The Role of Cancer-Associated Fibroblast as a Dynamic Player in Mediating Cancer Stemness in the Tumor Microenvironment

The enrichment of cancer-associated fibroblast (CAFs) in a tumor microenvironment (TME) cultivates a pro-tumorigenic niche via aberrant paracrine signaling and matrix remodeling. A favorable niche is critical to the maintenance of cancer stem cells (CSCs), a population of cells that are characterized by their enhanced ability to self-renew, metastasis, and develop therapy resistance. Mounting evidence illustrates the interplay between CAF and cancer cells expedites malignant progression. Therefore, targeting the key cellular components and factors in the niche may promote a more efficacious treatment. In this study, we discuss how CAF orchestrates a niche that enhances CSC features and the potential therapeutic implication.


INTRODUCTION
Tumors are illustrated as "wounds that do not heal" due to the enrichment of fibroblasts and immune cells at the tumor site, which highly mimics that of an inflammatory response of a nonneoplastic tissue (Dvorak, 1986). Indeed, cancer-associated fibroblast (CAF) constitutes the bulk of the stromal component in solid tumors. Cancer cells hijack the normal physiological function of activated fibroblast functions during wound recovery to fuel malignant development. Given cancer stemness accelerates disease progression and impinges therapy, understanding how CAFs cultivate a niche that bestows cancer stem cell (CSC) features in cancer cells may facilitate novel therapeutic intervention.

CANCER-ASSOCIATED FIBROBLAST AS A DYNAMIC PLAYER IN PROMOTING CANCER STEMNESS
Cancer-associated fibroblast, also known as tumor-associated fibroblast, is generally regarded as fibroblasts that exist within or enveloping the tumor (Öhlund et al., 2014). The presence of CAFs throughout tumor development, from the incipient stage to advanced stages, alludes to CAF participation in tumor initiation and progression (Erez et al., 2010). Compared with non-tumor fibroblast (NF) or peri-tumor fibroblast, CAFs are located more proximal to the neoplastic region and exhibit a greater activation marker, namely actin smooth muscle (αSMA) and fibroblast activation protein (FAP) (Qin et al., 2016(Qin et al., , 2019. However, compiling studies indicate CAF displays vast heterogeneity within a tumor with distinct functions, underscoring the importance of identifying the function of how each subtype contributes to malignancy (Öhlund et al., 2017;Costa et al., 2018;Lambrechts et al., 2018). A myriad of elements may affect CAF subtypes, for instance, proximity to the tumor. CAF subsets were distinguished based on their expression of αSMA and IL6 with high-αSMA localized nearer to the tumor and high-IL6 further, indicating juxtacrine and paracrine interaction between cancer cells and fibroblast may stimulate the CAF to differentiate to subtypes with distinct functions (Öhlund et al., 2017). Aside from molecular markers, the secretome of CAF and NF can be discerned as the former secretes a greater abundance of ligands that promote stemness properties of HCC Álvarez-Teijeiro et al., 2018). CAF plays an extensive role in shaping the extracellular matrix (ECM) of the tissues, allowing it to exert its oncogenic influence by modulating the physical properties of the TME. In addition to promoting cancer progression, fibroblasts are shown to be educated by neoplastic cells to their benefits, depicting the dynamic cellular network in the TME (Figure 1).
Cancer stem cells are generally deemed as a rare population of cancer cells that share similar features to normal stem cells, including the ability to self-renew and differentiate into lineages that comprise the tumor bulk, thus regarded as tumorigenic, as opposed to non-CSC cancer cells (Reya et al., 2001;Clarke et al., 2006). Common CSC markers include surface markers, such as CD24, CD44, CD90, CD133, EpCAM, LGR5, and aldehyde dehydrogenase (ALDH) ALDH activity (summarized in Walcher et al., 2020). Moreover, expressions of stemness genes, such as OCT4, SOX2, and NANOG, are often deployed to measure cancer stemness. Functional surrogate to assess cancer stemness properties includes in vitro sphere formation and in vivo limiting dilution in which cancer cells are injected at low doses into animal models. Cancer cells with greater stemness are generally considered to have augmented metastasis ability due to their enhanced ability to reconstitute a tumor at the secondary site, and the acquisition of EMT is coupled with elevated stem cell properties, including stem cell markers and tumorigenicity, signifying EMT mimics features of CSC (Mani et al., 2008;Baccelli and Trumpp, 2012;Oskarsson et al., 2014). Emerging evidence illustrates that stemness properties of cancer cells are dependent on their niche (Plaks et al., 2015;Batlle and Clevers, 2017). Factors derived from niche may promote plasticity of non-CSC cancer cells into CSC, whereas depletion of factors may reduce the CSC population, thereby implicating niche factors as potential therapeutic targets (Batlle and Clevers, 2017). Given the CAF fosters the niche by enriching pro-tumor factors and remodeling the matrix of the TME, understanding how a niche facilitates cancer initiation and progression will allow us to contrive therapy targeting the key cellular components and factors sustaining cancer stemness.

Cancer Cells Educate Cancer-Associated Fibroblast to Adopt a Pro-tumor Phenotype
Accumulating research illuminates the cancer cells instigate protumor CAF to foster a pro-favorable niche. Transforming growth factor-β3 (TGF-β3) derived from HNSCC cells can activate CAF to secrete POSTN, leading to greater metastasis ability of the neoplastic cells (Qin et al., 2016). Prostate cancer cells can recruit marrow-derived mesenchymal stem cells (MSCs) and activate them into CAF by TGF-β1 (Barcellos-de-Souza et al., 2016). Of significance, MSC-derived CAF can recruit and induce monocytes into M2 macrophages, illustrating the CAF capacity to govern the TME. Cardiotrophin-like cytokine factor (CLCF1) derived from CAF stimulates the production of C-X-C motif ligand 6 (CXCL6) and TGF-β, which not only escalate stemness properties of cancer cells but also stimulate CLCF1 expression in the CAF, thereby fostering a positive feedback loop that expedites malignancy . Furthermore, CXCL6 and TGF-β can enhance infiltration and development of pro-tumor N2neutrophils , corroborating CAF capacity to foster an immunosuppressive TME. Sonic hedgehog (SHH) ligand derived from CD24 + CD49f hi breast CSCs constructs a pro-tumorigenic TME by activating CAF (Valenti et al., 2017). Cancer cells-derived hedgehog (Hh) instigates the production of pro-tumor paracrine factors and ECM constituents in CAF (Cazet et al., 2018). Autocrine and colorectal cancer cells-derived paracrine signaling of IL34 facilitate the conversion of normal fibroblast into CAF that adopts a pro-tumor secretome that includes stemness-promoting factors, such as Netrin-1 and FGF2 protein (Giulianelli et al., 2019;Sung et al., 2019). The coculture of HNSCC cells and CAF stimulate CAF to generate WNT3A, leading to increased cancer stemness (Le et al., 2019).
Exosomal miRNAs are gaining interest as a mediator between cancer cell-fibroblast crosstalk. Activation of focal adhesion kinase (FAK) signaling by cancer cell-derived miRNA stimulates the production of CAF-derived exosomes that facilitate the spheroid formation and metastasis . Tumorderived exosomal miR-1247-3p triggers the activation of fibroblast by fueling NF-κB signaling consequently stimulates the CAF to secrete a greater abundance of IL6 and IL8, instigating the cancer cells to an EMT phenotype with enhanced lung metastases (Fang et al., 2018). Exosomal miR-9-5p elevates IL6 production in CAF, leading to enhanced spheroid-forming ability . In summary, targeting the aforementioned factors or exosomes may disrupt the ability of cancer cells to assemble a pro-tumor niche.

Paracrine Signaling of Cancer-Associated Fibroblast Enhances Cancer Stem Cell Features
In a tumor, fibroblast-secreted factors and cytokines, which serve for tissue recovery, are seized by cancer cells. Compiling studies have depicted the critical roles of CAF-derived factors in maintaining the CSC features of cancer cells by promoting stemness pathways (Table 1). HGF is reported to enhance cancer stemness by potentiating the frequency of CD44 + , CD47 + , and CD90 + HCC CSC and spheroid formation, stemness and EMT gene expressions (Lau et al., 2016;Ding et al., 2018;Yan et al., 2018). Mechanistically, HGF exerts its oncogenic influence via distinct mechanisms in a tissue-dependent manner. HGF augments ERK/FRA1/HEY1, STAT3/TWIST1, and YAP/HIF-1α in HCC, gastric cancer, and pancreatic cancer, respectively (Lau et al., 2016;Ding et al., 2018;Yan et al., 2018). CAF-derived IL6 is demonstrated to fuel stemness as evident in increased spheroid formation, stemness genes markers via STAT3 signaling (Xiong et al., 2018;Zhang et al., 2020). CAF paracrine signaling regulates the AKT pathway via insulin-like growth factor 2 (IGFII), thereby promoting NANOG expression, enhanced tumorigenicity in vivo, and ALDH activity . The AKT signaling is also mediated by cartilage oligomeric matrix protein (COMP) to fuel EMT and spheroid formation . CCchemokine ligand 2, 5, 7, and 16 (CCL2, CCL5, CCL7, and CXCL16) potentiate EMT via the Hh and TGF-β signaling (Liu et al., 2016). Serglycin (SRGN), a proteoglycan secreted from cancer cells and CAFs, renders chemoresistance and EMT features via a CD44-dependent NF-κB activation (Guo et al., 2017). The binding of CAF-derived periostin (POSTN) to protein tyrosine kinase 7 (PTK7) drives the β-catenin signaling, leading to increased expression of CSC markers and stemness genes . Annexin A1 (AnxA1) secreted from CAF potentiates stemness and EMT genes expressions, and spheroid formation (Geary et al., 2014). Prostaglandin E 2 (PGE 2 ) produced from fibroblasts drives the expansion of tumor-initiating cells by inducing YAP signaling (Roulis et al., 2020). The cellular crosstalk within the TME was displayed as CAFs recruit tumor-associated macrophages (TAMs) via CXCL12 secretion and induces the M2phenotype TAMs, subsequently promotes stemness and EMT signature genes . Alteration in specific genes may dictate the impact of CAF on cancer cells. Androgen receptor (AR)-depleted CAF enhances the production of IFNγ and M-CSF, leading to increased cancer stemness (Liao et al., 2017). Notch1-depleted CAFs augment CD271 + melanoma CSC frequency (Du et al., 2019) and metastasis to the lung (Shao et al., 2016). Together, blocking CAF-derived factors or a downstream signaling cascade may be a therapeutic revenue to diminish cancer stemness.

Cancer-Associated Fibroblast Mediates Cancer Stemness via Extracellular Matrix Remodeling
A fundamental function of activated fibroblast is its participation in ECM homeostasis by synthesizing ECM constituents, including collagen and fibronectin, and producing ECMdegrading proteases to aid communication and trafficking of inflammatory cells (Kalluri, 2016). While the impact of matrix stiffness remains controversial as studies indicate both softer matrix (Tan et al., 2014;Ng et al., 2021) and stiffer matrix (Pang et al., 2016;Tao et al., 2021) can instigate cancer stemness, the role of CAF in mediating the stiffness of the ECM illustrates its ability to affect stemness. Indeed, collagen deposited by CAF functions as a mechano signal to escalate stem cell markers and spheroid formation ability (Cazet et al., 2018). Additionally, hyaluronan (HA), a major constituent of the ECM that boosts CSC self-renewal and EMT (summarized in Chanmee et al., 2015), is shown to be highly produced by CAF (Affo et al., 2021). Given CAFs are a source of ECM-degrading proteases, namely matrix metalloproteinases (MMPs), they can affect CSC properties by remodeling the ECM. Studies demonstrated IL6 derived from prostate cancer cells elicits pro-tumor FAP + CAF (Giannoni et al., 2010). Consequently, these FAP + CAF secretes MMP2 and MMP9 to potentiate EMT and stemness genes. MMP3 that is predominantly produced by fibroblast (Witty et al., 1995) is demonstrated to promote stem cell population by enhancing canonical Wnt signaling, potentially leading to hyperplastic growth of normal tissues (Kessenbrock et al., 2013), depicting fibroblast can contribute to tumor initiation via ECM remodeling. Future research can focus on ablating the effect of CAF on the ECM, either by targeting ECM constitutes or MMPs, to attenuate the CAF-mediated oncogenic effect.
Another mechanism that CAF mediates chemoresistance resides in its ability to remodel the ECM. Studies have demonstrated HA upregulates NANOG/STAT signaling, leading to increased expression of multidrug transporter (MDR1) that contributes to chemoresistance (Bourguignon et al., 2008). In accordance, HA-CD44 interaction mediates stemness signaling that governs miRNA regulation of genes involved in chemoresistance in breast cancer cells (Bourguignon et al., 2009) and HNSCC cells (Bourguignon et al., 2012a,b). HA mediates CD44v3 high ALDH1 high HNSCC CSC via an epigenetic alteration to promote cisplatin chemoresistance (Bourguignon et al., 2016). The ability of CAF to mediate matrix stiffness via MMPs may be an area of interest. Given matrix stiffness regulates stemness and chemoresistance (Ng et al., 2021); how a CAF-mediated ECM fosters a niche that endows chemoresistance warrants further investigations.

Single-Cell RNA Sequencing Identified Cancer-Associated Fibroblast Subsets With Distinct Functions on Cancer Stemness
State-of-the-art single-cell transcriptomic profiling of intrahepatic cholangiocarcinoma (ICC) revealed six distinct fibroblast populations . Vascular CAFs (vCAF), the most abundant population, were found to secrete IL6 to promote self-renewal of cancer cells. Of note, the characterization of the function of other CAF subtypes identified such as matrix CAFs, which expressed high ECM signature genes, would be an interesting area of research. Lineage tracing coupled with single-cell RNA sequencing (scRNA-seq) in sophisticated animal models categorized CAF into myofibroblast (myCAF) and inflammatory and growth factor-enriched CAF (iCAF) populations that modulate tumor cell proliferation via distinct mechanisms; the former has a high level of hyaluronan synthase 2 (Has2) that can enhance pro-tumor HA; the latter secretes HGF (Affo et al., 2021). scRNA-seq identified prostaglandinendoperoxide synthase 2 (Ptgs2)-expressing fibroblasts that expand tumor-initiating stem cells via YAP signaling, whereas ablation of Ptgs2 diminished the occurrence of the small intestine and colon cancers (Roulis et al., 2020). Capturing cellular diversity at single-cell resolution ignites exciting research questions as follows: how cellular and non-cellular components of the TME govern CAF identity? How each subtype contributes to disease progression? How CAFs differentiate or undifferentiate into different subtypes?

POTENTIAL THERAPEUTIC STRATEGY AND CHALLENGES
Targeting Multiple Cancer-Associated Fibroblast-Derived Factors to Circumvent Stemness As aforementioned (Table 1), cancer-promoting bioactive molecules derived from stromal cells may be targeted using specific neutralizing antibodies, particularly in combination with chemotherapy. However, the promising preclinical results have yet to yield optimistic results in clinical settings. A Phase 2 trial (NCT00433446) examining siltuximab (CNTO328), an anti-IL6 antibody, showed siltuximab did not have clinical benefit for patients with advanced prostate cancer (Dorff et al., 2010). Similarly, the addition of siltuximab (CNTO328) to mitoxantrone/prednisone regimen did not yield improved clinical (NCT00385827) outcomes compared to mitoxantrone/prednisone alone in treating advanced prostate cancer (Fizazi et al., 2012). The Phase 2, single-arm, clinical trial (NCT00992186) explored the efficacy of carlumab (CNTO 888), antibody targeting (CCL2), in patients with metastatic castration-resistant prostate cancer (CRPC), who had undergone docetaxel treatment (Pienta et al., 2013). However, none of the patients treated with carlumab showed partial or complete remission. One key possibility is that, while the CCL2 level was depleted upon carlumab administration, the level rebounded rapidly within a week, thus suggesting carlumab cannot suppress CCL2 for the clinically meaningful duration (Pienta et al., 2013). Nonetheless, the addition of siltuximab to bortezomib-melphalan-prednisone (VMP) demonstrated marginal clinical benefits in multiple myeloma by statistically improving partial response rate (San-Miguel et al., 2014). Together, these imply more precise stratification of patients and usage of biomarkers may result in better clinical outcomes. Of note, several clinical trials are exploring the clinical utility of neutralizing antibodies that target HGF (NCT04368507), MET (NCT04077099), and IL6R (NCT03999749). Aside from neutralizing antibodies, a combination of inhibitors-targeting receptors, including Osimertinib (an EGFR inhibitor) and Savolitinib (MET inhibitors), are being tested in clinical trials (NCT03778229). Notably, MP0250, a drug candidate targeting HGF and VEGF, which have met the safety requirement and demonstrated clinical efficacy, is being tested in combination with existing drugs to treat multiple myeloma (NCT03136653). Given the TME is often accompanied by an upsurge of cytokines and chemokines, targeting multiple factors may promote treatment efficacy.

Depleting Pro-tumor Cancer-Associated Fibroblast
The ability of CAF to endow stemness of cancer cells makes it an intriguing therapy. CAF is shown to be more positively correlated with gene sets associated with poor prognosis compared with epithelial cancer cells, immune cells, and endothelial cells (Calon et al., 2015), substantiating targeting CAF as a potential therapeutic avenue. Studies have demonstrated that αSMA expression represents a marker for worse prognosis in colorectal cancer, indicating myofibroblast abundance is crucial to disease prediction (Tsujino et al., 2007). In accordance, studies conducted in tongue cancer and oral cancer depicted similar results (Vered et al., 2010;Marsh et al., 2011). However, the heterogeneity of CAF necessitates precise identification of more specific markers as tumor-restraining CAF exists. Genetic deletion of αSMA in mouse models enhances the progression of pancreatic ductal adenocarcinoma (PDAC) as evidenced by a lower survival rate in αSMA-depleted mice (Ozdemir et al., 2014). Interestingly, the myofibroblast-depleted tumor demonstrated enhanced spheroid-forming ability, indicating a greater proportion of CSC. Indeed, identification of CAF subtypes shows they promote or suppress tumor progression in a tissue-dependent manner (Galbo et al., 2021), corroborating the need for further investigation into more CAF-specific markers. Recent work has illuminated the potential therapeutic benefit of targeting tumor-promoting fibroblasts. Administration of GRP77 neutralizing antibody-targeting CAF depletes CAF-secreted IL6 and IL8, thereby abolishing a stem cell-supporting niche and sensitizing breast cancer cells to doxycycline, leading to significant shrinkage of the tumor volume (Su et al., 2018), highlighting the therapeutic potential of targeting precise tumor-promoting CAF.

Rewire Pro-tumor Cancer-Associated Fibroblast Into Quiescent or Anti-tumor Fibroblast
Reeducating pro-tumor CAF into a quiescent state or even antitumor CAF is a tempting strategy. Vitamin D metabolite 1α, 25-dihydroxyvitamin D3 [1,25(OH)2D3] is shown to deplete the oncogenic influence of stromal fibroblast to the cancer cells, while 1,25(OH)2D3-treated stromal fibroblast displayed a gene signature that favors a clinical outcome (Ferrer-Mayorga et al., 2017). VDR activation of the stromal fibroblast using calcipotriol, a vitamin D analog, diminished expression of genes involved in growth factors and cytokines, for instance, the IL6 and POSTN, thereby suppressing a tumor-promoting secretome (Sherman et al., 2014). Additionally, the combination treatment of calcipotriol and gemcitabine markedly prolonged survival in preclinical models. Moreover, high vitamin D receptor (VDR) expression in stromal cells predicts favorable survival (Ferrer-Mayorga et al., 2017). All-trans retinoic acid (ATRA), also known as tretinoin, a vitamin A metabolite, renders pancreatic stellate cells into a quiescent state, which, in turn, secreted a greater level of secreted frizzled-related protein 4 (sFRP4) that negatively modulates Wnt signaling of cancer cells in a paracrine manner (Froeling et al., 2011). Indeed, ATRA has passed the safety in Phase I clinical trial (NCT03307148) and will proceed to Phase II (NCT04241276) as encouraging therapeutic responses were observed (Kocher et al., 2020). Given vitamins are essential for healthy tissue and their toxicity is relatively lower compared to chemotherapy, repurposing vitamin analogs to rewire activated fibroblast into a quiescent state may present a viable therapeutic strategy that can be translated into the clinical settings.
Another strategy to rewire the population of fibroblasts is by reprogramming the fibroblast using growth factors as exemplified in the plasticity of CAF found in PDAC. While IL1 induces fibroblast into having an inflammatory phenotype categorized by an elevated cytokines production, TGF-β antagonizes an IL1induced phenotype and stimulates the fibroblast to adopt a myofibroblastic phenotype with less tumorigenesis, particularly reduced expression of factors promoting cancer stemness such as Il6 and Cxcl12 (Biffi et al., 2019). Together, this rationalizes the option to rewire tumor-promoting into tumorrestraining fibroblast. To account for the ability of the cancer cell to instigate tumor-promoting CAFs, inhibitors may be deployed to circumvent cancer cell-mediated activation of CAF. Breast cancer cells activate CAF via the hedgehog signaling to cultivate a stem cell niche by ECM remodeling and FGF5 secretion that promote docetaxel chemoresistance (Cazet et al., 2018). As such, targeting the crosstalk via smoothened inhibitors (SMOi), a hedgehog-signaling inhibitor, sensitizes triple-negative breast cancer (TNBC) to docetaxel. Concomitantly, inhibition of hedgehogs abrogates the activated stromal cells, thereby augmenting the efficacy of chemotherapy in treating pancreatic cancer (Olive et al., 2009). However, either genetic depletion or pharmacological inhibition of SHH, a ligand that activates pancreatic CAFs, resulted in less stromal composition but also a more aggressive tumor (Rhim et al., 2014), suggesting treatment should be tailored based on tissue and treatment.

FUTURE PERSPECTIVES
Advances in single-cell RNA sequencing may not only aid the characterization of cell types based on their molecular profiles and, subsequently, functions but also have the potential to be used to identify new biomarkers for patient stratification and tailorpersonalized medicine (Dominguez et al., 2020). Harnessing single-cell RNA-sequencing to profile CAF at the molecular level, future studies can address outstanding questions, including the origin and development of specific CAF subtypes, identification of biomarkers corresponding to each subtype, how each CAF subtype interacts with its niche, and amenable therapeutic opportunities to tackle tumor-promoting CAF. A more holistic approach investigating the CAF molecular profile, for instance, epitranscriptomic and epigenetics (Delaunay and Frye, 2019;Song et al., 2020), may unravel novel insights into how CAF shuttles between cell types. Recent studies have unmasked that the altered epigenetics profile between CAF and NF allows the former to generate greater levels of WNT5A that confers malignancy to the neoplastic cells (Maeda et al., 2020). A critical unmet knowledge gap in our understanding of CAF function is if juxtracrine signaling between CAF and cancer cells affects cancer stemness.
Given the TME is a dynamic region with various cell types actively contributing to tumor progression, merely focusing on targeting a single aspect seems unlikely to yield any longterm therapeutic benefit. For example, targeting niche factors alone may not be sufficient to eradicate the tumor due to the inclination of cancer cells to evolve into a niche-independent malignancy as the disease progresses (Fujii et al., 2016). Therefore, comprehensive characterization of cell types and their respective functions in the TME may pave the way for a multimodal approach to improve cancer treatment.

AUTHOR CONTRIBUTIONS
JL wrote the review article. SM edited the review article and provided funding support. Both authors contributed to the article and approved the submitted version.