|Home | About | Journals | Submit | Contact Us | Français|
The extracellular components of the cancer microenvironment play a critical role in tumor initiation, progression and invasion. In this review we examine the normal formation and function of the basement membrane and extracellular matrix. We characterize the interactions between the matrix and the epithelium and explore the causes and consequences of the extracellular remodeling that accompanies carcinogenesis. Finally, we address the therapeutic possibilities of incorporating matrix as well as epithelial strategies in the management of digestive cancer.
In our first review, “Stromal fibroblasts in digestive cancers”, we characterized the cancer microenvironment from the perspective of the important stromal cells that infiltrate and influence digestive tract carcinogenesis . The peritumoral stroma, however, also consists of a dynamic and integrated network of extracellular proteins stitched and woven together to support and sustain the overlying epithelium [2–5]. The gastrointestinal stroma provides more than just a structural base. It also provides context to the overlying epithelium in a process termed “dynamic reciprocity” [6, 7]. The embryological importance of the stromal–epithelial interaction was dramatically demonstrated in a series of experiments combining ectoderm and mesoderm from different species and from different tissue sites. In these studies it was shown that the stroma modified the epithelial phenotype, for instance corneal epithelium from a chick sprouted feathers when combined with mesoderm from a mouse . In another example mammary gland mesenchyme induced mammary gland development in overlying skin epithelium .
The regulatory relationship between epithelium and the underlying mesenchymal elements is interrupted in digestive carcinogenesis. The stromal changes in carcinogenesis such as the recruitment of inflammatory cells, remodelling of the extracellular matrix (ECM), activation of fibroblasts and promotion of angiogenesis co-evolve with the progressive epithelial transformation .
In this review we outline the normal formation and function of the basement membrane (BM) and ECM. We examine the interactions between the matrix and the epithelium and the extracellular remodelling that accompanies carcinogenesis. Finally, we explore the therapeutic possibilities of incorporating stromal as well as epithelial strategies in the management of digestive cancer [11–14].
Firstly, we shall place the BM and ECM in the context of the adjacent epithelium. The interactions between epithelial cells inform the physical and functional interface between the epithelium and matrix. Normal epithelial cells are polarized. They have an apical pole, which in most of the alimentary tract is the luminal surface with important barrier, absorption and secretion functions. There are lateral surfaces that form important intercellular connections to communicate and provide tensile strength along the epithelial sheet. Epithelial cells also have a basal surface that provides epithelial anchoring to the underlying BM through cell–matrix interactions. The key epithelial intercellular junctions include tight junctions, adherens junctions, desmosomes and a cell–matrix junction, the hemidesmosome. These junctions are comprised of cellular adhesion molecules such as cadherins, immunoglobulin (Ig) superfamily cellular adhesion molecules (CAMs), integrins and selectins . Cadherins and Ig-CAMs interact with like molecules on the opposite cell membrane. Integrins and selectins interact with important matrix proteins or cell surface sugar motifs, respectively . These CAMs are also important in cell–cell and matrix–cell signaling conveyed via conformational changes in their cytoplasmic domains, as discussed below.
The typical adherens junction involves an interaction between paired E-cadherin molecules across adjacent cell membranes. The cytoplasmic domains associate with α- and β-catenin and indirectly with F-actin. Within this complex there are at least two important mediators of gastrointestinal cancer, E-cadherin and β-catenin [16, 17]. The tight junction, another important intercellular interaction, contributes to the barrier function of epithelium, particularly relevant for the gut. Tight junctions contain members of the claudin family, occludin and junction adhesion molecule . Again, claudin proteins are implicated in several digestive cancer models, including esophageal, colon, gastric and hepatocellular cancers [18–21]. Finally, the desmosomes and hemidesmosomes, which anchor epithelial cells to adjacent cells and the BM, respectively, are associated with an intracellular network of intermediate fibers . The hemidesmosome utilizes the integrin α6β4 to secure the epithelium to laminin within the BM .
The BM is the immediate subepithelial extracellular layer (Fig. 1). The BM is a sheet-like network of extra-cellular proteins, compiled into a layer approximately 100–300 nm thick .The BM is comprised of three basic types of matrix molecules: type IV collagen, proteoglycans such as perlecan and multiadhesive matrix molecules including laminin, fibronectin and entactin (nidogen). The BM provides a structural and functional foundation to the overlying epithelium. BMs are not exclusive to the epithelium and also surround myocytes, endotheliocytes, adipocytes as well as serving in a specialized capacity in the glomerulus .
Type IV collagen is exclusively expressed within, and is the chief structural protein of, the BM . Type IV collagens assemble into very fine unstriated fibers, which associate to form a sheet-like structure. There are 6 type IV collagen genes (COL4A1 (α1), COL4A2 (α2), COL4A3 (α3), COL4A4 (α4), COL4A5 (α5), COL4A6 (α6)), but in the digestive tract the main type IV collagen is comprised of the heterotrimer α1α1α2 . Whilst most mutations in COL4A1 (α1(IV)) and COL4A2 (α2(IV)) are embryologically lethal, other chains are mainly expressed with maturation particularly within the developing kidney . COL4A1 and COL4A2 are located in a head-to-head orientation on chromosome 13q34. The type IV collagen network is strengthened by disulfide and aldimine bridges, as well as through its interactions with other molecules, as discussed below.
Perlecan is the major proteoglycan within the BM . Perlecan consists of repeating domains, which include laminin-like and Ig-like domains. Long glycosaminoglycan (GAG) molecules are attached to the core protein backbone. This combination of core protein and GAG branches provides perlecan with its vital binding capacity, linking scores of matrix proteins including entactin and laminin, cell surface receptors and epithelial growth factors .
Laminin is the primary multiadhesive matrix molecule in the BM. Laminin is a large, cruciform molecule with calcium dependent globular domains that bind to cell surface receptors, such as integrins . Entactin, another important matrix molecule, is a sulfated multidomain glycoprotein, which helps to cross-link BM components, including type IV collagen, perlecan and laminin .
Does the intestinal BM represent an extension of the epithelial or of the stromal compartment? Given that the BM exists as an interface between tissues of endodermal and mesodermal origin it is not surprising that the intestinal BM itself has a dual origin, from both intestinal epithelial and mesenchymal cells [13, 25, 26]. Mesenchymal intestinal subepithelial myofibroblasts produce the majority of the BM’s type IV collagen and entactin [13, 27]. Laminin is produced commensurately by epithelial and mesenchymal cells and perlecan primarily from the epithelial compartment [13, 27].
Beneath the BM is the lamina propria, which contains ECM interspersed with fibroblasts and other important stromal cells, such as recruited inflammatory cells. This is the next barrier that an invasive carcinoma must negotiate. The ECM shares many components with the BM, albeit that the matrix molecules in the ECM promote a 3-dimensional matrix rather than the sheet-like network of the BM [15, 23]. This difference results from the fibrillar collagens, particularly type I collagen, which replace type IV collagen as the defining structural glycoprotein of the ECM . Like type IV collagen, type I collagen is also a heterotrimer consisting of two α1(I) chains and one α2(I) chain. Type I collagen is chiefly produced by the resident stromal fibroblasts, although in the activated stroma recruited fibroblasts may also contribute . Once again, proteoglycans and the attached GAGs, such as heparin sulfate, play an important role in cell and matrix interactions. Fibronectin is an important polypeptide dimer within the ECM . Fibroblasts and endothelial cells are the major producers of local fibronectin, albeit that some epithelial cells including intestinal and mammary epithelium can produce smaller amounts . Fibronectin is an essential adhesive molecule that facilitates cellular-ECM attachments and can influence the morphology and motility of the associated cells. Fibronectins are also important for wound healing through their activation of clotting and chemotaxis of inflammatory cells into the activated stroma .
This section examines the specific mechanisms of matrix–cell signaling and thus the means by which the ECM influences epithelial biology. These mechanisms include ECM-cellular signaling through adhesion receptor intermediaries, such as integrins, ECM as a repository for epithelial growth factors and the contribution of matrix stiffness and deformity on adjacent cell behavior.
All cells have, or have had, a direct connection to the ECM during their development . The connection to the matrix helps to regulate progression through the cell cycle and cellular differentiation [30, 31]. Integrins, which recognize fibronectin, laminin and collagens, are critical in cell signaling and regulation of cellular growth (Fig. 2) . Integrins consist of an α and β-subunit, with different combinations of αβ providing binding and signaling specificity . The binding of integrins to the ECM promotes the reorganization of actin filaments into larger fibers and further aggregation of integrins, in a positive feedback loop . Integrin-associated cytoplasmic signaling proteins are recruited to the site to promote signaling . Integrins can activate a number of tyrosine kinases, including focal adhesion kinase (FAK), Src-family kinases, Abl, and integrin-linked kinase (ILK) . FAK interacts with the cytoplasmic tail of the β-subunit of integrin . Once activated, FAK undergoes autophosphorylation to produce a binding site for Src or Fyn. The Src kinase leads to further phosphorylation of key targets and activation of the Ras-Raf-MEK-ERK cascade, which leads to expression of several important proteins including cyclin D1 (Fig. 2) [31, 33]. As well as potentiating growth factor signaling through the Ras-Raf-MEK-ERK pathway integrins may also potentiate cellular proliferation by cycling ERK into the nucleus, a process that may depend on cellular adhesion . In addition, adhesion-related increases in phosphoinoside 3-kinase are important for the expression and stability of cyclin D1 [35, 36]. Integrins vary in their mitogenic capacity. Integrin αvβ3 signaling seems to be particularly important and can result in prolonged activation of ERK and activation of the NF-κB pathway . In several cell types expression of αvβ3 is associated with increased growth and even tumorigenesis .
Anoikis is a perfect illustration of the dependence of the normal epithelium on the matrix. Anoikis is a special example of apoptosis that results from the loss of cell–matrix interactions . Anoikis is a normal daily occurrence along the gut as senescent alimentary tract cells are shed into the lumen. In addition to the loss of the mitogenic pathways, outlined above, anoikis also results from the inactivation of the pro-survival protein Bcl-2, due to the release of normally cytoskeleton-bound Bmf . Failure of anoikis is an important process in carcinogenesis. In digestive cancers there are many examples of cancer mutations to circumvent anoikis, such as activating KRAS and BRAF mutations in colorectal cancers [39, 40].
Integrins connect the intracellular actin cytoskeleton to the ECM and thus mediate a rich network for detecting and responding to mechanical forces . Epithelial cells exert forces on their BM through adhesive, integrin-rich structures known as focal adhesions . These focal adhesions are likely to be the critical site of mechanotransduction . Fibrosis and carcinogenesis can increase matrix rigidity and, via integrin signaling, can cause the activation of Rho and other typical integrin-related signaling pathways as outlined above, including Ras-Raf-MEK-ERK [42, 43]. An interesting feature of fibroblast migration is that they tend to migrate towards more rigid matrix . There is a difference in tissue rigidity between the relatively stiff tumor and peritumoral stroma versus the comparatively less rigid surrounding normal tissue. It follows that this migratory mechanism, referred to as durotaxis, may be an important factor in fibroblast recruitment into the activated stroma .
In addition to the downstream synergy between growth factor and ECM signaling (Fig. 2), the ECM also regulates growth factors more directly . Many growth factors, such as VEGF and HGF, bind to important matrix molecules including heparin sulfate and fibronectin, within the ECM . The structured ECM reservoir of growth factors facilitates interstitial gradients, establishes a tissue-specific growth factor repository for release in case of tissue injury and ECM proteins also serve as important co-factors in the presentation of growth factors to their specific receptors, as is the case with FGF and TGF-β . Given the important role of TGF-β in the development of stromal fibroblasts it is worthwhile to consider the interaction between the ECM and this growth factor in more detail . TGF-β is secreted as a propeptide and following cleavage by furin protease, it associates with latency-associated peptide and latent TGF-β binding proteins . The latent TGF-β binding proteins also bind to ECM molecules, including fibrillins and fibronectin. Active TGF-β is released following degradation of the ECM or displacement or cleavage of latency-associated peptide . In addition, certain integrins such as αvβ6 and αvβ8 may activate ECM-bound latent TGF-β to facilitate localized activity of TGF-β . Finally, there is increasing evidence to suggest that the ECM itself, via specific growth factor-like domains, may be able to directly activate growth factor receptors .
Carcinomas are defined by the interruption of normal cellular compartments. Thus, “successful” neoplasia must acquire cells with the capacity to dissociate from their original tissue as well as invade adjacent and distant cellular communities . Invasion involves breaching several important barriers, including the 2-dimensional BM, the 3-dimensional ECM and ultimately the series of barriers posed by both proximate and distant endothelium and by the destination tissues . This involves neoplastic cellular changes but also changes in the surrounding matrix. As mentioned above, the type IV collagen network of the normal BM presents a resilient barrier. The BM has a regular pore size of about 50 nm, which limits cellular transmigration . Thus, the normal BM must undergo considerable remodeling in order to permit the migration of cells . In the lamina propria the 3-dimensional network, established by type I collagen, results in a structural barrier with effective pores of approximately 150 nm. Therefore, the ECM must also undergo considerable structural changes to permit invasion . These ECM compartments reflect a contribution from neoplastic cells, inflammatory cells and the resident and recruited stromal fibroblasts. All have important roles in ECM remodeling.
The integrity of the BM is determined by both the synthesis of new components by the contributing epithelial and stromal cells and the degradation of the BM by proteolytic factors. The cellular production side of this equation is important and has been alluded to above and in the context of our other review . The production and assembly of the BM is undoubtedly impaired in digestive cancers. Breaches in the BM are often most marked at the invasive front associated with epithelial–mesenchymal transition [47, 48]. The increased expression and activity of ECM-degrading enzymes, however, also promotes migration and metastasis [23, 49, 50]. These molecules are able to directly digest ECM proteins and are also capable of liberating growth factors and other signaling factors bound to ECM components .
Matrix metalloproteinases (MMPs) are one group of very important ECM proteases that have been implicated in carcinogenesis and metastasis [23, 51]. There are 25 MMPs that may be classified by their location of action. Eighteen MMPs are soluble enzymes secreted into the ECM, six are membrane-type MMPs (MT-MMPs) that are attached to a cell surface and one, MMP-23, is initially a membrane-bound MMP but can also be cleaved to form a soluble factor . Expression of important MMPs and MT-MMPs occurs within, but is not restricted to, the neoplastic cells . Stromal fibroblasts and endothelial cells also express these molecules to promote their own migration into the activated stroma and thus contribute to the microenvironment in terms of remodeling and angiogenesis [53–55].
MMPs are frequently over-expressed and over-activated in malignant cancers, particularly surrounding their advancing margin . Considerable MMP research has focused on MMP-2 and MMP-9. These enzymes are associated with tumorigenesis [56, 57], but are not sufficient to stimulate invasion through the BM . Furthermore, mice doubly deficient in these enzymes still achieve normal embryonic BM transmigration events such as pancreatic islet cell formation, suggesting that MMP-2 and MMP-9 are not essential for cellular invasion and migration [23, 59]. The roles of MMP-2 and MMP-9, as well as the other MMPs, within the cancer microenvironment are still being fully characterized [23, 60–62].
There has been greater evidence for a direct contribution from membrane-type-1 MMP (MT1-MMP), membrane-type-2 MMP (MT2-MMP) and membrane-type-3 MMP (MT3-MMP) in BM remodeling and migration . COS cells that were transfected with MT1-MMP, MT2-MMP, MT3-MMP but not MMP-2,-3,-7,-9,-11 or -13 led to BM effacement and migration in an ex vivo tumor invasion model .
ECM-proteases do not just break down physical barriers, they are integral biological mediators in both late and early carcinogenesis . For instance, in transgenic mice and cell culture experiments it was shown that ectopic MMP-3 expression in normal mammary epithelium induced carcinogenesis, primarily through upregulation of Rac1b, reactive oxygen species and resultant genomic instability [63, 64]. The concept of matrix proteases inducing cancer was supported by a transgenic mouse study in which increased matriptase activity in the epidermis caused cutaneous squamous cell carcinomas .
The contribution of invadopodia to digestive cancer is an evolving area of research . Invadopodia are cell–matrix contacts, similar to focal adhesions, established in the context of “invading” cellular protusions . Invadopodia can promote ECM proteolysis through their associated proteases, particularly MT1-MMP , and thus have been implicated in cancer invasion, both neoplastic cells moving out and cancer-associated fibroblasts moving in . The cause and consequences of invadopodia unites many of the key themes from the discussion above, including ECM remodeling, matrix proteases, cell motility and mechanical forces . Whilst, invadosomes have been examined primarily in 2-dimensional cell culture, there is now evidence for their role in 3-dimensional cell culture studies . Their in vivo contribution to the digestive cancer microenvironment, however, remains to be established .
2-dimensional cell culture experiments have been the in vitro cornerstone of much basic cancer research. Tumor biology, especially the interface between neoplasia and the ECM, however, extends to 3-dimensions . Therefore, 3-dimensional (3D) cell culture (i.e. organotypic) model systems have been developed. 3D cell culture systems can incorporate epithelial and stromal cell populations, such as cancer-associated fibroblasts, to better approximate the in vivo cancer microenvironment (Fig. 3). Whilst, 3D cell culture represents a great advance in cancer research it still cannot examine the recruitment of important cells, such as fibroblasts, endotheliocytes and leukocytes, into the cancer stroma. Furthermore, it is still unclear as to which ECM product best represents the in vivo situation, obviously a very important issue when trying to develop a biologically faithful model. Available options for ECM include collagen gels, ECM produced by stromal cells in vitro, or even a BM-type matrix known as Matrigel™ (also known as EHS matrix) [69, 70]. Matrigel™ is extracted from the EHS tumor, which is a mouse tumor derived from parietal endoderm (named after those that discovered and characterized the tumor, Engelbreth-Holm and Swarm) .
As in any in vitro model all of these ECM options represent a compromise between practical and biological considerations. For instance, collagen gels can mimic either loose or dense ECM depending on the concentrations used, but lack other important ECM components that influence cancer migration and growth, as outlined above. ECM that is produced by stromal cells in vitro may contain many of the important ECM components, but is limited by lower collagen concentrations and thus larger internal spaces than the true cancer stroma . And finally, whilst Matrigel™ has proven to be an important product in cancer research, it is a BM-type matrix containing BM components such as type IV collagen. Type IV collagen is physiologically-organized into and specific for the 2-dimensional BM rather than the 3D ECM, thus Matrigel™ may be best considered as a BM rather than as an ECM product, per se. In addition, Matrigel™ contains several angiogenic- and growth-promoting factors, which may in part explain the enhanced tumor growth when Matrigel™ is used as a vehicle in nude mice tumorigenicity studies [72, 73].
Organotypic cell culture has undoubtedly advanced our understanding of the ECM contribution to digestive carcinogenesis. Future in vitro systems, however, should incorporate physiological stromata that are dynamic and responsive to better reflect the changing ECM composition and cellular populations that accompany the initiation, progression and local invasion of digestive cancers.
Despite the fact that the BM and ECM contribute to tumor initiation, progression and invasion [6, 23], our approach to cancer prevention, diagnosis and management still focuses almost exclusively on the epithelium.
As discussed in our review on stromal fibroblasts, there are developing opportunities for incorporating stromal profiling in defining the pathological progression, and perhaps even the pathological potential, of carcinomas [1, 74]. Epithelial molecular analysis of CRCs is now mainstream and advanced CRCs are routinely tested for KRAS mutation prior to starting monoclonal antibody treatments targeting the epidermal growth factor receptor (EGFR) to ensure that only those most likely to respond are administered these therapies . Given the interaction between integrin signaling and the Ras-Raf-MEK-ERK pathway (Fig. 2), it is possible that integrin specific signaling pathways that feed into this pathway may also be relevant in predicting response to the EGFR-monoclonal antibodies cetuximab and panitumumab.
As discussed in our stromal fibroblast review, the tumor stroma could be manipulated as an adjunct to other therapy. For example, an inhibitor of hedgehog signaling was recently used in an animal model of pancreatic cancer to inhibit stromal desmoplasia and thus improve drug delivery to the tumor . As discussed above, agents that effect desmoplasia may also influence the mechanical properties of the tissue, perhaps with implications for integrin related signaling and epithelial growth . Targeting specific integrins or other specific molecules at the integrin-BM interface offers promise as a new approach in biological cancer treatment [74, 77–81].
The MMPs are being investigated as potential biomarkers and therapeutic targets in digestive cancer . MMP-2, -7 and -9 levels in blood, pancreatic juice and tissue have been used to distinguish malignant from benign pancreatic pathologies, as well as for prognostication [83–86]. In CRC, levels of MMP-1, -7, and -13 within the malignant epithelium and stroma have been associated with decreased survival . In gastric cancer, the presence of high MMP-7 staining on surgical specimens was associated with shorter survival . Finally, a study of plasma MMP-2 and MMP-9 suggested that these markers could even help detect patients with colorectal neoplasia . In the future, these stromal factors may serve as potential screening tests to help direct investigations or as a surveillance tool following colonoscopy or colectomy.
There has been a great deal of interest in developing and trialing MMP inhibitors in advanced cancers, including advanced CRC . AE-941 (Neovastat) is derived from shark cartilage and inhibits MMP-2, -9 and -12 [90, 91]. Dalteparin is a low-molecular weight heparin and inhibitor of MMP-9 . Other agents include chemically modified tetracyclines, as well as synthetic drugs structurally designed to block or reconfigure the activated MMP into its proenzyme state . Whilst, there was some initial promise in a phase II trial of Neovastat in advanced renal cell cancer, overall the therapeutic effectiveness of this approach has been disappointing [82, 93]. Nevertheless, with the development of more sophisticated drugs to selectively target the most critical MMPs, this area of research still presents exciting opportunities for the discovery of new chemotherapeutics and chemopreventatives.
The ECM and BM actively contribute to the normal development and maintenance of the gastrointestinal tract and undergo extensive remodeling during digestive carcinogenesis. These extracellular compartments contain important structural and regulatory molecules that actively signal to the overlying stroma to help regulate growth and differentiation. Important ECM proteases, such as MT1-MMP, influence neoplastic growth and dissemination. Along with the peritumoral stromal cells , the ECM is a critical and dynamic contributor to the epithelial–stromal sequence that characterizes digestive carcinogenesis . Cancer researchers are actively engaged in characterizing and translating these stromal events into better diagnostic, prognostic and therapeutic approaches to help reduce the burden of digestive cancer in our community.
The authors wish to thank Dr Michael Quante for supplying Fig. 3.
Sources of support
The Royal Australasian College of Physicians Cottrell Fellowship, the Queensland State Government Smart State PhD award and the NHMRC/RG Menzies fellowship (D.L.W).