Öhlund et al. (2017) conducted a unique study on CAFs. First, they analyzed human PDAC fibroblasts using fibroblast-activation protein (FAP) and α-SMA by immunofluorescence, and the results showed that most fibroblasts expressed FAP and low levels of α-SMA, whereas a small fraction expressed FAP and high levels of α-SMA. These results confirmed the heterogeneity hypothesis. Next, they invented a novel 3D coculture system that combined tumor-derived organoids with PSCs. They found that a subtype of CAFs with high expression of α-SMA and low expression of IL6 was activated only when tumor cells came into direct contact with PSCs, and they defined those CAFs as myCAFs. Meanwhile, myCAFs were observed in the periglandular region in vivo, again verifying that the formation of myCAFs requires direct interaction with cancer cells in PDAC. Otherwise, connective tissue growth factor (Ctgf) and collagen type I alpha 1 (Col1α1), which respond to actin alpha 2 (Acta2) and transforming growth factor-β (TGF-β), were upregulated in myCAFs (Öhlund et al., 2017). Elyada et al. (2019) reconfirmed the presence of myCAFs through sRNA-seq and identified a cluster of new marker genes, including transgelin (TAGLN), myosin regulatory light chain 9 (MYL9), and tropomyosin 1 (TPM1). The pathways enriched in myCAFs were identified by gene set enrichment analysis (GSEA), including smooth muscle contraction, focal adhesion, ECM organization, and collagen formation. Protein activity analysis showed activated proteins in myCAFs, including TGF-β1, SMAD family member 2 (SMAD2), and Twist family BHLH transcription factor 1 (TWIST1) (Elyada et al., 2019). Biffi et al. (2019) found that TGF-β promotes fibroblast transformation to myCAFs and inhibits fibroblast transformation to iCAFs by downregulating IL1R1 expression. The TGF-β/SMAD2/3 pathways could induce myCAF activation. They believed that myCAFs and iCAFs are interchangeable, depending on their location and exposure in the tumor (Biffi et al., 2019). Interestingly, Bernard et al. (2019) found that myCAFs were rare in low-grade IPMNs but were highly represented in high-grade IPMNs, suggesting that myCAFs could occur in non-invasive neoplasia. Sun et al. (2021) showed that chemokine (C-X-C motif) ligand 3 (CXCL3)-C-X-C motif chemokine receptor 2 (CXCR2) signaling converted CAFs into myCAFs, which hijacked low metastatic cancer cells for metastasis. They proposed that the “soil-carry-seed” concept for PDAC metastasis meant myofibroblasts (soil) hijack cancer cells (seed) for metastasis (Sun et al., 2021). A study revealed that fibroblasts switch into 15-leucine repeat membrane protein (LRRC15)⁺ myCAFs as a determinant of patient response to cancer immunotherapy, which was associated with poor clinical response (Dominguez et al., 2020). Recently, Chen et al. (2021) revealed that the deletion of Col1 in α-SMA⁺ myofibroblasts using mouse models decreases stromal Col1 content and accelerates the progression of pancreatic cancer. In summary, we conclude that myCAFs show α-SMA^high interleukin-6 (IL-6)^low myofibroblastic features, activated by direct contact with tumor cells, and are located in the periglandular region. The TGF-β/SMAD2/3 pathway could induce myCAF activation in PDAC. myCAFs are related to metastasis and immunosuppression in PDAC. At present, further experimental studies are needed to confirm whether these two cellular states can be converted into each other in vivo and the effect of promoting the occurrence and development of pancreatic tumors. We may build a target to intervene in the treatment of pancreatic cancer to improve the prognosis of pancreatic cancer.

In the 3D coculture system, Öhlund et al. (2017) found a subtype of CAFs with α-SMA^low IL-6^high inflammatory features and loss of myofibroblastic features. The subtype of CAFs was defined as iCAFs. iCAFs were activated by paracrine factors secreted from tumor cells, and their location was distant from tumor cells and myCAFs. They also revealed that IL-6, IL-11, CXCL1, CXCL2, and leukemia inhibitory factor (LIF) were particularly upregulated in iCAFs by scRNA-seq (Öhlund et al., 2017). Elyada et al. (2019) confirmed the presence of iCAFs through scRNA-seq in human and mouse tumors. They found that iCAFs upregulated new marker genes, such as IL6, IL8, and CXCL12. In addition, iCAFs specifically expressed hyaluronic acid synthase 1 (HAS1) and HAS2, two enzymes responsible for the synthesis of hyaluronic acid, which is considered to be an important barrier to the treatment of PDAC. They identified pathways enriched in iCAFs by GSEA, including interferon γ (IFN-γ) response, tumor necrosis factor (TNF)/nuclear factor kappa-B (NF-κB), IL2/signal transducer and activator of transcription (STAT) 5, and IL6/Janus kinase (JAK)/STAT3 in humans. To identify differentially activated proteins between iCAFs and myCAFs, protein-activity analysis was adopted, and IL1R1 and STAT3 were found to be differentially activated in iCAFs. Surprisingly, TGF-β receptors TGF-βR2 and TGF-βR3 were the top differentially activated proteins in iCAFs, which do not show an activated TGF-β program and may be caused by a loss of TGF signaling in iCAFs (Elyada et al., 2019). Biffi et al. (2019) proved that tumor-secreted IL-1 activated the LIF/JAK/STAT pathway to activate iCAFs in human and mouse PDAC and decreased myofibroblastic genes, including the connective Ctgf-TGF-β target gene Acta2-α-SMA gene. Shi et al. (2019) showed that LIF secreted by PSCs was associated with tumor progression in PDAC. Interestingly, Bernard et al. (2019) found that the iCAF population was abundant in PDAC but absent in low-grade and high-grade IPMN, which might show that iCAFs promote IPMN transformation into PDAC. Garg et al. (2018) found that iCAFs release CXCL12 in an NFκB-mediated manner, leading to immunosuppression and promoting tumor growth in PDAC. Above all, we conclude that iCAFs show α-SMA^low IL-6^high inflammatory features, are activated by paracrine factors from tumor cells, and are located away from cancer cells. The Il-1/LIF/JAK/STAT pathway could activate iCAFs, which suppress immune and ECM deposition and accelerate the occurrence and progression of PDAC. iCAFs are closely related to inflammation. Further studies on the relationship between inflammatory factors, iCAFs, and tumor cells can be conducted to determine whether the mechanism of pancreatic tumors can be discovered, and the inhibition of related factors combined with chemotherapy to improve the prognosis of pancreatic cancer is worth investigating in the future.

Elyada et al. (2019) identified a subtype of CAFs that expressed MHC class II family genes. The MHC II family is usually expressed in antigen-presenting cells; therefore, they named this group of fibroblasts “antigen-presenting CAFs” (apCAFs). apCAFs expressed unique genes, including H2-Aa (encoding α-chains of MHC II), H2-Ab1 (encoding β-chains of MHC II), CD74 (encoding invariant chains of MHC II), serum amyloid A3 (Saa3), and secretory leukocyte peptidase inhibitor (SLPI). The pathways enriched in apCAFs were identified, including antigen presentation and processing, fatty acid metabolism, MYC targets, and mTOR complex 1 (MTORC1) signaling. They found activated proteins in apCAFs, except the most highly activated proteins H2-Ab1 and CD74, which also included CD239, CD321, and interferon regulatory factor 5 (IRF5) (Elyada et al., 2019). Dominguez et al. (2020) also found that apCAFs highly expressed CD74, H2-Ab1, and Saa3. Hosein et al. (2019) proved that a subtype of CAFs could perform antigen processing and presentation through the MHC-II pathway and had complement activation functions. apCAFs could express costimulators such as professional antigen presenting cells (APCs), but their levels were significantly lower, indicating that apCAFs act differently from professional APCs in PDAC. In addition, apCAFs are regulated by IFN-γ signaling in vivo and exhibit an antioxidant response (Elyada et al., 2019). In summary, apCAFs show high expression of MHC II family genes; however, the activation conditions and their locations need to be further studied. There is a large space to explore the immune relationships between CAFs and the pancreatic tumor microenvironment and research on targeted drugs in PDAC (Figure 1).

It was believed that as long as fibroblasts were eliminated or inhibited, pancreatic tumor cell growth could be inhibited. Nab-paclitaxel could kill CAFs, thus reducing stromal content and increasing gemcitabine accumulation (Alvarez et al., 2013; Miyashita et al., 2018). However, this hypothesis is under debate due to the lack of therapeutic efficacy (Hidalgo et al., 2015). For a long time, it was customary to think that α-SMA- or fibroblast activation protein (FAP)-positive cells were fibroblasts, which were carcinogenic factors, and studies of therapeutics that targeted α-SMA and FAP have been carried out. Disappointingly, ablation of α-SMA⁺ CAFs caused lower survival rates in an animal model of PDAC due to increased Treg cells and suppressed immune surveillance (Özdemir et al., 2014). FAP in PDAC was highly expressed in CAFs (Aggarwal et al., 2008; Puré and Blomberg, 2018). Various FAP targeting agents, including bispecific antibodies (Wüest et al., 2001; de Sostoa et al., 2019), antibody-drug conjugates (Fang et al., 2016), vaccines (Duperret et al., 2018), and chimeric antigen receptor (CAR) T cells (Lo et al., 2015), all aimed to eliminate FAP⁺ fibroblasts, but anti-FAP therapy resulted in muscle loss, bone toxicities, cachexia and even death (Tran et al., 2013). Targeting α-SMA and FAP has not achieved ideal effects at present, but these preliminary studies also provide a reference for future studies. It has been suggested that depletion of myCAFs leads to the acquisition of EMT, a stem-like phenotype and an undifferentiated state in pancreatic cancer; therefore, inhibition of α-SMA can accelerate tumor progression (Özdemir et al., 2014). FAP⁺ CAFs may be multifunctional mesenchymal stem cells that are recruited into the tumor environment, thus leading to the above malignant results; therefore, the direct inhibition of the FAP⁺ CAF strategy needs to be discussed (Tran et al., 2013).

The STAT3 signaling pathway promotes tumor progression in PDAC (Laklai et al., 2016). Later, Biffi et al. (2019) proved that pancreatic tumor-secreted IL1 activated the LIF/JAK/STAT pathway to activate iCAFs in PDAC to promote tumor growth. This pathway is closely related to iCAFs, and iCAFs are considered to be inflammatory CAFs that promote tumor progression. Therefore, studies of therapies targeting this pathway have been carried out recently. Anakinra, an IL-1R antagonist, which was used to treat IL1⁺ pancreatic cancer tumor cells in mice combined with CAFs, reduced Th2 levels and improved overall survival of pancreatic cancer in a mouse model (Brunetto et al., 2019). IL-1R antagonist-based clinical studies are ongoing (NCT02021422). In another study, Zhang et al. proved that inhibiting IL-1β reduced the fibrosis level of pancreatic tumors and reduced drug resistance in KPC mice (Zhang et al., 2018). Das et al. (2020) found that neutralizing IL-1β with antibodies significantly enhanced the therapeutic efficacy of PD-1 and increased tumor infiltration by CD8⁺ T cells in mice. IL-1β inhibitor-based clinical studies are ongoing (NCT04581343). Targeting IL-1 or its receptors in the presence of CAFs may play a role in future therapies. STAT3 inhibition combined with gemcitabine obviously inhibited tumor growth in a mouse model (Nagathihalli et al., 2015). Ruxolitinib is an inhibitor of the JAK signaling pathway. A phase II study of ruxolitinib combined with capecitabine in clinical therapy has failed in pancreatic cancer. There are some limitations in this study. First, the benefit of ruxolitinib was observed in patients with elevated C-reactive protein (CRP) levels, and modest activity was observed in the intent-to-treat (ITT) population. Second, this was a proof-of-concept study with a limited sample size (Hurwitz et al., 2015). Based on the results of the phase II trial, phase III studies were initiated only in patients with metastatic pancreatic cancer who had failed gemcitabine treatment and had a systemic inflammatory response. Unfortunately, phase III studies have not shown improved overall survival (NCT02119663 and NCT02117479), which suggests that ruxolitinib is an ineffective drug in patients with high levels of CRP (Hurwitz et al., 2018). It is thought that a large number of signaling bypasses and compensatory feedback loops may affect the activity of JAK inhibitors (Hurwitz et al., 2018). Therefore, in the process of CAF activation, it is necessary to further study whether there are bypasses and feedback loops of JAK/STAT3 in the process of CAF activation (Figure 3).

Inflammatory CAFs show α-SMA^low IL-6^high inflammatory feature. CAFs secrete IL-6, which suppresses NK cell activity and leads to PDAC metastasis (Huang et al., 2019). IL-6 promotes the development of PDAC by increasing mitogen-activated protein kinase (MAPK) signaling activation (Zhang et al., 2013). Tumor-induced IL-6 impairs the ketogenic response to inhibit caloric intake, leading to a systemic metabolic stress response that hinders anticancer immunotherapy in PDAC (Flint et al., 2016). A recent study showed that high levels of IL-6 were also related to a reduced response to therapy (Neumann et al., 2018). In addition, it was found that IL-6 secreted by PSCs transformed non-invasive pancreatic progenitor cells into invasive PDAC (Nagathihalli et al., 2016). Blocking IL-6 signaling obviously inhibits primary tumor growth and recurrence in orthotopic xenograft models (Goumas et al., 2015). Mace et al. (2018) inhibited IL-6 and PD-L1 to suppress tumor progression and enhance overall survival in murine models of PDAC, and a combination of targeted therapies that inhibit IL-6 and PD-1 has also recently been launched (NCT04191421). Long et al. (2017) found that IL-6 could activate STAT3 and lead to chemical resistance in PDAC in a mouse model, and the efficacy of chemotherapy in PDAC was improved after the blockade of IL-6 signaling. Higher serum IL-6 was statistically found to be an independent risk factor for PDAC progression to extensive liver metastasis, although it was not associated with longer survival (Kim et al., 2017). During the early stage of pancreatic cancer in mice, liver cells show activation of STAT3 and increased SAA, both of which are dependent on the release of IL-6 into circulation by non-tumor cells. It has also been observed in patients that IL-6-STAT3-SAA signaling promotes the establishment of a prometastatic niche and liver metastasis (Lee et al., 2019). Therefore, inhibition of IL-6-STAT3-SAA signaling can be an effective therapeutic target in the prehepatic niche in PDAC. A clinical trial involving 140 patients with advanced pancreatic cancer was launched, and they were randomly assigned to receive either a chemotherapy plus the IL6 receptor inhibitor tocilizumab or a chemotherapy drug plus a placebo (NCT02767557). As the most important marker of iCAFs, IL-6 plays a very important role. It is worth exploring the effect of IL-6 inhibition or its receptor on tumor progression (Figure 4).

Transforming growth factor-β has been shown to reshape CAFs, which induce increased elongation, lamellipodia formation, cell spreading and spheroid invasion in pancreatic cancer (Stylianou et al., 2018). Another study has shown that CAF-secreted TGF-β drives invasive and proliferative phenotypes in PDAC cell lines (Ligorio et al., 2019). Biffi et al. (2019) found that TGF-β promotes the transformation from fibroblasts to myCAFs. Thus, TGF-β has also become a research target for the treatment of PDAC. Early on, TGF-β was targeted in a mouse model, but the result was a deterioration of tumor differentiation and stage (Hezel et al., 2012). Galunisertib is the first oral inhibitor of the TGF-β receptor, and galunisertib combined with gemcitabine improved overall survival in unresectable patients compared with gemcitabine (Melisi et al., 2018). Huang et al. (2019) found that inhibition of the TGF-β receptor reduced IL-6 production from CAFs, leading to decreased STAT3 activation in tumors and reversed immunosuppression in a mouse model. M7824 is a double-fusion protein that blocks both PD-L1 and TGF-β signaling, which inhibits tumorigenesis in mouse models (Lan et al., 2018). Further study should explore whether inhibition of TGF-β signaling in combination with PD-1 inhibitors or gemcitabine can improve the prognosis of pancreatic cancer, and side effects after the combination also need to be avoided.

Sonic hedgehog is thought to be highly associated with pancreatic cancer stroma formation. A combination of gemcitabine and the Shh inhibitor IPI-926 has been used in a pancreatic cancer mouse model, and the results showed increased gemcitabine concentrations in the tumor and stabilized the disease (Olive et al., 2009). Another study showed that cyclopamine, a Shh inhibitor, combined with chemotherapy in a mouse model of pancreatic cancer showed prolonged survival (Zhao et al., 2018). However, Shh inhibitors have also been shown to increase tumor growth in mouse models of pancreatic cancer and pancreatic intraepithelial neoplasia (PanIN) progression (Lee et al., 2014; Rhim et al., 2014). In a clinical trial, the Shh inhibitor vismodegib combined with gemcitabine vs. gemcitabine alone did not improve survival after PDAC treatment (Kim et al., 2014). In a randomized Phase Ib/II study of gemcitabine plus vismodegib or placebo to treat metastatic pancreatic cancer, there was no difference in overall response rate, PFS, or OS (NCT01064622) (Catenacci et al., 2015). Phase 2 of this study showed the same result (NCT01088815) (De Jesus-Acosta et al., 2020). A study found that Shh signaling was specifically activated in myCAFs. The smoothened antagonist LDE225 inhibited Shh signaling, reduced myCAF numbers and increased iCAF numbers in mice, which resulted in fibroblast composition and immune infiltration in the PDAC microenvironment (Steele et al., 2021). This study suggests explosive conclusions for the treatment of PDAC and provides great inspiration. Targeting a certain type of CAF specifically for heterogeneity of CAFs, rather than all of the CAFs, may lead to new directions for the treatment of PDAC.

Feig et al. (2013) found that FAP⁺ CAFs secreted CXCL12, which resulted in immunosuppression of PDAC. CXCR4 is the receptor for CXCL12, and an inhibitor of CXCR4, AMD3100, induces T-cell accumulation and works with PD-L1 blockade to reduce cancer cells in KPC mice (Feig et al., 2013). Interestingly, Garg et al. (2018) also found that CXCL12 prevented cytotoxic T-cell infiltration in tumors and killed cancer cells. Blocking the effect of CXCL12 on PDAC cells may enhance antitumor immunity (Garg et al., 2018). It has been found that the SDF-1/CXCR4/special AT-rich sequence binding protein 1 (SATB-1) axis induces malignant progression of PDAC, and inhibition of this axis may be a potential therapeutic target of PDAC (Wei et al., 2018). A phase IIa clinical trial was recently conducted to evaluate the efficacy, safety and immune effects of the CXCR4 antagonist BL-8040 with pembrolizumab and chemotherapy in metastatic PDAC (NCT02826486). The results showed that combined CXCR4 and PD-1 blockade expanded the benefit of chemotherapy in PDAC (Bockorny et al., 2020). We expect more amazing results in an expansion cohort trial combining BL-8040 and panitumumab for second-line treatment (Figure 5). CXCL3 has been shown to be highly expressed in PDAC, and its receptor CXCR2 is almost exclusively expressed in CAFs. CXCL3-CXCR2 signaling could stimulate the transformation of CAFs to myCAFs, which secrete type III collagen and accelerate tumor metastasis (Sun et al., 2021). Another study found that inhibition of CXCR2 reversed tumor progression promoted by type I collagen deletion in a PDAC mouse model (Chen et al., 2021). By summarizing the studies on the CXCL12-CXCR4 axis and CXCL3-CXCR2 axis, two potential inhibition pathways for PDAC are currently effective in preclinical studies. A phase IIa clinical trial showed that combined CXCR4 and PD-1 blockade expands the benefit of chemotherapy in PDAC.

Hyaluronidase (HA) has been shown to be very rich in human PDAC extracellular matrix, which can increase vascular pressure and interstitial fluid pressure, reduce immune cell infiltration and drug penetration, and promote tumor growth (Jacobetz et al., 2013; Wang-Gillam, 2019). Studies have found that patients with low levels of HA live significantly longer than patients with high levels of HA (Whatcott et al., 2015). In a mouse model, an HA inhibitor significantly prolonged survival after combination with gemcitabine (Provenzano et al., 2012). Surprisingly, a phase Ib/II-randomized study of FOLFIRINOX plus pegylated recombinant human hyaluronidase 20 (PEGPH20) versus FOLFIRINOX alone in metastatic PDAC patients showed significant side effects and shortened survival (Ramanathan et al., 2019). Then, a randomized phase III trial of PEGPH20 with nab-paclitaxel plus gemcitabine for patients with HA-high metastatic PDAC was conducted, and it could not improve overall survival or progression-free survival (Van Cutsem et al., 2020). PEGPH20 significantly increased gastrointestinal toxicity and thromboembolic events, increased dose delays and dose reductions, and reduced FOLFIRINOX drug exposure (Ramanathan et al., 2019; Van Cutsem et al., 2020). Therefore, HA inhibition is not recommended because it can induce unexpected clinical effects and serious side effects.

Pancreatic cancer has a dense ECM that insulates T cells and prevents drug penetration. Wang et al. used acidic tumor extracellular pH (PHe)-responsive clustered nanoparticles to deliver drugs in a mouse model (Wang et al., 2020). Wiley et al. (2018) found that CAFs increased the expression of GPR68 (a G-protein-coupled receptor), which could sense acidic environments to further increase fibrosis and IL-6 levels, leading to the development of PDAC proliferation. The acidic environment is closely related to the metabolism of CAFs and accelerates the development of PDAC. Adjusting the pH level of the microenvironment in which CAFs are located, we can explore the communication between CAFs and tumor cells to find targets that could inhibit tumor development. Ji et al. (2017) constructed a tailor-designed matrix metalloproteinase 2 (MMP-2)-responsive amphiphilic peptide and phospholipids, which could release pirfenidone at PDAC and downregulate the many components of the ECM expressed by CAFs. Finally, the penetration of gemcitabine into tumor tissue was significantly increased, and the efficacy of gemcitabine in pancreatic tumors was enhanced (Ji et al., 2017). Diop-Frimpong et al. (2011) found that losartan reduced the generation of type I collagen through CAFs and improved the penetration and therapeutic effect of nanoparticle drugs in tumors. The hexosamine biosynthesis pathway (HBP) is a glycolytic shunt pathway and a metabolic node of cancer cells that can promote survival pathways and influence hyaluronan synthesis in the ECM. Sharma et al. (2020) inhibited this pathway by a small molecule glutamine analog (6-diazo-5-oxo-l-norleucine (DON)). The results showed that DON reduced hyaluronic acid and collagen, leading to extensive remodeling of the ECM, increased CD8⁺ T infiltration, and enhanced sensitivity to PD-1, which ultimately reduced tumor growth and metastasis and prolonged survival. These studies suggest that tumor cells can reprogram the ECM to change the tumor-CAF interaction to promote tumor growth, and the mechanisms by which CAFs reverse ECM remodeling and inhibit tumor development should be explored.

Targeted therapies for CAFs in pancreatic cancer include focal adhesion kinase (FAK), neuregulin 1 (NRG1), LRRC15⁺ CAFs, etc., The FAK inhibitor suppressed tumor cells and CAFs in a mouse model (Stokes et al., 2011; Kanteti et al., 2018). Jiang et al. (2016) found that the FAK inhibitor VS-4718 was able to respond to immunotherapy and PD1 antagonists by overcoming immunosuppression and fibrosis levels. Inhibition of FAK could reduce ECM remodeling and increase sensitivity to chemotherapy and immunotherapy (NCT02428270, NCT02546531, NCT02651727, and NCT02758587) (Zaghdoudi et al., 2020). NRG1 is secreted by tumors and CAFs, and Ogier et al. (2018) demonstrated that 7E3 (NRG1 antibody) inhibited tumor growth and migration when cocultured with CAFs. LRRC15⁺ CAFs were associated with poor response to anti-PD-L1 treatment (Dominguez et al., 2020). The LRRC15^–targeted antibody-drug conjugate ABBV-085 demonstrated robust preclinical efficacy (Purcell et al., 2018). Meflin⁺ CAFs inhibited differentiation toward myCAFs and suppressed pancreatic carcinogenesis (Mizutani et al., 2019). Netrin G1 (NetG1)⁺ CAFs supported PDAC survival, and blocking NetG1 inhibited tumor formation in vivo (Francescone et al., 2020). These recent studies show that CAFs communicate with pancreatic tumor cells in various ways. At present, research on CAFs has changed from the complete inhibition of CAFs to the intervention of a subpopulation of CAFs, and some initial results have been achieved (Vennin et al., 2019; Table 1).