Recent advances in understanding cancer-associated fibroblasts in pancreatic cancer

Abstract

Pancreatic ductal adenocarcinoma (PDA) is a devastating disease with a poor survival rate. It is resistant to therapy in part due to its unique tumor microenvironment, characterized by a desmoplastic reaction resulting in a dense stroma that constitutes a large fraction of the tumor volume. A major contributor to the desmoplastic reaction are cancer-associated fibroblasts (CAFs). CAFs actively interact with cancer cells and promote tumor progression by different mechanisms, including extracellular matrix deposition, remodeling, and secretion of tumor promoting factors, making CAFs an attractive target for PDA. However, emerging evidences indicate significant tumor-suppressive functions of CAFs, highlighting the complexity of CAF biology. CAFs were once considered as a uniform cell type within the cancer stroma. Recently, the existence of CAF heterogeneity in PDA has become appreciated. Due to advances in single cell technology, distinct subtypes of CAFs have been identified in PDA. Here we review recent updates in CAF biology in PDA, which may help develop effective CAF-targeted therapies in the future.

INTRODUCTION

Pancreatic ductal adenocarcinoma (PDA) is a lethal disease that is projected to become the second leading cause of cancer-related deaths in the United States by the year 2030 (96). The majority of PDA patients typically present with locally advanced or metastatic disease, limiting access to the current most effective therapy, surgery (83, 112). Those patients who are candidates (10–20%) for resection often recur with local or distant disease. Thus radiation and systemic chemotherapy are mainstays for the therapy of PDA. However, a major challenge is that PDA is often resistant to standard therapies. Thus progress for the treatment of PDA is slow and to date, the combined 5-yr survival rate for PDA is still under 10% (111). Immune oncology approaches have revolutionized the treatment of many types of cancer. However, effective immune oncology strategies for PDA have yet to be identified (80). An area of intense effort and interest in PDA biology and therapy development is understanding how the microenvironment of PDA influences tumor development, progression, and response to therapy (34).

Carcinomas have been classically considered as “wounds that do not heal” (30). This is in part due to the similarities between the tumor microenvironment and the wound microenvironment. A substantial feature of many tumors including PDA is the formation of a dense desmoplastic reaction during tumor development (34). This stromal reaction produces a unique tumor microenvironment that consists of heterogeneous cell populations including cancer-associated fibroblasts (CAFs), immune cells, and vascular endothelial cells. These tumor-associated stromal cells exist in conjunction with abundant extracellular matrix (ECM) deposition, forming a fibrotic network that can make up 80–90% of the tumor volume and has important functions in PDA development, invasion, metastasis, immune evasion, and resistance to therapies (49, 82).

Among all the stromal cell populations of PDA, CAFs are probably the most important contributor to desmoplasia; however, they are also one of the least understood stromal cells in terms of their functions, origins and biology (53). For a long time, CAFs were considered as a group of relatively uniform cells within the tumor microenvironment. Most studies consider that these cells have an origin of mesoderm and that they promote PDA progression by producing ECM, chemokines, cytokines, and growth factors (97). Therefore, CAFs have been a therapeutic target of interest given their importance and abundance in PDA. However, the complexity of CAFs was highlighted when several studies ablated CAFs (90, 99). Results from these studies demonstrated that CAFs can have tumor-limiting functions. The emergence of high-throughput single-cell transcriptomic technology has facilitated acute dissection of the heterogeneity of CAFs. In this review, we will discuss the recent advances of CAF studies in PDA and the potential of exploiting CAFs as an effective therapeutic target.

CLASSIC ORIGINS AND FUNCTIONS OF CAFs

Definition of CAFs

Fibroblasts are a group of spindle-shaped cells that are most commonly present in the connective tissue in animals. In cancer, CAFs are classically considered to be all of the fibroblasts found within the tumor microenvironment. Although CAFs were recognized to have “active” and “inactive” states or phenotypes that could be regulated by cancer cells, CAF heterogeneity in terms of phenotype or function was not appreciated until recently (85, 97). Unlike immune cells, which are extensively characterized, precise characterization of CAF subtypes is lacking. Fibroblasts are characterized by mesenchymal features and can be isolated easily and propagated in adherent cell culture (8). Therefore, numerous studies of CAFs have been performed in vitro, under conditions that cannot fully recapitulate the complexity of in vivo microenvironment (33). As a result, there are many conflicting studies on CAF functions, which contribute to challenges associated with understanding CAF biology (68).

Cell of Origin of CAFs

CAFs have been shown to be derived from a variety of cellular sources (Fig. 1). One of the most classic CAF models in PDA is that cancer cells activate normal resident tissue fibroblasts and turn them into CAFs. For example, quiescent pancreatic stellate cells (PSCs) have been considered to be the main source of CAFs in PDA (5, 89). PSCs are resident cells of the pancreas located primarily in the periacinar, periductal, and perivascular spaces of the pancreas. PSCs comprise ~4–7% of total pancreatic cells under physiological conditions. PSCs were found to have a very similar transcriptional profile to hepatic stellate cells. However, the origin of PSCs is still poorly understood. One of the hallmarks of quiescent PSCs is abundant cytoplasmic lipid droplets that contain vitamin A (4). In addition, they also express specific markers such as glial fibrillary acidic protein (GFAP) and desmin. PSCs are thought to be important for maintaining pancreas homeostasis and regulating ECM turnover. In response to pancreatic injury including carcinogenesis, PSCs will be activated and display a myofibroblast-like phenotype by expressing the myofibroblast marker α-smooth muscle actin (α-SMA). At the same time, they will lose the vitamin A-containing lipid droplets (6, 40). In PDA, active PSCs are typically considered CAFs.

Fig. 1.

Cell of origin of cancer-associated fibroblasts. Cancer-associated fibroblasts have been reported to be derived from a wide variety of cells of origin, including organ resident fibroblasts, stellate cells, mesenchymal stem cells, pericytes, endothelial cells, fibrocytes, adipocytes, mesothelial cells, and even epithelial cells through epithelial-mesenchymal transition (EMT).

Besides PSCs, other sources of CAFs have also been reported. Another well-studied source is mesenchymal stem cells (MSCs) (43, 77). During embryonic development, MSCs circulate in the bloodstream to seed emerging sites of hematopoiesis. Circulating MSCs are present in a large number for the first 12 wk of gestation, and the number decreases markedly in adults but remaining MSCs maintain self-renewal capacity (12). However, during injury or other pathological conditions, MSCs can be mobilized and contribute to the wound-healing process (14). Both in vitro and in vivo tracing studies using animal models have demonstrated that MSCs can differentiate into a portion of CAFs in cancers such as gastric, breast, as well as PDA (77, 107, 118, 124). Under certain culture conditions, such as treatment with transforming growth factor-β (TGFβ) or tumor-conditioned media, MSCs can develop characteristics of myofibroblasts by expressing α-SMA and other fibroblast markers such as fibroblast surface protein (FSP), stromal cell-derived factor 1 (SDF-1), and vimentin (61). MSCs isolated from bone marrow and labeled ex vivo can localize to solid tumors after intravenous injection in murine cancer models (27, 81). Another potential source of CAFs that has been proposed are fibrocytes, which are a subpopulation of circulating leukocytes that express hematopoietic cell surface markers such as CD34 and CD45 and ECM proteins such as collagen (7, 24). During wound healing or tumor formation, fibrocytes can enter the tissues and differentiate into fibroblasts (38, 122). Fibrocytes are considered to be monocyte-derived progeny; therefore, they have many immune-related functions by secreting a wide variety of inflammatory cytokines (61, 84, 98). Besides differentiating into fibroblasts directly, fibrocytes can also activate local resident fibroblasts to enhance tissue fibrosis. For example, a population of CD45⁺Col I⁺CXCR4⁺ circulating fibrocytes has been shown to traffic to the lungs in response to CXCL12 and promote the fibrotic reaction in a murine model of bleomycin-induced pulmonary fibrosis (92). In addition, the treatment with a neutralizing antibody against CXCL12 significantly inhibited intrapulmonary recruitment of circulating fibrocytes and attenuated lung fibrosis.

Transdifferentiation from other cell types has also been proposed as an origin for CAFs. For example, lineage-tracing studies have shown that local epithelial cells can undergo phenotypic conversion and gain fibroblastic phenotype [epithelial to mesenchymal transition or (EMT)] (51). Importantly, epithelial cancer cells can also go through EMT and gain very similar features as CAFs (54). Therefore, it is difficult to distinguish mesenchymal cancer cells from CAFs in many instances. CAFs are normally considered to be genetically stable, which should be a distinct difference from mesenchymal cancer cells. However, multiple studies suggest that mutations may also exist in CAFs such as mutations in TP53 (64, 91). Another epithelial cell type that may contribute to the formation of CAFs is the mesothelial cell (60). Mesothelial cells form a continuous layer of epithelial cells known as the mesothelium. It is the largest epithelial organ in the adult mammal covering major body cavities and internal organs. During development, the mesothelium has multipotency and contributes substantially to growing organs by differentiating into various cell types (100). Recent studies have shown that in the adult a subset of mesothelial cells remains multipotent and can differentiate into different mesenchymal cells including fibroblasts during certain conditions such as wound healing, myocardial infarction, peritoneal fibrosis, surgical adhesions, and cancer (32, 101, 110, 115, 121). Other cell types such as endothelial cells, adipocytes, and pericytes have also been shown to be potential sources of CAFs (13, 39, 60, 110, 115).

Given the potential diverse sources of CAFs, it is likely that CAFs in PDA do not represent a uniform cell population. Instead, CAFs in PDA almost certainly represent a collection of multiple subtypes of cells. Furthermore, growing evidence suggests that the phenotypes and functions of CAFs are dynamically regulated and altered at different stages of tumor development.

Classic Functions of CAFs in PDA

Before CAF heterogeneity was appreciated, numerous studies reported the protumorigenic functions of CAFs in PDA (53). This included promoting cancer cell proliferation, invasion, metastasis and inflammation (11, 65, 70, 105, 113, 114). Indeed, in many cancer types including PDA, the accumulation of CAFs and ECM accompanies tumor initiation and progression (42). This suggests that CAFs and ECM deposition are important factors for tumor development. CAFs affect the tumor microenvironment through ECM production, ECM remodeling and the production of cytokines and growth factors (the CAF secretome) (49, 71). Major ECM constituents of PDA stroma include collagens, fibronectin, laminin, and hyaluronic acid (hyaluronan), which are actively produced and secreted by CAFs (119). The ECM forms a three-dimensional noncellular network that is present and essential in every organ (36). However, the excessive deposition of ECM in PDA results in an increase in tissue stiffness and elevated interstitial fluid pressure (IFP) (94). As a result, vascular patency is profoundly reduced and effective drug delivery can be blunted. This certainly influences chemoresponse in PDA, which as mentioned above is limited (47). In addition, the dense ECM is also believed to form a barrier for T-cell infiltration and contributes to tumor immune evasion (102). Interestingly, PDA that harbors TGFβ signaling deficiency tends to have increased matricellular fibrosis and tissue tension, which is activated by protumorigenic Yes-associated protein (YAP) signaling and results in more aggressive tumor phenotype and shorter survival (66).

Besides providing physical cues, the ECM also directly affect cancer cells through signaling. For example, integrins are transmembrane ECM receptors that consist of heterodimers of α- and β-subunits, the combination of which determines the specificity to different ECM proteins (56). As the most abundant ECM proteins in PDA stroma, collagens bind to integrin receptors on cancer cells and activate the downstream focal adhesion kinase (FAK) (78). Similarly, collagens can also activate receptor tyrosine kinases called discoidin domain receptors (DDR1 and DDR2) (117). DDRs are characterized by an extracellular discoidin domain, which can directly bind to the GVMGVO motif within fibrillar collagens. Upon collagen binding, the catalytic kinase domain near the COOH terminus will become activated (49). Unlike integrins, which do not have kinase activity, DDR1 once activated can initiate intracellular signaling events including the downstream FAK family protein PYK2 (48, 109). Integrin-FAK and DDR1-PYK2 signaling can eventually lead to the activation of cJUN and upregulation of mesenchymal adhesion molecule N-cadherin. As a result, cancer cells become more scattered and invasive. This unique EMT phenotype also contributes to resistance to chemotherapy (1), and collagen can also increase the survival of pancreatic cancer cells treated with chemotherapy (2, 21). Furthermore, collagen itself can act as a nutrient source for PDA cancer cells. In the tumor regions with limited fuel such as areas that are distant from vasculature, collagen can be degraded into proline-rich fragments and provide nutrients for cancer cells (87). Collagen degradation is normally mediated by matrix metalloproteinases (MMPs), a large family of zinc-containing proteolytic enzyme such as MMP-2 and MMP-9 (52). CAFs also actively secrete MMPs that degrade collagens (72).

Another way that CAFs promote PDA progression is through their secretome (68, 71). Many paracrine signals have been reported to be secreted by CAFs, these include a wide variety of chemokines, cytokines and growth factors. Many of these factors, such as FGF, HGF, TGFβ, IL-6, CXCL8, CXCL12, and CCL7, can directly act on cancer cells and enhance the proliferation, stemness, and metastatic capacity of cancer cells, resulting in therapy resistance (50, 55, 73, 76, 93). Moreover, paracrine signals from CAFs can also influence other stromal cells and alter the tumor microenvironment. CAFs closely modulate the immune landscape within the tumor through the production of cytokines and chemokines including IL-1, IL-4, IL-6, CXCL1, CXCL2, CXCL8, and macrophage colony-stimulating factor (M-CSF) to create an inflammatory environment and recruit acute responder cells (71, 79). This is similar to the acute phase of the wound-healing process, during which neutrophils and M1 proinflammatory macrophages are among the first responders to be recruited to clear infection and pathogens and ingest and degrade tissue debris to prepare for tissue repair (62, 106). M2 anti-inflammatory macrophages are recruited next in wound healing and these macrophages secrete anti-inflammatory cytokines that dampen inflammation and create a favorable environment for a fibrotic reaction including ECM deposition, remodeling and degradation to heal the wound (123). However, in a malignant lesion, a wound that never heals, these wound-healing-associated stromal cells are coopted by cancer cells to promote tumor development and maintenance. The milieu of the tumor microenvironment alters the functions of the immune cell acute responders such that they become tumor-associated neutrophils (TANs) and tumor-associated macrophages (TAMs) (57). TANs and TAMs secrete various factors to support the survival of cancer cells, and they directly suppress anti-tumor immunity by producing immunosuppressive chemokines/cytokines. For example, TAMs typically function like M2 macrophages and secrete IL-10 and TGFβ, which can directly suppress the activity of cytotoxic lymphocytes such as natural killer (NK) cells and cytotoxic T lymphocytes (CTLs) and induce the differentiation of regulatory T cells (Tregs) (19, 120). Coculture of monocytes with PDA cells and fibroblasts was shown to increase the production of immunosuppressive cytokines and promote the protumor polarization of M2-like macrophages and myeloid-derived suppressor cells (MDSCs) (63, 104). Another study found that culture of peripheral blood mononuclear cells (PBMCs) with PDA CAF conditioned media promoted PBMC differentiation into an MDSC phenotype that suppressed cytotoxic T-cell proliferation in an IL-6/STAT3-dependent manner (74). This study also highlighted IL-6 was an important mediator during this process as its neutralization inhibited the conditioned media-mediated STAT3 activation and MDSC differentiation.

CAFs were also found to produce factors that influence the migration and infiltration of T lymphocytes. In particular, CXCL12 secreted specifically by CAFs activates CXCR4 on CTLs and causes T-cell exclusion (16). Furthermore, CAFs promote immunosuppression by facilitating a predominant T-helper type 2 (Th2) over T-helper type 1 (Th1) lymphoid infiltrate in PDA (23). Thymic stromal lymphopoietin (TSLP), which is secreted by CAFs, can drive Th2 cell polarization and TSLP-dependent induction of Th2-type inflammation is closely associated with reduced patient survival. Taken together, these studies demonstrate that CAFs can directly support cancer cells and shape the immunosuppressive tumor microenvironment that leads to immune evasion in PDA.

Controversies in CAF Function

CAFs had been considered as tumor-promoting until several studies that attempted to target CAFs provided unexpected results. Paracrine Hedgehog (Hh) signaling from cancer cells is known to activate CAFs and promote stromal desmoplasia (88). Binding of Hh ligands to Patched1 receptor relieves repression of the transmembrane protein Smoothened, resulting in the activation of Gli family transcription factors in CAFs (17). Inhibition of Hh signaling was initially thought to be an effective method to prevent fibroblast activation and thus reduce desmoplasia in PDA (88); however, clinical trials to test Hh inhibitors have yielded disappointing results (15, 22, 69). For example, in a phase 2 double-blind placebo-controlled study of the Hh inhibitor saridegib, interim data analysis indicated that median overall survival (OS) for the saridegib plus gemcitabine arm was less than 6 mo whereas the median OS for the placebo plus gemcitabine arm was greater than 6 mo, resulting in early termination of the clinical trial. A later study demonstrated that genetic deletion of the Hh signaling ligand Sonic hedgehog (Shh) in a genetically engineered mouse model (GEMM) of PDA resulted in a reduction of α-SMA-positive CAFs and stroma formation; however, the loss of Shh also led to a more aggressive and undifferentiated tumor phenotype with shorter animal survival (99). In another study, genetic ablation of α-SMA-positive CAFs in a GEMM of PDA resulted in decreased survival with enhanced intratumoral hypoxia and an invasive and undifferentiated tumor phenotype, characterized by EMT, cancer stemness, and therapy resistance (90). In addition, tumor immune evasion was also promoted as evidenced by an increase of Tregs. Compared with the numerous studies reporting the tumor-promoting functions of CAFs, these conflicting studies for the first time highlighted the complexity of CAFs in PDA and underlined the possibility that CAFs might be dynamically regulated and different subtypes of CAFs with distinct functions might exist.

CAF HETEROGENEITY IN PDA

The breakthrough of single cell transcriptomic technology has enabled the characterization of CAF heterogeneity at single cell resolution. CAF heterogeneity in PDA was first identified using an in vitro organoid coculture model (86). In this study, classic CAFs, which have myofibroblast features and express α-SMA, were found located adjacent to cancer cells in mouse and human PDA tissue. This phenomenon was recapitulated by coculturing mouse PSCs and PDA organoids, in which PSCs started to express α-SMA and produce desmoplastic stroma. These α-SMA-positive CAFs were also located adjacent to cancer cells. However, another population of CAFs was also identified with low α-SMA expression and characterized by an inflammatory secretome including IL-6. Moreover, this CAF population tended to be located more distant from cancer cells. Furthermore, these two populations of CAFs appear to be mutually exclusive and coexist in PDA. α-SMA-positive population was named myCAF while the inflammatory population was named iCAF. Multiple groups including ours have applied single cell transcriptomics to further characterize different subtypes of CAFs in PDA. To date, at least four studies have attempted to resolve the features of the CAF populations at a single cell resolution (9, 26, 33, 45). Although CAF subtypes were termed differently in these studies, each study found the existence of the myofibroblastic and inflammatory CAF populations (Fig. 2).

Fig. 2.

Cancer-associated fibroblast (CAF) heterogeneity during pancreatic ductal adenocarcinoma (PDA) progression. Three major subtypes of fibroblast contribute to CAF formation in late-stage PDA. One population has a myofibroblastic feature and expresses myofibroblast markers such as α-smooth muscle actin (α-SMA) and transgelin. This population secretes extracellular matrix (ECM) and has tumor-suppressive function. A recent study (26) reported that a subpopulation of myofibroblastic CAFs marked by leucine-rich repeat containing 15 (LRRC15) is associated with poor response of immunotherapy in PDA. A second population is characterized by the expression of platelet-derived growth factor receptor-α (PDGFR-α) as well as an inflammatory feature that produces inflammatory and immunosuppressive cytokines/chemokines. A third population is derived from mesothelial cells of normal pancreas and can present antigens to T cells. MHC II, major histocompatibility complex II; TGFβ, transforming growth factor-β.

Different CAF Subtypes

The myofibroblastic CAF population was considered to be the only CAF population for a long time, as α-SMA had been used as a pan-CAF marker (53). Single-cell RNA sequencing (scRNA-seq) analyses of mouse and human PDA tumors have found that the myofibroblastic CAFs are marked by the elevated expression of various ECM constituents as well as myofibroblast markers that are involved in cell contraction (e.g., α-SMA and transgelin) (45). This suggests that the function of myofibroblastic CAFs mainly focuses on ECM deposition and remodeling. This is also similar to the fibroblast function during the late stage of the wound-healing process after the acute inflammation is dampened (106).

Importantly, disruption of Hh signaling has been shown to reduce α-SMA-positive fibroblasts and ECM deposition (88, 99). In addition, both the ablation of Hh signaling and α-SMA-positive CAFs resulted in a reduction of stromal content and a similarly undifferentiated tumor phenotype, further indicating the function of α-SMA-positive CAFs may be regulated by Hh signaling (90, 99). However, there were also differences between the phenotypes of the two studies. For example, Shh-deficient tumors had increased vascularity and administration of anti-angiogenesis therapy selectively improved the survival of animals bearing Shh-deficient tumors, while the ablation of α-SMA-positive CAFs led to a significant reduction of tumor vasculature. Another caveat to keep in mind is that besides CAFs, an important vasculature supporting cell type called pericyte is also marked by the expression of α-SMA, which might also be ablated in this study and thus potentially caused the vasculature reduction (29). Although further studies are still required to investigate whether Hh signaling is specific to myofibroblastic CAFs, results from ablation of Hh signaling or α-SMA-positive CAFs support that the myofibroblastic CAF population in PDA has tumor-suppressive functions. Stromal analyses of PDA tumors from clinical samples also provide evidence on the tumor-restricting function of ECM, in which they found that high stromal content in patients correlates with less likelihood of oncogenic mutations, less mesenchymal and metastatic features and longer survival (59). These studies suggest that although ECM secreted by myofibroblastic CAFs can support tumorigenesis in multiple ways, therapies that try to ablate or reduce myCAF function will also result in alteration of the ECM, which has been shown to be a trigger for more aggressive tumor growth.

A recent scRNA-seq study in PDA identified a myofibroblastic CAF population that expresses leucine-rich repeat containing 15 (LRRC15). This submyofibroblastic CAF population, which is programmed by TGFβ, surrounds tumor islets and is absent from normal pancreatic tissue. These findings highlight further complexity of the myofibroblastic CAF population (26). In spite of the potential tumor-suppressive function, analyses of immunotherapy clinical trials that comprise over 600 patients across 6 cancer types including PDA demonstrated that increased levels of the LRRC15-positive CAF signature correlated with poor response to immune checkpoint blockade. This suggests that earlier Hh signaling or α-SMA-positive CAF ablation studies may not be specific enough to reveal the diverse functions of this CAF population. Nevertheless, these studies have provided evidence that myofibroblastic CAF population is associated with immunosuppression. Although increased CD4⁺Foxp3⁺ Tregs were observed when total myofibroblasts were ablated, these tumors instead became more responsive to anti-CTLA4 immunotherapy, which could reverse the disease acceleration caused by α-SMA-positive CAF ablation and prolonged animal survival (90). Therefore, at this point, it might be too simplified to define the myofibroblastic CAF population as tumor-promoting or -suppressive. Although it is still not clear how these functions of myofibroblastic CAFs are dynamically regulated, the evidence is robust that myofibroblastic CAFs only comprise one subtype of CAFs in PDA (Fig. 2).

We have discussed how fibroblasts respond to an injury and secrete factors to recruit immune cells. The existence of inflammatory CAF population suggests that this might be the distinct CAF population that forms the inflammatory tumor microenvironment. Inflammatory CAFs express platelet-derived growth factor receptor-α (PDGFR-α) and high levels of inflammatory cytokines and chemokines (IL-6, CCL2, CCL7, and CXCL12) (33, 45, 86). Inflammatory CAFs hardly express myofibroblast markers and produce a low level of ECM proteins. The secreted factors produced by inflammatory CAFs contribute to cancer progression by recruiting tumor-associated myeloid cells, promoting immunosuppression, and directly stimulating tumor cell proliferation, invasion, and stemness, as well as angiogenesis (50, 68). Therefore, the inflammatory CAFs are highly relevant to tumor progression, immune evasion, and therapy resistance (Fig. 2).

Interestingly, before inflammatory CAFs were characterized by single cell transcriptomics, one study had isolated PDGFR-α-positive CAFs from PDA of GEMM and compared their transcriptional profile with normal pancreatic fibroblasts (25). They found that Saa3, a member of the serum amyloid A (SAA) apolipoprotein family, was the most differentially expressed gene and the major mediator of the tumor promoting activity of PDGFR-α-positive CAFs. Saa3-competent CAFs stimulate while Saa3-null CAFs inhibit the growth of tumor cells in an orthotopic model. In addition, SAA1, the ortholog of murine Saa3, is overexpressed in human CAFs and a high level of SAA1 in the stromal component correlates with worse survival. Previous studies that attempted to target fibroblast activation protein (FAP)-positive CAFs are highly relevant to the inflammatory CAF population (35). FAP-positive CAFs have been considered to be the main source of CXCL12. Depletion of FAP-positive CAFs increased the efficacy of anti-CTLA-4 and anti-PD-L1 therapies in PDA. Furthermore, the combination of CXCR4, the CXCL12 receptor, inhibitor with anti-PD-L1 induces rapid T-cell accumulation in PDA and markedly diminishes PDA cancer cells, presumably through direct T-cell killing. In addition, we recent reported that IL-6 secreted by inflammatory CAFs is a major contributor to tumorigenesis and immune evasion in PDA (50). Consistently, IL-6 inhibition has been shown to sensitize syngenic models of PDA to PD-L1 blockade (75). While the presence of inflammatory CAF features does not correlate with response to immune checkpoint blockade in PDA patients, there is strong evidence that this CAF population is crucial to tumor development and has potential as a therapeutic target.

Besides myofibroblastic CAFs and inflammatory CAFs, a third CAF population was also identified that expresses major histocompatibility complex II (MHC II) genes including the invariant chain CD74 (Fig. 2) (33). These cells can present antigen to and activate T cells in vitro. Given their unique features, these CAFs were termed antigen-presenting CAFs (apCAFs). It is very likely that these cells have important function in modulating tumor immunity. This is an area that needs further investigation. Intriguingly, apCAFs are marked by the expression of Saa3, while in our scRNA-seq data, they clustered together with the myofibroblastic CAF population, suggesting their close relationship or similar cell of origin with the other two CAF populations (33, 45). Importantly, apCAFs were found by another study to be mesothelial cells, which is also consistent with our scRNA-seq data (26). Mesothelial cells are known to have antigen-presenting function that is important in peritoneal immunity (37, 44, 116). During pathological conditions, mesothelial cells can gain mesenchymal cell features and contribute to the formation of stromal cells (60). Therefore, it is important to determine the contribution of mesothelial cells to the tumor microenvironment of PDA in the future.

CAF Subtypes During PDA Progression

By performing scRNA-seq in normal pancreas tissue, early neoplastic lesions and advanced PDA using GEMMs, we found that fibroblast heterogeneity exists in the normal pancreas (45). We identified a population representing the inflammatory fibroblasts and another population with mixed features of myofibroblastic and antigen-presenting fibroblasts. We also identified a third fibroblast population in normal pancreas and early neoplastic lesions, marked by specific gene expression such as Pi16, Ly6c1, and Nov. Gene Ontology analysis showed that the function of this population is associated with opsonization, regulation of epithelial and mesenchymal cell proliferation, and ECM organization, suggesting it may be important for the homeostasis of pancreas. However, in advanced tumors, this population of fibroblasts is not present. By exploring different commonly used GEMMs of PDA (KIC, KPC, and KPfC) that carried Cdkn2a or Tp53 as the second driver mutation, we found that the inflammatory and myofibroblastic CAFs exist regardless of different mutation profiles. In two other independent scRNA-seq studies of normal pancreas or early tumor lesions of mouse and human, the preexisting inflammatory and myofibroblastic fibroblasts were also identified, with one appearing more immunoregulatory and the other demonstrating expression feature of structural support (9, 26). Further differentiation trajectories of expression profiles as well as pseudotime analysis were performed to support that these two separate lineages evolved separately into IL-1- and TGFβ-driven CAFs in the context of PDA. Data from these studies including ours highly suggest that the two CAF populations in PDA are likely to be derived from preexisting fibroblasts of normal pancreas. Future lineage-tracing studies will allow us to further understand how fibroblasts evolve during PDA progression.

Understanding the signaling that drives the desmoplastic reaction and CAF evolution is important. IL-1 and TGFβ are two major paracrine signals that have been previously reported to regulate the phenotype of CAFs (Fig. 2) (10). IL-1 can directly activate IL-1 receptor on CAFs and promote an inflammatory CAF phenotype by downstream NF-κB and subsequent induction of LIF, which activates the JAK-STAT signaling in CAFs in an autocrine manner. In contrast, TGFβ signaling promotes a myofibroblastic phenotype. Further in vivo analysis of fibroblast evolution supports the function of IL-1 and TGFβ in driving the two CAF phenotypes (26). However, due to the dynamics of the CAF phenotypes and the fact that the cellular origins of these CAFs are still largely unknown, the in vivo contribution of IL-1 and TGFβ to CAFs is incompletely understood. For example, although TGFβ was reported to antagonize the inflammatory CAF phenotype, the inflammatory CAF population in established PDA tumors expresses a high level of TGFβ receptors including TGFβR1 and TGFβR2 (50). This suggests TGFβ might regulate certain functions of inflammatory CAFs. We also found that in inflammatory CAFs and mouse fibroblasts, TGFβ has combinatory effect with IL-1 on IL-6 induction, which contributes to tumorigenesis and immune evasion. In addition, induction of IL-6 and LIF by TGFβ has been reported in different types of fibroblasts including CAFs in other cancers (3, 20, 28, 31, 108). TGFβ is a multifunctional cytokine that this expressed robustly during injury and has been reported to be important for multiple steps of the wound-healing process, such as inducing a senescence-associated secretory phenotype and activating fibroblasts to promote wound healing (58). Further the function of TGFβ is highly lineage and transcriptionally dependent. Thus future studies should identify TGFβ-driven transcriptional networks that dynamically regulate specific functional genes in CAF populations and also explore the impact of such axes on PDA development.

TARGETING CAFs IN PDA

The previous disappointing results of targeting CAFs in PDA have made the field very cautious on the development of CAF-targeted therapies. However, due to the identification of CAF heterogeneity, we now have more understanding of CAF biology, which will enable the development of therapies specific for CAF-derived protumorigenic factors.

Although a deeper understanding of CAF functions in PDA and other tumors is needed, multiple studies have revealed potential strategies to target CAFs of CAF functions effectively. Despite the fact that reducing desmoplasia by inhibiting Hh signaling was not successful, another strategy that is being attempted is degrading a key ECM component hyaluronic acid (HA) (103). Preclinical studies and early clinical results suggest that degrading HA can reduce interstitial fluid pressure and improve delivery of chemotherapy into PDA. However, recent results from a Phase III study (HALO301) indicate that the study failed to meet its primary end point (41). A full accounting of the results from the trial is not yet available and will be of great interest to the field. Another strategy is to target CAF-secreted factors. Many CAF-secreted factors promote PDA progression. For example, IL-6 and CXCL12 are potentially viable targets for interrupting CAF driven PDA progression. Blocking the signaling of these factors is feasible and tools are available. A monoclonal antibody against the IL-6 receptor (tocilizumab) (ClinicalTrials.gov Identifier: NCT02767557) and olaptesed pegol a pegulated L-oligoribonucleotide (ClinicalTrials.gov Identifier: NCT03168139) that binds CXCL12 are being tested clinically in PDA patients. In addition, while ECM may have tumor-limiting function, the signaling activated by ECM proteins such as collagen has been shown to be protumorigenic (49). Therefore, specifically targeting ECM-driven signaling pathways is attractive. For example, we demonstrated that pharmacological inhibition of DDR1 in robust animal models of PDA can significantly improve the efficacy of chemotherapy (1). Thus, there are potentially viable strategies to combat CAF-driven PDA progression that might avoid the challenges associated with outright myofibroblast ablation or the challenges associated with Hh inhibition.

CAF HETEROGENEITY IN OTHER CANCER TYPES

Besides PDA, CAF heterogeneity has also been reported in other cancer types. This is not surprising given the diverse cell of origin of CAFs we have already discussed. In triple-negative breast cancer, two myofibroblastic CAF populations were found to be enriched in the microenvironment at the same time (18). One population was involved in cell adhesion, ECM organization, and immunosuppression, while the other was characterized by contracting capability and oxidative metabolism and similar as pericyte. Another study in head and neck cancer using single cell transcriptomic analysis has also revealed three CAF populations, with one expressing myofibroblast markers; a second expressing receptors, ligands, and ECM genes, including FAP and connective tissue growth factor (CTGF); and a third depleted of myofibroblast markers and demonstrating resting fibroblast features (95). Moreover, in a lung cancer scRNA-seq study, seven distinct CAF subtypes were identified (67). Each of these CAF subtypes expresses a unique repertoire of ECM molecules. Two populations were involved in inflammatory response while others were involved in a wide variety of biological processes, such as metabolic regulation and angiogenesis. Although the CAF subtypes of other cancers are not identical to PDA, it is consistent that these CAFs are mostly involved in complex structural and paracrine interactions and immune regulation.

CONCLUDING REMARKS

PDA remains a challenging human disease due to its late detection and unique stroma that contributes to therapy resistance but also contributes to tumor control. CAFs are central to the formation of stroma and tumor progression in PDA. Thanks to the advances in technology, our understanding of CAFs in PDA has improved. In this review, we have discussed the classic studies on CAFs and the recent discovery of CAF heterogeneity. However, many questions remain unanswered, for example, cellular origins of different CAF populations are still unknown. Although evolution analyses based on CAF transcriptomes have been done, robust lineage-tracing studies in GEMMs of PDA are needed to reveal the details of the evolution of different CAF populations. Moreover, functions of CAF subtypes still need to be identified and validated in a context specific manner. Genetic ablation of CAF populations can also provide more information of the distinct functions of each CAF subtype; however, using one CAF marker for genetic ablation might not provide enough specificity; thus genetic strategies such as applying split-cre recombinase may provide tools to address these challenges (46). Taken together, a new age of CAF investigation has begun, further studies on CAF biology in PDA will enable us to develop effective CAF-targeted therapies for this challenging disease.

GRANTS

This work was supported by NIH (National Cancer Institute Grants R01 CA192381 and U54 CA210181 Project 2), the Effie Marie Cain Fellowship. and the Jean Shelby Fund for Cancer Research at Communities Foundation of Texas (to R.A.B.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

H.H. drafted manuscript; R.A.B. edited and revised manuscript; H.H. and R.A.B. approved final version of manuscript.