The Matrix Revolution: Matricellular Proteins and Restructuring of the Cancer Microenvironment Free

The behavior of individual cells is influenced by a plethora of signals originating from the surrounding microenvironment, which includes the extracellular matrix (ECM). Previously regarded as merely a static scaffold for cell/tissue organization, the ECM is now viewed as a critical niche contributing to the regulation of cellular survival, proliferation, and migration. This realization has positioned the ECM at the center stage of normal physiologic processes such as development, tissue homeostasis, and tissue remodeling.

The dynamic nature of ECM signaling is determined by a secreted subset of nonstructural matricellular proteins (MCP; ref. 1), in contrast to the structural roles of “classical” ECM proteins such as collagen and fibronectin (2). MCP functional versatility is achieved by its multiple domains that either (i) bind ECM proteins and/or cell surface receptors, (ii) bind and regulate the activity or accessibility of extracellular signaling molecules such as growth factors, proteases, chemokines, and cytokines, or (iii) mediate intrinsic enzymatic activities to precisely orchestrate the assembly, degradation, and organization of the ECM. MCPs are tightly controlled, with expression promptly occurring in context-specific scenarios. Typically, they are highly expressed during early development, ultimately subsiding in adult tissues under physiologic conditions. However, transient reexpression is observed during injury repair, and can also be sustained in chronic pathologies such as cancer (2–7). Indeed, chronic unscheduled expression of various MCPs, either by tumor cells or the surrounding stromal cells (8), leads to abnormal ECM remodeling and stimulation of mitogenic pathways essential for cancer progression. This may underlie the correlation between the upregulation of many MCPs and poor prognosis in cancer patients (9) and, moreover, provide rationale for exploring the utility of MCPs as cancer biomarkers and therapeutic targets.

This review will focus on the burgeoning roles of the MCP families SPARC, CCN, SIBLING, tenascin, and Gla-containing proteins in both cancer development, and detection and treatment. Certainly, members of these particular families are aberrantly expressed in various tumor types, and moreover exhibit biochemical, biomechanical, and metastatic properties influencing cancer progression.

The ever-growing number of newly discovered MCPs has necessitated their classification into families. Members are grouped on the basis of shared domains, which in turn reflect the functional diversity between families.

The SPARC protein (secreted protein acidic and rich in cysteine; hereafter alternative protein names are included in parentheses; BM40, osteonectin), one of the original MCPs to be characterized, is considered prototypical due to its simple structure and rich functionality. The subsequent discovery of other MCPs with structural similarity revealed a broader family of SPARC-related proteins (10). Such SPARC family members share follistatin-like and extracellular calcium-binding (EC) domains, and are classified into five distinct groups based on sequence homology of their EC domains (10): SPARCs, SPARCL1, SMOCs, SPOCKs, and follistatin-like protein-1 (FSTL1). SPARC family members were shown to regulate ECM assembly and deposition, influence cytokine activity, inhibit cell adhesion and cell-cycle progression, regulate cell differentiation, and activate matrix metalloproteinases (MMP; ref. 10). While most SPARC members exhibit ubiquitous expression throughout early development, in adults, expression is largely limited to tissues that are diseased or undergoing wound repair/remodeling.

The vertebrate CCN (centralized coordination network) family is composed of six homologous cysteine-rich members (11): CCN1 (CYR61), CCN2 (CTGF), CCN3 (NOV), CCN4 (WISP-1), CCN5 (WISP-2), and CCN6 (WISP-3). Each is comprised of an N-terminal secretory peptide and four functional domains: insulin-like growth factor-binding protein domain (IGFBP), Von Willebrand factor type C domain (VWC), thrombospondin type-1 repeat module (TSR), and carboxy-terminal cysteine-knot (CT) motif (11). In response to tissue remodeling, CCN proteins are expressed principally in mesenchymal cells during development and in connective tissue pathologies (12). The postnatal role of CCN proteins is known for promoting collagen stability or organization (13).

Tenascins (TN) comprise a family of four large ECM glycoproteins, that is, TNC, -R, -W, and -X, which exist as either trimers or hexamers (14). Tenascins share a characteristic modular structure composed of tandem EGF-like domains, fibronectin-type III domains, and a C-terminal fibrinogen-related domain (FReD). Consequently, tenascins share functions in modulating cellular responses to the ECM and growth factors, specifically regulating growth, differentiation, adhesion, and migration during tissue remodeling events (15). However, each member has distinct spatial and temporal expression. TNC expression is typically present in all organs during fetal development and mechanical stress, whereas TN-W expression is restricted to developing/remodeling bone and certain stem cell niches (14). TN-R is expressed exclusively in the developing and adult nervous system, while TN-X represents a constitutive ECM component of most connective tissues, being hardly influenced by external factors (14).

The SIBLING (small integrin-binding ligand N-linked glycoprotein) family includes bone sialoprotein (BSP), osteopontin (SPP1, also known as OPN), dentin sialophosphoprotein (DSPP), matrix extracellular phosphoglycoprotein (MEPE), and dentin matrix protein-1 (DMP1). These proteins are primarily implicated in bone morphogenesis and biomineralization, and were thus thought to be exclusively localized to mineralized tissue such as bone and teeth (16). However, apart from these traditional functions, SIBLING members were also shown to influence cellular proliferation/survival pathways, collagen fibrillogenesis, MMPs activities, and response to injury (17–20).

The Gla-protein family members contain vitamin K–dependent γ-carboxyglutamic acid residues (21), which have high affinity for calcium ions, thus conferring important roles in coagulation and bone homeostasis (22). Among the 17 Gla-protein members, periostin (POSTN) and matrix Gla-protein (MGP) are known to affect ECM cross-linking and various cellular behaviors, such as migration, adhesion, and proliferation in epithelial, endothelial, fibroblast, osteoblast, and myocyte cells (23–27). POSTN is expressed in osteoblast, mesangial, fibroblast, mesenchymal, and vascular smooth muscle cells (22), while MGP is typically secreted and localized in the surrounding ECM of chondrocytes or endothelial cells (28).

Considering that MCP expression is context dependent, MCP-knockout mouse models generally lack any postnatal phenotype unless challenged by injury or disease, in which case they exhibit an impaired yet subtle response [see references for further details: SPARC family (29–34); CCN family (11, 35); tenascin family (36–38); SIBLING family (39–42); and POSTN (43–45)]. However, some MCP mouse knockouts are characterized by severe complications. For example, FSTL1- and CCN2-null mice die shortly after birth, while CCN1 and CCN5 whole-body knockouts are embryonic lethal, showing that these proteins are essential for development (11, 46, 47). As for MGP-knockout mice, they show severe vascular calcification, arteriovenous malformation, and craniofacial anomalies, and die within 8 weeks after birth (48–50).

MCP overexpression is characteristic of tissue remodeling processes, including those occurring during carcinogenesis, as opposed to low/undetectable levels in normal tissue. Tumor cells and the surrounding activated stromal cells are the major cell types that aberrantly secrete MCPs into the tumor microenvironment, in turn promoting cancer development (5, 51). Nonetheless, we note there are certain cases where MCP expression has been shown to oppose cancer development (51, 52).

SPARC protein is highly expressed in cancer cells and the stroma of certain cancers, including glioma, breast, and cervical melanoma (53–56), where it exhibits oncogenic roles in cell growth, invasion, and apoptosis. Interestingly, SPARC has also been associated with tumor suppression by influencing these same processes (57). This discrepancy might be explained by cancer type and stage, and/or the concentration of SPARC in the tumor microenvironment (57). Like SPARC, the role of FSTL1 in carcinogenesis has generated significant controversy. Endometrial and ovarian cancers exhibit low FSTL1 levels; moreover, ectopic FSTL1 expression exerts antineoplastic activity by inducing apoptosis (58). Among SPOCK isoforms (SPOCK1–3), SPOCK1 is upregulated in different tumor types, and its expression positively correlates with invasive/metastatic potential and hence poor prognosis (59–61). However, in brain tumors, expression of all SPOCK family members decreases with increasing tumor grade (62). SMOC2 was shown to be upregulated in hepatocellular, endometrial, and colorectal cancers where it modulated proliferation, chemoresistance, and metastasis, respectively (63–65). Very little is known regarding any role for SMOC1 in carcinogenesis, although its expression is increased in brain tumors, where it interacts with TNC to counteract the chemo-attractive effect of the latter on glioma cells in vitro (66).

Among the CCN family, CCN1 and CCN2 are the most studied in cancer (11). Specifically, CCN1 expression is elevated in many tumor types including brain, breast, prostate, and pancreas (67–70); similarly, CCN2 upregulation is implicated in proliferation, apoptosis, and migration for numerous cancers (71), including gastric (72), pancreatic (73), melanoma (74, 75), and breast (76). In addition to cancer cells, a potential origin of these MCPs may be cancer-associated fibroblasts (CAF; ref. 77), and indeed this cell type was shown to be the source of CCN1/CCN2 expression in murine models of skin cancer (78, 79). Although unscheduled expression of CCN1 and CCN2 are generally associated with tumor promotion, in some cases these proteins were reported to inhibit cancer development (80, 81). Like CCN1/2, CCN3, and CCN4 exhibit a mixture of pro- versus antitumorigenic effects, whereas CCN5 and CCN6 are predominantly regarded as tumor suppressors (11, 82).

Each tenascin family member differs substantially in spatial (tissue specificity) and temporal expression patterns (14). In the case of TNC and TN-W, de novo expression is prominent in tumors versus healthy tissue, where they promote tumor progression on multiple levels, that is, proliferation, invasion, metastasis, and angiogenesis. TNC is recovered in the stroma of most solid cancers, while TN-W is primarily restricted to brain, colon, kidney, and lung cancers (14). In contrast, TN-R and TN-X are constitutively expressed and largely unaffected by tumorigenic signals, that is, to date have not been reported to play a substantial role in carcinogenesis (14).

Among SIBLING proteins, SPP1 and BSP have been the most extensively studied in the context of cancer (16). Consistent with their roles in osteogenesis, SPP1 and BSP have been implicated in bone malignancy (16). However, while these proteins were initially thought to be expressed only during bone morphogenesis, both were subsequently shown to be broadly expressed in human epithelial carcinomas, including but not limited to breast (83, 84), lung (85, 86), prostate (87), liver (88), pancreas (89), and colon (87, 90), where their pathophysiologic roles have recently been thoroughly reviewed (16, 91). Furthermore, CAFs have been shown to produce and secrete SPP1, which contributes to melanoma tumor growth (92).

POSTN, the most well-characterized Gla-protein family member, was shown to be a major determinant in proliferation for a number of aggressive, advanced solid tumors with poor prognosis (22). MGP is much less understood than POSTN, but is gradually emerging as a determinant in cancer progression, exhibiting increased expression in colorectal, glioblastoma, breast, cervical, osteosarcoma, and skin cancers with unfavorable prognosis (93–97).

In general, MCPs are capable of regulating a variety of mechanisms necessary for tumorigenesis, such as survival, proliferation, migration, matrix stiffness, and development of a signal reservoir and metastatic microenvironment. These versatile functions depend on the diverse biochemical, biomechanical, and metastatic niche effects induced by MCPs (Fig. 1), as discussed in more detail immediately below.

Much evidence has shown that MCPs possess biochemical properties essential for regulating various cellular behaviors, including ones implicated in tumor development. These properties mainly pertain to the ability of MCPs to activate a number of cell surface receptors and elicit their downstream signaling (Fig. 1A). Most MCPs are well-known to directly bind integrins, which are αβ heterodimers composed of 18 α subunits and 8 β subunits (98). Integrins are commonly bound by members of the SPARC, CCN, SIBLING, tenascin, and Gla-protein families (99–101), each bound to varying heterodimer combinations. Other than integrins, members of CCN and tenascin families can also bind syndecans, while SPP1 is reported to also bind CD44 receptors (99, 100). In addition, MCPs can act indirectly by binding a variety of ligands (i.e., growth factors and cytokines), thereby affecting ligand distribution and accessibility, and/or coactivating or inhibiting their function (101).

SPARC has been reported to mediate a variety of signaling pathways. For example, SPARC can bind directly to integrin receptors (αvβ1, αvβ3, and αvβ5), resulting in the activation of the proximal intracellular kinases Akt, focal adhesion kinase (FAK), and integrin-linked kinase (ILK; refs. 102–105). These kinases were associated with SPARC-mediated invasion and survival of glioma cells (105). SPARC may also directly interact with the TGFβ1 receptor to mediate Smad signaling, as shown in lung cancer cells (106). Recently, SPARC was reported to bind TGFβ1 to regulate its deposition in the ECM (107). In addition, SPARC may bind other growth factors but with unknown effects (108, 109). Interestingly, SPOCK1 was identified as a downstream target of TGFβ1 and a key player in lung cancer metastasis and proliferation (110), as well as in antiapoptosis via activation of the PI3K/Akt pathway (111, 112). SMOC2 acts to maintain ILK activity during G₁-phase, which in turn influences cell-cycle progression by modulating cyclin D1 expression and DNA synthesis (113). This possibly involves ILK interaction with integrin β1 and β3 cytoplasmic domains, which also leads to inhibition of anoikis and apoptosis through activation of PI3K/Akt signaling (114). Studies by Maier and colleagues suggested that SMOC2 can bind directly to integrins αVβ1 and αVβ6 (115), consistent with recent data showing that SMOC2 binds integrin β1 to activate FAK in kidney fibroblasts (29).

CCNs act through multiple mechanisms to regulate a plethora of dynamic cellular processes (11, 101, 116). In particular, these proteins activate ILK/Akt, MAPK, and associated growth-promoting pathways in cancer, with each CCN member exerting distinct effects and temporal expression profiles. For example, CCN1 signals through integrin αVβ3/Sonic hedgehog to promote motility in vitro and tumorigenic growth in vivo (117), as well as integrin α6β1-mediated invasion (118), in pancreatic cancer. In glioma, CCN1 overexpression enhances tumorigenicity through integrin αVβ3- and αVβ1-linked ILK-mediated activation of Akt, β-catenin-TCF/Lef, and associated survival and proliferation pathways (119). In breast cancer cells, CCN1 can promote (i) resistance to anoikis, partly via integrin β1 (120), as well as (ii) proliferation, survival, and apoptosis resistance through the αvβ3-activated ERK1/2 pathway (121). Similar to CCN1, ectopic expression of CCN2 (i) promotes migration and angiogenesis (122), and (ii) confers apoptosis resistance through integrin αvβ3/ERK1/2 upregulation of antiapoptotic Bcl-xL and cIAP (76) in breast cancer cells. Although most CCNs act primarily through binding various integrin heterodimer combinations, they also bind several other receptors (11, 101, 123, 124), for example syndecan-4 and Notch in the case of CCN1/2 and CCN3, respectively. Interestingly, CCN proteins may be activated by proteolytic cleavage (125, 126).

The opposing effects of CCN3 and CCN4 in different cancers raise the question of which biochemical pathways are responsible for their signaling diversity. In colorectal cancer cells, CCN3 inhibits survival by regulating caspase-3/-8 while inhibiting JNK-mediated migration (127). On the contrary, CCN3 promotes osteoclastogenesis through the FAK/Akt/p38/NF-κB pathway (128). CCN4 also promotes FAK and p38 signaling through αvβ1 integrin in prostate cancer cells; however, this pathway specifically induces migration and vascular cell adhesion molecule-1 (VCAM-1) expression by downregulating miR-126 (129). Furthermore, osteoblast-derived CCN4 plays a key role in prostate cancer cell adhesion to bone through VCAM-1/integrin α4β1 (130). CCN4 also promotes lymphangiogenesis in oral squamous cell carcinoma (SCC) via integrin αvβ3/Akt signaling and upregulation of VEGFC expression, as well as promotes integrin αvβ3/FAK/JNK signaling to induce VEGFA activation of angiogenesis in osteosarcoma cells (131, 132). Conversely, CCN4 inhibits migration in melanoma and lung cancer cells by inactivating the family of Rho-like GTPases (133, 134).

TNC has been shown to stimulate proliferation and survival in a variety of cancers by activating several pathways downstream of integrins and syndecans (135), including integrin α9β1 activation of Akt and MAPK (136) and αvβ3 activation of FAK and paxillin (137). However, a recent study showed that TNC signaling through integrin α2β1, but not α9β1 or αvβ3, induced autocrine growth in brain tumor cells (138). Through an indirect mechanism of tumorigenesis, TNC is able to compete with syndecan-4 binding to fibronectin, thereby interfere with fibronectin inhibition of proliferation (139). Instead, the FReD domain of TN-X was reported to convert latent TGFβ1 into its biologically active form to indirectly control mesenchymal differentiation (140).

POSTN is primarily known for binding integrins αvβ3 and αvβ5 to elicit activation of FAK/JNK and PI3-K/Akt signaling pathways controlling cell proliferation, survival, or migration in various cancers (141–143). POSTN may also signal through EGFR to influence migration in esophageal SCC (144), potentially through cross-talk with integrin αvβ5. Unlike POSTN, little is known regarding the mechanism of MGP in cancer development, although the latter can influence the TGFβ superfamily, including activation of TGFβ1 receptor and inhibition of the bone morphogenetic proteins BMP-2 and BMP-4 (27, 145).

The SIBLING family members BSP and SPP1 exhibit similar activities in cancer development. BSP supports adhesion, proliferation, and migration through αvβ3 and αvβ5, and the prometastatic activity of TGFβ1 in breast cancer cells (146, 147). SPP1 can interact with several integrin receptors (αvβ1, 3, and 5, α8β1, α9β1 and 4, and α4β1) to regulate cell proliferation, angiogenesis, adhesion, and migration (116, 148). SPP1 can also signal through CD44 (149) to activate HIF2α-induced stemness in hepatocellular carcinoma and glioblastoma cells (150, 151), and Akt-mediated cell survival in mesothelioma and colorectal cancer cells (152, 153).

Remodeling of the ECM is an integral process in cancer development that accommodates the structural architecture of the tumor and provides necessary physical changes such as increased matrix and tissue stiffness to promote and sustain neoplastic transformation (154). Mechanotransduction is a process in which perturbations in ECM mechanical stiffness are transduced into biochemical signals. ECM stiffness can communicate with cells through mechano-responsive integrins (98). In a normal setting, the ECM forms a structural microenvironment of relaxed nonoriented fibrils that exerts homeostatic stiffness on embedded cells. In cancer, disruption of this local ECM structure can occur through MCP-mediated remodeling (5, 8, 155–157), which results in structures that are often stiffer, more highly linearized, and have a different orientation relative to normal stroma (7). In response to this, matrix bound integrin structures convert these physical mechanical signals into conventional integrin biochemical signals to influence survival, proliferation, and growth (158). Moreover, integrin and their associated intracellular cytoskeleton mature into reinforced focal adhesions and stress fibers, respectively, to compensate for changes in ECM stiffness (Fig. 1B). Stress fibers are formed from the bundling of actin, and generate a counter-force, both of which are regulated by phosphoactivating myosin through the stiffness-induced integrin–RhoA–ROCK axis (159). Some of the biomechanical processes regulated by MCPs that affect ECM stiffness include increased matrix organization, collagen cross-linking, and deregulation of enzymatic activity.

SPARC is well-known to be implicated in rearranging the matrix through collagen cross-linking. SPARC binds to several fibrillar collagens (I, II, III, and V) as well as to collagen IV, a prominent constituent of basement membranes (160, 161), and is critical for organization of collagenous ECMs. SPARC-knockout mice manifest significant changes in collagen fibril morphology, as well as a substantial decrease in adult tissue concentrations of collagen (32). SPARC also influences the response of host tissue to implanted tumor cells and a lack of endogenous SPARC engenders decreased capacity to encapsulate the tumor, as well as a reduction in the deposition of collagen (162). SPARC exerts at least two roles in collagen fibril assembly, that is, by modulating interactions of collagen with cell surface receptors and directly regulating collagen incorporation into fibrils (163). Loss of SPARC also disrupts the homeostasis of basement membranes and alters tissue biomechanics and physiologic function (164). Finally, SPARC can act as an extracellular chaperone for collagens that enhance the tumorigenic environment (164–166).

In a recent study, TNC significantly colocalized with aligned collagen fibers in patients with breast cancer, compared with the wavy and randomly organized layout of collagen (167) typically observed in normal tissue (168). TNC contains multiple ECM-binding partners, including collagen; however, its involvement in collagen alignment may be mediated through binding to fibronectin, which serves to direct collagen organization (169–171). Similarly, POSTN plays a mechanistic role in intermatrix interactions through formation of a POSTN–BMP-1–LOX complex, where BMP-1 promotes LOX activity for collagen cross-linking (172, 173). In fact, POSTN-knockout animal models exhibit aberrant collagen fibrillogenesis (174). Furthermore, the mechanotransduction pathways of both ROCK in SCC and the transcription factor TWIST from various mechanical stress models are known to increase POSTN deposition (24, 175). The POSTN family member MGP was recently shown to be incorporated into cross-linked multimers of fibronectin, which enhanced cancer cell attachment to fibronectin (23). As for the CCN family, recent studies have shown CCN1, CCN2, and CCN4 to promote alignment and stability of collagen fibers (13, 157, 176).

The matrix environment needs to achieve a level of plasticity for cellular displacement during metastasis. To disseminate, cancer cells require a local ECM niche to support cellular differentiation and intravasation, and an ECM at the secondary metastatic site to permit invasion and colonization (Fig. 1C). There are various ways in which MCPs are able to establish a metastatic niche by influencing the ECM and its embedded cells. First, MCPs induce cancer cells to undergo an epithelial-to-mesenchymal transition (EMT), a genetic program that promotes metastatic dissemination of cancer cells from primary epithelial tumors (177). Second, MCPs reorganize the ECM architecture and integrity to promote cancer cell accessibility into intact structures, that is, basement membrane (178). MCPs can also affect physical properties of the ECM, including spatial arrangement, orientation, rigidity, permeability, and solubility, in such a way as to alter anchorage sites and create motility tracks suitable for metastasis (178).

Normally, epithelial cells maintain their polarity, intercellular tight junctions, and adherence to the basement membrane necessary for proper tissue architecture and function (179). During EMT, epithelial cells undergo reorganization of adhesion and cytoskeletal structures to acquire a mesenchymal morphology. This allows cells to detach, which, in conjunction with enhanced migratory capacity associated with the mesenchymal phenotype, stimulates metastasis (179).

SPARC family members promote EMT in a variety of cancers (Fig. 1C). Recently, SMOC2 was shown to participate in a prometastatic secretome mediated by the ARNTL2 transcription factor in lung adenocarcinoma (180), and SMOC2 induction is required for colon cancer invasion by stimulating EMT (65). Several studies also show that SPOCK1 promotes EMT (110, 181). Among SPARC family members, SPARC is the most characterized for influencing EMT (106, 182, 183): (i) in lung cancer cell lines, TGFβ1 activation of migration and EMT is in part through SPARC (106), (ii) in head and neck cancer cells, SPARC enhances EMT signaling via activation of Akt (182), and (iii) overexpression of SPARC in melanoma cells increases invasiveness mediated by phosphorylation of FAK and Snail repression of E-cadherin promoter activity (184).

The CCN family exerts varying effects on EMT. An early study using pancreatic cancer cells reported that CCN1 promotes EMT and stemness, and that silencing this MCP forestalled aggressive tumor cell behavior by reversing the EMT phenotype (67). Recent studies have continued to dissect CCN1 signaling leading to EMT. In osteosarcoma, pharmacologic or gene knockdown of integrin αvβ5/Raf-1/MEK/ERK signaling components inhibited CCN1-induced EMT (185), as well as CCN1-mediated expression of EMT markers and cell spreading through an IGF1Rβ-JNK–dependent pathway (186). In contrast, CCN5 and CCN6 exert opposing effects on EMT. In triple-negative breast cancer cells, CCN5 activates the Bcl-2/Bax apoptotic pathway and inhibits both EMT and migration (187), while activation of the JAK/Akt/STAT pathway reverses such CCN5-mediated events (188). Similarly, CCN6 reversed the EMT features and inhibited metastasis of breast cancer cells in vivo, but through a Slug signaling axis that regulates Notch1 activation (189). Another mechanism involves CCN6-BMP-4 binding in breast cancer cells, which reduces BMP-4 signaling through p38/TAK1 and subsequent downstream activation of invasion and migration (190).

TNC and POSTN have also been associated with metastasis (191–195). While the influence of TNC (196, 197) and POSTN (198, 199) can be exerted through the EMT process, interestingly, these MCPs are also capable of remodeling the ECM to form migratory tracks that support rapid dissemination of cancer cells (Fig. 1C). TNC is frequently observed to be expressed along the border of matrix tracks in skin (200), pulmonary (201), colorectal (202), and breast (203) cancers. In fact, TNC assembles into matrix tracks with ECM molecules such as fibronectin, laminins, and several collagens (200, 204), which are also linked to metastatic potential (200, 205, 206). Evidence reveals that these TNC matrix tracks have a functional purpose in metastasis. In coculture experiments, leading fibroblasts were able to create matrix tracks composed of TNC and fibronectin, which were left behind for the movement of SCC cells (207). For fibronectin and TNC to coassemble into such tracks, POSTN is responsible for incorporating TNC into the meshwork architecture (208). While integrating TNC, it is possible that POSTN could also serve as a scaffold for BMP-1, LOX-1, and collagen to accelerate collagen cross-linking into migratory tracks during metastasis (172). Mechanistically, track mobility involves TNC competing for syndecan-4 binding to fibronectin, which blocks integrin α5β1–mediated cell adhesion for detachment (139), followed by TNC promotion of migration through integrin α9β1 following YAP inactivation (209). As previously discussed, TNC and POSTN can bind multiple ECM proteins (i.e., fibronectin and various collagens) and enzymes (i.e., LOX) to serve as scaffolds of collagen cross-linking needed for cancer cell proliferation and survival. This is a similar process that comes into play when TNC and POSTN interact to build ECM scaffolds for migration tracks (24, 167, 204, 208, 210).

Finally, for metastatic cells to exit the embedded state for intravasation, and then return into the ECM for colonization after invasion at a distant site, a degree of matrix plasticity is required. Such ECM remodeling is achieved by the degradative activity of extracellular proteases (Fig. 1C), in particular MMPs. Like several other MCPs, SIBLINGs bind and activate MMPs to promote metastasis. SPP1 binds CD44 to activate MMP-3 while BSP binds to integrin αvβ3 to activate MMP-2 to increase invasiveness in various cancer cell types (20, 211). Furthermore, SPP1 and BSP bind and activate MMP-3 and MMP-2, respectively (20). Early studies reported that SPARC upregulates the expression and activity of MMP-2 and MMP-14 in glioma cells and MMP-2 in breast cancer cells (212, 213). On the other hand, the SPARC family member SPOCK2 was recently shown to inhibit the expression of MMP-2 and MMP14, and activation of MMP-2 in endometrial cancer cells (214, 215). TNC may also influence the invasion of chondrosarcoma, colon cancer, and glioma cells by interacting with and upregulating MMP-1, -2/9, and -12, respectively (216, 217).

MCPs can also serve as a substrate for MMPs (218), that is, SPARC and SPP1 in the case of MMP-2, -3, -7, -9, -12, and -14, and MMP-3, -7, -9, and -12, respectively. From the SPARC family, SPARCL1 and SPARC are cleaved by MMP-3 in gliomas (219) and cathepsin K in bone cancer (220), respectively, whose fragments could affect SPARC activity. Recently, MMP-9 cleavage of SPARC was reported to enhance SPARC-collagen binding, preventing collagen degradation by MMPs in lung cancer (166). As for SPP1, thrombin and plasmin can cleave its C-terminal, which increases adhesion of melanoma cells (221) and migration of breast cancer cells (222), while cleavage of SPP1 by MMP-9 is essential for hepatocellular carcinoma invasion, which correlates with metastatic potential (223).

Apart from targeting MMPs, the role of TNC in influencing the ECM to promote invasion is multifaceted. In Ewing sarcoma, TNC expression and Src activation cooperate to promote invadopodia formation, an actin-rich protrusion of the plasma membrane involved in degradation of the ECM during cancer cell extravasation (224). In order for distant sites to accommodate disseminated tumor cells, MCPs are also required at the secondary target tissue to prime the metastatic niche for colonization. TNC has been shown to be involved in metastatic colonization because loss of this MCP in breast cancer, melanoma, or metastatic niche stromal cells inhibited colonization in the lungs (225–227). Gla-containing proteins have also been implicated in establishing a metastatic niche. Tumor-derived POSTN was reported to form a microenvironmental niche supportive of breast cancer stem cells via the integrin αvβ3/ERK pathway (228). In various mouse models, POSTN was responsible for metastatic colonization of the lung by breast and melanoma cells as evidenced by POSTN-neutralizing antibodies, antisense oligonucleotides, and knockout mice, all independently inhibiting metastasis (229–231). Given their significance for breast cancer cell dissemination to the lungs, it remains possible that both POSTN and TNC are interdependent in promoting colonization of the metastatic niche, because POSTN anchors TNC to the ECM (208). MGP was also recently shown to influence the metastatic niche by promoting osteosarcoma adhesion, extravasation, and MMP activities in murine lung endothelium in vitro (94).

MCPs are generally expressed at low levels in adult tissues but highly upregulated in various pathologies or injuries (4–6). This has prompted researchers to elucidate the potential functions of different MCPs in diseases such as cancer. As discussed throughout this review, numerous studies have shown that MCPs play critical roles in cancer development. In addition, the presence of certain MCPs in circulation as well as diseased tissue indicates their utility as noninvasive diagnostic and prognostic cancer biomarkers. Furthermore, their extracellular location and involvement in cancer pathology indicate that MCPs represent accessible and potentially effective therapeutic targets. In the following sections, we discuss various preclinical studies and clinical trials exploring the above possibilities.

SPARC has been suggested as a prognostic biomarker for certain cancers such as soft tissue sarcoma, esophageal SCC, and glioblastoma because its expression correlates with poor survival (232–234). In addition, SPP1 may be prognostic for breast, lung, gastric, liver, and colon cancers because it is associated with tumor progression and decreased patient survival (235–238). Subsequently, a number of ongoing clinical trials have been established to validate their application. Recently, SPARC has been the subject of a clinical study probing its utility as a diagnostic marker for brain cancer [registered number clinical trial (NCT) 01012609], given prior investigations correlating increased tumor vascular SPARC expression with decreased brain cancer patient survival (239). Several groups have also reported that high plasma SPP1 concentrations might be predictive of poor outcome for several cancers, including breast cancer (240). Consequently, one clinical study is currently probing the relevance of SPP1 serum levels for diagnosis of breast cancer (NCT 02895178). Other MCP families await successful clinical trials since the expression of several CCN family members in pancreatic, breast, oral, esophageal, and brain cancers (241–245), TNC in colorectal, glioma, pancreatic, and bladder cancers (196, 246–248), and POSTN in various solid cancers (249) have all been touted as potential diagnostic and prognostic biomarkers.

Targeting MCPs for therapeutic purposes has received relatively little attention, primarily because of limited data concerning mechanisms of action. The fact that MCPs are located in the extracellular space during cancer development renders them attractive as accessible targets for drug delivery; moreover, their context-specific expression implies that targeting these proteins would result in minimum pleiotropic side-effects. Neutralizing antibodies against MCPs have shown success in various preclinical settings; however, translation to the clinic has been difficult. One group showed that an SPP1 mAb (AOM1) significantly inhibited tumor growth and metastasis in a mouse model of non–small lung cancer (250). In addition, a commonly used mAb for antagonizing CCN2 (FG-3019) has reportedly been used in preclinical models with success in both monotherapy and combination therapy for different tumor types, including pancreatic and melanoma (251–256). With such progress, neutralizing antibodies targeting MCPs have advanced to registered clinical trials. For example, FG-3019 is currently in phase III for CCN2-targeted treatment of pancreatic cancer (LAPIS, NCT03941093).

Alternatively, MCPs could be targeted by inhibiting gene expression in patients. In fact, one of the first studies using RNAi to treat cancer with promising results involved targeting TNC in 11 patients with glioma (257). This was followed up with an investigation of a larger cohort of 46 patients, which reported significant improvement in overall survival (258). Other promising MCP targets for posttranscriptional gene silencing include SPP1, POSTN, CCN1, and CCN2, where inhibition of RNA expression was shown to reduce cancer progression in various animal models (68, 73, 229, 235). Another therapeutic approach involves exploiting the high expression of MCP within the tumor environment as a strategy to deliver therapeutic molecules. Using a high affinity antibody to deliver radiotherapy (mAb 81C6), TNC was targeted for treatment of glioma and lymphoma (259, 260), showing safe and promising antitumor benefit.

Upon perturbation of tissue homeostasis during multistage carcinogenesis, MCPs are upregulated in the tumor microenvironment to become key mediators of cell–ECM communication that in turn promotes cellular proliferation, survival, and metastasis. The functional diversity of MCPs stems from their ability to interact with a variety of extracellular signaling molecules such as ECM components and growth factors. Moreover, as emphasized in this review, many MCPs have been implicated in cancer development, and thus may certainly exert additive and/or antagonistic effects in this process. Our current understanding of MCP pathways gives an impression of redundancy, and so a primary aim in the ECM field is to elucidate the precise manner in which MCPs mechanistically converge, both functionally and temporally, to remodel the tumor microenvironment and orchestrate critical neoplastic processes.

This overarching goal highlights a major challenge, that of developing experimental systems that better model the physical state of the native interstitial ECM. The usefulness of various existing models, including 2D monolayers (261), 3D Matrigel (262), and tissue-extracted ECM (263), is limited because these models fall well short of fully recapitulating the complexities of tissue ECM in vivo. This inadequacy may underlie some of the discrepancies in the literature regarding MCP functionality in cancer. Furthermore, the identification of naturally occurring protein–protein interactions and posttranslational modifications among MCPs in the ECM have been difficult to characterize. Overall, as concisely reviewed elsewhere (264), new approaches are clearly needed to dissect the daunting complexity of the ECM environment and its role in carcinogenesis.

While confronting the above challenges it remains important to concomitantly work toward characterizing particular MCPs, alone or in combination, as impactful diagnostic/prognostic cancer biomarkers and therapeutic targets. In fact, given their burgeoning roles in cancer development and extracellular accessibility, MCPs have long been regarded as potentially useful for diagnosing and treating various pathologies such as fibrosis and cancer; nonetheless clinical data supporting this notion have been relatively scant. Toward addressing this knowledge gap, over the past decade, progress has been made in defining better the fundamental mechanisms of MCPs, opening new questions that entice the generation of the next needed tools to understand sufficient detail for optimal therapeutic design.

Herein we have summarized some important ways in which misregulation of MCP expression promotes cancer development, including perturbation of intracellular signaling and aberrant coordination of ECM remodeling. Although we focused on MCP families with the most well-characterized roles, others are emerging as potentially important players, such as the EMILIN and R-Spondin families. Recently, R-Spondin-1 and 2 were shown to promote liver, glioblastoma, and ovarian cancer through their well-defined influence on Wnt/β-catenin signaling (265–267). In addition, EMILIN2 promotes the formation of tumor-associated vessels in melanoma (268), and EMILIN1 exerts an oncosuppressive role in colon and skin (269, 270). Clearly, there is still much to be discovered regarding the exquisite spatiotemporal regulation of MCP expression patterns and functions in the extracellular space during cancer tissue remodeling, similar to the approach taken in fibrosis (17). In this respect, as more and more knowledge accumulates, it should be possible to design appropriate clinical studies that could firmly establish MCPs as useful biomarkers and therapeutic targets in cancer.

A. Leask has ownership interest in Fibrogen. No potential conflicts of interest were disclosed by the other authors.

This work was supported by the Operating Grant Funding Program 24347 (co-funded by Cancer Research Society and the Kidney Cancer Research Network of Canada) and start-up funds from Hôpital Maisonneuve-Rosemont Foundation (all to C. Gerarduzzi). C. Gerarduzzi is a recipient of the Kidney Research Scientist Core Education and National Training (KRESCENT) Program New Investigator Award KRES180003 (co-funded by the Kidney Foundation of Canada, Canadian Society of Nephrology, and Canadian Institutes of Health Research) and the Cole Foundation Early Career Transition Award.