Fibrous stroma: Driver and passenger in cancer development

Abstract

Cumulative evidence shows that fibrogenic stroma and stiff extracellular matrix (ECM) not only result from tumor growth but also play pivotal roles in cellular transformation and tumor initiation. This emerging concept may largely account for the increased cancer risk associated with environmental fibrogenic agents, such as asbestos and silica, and with chronic conditions that are fibrogenic, such as obesity and diabetes. It may also contribute to poor outcomes in patients treated with certain chemotherapeutics that can promote fibrosis, such as bleomycin and methotrexate. Although the mechanistic details of this phenomenon are still being unraveled, we provide an overview of the experimental evidence linking fibrogenic stroma and tumor initiation. In this Review, we will summarize the causes and consequences of fibrous stroma and how this stromal cue is transmitted to the nuclei of parenchymal cells through a physical continuum from the ECM to chromatin, as well as ECM-dependent biochemical signaling that contributes to cellular transformation.

INTRODUCTION

The presence of dense collagenous stroma is a strong indicator of tumor aggressiveness and therapeutic resistance. Desmoplasia, the formation of fibrous tumor stroma, accounts for the hard, lumpy appearance of many types of carcinomas (1) and is a hallmark of cancer presumed to result from the stromal response to malignant progression (2). However, cumulative evidence has shown that desmoplasia may actually precede the emergence of malignant cells. Desmoplasia is often detected in normal breast tissues by mammography and is a strong risk factor for breast cancer (3). Myofibroblasts (MyoFbs), the cells responsible for desmoplasia, are often found in premalignant lesions or tumor-adjacent normal tissues and are proposed as key players in tumor initiation (4-6). The presence of MyoFbs in noncancerous tissues is considered an indicator of cancer predisposition, and various types of tissue injuries and chronic wounds harboring abundant MyoFbs are considered premalignant (7).

MyoFbs, which are associated with tissues undergoing repair and with tumors, are highly contractile and secretory cells derived from local fibroblasts, epithelia, and other cell types (8). Although their major function is to promote wound repair, their excessive activation dysregulates the production of extracellular matrix (ECM) and growth factors, ultimately inducing tissue fibrosis and hypertrophic scars. The abundance of MyoFbs in tumor stroma led to the description of tumors as “wounds that never heal” (9). It is well established that the abundant MyoFbs, indicative of reactive stroma, greatly contribute to tumor progression and aggressiveness, and emerging evidence substantiates the direct involvement of MyoFbs in tumor initiation (6, 10).

The establishment of fibrous stroma by MyoFb accumulation stimulates parenchymal cells biochemically and mechanically. The stromal signals are eventually transmitted to the parenchymal nuclei through the physical continuum between the ECM and chromatin, which modulates the expression of specific genes to adapt to the changes or to remodel the stroma. Parenchymal and stromal cells thus coevolve through such dynamic and reciprocal regulation. Here, we present an overview of the advances in our understanding of the roles of fibrotic stroma in cellular transformation and tumor initiation and the mechanistic bases of these effects.

PHENOTYPIC INFLUENCE BY STROMAL FIBROBLASTS

Fibroblasts, one of the most abundant cell populations in the stroma, are a heterogeneous group of spindle-shaped cells of mesenchymal origin with phenotypes ranging from quiescent, noncontractile fibroblasts to activated, contractile MyoFbs (11). These subpopulations produce distinct sets of ECM proteins, including collagens (types I, III, and V), proteoglycans, fibronectin, elastin, and glycosaminoglycans, which altogether constitute the interstitial matrix (IM), as well as paracrine factors and ECM-modifying enzymes for establishing the IM scaffolds (12). The phenotype of fibroblasts strongly influences both the microenvironment and the phenotype of parenchymal cells. Fibroblasts manifest distinct tissue- and context-specific phenotypes and gene expression profiles through differential epigenetic mechanisms (13). Even within the same tissue, there are functionally distinct fibroblast subpopulations (Fig. 1) (14). There are three general categories of fibroblasts: normal fibroblasts (NFs), cancer-associated fibroblasts (CAFs), and MyoFbs.

Fig. 1. Heterogeneity of fibroblasts.

Normal tissue stroma mainly contains normal fibroblasts (NFs), which are quiescent and spindle-shaped. The cancer-associated fibroblasts (CAFs) in tumor stroma include NF-like cells and myofibroblasts (MyoFbs), which are proliferative and contractile. Several major markers distinguish NFs from CAFs and between different NF and CAF subtypes. NFs are involved in tumor suppression through neighbor suppression (the direct inhibition of the growth of abnormal cells through cell-cell contacts), paracrine signaling that inhibits cell growth and migration, and maintenance of normal ECM. CAFs promote tumor initiation and metastasis through the production of growth factors, chemokines, and fibrous ECM, as well as immunosuppression.

Subgroups of fibroblasts

NFs are quiescent mesenchymal cells, comprising several distinct subpopulations (Table 1). In contrast to the CAFs that assist tumor progression, NFs suppress tumor initiation and metastasis. Coculturing premalignant or malignant breast epithelial cells with NFs in three-dimensional (3D) conditions reverts the phenotype of the epithelia, restoring quiescence (15). In animals, NFs cotransplanted with breast cancer cells suppress tumor growth (16). Such tumor-suppressive functions of NFs are sustained through the production of a group of proteins [including bone morphogenetic protein receptor 2, transforming growth factor β (TGF-β) receptor 2, and forkhead box F1 (FOXF-1) and FOXF-2] that protect NFs from becoming MyoFbs under the influence of cancer cells (17).

Table 1.

NF subpopulations identified by single-cell RNA sequencing.

Cell type	Marker expression	Morphology	Potential functions	Spatial distribution	Reference
–	lin− CD90+ CD39+ CD26− COL6A5+	Spindle shaped	Anti-inflammatory	Upper dermis	(245)
Preadipocytes	lin− CD90+ CD36+	Epithelioid	–	Lower dermis
Pericytes	lin− CD90+ CD39− RGS5+	–	–	Reticular dermis
Uncharacterized	lin− CD90+ CD39+ CD26+	Spindle shaped	–	Reticular dermis
Uncharacterized	lin− CD90+ CD39− RGS5−	–	–	Reticular dermis
Mesenchymal cells	SFRP2/DPP4	Small and tubular with narrow elongated nuclei	Matrix deposition	Between collagen bundles	(246)
Bone marrow stromal cells	FMO1/LSP1	Large with larger nuclei	Inflammatory cell retention	Interstitial and perivascular
Progenitor or stem cells	CRABP1	–	–	–
	COL11A1	–	Connective tissue cell differentiation	–
	FMO2	–	–	–
	PRG4	–	–	–
	C2ORF40	–	–	–
Secretory reticular	WISP2, SLPI, CTHRC1, MFAP5, and TSPAN8	–	Collagen/ECM production and organization	Reticular dermis	(247)
Secretory papillary	APCDD1, ID1, WIF1, COL18A1, and PTGDS	–	Collagen/ECM production and organization	Papillary dermis
Proinflammatory	CCL19, APOE, CXCL2, CXCL3, and EFEMP1	–	Inflammatory response	Whole dermis
Mesenchymal	ASPN, POSTN, GPC3, TNN, and SFRP1	–	ECM organization and skeletal system development	Reticular dermis

The antitumor functions of NFs are mediated by at least four mechanisms. First, NFs suppress the growth of abnormal cells by direct cell-cell contact through a phenomenon termed “neighbor suppression” (18). Second, NFs secrete tumor-suppressive paracrine factors, such as tumor necrosis factor (TNF) and interleukin-6 (IL-6); protease inhibitors, such as whey acidic protein four-disulfide core domain 1 (WFDC1), which is an inhibitor of matrix metalloproteinase 9 (MMP-9), and tissue inhibitors of metalloproteases (TIMPs), which inhibit both MMPs and a disintegrin and metalloproteinases (ADAMs); and inhibitors of protumorigenic TGF-β–like proteoglycans and fibrillins (19-21). Third, NFs help maintain the IM integrity that elicits tumor suppression in a manner dependent on the cytoskeletal regulator Ras homolog family member A (RhoA). RhoA knockout in NFs reorganizes vimentin, increases the cells’ cytoplasmic rigidity, and induces the expression of proinflammatory genes that promote tumor growth (22). Last, NFs produce tissue-specific ECM that is substantially softer than CAF-derived ECM and suppresses tumor growth (23). Culturing cancer cells on NF-derived ECM inhibits the histone demethylase JMJD1a, an activator of Yes-associated protein 1 (YAP) and Tafazzin (TAZ), both of which encode mechanosensitive transcription factors that promote cancer cell growth. Conversely, culturing cancer cells on CAF-derived ECM induces YAP and TAZ expression and exacerbates cancer cell growth (24).

NFs of different tissue origins, host species, and ages show different degrees of tumor suppression. For example, NFs from naked mole rats, which have an extremely long life span and high resistance to developing cancers, produce high molecular weight hyaluronan (HMWHA) that is five times larger than human or mouse hyaluronic acid (HA). HMWHA protects tissues and organs from aging and oncogenic transformation through early contact inhibition (25). Conversely, NFs from aged humans produce decreased amounts of hyaluronic and proteoglycan link protein (HAPLN1), which enhances ECM alignment to stimulate tumor cell invasion and protumor immunity (26).

The efficiency of NF-mediated tumor suppression is also influenced by the nature of the tumor cells. NFs produce the TGF-β inhibitor asporin, a small leucine-rich proteoglycan, in response to TGF-β signaling from adjacent cancer cells. However, asporin production is influenced by the subtype of cancer cells. For example, luminal-type breast cancer induces asporin in NFs, whereas triple-negative breast cancer inhibits it through production of high amounts of IL-1β (27). Similarly, NFs secrete an antitumor protein, slit homolog 2 (SLIT2), which binds to the receptor roundabout homolog 1 (ROBO1) on cancer cells. Although the SLIT2-ROBO1 axis exerts antitumor functions in breast cancer by inhibiting nuclear translocation of β-catenin (28), it exerts protumor functions in osteosarcoma by increasing the glycolytic enzyme, 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 2 (PFKFB2) (29).

CAFs are fibroblasts that have been activated by fibrotic signals from neoplasms and exhibit highly proliferative and migratory properties (30). Compared with NFs, CAFs are larger and contain more indented nuclei, conspicuous nucleoli, and cytosolic filament bundles (31). Similar to NFs, CAFs are a heterogeneous group of cells with different gene expression profiles and cellular origins (Table 2). They include myofibroblastic (myCAF) and inflammatory (iCAF) types as well as some intermediate types (32, 33). CAFs isolated from tumors promote tumor initiation, growth, and progression under experimental conditions in vitro and in vivo (16). Coculturing normal or premalignant breast epithelia with CAFs induces aberrant epithelial proliferation and morphogenesis (15), and cotransplantation of premalignant breast cells with either genetically engineered CAFs overexpressing hepatocyte growth factor (HGF) and TGF-β1 or fibroblasts from inflammatory lesions into mice induces mammary tumors (6, 16). Such protumor effects of CAFs are attributed to several major functions. Most notably, CAFs produce paracrine factors [such as HGF, TGF-β, and C-X-C motif chemokine ligand 12 (CXCL12)] that promote tumor growth and invasion (34). They also produce fibrous ECM proteins (such as collagen and elastin) that align and confine the epithelia, thus altering their morphology and polarity, ultimately inducing epithelial-mesenchymal transition (EMT) (16). These ECM fibers also exert mechanical force on the epithelia, activating oncogenic signals (34). CAFs also activate metabolic pathways to buffer toxic metabolites produced by cancer cells, such as lactate, thus sustaining tumor growth (35). By producing MMPs that remodel the ECM as they move, CAFs create a path that facilitates collective cancer cell invasion (36). They induce the angiogenesis that enables continued tumor growth through the production of proangiogenic factors, such as fibroblast growth factor 2 (FGF2) and vascular endothelial growth factor 1 (VEGF1) (37, 38). Although some CAFs help establish an anti-inflammatory tumor microenvironment (TME) by recruiting immunosuppressive cells, such as M2-type macrophages and myeloid-derived suppressor cells (MDSCs; immature myeloid cells) (39), iCAFs produce proinflammatory factors like CXCL2, leukemia inhibitory factor (LIF), and IL-6, highlighting the heterogeneity among CAFs (32, 40, 41). Last, CAFs help induce drug resistance in tumors (see below for more details) (42).

Table 2. CAF subpopulations identified by single-cell RNA sequencing.

LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma.

CAF subpopulations	Functions	Marker expression	Cell types	Cancer type	Reference
MyoFbs (myCAFs)	Actin cytoskeleton organization	α-SMAhi	Pancreas	Adenocarcinoma	(32)
Inflammatory CAFs (iCAFs)	Immunomodulation and monocyte recruitment	α-SMAlow and IL-6+
Antigen-presenting CAFs (apCAFs)	Immunomodulation	MHC class II and CD74			(248)
Cancer-promoting CAFs (pCAFs)	Promotion of tumor growth	Meflinneg α-SMApos	Pancreas	Adenocarcinoma	(57)
Cancer-restraining CAFs (rCAFs)	Suppression of tumor growth	Meflinpos α-SMAneg
PDPN CAFs (pCAFs)	Cell migration, immune regulation, wound healing, and extracellular fiber organization	PDPN, Cxcl12, Saa3, Cxcl1, IL16, Acta2, and Fbn1	Breast	Adenocarcinoma	(249)
S100A4 CAFs (sCAFs)	Antigen presentation and ECM remodeling and protein folding and metabolic regulation	Spp1hiS100A4low
		Spp1lowS100A4hi
Matrix CAFs (mCAFs)	ECM structure	PDGFRα and FBLN1	Breast	Adenocarcinoma	(250)
Vascular CAFs (vCAFs)	Vascular development and angiogenesis	NID2, CD31, αSMA, and PDGFRβ
Cycling CAFs (cCAFs)	Cell cycle, cell division, and angiogenesis	Ki-67 and PDGFRβ
Developmental CAFs (dCAFs)	Cell differentiation and development	SCRG1 and SOX9
CAF-S1	Immunosuppression	CD29Med FAPHi FSP1Low-Hi αSMAHi PDGFRβMed-Hi CAV1Low	Breast	Adenocarcinoma	(40, 230)
CAF-S2	–	CD29Low FAPNeg FSP1Neg-Low αSMANeg PDGFRβNeg CAV1Neg
CAF-S3	–	CD29Med FAPNeg FSP1Med-Hi αSMANeg-Low PDGFRβMed CAV1Neg-Low
CAF-S4	Oxidative metabolism	CD29Hi FAPNeg FSP1Low-MedαSMAHi PDGFRβLow-Med CAV1Neg-Low
CAF-A	–	MMP-2, decorin, and collagen type 1 α2	Colorectal	Adenocarcinoma	(251)
CAF-B	–	α-SMA, PDGFA, and transgelin
CAF-1	Tissue remodeling	MMP-11, CAV1, and THY1	Head and neck	Squamous	(252)
CAF-2	Tissue remodeling	JUN, FOS, TGFBR3, and FGF7
Cluster 1	–	COL10A1	Lung	Squamous/adenocarcinoma	(253)
Cluster 2	–	COL4A1, α-SMAhi, MEF2C, MYH11, and ITGA7
Cluster 4	–	PLA2G2A
Cluster 5	–	MMP-3 and mTOR
Cluster 6	–	FIGF
Cluster 7	–	CCL2 and mTOR
Proto-inflammatory fibroblasts (CAF-1)	Inflammation	SGK1, APOE, CXCL2, and PLA2G2A	Lung	Adenocarcinoma (LUAD)/squamous cell carcinoma (LUSC)	(33)
Proto-MyoFbs (CAF-2)	Matrix remodeling	CXCL12, IGF1, and COL1A1
Catabolic proto-MyoFbs (CAF-3)	Catabolism and matrix remodeling	POSTN, FAP, and COL1A1
MyoFbs (CAF-4)	Matrix remodeling	α-SMA, POSTN, and FAP

Protumor activities of CAFs could also be attributed to mutations. Although CAFs were initially reported to be genetically normal (43, 44), next-generation sequencing technology has revoked this notion. For example, 10 to 50% of CAFs from colorectal tumors were found to harbor somatic copy number alterations, the most common of which is the gain of a whole chromosome 7, indicating clonal expansion (45). In addition, CAFs from skin squamous cell carcinoma often have NOTCH1 amplification and overexpression, which prevents the kinase ataxia-telangiectasia mutated (ATM) from initiating the DNA damage response (DDR) and growth arrest upon ultraviolet exposure, thus promoting their proliferation potential (46). Moreover, prostate cancer cells often promote Tumor protein p53 (TP53) expression in CAFs, which suppresses their growth and confers selective pressure on a subpopulation of CAFs that are TP53-null and therefore highly proliferative (47). Likewise, CAFs of more than 25% of sporadic breast tumors harbor TP53 mutations in association with lymph node metastases (48). Such stromal mutations could be, in part, due to high amounts of MMP-3, which increases reactive oxygen species (ROS) that elicit genotypic stress (49). Mutant stromal cells could also be derived from cancer cells undergoing EMT (see below).

In addition to mutations, the epigenetic profiles of stromal fibroblasts greatly change during cancer development (50). Such changes manifest as global genomic hypo- and hypermethylation of select genes (51). For example, TGF-β signaling induces NFs to increase their production of LIF, which then promotes expression of the DNA methyltransferase 3b (DNMT3b). DNMT3b–mediated repression of the gene encoding Src homology region 2 domain–containing phosphatase-1 (SHP-1) causes constitutive activation of Janus kinase 1 (JNK1)–signal transducer and activator of transcription 3 (STAT3) signaling, which promotes the conversion of NFs to CAFs (52). Epigenetic reprogramming also mediates the conversion of mesenchymal stem cells (MSCs) to CAFs in pancreatic ductal adenocarcinoma (PDAC). Lactate that accumulates in the PDAC TME induces α-ketoglutarate in MSCs, leading to the activation of the demethylase ten-eleven translocation (TET), which replaces cytosine methylation with hydroxymethylation, thus triggering MSC differentiation to CAFs (53). Furthermore, ~70% of human epidermal growth factor (EGF) receptor 2 (HER2)–positive breast tumors harbors CAFs with hypermethylation of the genes encoding the progesterone receptor (PGR), hydroxysteroid 17-β dehydrogenase 4 (HSD17B4), and cadherin 13 (CDH13), compared with 10% of HER2-negative tumors, highlighting the existence of cancer subtype–specific CAF profiles (54). See Polyak et al. (55) for a comprehensive review of CAF genetic and epigenetic profiles.

The heterogeneity among CAFs implies that some could promote tumor growth and others inhibit it, thereby contributing to the complexity of the TME. Whereas most studies report tumor-promoting effects of CAFs, one study reports that the depletion of α–smooth muscle actin (α-SMA)–positive CAFs worsens PDAC metastasis and prognosis (56). Other studies have identified cancer-promoting (pCAF; Meflin⁻/α-SMA⁺) versus cancer-restraining (rCAF; Meflin⁺/α-SMA⁻) CAF subpopulations in PDAC (57, 58) and linked Meflin⁺ CAFs to improved responsiveness of non–small cell lung carcinoma (NSCLC) to immune checkpoint inhibitors (57, 58).

MyoFbs comprise a subpopulation of CAFs (myCAFs) that are highly contractile and secretory. MyoFbs are also present at wound sites in the absence of tumors. MyoFbs arise from the activation of tissue-resident fibroblasts and other cell types, which are quiescent during tissue homeostasis, by tissue damage and wound healing signals. They produce paracrine factors and ECM proteins that promote wound healing (59). After wound healing is complete, as indicated by stress release through ECM softening or by the formation of granulation tissue, MyoFbs are cleared through apoptosis (60). Under the continuous wound healing signals present in fibrotic lesions and tumors, MyoFbs evade the termination signal and persist, exacerbating the disease conditions (9, 60). Activated fibroblasts in fibrotic tissues and tumor stroma are collectively referred to as MyoFbs because they are highly similar to one another both morphologically and functionally, with no definitive phenotypes or markers to distinguish between the two (61). Single-cell RNA sequencing has uncovered some differences between the transcriptomes of MyoFbs in fibrotic tissues and those in tumor stroma (Table 3), but the functional contributions of these differences are yet to be determined.

Table 3.

MyoFb subpopulations in fibrotic stroma and tumor stroma.

	MyoFbs in fibrotic stroma				MyoFbs in tumor stroma
Tissue	Disease	MyoFb precursors	Marker expression	Reference	Disease	MyoFb precursors	Marker expression	Reference
Kidney	Kidney fibrosis*	Pericytes	NOTCH3+ RGS5+ PDGFRα− PDGFRβ+	(254)	Kidney cancer–Wilms tumor*	–	α-SMA+ PLVAP+ PDGFRβ+ EMILIN1+ SLC14A1−	(255)
		Fibroblast	MEG3+ PDGFRα+ PDGFRβ+
		Fibroblast	COLEC11+ CXCL12+ PDGFRβ+
Lung	Pulmonary fibrosis*	Mesenchymal cells	α-SMA	(256)	Lung cancer (NSCLC)*	Mesenchymal cells	α-SMA, POSTN, and FAP	(33)
	Idiopathic pulmonary fibrosis*	Unknown	PDGFRB, MYLK, NEBL, MYO10, MYO1D, RYR2, ITGA8, α-SMA, and COL	(257)
	Systemic sclerosis-associated interstitial lung disease (SSc-ILD)*	MFAP5hi fibroblasts	SFRP2, DPP4, COL10A1, DPEP1, TSPAN2, POSTN, and CTHRC1	(258)
Liver	Liver fibrosis*	Hepatic stellate cells	COL1A1, COL1A2, and COL3A1	(259)	HCC†	Hepatic stellate cells	GFAP, LRAT, and PDGFRα	(260-263)
	Liver cirrhosis*	Mesenchymal cells	PDGFRα+ and collagen	(264)		Fibrocytes	CD34 and collagen type 1	(263, 265)
Pancreas	Chronic pancreatitis‡	–	Tenascin Clo, α-SMA, desmin, and vimentin	(266)	PDAC‡	–	Tenascin Chi, α-SMA, desmin, and vimentin	(266)
					PDAC*	Fibroblasts	LRRC15	(267)
					High-grade IPMN*	Fibroblasts	α-SMA+ CXCL12− DES−	(268)
Skin	Cutaneous fibrosis§	Adipocytes	COL5A1, COL5A2, and COL1A2	(269)	Skin cancer§	–	GREM1	(270)
	Bleomycin induced fibrosis†	Mesenchymal cells	CD29hi	(271)

^*Single-cell RNA sequencing.

^†Flow cytometry.

^‡Transcriptome array.

^§Reverse transcription polymerase chain reaction.

MyoFbs are characterized by the production of α-SMA, which forms contractile stress fibers, and are distinct from other types of fibroblasts devoid of a contractile cytoskeleton (62). The high contractility of MyoFbs, along with increased ECM deposition and cross-linking, contributes to ECM stiffening (63). MyoFbs also manifest a stellate cell shape, unlike the spindle shape of NFs, as well as enhanced proliferation and migration, and the production of fibroblast activation protein (FAP), a marker of reactive stromal fibroblasts (61). MyoFbs are largely heterogeneous and originated from various types of locally residing cells; NFs are the main source in most tissues, but MyoFbs can also arise from epithelia, endothelia, pericytes, mesenchyme, smooth muscle, bone marrow, and immune cells (8, 64). In liver fibrosis, the major source of MyoFbs is the resident pericyte population, hepatic stellate cells (65). Circulating fibrocytes (bone marrow–derived CD34⁺ mesenchymal cells) can also be recruited and differentiate into α-SMA⁺ (CD34⁻) MyoFbs in the skin, liver, and lung (66).

MyoFb differentiation is primarily induced by paracrine factors, such as TGF-β and platelet-derived growth factor (PDGF), secreted from tumor cells and stromal cells (67). Differentiation of NFs to MyoFbs depends on the activation of the gene encoding α-SMA, ACTA2, which is otherwise inactivated by methylation of the promoter at three CpG islands. TGF-β signaling inhibits DNA methylation and derepresses ACTA2 transcription (68). Furthermore, TGF-β signaling promotes the activity of myocardin, a smooth muscle–specific transcriptional regulator involved in the expression of ACTA2 (69, 70). High expression of ACTA2 increases contractility and focal adhesions (FAs), both of which are hallmarks of MyoFbs (64). In contrast, transdifferentiation of epithelial cells into MyoFbs is driven by a form of TGF-β–induced EMT (type 2 EMT) that is associated with wound healing, wherein epithelial apical-basolateral polarity is converted into the mesenchymal front-rear polarity (71). This change is accomplished by a reduction in the expression of epithelial markers like E-cadherin, the acquisition of a spindle shape, and the increased expression of mesenchymal markers such as vimentin and ACTA2 (72). Type 2 EMT is implicated in various types of tissue and organ fibrosis associated with overactivation of TGF-β signals due to a reduction in SMAD specific E3 ubiquitin protein ligase 2 (SMURF2), a ubiquitin ligase that targets the TGF-β downstream effectors mothers against decapentaplegic homolog 2 (SMAD2) and SMAD3 for degradation (73). Other causative agents of type 2 EMT include hypoxia, collagenous ECM, cytokines (such as TGF-β and FGF2), inflammation, and ROS (72).

Protumor activities of MyoFbs

The abundance of α-SMA⁺ MyoFbs predicts poor prognosis and cancer recurrence (74). Coculturing cancer cells with MyoFbs or with MyoFb-conditioned medium promotes cancer cell invasiveness (75), and cotransplantation of MyoFbs with cancer cells into animals exacerbates tumor growth (76). This pathogenicity is attributed to the fibrogenic effects of MyoFbs, which are amplified by the TGF-β–induced infiltration of MyoFbs into the ECM and their recruitment α-SMA⁻ stromal cells to become α-SMA⁺ MyoFbs (77). MyoFbs also produce large amounts of growth factors, ECM proteins, and ECM-modifying enzymes that together increase the rigidity and stability of the ECM, which both promotes tumor progression and invasion (see below) (8, 62) and compromises the termination of MyoFb activity (60). Moreover, as discussed above, MyoFbs and other CAFs activate specific metabolic pathways that buffer and recycle toxic metabolites produced by cancer cells, thus enabling continued cancer growth (35).

In addition to these roles in promoting tumor growth and progression, MyoFbs play pivotal roles in tumor initiation. Premalignant cells cotransplanted with MyoFbs establish tumors in animals, whereas premalignant cells alone fail to do so (6, 16). This is primarily because MyoFbs help construct the “premalignant niche,” wherein premalignant cells accumulate mutations that favor their transition into malignancy (78-80). Indeed, various types of chronic wounds and damaged tissue populated with MyoFbs can eventually give rise to neoplasms (7, 81), and the presence of MyoFb-like, α-SMA–high cells in dense breast tissue increases breast cancer risk by four- to sixfold (3, 82). It has long been known that α-SMA–high cells that resemble those in the breast cancer stroma are present in the stroma of nonmalignant breast lesions but are virtually absent from normal breast tissues (4). It is thus proposed that the emergence of MyoFbs in nonmalignant tissues indicates a predisposition to tumor initiation (6, 10), concordant with the analogy of tumors to “nonhealing ulcers” (9, 81) and the association of up to 20% of cancers with the chronic inflammation and tissue damage that also favor fibrosis (7, 83). Many types of tissue injuries harbor excessively activated MyoFbs and could develop into malignant scars. These include nonhealing ulcers in diabetic and immunocompromised patients, Marjolin ulcers, surgical incisions, stomach ulcers, ulcerative colitis, bronchitis, and cirrhosis (83-85). In fact, cirrhosis, which is fibrosis of the liver caused by viral infection, metabolic disease, drug toxicity, or alcoholism, has a high chance (20%) of becoming malignant (86). Other causes for MyoFb activation linked to tumor initiation and promotion include chronic metabolic disorders (obesity and diabetes), environmental agents (asbestos and silica), and chemotherapeutic drugs (bleomycin, methotrexate, and cisplatin) (87-90).

The tumor-initiating capabilities of MyoFbs are largely attributed to their production of large amounts of paracrine factors, ECM, and ECM-modifying enzymes (34). These paracrine factors include TGF-β; FGFs; insulin-like growth factors (IGF1 and IGF2); HGF; PDGF; VEGF; stem cell factor; TNF; stromal cell–derived factor 1 (also called CXCL12); and IL-1, IL-6, IL-8, and IL-10 (91). In particular, fibroblasts overexpressing TGF-β and HGF induce tumor formation from premalignant cells cotransplanted into animals (6). This is partly because these factors induce global epigenetic changes within the receiving cells, particularly changes in the expression of genes involved in stemness and tumorigenesis that can drive epithelia to become tumor-initiating cells (92-94). The growth of tumor-initiating cells would also be assisted by MyoFbs metabolizing toxic metabolites produced by proliferating cells (35). Moreover, MyoFbs produce copious amounts of ECM proteins, including collagen I, fibronectin, fibrillin, elastin, proteoglycans, and tenascins (91), which serve not only as the building blocks for the stroma but also as ligands for cell surface receptors promoting cell proliferation and other tumor-initiating events (95). In addition, MyoFbs produce ECM–cross-linking enzymes, such as lysyl oxidase (LOX), LOX-like (LOXL), and transglutaminase (TG) as well as proteases that remodel the ECM, such as MMPs (8, 34, 62, 96, 97). These ECM components and ECM-modifying enzymes together can increase stromal rigidity, which mechanically induces tumor initiation and progression. Stiff ECM with high collagen density alone is sufficient to induce the malignant transformation of cultured normal breast cells (98, 99), demonstrating that the mechanical properties of the fibrotic stroma play a key role in tumor initiation. MMPs produced by MyoFbs could also increase ROS production in the tissue, promoting genomic instability that facilitates tumor initiation (49). Furthermore, the MyoFb-rich stroma compromises antitumor immunity, while also assisting in the development and survival of tumor-initiating cells (100, 101).

PHENOTYPIC INFLUENCE BY THE ECM

The ECM is a network of fibrous proteins embedded in a gel-like substrate made of proteoglycans. The core matrisome of the mammalian ECM includes ~300 proteins secreted from different resident cells that biochemically and mechanically regulate cell and tissue functions (102, 103). It not only supports tissues structurally but also provides or sequesters various signals to dynamically regulate diverse cellular functions, including growth, survival, morphogenesis, and differentiation (104). The ECM is constantly degraded and remodeled in response to growth, disease, and injury (105-107), diversifying its components and functionalities across tissues. Such variations are derived from differences in protein isoforms or homologs, post-translational modifications, ratios of individual components, and turnover rates, resulting in tissue-specific mechanical and biochemical properties (105, 108). As discussed above, MyoFbs have a profound effect on ECM structure and composition, although other cell types, such as adipocytes, macrophages, MSCs, and endothelial cells, also contribute to the ECM (109-113).

Properties of the ECM

The ECM is composed of several distinct families of evolutionarily conserved molecules. These include glycoproteins (such as collagen, laminin, and elastin), which form fibrous macromolecules and interact with cell surface receptors, and proteoglycans (such as heparin sulfate), which are attached to anionic glycosaminoglycans, exert hydrodynamic functions, and form linkages to macromolecules (114). The basement membrane (BM) component of the ECM surrounds the parenchyma and demarcates it from the stroma, whereas the IM constitutes the bulk of the ECM in the body (Fig. 2) (115).

Fig. 2. Properties of the ECM.

The ECM, here illustrated using breast tissue as an example, consists of the interstitial matrix (IM) and the basement membrane (BM). The IM, which constitutes the bulk of the ECM, is produced by resident stromal cells, serves as a structural scaffold, controls the availability of growth factors and other signaling molecules, and imparts mechanical stimuli to cells. The other component of the ECM, the BM, is produced by basal epithelial cells and is more compact and less porous than the IM. In many disease conditions, the IM becomes fibrous and stiff, and the BM is remodeled or compromised.

BM is more compact and less porous than the IM, containing type IV, XV, XVII, and XVIII collagens; laminins; entactin; and proteoglycans (105). BM is synthesized and organized by basal epithelial cells (116); in the mammary gland, it is primarily produced by myoepithelial cells, the outer cells of the bilayered epithelium (117). The deposition of BM proteins, especially laminins, is facilitated by Rho-associated kinase 1 (ROCK-1), which promotes the localization of the polarity protein partitioning-defective 1b (PAR-1b) to the basal surface of the epithelium (118). Deposited laminins bind to cell surface receptors (such as integrins), initiating the self-assembly of polymeric networks and cross-linking of other BM proteins (such as type IV collagen and nidogen) to establish the BM (119). The BM directly contacts the parenchymal cell surface, providing protection against mechanical and chemical stresses, mediating interactions between cells and the IM, and acting as a barrier between tissues. The BM also plays key roles in tissue differentiation, polarity, and homeostasis (120). Some BM components, such as laminins and type IV collagens, serve as reservoirs for endothelium-derived antiangiogenic factors (endostatin, arrestin, canstatin, and tumstatin), contributing to its tumor-suppressive roles (121). During embryogenesis, BMs are reprogrammed by spatiotemporally regulated MMPs to generate anisotropic patterns that mechanically influence tissue and organ morphogenesis (122). In tumors, vascular and epithelial BMs are actively remodeled during angiogenesis and cell invasion, respectively (121). BMs can also be mechanically ruptured by the protrusive force of dividing cells or the compressive force of an expanding epithelial mass, contributing to tumor cell invasion (123).

The IM constitutes a large part of the stroma, containing type I, III, V, VI, VII, and XII collagens as well as fibronectin, elastin, tenascin, glucosaminoglycans, mucopolysaccharides, and proteoglycans. IM proteins are synthesized by the resident stromal cells, in particular, fibroblasts. Fibrillar collagens (type I, II, III, V, and XI) are the most abundant proteins in the IM (124). Newly translated collagen forms triple-helical procollagen in the Golgi apparatus before it is secreted into the extracellular space, where the N and C termini are cleaved by MMPs to release the mature collagen that is cross-linked by LOX to stabilize the fibrillar structure (125). The IM plays roles in tissue hydration, growth factor binding, structurally supporting tissues, and linking neighboring tissues (115). IMs are bathed in interstitial fluid, which transports nutrients, oxygen, and metabolites between vessels and cells. Unlike the BM’s role as a selective tissue barrier, the IM is the site of a tissue’s first response to external cues. For example, TGF-β forms a latent complex that is sequestered in the IM until biochemical or mechanical stimuli trigger its release from the IM, thus initiating fibrogenic signaling (126). In addition, the IM itself mechanically stimulates cells. Increased collagen deposition and cross-linking increase the stiffness of the IM, which induces cell motility and proliferation (105). During tissue and organ morphogenesis, local increases in collagen deposition create mechanical anisotropies that drive cell migration in a particular direction (115).

Mechanical continuum from the ECM to the nucleus

Cells sense the biochemical and mechanical properties of the ECM and, in turn, regulate the expression of genes encoding ECM proteins and ECM-modifying enzymes (106, 107). Such dynamic and reciprocal ECM-nucleus interactions were initially proposed more than 40 years ago (127, 128) and have since been corroborated by the discovery that the ECM-nucleus linkage is not only biochemical but also mechanical through physical continuums that hardwire the ECM to the nucleus (Fig. 3) (129, 130). The viscous and elastic properties of the ECM, defined by Young’s elastic modulus, depend on the composition and organization of its constituents (131). Mechanotransduction transforms these physical properties into biochemical signals that influence cellular functions and tissue specificity. Increased ECM stiffness, in particular, is closely associated with pathological conditions (115).

Fig. 3. Physical continuum between the ECM and chromatin.

The signals received at cell-cell junctions (tight junction, adherens complexes, and desmosomes) and at cell-ECM junctions [focal adhesions (FAs) and hemidesmosomes] are transmitted to the nucleus through microfilaments, microtubules, and intermediate filaments. At the nucleus, these elements of the cytoskeleton are connected to the LINC complex (KASH and SUN) through nesprin. The LINC complex spans the outer (ONM) and inner (INM) nuclear membranes to connect the cytoskeleton to lamin in the nuclear lamina, which associates with chromosome in the nucleoplasm.

Integrins, heterodimeric transmembrane receptors composed of α and β subunits, are the most well-studied cell surface receptors transmitting mechanical ECM signals into cells. The extracellular domain of the α subunit determines ligand specificity, and the cytosolic tail of the β subunit affects a multitude of downstream signaling pathways through its connection to the cytoskeleton (132). Whereas integrins αvβ3 and α3β1 support the formation of FA complexes at the cell-IM interface, integrin α6β4 supports the formation of hemidesmosomes (HDs) at the cell-BM interface (133-135).

FAs are multiprotein complexes involved in cellular contractility and invasiveness. They form when cells contact a stiff substrate and are the primary sites to receive and transmit mechanical cues from the ECM. During cell migration, FAs activate Rho, Rac, and Cdc42 guanosine triphosphatases (GTPases) to trigger cytoskeletal reorganization, generating tension within a cell to form filopodia (134, 136). FA formation is triggered by clustering integrin heterodimers that connect their ECM ligands to the actin cytoskeleton, enabling the bidirectional signaling and force transmission (134, 137). Formation and maturation of FAs involve “inside-out” integrin signaling. At the nascent adhesions, integrin heterodimers are in an inactive state with low binding affinity. Integrin-ECM interaction then leads to the generation of phosphatidylinositol-4,5-bisphosphate [PtdIns (4, 5) P2] that anchors FA kinase (FAK) to the nascent adhesion. FAK, in turn, recruits talin, an adaptor protein linked to actin cytoskeletons. Talin then binds to the cytosolic domain of the integrin β subunit, converting the receptor dimers to an active, high-affinity state. Activated integrins form clusters, further promoting ligand affinity and accumulation to establish mature FAs. Integrin dimers at this stage can receive and transmit extracellular stimuli through “outside-in” signaling (137-140). In stiff ECM, the mechanical tension reinforces FAs by promoting their size and maturation rate to produce respective cytoskeletal activities for force transmission (141-143).

HDs, the other type of integrin-mediated adhesions between epithelial cells and the ECM, link the ECM to the cytokeratin network and are essential for tissue differentiation and polarity. Type I HDs are found in basal stratified epithelia, such as skin, and are composed of integrin α6β4, CD151 (tetraspanin), plectin, BP180, and BP230. The integrin ligand in this case is laminin-332, and plectin and BP230 bind cytokeratin 5 (CK-5) and CK-14. Type II HDs are found in simple epithelia, such as the intestine, and are composed of integrin α6β4 and plectin only. In this case, the integrin ligand is also laminin-332, but plectin binds CK-8 and CK-18 (144). HDs provide epithelial tissues with structural stability and integrity against mechanical stress while also playing a role as mechanosensors (145). Mechanical tension generated during tissue morphogenesis activates the kinase PAK-1, which phosphorylates cytokeratins to induce maturation of HDs to resist mechanical stress (135). The formation of stable HDs also depends on the proteolytic processing of laminin-332 (146). Whereas laminin-332 is processed at the α3 chain and binds integrin α6β4, leading to the formation of stable HDs in normal cells, laminin-332 is often unprocessed at the α3 chain in tumors and instead binds integrin α3β1, which is normally involved in FA formation (147). HDs are rapidly disassembled and reassembled during migration and invasion (148).

FAs and HDs link the ECM to the cytoskeletal network of structural and regulatory proteins that can both generate and transmit forces to control cellular movement and shape, as well as nuclear morphology and integrity (149, 150). The three structural components of the cytoskeleton—microfilaments (MFs), microtubules (MTs), and intermediate filaments (IFs)—regulate one other (151). For example, at lamellipodia, the growth of MTs activates the GTPase Rac to promote polymerization of the actin molecules that constitute MFs (152). In response to external cues, these polymers are rapidly disassembled and reassembled into different cytoskeletal structures (149, 151). Such pliability makes them highly dynamic, enabling changes in cell shape, mobility, and rigidity (151).

The assembly and disassembly of the actin filaments that constitute MFs depend on the Rho family GTPases that regulate the balance between monomeric actin (G-actin) and filamentous actin (F-actin) (153). These dynamic filaments drive cell movements and form cross-linked networks that establish cell shape and rigidity (151). During cell migration, the MF network extends toward the leading edge of the cell, forming membrane protrusions that attach to the substrate through FAs. This triggers myosin-mediated contraction of the actin filaments, generating traction forces that push the cell forward (154). MFs are also critical for mechanosensing and mechanotransduction because tensional stress causes MFs to change their conformations and interactions with regulatory proteins. In response to stiff ECM, Rho-dependent actin polymerization increases, which, in turn, stiffens the cell interior (155). MFs contribute to the membrane skeleton, the cortical actin rim, and stress fibers. MFs of the membrane skeleton, which lies beneath the plasma membrane and controls membrane architecture and cell shape, are cross-linked by spectrin and associate with several proteins (ankyrin, protein 4.1, adducin, and α-catenin) that link them to transmembrane proteins (156). In the cortical actin rim, which is associated with cell-cell and cell-ECM adhesion complexes that strengthen cell adhesions and generate the outward tension that controls cell shape, MFs constitute the cortical bundles that establish adhesion belts linked to adherens junctions and tight junctions. The cortical actin rim is organized by various actin-binding proteins (spectrin, α-actinin, α-catenin, vinculin, filamin, cortactin, Wiskott–Aldrich syndrome protein (WASP), vasodilator-stimulated phosphoprotein (VASP), gelsolin, cofilin, ezrin-radixin-moesin, and heat shock protein 27) (156). Stress fibers, which are linked to FAs, adherens junctions, and the nuclear membrane, are actomyosin bundles in which short F-actins with alternating polarity are cross-linked by α-actinin and bound by vinculin, talin, and filamin. Stress fibers generate inward tension to induce cell contraction and to counteract outward tension generated by the cortical actin rim (156, 157).

MTs are rigid, polymeric hollow tubes that are composed of dimers of α- and β-tubulin and serve as tracks for moving cargoes and as scaffolds for intracellular trafficking, mitosis, and organizing organelles (151). The slow-growing minus ends of MTs are anchored to MT organizing centers (MTOCs, the centrosomes), whereas the fast-growing plus ends extend toward the cell surface (158). The layout of MTs in a cell depends on the context. For example, in differentiating epithelia, centrosomes are localized subapically, orienting the MT minus ends toward the apical surface with the plus ends toward the basal surface, generating apicobasal polarity (159). In migrating fibroblasts, centrosomes are localized between the nucleus and the leading edge, with the plus ends extending toward the leading edge (160). MTs also play roles in mechanosensing and mechanotransduction. Tension-induced deformation of MTs activates Src signaling to restrict MT growth at the plus ends, which helps control organ size and shape (161).

IFs are the major structural components of both the cytosol and nucleus, comprising at least 65 distinct proteins. IFs help establish structural integrity and absorb mechanical stress. Unlike MFs and MTs, which are dynamic polymers of globular proteins, IFs are stable, insoluble filaments composed of long continuous segments of α helices. The assembly of IFs is regulated developmentally and in cell- and organelle-specific manners, as seen for cytokeratins (epithelia), desmin (muscle), vimentin (mesenchymal cells), glial fibrillary acidic protein (GFRAP, glia cells), neurofilaments (neurons), and lamins (nuclear envelope) (162). Cytosolic IFs help stabilize cell shape by contributing to various adhesion complexes (163). In cell-cell adhesions (desmosomes), IFs are anchored to adhesion plaques (desmoplakin) linked to adhesion proteins (desmoglein and desmocollin), connecting the plasma membranes of two adjacent cells through linker proteins (plakoglobin and plakophilin) (164). In cell-ECM adhesions (HDs), IFs are linked to the ECM laminin receptor integrin α6β4 through plectin and BP230 (144).

Nuclear IFs are solely assembled by lamins and constitute the nuclear lamina, a fibrillar network associated with the inner nuclear membrane. The nuclear lamina provides a platform for chromatin organization and gene transcription (165) and connects the nucleus to the cytoskeleton, which provides both structural support for the nucleus and a mechanical continuum transmitting external force from ECM to nucleus (Fig. 3) (165, 166). The linker of nucleoskeleton and cytoskeleton (LINC) complex mechanically connects the nuclear lamina to the cytoskeleton and is composed of the Sad1 and UNC-84 domain proteins SUN1 and SUN2, which are linked to lamins and span the inner nuclear membrane, and the Klarsicht, nuclear anchorage protein 1 (ANC-1), and Syne homology domain (KASH) proteins, which are linked to SUN and span the outer nuclear membrane. Nesprins embedded in the outer nuclear membrane connect LINC to the cytoskeleton by interacting with both KASH and cytoskeletal proteins. Nesprin-1 and nesprin-2 link to MFs, nesprin-3 links to IFs, and nesprin-4 links to MTs (166, 167). In addition to being important for DNA damage repair and chromosome anchoring during meiosis, LINC is also required for maintaining nuclear shape, cell polarization, and migration under mechanical stimuli (167). SUN mediates the connection between LINC and the nuclear lamina by interacting with lamin A. Whereas the amounts of B-type lamins, which assemble into randomly organized filaments and confer elasticity and deformability to the nucleus (167-169), are fairly constant, the amounts of A-type lamins (lamin A and lamin C) vary and positively correlate with tissue stiffness because they increase the rigidity of the nuclear envelope (169). Because of its interaction with LINC, lamin A serves as a molecular shock absorber that protects nuclear architecture from mechanical stress (167). In response to tension, lamin A amounts increase and LINC is stabilized through additional recruitment of the complex proteins and actin filaments, imparting stiffness to the nucleus to resist the force (167, 168). Because lamin A binds to chromosomes, it also links mechanical tension in the cytoskeleton to gene regulation (167).

Nuclear responses to mechanical stress

Although the nucleus is generally 2 to 10 times stiffer than the cytosol, it is elastic and deformable (170). As discussed above, A-type lamins impart stiffness to the nuclear envelope, whereas B-type lamins confer the elasticity and deformability that allow for cell survival, migration, and differentiation (167-169). Chromosomes in the nucleoplasm also contribute to the viscoelasticity of the nucleus: Chromosome condensation during cellular differentiation stiffens the nucleus, whereas chromosome decondensation during mitosis softens the nucleus (170). Thus, the mechanical properties of the nucleus are determined by the elastic, deformable nuclear lamina and the viscoelastic nucleoplasm (170).

Cellular and nuclear deformation by external stresses can induce either local chromatin stretching to absorb the stress or global chromatin decondensation to dissipate the stress. This is mediated by a shift in A-type lamin interactions from highly compact (H3K9me3-marked) to less-compact (H3K27me3-marked) heterochromatin, which activates the mechanosensitive transcriptional regulators megakaryoblastic leukemia 1 (MAL)/serum response factor (SRF) complex, YAP, and TAZ (171). For long-term effects, however, the nucleus eventually increases its deformability, for example, by increasing the lamin B–to–lamin A ratio (171). When a cell is extremely deformed, such as during migration through densely packed ECM fibers, the nucleus becomes extremely compressed, leading to nuclear membrane rupture that causes DNA damage (170, 172).

Pathogenic effects of stiff ECM

There are three major causes of the ECM becoming fibrous and stiff in diseases such as fibrosis and cancer (Fig. 4, A to C). The first is dysregulated matrix biosynthesis and remodeling by MyoFbs. As discussed above, MyoFbs produce copious amounts of fibrogenic factors, ECM proteins, and ECM-modifying enzymes that are required for tissue repair, but excessive amounts of these products increase the rigidity and stability of the ECM (Fig. 4A) (8, 96). The second major cause of ECM stiffening is the increased contractile forces generated by MyoFb α-SMA stress fibers. Stress fibers within cells transmit contractile forces to the ECM through their connections to the integrins in FAs (αvβ5 and αvβ3). This induces the remodeling and stabilization of the local ECM to increase the stiffness (Fig. 4B) (63). MyoFb contraction further stimulates fibrogenic signaling through positive feedback mechanisms, such as liberation of TGF-β from the ECM-bound latent complex (173). Moreover, the increased ECM stiffness increases the traction forces of MyoFbs against the ECM, also triggering the release of TGF-β from the ECM (174). The third major cause of ECM stiffening is advanced glycation end products (AGEs), strong oxidizing agents that induce nonenzymatic collagen cross-linking (175). AGEs are produced by the conjugation of sugars to various biomolecules, such as proteins and nucleic acids, and their amounts increase during aging and in chronic diseases such as diabetes and cardiovascular disease (176). AGEs also induce MyoFb formation from NFs (differentiation) and epithelia (type 2 EMT) (Fig. 4C) (177). In fact, increased MyoFb formation by AGEs partly accounts for higher cancer risks in older people and individuals with chronic disorders (178).

Fig. 4. Three major mechanisms of ECM stiffening.

(A) Cancer cells and stromal MyoFbs secrete large amounts of fibrogenic factors (such as FGF and TGF-β), fibrillar collagen, and ECM cross-linking (LOX) and remodeling (MMP) enzymes. (B) Lockstep model of ECM stiffening by MyoFb contraction. ECM fibers are connected to MyoFb stress fibers through FAs. The contraction of MyoFbs loosens adjacent ECM fibers, which activates local ECM remodeling and stabilizing factors to increase local ECM rigidity. As MyoFbs retract, they stretch the ECM, increasing ECM tension. Repetition of the cycle further increases ECM rigidity. (C) ECM stiffening is triggered by advanced glycation end products (AGEs). AGEs nonenzymatically cross-link collagen, which stiffens the ECM, and stimulate the receptor RAGE, which increases the MyoFb population by both activating NFs and promoting type 2 EMT in epithelia.

Excessive synthesis or cross-linking of collagen I stiffens the ECM and underlies the pathogenesis of many diseases, including cancer, fibrosis, and chronic disorders. Stiff ECM exerts tensional stress that disrupts tissue structures and promotes transformation, metastasis, and cell contractility through various signaling pathways, including integrin β1, FAK, extracellular signal–regulated kinases (ERKs), and RhoA (105). Enzymatic cross-linking of collagen I is catalyzed by enzymes, like LOX and TG, that are increased in metastatic tumors (108, 179). Conversely, nonenzymatic cross-linking of ECM proteins occurs through glycation (180). AGEs not only stiffen the ECM but also activate their cognate receptor, RAGE, which promotes cancer progression by triggering proliferative signaling and inhibiting cell death pathways (181, 182). Furthermore, the mechanical tension of stiff ECM itself plays critical roles in cancer initiation by inducing the transformation of normal cells (183, 184). Stiff ECM also promotes the Warburg effect, a shift to aerobic glycolytic metabolism (98, 185, 186), and nuclear translocation of YAP and TAZ, which stimulate the expression of growth-associated genes (168, 187). The tensional force of stiff ECM that is ultimately transmitted to the nucleus through LINC (129, 130, 168) also leads to nuclear stiffening and genomic instability (150, 168).

In addition to highly cross-linked ECM fibers, tumor stroma also contain abnormally high amounts of interstitial fluid due to leaky vasculature (188) and high amounts of the glycosaminoglycan HA (189), both of which contribute to increased interstitial pressure. The resulting pressure gradient between tumors and normal tissues creates fluid flow, activating integrin signals that promote tumor cell invasion (188).

Stiff ECM drives hallmarks of cancer

Experimental evidence attests to the notion that normal or nonmalignant cells can acquire a malignant phenotype after encountering stiff ECM. Culturing normal mammary epithelia in high-density collagen matrix alone helps them acquire a proliferative and invasive phenotype (98, 99), and crossing a mouse strain that develops spontaneous mammary tumors with a strain overexpressing collagen I greatly exacerbates tumor development (10). These effects are largely due to the increased mechanical—not biochemical—stimuli of stiff ECM, as demonstrated by in vitro studies using inert synthetic matrices (such as polyacrylamide hydrogel) instead of bioactive collagen (98, 184). To understand how the mechanical stimulus of stiff ECM contributes to tumor initiation, several studies describe its roles in driving different hallmarks of cancer (95, 190), the processes necessary for the transformation of normal human cells to malignant tumors (95, 190). Here, we summarize some of the insights into how cancer hallmarks are ascribed to stiff ECM (Fig. 5, A and B).

Fig. 5. Mechanisms of stiff ECM–induced tumor initiation.

(A) Fibrogenic signals increase the numbers of MyoFb-like cells and cause ECM stiffening, which eventually leads to malignant transformation of the parenchyma characterized by sustained proliferation, evasion of growth suppression, resistance to cell death, and replicative immortality. (B) The tension of the stiff ECM is transmitted to the nuclear membrane of parenchymal cells. This may deform or even rupture the nuclear membrane, leading to genome instability through chromosome segregation errors, DNA damage, defective damage repair, and/or reactivation of transposons.

Sustained proliferative signaling and the evasion of growth suppression are the major consequences of stiff ECM. Normal cells embedded in compliant matrices differentiate and maintain homeostatic tissue architecture and functions, with their proliferation controlled by mitogenic signals from ligand-activated growth factor receptors that promote cell cycle progression and contact inhibition mediated by the cell-cell junction proteins E-cadherin and zonula occludens-1 (ZO-1), which induces G₁ arrest at high cell density. In cancer cells or normal cells surrounded by stiff ECM, various mechanisms can block negative feedback of growth factor signaling, cause mitogenic signaling to become autonomous, and/or override or bypass contact inhibition. The internalization of cell surface receptors through clathrin-mediated endocytosis is a negative feedback mechanism that limits growth factor signaling (191). However, stiff ECM increases the affinity of integrin αvβ5 for its ECM ligand, forming long-lived plaques on the cell surface that are resistant to endocytosis (192). Stiff ECM also activates FAK-Rac signaling, which promotes mitogen-independent cell cycle progression by transcriptionally inducing cyclin D1 (193), and induces translocation of the mechanosensitive transcription regulator YAP into the nucleus, where it activates the expression of genes encoding cell cycle promoters, such as E2F1 (194). Furthermore, stiff ECM is also associated with the mislocalization of E-cadherin and ZO-1 and greatly lowers the threshold of EGF necessary to counteract contact inhibition (195). Induction of the DDR upon genotoxic stress is another major growth suppressor that is inhibited by stiff ECM. Under normal conditions, DNA damage induces cell cycle arrest through tumor suppressors such as Rb, p53, breast cancer 1 (BRCA1), and phosphatase and tensin homolog (PTEN), which enables to DNA repair; however, the expression and function of these proteins are reduced by stiff ECM, thus impairing cell cycle arrest upon DNA damage (193, 196-198).

Resistance to apoptosis is another effect of stiff ECM that is closely associated with tumor initiation. The survival of normal adherent cells requires their attachment to the ECM through integrins, and detachment induces a form of caspase-mediated apoptosis called anoikis. Anoikis is essential for the formation and maintenance of normal tissues, such as lumen formation in alveoli, and protects against dysplasia by preventing anchorage-dependent cells from growing outside their normal setting (199). Cancer cells acquire resistance to anoikis through various mechanisms, allowing them to metastasize to different environments. One way is to induce an “integrin switch” that modulates integrin profiles and therefore sensitivity to ECM detachment (200). The second way is to constitutively activate downstream mediators of integrin signaling, such as phosphoinositide 3-kinase (PI3K) and AKT (201). The third way is to induce EMT, which, in turn, increases the production of antiapoptotic (Bcl-2 family) and prosurvival proteins (PI3K and Akt) (201). A fourth way is to directly increase the production of antiapoptotic proteins, such as survivin (202). All these mechanisms are induced in normal cells encountering stiff ECM (95). Furthermore, resistance to apoptosis owing to stiff ECM is a major cause for reduced efficacy of antitumor therapies. For example, stiff ECM inhibits doxorubicin-induced, p53-mediated apoptosis in breast tumors (203).

Replicative immortality is another consequence of stiff ECM that may contribute to tumorigenesis. Immortality is largely attributed to the activation of telomere maintenance mechanisms (204). Telomeres, repetitive DNA sequences at the ends of chromosomes, progressively shorten as normal cells divide because of incomplete DNA synthesis or damage, leading to replicative senescence and aging. However, a rare cell that constitutively produces telomerase, the enzyme that repairs the shortened telomere, can divide indefinitely (205). Telomerase promotes EMT, is essential for stem cell self-renewal and pluripotency (206), and is closely associated with stiff ECM (190). For example, telomerase is increased in idiopathic pulmonary fibrosis, which is characterized by stiff ECM, abundant MyoFbs, and hypertrophy and is directly associated with increased risk of developing lung cancer (207).

Last, genome instability and mutations, early events in tumorigenesis, can also result from stiff ECM. In fact, several critical mutations found in transformed cells are already present in nonmalignant cells from benign fibrotic breast tissue biopsied from women who later developed breast cancer (208). Genome instability arises from many different causes, including chromosome segregation errors, DNA damage, defective DNA damage repair, and increased activity of transposable elements. Faithful chromosome segregation is essential for genome stability and survival of the organism. However, the fidelity of chromosome segregation depends on proper tissue architecture, which can be disturbed by stiff ECM (184, 209). Cells confined between the rigid ECM fibers can undergo asymmetric multidaughter mitosis that generates aneuploid cells (210). Migration of cells through densely packed, stiff ECM also compresses the nuclei, which can result in nuclear deformation or nuclear membrane rupture that increases replication stress and DNA damage, thereby compromising genomic integrity (170, 172). Compression of the nucleus by stiff ECM can also inhibit major DNA repair proteins, such as BRCA1 and p53BP1, from entering the nucleus (211). Thus, the rate of incurred mutations is predicted to be directly proportional to the frequency and magnitude of mechanical stress exerted on the nuclei (212). Cumulative stresses caused by the chronic disruption of nuclear architecture by mechanical tension from stiff ECM could alter chromatin structure and function, potentially leading to the dysregulation of oncogenes or tumor suppressors or to the reactivation of transposable elements that are otherwise repressed, which would promote mutagenesis and genomic instability (170).

FIBROBLASTS IN RESISTANCE TO CANCER THERAPIES

Mechanisms of CAF-induced therapeutic resistance

In addition to promoting tumor initiation and progression, MyoFbs and other CAFs help induce therapeutic resistance (42). Chemotherapy itself can also stimulate CAF differentiation (213). There are several major mechanisms through which CAFs promote drug resistance. First, CAF-derived paracrine factors can not only induce epigenetic changes in stemness genes but also trigger the expression of stem factors (such as Wnt and HGF) within epithelia, driving them to become tumor-initiating cells or cancer stem cells (92-94, 214), which are inherently resistant to chemotherapeutic agents due to having an EMT phenotype and the ability to inactivate drugs (215). Second, increases in mechanical stress within a tumor, driven by CAF-derived ECM proteins and ECM-modifying enzymes, could collapse blood vessels, which not only impairs drug delivery but also causes hypoxia and subsequent acidosis that reduces the efficacy of anticancer drugs (216, 217).

A third mechanism is through the activation of survival pathways by nuclear factor κB (NF-κB) signaling. Most anticancer drugs activate NF-κB signaling in CAFs, which, in turn, increases the release of paracrine factors (such as WNT16B, secreted frizzled related protein 2 (SFRP2), IL-6, IL-17A, and HGF) that induce survival signaling in cancer cells through ERK1/2 and PI3K–ATK–NF-κB pathways (218, 219). This drives the expression of antiapoptotic proteins, such as B-cell lymphoma-2 (BCL-2), and the efflux pump multidrug-resistant protein 1 (220). In aggressive cancers, NF-κB is constitutively active in both CAFs and tumor cells (221). Fourth, CAF-secreted MMPs promote a proinflammatory signature (COX-2, NF-κB, and HIF-1 expression) in cancer cells, leading to the production of ROS, which induces EMT that, in turn, promotes invasiveness and stemness and lowers the efficacy of certain anticancer drugs (222-224). Fifth, under drug treatment, both CAFs and tumor cells undergo a metabolic switch to glycolysis. This induces the expression of antiapoptotic proteins and drug efflux pumps, EMT, autophagy, and acidification, leading to multidrug resistance (225, 226). Furthermore, a glycolytic metabolite, lactate, serves as a signaling molecule instructing CAFs to secrete HGF, which also induces drug resistance (225, 227).

Last, CAFs suppress antitumor immunity and compromise immunotherapy through multiple mechanisms. A subset of CAFs (FAP⁺ CAFs) facilitate the formation of protumor myeloid cells, including M2-type macrophages and MDSCs that inhibit CD8⁺ cytotoxic T cells (228, 229). Another subpopulation of CAFs (CAF-S1 cells; see Table 2) promotes the formation of regulatory T cells that inhibit both CD4⁺ helper and CD8⁺ cytotoxic T cells (230). CAFs also produce paracrine factors that inhibit natural killer cells [prostaglandin E2 (PGE2) and indoleamine 2,3-dioxygenase (IDO)] and dendritic cells (IL-6), as well as immune checkpoint proteins [programmed death-ligand 1 (PDL-1) and PDL-2] that inhibit T cell activation (229).

Epithelial-mesenchymal plasticity of cancer cells

In addition to assisting cancer cells in acquiring resistance to chemo- and immunotherapies, CAFs may also play direct roles in the resistance. Although CAFs have long been considered as mere assistants of cancer cells, CAFs could in fact be cancer cells that have camouflaged to evade anticancer immunity and treatments. Cancer cells often undergo EMT to acquire therapy resistance (222-224, 231) but may revert to epithelia through MET (mesenchymal-epithelial transition) after the drug is withdrawn (232, 233). These transitions can cause the cancer cells to adopt mixed epithelial-mesenchymal features, a phenomenon termed “epithelial-mesenchymal plasticity,” which is associated with enhanced aggressiveness and therapeutic resistance (234). One such clinical manifestation is metaplastic breast cancer (MpBC), a rare, aggressive cancer accounting for less than 1% of the total breast cancer cases. MpBC carcinomas are composed of multiple cell populations that have undergone metaplastic differentiation into nonglandular cells, such as spindle-shaped cells. These cells produce high amounts of EMT markers and exhibit stem cell (tumor-initiating cell)–like features, including drug resistance (235). MpBC, being positive for the epithelial marker cytokeratin, is thought to arise from the epithelial compartment through metaplasia or EMT as cells adapt to tissue-damaging stimuli (235-237). Thus, such transformation could be triggered by drug treatment for a less aggressive epithelial cancer. It is plausible that this type of aggravated cancer progression and the resulting drug resistance would be blocked by a combinatory therapy of an anticancer drug and an EMT inhibitor or a drug targeted to tumorigenic stroma (238-240).

CONCLUSIONS

Dynamic reciprocity between the ECM and nuclear matrix was initially proposed by M. Bissell and P. Bornstein 40 years ago. In this model, the ECM exerts mechanical and biochemical influences on the nucleus through cell surface receptors, leading to changes in gene expression that, in turn, affect the ECM, resulting in a cycle of dynamic cell-ECM interactions (127, 128). Their provocative hypothesis was finally validated 15 years later by the pioneering studies of D. Ingber, substantiating the physical continuum that hardwires the ECM to the nucleus (129, 130). These decades of ground-breaking studies provoked a paradigm shift in the fields of life sciences and medicine, eventually giving forth the previously unknown concepts of tissue microenvironment and mechanobiology playing pivotal roles in regulating tissue and organ homeostasis and disease pathogenesis (241).

These concepts have reshaped cancer research and treatment based on the well-established fact that most tumors harbor stiff, desmoplastic stroma. The fibrous tumor stroma was initially thought to be the mere consequence of malignant progression; however, cumulative evidence has shown that stiff ECM also drives tumor initiation. This knowledge has pushed forward the development of new cancer therapeutics targeting the stiff stroma. Most of these approaches aim to ameliorate the stiffness of tumor stroma with inhibitors of fibrogenic factors (such as TGF-β), ECM cross-linking enzymes (such as LOX and TG), ECM remodeling enzymes (MMPs), or collagen biosynthesis enzymes (such as prolyl hydroxylase) (242). Another method uses an N-terminal α-SMA peptide that inhibits MyoFb contraction or inhibitors of molecules involved in signal transduction induced by stiff ECM (such as integrins, FAK, RhoA-ROCK, and YAP/TAZ) (242). There also have been new technologies that deliver or activate chemotherapeutics specifically in stiff ECM (243, 244). These new therapeutics, however, have, so far, only been used for cancer treatment not for prevention. We speculate that combining a mechanoresponsive delivery or activation system with a mitigator of stromal stiffness could possibly be applied to preventing cancers associated with tissue fibrosis or chronic wounds. Such an exciting possibility warrants further investigation.