible for maintaining an inflammatory state [53] and inducing the recruitment of tumor-associated macrophages (TAMs) [54].

TGF-β1 stimulation was shown also to indirectly induce iCAF in esophageal squamous cell carcinoma (ESCC). Indeed, TGF-β1 stimulates tumor cells to release the chemokine (C-X-C motif) ligand 1 (CXCL-1), which, in turn, has been observed to correlate to iCAF formation [55]. Conversely, others reported the capacity of TGF-β to inhibit Interleukin 1 receptor-like 1 (IL1-R1) expression, thus antagonizing interleukin 1 alpha (IL1-α) responses and blocking iCAF generation [56].

A bioinformatic deconvoluted analysis was shown to discern CAF subtypes providing new information on CAF phenotype. Indeed, in this study, the authors identified several genes associated with myCAF and iCAF in bladder carcinoma (BLCA), squamous cell cancer in the head and neck region (HNSC), and lung adenocarcinoma (LUAD), including 4 overlapping genes between myCAF and iCAF. Notably, gene ontology analysis reported that genes co-expressed in both myCAF and iCAF were involved in the formation of ECM [57]. Furthermore, in this study, an elevated presence of myCAF correlated to poor prognosis in different types of cancer [57]. The poor prognostic value of myCAF has been reported also by others ^[58][59][58,59], and related to the ability of myCAF to decrease the sensitivity to specific antineoplastic drugs, therefore worsening disease prognosis. On the contrary, the role of iCAF is still unclear, as their presence does not determine a significant difference in drug sensitivity, but rather has been associated with a better prognosis regardless of their number [57].

In addition to myCAF and iCAF, another population of CAF has been identified and termed antigen-presenting CAF (apCAF) [60]. These cells express major histocompatibility complex II (MHC-II) and CD74, lack the canonical co-stimulatory molecules, and release prostaglandin E₂ (PGE2). On one hand, the concomitant lack of costimulatory molecules and release of PGE2 was shown to enhance the ability of apCAF to promote the expansion of CD4⁺ regulatory T cells (Tregs) [60]. On the other hand, apCAF were shown to activate CD4 and CD8 T lymphocytes present in the tumor tissue in an antigen-specific manner, thus enriching their phenotype with diverse immunomodulatory functions ^[60][61][60,61]. The expression of programmed death-ligand 2 (PDL2) and Fas ligand (FasL) on these cells allows them to inhibit the cytotoxic activity of CD8 lymphocytes. Indeed, the use of monoclonal antibodies to neutralize the in vivo interaction between CD8 and PDL2 or FasL expressed on CAF was shown to restore the cytotoxic action of T lymphocytes [62].