The contribution of vascular basement membranes and extracellular matrix to the mechanics of tumor angiogenesis^†

While normal tissues are dependent on oxygen delivery for survival, it is now well established that pathological states such as tumor growth are also dependent on oxygen delivery via recruitment of new blood vessels and capillaries. A colony of cancer cells cannot form a tumor larger than 1–2 mm³ unless it stimulates development of its own vasculature (1). Tumor vasculature arises via the process of angiogenesis and vasculogenesis, with the help of local cells and the bone marrow-derived stem cells. Angiogenesis is a process whereby new vessels are formed via sprouting events from the pre-established vasculature. In this process endothelial cells detach from pre-existing blood vessels, are surrounded by provisional extracellular matrix (ECM), form a lumen, adhere to newly synthesized vascular basement membranes (VBM) and are enveloped by pericytes, thus forming new capillaries which enable the flow of blood (1–5). An important factor that determines whether a new blood vessel will form is the physiological balance between pro-angiogenic factors and the anti-angiogenic factors in the body and their local influence on the endothelium. Interestingly, endothelial cells in close proximity to each other may respond differentially to angiogenesis stimulation, indicating that non-angiogenic tumor microenvironment-related factors influence how the endothelium of the pre-established vasculature responds to angiogenic stimulus.

In recent years our understanding of VBM has dramatically changed from their being mere structural components of tissues and barriers of infiltration, to their being an active modulator of blood vessel formation and function. VBM surround normal blood vessels of all sizes, providing a scaffold between the endothelial cells that line the vessels and smooth muscle cells/pericytes that form a peripheral supporting cell layer. A basement membrane, although possibly abnormal and leaky, also surrounds all the tumor-associated blood vasculature (6). Endothelial cell and pericyte adhesion to the scaffold provided by the VBM is predominantly mediated by integrins and other matrix adhesion molecules (7–9). This adhesion to VBM and potential changes to the VBM structure influence cellular decisions associated with growth, differentiation or apoptosis (10). Apart from their role in regulating pivotal cell-matrix interactions, the ECM/VBM have been found to sequester angiogenic factors and inhibitors of angiogenesis, which become bio-available upon proteolysis of matrix proteins by enzymes such as matrix metalloproteinases, cathepsins and elastases (8). Therefore, VBM constitutes a solid-state depot source of critical factors that can regulate angiogenesis. In this review, we will highlight the role of VBM in the regulation of angiogenesis and tumor growth. We will approach this subject from the viewpoint of VBM as a solid-state structure (its assembled form), and its solution state form generated by proteolysis within the tumor microenvironment and elsewhere. We will review how the same set of proteins, in different forms, exert distinct functional influences upon the endothelial cells of the blood vessels.

MECHANICAL VIEW OF THE VBM

In addition to uncontrolled growth, malignant transformation of cells results in the loss of normal tissue architecture and contributes to the breakdown of tissue boundaries. In this regard, progression of cancer is considered to be a collection of influences from neoplastic cells and their microenvironment. Angiogenesis is one such host-derived contributor to tumor progression. Several growth factors, ECM molecules and membrane receptor signaling pathways have been implicated as important for regulating tumor angiogenesis (1, 11). These factors can reproducibly in various in vitro experiments either induce or inhibit angiogenesis. During angiogenesis, individual endothelial cells bind to both angiogenic stimulators and inhibitors, and attach to and detach from the surrounding matrix components. What determines endothelial cell sprouting and what particular endothelial cells of a growing capillary/blood vessel contribute to it is unknown. Where and how new angiogenic sprouts originate in both physiological and pathological settings should be an active area of angiogenesis research.

In a camera lucida study of physiological neovascularization by Clark and others in 1938, it was demonstrated that growing sprouts become functional capillary tubes when the perivascular matrix changes to a tissue substance resembling a ‘soft gel’ (12). This observation possibly represents the first description of the role of ECM and BM-like material in the angiogenic response. Subsequent studies further suggest that FGF-stimulated endothelial cells may be ‘switched’ between growth, differentiation and involution modes during angiogenesis by altering the adhesivity or mechanical integrity of their ECM (13).

Ingber and coworkers have studied the role of changes in ECM structure and mechanics in tissue differentiation/remodeling during angiogenesis and vascular patterning using the cellular tensegrity model (11, 14–16). This hypothesis is based on the fact that cells possess a filamentous cytoskeletal (CSK) network containing contractile microfilaments, microtubules, and intermediate filaments (11, 14, 16, 17). The CSK is in direct connection with the ECM through integrins, providing continuous tension between the cell and its surroundings. In the tensegrity model, the forces generated by the CSK and the binding to the ECM generate a dynamic force balance and establish an internal pre-existing tensile stress that stabilizes the cell (11, 14–16). The notion here is that local variations in ECM remodeling that are observed during angiogenesis change ECM structure and mechanics, and thus influence cell behavior by altering the tensile stress. High matrix molecule turnover in localized areas causes thinning of the VBM at the tips of new capillary sprouts during active angiogenic recruitment (18). A thinner VBM becomes compliant and stretches out more than the VBM associated with neighboring endothelial cells. Such structural differences induce a change in the balance of forces that are transferred across cell surface adhesion receptors that link the ECM to the CSK. It is thus conceivable that changes in endothelial cellular physics result in localized growth and motility that drives new blood vessel formation. The continuation of this process along the invasive edge of newly formed capillary vessels leads to the formation of branching patterns that are characteristic of all growing vascular networks (Fig. 1) (11).

Interesting results were also obtained by developing a method in which cell distortion becomes an independent variable of cell behavior. Micron-sized adhesive islands were created and coated with a saturating density of different immobilized ECM molecules and surrounded by barriers of non-adhesive regions containing polyethylene glycol to prevent protein adsorption. Varying the size and shape of the adhesive island produced different effects on cell behavior. Increasing the surface area (and promoting cell spreading) caused cell growth, while decreasing it caused apoptosis, even when endothelial cells were cultured in a growth promoting medium (19). On the other hand, moderate restraining of spreading caused differentiation of endothelial cells (20). Most interestingly, various ECM molecules (type I and IV collagens, FN, laminin and vitronectin), which all use different types of integrin receptors, displayed the same effects on cell shape and function when immobilized under similar conditions, highlighting the matrix component- independent effect of cell distortion and alteration in tensile stress on cellular behavior (19, 21–24). Using the same method, it has also been shown that the effects of directional migration of endothelial cells are distortion dependent. When a cell adherent to ECM islands was exposed to a mitogen such as platelet-derived growth factor (PDGF) or fibroblast growth factor (FGF), the cell sent out lamellipodia, filopodia and microspikes, driving cell migration only in a mode allowed by the geometrical constraints of the island, such as the corners of a square cell. This means that contact with the ECM and subsequent distortion of the cells strongly drives the directionality of cell movement (25).

These studies indicate that by varying the degree to which cells can spread over ECM in an in vitro setting, endothelial cell responses to soluble stimuli can be mimicked (13, 19–21, 25–27). Such observed cellular behavior is very similar to that seen during vascular pattern formation in vivo and most likely plays a fundamental role during physiological and pathological angiogenesis.

THE SOLID STATE FORM OF VBM

Basement membranes (BM) are highly specialized 50–100 nm-thick sheet-like extracellular matrices that surround epithelium, endothelium, muscle cells, peripheral nerves and fat cells, separating them from the stroma (28, 29). In blood vessels, vascular basement membranes (VBM) are located between the endothelial cell lining and the pericytes that make up the outer wall of the vessel (Fig. 2) (30–35). BM are composed of many different proteins that are produced by most cell types. The main components of BM are type IV collagen, laminin, and heparin-sulphate proteoglycans (HSPGs) such as perlecan and nidogen/entactin. Other proteins that are found in smaller amounts are type XV and XVIII collagens, fibulins, agrin, and SPARC/BM-40/osteopontin (29, 34 36). BM self-assemble through a complicated process that involves interaction with cell-surface proteins such as integrins to form a laminin network that is central to BM formation. A separate type IV collagen network is also formed, providing the scaffold network structure, and subsequently the laminin and type IV collagen networks interact to form a mature BM scaffold (30–35). Although most BM contain the same set of proteins, several different variants of these BM proteins exist, which generates tissue-specific BMs with possible roles in diverse tissue morphogenesis and function. Angiogenesis depends on VBM components that support growth and survival of vascular endothelium. However, in recent years cryptic domains within these VBM proteins, which are released during BM turnover and breakdown, have also been identified as being involved in inhibiting angiogenesis (37–50).

TUMOR VBM AND VESSEL STRUCTURE

Blood vessels associated with tumors are considered abnormal, displaying irregular diameters, altered branching and leakiness (5, 51, 52). The VBM of tumor vessels have been described as incomplete or absent (53, 54). Endothelial cells in tumor vessels have loose interconnections and areas of intercellular gaps, which could also explain the observed leakiness (53, 54). Tumor-associated blood vessels have an incomplete coverage of pericytes when compared to normal blood vessels (5, 55, 56). The lack of adequate pericyte coverage is most likely one of the explanations for the vascular heterogeneity, because these cells have an important role in conferring structural support to the blood vessel (5, 57, 58). Betsholtz and coworkers studied the role of platelet-derived growth factor B (PDGF-B) and its receptor PDGF-Rβ in pericyte recruitment to blood vessels during developmental and tumor angiogenesis (5). Mice deficient in PDGF-B or its receptor, or lacking PDGF-B specifically in the endothelial cells, display developmental microvascular defects due to inadequate pericyte recruitment. Such defects were very similar to those observed in tumor blood vessels (59–61). By using mice with genetically altered PDGF-B or providing PDGF in the in vitro angiogenesis assays, it was shown that functional abnormalities associated with tumor-associated blood vessels were partially overcome due to successful pericyte recruitment. However, these pericytes still displayed loose connections with the endothelium, possibly due to altered VBM composition, structure and adhesion (62).

Benjamin and coworkers elegantly demonstrated the importance of pericyte coverage for blood vessel survival in their studies with VEGF and also pericyte dependence for tumor vasculature (63). The authors analyzed the ratio of vessels lacking pericyte coverage to that of vessels covered by pericytes and demonstrated that during tumor progression the number of immature blood vessels lacking pericyte coverage increases (63). Constructing a C6 glioma cell line that expresses VEGF gene under the control of tetracycline promoter enabled the authors to turn VEGF expression on or off in these cells at will. The cells were used for tumor xenograft studies in nude mice and, when VEGF expression was turned off, increased endothelial cell apoptosis and blood vessel regression was observed. Interestingly, the blood vessels susceptible to apoptosis by VEGF withdrawal were the cells lacking pericyte coverage (63). The same was subsequently shown to take place in human tumor samples from prostate cancer patients who had undergone androgen ablation (known to reduce VEGF levels). Lower numbers of immature blood vessels lacking pericyte coverage were observed as compared to in patients who had not undergone androgen ablation (63). These results stress the importance of proper interactions between the various cellular and matrix constituents for appropriate blood vessel formation.

By analyzing the composition and structure of VBM in different mouse tumor models, Baluk et al. 2003 could show that VBM identified by type IV collagen immunoreactivity covered >99.9% of blood vessels in all analyzed tumors. Type IV collagen-immunoreactive zones were also present on endothelial sprouts, supporting the idea that basement membrane components are present as sprouts grow and regress (6). Other VBM components such as laminin, entactin/nidogen and fibronectin were found surrounding the tumor vessels. These results indicate that involvement of VBM components is essential for angiogenesis. Several earlier studies have indicated frequent holes in VBM, but in this study using 1 μm confocal optical sections, the gaps were <2.5 μm in diameter and involved only 0.03% of the vessel surface (6). However, the VBM of the tumors was not entirely normal; structural abnormalities such as a loose association with endothelial cells and pericytes, broad extensions away from the vessel wall, and multiple layers observed by electron microscopy were documented.

SOLUTION STATE FORM OF VBM

Basement membranes including the VBM are highly insoluble structures that are tightly cross-linked. Endothelial cells and pericytes are bound to such insoluble structures in a reversible manner. Therefore, it is conceivable that turnover of VBM, endothelial cells and pericytes is in synchrony, to facilitate cell-matrix interactions. An appreciation of the contribution of VBM to tumor angiogenesis has dramatically changed in the last few years. Tumor microenvironment-associated proteases break down VBM and allow fragments of VBM to become soluble and enter a solution state, which results in the exposure of endothelial cells to different fragments and molecular structures of VBM molecules, that could either have a pro-angiogenic or anti-angiogenic effect on proliferating endothelial cells. Such a view of one gene product either supporting endothelial growth (in assembled and intact form) or having anti-angiogenic function (degraded unassembled form) stresses the active and dual role of VBM during tumor angiogenesis. Interaction of such cryptic protein fragments with endothelial cell surface receptor/s leads to cell-signaling events resulting in angiogenic effects. Anti-angiogenic fragments derived from well-known VBM proteins such as type IV collagen, type XVIII collagen and type XV collagen will be discussed in detail below.

TYPE IV COLLAGEN-DERIVEDANTI-ANGIOGENIC FRAGMENTS – ARRESTEN, CANSTATIN AND TUMSTATIN

Type IV collagen is the main protein component of all basement membranes and it is crucial for the stability and assembly of this specialized connective tissue structure (64, 65). Six different type IV collagen α-chains have been identified in mammals. The α1- and α2-chains are found in most basement membranes, whereas the other chains display a more restricted tissue expression pattern. The genes that encode type IV collagen α-chains are located in an intriguing head-to-head fashion on three different chromosomes, so that α1- and α2-chains, α3- and α4-chains, α5- and α6-chains share potential bifunctional promoters and are transcribed on opposite strands in different directions (66). The means by which such a genomic structure regulates transcription and promotes subsequent α-chains trimerization and assembly into higher macromolecular structures resulting in a scaffold of type IV collagen networks in different tissues is currently unknown. Each type IV collagen α-chain consists of three domains: an N-terminal 7S domain, a middle triple-helical domain, and the C-terminal globular non-collagenous (NC1) domain (66). The NC1 domain is considered important for the assembly of the type IV collagen, which is then assembled into the suprastructure of type IV collagen. The importance of a normal and appropriate type IV collagen network for the function of BM is highlighted by diseases such as Alport syndrome, a hereditary kidney disease which is caused by mutations in the type IV collagen α3, α4 or α5 -chain (66–70).

Goodpastures syndrome, on the other hand, is caused by autoimmunity against type IV collagen α3 -chain, present in large amounts in the glomerular BM of the kidney (66, 71, 72). The gene for the collagen α3 -chain has been inactivated in mice and the mice lacking this gene develop normally, but succumb to early renal failure (70). By degrading VBM preparations with tumor-associated enzymes such as matrix metalloproteinases (MMPs), elastases, and cathepsins, our laboratory identified several fragments from VBM proteins with anti-angiogenic activity (Table 1) (8, 73).

Arresten (NC1 domain of α1) is a 26 kDa molecule that was isolated from degradation of human placenta BM preparation and recombinantly produced in Escherichia coli and 293 embryonic kidney cells (43). Arresten inhibits endothelial cell proliferation, migration, tube formation and neovascularization of Matrigel plugs. The growth of two human xenograft tumors in nude mice and the development of tumor metastases is inhibited by arresten (43). The anti-angiogenic activity of arresten is potentially mediated via interaction with α1β1-integrin on endothelial cells (43). Similarly, canstatin (NC1 domain of α2 -chain) is a 24 kDa molecule that was purified and recombinantly produced, and shown to significantly inhibit human endothelial cell migration, murine endothelial cell tube formation, endothelial cell proliferation, and induce apoptosis of these cells (50). This molecule also suppressed in vivo growth of large and small size tumors in two human xenograft mouse models (50). Peticlerc et al. have later validated the anti-angiogenic activity of canstatin and they also describe anti-angiogenic activity of the NC1 domain of the α6-chain (44). Tumstatin (NK1 Domain of α3-chain) is a 28 kDa molecule and the most extensively studied type IV collagen-derived anti-angiogenic molecule and will thus be discussed in more detail.

Human tumstatin exerts its anti-angiogenic effect by specifically inducing apoptosis of proliferating endothelial cells (45). Tumstatin can be found in the blood at physiological levels of approximately 300–350 ng/ml. This fragment is most likely generated by MMP cleavage of BM containing the α3-chain of type IV collagen, as part of the regular BM turnover process (45). Tumstatin interacts with αvβ3 integrin on the endothelial cell surface and this binding causes an inhibition of CAP-dependent protein translation mediated through negative regulation of mTOR (mammalian target of rapamycin) signaling (45–49). mTOR is a kinase that phosphorylates 4E-BP1 protein and activates the cap protein complex to initiate translation of mRNA. The ability of tumstatin to inhibit mTOR mimics that of rapamycin (a small-molecule inhibitor of mTOR with effect on all cell types) although in the case of tumstatin affecting only proliferating endothelial cells expressing αvβ3 integrin (49).

Tumstatin inhibits angiogenesis in several in vitro and in vivo models and suppresses tumor growth in various mouse tumor models (45–49). Mice with inactivation of the type IV(α3) collagen gene, thus also lacking tumstatin, display accelerated tumor growth associated with enhanced pathological angiogenesis. When these mice were supplemented with recombinant tumstatin, to reconstitute normal physiological concentrations, the increased rate of tumor growth was normalized. Interestingly, physiological angiogenesis during development and tissue repair proceeds normally in these mice. Proliferation of mouse lung endothelial cells (MLEC) isolated from mice with an inactivation of the β3 integrin gene was not altered by treatment with recombinant tumstatin (74). Tumstatin also displayed insignificant inhibition of VEGF-induced neovascularization of Matrigel plugs implanted into β3 integrin-deficient mice, indicating a requirement for interaction with β3 integrin for the anti-angiogenic effect (74). The tumor-associated enzyme MMP-9 has been found to cleave tumstatin most efficiently from type IV collagen chain. The in vivo importance of this cleavage is indicated by the fact that mice deficient in MMP-9 display decreased circulating levels of tumstatin and accelerated tumor growth (74).

TYPE XVIII COLLAGEN-DERIVED ENDOSTATIN

Type XVIII collagen is a BM protein of unknown function, but is believed to contribute to the normal development of the vasculature in the retina. Mutations in human collagen XVIII have been associated with Knobloch syndrome, an autosomal recessive disorder characterized by high myopia, vitreoretinal degeneration with retinal detachment, and congenital encephalocele (75, 76). Mice lacking type XVIII collagen are viable, healthy and reproduce well. However, these mice exhibit abnormalities in the outgrowth of retinal vasculature and delayed regression of hyaloid capillaries in the eye, suggesting an important role in the development of eye vasculature and providing an explanation for the ocular defects observed in Knobloch syndrome patients (77, 78). Recently type XVIII collagen-deficient mice were shown to have massive accumulation of subretinal pigment epithelium (RPE) deposits with striking similarities to basal laminar deposits, abnormal RPE and age-dependent loss of vision similar to what was observed in patients with age-related macular degeneration, the leading cause of blindness in the Western world (76).

Endostatin is a 20–22 kDa molecule that was isolated and sequenced from conditioned medium of a mouse haemangioendothelioma, and shown to be a C-terminal fragment of the α1-chain of type XVIII collagen (Table 1) (41). This anti-angiogenic molecule has been extensively studied and is currently in Phase II clinical trials (8). The endostatin domain can be cleaved from type XVIII collagen downstream of a trimerization region in an area containing a protease-sensitive hinge by cathepsin-L, elastin and matrilysin (79–81), and accumulates in blood plasma (20–35 ng/ml) and tissue extracts (82 83). Three different variants of type XVIII collagen are expressed due to the use of alternative promoters and vary as to their N-terminus composition, but all share a common C-terminus structure and contain the endostatin domain (84). The shortest form of type XVIII collagen strongly localizes to several VBM structures (84–86).

Endostatin possesses anti-angiogenic activity and causes tumor regression by acting as an inhibitor of endothelial cell proliferation and migration, inducing apoptosis of proliferating endothelial cells and causing G1 arrest of endothelial cells (37, 38, 41, 87). Endostatin has been shown to bind α5β1, αvβ3, αvβ5 integrins and glypicans on the cell surface in vitro (88–90). Endostatin has been found to block the binding of vascular endothelial growth factor (VEGF) isoforms VEGF₁₆₅ and VEGF₁₂₁ to the receptor Flk-1, and the subsequent tyrosine phosphorylation of this receptor, even though it does not bind to VEGF itself (91). Endostatin also stabilizes cell-cell adhesions and cell-matrix adhesions, and this inhibits angiogenesis as loosening of these junctions is required for new sprout formation (92). Endostatin has furthermore been described to affect matrix breakdown by inhibiting MMP-2 activity (93). The binding of soluble endostatin to 5β1 integrin has been studied by our laboratory. This binding leads to inhibition of the focal adhesion kinase/c-Raf/MEK1/2/p38/ERK1mitogen-activated protein kinase pathway in vitro, with no effect on phosphatidylinositol 3-kinase/Akt/mTOR/4E-BP1 and CAP-dependent translation (90). Therefore the anti-angiogenic action of endostatin is clearly different from that of tumstatin.

THE ENDOSTATIN HOMOLOGUE IN TYPE XV COLLAGEN

Type XV collagen is a protein highly homologous to type XVIII collagen and together these two collagens form a subfamily within the collagen family of proteins (84). Type XV collagen is a non-fibril-forming collagen, and is thought to be a homotrimer that consists of three α1(XV)-collagen chains (84). This protein is located to BM zones of many tissues, including the VBM (94–98), and is strongly expressed in heart and skeletal muscle (98–100). The highest homology between type XVIII and XV collagens is found in the C-terminal globular NC domain containing the endostatin domain (39, 40, 94).

The role of type XV collagen in the regulation of angiogenesis is unknown, although the endostatin-like domain of this protein has been shown to possess anti-angiogenic activity (Table 1) (40, 101). Type XV collagen might also be involved in cancer by other unknown mechanisms (102). Mice with inactivation of type XV collagen gene undergo normal development of the vasculature and are viable and fertile (103). Strongest expression of type XV collagen is found in heart and skeletal muscle, and in ultrastructural analyses, collapsed capillaries and endothelial cell degeneration were found in mice lacking this collagen (103). These studies indicate that type XV collagen has an important structural role in VBM, thus stabilizing skeletal muscle cells and capillaries (103). Recently, double knockout mice for collagen types XV and XVIII were generated and surprisingly these mice reveal no new major defects compared to the individual gene inactivated mice. This indicates that no major functional overlap takes place in vivo between these highly similar collagens with similar expression pattern in BM (104).

ENDOREPELLIN

The latest addition to the growing number of VBM protein-derived fragments with anti-angiogenic activity is endorepellin (Table 1). Endorepellin is an 81 kDa C-terminal fragment cleaved from the ubiquitous basement membrane protein perlecan. This fragment was found to inhibit endothelial cell migration, collagen-induced tube formation, and blood vessel growth in Matrigel plug as well as chorioallantoic membrane assays (105). It was also shown to bind endostatin and potentially counteract the anti-angiogenic effect of this molecule. The endothelial cell receptor/s or the intracellular events triggered by endorepellin still remain unknown (105).

SUMMARY

A specialized connective tissue structure, the vascular basement membrane, surrounds all blood vessels. Basement membrane provides a structural scaffold for the endothelial cell and pericyte interactions, and changes in the composition of the VBM provide the cells with signals that affect cell growth, differentiation and apoptosis. The basement membrane thus constitutes an important extracellular microenvironment sensor for the endothelial cells and pericytes. During tumor growth and invasion, the basement membrane barriers are broken down and cryptic domains of the VBM proteins are potentially released. These fragments have been found to affect endothelial cell behavior by acting to reduce cell proliferation, migration or induce apoptosis. These VBM- derived endogenous inhibitors of angiogenesis are attractive candidates for potential cancer therapy.

This work was partially supported by NIH grants DK62987, DK55001, and research funds from the Center for Matrix Biology. LX is funded by the Stop and Shop Pediatric Brain Tumor Foundation. MS is the recipient of the Sigrid Juselius Fellowship and is funded by the Sigrid Juselius Foundation, the Maud Kuistila Foundation, the Finnish Medical Society Duodecim and the Emil Aaltonen Foundation. We would also like to acknowledge the generous financial donations to the Kalluri Laboratory from the Ann L. and Herbert J. Siegel Philanthropic Jewish Communal Fund and the Joseph Lubrano Memorial Goldfield Family Charitable Trust Fund.