The Role of the Extracellular Matrix in Cancer Stemness

1The extracellular matrix (ECM) is a major structural component of the tumor microenvironment and comprised of a network of biochemically distinct components, including fibrous proteins, glycoproteins, proteoglycans, and polysaccharides. The ECM is a highly dynamic structure, constantly undergoing a remodeling process where ECM components are deposited, degraded, or modified. Increasing evidence suggests that the ECM serves as a niche for normal and cancer stem cells (CSCs). CSCs, also called tumor-initiating cells, are a small population of cells within tumors that have capabilities of self-renewal properties, tumor initiation and chemoresistance. As one of the CSC niches, the ECM provides both structural and biochemical support to regulate proliferation, self-renewal, and differentiation of CSCs. In this review, we will cover the current understanding of how different ECM components affect the cancer “stemness” phenotype.

2Collagens constitute the main structural element of the ECM and are the most copious type of fibrous proteins within the interstitial ECM. Collagens play a role in tissue development by providing mechanical strength, altering cell adhesion, promoting cell migration. Studies have reported that several collagens (e.g., COL3A1, COL4A2, COL7A1, COL17A1) are overexpressed by CSCs (Table 1). Multiple collagen subtypes have been shown to increase epithelial-mesenchymal transition (EMT), tumor-initiating potential, drug resistance and self-renewal of CSCs (Table 1 and Figure 1).

3Glycoproteins, which make the ECM a cohesive network of molecules by linking cells together with structural components, include fibulin, fibrillin, laminin, fibronectin, vitronectin, tenascin-C, Secreted Protein Acidic and Rich in Cysteine (SPARC), periostin (POSTN), thrombospondin, mucins (MUCs) and nidogen (Table 1). CSCs overexpress several glycoproteins (e.g., tenascin-C, POSTN, MUC1) and their receptors (e.g., integrins αVβ3 and α9β1, CD47). Adhesive glycoproteins bind to integrins, non-integrin receptors, growth factors, and other ECM components to activate downstream signaling pathways to regulate EMT, self-renewal, and drug resistance of CSCs (Table 1). For example, fibronectin, a major adhesive ECM glycoprotein that attaches cells to a variety of ECM components, has been shown to increase EMT, self-renewal, expression of CSC markers and drug resistance of CSCs. Laminins, another class of adhesive glycoproteins that constitute structural scaffolding of all basement membranes, support self-renewal of CSCs through their interaction with integrins. Some glycoproteins have dual roles in cancer stemness depending on the cancer type. For instance, fibulin-3, an ECM glycoprotein associated with basement membranes, inhibits self-renewal in lung and pancreatic CSCs while stimulating breast CSC self-renewal (Table 1).

4Proteoglycans are glycosylated proteins composed of a core protein and one or several covalently attached sulfated glycosaminoglycan chains and are present in the ECM of connective tissues. Proteoglycans play a crucial role in ECM assembly and cell signaling. They bind to growth factors, cytokines and other ECM molecules and act as co-receptors to assist ligand and cell surface binding to modulate downstream signaling. Several proteoglycans (e.g., decorin, lumican, biglycan, versican, aggrecan) are highly expressed by CSCs and their roles in cancer stemness are summarized in Table 1.

5Polysaccharides, a chain of monosaccharide repeats linked through glycosidic bonds, fill the interstitial space and buffer physical stress on the ECM. Hyaluronic acid (HA or “hyaluronan”) is a high-molecular-mass polysaccharide that constitutes a major component of interstitial gels, especially in soft connective tissues. In tumors, HA is produced by both tumor stroma and tumor cells, and its binding to the cellular receptor CD44 activates intracellular signaling (e.g., PI3K/Akt and Erk pathways, RhoA and Rac, Ras, NF-kB and Src signaling) to promote cell survival, cancer stemness, motility and invasion by cytoskeletal reorganization. High levels of HA are produced by CSCs and HA–CD44 interaction has been shown to promote acquisition of CSC characteristics and chemoresistance in breast, ovarian and head and neck CSCs (Table 1).

6Physical properties of the ECM such as rigidity, porosity and topography impact various anchorage-dependent CSC functions. The interstitial ECM, mainly composed of collagens, proteoglycans and hyaluronan, provides a physical barrier that hinders the transport of solutes, water and chemotherapeutic drugs. In this regard, it has been shown that cisplatin, a chemotherapeutic drug frequently used to treat various solid tumors, extensively binds to collagen fibers in tumors. Binding of chemotherapeutic drugs to the ECM prevents drug penetration into tumors, thereby increasing CSC survival. The ECM also provides sites for adhesion of CSCs in the tumor microenvironment. ECM-CSC interaction via CSC receptors such as integrins (e.g., β1, α6, β3, β4), discoidin domain receptors (DDR1, DDR2), CD44 (HA receptor) and CD47 (thrombospondin 1 receptor) enhances CSC properties. For example, CSCs bind to HA through CD44 and this increases not only the expression of stemness factors NANOG and SOX2 but also MDR1 (Multi Drug Resistance 1) expression and drug resistance in breast and ovarian CSCs. The ECM also provides anchorage and homing sites for CSCs in pre-metastatic niches, initiating metastatic colonization and organotropism of cancer cells. For instance, infiltrating breast tumor cells induce the expression of POSTN in the stroma of the secondary target organ (e.g., lung). By recruiting Wnt ligands and increasing Wnt signaling in CSCs, POSTN sustains CSC population in the secondary site and promotes metastatic colonization. Changes in the ECM topology also affects CSC self-renewal by controlling the balance between symmetric and asymmetric cell divisions. The spatial distribution of the ECM has been shown to guide the orientation of the cell division axis by controlling the location of actin polymerization at the membrane through focal adhesions and the segregation of cortical components in the interphase. The β1 sub-family of integrins also regulates stem cell self-renewal by controlling the balance between symmetric and asymmetric cell divisions. Furthermore, ECM distribution affects migration of cancer cells and immune cells. During tumor progression, wavy collagen fibers become straightened and align perpendicular to the tumor boundary. It has been shown that linear collagen fibers oriented perpendicular to the tumors facilitate high-speed migration of breast cancer cells and paired macrophages to promote metastasis to distant organs.

7Tumor ECM is typically stiffer than normal tissue ECM due to overexpression of many ECM components (e.g., collagens I, II, III, V, IX, and XI, heparan sulfate proteoglycans) and ECM-modifying enzymes [e.g., lysyl oxidase (LOX)]. Mechanical properties conferred by ECM stiffness are transmitted to CSCs through the formation of focal adhesions and subsequent activation of mechanotransduction pathways (e.g., Rho/ROCK, YAP/TAZ). ECM stiffness plays a crucial role in regulating stem cell self-renewal and differentiation. Several studies have demonstrated that ECM stiffness directs human mesenchymal stem cells (MSCs) and neural stem cells to differentiate into different cell lineages. Human MSCs cultured on hydrogel with an elastic modulus very similar to bone marrow, exhibit enhanced self-renewal and multipotency. In the case of melanoma CSCs, three-dimensional (3D) soft fibrin matrices promote histone 3 lysine residue 9 (H3K9) demethylation and increase SOX2 expression and self-renewal, whereas stiff matrices exert the opposite effects. Conversely, breast CSCs increase CSC marker expression on stiff matrix through integrin linked kinase (ILK) signaling, suggesting that the effect of matrix stiffness on stemness is cancer type specific.

8The ECM can provide external cues that induce EMT, one of the cellular transformation processes that has been shown to route some cancer cell types from a differentiated to a stem cell state. Collagen I has been shown to induce EMT through activation of ILK and subsequently NF-κB-dependent inactivation of GSK-3β, along with the nuclear translocation of β-catenin. Collagen XVII and laminin-5 can also induce EMT-driven cancer stemness through the activation of FAK/Akt paired with inhibition of GSK-3β. The induction of EMT and CSC phenotypes by the ECM seems to be driven by a master regulator, Akt. Akt activation, which can be achieved via intracellular focal adhesion proteins such as FAK and ILK, subsequently modulates the activity of downstream effectors. For instance, Akt can activate NF-κB, which has been shown to upregulate the expression the stemness genes SOX2, NANOG and KLF4 in breast and prostate cancer cells. Akt, as well as ILK, can also inactivate GSK-3β, which increases the nuclear translocation of β-catenin, a transcription factor that is associated with stemness and is also an activator of NOTCH and Wnt signaling. Therefore, ECM regulates the switch between CSC and non-CSC states by inducing EMT.

9The ECM also promotes CSC self-renewal. In this regard, collagen I has been shown to preserve stemness in malignant and non-malignant stem cells by activating transcriptional programs that induce self-renewal. Binding of collagen to α2β1 integrin results in the nuclear translocation of Bmi1, a stemness-inducing transcription factor downstream of Hedgehog signaling. Studies have shown that Bmi1 is a transcriptional target of Gli1, a stemness related gene, and that FAK/Ras signaling enhances the expression of Gli1. Akt/p-S6K1 signaling has also been shown to play a regulatory role in activity of Gli1. Laminin and fibronectin signaling also plays a crucial role in CSC self-renewal. Laminin 511 can sustain breast cancer stemness through activation of α6β1 integrin, in a TAZ-dependent manner. TAZ expression and nuclear localization induce the expression of the stemness transcription factors, OCT4, SOX2 and NANOG in non-malignant and malignant cells. Fibronectin’s extra domain A (EDA) has also been demonstrated to positively regulate CSC self-renewal through activation of α9β1 integrin/FAK/ERK/Akt/β-catenin pathway.

10The ECM might serve as a reservoir for factors that aid in the sustenance of CSCs. Embryonic stem cells (ESCs) have been shown to utilize matrix metalloproteases 1 (MMP1) to release ciliary neurotropic factor (CNTF) from an ESC-derived matrix, which enhances ESC self-renewal though JAK/STAT3 signaling, a pathway that has also been implicated in promoting self-renewal of breast CSCs. Hematopoietic stem cells (HSCs) also upregulate MMP-9 to release soluble kit-ligand, also known as stem cell factor (SCF), which promotes survival signaling and chemoresistance in many types of cancers. CSCs are thought to remodel their matrices more significantly than their non-cancer stem cell counterparts as CSCs upregulate expression of different MMPs. This may enable them to effectively degrade and remodel ECM matrices to release growth factors and cytokines to promote their survival.

11The ECM serves as a functional repository for a plethora of factors that dynamically modulate the tumor microenvironment to promote CSC metabolism. Focal adhesion formations transduce ECM signaling into the tumor cells and activate the PI3K pathway which increases glycolysis, in addition to activating glutamine signaling in a Ras- and Myc- dependent manner. Furthermore, a stiff ECM acts as a driver of glycolysis in CSCs. On the contrary, accumulating evidence suggests that CSCs also utilize OXPHOS, fatty acid oxidation and glutaminolysis. In this regard, it has been demonstrated that CSCs with high telomerase activity upregulate glycolysis and OXPHOS in lung and ovarian cancers. Given the diversity of tumors and their microenvironments, it is possible that based on the availability of nutrients, CSCs can manipulate their metabolism. For example, while CSCs in a hypoxic microenvironment may survive by means of glycolysis, CSCs in a normoxic environment use oxidative metabolism. Furthermore, CSCs utilize metabolites secreted by cancer-associated fibroblasts such as lactate and ketone bodies to fuel OXPHOS. Recycling of nutrients via autophagy is another way by which CSCs not only self-renew but also acquire drug resistance. Autophagy impairment downregulates the expression of CSC markers and consequently the CSC self-renewal capacity in breast, liver, ovarian and pancreatic cancers, osteosarcoma and gliobastoma. ECM-receptor ligation has been shown to induce autophagy. Collagen VI, a promoter of tumorigenesis and a supporter of stem cell niches, also functions as an autophagy inducer in skeletal muscle stem cells by functionally interacting with decorin, a small leucin-rich proteoglycans (SLRP) that has been shown to induce stemness in glioblastoma. A growing number of studies indicate that collagen VI directly maintains CSCs by activating the Akt–GSK-3β–β-catenin–TCF/LEF axis, which is required for activation of autophagy. Decorin signaling, independent of Collagen VI, can also maintain stemness of trophoblasts and prevent their differentiation.

12Solid tumors frequently contain highly hypoxic regions and tumor hypoxia is positively associated with poor prognosis. Hypoxic tumor cells express stem cell markers, are highly undifferentiated and exhibit enhanced clonogenic potential in vitro and tumor initiating potential in vivo. Furthermore, hypoxia can lead to increased ECM deposition and remodeling. Histological studies on clinical tumor samples have shown increased collagen deposition resulting in fibrosis in hypoxic regions of tumors. In addition to cancer cells, fibroblasts cultured under hypoxic conditions show increased type I procollagen α1 mRNA. Abrogating HIF1α expression inhibits collagen deposition from both breast cancer cells and fibroblasts in vitro and in vivo. ECM remodeling enzymes such as LOX, LOX-like protein 2 (LOXL2), LOXL4, MMP2, MMP9 and MMP14 and growth factors inducing collagen deposition (e.g., VEGF) are HIF-regulated genes that are involved in tumor fibrosis. Since all these factors have been previously implicated cancer stemness, it is not surprising that the ECM acts a functional conduit for hypoxia-derived signals that foster cancer stemness.

13Extracellular matrix can profoundly influence recruitment of immune cells into the tumor microenvironment. CSCs can evade immune surveillance by altering this microenvironment to favor their survival. For example, ECM drives the activation of pro-survival pathways such as PI3K/AKT, which has been shown to facilitate immune evasion in CSCs. ECM proteins can recruit immunosuppressive cells such as tumor-associated macrophages (TAMs) and regulatory T cells that have been known to promote CSC survival, while simultaneously blocking the recruitment of antitumorigenic immune cells such as cytotoxic T cells. In addition, the ECM composition can dramatically modulate the activation state of the tumor infiltrating immune cells. For instance, a stiff collagen-rich or POSTN-rich ECM allows macrophage polarization to a pro-tumorigenic M2 phenotype. Following recruitment, the M2 macrophages activate several CSC survival signaling pathways including Src, NF-κB, STAT3/SOX2 and Hedgehog. ECM can also impair proliferation and activation of T cells, that are required for capturing and killing CSCs. A collagen-rich ECM can inhibit T-cell proliferation and activation through type I collagen-dependent fusion of LAIR receptors in addition to sequestering growth factors required for T cell proliferation. Furthermore, TAMs and neutrophils that can selectively reorganize the ECM to promote malignant growth of cancers are preferentially recruited to the microenvironment.

14Currently, there are several inhibitors targeting various aspects of ECM-induced cancer stemness that are undergoing clinical testing. For example, the CD47 blocking protein TTI-621 is currently being assessed in a number of phase I clinical trials (NCT03013218, NCT02663518, NCT02216409, NCT02678338) for various types of cancers. Other groups have targeted FAK with the inhibitor VS-6063 (Defactinib), which has completed clinical phase I and II trials (NCT01778803, NCT01943292, NCT01951690) with one of those clinical trials assessing for CSCs as an endpoint (NCT01778803). Other inhibitors of stemness-related molecules further downstream of ECM signaling are also being tested in clinical trials, such as the STAT3 inhibitor BBI-608 in a phase II trial that will test for presence of CSC as an endpoint (NCT02279719) and in a phase III clinical trial aimed at reducing CSCs by targeting phosphorylated Stat3 positive cancer cells (NCT02753127). The β-catenin pathway inhibitors PRI-724 and CWP232291 are currently being tested in two phase I clinical trials (NCT01764477, NCT01398462). Inhibition of the Hedgehog pathway with the inhibitor GDC-0449 (Vismodegib), is also currently being clinically evaluated in a phase II trial which will test for the presence of pancreatic CSCs (NCT01088815).

15Although the above drugs may effectively reduce the number of CSCs, there are still many potential challenges that ECM components in a tumor microenvironment may set that could interfere with an otherwise successful treatment regimen. Firstly, ECM proteins have been shown to act as a physical barrier, making drug delivery to cancer cells more difficult. Secondly, ECM proteins can de-differentiate non-CSCs into CSCs, which makes eliminating all CSCs more challenging. Thirdly, ECM plays a role in modulating immune cell recruitment, hence, potential immunotherapeutic strategies could be hindered by dysregulated ECM components. Finally, the ECM has a very complex and dynamic nature: different ECM molecules are expressed in a time and tissue-specific manner where various isoforms of the same molecule can play opposing functions in cancer stemness in a context-dependent manner. Considering these concerns, it is crucial that future studies further elucidate the role of ECM components on cancer stemness in order to design therapies that effectively eradicate all CSCs.

16All authors conceptualized the content and wrote the manuscript.

17The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.