How Do Cancer Cells Create Cancer‐Associated Fibroblast Subtypes? Impacts of Extracellular Vesicles on Stromal Diversity

ABSTRACT

Cancer‐associated fibroblasts (CAFs) are the major component of the tumor stroma. They mediate various attributes of tumor cells, such as cell growth, migration, invasion, angiogenesis, metabolic reprogramming, apoptosis, immune regulation, and extracellular matrix reconstitution, all related to cancer progression and treatment resistance. Although many researchers have recognized CAF heterogeneity, recent technological advances have emphasized the functional and phenotypic diversity of CAFs in cancer progression. Why are these CAF subtypes generated within tumor tissues? And how do cancer cells dictate such heterogeneous subtypes of CAFs? This review will highlight the CAF subtypes within the tumor microenvironment and their role in tumor progression. CAF subtype induction by extracellular vesicles (EVs) and their significance, which we reported previously, is also discussed.

Cancer‐associated fibroblasts (CAFs) are the major component of the tumor stroma. This review will highlight the CAF subtypes within the tumor microenvironment and their role in tumor progression. CAF subtype induction by extracellular vesicles (EVs) and their significance, which we reported previously, are also discussed.

Abbreviations

ANTXR1

Anthrax Toxin Receptor 1

apCAFs

antigen‐presenting CAFs

BM‐MSCs

bone marrow‐derived mesenchymal stem cells

CAFs

cancer‐associated fibroblasts

CXCL12

CXC‐chemokine ligand 12

DPP4

dipeptidyl peptidase 4

Dpt

dermatopontin

ECM

extracellular matrix

EMT

epithelial‐mesenchymal transition

EVs

extracellular vesicles

FAP

fibroblast activation protein

FSP1

fibroblast‐specific protein‐1

GC

gastric cancer

HGF

hepatocyte growth factor

Hh

Hedgehog

Hmgb1

high‐mobility group box 1

iCAFs

inflammatory CAFs

ICC

intrahepatic cholangiocarcinoma

ICI

immune checkpoint inhibitor

IL

interleukin

IL1R1

IL‐1 receptors

Irf1

interferon regulatory factor 1

ITGA11

integrin subunit alpha 11

LIF

leukemia inhibitory factor

LRRC15

leucine‐rich repeat containing 15

MET

mesenchymal‐epithelial transition

miRNAs

microRNAs

MMPs

metalloproteinases

MVBs

multivesicular bodies

myCAFs

myofibroblastic CAFs

NG2

neuron‐glial antigen 2

PDAC

pancreatic ductal adenocarcinoma

PDGFR

platelet‐derived growth factor receptor

PD‐L1

programmed death ligand 1

PDPN

podoplanin

Pi16

peptidase inhibitor 16

Shh

Sonic hedgehog

TGFBR2

transforming growth factor‐beta receptor 2

TGF‐β

transforming growth factor‐beta

TME

tumor microenvironment

Tslp

thymic stromal lymphopoietin

YAP

yes‐associated protein

α‐SMA

alpha‐smooth muscle actin

1. Introduction

Over the past 3 decades, the tumor microenvironment (TME) has attracted considerable attention as the essential regulator of tumor progression. The TME comprises noncellular components, such as the extracellular matrix and various cellular components, including fibroblasts, endothelial cells, and immune cells. In healthy circumstances, such stromal cells sustain physiological homeostasis and prevent disease progression, such as carcinogenesis [1, 2, 3, 4, 5]. However, cancer cells that evade such normal homeostatic mechanisms can convert phenotypes of these noncellular and cellular components into TME characteristics favoring their progression [2, 6]. Of these TME components, cancer‐associated fibroblasts (CAFs) play an essential role in tumor progression by promoting tumor metastasis and the mechanical alteration of tumor tissues, influencing angiogenesis, immune cell infiltration and activation, and therapeutic resistance [7, 8]. Therefore, understanding the biological basis underlying the intercellular interactions between cancer cells and CAFs can provide new avenues to improve tumor treatment.

However, there is a challenge that needs to be addressed to target the TME, particularly CAFs, as a novel treatment modality. That is, the phenotypic and functional diversity of CAFs. Several researchers so far have recognized the heterogeneity of CAFs through various types of analyses, including immunohistochemical techniques [9, 10, 11, 12]. In addition, CAFs have historically been considered a tumor‐promoting factor, but some evidence has shown that some CAFs also retained tumor‐suppressive functions [13, 14, 15]. Consistent with these observations, current single‐cell resolution platforms have demonstrated that CAFs are heterogeneous populations with various subtypes. Multiple factors, including their cellular origin, humoral factors, extracellular vesicles (EVs), and physical stimuli, may be responsible for generating such CAF heterogeneity. CAF heterogeneity is thus engaged in the different aspects of cancer biology, making understanding the biology of the TME and its therapeutic applications more challenging. In this review, recent advances in understanding the mechanisms of creation of CAF heterogeneity, the functions of CAF subtypes in cancer progression, and the role of extracellular vesicles in CAF subtype induction and function are discussed.

2. What Are Fibroblasts? Their Functions Are in Normal Homeostasis

Fibroblasts are key players in maintaining tissue and organ homeostasis. One of their essential functions is the provision of molecules that form the extracellular matrix (ECM), the scaffolding of the cell, such as collagen, fibronectin, laminins, elastins, integrins, and metalloproteinases (MMPs), and other tissue‐specific ECM proteins [7, 8, 16]. By providing these proteins, fibroblasts sustain cellular scaffolds and the architectures of tissues and organs to promote correct cellular proliferation, differentiation, and tissue morphogenesis. For example, fibroblasts are temporarily activated during lung development to support alveolarization by producing elastin and tenascin frameworks [17]. The generation of mechanical force is also a typical function of fibroblasts. Physical contraction by activated fibroblasts appears to involve alveolar septum formation in alveolarization [18]. In particular, during the wound‐healing process, fibroblasts migrate into the wound and then enter the activated state, termed myofibroblasts, that are the primary source of ECM molecules to support epithelial cell migration into the wound [1, 19]. Along with these functions of fibroblasts, they also serve as a source of humoral factors, such as growth factors, hormones, and chemokines [1, 8, 16]. In the pancreas and liver, other types of fibroblasts, termed stellate cells, have lipid droplets containing retinoic acid [20]. The activated state of stellate cells is closely involved in the progressive pathogenesis in the liver and pancreas, such as fibrosis, carcinogenesis, and the chemoresistance of tumors [21, 22, 23]. Within the population of PDGFR‐expressing myofibroblasts at the crypt–villus junction of intestinal tissue, two distinct mesenchymal cell types have been identified. Telocytes, characterized by platelet‐derived growth factor receptor alpha (PDGFRα)‐high Foxl1‐positive, are located at the villus base and function as a reservoir for BMP ligands. In contrast, trophocytes, defined by PDGFRα‐low CD81‐positive, reside beneath crypts and express the BMP inhibitor Grem1, thereby supporting and maintaining intestinal stem cells [24]. Fibroblasts are, therefore, essential for the maintenance of proper biological functions.

Although fibroblasts are often considered to have homogeneous characteristics, their properties and cell lineages differ considerably among tissues and organs. Indeed, single‐cell resolution analysis showed the phenotypical and functional differences of fibroblast lineages among all mouse organs [25]. Stellate cells have unique features, such as lipid droplet storage, distinct from other fibroblasts within tissues other than the liver and pancreas. The origin of most fibroblasts is the primitive mesenchyme that develops from the mesoderm, and other mesenchymal cells, including adipocytes and osteoblasts, share the same embryonic origin [1, 7, 8]. Therefore, it is not surprising that the fibroblasts comprise heterogeneous subtypes. Importantly, no specific markers of fibroblasts have been found so far that are not expressed in other cell types. It is necessary to bear in mind the histological and embryological heterogeneity of the fibroblasts from which they originate to understand how the heterogeneity of CAFs comes about.

3. Definitions of CAFs

Generally, CAFs refer to the fibroblasts or stellate cells (in the pancreas and liver) that reside within the tumor stroma [26]. CAFs are similar in nature to activated fibroblasts during wound healing [1, 7]. The wound‐healing mechanism, an intrinsic function of fibroblasts, acts in response to tissue destruction by cancer cells. However, as Harold Dvorak postulated in 1986, “tumours are non‐healing wounds,” [27] and tissue destruction by cancer cells is persistent, unlike wound healing, creating a unique environment in which inflammatory and wound‐healing responses are prevalent. Such a “non‐healing wounds” response creates CAFs stimulated by variant signals from cancer cells, other cell types, and ECMs within the TME. Along with the heterogeneity of CAFs in their origin, the variance of extracellular signals can generate highly complex CAF subtypes within the tumor stroma.

The same issue of the lack of comprehensive fibroblast markers applies to CAFs. Therefore, defining the entire CAF population precisely and uniformly using simple marker expressions is challenging. So far, the definition likely depends on the study's design and purpose. For instance, single‐cell resolution analysis uses gene expressions of collagens, decolin, and complement C1r to isolate fibroblast lineages within tumor tissues [28]. For in vitro experimental use, fibroblasts that crawl out from tumor tissues in culture are exploited [29, 30]. Combining several antibodies or gene expressions is also used to isolate specific CAF subtypes [9, 31, 32, 33, 34, 35, 36, 37, 38]. Conversely, combining multiple gene or protein expressions for other cell lineage markers can separate CAFs that do not express these lineage markers [8, 39]. Another criterion of CAFs is cells that lack genetic mutation within a tumor to discriminate them from cancer cells showing epithelial–mesenchymal transition (EMT) [8]. A self‐developed fibroblast single‐cell atlas is also used to identify universal fibroblast lineages across the whole organ [25]. It should clarify and accumulate information on how CAFs are defined and separated and what CAF functions have been identified to understand CAF biology and develop novel therapeutic modalities.

4. Definition of CAF Subtypes and Their Functions for Tumor Progression

Recent advances in single‐cell resolution techniques have identified numerous CAF subtypes. The well‐known subtypes are myofibroblastic CAFs (myCAFs), inflammatory CAFs (or immunomodulatory CAFs: iCAFs), and antigen‐presenting CAFs (apCAFs) (Figure 1 and Table 1). However, finer subpopulations with specific gene expressions appear even within these subtypes. Some subtypes do not belong to these three groups (Other CAF subtypes in Figure 1 and Table 1). This review focuses on particular CAF subtypes whose molecular mechanisms have been investigated for their role in cancer progression.

FIGURE 1.

Functions and markers for each CAF subtype. Cancer‐associated fibroblasts consist of various subtypes. MyCAFs are similar to the myofibroblast phenotype and express ECM‐related or mechanical contraction‐related genes. They promote cancer cell growth, migration, invasion, and inhibition of T‐cell infiltration via mechanotransduction and humoral factors. However, contrary to the tumor‐promoting functions, myCAFs also include tumor‐suppressive subgroups via a mechanical ECM barrier to suppress cancer cell invasion. The iCAFs are a subtype that produces cytokines and chemokines to enhance tumor growth and immunosuppression. Distinct from myCAFs, they have no myofibroblast‐like features, such as ECM‐related genes. The apCAFs express MHC Class II molecules involved in antigen presentation by immune cells and act as a decoy receptor that differentiates naïve CD4‐positive T cells from Tregs. As with other CAF types, many subtypes have been elucidated, including vCAFs and FAP+ PDPN+ CAFs. “Steady‐state” CAFs are also found and likely include both the tumor‐promoter and the tumor‐suppressor subgroups.

TABLE 1.

Functions of CAF subtypes in each tumor type discussed in this review.

Tumor type	CAF subtype (subpopulation)	Functions	Origin of CAF subtype	Induction mechanisms	Subtype markers	References
Pancreatic ductal adenocarcinoma (PDAC)	myCAF	Not investigated	Pancreatic Stellate Cells	Jaxtacrine interactions	α‐SMA	[33]
	iCAF	Sustain the survival of PDAC organoids	Pancreatic Stellate Cells	Paracrine interactions	IL‐6, IL‐11, LIF etc.	[33]
	myCAF	Not investigated	Not investigated	Not investigated	ACTA2, TAGLN, MMP11, MYL9, HOPX, POSTN, TPM1, TPM2	[31]
	iCAF	Not investigated	Not investigated	Not investigated	IL6, PDGFRA, CXCL12, CFD, DPT, LMNA, AGTR1, HAS1, CXCL1, CXCL2, CCL2, IL8	[31]
	apCAF	Induce CD25 and CD69 in cocultured T cells	Not investigated	Not investigated	COL1A1, CD74, HLA‐DRA, HLA‐DPA1, HLA‐DQA1, SLPI	[31]
	EGFR/ERBB2 signaling activated myCAF	Promote PDAC metastasis	Pancreatic Stellate Cells	TGF‐β signaling	EGFR/ERBB2 activation, AREG, Dusp6, ECM‐associated and TGF‐β dependent gene signatures, and KRAS signaling related genes etc.	[40]
	iCAF	Not investigated	Pancreatic Stellate Cells	IL1 signaling	JAK/STAT signaling and NF‐κB related genes	[40]
	apCAF	Not investigated	Not investigated	Not investigated	Ly6C‐ and MHCII+	[40]
	LRRC15+ myCAF	Induce therapeutic resistance of cancer cells against immune checkpoint inhibitors	Pancreatic Stellate Cells	TGF‐β signaling	LRRC15 and PDPN	[41]
	iCAF	Not investigated	Pancreatic Stellate Cells	May induced by IL1 signaling	CXCL1, HAS1 etc.	[41]
	LRRC15+ myCAF	Suppress CD8‐T cell functions to support the immune evasion of cancer cells	Dermatopontin (DPT) + fibroblasts	TGF‐β signaling	LRRC15 and PDPN	[42]
	apCAF	Act as a decoy receptor that differentiates naïve CD4‐positive T cells into regulatory T‐cells (Tregs)	Mesothelial cells	Both IL1 and TGF‐β signaling	Cd74, H2‐As, H2‐Ab1, H2‐Eb1, Msln, Uok3b, Lrrn4, Krt19	[43]
	FAP+ CAF	Suppress antitumor immune effects via CXC‐chemokine ligand 12 (CXCL12) production	Not investigated	Not investigated	FAP	[44, 45]
	Meflin+ CAF	Suppress tumor progression and enhance the response to immune checkpoint inhibitors	Mesenchymal stem cells	Depending on Meflin expression	Meflin	[46, 47, 48]
Breast cancer	CAF‐S1	Enriched in triple‐negative breast cancer and associated with an immunosuppressive microenvironment and promote Treg differentiation via CXCL12	Not investigated	Not investigated	CD29Med, FAPHi, FSP1Low‐Hi, α‐SMAHi, PDGFRbMed‐Hi, CAV1Low	[34]
	CAF‐S2	Enriched in Luminal A breast cancer	Not investigated	Not investigated	CD29Low, FAPNeg, FSP1Neg‐Low, α‐SMANeg, PDGFRbNeg, CAV1Neg	[34]
	CAF‐S3	Enriched in Juxta‐tumor	Not investigated	Not investigated	CD29Med, FAPNeg, FSP1Med‐Hi, α‐SMANeg‐Low, PDGFRbMed, CAV1Neg‐Low	[34]
	CAF‐S4	Enriched in HER2 breast cancer	Not investigated	Not investigated	CD29Hi, FAPNeg, FSP1Low‐Med, α‐SMAHi, PDGFRbLow‐Med, CAV1Neg‐Low	[34]
	CAF‐S1	Stimulate cancer cell migration and invasion through CXCL12 and TGF‐β signallings to promote metastasis	Not investigated	Not investigated	FAPHigh, CD29Med‐High, α‐SMAHigh, PDPNHigh, PDGFRβHigh	[49]
	CAF‐S4	High contractile capacity to support cancer cell invasion	Not investigated	Not investigated	FAPLow‐Med, CD29High, α‐SMAHigh, PDPNLow, PDGFRβMed	[49]
	vCAF	Associated with microvasculature signatures	Perivascular cells (pericytes)	Not investigated	Vascular development and angiogenesis‐related genes	[39]
	matrix CAF	Associated with invasion sigunatures of cancer cells	Resident fibroblasts	Not investigated	ECM and EMT‐related genes	[39]
	cycling CAF	May be proliferating segment of vCAFs	vCAF subpopulation	Not investigated	Miki67, cell cycle related genes	[39]
	developmental CAF	Not investigated	May derived from cancer cells possessing EMT	Not investigated	Tissue development, cell differentiation, and ECM‐related genes; Similar expression pattern to tumor epithelium	[39]
Colon cancer	myCAF within liver metastatic sites	Possess hyaluronan‐mediated tumor‐promotive function and may also have tumor‐suppressive function via creating ECM barrier	May derived from hepatic stellate cells	Not investigated	Acta2 and several collagen isoforms	[50]
	iCAF within liver metastatic sites	Secret HGF to enhance metastatic tumor growth	May derived from hepatic stellate cells	Not investigated	Growth factor and inflammatory gene expressions	[50]
	mesothelial CAF (mesCAF) within liver metastatic sites	Not investigated	May derived from portal fibroblasts in liver	Not investigated	Mesothelial marker expressions (Msln, Upk1b, and Up3kb)	[50]
	myCAF	May have tumor‐suppressive function but not investigated precisely	Not investigated	Not investigated	ACTA2, RGS5, CSPG4, and AOC3	[51]
	IL1R1+ iCAF	Suppress CD8‐positive killer T cells and induce tumor‐promoting macrophages	Resident fibroblasts	IL1 signaling	IL1R1, CXCL14, CXCL12, CXCL16	[51]
	IL11+ CAF	Promote tumor formation and growth	Resident fibroblasts	IL11 induced STAT3 signaling	IL11	[51]
Gastric cancer	myCAF‐like	Not investigated	Resident fibroblasts	TGF‐β signaling	ACTA2, COL4A1	[29]
	iCAF‐like	Create a favorable tumor microenvironment for cancer progression via chemokines	Resident fibroblasts	miRNAs (miR‐193b etc.) packaged into extracellular vesicles	CXCL1, CXCL2, CXCL8	[29]
Intracolangio carcinoma (ICC)	myCAF	Promote tumor growth through hyaluronan synthase 2	Hepatic Stellate Cells	Not investigated	ECM‐related genes, Col1a1, and SERPINF1	[52]
	iCAF	Promote tumor growth through HGF‐MET signaling	Hepatic Stellate Cells	Not investigated	Inflammatory and growth factor related genes and Rgs5	[52]
	mesCAF	Not investigated	Portal fibroblasts	Not investigated	Msln, Upk1b, Upk3b, and Gpm6a as markers of portal fibroblasts/mesothelial cells	[52]
	vCAF	Contribute tumorigenesis and cancer stemness through the IL‐6/IL‐6R signaling	Not investigated	Not investigated	IL6, MYH11, CD146, CCL8, RGS5, ACTA2, COL1A2, PDGFRB	[53]
	matrix CAF	Associated with ECM and stroma‐ related gene signatures	Not investigated	Not investigated	POSTN, COL5A1, COL5A2, COL6A3, FN1, LUM, DCN, VCAN, ACTA2, COL1A2, PDGFRB	[53]
	iCAF	Not investigated	Not investigated	Not investigated	FBLN1, C3, IGF1, CXCL1, IGFBP6, SLP1, SAA1, ACTA2, COL1A2, PDGFRB	[53]
	apCAF	Not investigated	Not investigated	Not investigated	HLA‐DRB1, CD74, HLA‐DRA, ACTA2, COL1A2, PDGFRB	[53]
Lung cancer	PDPN+ CAF	Increase EGFR‐TKI resistance in lung cancer cells	Not investigated	Not investigated	PDPN	[54]
	apCAF	Directly activate the TCRs of CD4 T cells and act as tumor suppressor.	May originate from ATII cells	Not investigated	MHCII, SLP1, IL6, CFD, C1QA, C1QB	[55]
	myCAF	Not investigated	Not investigated	Not investigated	COL1A1, COL1A2, COL3A1, C1QTNF3, FAP, PDGFRA, CXCL12, ACTA2, MMP11, MMP13	[56]
	iCAF	May participate in brain metastasis of cancer cells through MET‐HGF signaling pathway	Not investigated	Not investigated	FAP, PDGFRA, CXCL12, CXCL2, C7, APOD	[56]
	apCAF	May participate in bone metastasis of cancer cells through SPP1‐CD44/SPP1‐PTGER4 signaling pathway	Not investigated	Not investigated	CXCL2, CD74, HLA‐DRA, HLA‐DPB1 and HLA‐DPA1, RGS1, HLA‐DRB1, CCL4	[56]
	myCAF	Not investigated	Resident fibroblasts	Extracellular vesicles	Proliferation‐ and ECM‐related genes	[57]
Bladder cancer	ITGA11+ PDGFRα+ myCAF	Promote lymphangiogenesis and assist in intravasation to aid lymphatic metastasis	Resident fibroblasts	TGF‐β signaling	ITGA11 and PDGFRα	[58]
	iCAF	Induce EMT in cancer cells through IL‐6 signaling	Resident fibroblasts	Cancer‐derived extracellular vesicles	IL6, ACTA2, FAP	[59]
Ovarian cancer	CAF‐S1	Increasing attraction, survival, and differentiation of T lymphocytes	Not investigated	Not investigated	CD29Med‐Hi, FAPHi SMAMed‐Hi, FSP1Med‐Hi, PDGFRβMed‐Hi, CAV1Low	[35]
	CAF‐S2	Not investigated	Not investigated	Not investigated	CD29Low FAP‐Neg SMANeg‐Low FSP1Neg‐Low PDGFRβNeg‐Low CAV1Neg	[35]
	CAF‐S3	Not investigated	Not investigated	Not investigated	CD29Med FAPLow SMALow FSP1Med‐Hi PDGFRβMed CAV1‐Neg‐Low	[35]
	CAF‐S4	Not investigated (Could not be maintained in culture)	Not investigated	Not investigated	CD29Hi FAPLow SMAHi FSP1Hi PDGFRβMed‐HiCAV1Neg‐Low	[35]
	ANTXR1+ ECM‐myCAF	Associated with chemotherapy resistance and suppression of CD8‐positive T‐cell activation	Not investigated	Not investigated	ANTXR1, FAP, ACTA2, CD29, FSP1, LRRC15, YAP1	[36]
	Detox‐iCAF	Enhance CD8‐positive T‐cell migration	Not investigated	Not investigated	ANTXR1‐negative, lower expression of FAP and α‐SMA	[36]
Skin cancer	Matrix CAF	Form ECM barrier to restrict T‐cell invasion into tumors	May derived from resident fibroblasts	Not investigated	FAP, α‐SMA, COL1A1, COL1A2, COL3A1, Lumican, Periostin, Tenascin‐C	[60]
	iCAF	Have immunomodulatory functions and may contribute to the induction of T‐cell exhaustion	May derived from resident fibroblasts	Not investigated	IL6, IL8, IDO1, MMP1, MMP3, FAP, α‐SMA	[60]
	RGS5+ CAF	Not investigated	May derived from resident fibroblasts	Not investigated	FAP, α‐SMA, RGS5, TAGLN, DES	[60]
	iCAF‐like	Enhance recruitment of neutrophil into the lung metastatic niche	Resident fibroblasts in lung	RNAs (Hmgb1, Tslp, Irf1) in extracellular vesicles	Inflammatory genes including IL1α, IL1β, CCL2, CCL3, CCL5, CXCL10, CXCL12	[61]

Note: For some CAF subtype markers, classifications by expression level were used. In these cases, “Hi” for high expression, “Med” for medium expression, “Low” for low expression, “+” or “positive” for expressing marker and “‐” or “negative” for no marker expression were described. In addition to CAF subtypes discussed in this review, various other subtypes have been reported.

5. Myofibroblastic CAFs (myCAFs)

MyCAFs are similar to the myofibroblast phenotype described above, which arises during wound healing and expresses ECM‐related or mechanical contraction‐related genes, such as alpha‐smooth muscle actin and collagens (Figure 1 and Table 1). Thus, myCAFs are likely capable of cancer cell migration and invasion through ECM stiffening [62, 63]. In particular, myCAFs play a principal role in the mechanotransduction regulated by ECM hardening, leading to yes‐associated protein (YAP) activation and nuclear localization to induce survival and proliferation signals in cancer cells [64]. Paracrine signals derived from myCAFs, such as transforming growth factor‐beta (TGF‐β) signaling, mediate not only the activation of surrounding benign stromal fibroblasts but also the induction of EMT, cell migration, invasion, and immune evasion of cancer cells [8, 40, 65, 66]. Collagen Type I+ myCAFs are involved in intrahepatic cholangiocarcinoma (ICC) cell growth via hyaluronan production [52]. Thus, myCAFs can be involved in tumor progression via humoral factors and mechanical interactions.

Regarding other functions of myCAFs, alpha‐smooth muscle actin (α‐SMA)‐positive and leucine‐rich repeat containing 15 (LRRC15)‐positive podoplanin (PDPN)‐positive CAFs induce therapeutic resistance of cancer cells against immune checkpoint inhibitors (ICIs), such as antiprogrammed death ligand 1 (PD‐L1) antibodies [41], and they suppress CD8‐T cell functions to support the immune evasion of cancer cells [42]. Similarly, the Anthrax Toxin Receptor 1 (ANTXR1)‐positive fibroblast activation protein (FAP)‐positive CAF subtype in high‐grade serous ovarian cancer is identified as an ECM‐producing myCAF, which is associated with chemotherapy resistance and suppression of CD8‐positive T‐cell activation in a YAP1‐dependent manner [36]. The myCAFs expressing integrin subunit alpha 11 (ITGA11) and PDGFRα promote lymphangiogenesis of endothelial cells and assist in the intravasation of bladder cancer cells to aid lymphatic metastasis [58]. Various subpopulations in the myCAF subtype might have shared and distinct tumor‐promoting functions.

In contrast to the tumor‐promoting functions, tumor‐suppressive functions of myCAFs have been found in the pancreatic ductal adenocarcinoma (PDAC) mouse model. Ganciclovir‐induced depletion of α‐SMA+ CAFs in mice that spontaneously develop PDAC unexpectedly results in poorly differentiated tumors with enhanced hypoxia and EMT and increased regulatory T cells to suppress immune surveillance [13]. An experimental system that deleted the Collagen Type I gene specifically from α‐SMA+ CAFs has also been reported by the same research team, and tumor progression and induction of immunosuppression have been shown even in this case [15]. This tumor‐promotive effect of collagen depletion from myCAFs can be due to the loss of the collagen‐based mechanical ECM barrier surrounding the cancer cells [50]. In the model of ICC [52], the presence or absence of Collagen Type I expression in myCAFs did not impact tumor growth, suggesting that the mechanical actions of myCAFs do not universally assist cancer progression. These mechanotransductive effects may differ depending on the cancer cell type, CAF origin, and the surrounding tissue environment. The genetic engineering loss of sonic hedgehog (Shh) has also been reported. Shh ligand and Hedgehog (Hh) signaling contribute to desmoplastic stroma formation in many solid tumors. Intriguingly, loss of Hh signaling decreases α‐SMA+ stromal cells, but it extensively accelerates tumor progression in the PDAC model [14]. In this PDAC model, the α‐SMA+ CAFs with heterogeneous expression of the Hh signaling‐targeted gene are also found, suggesting that myCAFs comprise finer subpopulations with conflicting functions in cancer progression. Considering that the features of myCAFs resemble those of activated fibroblast features, they can result from the contradictory wound‐healing action of benign‐resembling subtypes and the tumor‐supporting function of other subtypes, although this remains unclear.

6. Inflammatory CAFs (iCAFs)

The iCAFs are a subtype that produces cytokines and growth factors, thereby inducing an inflammatory microenvironment and supporting tumor progression (Figure 1 and Table 1). They were first defined in the PDAC model [33] and then have also been reported in various cancer types, including breast, gastric, and colorectal cancers. The iCAFs in PDAC arise from pancreatic stellate cells cocultured with cancer organoids. Functionally, they produce inflammatory cytokines such as interleukin (IL)‐6, IL‐11, and leukemia inhibitory factor (LIF) to sustain the survival of PDAC organoids [33]. In the ICC, iCAF‐producing hepatocyte growth factor (HGF) enhances ICC cell proliferation via the cancer‐expressing HGF receptor MET [52]. Similarly, a study using liver metastasis models of colorectal cancer and PDAC reported that HGF derived from iCAFs within metastatic sites contributes to metastatic tumor growth [50]. Interestingly, this study also suggested that myCAFs derived from hepatic stellate cells, also the origin of iCAFs, possess a tumor suppressor aspect by producing Collagen Type I. Thus, iCAF subtypes have distinct functions from myCAFs but can arise from a common cell type.

As for the other functional role of iCAFs, several reports show the immunosuppressive function of iCAFs. For instance, iCAF‐expressing IL‐1 receptors (IL1R1) have been reported in CMS4 colorectal cancer. IL1R1+ iCAFs can suppress CD8‐positive killer T cells and induce tumor‐promoting macrophages [51]. The iCAFs in skin cancers also have immunomodulatory functions and may contribute to the induction of T‐cell exhaustion [60]. These findings suggest that iCAFs induce tumor growth and antitumor immune effects as principal functions through humoral factors such as inflammatory cytokines. Although myCAFs also mediate tumor growth and immune surveillance, iCAFs are partially distinct from myCAF functions with ECM stiffness and mechanotransduction.

7. Antigen‐Presenting CAFs (apCAFs)

The apCAF is a subtype identified by single‐cell RNA‐seq analysis using pancreatic cancer tissues and expresses MHC Class II molecules involved in antigen presentation by immune cells (Figure 1 and Table 1). Functionally, it can act as a decoy receptor that differentiates naïve CD4‐positive T cells into regulatory T cells (Tregs) [31, 43]. However, unlike apCAFs in pancreatic cancer, a subpopulation of apCAFs in lung cancer promotes MHC Class II immunity and acts to suppress tumors [55]. As for apCAFs in pancreatic cancer, they can originate from mesothelial cells, and indeed, a monoclonal antibody to the mesothelial cell marker mesothelin reduced the abundance of apCAFs [43]. Recently, it has also been reported that apCAFs in nonsmall cell lung cancer are involved in bone metastasis of cancer cells, whereas iCAFs contribute to brain metastasis [56]. Whether cancer cells with bone and brain metastasis capabilities have converted CAFs to their properties remains unclear, but CAF subtypes may directly manage the organ tropism in tumor metastasis.

8. Other CAF Subtypes

Several reports have identified other CAF subtypes besides the CAF subtypes discussed above (Figure 1 and Table 1). In 2006, immunohistochemical analysis using mouse tumor tissues visualized CAF phenotypic heterogeneity [9]. This study showed a fibroblast‐specific protein‐1 (FSP1/S100A4)‐positive CAF subtype and other subpopulations expressing α‐SMA, PDGFRβ, and neuron–glial antigen 2 (NG2). This finding suggests CAF heterogeneity due to differences in marker protein expression and cell lineage of origin [9]. A study using a mouse PDAC model that specifically eliminates FAP‐positive cancer stromal cells from primary tumors showed that FAP‐positive CAFs suppress antitumor immune effects via CXC‐chemokine ligand 12 (CXCL12) production [44, 45]. CAF‐secreted CXCL12 is also closely associated with tumor angiogenesis [67, 68]. Related to this, CAF subtypes with angiogenesis‐related gene expressions, termed vascular‐CAFs (vCAFs), are found [39, 53]. As for other CAF subtypes, PDPN‐positive CAFs have been reported to increase EGFR‐TKI resistance in lung cancer cells [54]. However, these FAP+ or PDPN+ CAFs may be “subgroups” subdivisions of CAF subtypes such as myCAF and iCAF described above. For instance, five CAF markers, including FAP and PDPN, are used to identify the CAF‐S1 subtype, a key modulator in breast and ovarian cancer progression [34, 35, 49]. The subsequent single‐cell analysis showed that the FAP‐high CAF‐S1 subtype defined in breast cancer includes five myCAF and three iCAF subgroups [37]. Whether subpopulations of FAP+ or PDPN+ CAF subtypes are common to all solid tumors remains unclear. Further investigation, including other cellular lineage markers, would aid in defining precise CAF subtypes.

Normal fibroblasts are known to function in an inherently cancer‐suppressive manner [69, 70]. Interestingly, such “steady‐state” normal fibroblast subtypes with peptidase inhibitor 16 (Pi16)‐positive and dermatopontin (Dpt)‐positive expressions are found in both healthy and diseased tissues [25]. Experimental depletion of LRRC15‐positive myCAFs has been reported to increase the number of Pi16+ subtypes within tumor tissues [42]. Moreover, synthetic retinoids may enhance the response to immune checkpoint inhibitors by artificially inducing the cancer‐suppressive Meflin+ CAF subtype [46, 47, 48]. Meflin is a marker of mesenchymal stem cells (MSCs) and fibroblasts in normal tissues [71]. These findings suggest that direct targeting of cancer‐promoting CAF subtypes induces tumor‐suppressing CAF subtypes within the TME. However, the CAF subtype with a secretome pattern similar to normal fibroblasts is also found in oral squamous cell carcinoma. It can promote cancer cell invasion in a hyaluronic acid‐dependent manner [72]. Therefore, not all “steady‐state” fibroblast subtypes likely behave as tumor suppressors. Determining whether they promote or suppress cancer may be complicated by subtyping on simple transcriptome and proteome profiling.

9. Why Do CAF Subtypes Arise?

9.1. Cellular Diversity and Cancer‐Derived Signals Determining CAF Subtypes

As discussed above, various factors are involved in the creation of CAF subtypes within the TME. One of these factors is the cellular lineage‐dependent mechanism. Although resident fibroblasts, stellate cells, or astrocytes within specific organs are the primary origin of CAFs, various cell lineages have been proposed as the origin of CAFs, including adipocytes, bone marrow‐derived mesenchymal stem cells (BM‐MSCs), vascular endothelial pericytes, epithelial cells, and mesothelial cells (cell lineage‐dependent, Figure 2) [6, 26]. For instance, a study with an in vivo model of gastric cancer (GC) showed that BM‐MSCs are recruited in dysplastic stomach and subsequently converted into CAFs [73]. Consistent with this study, BM‐MSC‐derived CAFs are responsible for cell proliferation and angiogenesis in the lung metastasis model of breast cancer. However, PDGFRα‐positive CAFs in metastatic foci are derived from resident fibroblasts and express Col1a, Acta2, and extracellular matrix‐related genes [74]. Mesothelial cells and pericytes can be the origin of apCAFs and vCAFs, respectively [39, 43, 53]. These findings show that cellular lineage‐dependent mechanisms can shape part of the functional and phenotypic heterogeneity of CAFs.

FIGURE 2.

How do cancer cells produce distinct CAF subtypes? Various factors are involved in CAF subtype creation within the tumor microenvironment (TME). One is the cellular lineage‐dependent mechanism. Many origin cells can convert into CAFs, so multiple features emerge as CAF subtypes. Second is the environmental signaling‐dependent mechanism. A single‐cell lineage can differentiate into different CAF subtypes via various signaling pathways within the TME. Interconversion between CAF subtypes is also found. Third is the epigenetic regulation‐dependent mechanism. Methylation patterns of histones are associated with CAF subtype marker gene coding regions, suggesting that epigenetic regulation also contributes to CAF heterogeneity.

As other factors involved in the creation of CAF subtypes, the features of cancer cells, such as genetic background and metastatic properties, also affect CAF subtype composition (environmental signaling‐dependent, Figure 2). Differential p53 mutation status induces different CAF phenotypes to create a premetastatic and chemoresistant microenvironment [75]. We also demonstrated that highly metastatic GC cells educate stromal fibroblast phenotypes in transcriptome and metabolomic levels [29, 76]. Thus, both intrinsic properties of cancer cells and intercellular signaling from the TME contribute to establishing CAF heterogeneity.

Besides the examples for creating CAF subtypes described above, inter‐subtype conversions of CAF subtypes have also been reported (environmental‐signaling dependent, Figure 2). In the PDAC model, the induction of myCAFs depends on TGF‐β signaling, but the iCAFs are determined by IL‐1B‐inducing NF‐κB/STAT3 signaling. A model of the antagonistic relationship of these signals that regulate the CAF subtype has been proposed [32, 33]. TGF‐β signaling and NF‐κB are well known as antagonists, suggesting that the balanced relationship between these signaling pathways creates CAF heterogeneity. Indeed, the induction of apCAFs from mesothelial cells in cancer tissue requires both TGF‐β and IL‐1α signals. Thus, apCAFs might have the properties of both myCAFs and iCAFs and represent an intermediate state in which both signals are received. Related to these findings, dipeptidyl peptidase‐4 (DPP4) and YAP1‐dependent signaling also mediate the conversion of iCAF subgroups, Detox‐iCAFs, to a myCAF subpopulation, ECM‐myCAFs [38]. The DPP4‐mediated pathway directly converts Detox‐iCAFs into ECM‐myCAFs, but the YAP1‐mediated pathway appears to induce other intermediate myCAF subgroups resembling features of activated fibroblasts during wound healing. TGF‐β signaling via transforming growth factor‐beta receptor 2 (TGFBR2) can also induce ECM‐myCAFs from Detox‐iCAFs [38]. Highly complex intercellular signaling may organize stromal heterogeneity in solid tumors.

Epigenetic changes have been proposed as the other way for CAF subtype regulation to be independent of cell‐derived signals and fibroblast lineages (epigenetic regulation‐dependent, Figure 2). Single‐cell resolution ATAC‐seq analysis shows that open chromatin status correlates with some CAF subtype‐related gene loci, suggesting that chromatin accessibility affects transcriptomic features and may determine some CAF subtypes [77]. Therefore, epigenetic changes within the tumor background tissue may mediate CAF subtype regulation. Further mechanistic analysis is needed to clarify CAF subtype creation.

9.2. Contribution of Extracellular Vesicles in CAF Subtype Induction

EVs are lipid bilayer membrane vesicles that contain various bioactive molecules, such as mRNAs, microRNAs (miRNAs), and proteins. These molecules in EVs are transferred to the recipient cells within the proximal tissues and distal organs and mediate numerous cellular phenotypes. In particular, many cancer cells crosstalk with stromal cells through EVs to regulate tumor initiation, growth, invasion, metastasis, and treatment resistance. In addition, EVs are detectable in various body fluids, such as blood, saliva, and urine, besides their striking functions in intercellular communications [78, 79]. The composition of these molecules in EVs varies according to the gene expression profiles of their origin cells. Thus, EVs are attractive targets as an avenue to improve cancer diagnosis if they package tumor‐specific molecules.

EVs include various subtypes with different components, sizes, and biogenesis processes. For example, exosomes (100–200 nm) originate from the exocytosis of multivesicular bodies (MVBs) [79, 80]. Ectosomes (150–1000 nm) are generated directly from the plasma membrane [80]. Apoptotic bodies (100–5000 nm) arise during programmed cell death [80]. However, the expert consensus in EV research encourages using the nomenclature “EVs” unless subcellular origin can be demonstrated. This review will use the term “EVs” for all subtypes of membrane vesicles in the extracellular space to avoid misleading readers.

So far, many researchers have demonstrated that cancer‐derived EVs convert different stromal cell types, including fibroblasts, macrophages, and mesenchymal stem cells, into CAFs [79, 81]. However, most of these reports regarding the induction of CAFs via EVs remain to clarify whether cancer EVs can create heterogeneous subtypes of CAFs. Given that cancer‐derived signals can reprogram a single‐cell lineage into two types of CAF subtypes [32, 33], it is highly plausible that cancer EVs are also responsible for CAF subtype creation. Indeed, our group found that gastric cancer (GC) cell‐secreting EVs induce chemokine expressions such as IL‐1 and IL‐8 in stomach fibroblasts, similar to the property of iCAFs [29]. As mechanisms, these GC‐derived EVs transport a series of miRNAs, including miR‐193b, into stomach fibroblasts to induce chemokine expressions. These GC cell‐derived EVs cannot regulate the expression of α‐SMA and collagen, which are part of the myCAF subtypes. Importantly, high‐metastatic GC cells are inducible in both α‐SMA and chemokine expressions in fibroblasts compared with low‐metastatic GC cells. These findings suggest that cancer‐derived EVs selectively confer iCAF‐like features in fibroblasts. IL‐8‐expressing CAFs within tumor stroma are closely associated with poor outcomes in patients with GC, suggesting that GC generates a microenvironment favorable for their progression. Consistent with our findings, other research groups also provided evidence regarding CAF subtype creation via cancer‐derived EVs. Bladder cancer cell‐derived EVs convert the primary bladder fibroblasts into iCAF features [59]. Compared with TGF‐β treatment, these cancer‐derived EVs have a lesser effect on the induction of α‐SMA in fibroblasts [59]. EV‐induced iCAFs express IL‐6 and activate the STAT3 signaling pathway that leads to the EMT in bladder cancer [59]. In addition, metastatic melanoma cell‐derived EVs activate proinflammatory signaling in the fibroblasts of metastatic niches [61]. These EVs contain RNAs capable of instigating inflammatory signaling in the fibroblasts, such as high‐mobility group box 1 (Hmgb1), thymic stromal lymphopoietin (Tslp), and interferon regulatory factor 1 (Irf1) [61]. Interestingly, EVs derived from lung adenocarcinoma cells induce ECM‐myCAF‐like features but decreased iCAF‐related gene expressions in lung fibroblasts [57]. These findings suggest that cancer cells may confer favorable features to surrounding stroma via EVs and change the CAF subtype balances within the TME (Figure 3).

FIGURE 3.

Impact of extracellular vesicles on CAF subtype construction. Extracellular vesicles (EVs) also contribute to CAF subtype construction. The composition of these molecules in EVs varies according to the gene expression profiles of their origin cells. Thus, the impact of cancer EVs on CAF subtypes also apparently depends on cancer cell features. EVs derived from GC and bladder cancer can induce iCAF subtypes through transferring miRNAs and lncRNAs. In contrast, lung cancer‐derived EVs have a capacity for myCAF induction, but not iCAF induction. Cancer cells may induce CAF subtypes that have adapted to their features via EVs.

10. Conclusion

Accumulating evidence has clarified the nature of CAFs and their function in tumor progression. However, many aspects of CAF biology, in particular, how CAF heterogeneity is created and how they mediate tumor progression and therapeutic resistance remain unclear. In addition, many CAF subtype classifications have emerged from single‐cell analyses and are likely not yet well organized. The classification of subtype markers and their precise function in tumor progression will need to be addressed in future studies. The fact that CAFs have both cancer‐promotive and inhibitory subtypes has led to the key concept that the balance between them determines the role of CAFs in the TME [46, 48]. However, given that different cell lineages and subtypes have been reported for each organ and tissue of origin, we believe a systematic understanding of CAF subtypes in all cancer types is required to prove this concept. Understanding the nature of the CAF subtypes and their function may facilitate the development of new therapeutic modalities and diagnostic methods by leveraging the profile of CAF subtypes.

Author Contributions

Yutaka Naito: conceptualization, funding acquisition, project administration, visualization, writing – review and editing.

Ethics Statement

The author has nothing to report.

Consent

The author has nothing to report.

Conflicts of Interest

The author declares no conflicts of interest.