Biomarkers for cancer-associated fibroblasts

Cancer-associated fibroblasts (CAFs) are the key component of tumor stromal. High heterogeneity of CAFs reflects in their origin, phenotype and function. Biological function which can be suggested by biomarkers of distinct CAF subgroups may be different, even opposite, just like water and fire. Identifying CAF subpopulations expressing different biomarkers and reconciling the relationship of the “water and fire” among distinct CAF subsets may be a breakthrough in tumor therapy. Herein, we briefly summarize the biomarkers commonly used or newly identified for distinct CAFs in terms of their features and potential clinical benefits.

passage number of cultured CAFs between 1 to 6 is suitable for experiments.
Neutral biomarkers with dual functions α-SMA: a snapshot of CAF heterogeneity As the marker of activated fibroblasts, α-SMA, a cytoskeletal protein associated with TGF-β production and highly contractile phenotype, is the first identified and most frequent CAF biomarker [8,9]. Studies have verified that α-SMA + CAFs could promote tumor progression, confer therapeutic resistance and mediate immunosuppressive TME in multiple ways, such as paracrine and ECM remodeling [2,10,11].
However, depleting α-SMA + CAFs directly has been failed to treat pancreatic ductal adenocarcinoma (PDAC) in murine models, probably due to the immune surveillance suppression [4]. Patients with high desmoplasia defined by expression of collagen I and CAF markers including α-SMA have better prognosis in PDAC, breast and lung cancer patients [12,13]. These results show that the fibrotic response of α-SMA + CAFs may be a host protection mechanism against tumor progression, which needs further investigation. The contradictory conclusions above indicate the α-SMA + CAF heterogeneity which has been confirmed in a recent study classifying different subtypes of α-SMA + CAFs. Typically, inflammatory CAFs (iCAFs) being α-SMA low IL-6 high are identified in PDAC and located more distantly from neoplastic cells than α-SMA high IL-6 low myofibroblastic CAFs (myCAFs) (Fig. 1a). ICAFs appear to be more tumor-promoting than myCAFs by producing chemokines and cytokines [14] and indicate a higher malignancy in pancreatic tumorigenesis [15]. On the other hand, myCAFs may deposit ECM extensively to impede drug delivery despite less cancer-stimulating [16]. Considering the biological effects of myCAFs and iCAFs, the composition of myCAFs and iCAFs in the TME may have different implications for treatment. High iCAF content may indicate a higher degree of malignancy, while high myCAF content may indicate a poor response to treatment.
Accordingly, combination of reducing the upstream formation of iCAFs with inhibiting the downstream desmoplasia derived from myCAFs may design the treatment strategy. Furthermore, IL-1/JAK/STAT signaling cascades and TGF-β have been found involved in the formation and mutual transformation of these two CAF subtypes. JAK inhibitors can suppress tumor growth as well as shift iCAFs to myCAFs while TGFBR inhibition could partially attenuate the function of myCAFs without influencing tumor growth [16]. Patients might benefit from combination therapy of these two drugs. Altogether, classification of α-SMA + CAFs deepens our understanding of CAF heterogeneity as well as provides new ideas for CAF targeted therapy which remains further study. Fig. 1 The "double face" of cancer-associated fibroblasts (CAFs). CAF subsets identified by different biomarkers play distinct roles in tumor microenvironment. a, several studies have shown that CAFs expressing α-SMA or S100A4, which are considered tumorigenic previously, are also potentially anti-tumor. Identification of myCAFs and iCAFs among α-SMA + CAFs may represent an aspect of CAF heterogeneity, which is believed that two subsets can convert into each other. b, most biomarkers represent tumor-promoting CAFs including traditional CAF biomarkers (FAP, PDGFR, Vimentin, PDPN and CD70) as well as some newly identified markers (CD49e, CD10/GRP77 and MHCII/CD74). c, CD146 + CAFs, CAV1 high CAFs and PDGFRα + Saa3 − have been identified as tumor-suppressive CAF subsets Bidirectional roles of S100A4 S100A4, also called fibroblast-specific protein-1, is usually expressed on CAFs transited from epithelial or endothelial cells [17,18]. However, the biological effects of S100A4 + CAFs are controversial (Fig. 1a). S100A4 + CAFs promote tumor metastasis through secretion of VEGF-A and Tenascin-C [19]. Chemokine ligand 2 (CCL2) derived from S100A4 + CAFs contributes to immune evasion by maintaining macrophage infiltration [20]. On the other hand, fibroblasts expressing S100A4 can product collagen and encapsulate carcinogens to enhance immune surveillance ability [21]. α-SMA + S100A4 + CAFs can activate tumor immune response by promoting CD8 + T cell activation through fusion with dendritic cells [20]. These findings confirmed the CAF heterogeneity, and S100A4 + CAFs remain further characterization.
Accomplices: pro-tumorigenesis biomarkers FAP: a promising therapeutic target FAP is another wide-spread biomarker for CAFs, as well as a serine protease participating in ECM remodeling and fibrogenesis, thereby accelerating tumor progression [22]. FAP + CAFs can shape the immunosuppressive TME by secreting distinct chemokines and cytokines [2]. A recent study in ovarian cancer found that FAP high CAFs are correlated with poor patient outcomes [23].
However, sibrotuzumab, a FAP-specific antibody, which has been found clinically safe and effective in a phase I trial of advanced cancers, showed no beneficial effect in a phase II trial of metastatic colorectal cancer [35][36][37]. Another phase II trial utilizing PT-100 in metastatic colorectal cancer also failed [38]. It is noted that the patients involved in both of these phase II trials were heavily pre-treated and represented a refractory patient population, which may account for the failure. FAP may probably contribute more in the earlier-stage tumors rather than late-stage metastasis [39]. Nevertheless, FAP is still the most promising CAF therapeutic target and more exploration is needed.

Other traditional biomarkers: PDGFR, Vimentin, PDPN and CD70
PDGFRα/β are both upregulated in multiple tumors [40], and PDGFRβ is correlated with poor survival in breast cancer [41]. PDGFRα/β + CAFs participate in immunomodulation by inducing macrophage migration and M2 polarization [2]. Blocking PDGFR signaling can suppress angiogenesis and tumor growth in human cervical cancer [42]. Furthermore, a PDGFR inhibitor, Dasatinib can partially reverse the pro-tumorigenic CAFs of lung adenocarcinoma (LUAD) to a quiescent state [43], which might be a potential treatment strategy for LUAD.
Vimentin is a biomarker for epithelial-to-mesenchymal transition (EMT) maintaining structure and motility during cell migration [44], involved in CAF motility to lead cell invasion in LUAD [45]. Podoplanin (PDPN) + CAFs in LUAD have been reported as the inducer of immunosuppressive microenvironment [46]. Moreover, CAFs expressing CD70 enrich regulatory T cells to invasive colorectal cancer and CD70 expression is negatively correlated with survival of patients with colorectal cancer (Fig. 1b) [47]. PDPN seems to be a potential therapeutic target while CD70 and vimentin might be more suitable for prognostic indicators and more evidences are warranted.

Defenders: tumor-suppressive biomarkers
The potential anti-tumor function of CAFs exhibits when we attempt to inhibit hedgehog pathway which has been proved to activate fibroblasts [48][49][50], indicating the existence of tumor-suppressive CAF subsets. Studies on CAF heterogeneity has identified several potential anti-tumor CAF subpopulations and biomarkers (Fig. 1c).
CD146 expression is found to distinguish at least two CAF subpopulations in ER + breast cancer, among which CD146 + CAFs could promote tamoxifen sensitivity by continuously expressing ER, whereas CD146 − CAFs opposite [51]. Another stroma-derived gene expression signature of breast tumor shows that CAV1 low CAFs are correlated with poor prognosis [52]. Serum amyloid A3 (Saa3) is also identified as new biomarker of mouse PDGFRα + CAFs in PDAC, as PDGFRα + Saa3 + CAFs could facilitate tumor growth while PDGFRα + Saa3 − CAFs impairing tumor proliferation [53]. In summary, increasing proportion of CD146 + CAFs in ER + breast cancer and targeting Saa3 specifically in PDAC might be promising strategies to revert the TME to an anti-tumor environment. So far it might be too immaturely to use the defensive CAFs as a cellular therapeutic strategy, however at least these biomarkers might be benefit for prognostic diagnosis.

Newly identified CAF subpopulations and biomarkers
Recent advances using single-cell RNA sequencing (scRNA-seq) provide us technical advantages to better understand CAF heterogeneity and identify novel biomarkers. For example, CD49e has been identified as a new cell surface pan-CAF biomarker in ovarian cancer recently, just like α-SMA [23]. Herein, we list the CAF subpopulations and biomarkers newly identified in distinct tumor tissues (Table 2).
GPR77 and CD10 are potential targeted biomarkers as the infiltration of CD10 + GPR77 + CAFs indicates chemotherapy resistance and poor survival, especially in the ER − HER2 − subtype and high-grade breast tumors. Blocking GPR77 substantially can suppress tumorigenesis along with enhancing chemosensitivity in a patientderived xenograft model [54].
The existence of iCAFs and myCAFs has been verified in triple-negative breast cancer (TNBC) and bladder urothelial carcinoma by scRNA-seq, despite the biomarkers identified are different. For specific, CXchemokine ligand 12 (CXCL12) is the biomarker of TNBC iCAFs [55]. RGS5 and PDGFR are the biomarkers for myCAFs and iCAFs in bladder tumor, respectively [56].
Major histocompatibility complex (MHC) class II family genes and CD74 have been identified as biomarkers of another PDAC CAF subpopulation termed as antigen presenting CAFs (apCAFs) besides myCAFs and iCAFs. ApCAFs process an immunomodulatory role by interacting with CD4 + T cells [57]. All of these three CAF subpopulations with another three are verified in TNBC. It is noted that apCAFs are also found in normal tissues, indicating that the phenotype is not TMEinduced. Furthermore, PDGFRα is found highly expressed in iCAFs while PDGFRβ is found highly expressed in myCAFs [58].
As we summarized above, increasing CAF subsets with distinct biomarkers have been identified to coexist in tumor tissues and play different biological functions. However, due to the tissue heterogeneity, distinct classification criteria, biomarkers and nomenclature selected by different laboratories, the identification of CAF subsets is somewhat messy and intersecting at present. Among the CAF subsets identified, myCAFs and iCAFs both seem prevalent across-organ despite the biomarkers identified are different, probably due to the tissue heterogeneity. Furthermore, iCAFs have been widely proved to promote tumor progression by secreting chemokines and cytokines such as IL-6 and CXCL12. On the other hand, several CAF subgroups identified from different tissues have been named the same by different studies, such as vCAFs and mCAFs. However, given their different biological functions and biomarkers, whether they are the same subgroup is worth further discussion. We believe that with the further research on the CAF heterogeneity, there will be a unified standard for the selection of CAF biomarkers and nomenclature of CAF subpopulations.
With the further development of the technology, we believe that our understanding of the CAF subsets and biomarkers will be more profound. CAF biomarkers can be utilized as both prognostic indicators and therapeutic targets for clinical benefits. Hopefully, we can reverse the accomplice CAFs to defender ones by targeting appropriate molecules.