CAF-immune cell crosstalk and its impact in immunotherapy

It was initially observed that fibroblasts in the TME behaved like reactive fibroblasts that become activated during the process of wound healing [28]. Although a universal marker that defines all fibroblasts in the TME is lacking, numerous markers are described to be expressed by activated fibroblasts in the tumours, among which the two most prominent are fibroblast-activation protein (FAP) and alpha-smooth muscle actin (αSMA) [29]. CAFs are important producers of ECM and growth factors that can directly or indirectly affect tumour cell biology and drive a variety of pro-tumourigenic processes, such as proliferation and invasion [29]. The first hints of CAF heterogeneity arose when researchers tried to eliminate CAFs from tumours and observed contrasting results in preclinical models. While depletion of FAP+ CAFs from the tumour stroma led to tumour regression and improved survival in mouse models of breast and colon cancers [30, 31], targeting αSMA+ fibroblasts or the sonic hedgehog (Shh) signalling in CAFs to reduce the fibrotic tissue around the tumour, also known as desmoplasia, resulted in accelerated tumour growth in pancreatic ductal carcinoma (PDAC) [32, 33]. Interestingly, opposite effects on immune cell composition were observed when the distinct CAF populations were eliminated, with enhanced anti-tumour immunity and an immunosuppressive environment developing when FAP+ or αSMA+ fibroblasts were targeted, respectively. It was now clear that targeting CAFs for cancer therapy would not be an easy task and that a deeper understanding of this cell population would be necessary to make any progress in this field.

CAFs have been traditionally studied using either bulk omics methods, which lack single-cell resolution, or at the single-cell level by immunohistochemistry (IHC) or flow cytometry, which only allows the investigation of a limited number of markers. Advances in single-cell technologies, among which single-cell RNA sequencing and imaging mass cytometry, provided the boosting platform that was necessary. By employing single-cell technologies, different cancer entities have been investigated, with a strong emphasis on PDAC and breast cancer (BC), likely due to their high content of desmoplasia. Among the studied tumour types, numerous subpopulations of fibroblasts have been identified. For simplicity, CAFs are often categorised into three main subpopulations, namely myofibroblasts (myCAFs), inflammatory CAFs (iCAFs), and antigen-presenting CAFs (apCAFs) (reviewed in [5]). Very briefly, myCAFs express high levels of αSMA, secrete ECM proteins in abundance, and are driven by TGFβ. iCAFs on the other hand, secrete high levels of pro-inflammatory cytokines, and their differentiation is induced by IL-1β. The latter subtype, which is often found in less abundance in the TME, is characterised by the expression of MHC class II molecules at the cell surface. The cell of origin (reviewed in [29]) and factors to which fibroblasts are exposed throughout tumour development and progression are some of the factors contributing to the high heterogeneity and plasticity observed in this cell type, which explain differences observed in composition throughout the tumour evolution and between tumour entities.

CAF-immune cell interactions

The mechanisms by which CAFs may alter the tumoural immune landscape are summarised in Fig. 2. Interactions between fibroblasts and immune cells that drive immunosuppression and, therefore, might contribute to the failure of immunotherapies are emphasised throughout the next sections. Nevertheless, examples in which CAFs have notable anti-tumour effects are also provided to highlight the complexity of these interactions and the difficulty of targeting CAFs for cancer treatment. Furthermore, we will mostly focus on the latest findings where heterogeneity of CAFs in the TME was investigated with single-cell technologies since their heterogeneity and in-depth study are of utmost importance.

Fig. 2figure 2

CAF subtypes, their impact on the immune milieu and on the cancer-immunity axis. Green arrows describe a positive correlation or effect while red arrows show an inhibitory effect. Colour-coded squares show the cancer-immunity axis steps which are affected by the TME components shown above. Created with BioRender.com

CAFs and myeloid suppressor cells

CCL2, which has been shown to be secreted by CAFs but also by other cells in the TME, controls the recruitment of monocytes and myeloid-derived suppressor cells (MDSCs) [34, 35]. Interestingly, CCL2-mediated recruitment of myeloid cells was associated with resistance to checkpoint inhibition [35]. In mouse models of different tumour entities, CXCL1, which seems to be exclusively produced by iCAFs [36], promoted the infiltration of polymorphonuclear (PMN)-MDSCs into the TME and drove tumour progression [37]. Importantly, inhibition of CXCR2, the CXCL1 receptor, prevented the migration of PMN-MDSCs to the TME [37]. Selective inhibition of CXCR2 might be an interesting option since this receptor is highly expressed in CAFs and CXCL1-CXCR2 signalling controls the expression of numerous cytokines involved in the recruitment of neutrophils. Moreover, CAF-secreted IL-6, which is primarily associated with iCAFs, promotes the differentiation of myeloid cells into MDSCs in the TME [38, 39]. On the other hand, blocking of TGFβ, a molecule secreted by myCAFs, in preclinical models also resulted in a significant decrease in the amount of myeloid suppressor cells in the TME [40]. A recent study in BC using an orthotopic mouse model showed that fibroblasts in the lung metastatic microenvironment express high levels of CXCL1, IL-6, and CCL2, as well as cyclooxygenase (COX)-2 upon exposure to IL-1β. COX2high CAFs secrete high amounts of prostaglandin E2 (PGE2), which induces the downregulation of molecules important for antigen presentation, including MHC-class II in DCs, and, consequently, impairs CD4+ and CD8+ T cell responses against tumour cells [41]. Moreover, several immunosuppressive genes (e.g. Arg1, Ptg2, Nos2, and Il-10) were also upregulated in DCs and other myeloid-derived cells such as monocytes upon exposure to COX2high fibroblast-conditioned media. It is worth mentioning that COX2high fibroblasts were present in healthy lungs and had an intrinsic immunosuppressive capacity, even in the absence of cancer. It appears that this subpopulation is more predominant in lung tissues compared to all other tissues studied. Importantly, the blockade of COX2-PGE2-EP signalling improved the efficacy of DC therapeutic vaccination as well as PD-1 inhibition [41].

CAFs and regulatory T cells (Tregs)

In BC, Costa et al. identified a subpopulation of fibroblasts characterised by high expression of αSMA and with immunosuppressive properties - CAF-S1. Not only do CAF-S1 secrete high levels of CXCL12, which attracted CD4+CD25+ T cells to the tumour site, but they also induced the differentiation of these cells into CD25highFOXP3high Tregs via high expression of B7-H3, CD73, and dipeptidyl peptidase-4 (DPP4, also known as CD26) in this CAF subpopulation [42]. This goes in line with previous observations that reported a synergistic effect between the targeting of CXCL12 and PD-L1 immunotherapy in pancreatic cancer [43]. Interestingly, Elyada et al. defined CXCL12 as a marker for iCAFs rather than myCAFs in PDAC [44]. Additional dissection of CAF-S1 in BC by Kieffer et al. revealed high levels of heterogeneity within this subpopulation. Eight different CAF-S1 fibroblast clusters were defined, with some clusters actually being classified as iCAFs. The authors further show that differentiation in Tregs is mediated by a myCAF subcluster (ECM-myCAF) rather than iCAFs and that CD4+CD25+ T cells can in their turn affect the phenotype of myCAFs [45]. These findings highlight the difficulty in defining CAF subtypes and could underscore the need for more in-depth studies to understand this cell type and how it can be efficiently targeted in patients. The authors showed a correlation between the presence of ECM-myCAF cluster-specific signatures and the lack of response to PD-1 inhibitors in humans [45]. In another study, a TGFβ signature, which defines myCAFs, was also shown to associate with poor response to ICIs across several cancer types [46]. TGFβ secretion by CAFs is an important regulator of immunity, which in addition to promoting differentiation of Tregs can also directly inhibit cytotoxic T cells [47,48,49] and, consequently, hinder anti-tumour immunity. Furthermore, TGFβ has been shown to induce expression of PD-1 in tumours [50], and engagement of the PD-L1-PD-1 axis can, on its own, drive the formation of Tregs [51]. Indeed, targeting TGFβ in numerous models alters the immune landscape of the tumour and strongly synergises with checkpoint inhibitors [40, 52,53,54]. Additionally, gene expression analysis of CAFs from cancer patients shows a positive and negative correlation between myCAF-signatures and the infiltration of CTLA-4+CD4+ T cells and CD8+ T cells, respectively [42, 45]. It is important to note that TGFβ secretion in the TME is not exclusive from CAFs, and therefore, targeting TGFβ-secreting CAF might not be enough to deplete this molecule from the TME.

Another CAF subtype that has been shown to control Treg differentiation and promote their expansion in the TME is the apCAFs. These mesothelial-derived cells, whose differentiation has been attributed to several factors (IL-12, IFN-γ, IL-1β, and TGFβ), are characterised by the expression of MHC-class II molecules but lack expression of traditional co-stimulatory proteins (e.g. CD80, CD86, and CD40) [45, 46, 55,56,57,58]. Antigen-presentation by apCAFs in the absence of co-stimulatory molecules likely drives an anergic or regulatory state in T cells upon interaction. Interestingly, in human PDAC, the presence of apCAFs positively correlated with Tregs levels, although the authors lacked to show a link with immunotherapy outcome [55].

A recent study shows evidence that Tregs can also modulate the phenotype of CAFs in an IL-1 signalling-dependent manner [59]. The authors show that IL-1R2, a decoy receptor for IL-1β, is exclusively expressed by tumour-infiltrating Tregs in several murine and human cancer types. This results in the inhibition of IL-1β signalling through its main receptor, IL-1R1, which is mostly expressed by CAFs. Inhibition of IL-1β signalling in CAFs results in increased expression of MHC-class II, indicating that the presence IL-1R2+-Tregs in the TME can drive the differentiation of apCAFs, which the authors further describe, promoting the additional accumulation of Tregs. Supporting this, specific blockade of IL-1R2 in Tregs in their murine models improved anti-tumour immunity upon ICI therapy in several murine models [59].

CAFs and effector T cells

CAFs can express checkpoint ligands, such as PD-L1 and PD-L2, and in this way, impact T cell activation [60, 61]. Moreover, secretion of CXCL5 by CAFs in melanoma and colorectal cancer (CRC) mouse models regulates the expression of PD-L1 in tumour cells in a PI3K/AKT signalling-dependent manner [62].

In contrast to the observations reported in the previous section, a tumour-suppressive effect of apCAFs has also been described. A recent study showed a direct effect of this CAF subtype on CD4+ T cells, which was important to control tumour growth. In mouse models of non-small cell lung carcinoma (NSCLC), depletion of apCAFs led to accelerated tumour growth accompanied by decreased numbers of tumours infiltrating CD4+ and CD8+ T cells. apCAFs were shown to promote the survival of effector CD4+ T cells by inhibiting their apoptosis in a C1q-dependent manner [57]. Another interesting observation from Kerdidani et al. was that the tumour-suppressive effect of apCAFs, although observed in different models of lung cancer, could not be replicated in apCAFs derived from BC, indicating a possible tissue-dependent function of this CAF subtype [57]. A study by Hutton et al. has shown that in a PDAC mouse model, CD105 (endoglin) distinguishes two populations of CAFs with contrasting effects on immunity [58]. CD105neg CAFs, which encompassed apCAFs, were able to restrict tumour growth in an adaptive immunity-dependent manner. However, the described effect was independent of the antigen-presenting capacity of apCAFs since depletion of MHC-class II, CD74, and CD80 did not abolish the tumour suppressive effect of the cells. Tumours co-injected with CD105neg CAFs were more infiltrated by T and dendritic cells with higher anti-tumour response signatures compared to their CD105pos counterparts. The authors further showed that these dichotomous populations of CAFs exist in human samples, although CD105 did not bear any prognostic value in human tumours [58]. The contrasting effects of apCAFs in tumour immunity could once again point to a hidden heterogeneity within this CAF subpopulation and would warrant further investigation of which mechanisms and molecules are involved in their activity prior to any attempt to target this subpopulation.

Fibroblasts from several tumour entities (lung, melanoma, and CRC) can also process and present HLA-class I peptides to CD8+ T cells and suppress T cell cytotoxicity through distinct mechanisms. Lakins et al. described a PD-L2 and FAS-L-induced apoptosis of T cells upon antigen cross-presentation by CAFs [61]. Although this effect was not reproduced by Harryvan et al., they observed an increase in the expression of inhibitory molecules (TIM-3, LAG3, and CD39) on the surface of CD8+ T cells after interaction with CAFs [63].

CAFs and tertiary lymphoid structures (TLS)

An interesting structure in the TME that has gained some attention in recent years is tertiary lymphoid structures (TLS). These are well-organised lymph node-like structures formed by immune cells, which can be found in non-lymphoid tissues and often develop in chronic inflammatory diseases but have been reported in certain tumours (reviewed in [64]). In the context of cancer, TLS seem to support anti-tumor immunity and are mostly associated with a favourable prognosis. Interestingly, in chronic inflammation, PDPN+/FAP+ fibroblasts are essential in the formation of these structures through a multistep process involving the secretion of numerous cytokines and chemokines (e.g. IL13, CXCL13, CCL19, and CCL21), and they also drive pathology [65]. In lung cancer, a CCL19-producing population of fibroblasts was associated with enhanced anti-tumour T cell responses and decreased tumour growth [66]. In another recent study, Rodriguez et al. showed a more direct effect of CAFs, with the fibroblast landscape determining the formation of TLS with FAPneg CAFs promoting the assembly of these structures [67]. Although TME-associated fibroblasts have been implicated in the development of TLS, our understanding of this process is still bleak, and this association needs to be further addressed.

ECM impact on the tumour immune milieu

The ECM is present in all healthy tissues, and it is composed of a complex non-cellular mesh of proteins (approx. 300 macromolecules), including collagens, glycoproteins (e.g. laminins, elastin, and fibronectin), proteoglycans (e.g. versican and hyaluran), and polysaccharides [68]. ECM biology has been consistently reported as strongly altered in the tumour context [69,70,71] and is often correlated with the patient outcome, with several studies throughout the years showing a prognostic value of ECM signatures in several cancer entities [72,73,74,75].

Although virtually every cell is capable of secreting ECM components, CAFs are the main architects of the ECM, with myCAFs being the main responsible for the secretion and deposition of ECM [36, 44, 45].

Direct impact of ECM on T cells

Immune cell-expressed inhibitory leukocyte-associated Ig-like receptor 1 (LAIR-1) has been shown to directly bind collagens in vitro [76, 77], which led to the inhibition of LAIR-1-expressing cells, including T cells. High mRNA expression of collagens, as well as LAIR-1, is associated with bad prognosis in multiple tumour types [78, 79]. Moreover, the degradation of collagen by matrix metalloprotease 1 (MMP1), which can be produced by CAFs [80], has been shown to generate LAIR-1-binding fragments. MMP1, collagen, and LAIR-1 expression were also associated with poor prognosis [79]. Importantly, collagen-driven activation of LAIR-1 has also been shown to drive CD8+ T cell exhaustion and dictated the response to PD-1/PD-L1 inhibition in a genetic lung cancer mouse model. In the same study, the authors showed that LAIR-1 and collagen expression in melanoma patients is predictive of ICI success, with higher levels of these markers defining poorer response to therapy and survival [81]. This goes in line with other studies that have identified ECM signatures correlated with CAF activation as markers of immunosuppression and predictors of checkpoint inhibitor response [69]. Transcriptomic analysis of T cells cultured in a 3D model revealed that high-density matrixes characterised by high collagen content drove a TGFβ-induced regulatory-like program in cytotoxic T cells while leading to the downregulation of cytotoxic markers and impairment of autologous cancer cell killing [82]. Tenascin C, another ECM protein, has also been described to inhibit the interaction between T cell-expressed integrin β1 and fibronectin, impairing T cell migration [83]. Galectins, which can be secreted by a variety of cells in the TME, including CAFs, have also been described to modulate the activity of T cells. When covered by galectin-3, TILs failed to trigger lymphocyte function-associated antigen 1 (LFA-1) and, consequently, were unable to establish a functional secretory synapse and to secrete cytokines [84]. Moreover, direct binding of galectin-3 to the TCR can prevent TCR-CD8 colocalisation in TILs and impair cytokine secretion [85].

ECM and immune cell exclusion

The ECM can act as a physical ‘barrier’ to drive immune exclusion [54]. T cells move along collagen matrixes using amoeboid migration. Therefore, perpendicularly oriented and densely packed collagen fibres, which are often found in the tumour periphery, can impair T cell migration. Compared to softer matrixes, T cells migrated slower in vitro when seeded in high-density collagen matrixes [86,

留言 (0)

沒有登入
gif