The extracellular matrix (ECM) acts both as a physical scaffold for cells and as a repository for growth factors. Moreover, ECM structure and physical–chemical properties convey precise information to cells that profoundly influences their biology by interactions with cell surface receptors termed integrins. During angiogenesis, the perivascular ECM plays a critical role in determining the proliferative, invasive and survival responses of the local vascular cells to the angiogenic growth factors. Dynamic changes in both the ECM and the local vascular cells act in concert to regulate new blood vessel growth. The digestion of ECM components by proteolysis is critical for the invasive capacity of endothelial cells, but also creates ECM fragments, which antagonize the mechanosensory function of integrins, and can be apoptogenic. Here, we discuss the roles of integrins in modulating cellular responses to a changing ECM, in particular the regulation of survival and invasion among invasive endothelial cells.
The initiation of angiogenesis
Angiogenesis as a determinant in disease
New blood vessel growth, or angiogenesis, occurs during the development and maturation as well as during critical physiological processes, including wound healing and reproduction. However, angiogenic processes are also usurped in many pathologies, and thereby contribute to an array of disorders including cancer, inflammation and autoimmune diseases (Folkman, 2006). In these cases, angiogenesis typically supports expanding aberrant tissues, such as a tumor or the rheumatoid pannus. However, the hypervascularization of these tissues often promotes significant aspects of disease pathology. For example, it may exacerbate inflammation in autoimmune disease or facilitate tumor progression or dissemination in cancer. In other cases, angiogenesis itself is the principal factor responsible for disease pathology (Andreoli and Miller, 2007), such as is the case in diabetic retinopathy and macular degeneration (Andreoli and Miller, 2007), or even endometriosis (Sha et al., 2007). In these cases, it is now well established that an aggressive neovascular response can serve as a diagnostic and/or prognostic indicator of disease progression (Craft and Harris, 1994; Gasparini et al., 1996).
In contrast, the loss of vascular function that follows stroke or infarction leads to tissue ischemia and a variety of harmful sequelae (Marti and Risau, 1999). In these cases, revascularization of the local tissues is critical to restoring tissue function. Thus, it may sometimes be desirable to promote neovascularization to treat disease. Whether one considers the promotion or inhibition of angiogenesis as a therapeutic strategy, an increased understanding of the highly complex mechanisms that regulate the survival and invasion of angiogenic vascular cells is critical.
Architecture of the resting microvasculature
The vascular cells resident within the arterioles, venules and capillaries of the body are typically quiescent. The endothelium acts as vascular lining cells, which facilitate smooth blood flow. Capillary endothelial cells are anchored in place through interactions with the underlying vascular lamina over a very large surface area, which encompasses nearly half the cell membrane. In addition, endothelial cells interact with each other in cell–cell junctions that often partially overlap each other. These junctions limit vascular leak, permitting only molecules below 75 kDa to escape the circulation (Firth, 2002; Weis et al., 2007). The junctions also served to provide structural continuity, connecting the cytoskeletal structure of one cell to the next. Specialized junctions, such as those in the brain, can limit the escape of blood-borne elements further (Ueno, 2007). Interspersed along the exterior of capillaries is a second type of vascular cells; the pericyte, which extends cellular processes to contact several different endothelial cells in its immediate microenvironment. These processes interact directly with the endothelial cell surface and both support and stabilize the capillary structure. The interaction between pericyte and endothelial cells is important; loss of contact provides one criteria for the transition from mature, nonproliferative vessels to immature, unstable angiogenic vessel.
The basal proliferation of vascular cells is low, among the lowest of any cell in the body (Folkman, 2006). However, the combination of low proliferation, strong anchorage, tight cell junctions and supporting pericyte contact defines the infrastructure that renders mature endothelium highly resistant to proapoptotic insults. However, this privileged ‘resistant’ status changes following the induction of angiogenesis.
Early signaling events during the initiation of angiogenesis
Angiogenesis is a local process that can be induced by paracrine and autocrine growth factors. The list of proangiogenic agents that can stimulate quiescent endothelial cells to become angiogenic is still growing and includes protein growth factors, bioactive lipids and even complex polysaccharides. Angiogenesis can be considered to occur in a series of stages, with immediate early events triggered by the binding of growth factors and cell signaling, followed by early events that change genetic programming of vascular cells, followed by invasive angiogenesis and finally a resolution stage, where cells revert to their quiescent stage. At each step, cell interaction with the local microenvironment plays a critical role in determining the angiogenic response.
Vascular endothelial cell growth factors (VEGFs) represent a family of cytokines that stimulate vascular and lymphatic endothelium to form new vascular or lymphatic structures, respectively. VEGFs are well-studied growth factors that bind to cell surface heparin, neuropilin and to the receptor tyrosine kinases Flk-1, Flt-1 and Flt-3, depending upon the particular VEGF isoform (A, B, C and D) or splice variant (Ferrara et al., 2003). VEGF-A binding to Flk-1 on endothelial cells triggers a cascade of immediate-early signaling events, in which the activation of nonreceptor tyrosine kinases play key roles. In particular, the Src-family kinases and the members of the focal adhesion kinase (FAK) family are important.
These early Src-family kinase-mediated events include Src-dependent phosphorylation of VE-cadherin, which promotes the disruption of endothelial cell junctions (Potter et al., 2005), a critical factor in inducing vascular permeability (Figure 1). The induction of permeability slows blood flow, decreasing the clearance of locally produced angiogenic factors, while facilitating the leakage of plasma-borne components into the local interstitial tissues. Moreover, Src-family kinases coordinately phosphorylate FAK, which promotes the turnover of the stable focal contacts used by resting vascular cells for anchorage (Eliceiri et al., 2002). The coordinated cytoskeletal remodeling goes beyond simple mobilization of integrins and disruption of cell–cell junctions. The activation of phosphoinositide 3′ kinases (PI3K) is critical for signaling small GTPases of the Rho family (Cheresh et al., 1999; Hoang et al., 2004; Nagy and Senger, 2006), whereas the initiation of mitogen-activated protein kinase (MAPK) signaling pushes cells towards proliferation, and may mobilize contractile elements important for cell invasion (Klemke et al., 1997; Zhai et al., 2003). Together, the pathways permit dynamic interaction between cytoskeleton and newly forming focal contacts initiated by ligation of integrins, the principle adhesion receptors for the extracellular matrix (ECM).
Integrin mobilization and subsequent interactions with ECM proteins potentiate the signaling events critical to immediate early events in angiogenesis. In fact, the signaling activity of receptor tyrosine kinases is dependent upon integrin engagement of ECM (Eliceiri et al., 1998). This is one important reason that the repertoire of integrin receptors on angiogenic vascular cells changes during angiogenesis; to maintain appropriate cellular interaction with the remodeling ECM. The earliest interactions occur through pre-existing integrins, such as αvβ5, which are activated downstream of growth factors (Klemke et al., 1994) and proceed through Src family kinases and Rho kinase, which facilitate endothelial cell retraction in concert with release of junctional complex. This permits circulating platelets direct access to the lamina underlying the endothelial cell margin (Weis et al., 2004). Platelet binding to these substrates activates them, leading to the discharge the contents of their α-granules, which include a variety of molecular effectors ranging from protease/lipases to ECM components to numerous angiogenesis-regulating factors such as VEGF and PDGF (platelet-derived growth factor) or S1P (sphingosine-1 phosphate).
It is important to consider that these immediate-early events in angiogenesis proceed as a concerted intracellular and extracellular cascade. Events within the cells are coordinated to permit changes in the cells microenvironment, and to permit the cell to adapt to those changes. Subsequent signaling within the vascular cells proceeds, releasing the ‘quiescent’ transcriptional program and promotes cell cycle entry, the expression of new cell surface receptors and proteases, and the upregulation of several apoptosis-regulating genes.
Changes to endothelial cell programming during angiogenesis
The switch from quiescent to angiogenic cell that occurs downstream of growth factor signaling is coordinated by changes in homeobox gene expression. In resting endothelial cells, the homeobox genes Gax, HoxA5 and HoxD10 coordinately suppress the expression of proangiogenic genes such as Flk-1, ephrin A1, Hif1α and COX2, whereas promoting expression of LRP1 (a negative regulator of proteolytic activity) and thrombospondin 2, a matricellular protein that inhibits angiogenesis (Myers et al., 2002; Patel et al., 2005). HoxA13 and its targets Ephrin A6 and A7 have also been implicated in maintaining the quiescent status of at least some endothelial cells (Shaut et al., 2007). This programming favors tight interaction with neighboring cells, no degradation of the surrounding ECM and low proliferation.
Deactivation of the quiescent program is initiated by the induction of proangiogenic Hox genes as well as by suppression/retargeting of pre-existing Hox genes. This includes transcriptional repression but also involves the expression of microRNA such as miR130a, which suppress the expression of both GAX and HoxA5 (Chen and Gorski, 2008). HoxD3 was the first proangiogenic Hox gene described. HoxD3 induces the expression of several invasion-promoting genes including ECM receptors such as integrin αvβ3 and proteases such as urokinase plasminogen activator (Boudreau et al., 1997), whereas its paralogs HoxB3 and HoxA3 regulate morphogenesis through its regulation of EphA1 or matrix metalloproteinases and UPA receptor, respectively (Myers et al., 2000; Mace et al., 2005). HoxA9 has also been identified as critical for upregulation of EphB4 during angiogenesis; interestingly, suppression of either HoxA9 or EphB4 block neovascularization (Bruhl et al., 2004). The common recurring theme among these changes in Hox genes is the upregulation of elements that guide cellular interactions with the local microenvironment, including proteases, integrins and eph receptors.
Changes to the vascular cell surface during angiogenesis
Among the changes occurring in the programming on the endothelium, the changes in the expression of cell surface adhesion receptors play prominent and guiding roles in angiogenesis for several reasons. First, the adhesion receptors are in direct contact with the extracellular milieu and are ligated to varying degrees based on counter-receptors on neighboring cells or on ECM content. These receptors are therefore constantly sampling the extracellular environment and conveying signals to the cell in a dynamic manner. Secondly, interactions with the ECM by adhesion receptors, and integrins in particular, modulates signaling events elicited by other cell surface receptors, such as protease-activated receptors and growth factor receptors (Giancotti and Ruoslahti, 1999). The combined interactions regulate downstream targets, which include transcription factors and epigenetic elements; these signals further optimize cellular programming for the current microenvironment. Among the alterations occurring on the endothelial cell surface during angiogenesis, the induction of integrin αvβ3 expression provides a well-studied example of an integrin that dramatically modifies cell behavior.
Integrin αvβ3 as a prototype integrin ‘biosensor’
A principle change following stimulation with angiogenic growth factors is the de novo expression of integrin αvβ3, an oncofetal receptor not commonly expressed in adult tissues, but expressed abundantly on vascular cells selectively during angiogenesis. Integrins are composed of 1 of 18 different α and 8 different β-subunits from at least 24 distinct α/β heterodimers; a single ECM protein can serve as a ligand for several different integrins (for example, laminin is bound by α3β1, α6β1, α6β4, α7β1 as well as by α1β1, and α2β1 to a lesser degree). Most integrin heterodimers on the endothelial cell surface recognize a relatively narrow range of ligands, determined in part by the particular α–β pair being used. For this reason, endothelial cells express a number of different integrins, such as α5β1 (a fibronectin receptor), α2β1 (a collagen receptor) and α6β4 (a laminin receptor). In each case, from tens of thousands to hundreds of thousands of copies of each integrin heterodimer are present on the cell surface.
By contrast, integrin αvβ3 is a relatively promiscuous receptor that recognizes a wide range of ligands. However, each of the ligands recognized by αvβ3 is either a provisional ECM component, such as vitronectin, fibrinogen or fibronectin, or is a fragment of an interstitial/anatomical ECM protein that is produced by proteases during ECM remodeling (including proteolytic fragments of collagen or laminin). Thus, the increase in αvβ3 expression is timed to coincide with alterations in the local ECM that provide a number of αvβ3 ligands. These results, and the fact that αvβ3 was increased in the vasculature of angiogenic endothelium in all chordates examined, suggested that αvβ3 played a critical role in regulating angiogenesis.
This idea was confirmed in early preclinical studies, which demonstrated that blockade of αvβ3 with monoclonal antibodies (Brooks et al., 1994a) or small molecules (Brooks et al., 1994b; Storgard et al., 1999) led to endothelial cell apoptosis and vascular regression. By contrast, no vascular abnormalities were described in human patients with a defective integrin β3 gene (Coller et al., 1991), nor were gross vascular abnormalities observed when the gene was later knocked out in mice (Hodivala-Dilke et al., 1999). It became clear that these endothelial cells were not completely normal (Reynolds et al., 2002; Weis et al., 2007), and in fact, increased neovascularization was observed during pathological forms of angiogenesis. These results were consistent with prior suggestions that antagonized or unligated integrins could signal negatively into the cell, promoting apoptosis (Stupack et al., 2001). The conclusions that we arrived at these studies is that αvβ3 integrin acts as a biosensor, providing prosurvival signals to a cell when the integrin is ligated, but eliciting proapoptotic signals when unligated (or antagonized). This model has subsequently been extended to other integrins, and appears to hold true in a variety of models, depending upon the specific integrin and the cellular context studied (Todorovicc et al., 2005; Stupack et al., 2006; Davis and Senger, 2008).
This integrin model matches that proposed for dependence receptors (Bredesen et al., 2005); those receptors that signal positively in the presence of ligand, but which promote apoptosis when an appropriate ligand is absent. Accordingly, increased expression of a dependence receptor in the absence of an appropriate ligand can promote apoptosis. This situation contrasts with a ‘classical’ growth factor receptor tyrosine kinase such as the EGF receptor or ErbB2, where overexpression per se can be sufficient to mediate cell-signaling events and promote survival. By contrast, absence of a dependence receptor frequently promotes aberrant cell survival. In this respect, it is interesting that mice lacking integrin αvβ3 exhibit increased vascularity during pathological forms of angiogenesis such as tumor growth (Reynolds et al., 2002).
Integrins vary from the dependence model in one particularly important detail; integrins are mechanosensitive receptors (Ingber, 2002; Mammoto et al., 2008). The application of tension increases integrin affinity for ligand, and also serves as a critical factor in modulating integrin-mediated signaling within the cell. In fact, the same ‘ligand’ can function either as an integrin antagonist or agonist, depending upon whether the ligand is soluble or affixed to an immobilized substrate. Translating these in vitro observations in the context of in vivo findings with small molecule inhibitors, it becomes clear why drugs that bind to the ligand-binding site of integrins do not promote productive signaling; they are compromised in their capacity to present an opposing tensile force.
Cell adhesion and ECM degradation as a central mechanism in regulating survival
The requirement for mechanical force to facilitate integrin signaling has additional important implications to the cell. On one hand, productive signaling can be propagated by ligated integrins, either through cyclic application of mechanical tension (such as would be transmitted through normal blood flow) or through the activation of internal contractile forces, such as those effected downstream of the Rho family of small GTPases (Wu et al., 2007a; Mammoto et al., 2008). These signals play an important role in regulating cell survival, particularly when growth factor signaling is limited. On the other hand, growth factor initiated (Giancotti and Ruoslahti, 1999) or mechanically propogated (Li et al., 1997) signaling events depend upon cell anchorage and co-signaling by integrins. This requirement provides a mechanism for attenuation or ablation of signaling events based on the current status of the local ECM microenvironment. Such a mechanism might be expected to be central to a rapidly remodeling ECM, wherein integrin ligation of immobilized ECM components is suppressed (or prevented).
In this respect, integrin mediated adhesion to the ECM can be strongly influenced by the actions of MMPs and serine or cysteine proteases, including UPA, MT1-MMP, soluble MMPs and cathepsins (van Hinsbergh et al., 2006). These enzymes influence cell adhesion by two mechanisms; first and most obvious, they digest the underlying ECM, and thereby compromise stable integrin interactions with it. Secondly, the by-products of ECM digestion are frequently small, soluble ECM fragments that subsequently compete with substrate-immobilized ligand for integrin occupancy. Studies have identified a growing list of these, including angiogstatin, endostatin, kininostatin, restin, tumstatin and vastatin (O'Reilly et al., 1997; Jimenez et al., 2000; Xu et al., 2001; John et al., 2005; Chen et al., 2006; Wu et al., 2007b), and it is likely that many as yet uncharacterized fragments remain to be discovered.
When present in modest concentrations, these integrin ‘antagonists’ can actually promote cell signaling and migration by facilitating integrin ‘turnover.’ The effect can be compared to ‘greasing’ the wheels on a car; integrin binding to ECM is compromised in a very limited manner, permitting somewhat fewer bound integrins on the cell surface at any given time, but increasing integrin cycling through focal adhesion contacts (that is, lowering their residency within these sites). This, increased rate of re-engagement of the ECM, in turn, enhances integrin signaling and promoting cell survival. However, as proteolytic activity increases and the local concentration of soluble protein fragments rises, integrin-ligand binding events are increasingly likely to involve antagonistic fragments rather than immobilized substrates. In this case, positive signaling is compromised, and proapoptotic signaling is elicited. In fact, mice that lack endogenous protease inhibitors, such as the MMP inhibitor PAI-1, often exhibit deficiencies in angiogenesis because of unregulated protease activity (Bajou et al., 1998). Consistent with these results, invasion-promoting proteases, such as MMPs, can also promote apoptogenic signals among invasive endothelial cells (Boger et al., 2001; Ben-Yosef et al., 2005).
The expression of cell surface proteases, such as MT-MMP1, or integrin-associated protease receptors, such as uPAR, is dramatically increased on the surface of angiogenic endothelial cells (van Hinsbergh and Koolwijk, 2008). Integrin αvβ3 serves as a scaffold for MMP-2 (Silletti et al., 2001), reinforcing the notion that the processes of adhesion, migration and proteolysis are tightly linked during vascular cell invasion. Angiogenesis should therefore be viewed as a highly orchestrated process in which proinvasive functions are checked and balanced by cellular apoptotic pathways.
Mechanisms of integrin-mediated regulation of vascular cell apoptosis
The expression of integrins such as αvβ3 compromise cell survival when the cells were placed in an environment that provided no ligands for this integrin; native collagen gels (Stupack et al., 2001). This occurs despite the fact that these cells expressed functional receptors for collagen, permitting cell adhesion to, and migration through, the gels. However, these cells did not ligate integrin β3 and they underwent apoptosis. Thus, antagonism of integrins may not be a prerequisite for apotposis; however, the addition of soluble ligands of αvβ3 to cells within the collagen gels enhances cell death; and these data therefore suggest ‘antagonized’ integrins as more efficient transducers of apoptotic signals than simple, unligated integrins. In fact, we also found this to be true in two-dimensional culture on collagen gels as well; in this case we could rescue the survival of endothelial cells by first cross-linking integrin ligands to the collagen gels.
Integrin αvβ3 antagonists may induce apoptosis among cells cultured on tissue culture plastic; in some cases, apoptosis is initiated selectively by a caspase 8-dependent mechanism, which then promotes cell detachment (Brassard et al., 1999; Stupack et al., 2001; Meerovitch et al., 2003; Erdreich-Epstein et al., 2005; Zhao et al., 2005). In other cases, (particularly when cells are attached to a substrate integrin αvβ3 ligand) detachment can be rapid and precedes the initiation of apoptosis (Ruegg et al., 2002; Kim et al., 2007). Here, cell death proceeds through anoikis, and can involve any number of different apoptotic pathways (Frisch and Screaton, 2001). Accordingly, the anoikis pathway works best when other cell adhesion mechanisms (that is, other integrins and ligands) are limiting to the cell.
Lastly, there is the phenomenon where cells manage to survive despite bearing unligated or antagonized integrins. This case is cell-type and cell-context dependent, and therefore may be the most interesting, as dissection of these survival pathways may give clues as to how integrins regulate apoptosis. For example, endothelial cells attached and spread on the surface of three-dimensional collagen gels are more sensitive to integrin αvβ3 antagonists than cells on two-dimensional collagen surfaces (Stupack et al., 2001). Ongoing studies using similar systems have provided clues as to the specific molecular mechanisms regulating vascular cell survival. These involve both early signaling events and downstream transcriptional targets.
Modification of apoptotic effectors: signaling events at the focal adhesion complex
The type of substrate present for an integrin, as well as the given quantity and type of integrin heterodimers on a cell will influence survival directly. For example, integrin αvβ3 and α5β1 appear to influence each other's ligand-binding activity (Blystone et al., 1999), as do several other integrin pairs, although the precise molecular mechanisms responsible for this may vary from case to case. As suggested above, integrin signaling occurs in a manner dependent on the mechanical rigidity of the substrate (Giannone and Sheetz, 2006). Therefore, the fact that three-dimensional ECM gels and tissue culture plastic differ in their rigidity by several orders of magnitude is significant, and it is perhaps not surprising that focal adhesion complexes that accumulate in tissue-cultured cells are much larger than those found among cells migrating through a three-dimensional ECM (Larsen et al., 2006). The signaling potentiated by these ‘exaggerated’ focal adhesions in two-dimensional cultured cells impacts both the caspase 9- and caspase 8-mediated signaling pathways in two-dimensional cultures, and generally makes cells highly resistant to apoptotic stimuli in vitro. Although apoptosis can be triggered by different initiator caspase pathways, most are influenced by integrin signaling (Stupack and Cheresh, 2002). The end result of an apoptotic stimuli, therefore, will often depend upon both the local ECM milieu and the specific integrin heterodimers that are expressed on the cell surface.
During cell spreading, integrin ligation activates the MAPK pathway downstream of Src, FAK and Ras (Schlaepfer et al., 1994). This pathway is critical for angiogenesis (Eliceiri et al., 1998), and this is in part due to a role in promoting endothelial cell survival. The MAPK pathway provides resistance to starvation-mediated cell death and does so in part through the translocation of Raf to the mitochondria, where it prevents loss of mitochondrial membrane permeability (Alavi et al., 2003). Activation of MAPK opposes also the proapoptotic ‘side effects’ of Rho-kinase activation (Mavria et al., 2006), likely through its capacity to signal to the acto-myosin system (Klemke et al., 1997). Integrin ligation also activates PI3K by Ras and subsequent downstream targets such as p21-activated kinases 1 and 2 (Kiosses et al., 2002; Hood et al., 2003) and Akt1 and Akt2 (Maeshima et al., 2002; Ruegg et al., 2002). The activation of Akt may be a highly potent survival signal, as dominant active forms of Akt protect against a wide range of cellular insults. Although expression of a dominant active kinase is probably not physiologically relevant to endothelial cell biology, such studies nonetheless suggest that lower levels of Akt activation observed in endothelial cells do play some role in maintaining cell survival, particularly under conditions of nutrient deprivation. In these cases, major targets are bcl-2 proteins, p53 (Mayo and Donner, 2002) and mTOR (Maeshima et al., 2002).
Integrin ligation also protects against apoptosis mediated through caspase 8 and the extrinsic cell death pathway. VEGF-mediated signaling is integrin αvβ5-dependent, and both integrin ligation and VEGF-signaling can prevent endothelial caspase 8 activation mediated by death receptors (Ruegg et al., 1998; Alavi et al., 2003; Bieler et al., 2007), whereas the loss of integrin adhesion can sensitize cells to apoptosis mediated by death receptors (Aoudjit and Vuori, 2001). The mechanisms for this are unclear; lack of adhesion may upregulate recpetors for cell surface death receptors. The protection also appears dependent on Src kinase activity (Alavi et al., 2003), which may function by altering death ligand expression or more likely may have a direct role in regulating cell death downstream of the integrins (Figure 2).
In this respect, we previously showed that caspase-8 was recruited to a complex containing (unligated) integrins, where it was activated, leading to apoptosis (Stupack et al., 2001). One might speculate that unligated integrins could therefore induce apoptosis through caspase 8 on their own or might otherwise lower the threshold required for caspase 8 activation by death receptors. Conversely, caspase-8 mediated killing by death receptors, or unligated integrins may be prevented by ligated integrins and focal adhesion complex signaling. In fact, caspase 8 is phosphorylated on tyrosine 380 following growth factor stimulation (Cursi et al., 2006; Senft et al., 2007) or integrin-mediated adhesion (Barbero et al., 2008). Phosphorylation of caspase 8 is mediated by Src family kinases (and possibly other kinases) within an SH2-recognition motif that can be bound by Grb2 and Src SH2 proteins (Barbero et al., 2008), as well as the SH2 domains of the p85 subunit of PI3K (Senft et al., 2007). Phosphorylation of this site inhibits proteolytic activation of the caspase 8 zymogen (Cursi et al., 2006). It is not clear whether this is due to phosphorylation per se, due to SH2-mediated steric blockade of a critical enzymatic site or simply due to sequestration of procaspase 8.
It is worth noting that the link between caspase 8 and integrins appears to function in both directions; integrins can influence caspase 8 activity, and caspase 8 can influence integrin activity. Indeed, among adherent cells, the expression of caspase 8 can promote motility independent of its protease activity (Senft et al., 2007), and phospho-caspase 8 is found to be enriched at the leading edge of migrating cells (Barbero et al., 2008). Therefore, caspase 8 phosphorylation or sequestration by focal adhesion complexes (downstream of ligated integrins) likely represent a mechanism to oppose caspase 8 activation, such as that promoted by death receptors or unligated integrins (Stupack, 2005; Cursi et al., 2006).
In addition to these immediate early interactions that prevent activation of the caspase 8 and caspase 9 signaling pathways, the ligation of specific integrins triggers selective signaling elements that promotes the transcriptional regulation of prosurvival proteins. For example, αvβ3 or α5β1, but not αvβ1, promotes upregulation of protective bcl-2 (Matter and Ruoslahti, 2001), whereas ligation of αvβ3 promotes Bcl-2 expression, but suppresses p53 activity and thereby attenuates Bax and p21WAF1/CIP1 expression (Stromblad et al., 1996). These prosurvival signals appear to depend in part upon the MAPK cascade (Hood et al., 2003) as well as on integrin signaling through the nuclear factor-κB pathway (Scatena and Giachelli, 2002; Courter et al., 2005). The nuclear factor-κB pathway is activated by death receptors, integrins and toll-like receptors—and each of these receptor groups also initiates apoptosis through caspase 8. The fact that caspase 8-deficient cells can be deficient in nuclear factor-κB signaling seems to bind these groups together at the functional level (Zheng et al., 2006). In fact, endothelial cells appear highly dependent upon caspase 8 during development, as selective deletion of caspase 8 with ECs is lethal (Kang et al., 2004).
Interactions with non-integrin receptors that influence vascular cell survival
The endothelial cell surface contains a variety of cell adhesion molecules, including ECM proteins and transmembrane cell adhesion molecules (CAMs). These CAMs belong to a number of families, with prominent molecules on the cell surface belonging to the immunoglobulin and cadherin families. As described above, VE cadherin is phosphorylated and internalized during angiogenesis, and cell surface expression is concomitantly dramatically decreased. This occurs as part of the process that facilitates vascular permeability. By contrast, immunoglobulin superfamily cell adhesion molecules, such as ICAMs, VCAM-1, L1-CAM and neuropilin are increased on the cell surface. VCAM is a ligand for integrins α4β1 and α4β7, where it plays a role in the recruitment of circulating endothelial cell precursors and in regulating cell–cell interactions with stromal cells such as pericytes (Garmy-Susini et al., 2005; Jin et al., 2006a, 2006b). ICAM-1 binds to fibrinogen fragment D (Lominadze et al., 2005) and similar to other ICAMs serves as an endothelial cell ligand for the β2 family of integrins. The β2 integrins are present on hematopoietic cell populations, and play a critical role in the recruitment of circulating cell populations to the angiogenic milieu. ICAMs are upregulated as part of the inflammation associated with angiogenesis, either by chemokines, interleukins or VEGF itself (Croll et al., 2004). L1-CAM is a homotypic adhesion molecule that also serves as a ligand for α5β1 and αvβ3 (Felding-Habermann et al., 1997; Hall and Hubbell, 2004; Hall et al., 2004). Neuropilin plays a central role as coreceptors for VEGFs and semaphorins (Vieira et al., 2007), thereby regulating permeability and angiogenesis (Acevedo et al., 2008), but can also serve as a cellular ligand for integrins (Fukasawa et al., 2007). Most of these immunoglobulins have short cytosolic domains, which interact with adaptor proteins and the actin cytoskeleton; thus, they serve as agonistic integrin ligands due to their capacity to provide an opposing mechanical force. However, each can be dissociated from the cell surface by proteases (Silletti et al., 2000; Garton et al., 2003; Tsakadze et al., 2004; El-Sheikh et al., 2005; Essick et al., 2008); thus, these cell surface adhesion receptors also have soluble forms that are antagonistic towards integrins. Thus, with ECM ligands, shed forms of these CAMs can influence cell survival during angiogenesis.
Integrin conformation and activity
Integrins adapt different conformations that influence their ligand binding and signal-transduction activity. The conformations were originally characterized on the surface of lymphoid cell lines before the ability of integrins to signal was appreciated (Dustin and Springer, 1989; Dransfield et al., 1992; Stupack et al., 1992). However, it was somewhat intuitive that peripheral blood leukocytes required some mechanism by which to regulate their dichotomous functions; on the one hand, present in the circulation, and on the other, arresting and extravasation into the surrounding tissues. This regulation was central to the concept of immune surveillance and lymphocyte trafficking, and has direct relevance to angiogenesis both with respect to the recruitment of endothelial cell precursors and cellular mediators of inflammation (Rose et al., 2007). These studies described a number of conformation-specific antibodies (van de Wiel-van Kemenade et al., 1992; Stupack et al., 1994; Pampori et al., 1999), which revealed that integrins on nonhematopoietic cells were also regulated at the level of conformation and avidity, although not to the same extreme degree as the hematopoietic cell populations originally studied. This regulation of conformation is influenced by the external milieu, and in particular, divalent cations can independently promote integrin maintenance of the most active conformation, as can the presence of bound ligand (Luscinskas and Lawler, 1994).
However, integrin conformation is also influenced by cytosolic effectors. For example, although integrin ligation can promote H-Ras activity and signaling through the MAPK pathway, the activation of H-Ras in turn promotes the acquisition of the less active conformation of integrins by deactivation of the small GTPase Rap (Kinbara et al., 2003; Banno and Ginsberg, 2008). Rap regulates the interaction of the cytoskeleton protein talin with the cytosolic domain of integrins; this may be a common mechanism to regulate integrin function, as paxillin binds to integrin α4 (Rose et al., 2003). This is significant in the context of angiogenesis, because most angiogenic growth factors, including bFGF and VEGF, activate the MAPK pathway through H-Ras and Raf (Eliceiri et al., 1998; Alavi et al., 2003; Hood et al., 2003). Thus, in tandem with positive survival signaling, a feedback mechanism that could promote integrin dissociation from ligand is provided. This may be critical for mobilization of integrin-mediated anchorage in anticipation of cellular migration, and/or it may prime the cells so that survival depends now upon ligation of a smaller subset of integrins. As mentioned above, antagonized integrins (which are, by definition, in the active conformation) are better at promoting apoptosis than unligated integrins. Thus, the regulation of integrin conformation may influence proapoptotic, or ‘negative’ signaling as well.
Aside from growth factor receptors, other cell surface receptor tyrosine kinases also influence integrin-ligand binding. In particular, ephrins are known to regulate integrin activity. Ephrins work in tandem with Eph receptors on neighboring cells. Ephrin-B1 transduces signals to activate integrin-mediated migration, attachment and angiogenesis (Huynh-Do et al., 2002). The downstream signaling pathways appear to use reactive oxygen species as an integration point. EphrinA1 inhibits Rac1 GTPase activity, which promotes LMW-PTP dephosphorylation of p190RhoGAP and contractility through RhoA activation (Parri et al., 2007). This promotes loss of adhesion, as EphA1 also promotes SPAR-mediated deactivation of Rap-1 (Pasquale, 2008). These changes reduce integrin interactions with the ECM, but may also ameliorate, to some degree, their capacity to bind soluble ligands.
Critical roles for hematopoietic, mural and tumor cells in angiogenesis
Endothelial cells interact with the ECM and with other endothelial cells during angiogenesis; however, endothelial cell interactions with other cell types, including pericytes, endothelial cell precursors and inflammatory cells may be at least as important in regulating angiogenesis. In quiescent endothelial cells, cadherins mediate intercellular interactions among ECs, whereas angiopoietin 1 (Ang1) mediates interactions between pericytes and the endothelial cells. The overall levels of cell surface immunoglobulin superfamily CAMs are low, and arrest and extravasation of circulating cells is rare.
Endothelial cells precursors are a circulating population of myeloid cells, which are recruited to sites of angiogenesis in part by the upregulation of cell surface CAMs such as ICAM and VCAM. Recent studies suggest that (Salven et al., 2003; Garmy-Susini and Varner, 2005; Hristov et al., 2007) a measurable fraction of ECs, both in physiological and pathological forms of angiogenesis, can be derived from circulating endothelial precursor cells. The paradigm of leukocyte recruitment to an inflammatory site is mimicked in this process, and EC precursors interact with cell surface CAMS through selectins and integrins; in particular, α4β1 (Jin et al., 2006b). Although it is not yet clear how the survival of circulating endothelial precursor cells is regulated during angiogenesis, it is assumed that these cells do not require adhesion during circulation for survival. Integrin-mediated events may be critical for targeting and activating these cells, at which point they become adherent and dependent upon the underlying ECM as a survival factor (Salven et al., 2003). This requirement is gained as part of the ‘precursor to EC’ differentiation process, as a requirement for ECM interaction during this process has been proposed to occur during this process in vitro and in vivo (Moldovan, 2003); its likely that this process is controlled by master differentiation genes of the ID1/ID3 family, which are essential for ECP recruitment (Lyden et al., 1999). This dependence on adhesion may occur very early in the recruitment process, as small molecule antagonists of integrins limit the accumulation of ECP in angiogenic sites (Loges et al., 2007).
Similar to ECPs, other populations of myeloid cells are recruited to angiogenic sites, and can be found within the interstitium. In particular, tumor-associated macrophages play a significant role in maintaining an inflammatory angiogenic milieu (Porta et al., 2007; Allavena et al., 2008). These cells secrete angiogenic growth factors directly or induce their expression in other cells in response to interleukin-1 or tumor necrosis factor-α. The situation is similar in tumor-like tissue, such as the rheumatoid pannus in rheumatoid arthritis (Szekanecz and Koch, 2007), where macrophage-derived factors provide a significant portion of the angiogenic stimulus. These factors can act as survival factors for the angiogenic endothelial cells. These cells bind to ICAMs and VCAM1 on the surface of endothelial cells through integrin-mediated interactions, subsequently invading the loal ECM. Within this milieu, tumor-associated macrophages further contribute by expressing ECM components, such as fibronectins. However, the prosurvival signals are complemented by the potentially apoptotic signals arising from ECM degradation by uPA and MMPs produced by these cells. In some cases, the leukocytes will even trigger a caspase cascade directly, through presentation of death ligands such as Fas (Ishida et al., 2003).
The critical role of the pericyte in angiogenesis has become appreciated in recent years. Pericytes play a critical role in stabilizing the vasculature through interactions at sites of pericyte: EC contact. These contacts involve interactions through the receptor tyrosine kinase Tie2 and Ang1 present within the quiescent vasculature (Fukuhara et al., 2008; Saharinen et al., 2008). After stimulation with an angiogenic growth factor, endothelial cell secretion and accumulation of angiopoietin 2 (Ang2) leads to direct competition with Ang1 for binding to the Tie2 receptor, disrupting the EC-pericyte junction. This leads to pericyte to retraction and facilitates endothelial cell invasion of the local interstitial matrix. Ang2 seems to function as more than a simple decoy ligand, as Ang2 binding to the Tie2 receptor can function to promote signaling in certain cell types, although not among ECs (Shim et al., 2007). Integrins act as a secondary receptor for Ang2, and Ang2 can promote cell invasion through integrin-mediated signals (Imanishi et al., 2007). Integrins also mediates some Ang1-mediated survival signaling, at least in myocytes, (Dallabrida et al., 2005) but this does not appear to be the case for all cells (Brindle et al., 2006). It remains unclear whether integrin ligation is required for Tie2-mediated tyrosine kinase activity, as it is for other receptor tyrosine kinases. Nonetheless, integrin binding to Ang2 could play a role in sequestering it, thereby favoring Ang1 interactions with Tie2.
These interactions are critical for the resolution of physiological angiogenic processes, as re-entry into a quiescent state and vessel maturity ultimately requires pericyte interaction. Thus, its perhaps not surprising that PDGF has been identified as an angiogenic growth factor. PDGF does not act directly upon endothelial cells, which lack PDGF receptors, but rather activates pericytes and other mural cells to produce VEGF secretion; this mechanism is analogous to the induction of VEGF by inflammatory mediators. VEGF is then capable of maintaining cell survival to some degree, although VEGF alone is insufficient for vascular maturation. What is somewhat surprising is that although PDGF promotes pericyte proliferation, endothelial cells numbers far exceed pericytes numbers in an angiogenic environment. Given that pericytes are ultimately critical to blood vessel stabilization and maturation, it is tempting to speculate that some mechanism exists that limits pericyte accumulation, thus preventing ‘overvascularization’ of tissues. Supporting this notion, the addition of VEGF or angiogenic growth factors can transiently induce angiogenesis in vivo, but is typically insufficient for a lasting (that is, mature) angiogenic response in vivo (Gounis et al., 2005).
Endothelial cell apoptosis as a natural conclusion to angiogenesis; integrating signals
Why should there be a balance that ultimately favors apoptosis? During angiogenesis events, increased vasculature accumulates relative to the surrounding tissues, although composed of immature vessels. However, in a physiological angiogenesis event, this is a temporary situation, and the natural process termed ‘vascular pruning’ subsequently occurs, wherein ‘extraneous’ blood vessels undergo apoptosis (Figure 3). In fact, apoptosis can be detected among endothelial cells during angiogenesis before pruning, and the ultimate dieback, which occurs is programmed into the cells. Angiogenic blood vessels require growth factor stimulation, which is dependent upon ongoing interactions with their microenvironment through integrins; active death is induced in those cells, which are in a highly proteolytic microenvironment wherein integrins are antagonized, or among cells in a microenvironment inappropriate to their integrin complement. In the pathological angiogenesis scenario, present within a tumor, the rheumatoid pannus or a diabetic retina, this is essentially an ongoing state, wherein cells are constantly proliferating and dying, remodeling their microenvironment and in turn being dependent on their addiction to growth factors and integrin-mediated signaling.
In these cases, the maturation associated with physiological angiogenesis does not occur. Maturation involves a depletion of growth factors, replacement of the provisional insterstitial ECM with an anatomically appropriate ECM and the reestablishment of pericyte contacts. In fact, depletion of growth factors such as VEGF using therapeutic antibodies or decoy receptors does induce endothelial apoptosis, yet ultimately results in establishment of a stable vasculature. Could the effect of these drugs simply result from facilitating a natural conclusion to the angiogenic event?
Ultimately, the fact that so many interacting factors balance the decision between life and death in angiogenic ECs leads one to the conclusion that angiogenic endothelial cells are heavily ‘loaded’ with respect to molecules determining cell fate. A given endothelial cell derives numerous signals that promote apoptosis, and thus becomes dependent upon survival factors to counteract them. This lends new credence to the concept originally proposed by Dr Folkman; interventions that impact a single factor strongly enough may be sufficient to ‘tip the balance’ irrevocably towards EC apoptosis (Folkman, 2006).
Conclusions and therapeutic outlook
Despite spectacular results in a variety of preclinical models, antiangiogenic agents have so far met with limited success in the clinic. Dramatic advances have been achieved in ocular indications using anti-VEGF-based approaches as single therapies (Algvere et al., 2008), whereas in the cancer clinic no single-agent activity has been achieved (Grothey and Ellis, 2008). In part, one might attribute this to the fact that human tumors display high levels of ongoing VEGF-production, whereas ablating VEGF activity in the retina can have a more significant impact on endothelial proliferation. But the answer may be far simpler; tumors may make more blood vessels than they need. Thus, ablation of VEGF signaling leads many vessels to die, but in doing so promotes maturation of the others (Willett et al., 2004; Winkler et al., 2004). In the back of the eye, this will restore the normal (pre-existing) physiology. However, it remains unclear how this plays out in the tumor microenvironment. Immature and leaky tumor vasculature is highly inefficient at drug deliver, but mature vessels yield sufficient increases in drug delivery that the lower vascular volume is inconsequential (Dickson et al., 2007). By reducing VEGF in the tumor blood vessels show increased maturation and blood flow associated with enhanced pericyte coverage. This may facilitate drug delivery to tumors and explains in part the need to give anti-VEGF therapy in combination with cytotoxic drugs (Willett et al., 2004)
Can integrins, as the critical receptors for the ECM and central regulators of survival and apoptosis, be exploited as targets as well? It would appear to be possible. Small molecule antagonists of integrins, as well as humanized monoclonal antibodies, have shown success in preclinical models, and appear safe in clinical trials (Beekman et al., 2006) (with some reports of dramatic effects as single agents) (Raguse et al., 2004). But this story is very similar to that of the VEGF-inhibitors, and trials with additional drugs are likely warranted to recaptitulate synergies already seen in preclinical models. One advantage of targeting integrins may be that many growth factor signaling pathways are dependent upon them, and disrupting integrins can thereby disrupt growth factor signaling (Eliceiri et al., 1999). Conversely, one possible limitation of the VEGF-inhibitors is that a tumor may eventually develop a significant capacity to induce angiogenesis through other growth factors, such as bFGF, rendering VEGF-targeted therapies less effective. Moreover, as EC recruitment or survival requires integrins (Bussolino et al., 2006; Kohl et al., 2007), it is possible that small molecule antagonists could also block EC accumulation, providing another antiangiogenic mechanism.
There have been many surprises arising from recent studies of angiogenesis. In spite of our recent revelations, angiogenesis remains a highly complex biological process, and ultimately poorly understood process. In fact, we are still only beginning to understand the effects of how the regulation of vascular cell survival can ultimately influence disease outcome. Nonetheless, initial results continue to be encouraging, and the recent direct evidence of impact in the clinic provides excitement for the future prospects of controlling neovascularization.