Molecular basis for endothelial lumen formation and tubulogenesis during vasculogenesis and angiogenic sprouting

2.1. Fundamental role for extracellular matrices and 3D matrix environments for endothelial lumen formation mechanisms

ECM is a critical regulator of vascular tube morphogenesis that serves as a signaling scaffold to affect survival, motility, invasion and morphogenic events for endothelial cells (ECs) and mural cells, including pericytes and vascular smooth muscle cells (Arroyo and Iruela-Arispe, 2010; Davis and Senger, 2005; Davis and Senger, 2008; Germain et al., 2010; Hynes, 2007; Rhodes and Simons, 2007; Senger and Davis, 2010). Different types of ECM have unique effects (i.e. basement membrane matrices vs. interstitial matrices vs. provisional matrices) depending on the biologic context with evidence for pro-morphogenic, pro-stabilization and pro-regressive activities (Davis and Senger, 2005; Davis and Senger, 2008; Senger and Davis, 2010). Certain ECM environments such as vascular basement membrane matrices appear to present quiescence signals to vascular cells to facilitate vascular stabilization. In contrast, ECM alterations through proteolysis or conformational changes (i.e. which exposes matricryptic sites) can activate vascular cells. Thus, ECM is a critical regulator of how cells within the vascular wall sense an injurious stimulus (Davis, 2009; Davis et al., 2000) and then acutely respond by affecting processes such as vascular permeability, inflammation, angiogenesis and thrombosis (Arroyo and Iruela-Arispe, 2010; Davis and Senger, 2005; Senger and Davis, 2010). The ECM also is a critical regulator of chronic EC responses to long-standing injurious stimuli such as that observed in diseases such as diabetes, atherosclerosis and hypertension. In all of these latter cases, considerable ECM remodeling occurs that affects vascular cell function and behavior (Greene et al., 1989; Hill et al., 2007; Schiffrin, 2004). The transition from quiescent ECs to activated ECs (which then undergo morphogenesis) has been functionally linked to degradation of the vascular basement membrane matrix to induce angiogenesis in response to a variety of stimuli. In this instance, the signaling events are likely to be complex with loss of inhibitory signals, gain of activating signals or both to induce ECs to undergo sprouting and tube morphogenesis.

The ECM scaffold possesses adhesive signals for cells by binding to cell surface adhesion receptors but it also presents specific growth factors to cells (Davis and Senger, 2005; Hynes, 2009; Somanath et al., 2009). A critical point is that different growth factors show affinity for unique ECM proteins present in the basement membrane matrix, interstitial matrix or elastin-rich microfibrillar matrix (Hynes, 2009; Ramirez and Dietz, 2009; Senger and Davis, 2010). Furthermore, the ECM modulates the activation of specific growth factors and thus, can modulate growth factor action to control vascular development or pathologic neovascularization (Hynes, 2009). Also, co-signaling between integrins and growth factor receptors are known to control vascularization responses during development and postnatal life (Hynes, 2009; Senger and Davis, 2010; Somanath et al., 2009). Interestingly, modulation of ECM through proteolysis or application of cellular mechanical force to ECM, which regulates matrix assembly and remodeling, may control the organization of growth factor and ECM interactions which then affect vascular signaling under normal or diseased conditions.

Particular ECM components are potent stimulators of vascular tube morphogenesis while others appear to be inhibitory. Collagen type I, the most abundant ECM component particularly in adult animals, and fibrin are potent pro-morphogenic stimuli for vascular tube morphogenesis in 3D matrices (Figures 1–3) (Aplin et al., 2008; Bayless et al., 2009; Bayless et al., 2000; Davis and Camarillo, 1996; Koh et al., 2008b; Nakatsu and Hughes, 2008; Whelan and Senger, 2003). The collagen-binding integrin, α2β1 has been shown to control EC tube morphogenic events in vitro and in vivo in collagenous matrices (Bayless and Davis, 2003; Davis and Camarillo, 1996; Davis and Senger, 2005; Senger et al., 1997) while the fibrin/fibronectin-binding integrins, α5β1 and αvβ3, have been shown to control tube morphogenesis in fibrin matrices (Bayless and Davis, 2003; Bayless et al., 2000). Due to the strong pro-morphogenic influence of collagen and fibrin matrices, these have been primarily used to establish EC tube morphogenic models in 3D ECM environments (Figures 1–3) (Aplin et al., 2008; Bayless et al., 2009; Koh et al., 2008b; Nakatsu and Hughes, 2008). Such models have strongly enhanced our knowledge concerning the molecular basis for EC tubulogenesis, sprouting, and tube maturation events. Overall, the ECM and integrin data demonstrates that vascular tube morphogenesis is directly connected to integrin-mediated recognition of these pro-morphogenic ECM components in 3D matrices.

2.2. Relevance of lumen formation mechanisms to vascular morphogenesis in development and disease

It is clear that the mechanisms underlying EC lumen and tube formation are necessary in order for EC networks to function as conduits to propel blood flow and to perform as exchangers for oxygen, carbon dioxide, nitric oxide and nutrients (Figure 4). This fundamental function of EC tubes is required for a closed circulatory system connected to a heart pump (Wagenseil and Mecham, 2009). Lumen formation mechanisms are also critical for the development of the lymphatic vasculature, although our understanding of these events is less developed (Adams and Alitalo, 2007). Also, the connection of blood EC vascular tubes with a pressurized and high flow system suggests that lumen remodeling events can occur when such flow responses are initiated or changed (Iruela-Arispe and Davis, 2009). This is a property of blood EC tubes that distinguishes it from other types of biological tubes such as epithelial tubes or less pressurized lymphatic EC tubes (Iruela-Arispe and Davis, 2009).

Recent work concerning the tumor vasculature is illustrative of how lumen formation mechanisms play a fundamental role in tissue perfusion that affects tumor cell growth and progression. The tumor cell vasculature is known to be abnormal and a functional microcirculation is absent within this microenvironment. Blood flow is abnormal, tissue interstitial pressure is elevated and there is a fundamental lack of appropriate vascular remodeling events necessary to create a functional microcirculatory circuit (Jain, 2005). Interestingly, tumor cells tend to overproduce VEGF which leads to increased expression of delta-like-4 (DLL4), a Notch inhibitor, on tumor ECs (Noguera-Troise et al., 2006). DLL4 has recently been found to control EC sprouting through its expression on EC tip cells (i.e. invasive cells at the leading front of sprouts) (Gerhardt et al., 2003; Gridley, 2010; Hellstrom et al., 2007; Holderfield and Hughes, 2008; Sainson et al., 2005) (Figures 1 and 2). When DLL4 is inhibited within the tumor microenvironment, there is a marked increase in angiogenic sprouting (Noguera-Troise et al., 2006). However, these treated tumors are smaller despite the increase in vascularity (Noguera-Troise et al., 2006; Thurston et al., 2007). Thus, although sprouting is markedly increased, the tumor is very hypoxic and there appear to be few functional EC lumen and tube networks within these tumors (Noguera-Troise et al., 2006). This work suggests that inhibition of this molecule within the tumor microenvironment creates a functional disconnect or discoordination of EC tip cell formation from EC lumen formation mechanisms (Benedito et al., 2008; Hellstrom et al., 2007; Noguera-Troise et al., 2006; Thurston et al., 2007), suggesting that the signaling stimuli for these cellular events are distinct. Thus, the lack of a functional microcirculation within tumors is responsible in part for the inability of anti-tumor drugs to penetrate effectively into this microenvironment (Jain, 2005). These findings have lead to the concept of normalization of the tumor vasculature (Jain, 2005) which could increase tumor perfusion and the effectiveness of tumor treatment regimens. A very recent study demonstrates that modest inhibition of GSK-3β, leads to increased vessel normalization by increasing microtubule stability within EC-lined tubes (Hoang et al., 2010c). Similarly, the same group demonstrated that delivery of either activated Cdc42 or separately, Rac1, leads to normalization of vessels in the face of chronic VEGF delivery within tissues (Hoang et al., 2010a; Hoang et al., 2010b), a significant problem within the tumor microenvironment. Thus, one strategy to normalize these abnormal tube networks is to modulate Rho GTPases by affecting the microtubule and actin cytoskeletons (Hoang et al., 2010a; Hoang et al., 2010b; Hoang et al., 2010c). Overall, the organization of the vasculature is fundamentally characterized as a connected series of vascular tubes whose establishment and maintenance depends on a variety of factors and mechanisms. These mechanisms include tube formation, flow responsiveness and tube stabilization that involve homotypic EC-EC adhesion signaling, heterotypic EC-mural cell interactions and EC interactions with vascular basement membrane matrices (Davis et al., 2007; Dejana et al., 2009; Iruela-Arispe and Davis, 2009; Lampugnani and Dejana, 2007; Mazzone et al., 2009; Senger and Davis, 2010; Wagenseil and Mecham, 2009; Warren and Iruela-Arispe, 2010). Thus, an understanding of the molecular mechanisms underlying these separate processes is critical to be able to appropriately manipulate vessels in a positive or negative direction for therapeutic applications in human disease.

2.3. Different model systems and species to investigate endothelial lumen and tube formation

A major effort of our laboratory as well as others has been to elucidate the molecule and signaling requirements for human ECs to form vascular tube networks in 3D matrices (which mimics vasculogenic tube assembly) (Davis et al., 2007; Davis and Senger, 2005; Holderfield and Hughes, 2008; Iruela-Arispe and Davis, 2009). Similarly, we have performed assays using ECs attached to a monolayer surface which then sprout to form tubes in 3D matrices (Bayless and Davis, 2003; Koh et al., 2008b) (Figures 1–3). Other laboratories have developed related systems which complement those developed by our group (Alajati et al., 2008; Aplin et al., 2008; Bayless et al., 2009; Korff and Augustin, 1998; Nakatsu and Hughes, 2008; Whelan and Senger, 2003). In addition, a number of laboratories have developed EC-pericyte or EC-fibroblast coculture systems that mimic EC-mural cell interactions that control tube assembly, remodeling and stabilization (Aplin et al., 2008; Bryan and D'Amore, 2008; Howson et al., 2005; Korff et al., 2001; Liu et al., 2009; Nakatsu and Hughes, 2008; Stratman et al., 2009a; Stratman et al., 2010). Overall, these 3D matrix microassay systems have allowed for detailed analyses of these events.

Considerable advances in our understanding of the molecular basis for EC lumen and tube formation have been obtained using in vitro models of this process with human ECs in 3D matrices (Davis et al., 2007; Egginton and Gerritsen, 2003; Holderfield and Hughes, 2008; Iruela-Arispe and Davis, 2009). Also, work using a variety of in vivo models including Zebrafish, mouse and quail has lead to important advances delineating these events as well (Drake et al., 1992; Egginton and Gerritsen, 2003; Iruela-Arispe and Davis, 2009; Kamei et al., 2006; Rupp et al., 2003; Sato et al., 2010; Strilic et al., 2009; Wang et al., 2010; Zovein et al., 2010). Lumen formation mechanisms have been investigated using 3D matrix models since EC lumen formation and tubulogenesis is a 3D matrix-specific process that does not occur on 2D matrix surfaces. Important advances have been made in the technologies necessary to approach such questions in terms of manipulating gene expression by transduction with viral vectors, suppression of gene expression using siRNA technology or morpholino oligonucleotides, performing signal transduction experiments in 3D matrix environments with primary human ECs, and performing time-lapse light and fluorescence microscopy to characterize the dynamic cellular morphologic changes that are necessary to form tube networks (Davis et al., 2007; Filla et al., 2004; Herbert et al., 2009; Iruela-Arispe and Davis, 2009; Kamei et al., 2006; Koh et al., 2008a; Koh et al., 2009; Larina et al., 2009; Lee et al., 2009; Lucitti et al., 2007; McKinney and Weinstein, 2008; Sacharidou et al., 2010; Saunders et al., 2006; Senger and Davis, 2010; Stratman et al., 2009a; Yaniv et al., 2006; Zamir et al., 2006). In addition, broad screening strategies using DNA microarray analyses and proteomic analyses as well as chemical modulatory screens are being used to understand vascular morphogenesis (Bell et al., 2001; Del Toro et al., 2010; Kalen et al., 2009; Lilly and Kennard, 2009; Sacharidou et al., 2010; Strasser et al., 2010; Su et al., 2010; Su et al., 2008).

It is clear that these developing technologies have lead to rapid advances in our understanding of EC lumen formation and tubulogenesis in 3D matrix environments. This work led to the primary discovery that Cdc42 controls lumen formation which was first demonstrated in ECs (Bayless and Davis, 2002) (Figure 5). In addition, Rac1, integrins such as α2β1 (in collagen matrices), α5β1 and αvβ3 (in fibrin matrices), MT1-MMP, Pak2, Pak4, Par3, Par6b, Jam-B, Jam-C, VE-cadherin, cerebral cavernous malformation protein (CCM)1, CCM2, atypical PKC, PKCε, Src family kinases, B-Raf, C-Raf and Erk kinases have all been shown to be required for EC lumen formation and tubulogenesis (Bayless and Davis, 2002; Kiosses et al., 2002; Koh et al., 2008a; Koh et al., 2009; Lampugnani et al., 2010; Sacharidou et al., 2010; Saunders et al., 2006; Stratman et al., 2009b; Whitehead et al., 2009; Yang et al., 1999) (Figure 6). Of great interest is recent work that demonstrates the presence of an EC lumen signaling complex of proteins that are physically connected to control EC tubulogenic events in 3D matrices (Blobel, 2010; Sacharidou et al., 2010) (Figure 6). These studies are examples of how rapidly the field of lumen formation in ECs has advanced over the last decade and this has been largely due to experiments performed using in vitro systems. It also should be pointed out that many new gene targets that control lumen formation such as CCM2 and CCM1 (and which are mutated in human diseases) have been demonstrated to play a role in lumen formation and vascular development (Gore et al., 2008; Kleaveland et al., 2009; Lampugnani et al., 2010; Liu et al., 2010; Whitehead et al., 2009). Knockouts of these genes in mice or morpholino knockdown of genes in Zebrafish lead to lumen defects in developing vessels for both CCM2 and CCM1. Importantly, both have been demonstrated to be involved in human EC lumen formation in 3D collagen matrices further demonstrating the powerful technology that these assay systems bring (Lampugnani et al., 2010; Whitehead et al., 2009). Similarly, recent experiments in mice and Zebrafish also confirm the role of Pak2, Pak4 and Par3 (Galan Moya et al., 2009; Liu et al., 2007; Tian et al., 2009; Zovein et al., 2010), the latter two were first demonstrated using in vitro models in 3D collagen matrices (Koh et al., 2008a). Pak2 was also demonstrated to play a critical role during these processes in vitro (Koh et al., 2008a). Thus, it is becoming increasingly clear that particular in vitro models are very accurately predicting the role of genes that control EC lumenogenesis in vivo. This was also the case with respect to the functional role of Notch signaling and EC sprouting and tip cell formation which was elegantly shown first using human ECs and an in vitro fibrin morphogenesis system (Sainson et al., 2005). Inhibition of Notch led to increased sprouting with marked increases in tip cell formation and increased EC proliferation (Sainson et al., 2005), which are also observed in vivo when Notch is inhibited (Leslie et al., 2007; Siekmann and Lawson, 2007; Uyttendaele et al., 1996). Later these studies were extended primarily using the in vivo postnatal retinal sprouting model or Zebrafish intersegmental vessels where delta-like-4 (DLL-4) was shown to be produced by tip cells and inhibit stalk cells through Notch signaling which trail the invading sprout (Hellstrom et al., 2007; Noguera-Troise et al., 2006; Siekmann and Lawson, 2007). Importantly, many of the fundamental concepts in these later studies were described first using the fibrin in vitro model system (Sainson et al., 2005). As more quality in vitro and in vivo systems are merged to investigate molecular mechanisms underlying vascular morphogenesis, the faster the advances in the field will proceed.

2.4. Relationships of EC lumen formation to epithelial lumen formation- similarities and key distinctions

It is apparent that both ECs and epithelial cells are able to make cell-lined tube networks, but these tubes are functionally very distinct. Both sets of tube networks serve as conduits for transport of materials, but interestingly, epithelial cells and their tubes are by their nature either secretory or absorptive and thus, have very specialized apical and basolateral protein targeting mechanisms to control these events (Bryant et al., 2010; Bryant and Mostov, 2008; Hao et al., 2010; Jaffe et al., 2008; Lubarsky and Krasnow, 2003; Madrid et al., 2010; Martin-Belmonte et al., 2007). Epithelial cell-cell junctions are unique and are more specialized than typical EC-EC junctional contacts (Dejana et al., 2009; Iruela-Arispe and Davis, 2009; Lampugnani and Dejana, 2007). In general, the membrane contact area for epithelial cell junctions is more extensive than for EC junctions which most likely plays a role in specializing these membranes for the apical-basolateral sorting that is required for selective secretion and absorptive functions (Bryant et al., 2010; Bryant and Mostov, 2008; Gassama-Diagne et al., 2006). Another critical difference is the fact that blood EC tubes are exposed to significant flow and blood pressure forces (Culver and Dickinson, 2010; Iruela-Arispe and Davis, 2009; Lucitti et al., 2007; Segal, 2005; Wagenseil and Mecham, 2009). Lymphatic EC tubes are exposed to lymph fluid and represent a lower pressure circuit. EC tubes are primarily conduits for plasma, blood cells or lymph and are not typically secretory in nature, thus, in combination with their exposure to high flow and blood pressure (i.e. in blood EC-lined tubes), suggests very different functions as it relates to their junctions as well as apical-basolateral polarity mechanisms. Another interesting difference is the presence of mural cells on the basal surface of most EC-lined tubes along with vascular basement membranes particularly in blood EC-lined tubes. The presence of mural cells as well as the vascular basement membrane matrix may be the defining feature that characterizes the basal versus apical surface of ECs (Stratman et al., 2009a; Stratman et al., 2010). We have recently shown that pericyte recruitment to EC-lined tubes stimulates vascular basement membrane matrix assembly selectively on the basal surface, a surface that pericytes also occupy (Stratman et al., 2009a; Stratman et al., 2010) (Figures 1, 7 and 8). In addition, pericytes migrate along this surface during basement membrane assembly events (Stratman et al., 2009a). Thus, the EC apical and basal surfaces are distinct and can be easily distinguished during in vitro tubulogenesis when ECs and pericytes are co-cultured in 3D matrices (Figures 1, 7 and 8). These surfaces are also readily identifiable by the presence of ECM on the basal surface as well as a fluid compartment exposed to the apical surface (Figures 1, 3, and 7). Thus, the lumen formation process rapidly creates such polarized surfaces in 3D matrices using the EC lumen signaling complexes and kinase signaling cascades mentioned above (Figure 6).

There is considerable evidence that epithelial lumen formation involves apical targeting of proteins such as Cdc42, Par3, annexin II, PIP3, and PTEN (Bryant et al., 2010; Bryant and Mostov, 2008). In addition, there is apical targeting of membrane to expand the lumenal surface during tubulogenic events. More recent studies have identified a complex of proteins involving Rab11 and Rab8 which facilitate the targeting of vesicles to this developing apical surface in a Cdc42-dependent process (Bryant et al., 2010). Recent studies have also identified that two Cdc42 guanine exchange factors (GEFs), Tuba and Intersectin 2, play a role in formation of epithelial cell lumens and proper positioning of mitotic spindles in 3D matrices (Hao et al., 2010; Qin et al., 2010; Rodriguez-Fraticelli et al., 2010).

One question is how related EC lumen formation is with respect to these processes in epithelial cells. We have shown a critical functional role for Cdc42 in EC lumen formation as well as its effectors, Pak2, Pak4 and Par6b (Koh et al., 2008a; Sacharidou et al., 2010). Ongoing studies are focused on identifying the relevant Cdc42 GEFs that are involved in the controlling EC lumen and tube formation (Sacharidou and Davis, in preparation), but thus far these appear to be distinct from those utilized by epithelial cells. Previously, we implicated FGD4 (a Cdc42-selective GEF) (Ono et al., 2000) in this process due to the fact that an FGD4 siRNA disrupted lumen formation in a manner similar to siRNA suppression of Cdc42 (Davis et al., 2007). Finally, we have shown that Cdc42 targets to membrane surfaces including intracellular vacuoles that control EC lumen formation (Bayless and Davis, 2002; Kamei et al., 2006) and also, Cdc42-GTP is a key component of EC lumen signaling complexes which links to Par6b, Par3, JamB, JamC, MT1-MMP and α2β1 and which together control EC lumen and tube formation (Sacharidou et al., 2010). Also, due to the fact that Rac1-GTP can also interact with Par6 isoforms (Brazil and Hemmings, 2000), it is likely that it along with Cdc42-GTP mediates the function of this signaling complex to control EC tubulogenesis. Importantly, both Cdc42 and Rac1 control EC lumen formation and Rac1 appears to be a component of the EC lumen signaling complex that we have recently identified (Sacharidou and Davis, unpublished observations).

Despite the clear apical membrane targeting of specific proteins in epithelial cells during lumen formation, this has been much more difficult to demonstrate in ECs during these events. Past and recent studies demonstrate apical targeting of podocalyxin in EC tubes both in vivo and in vitro (Dekan et al., 1990; Lampugnani et al., 2010; Strilic et al., 2009). How critical this targeting event is for the lumen formation process remains unclear. Podocalyxin along with cortical actin, moesin and other proteins are strongly targeted to apical surfaces during epithelial tube formation (Martin-Belmonte et al., 2007; Martin-Belmonte et al., 2008) but this is more difficult to demonstrate during EC tubulogenesis. One study showed that VE-cadherin based adhesions play a role in podocalyxin targeting to the apical surface in a manner dependent on CCM1 (Lampugnani et al., 2010). Furthermore, a recent study demonstrated modest moesin targeting to apical membranes during Zebrafish intersegmental vessel tube formation (Wang et al., 2010). This study also revealed the presence of intracellular vacuoles (labeled with GFP-moesin) that were present and targeted to this apical surface (Wang et al., 2010). More work is necessary to understand how EC apical targeting of proteins is controlled and how it relates to EC junction formation during morphogenic events. Also, the question of whether EC-EC junctional contacts are primary stimulators of EC lumenogenesis has been raised in a few studies (Strilic et al., 2009; Wang et al., 2010) and in our view, this remains an open question. Although it has been suggested to be a primary requirement for lumen formation (Strilic et al., 2009; Zeeb et al., 2010), this result is based on a central assumption that molecules such as VE-cadherin act solely or primarily as cell-cell adhesion molecules. In fact, single ECs can form lumen structures, without the need for contact with adjacent ECs (Davis and Camarillo, 1996). Jam proteins, which are known to participate in cell-cell adhesion of ECs (and of ECs and leukocytes), play a fundamental role in EC lumen formation from either single or aggregated ECs (Sacharidou et al., 2010). In fact, JamA has been reported to affect chemotactic responses of neutrophils migrating as single cells (Corada et al., 2005). Furthermore, it is clear that VE-cadherin has diverse effects on other EC functions including survival, maturation, proliferation, flow-induced EC cytoskeletal arrangements and signal transduction (Dejana et al., 2009; Harris and Nelson, 2010; Lampugnani and Dejana, 2007; Tzima et al., 2005). Clearly, it is important to examine these issues in more detail and assess whether VE-cadherin may act independently of cell-cell junction assembly to influence EC lumen and tube formation. VE-cadherin has been more convincingly linked to EC-EC junctional stability to maintain EC tubes that are formed (Crosby et al., 2005; Dejana et al., 2009; Montero-Balaguer et al., 2009; Rudini et al., 2008) and this function may also be particularly critical under flow and pressure conditions. Thus, it remains to be seen whether the vascular morphogenic defects observed in VE-cadherin mouse knockouts or morpholino knockdown Zebrafish are due to deficiencies in cell-cell adhesion that prevent lumen formation or actually represent defects in other critical signaling pathways that interfere with EC tube morphogenesis independently of effects on cell-cell adhesion.

3.1. Role of intracellular vacuoles and intercellular adhesion mechanisms in lumen formation

In a review on tubulogenesis a number of years ago, Krasnow and colleagues suggested five different mechanisms of lumen and tube formation including intracellular lumen formation and cord hollowing which appear to be the two major mechanisms controlling EC lumen formation (Lubarsky and Krasnow, 2003) (Figure 4). These two primary mechanisms appear to play a role in both initiating and sustaining EC lumen and tube formation (Davis and Bayless, 2003; Davis et al., 2007; Iruela-Arispe and Davis, 2009). An important consequence of either mechanism by which ECs form tubes is that they will create physical spaces within the ECM termed vascular guidance tunnels (Stratman et al., 2009b) (Figures 3 and 8). Tunnel formation appears to be directly coupled to the signaling pathways that control tubulogenesis in 3D matrix environments (Sacharidou et al., 2010; Stratman et al., 2009b).

Recent controversies have arisen over whether intracellular lumen formation mechanisms are utilized by ECs to form lumens and tubes (Blum et al., 2008; Davis et al., 2002; Davis et al., 2007; Iruela-Arispe and Davis, 2009; Kamei et al., 2006; Strilic et al., 2009). In fact, some investigators have stated that they are either not involved or are not relevant for lumen formation (Blum et al., 2008; Strilic et al., 2009; Zeeb et al., 2010). It is important to consider this issue by examining multiple studies from separate laboratories that involve both in vitro and in vivo analyses of these events. When this issue is examined critically and when both types of studies are considered, intracellular vacuoles clearly have been shown to play a role during EC lumenogenesis. Vacuoles have been demonstrated to be present during EC lumen formation both in vitro and in vivo (Bayless and Davis, 2002; Bayless et al., 2000; Davis and Bayless, 2003; Davis et al., 2002; Davis and Camarillo, 1996; Folkman and Haudenschild, 1980; Kamei et al., 2006; Liu et al., 2010; Wang et al., 2010; Yang et al., 1999; Zovein et al., 2010). Disruption of either moesin or CCM1 in Zebrafish leads to a decrease in intracellular vacuoles within ECs in vivo, and in both cases, these treatments result in marked deficiencies in EC lumen and tube formation (Liu et al., 2010; Wang et al., 2010). Also, blockade of intracellular vacuole formation in human ECs using anti-integrin antibodies or disruption of Cdc42 or Rac1 signaling markedly inhibits EC lumen formation in 3D matrices (Bayless and Davis, 2002; Bayless et al., 2000; Davis and Camarillo, 1996). It has also been demonstrated that either GFP-Cdc42 or GFP-moesin label intracellular vacuoles and furthermore, vacuoles can be demonstrated to dynamically form and fuse during lumen formation events in developing Zebrafish intersegmental vessels (Bayless and Davis, 2002; Kamei et al., 2006; Wang et al., 2010). Multiple independent groups have observed intracellular vacuoles during tube formation events in Zebrafish, mouse, and avian species. Furthermore, they have been demonstrated to play a critical role during lumen formation using human ECs in 3D matrix environments in vitro. Images of EC tubes during quail vascular development at E5 shows obvious evidence of intracellular vacuolation of ECs within tube walls and also with ECs that appear to be joining tubes during multicellular tube formation (Figure 9). Thus, intracellular vacuoles are observed in vitro and in vivo during EC tube morphogenic events and clearly play a functional role during lumen formation and tube assembly.

Although vacuoles are present in ECs and play a role during these events, they are clearly not the sole mechanism which controls lumen and tube formation. For example, it has been demonstrated that EC lumen expansion from single or aggregated ECs involves cell surface MT1-MMP activity, a mechanism that is clearly distinct from the intracellular vacuole and coalescence mechanism (Davis and Saunders, 2006; Saunders et al., 2006; Stratman et al., 2009b). Interestingly, MT1-MMP activity appears to facilitate vacuole-vacuole fusion events to help expand the lumenal compartment during this process. This has been repeatedly shown in multiple real-time movies that have been published from our laboratory during vasculogenic tube assembly or EC sprouting events (Iruela-Arispe and Davis, 2009; Saunders et al., 2006; Stratman et al., 2009b). This latter mechanism is fundamental toward the creation of vascular guidance tunnel spaces within the ECM (Stratman et al., 2009b) and is likely to represent a primary regulator of the EC cord hollowing process. In order for solid cords or even vacuolating cords of ECs to form lumens, they will need to be able to create physical spaces within the 3D ECM (i.e. vascular guidance tunnels) (Figures 3, 4, 6, 8). In fact, we have demonstrated that MT1-MMP is required for EC aggregates to form lumens which are essentially forming through a cord hollowing or cell-cell aggregate hollowing mechanism (Davis and Saunders, 2006). In both cases, MT1-MMP activity is necessary to create spaces in the ECM so that the lumens and tubes can expand properly. We also observe intracellular vacuoles within these EC-EC aggregates, however, vacuoles are more obvious when ECs are suspended individually within 3D matrices, likely reflecting the dependence of intracellular vacuole formation on integrin-based EC adhesion in these matrices (Davis and Bayless, 2003; Davis et al., 2002; Davis and Camarillo, 1996). In support of this comment, we have previously shown that intracellular vacuole formation requires integrin-ECM interactions and different integrins control their formation depending on the ECM enviroment in which the ECs are suspended (i.e. α2β1 requirement in collagen matrices, αvβ3 and α5β1 requirement in fibrin matrices) (Bayless et al., 2000; Davis and Camarillo, 1996).

It has recently been suggested that EC cord hollowing in vivo is mediated by cell mechanical mechanisms through Rho kinase (ROCK) to pull apart two opposing EC surfaces to form lumens (Strilic et al., 2009). In contrast, ROCK has been demonstrated by other groups to mediate tube collapse mechanisms that control vascular tube regression (Bayless and Davis, 2004; Im and Kazlauskas, 2007; Mavria et al., 2006) (Figure 10). One distinction in the former study is that it addresses aortic tube formation and a central assumption of this work is that the effects observed are in fact, EC autonomous, a question that was not directly addressed in this in vivo study. It is certainly possible that the addition of ROCK inhibitors might affect surrounding cells (i.e. non-ECs) as well as ECs so it is not possible to ascertain the relevant cellular target of the inhibitor. Developing aortic ECs are known to be in contact with other cell types including mesodermal cells and also definitive endoderm. Interestingly, siRNA suppression of RhoA (a critical activator of ROCK) or addition of dominant negative RhoA does not block lumen formation of human ECs in 3D matrices (Bayless and Davis, 2002; Koh et al., 2008a) (Figure 5). Furthermore, addition of ROCK inhibitors such as Y-27632 (which blocks both ROCK-1 and ROCK-2) fails to block EC lumen formation from either single or aggregated ECs in 3D collagen matrices (Figure 11). Thus, there is little support for the concept that RhoA and ROCK are primary regulators of EC lumen formation in 3D matrices through direct actions on ECs. In contrast, much of the evidence suggests that RhoA and ROCK need to be actively inhibited during EC lumen formation and tube maintenance events (Figure 10). Interestingly, both Src and Erk kinases which are activated to promote EC lumen and tube formation, both participate in inhibiting Rho/ROCK signaling to maintain EC-lined tubes (Bayless and Davis, 2004; Im and Kazlauskas, 2007; Mavria et al., 2006) (Figure 10). Also, there is considerable support for the role of Rho/ROCK in diassembly of EC-EC junctions and collapse of EC-lined tubes in 3D matrices in vitro and in vivo. Due to the ability of Rho/ROCK to cause EC junctional disassembly (Liu and Senger, 2004; Wojciak-Stothard et al., 2001), it is possible that activation of this pathway could facilitate the transition of quiescent ECs to morphogenically active ECs (Liu and Senger, 2004).

Another issue that should be mentioned is the possibility that ECs within different vascular beds (i.e. aorta vs. small vessels) might form lumens and tubes by distinct mechanisms. Although this is certainly a possibility that needs to be investigated, it is also possible that external influences of other cell types on ECs may underlie these apparent differences. Interestingly, there is considerable evidence that embryo growth and motion are occuring during these stages of development which might exert mechanical forces on both cells and surrounding ECM to affect these cellular responses (Filla et al., 2004; Sato et al., 2010; Zamir et al., 2006). This is one good example of an in vivo phenomenon that would be very difficult to mimic through in vitro morphogenesis models. As mentioned above, the developing aortic endothelium is in direct contact with underlying mesodermal cells as well as definitive endodermal cells, thus, it is possible that cell-cell or cell-ECM contacts which characterize these unique heterotypic interactions could play a major role in determining how aortic ECs assemble into lumen structures. Thus, the adhesive template for aortic EC lumen formation may be based on interactions with heterotypic cells, ECM, or both. These possibilities need to be strongly considered as possible reasons for the distinctions described in different studies. In contrast, ECs in the periphery, which assemble into small vessel capillary networks, are likely to have more direct contact with surrounding ECM, thus, mimicking the behaviors observed in the in vitro models that have been developed. Overall, such issues need to be carefully considered to understand the molecular basis for EC lumen and tube formation in different vascular beds.

3.2. Role of cell polarity mechanisms in controlling EC lumen formation

As discussed above, EC lumen and tube formation results in networks of EC-lined structures that have a central lumen space exposed to fluid and a basal surface exposed to ECM. Similar events occur during epithelial lumen formation, except in the Drosophila tracheal system where the lumen and apical compartment contains abundant chitin (Devine et al., 2005). Thus, it is clear that EC and vertebrate epithelial cell lumen formation mechanisms must have an ability to create these unique apical fluid and basal ECM relationships. In addition, in the case of ECs, mural cells are specifically recruited to this basal surface and specialized ECM deposition occurs there (i.e. formation of the vascular basement membrane matrix) (Figure 7). Interestingly, fluid and ion pumps have been shown to propel fluid into the lumen space but in order for this to stay within this compartment, appropriate cell-cell junctional contacts need to be in place such as tight junctions (Bryant and Mostov, 2008). This is particularly observed in epithelial cells where multicellular cell lumenal structures can be induced to acutely swell when treated with agents such as cholera toxin (to stimulate Cl⁻, Na⁺ and H₂O transport into the lumen space through the apical membrane) (Montesano et al., 2009). Interestingly, there is evidence for the involvement of CLIC Cl⁻ channels during intracellular vacuole formation and EC lumen formation suggesting that similar events could occur during vascular morphogenesis (Ulmasov et al., 2009). An important distinction is that EC tubes primarily propagate fluids being pumped from the heart, while many epithelial tubes are distinct and actively secrete fluids and ions from their specialized apical membrane domains.

The molecular basis for how EC tubes polarize is a question of great importance and there is little information that currently addresses this issue. It is quite clear that epithelial cell polarity is distinct from EC polarity as the functions of the two types of cells as well as their apical versus basal domains are also distinct (Bryant and Mostov, 2008; Iruela-Arispe and Davis, 2009). In general, it appears that apical membrane protein targeting is much less obvious in ECs compared to that of epithelial cells. Thus, it is very difficult to identify membrane proteins that are uniquely targeted apically in ECs while there are many proteins and cytoskeletal elements that show strong apical targeting in epithelial cells. Despite this fact, it appears that EC tube formation occurs much more rapidly than does epithelial tube formation in 3D matrices in vitro. EC tube formation occurs within several days in EC 3D cultures while it generally takes weeks to form extensive tube networks for epithelial cell cultures (Koh et al., 2008b; O'Brien et al., 2006). Thus, EC tubulogenesis appears to occur much more rapidly and extensively when compared to epithelial cells despite the clear differences in apical membrane targeting of proteins which strongly favors epithelial cells. Thus, at least with respect to this property, this could suggest that apical membrane protein targeting mechanisms may not correlate well with the ability of cells to form lumen structures in 3D matrices. In contrast, a signaling system that is known to affect cell polarization such as Cdc42-GTP binding to Par6 is clearly involved in both systems and is required for both EC and epithelial lumen formation (Bryant et al., 2010; Koh et al., 2008a; Martin-Belmonte et al., 2007; Sacharidou et al., 2010). Perhaps the cytoskeleton as well as intracellular structures such as centrosomes and the Golgi apparatus need to be polarized to be able to form lumen structures. Centrosome reorientation and nuclear repositioning during cell motility events is known to be controlled by Cdc42 and its effectors (Gomes et al., 2005). Such issues like these are much more likely to be relevant during EC lumen formation than whether particular apical membrane proteins such as podocalyxin can be targeted to the apical membrane surface. Clearly, much new information is necessary to understand how ECs polarize to form lumen structures in 3D matrix environments.

3.3. Mechanisms controlling lumen formation during EC sprouting events

A question that has not been addressed in detail concerns differences in lumen formation events during vasculogenic tube assembly or angiogenic sprouting. Of interest, particularly during the latter process, is the issue that lumen formation appears to occur selectively in the ECs that trail the invading EC tip cells (Sainson et al., 2005). However, we have observed intracellular vacuole formation at the rear of tip cells, and vacuoles are very clearly observed in the cells directly trailing the tip cell (Figure 2). How lumen formation occurs in the cells trailing the EC tip cells is not understood and has not been investigated in detail at this point. Of great interest is how EC tip cells interact with their trailing cells to control processes such as lumen formation. There is considerable evidence that DLL4-Notch signaling plays a role in defining the position of the tip cells and some suggestions have been made that this might control the process of lumen formation (Benedito et al., 2008; Hellstrom et al., 2007; Sainson et al., 2005; Siekmann and Lawson, 2007). Importantly, EC invasion depends on cell surface MMP activity and EC tip cells require such activity to invade into 3D collagen or fibrin matrices (Bayless and Davis, 2003; Saunders et al., 2006). Signals from molecules such as stromal-derived factor (SDF)-1α and sphingosine-1-phosphate (S1P) stimulate this process during EC invasion (Bayless and Davis, 2003; Saunders et al., 2006). Interestingly, lysophosphatidic acid (LPA) can stimulate invasion of tumor cells in 3D matrices by a process that also depends on cell surface MT-MMP activity (Fisher et al., 2006; Fisher et al., 2009; Sabeh et al., 2009). In all cases, these invading cells create physical spaces during this process and leave such spaces behind them as they invade forward (Bayless and Davis, 2003; Fisher et al., 2009; Stratman et al., 2009b). In the case of ECs, the cells that follow tip cells already occupy a physical space that can contribute to the developing lumenal space, by flattening out within this space (Stratman et al., 2009b). In addition, the cells behind the tip cells can expand the lumenal space by accumulating intracellular vacuoles or expand through proteolytic events from lateral membrane surfaces (Figures 2 and 4). This latter process is highly related to lumen formation that occurs from groups of aggregated cells (i.e. cord hollowing mechanisms) which also depend on cell surface MT-MMPs in 3D collagen matrices (Davis and Saunders, 2006). MT1-MMP regulates EC lumen formation during vasculogenic tube assembly from single cells or aggregated ECs that form spherical multicellular tubes (Davis and Saunders, 2006; Sacharidou et al., 2010; Saunders et al., 2006; Stratman et al., 2009b). In either case, they can sprout to connect with neighboring cells (or aggregates) to form multicellular lumen and tube structures as well as tube networks. This latter process also depends on MT1-MMP proteolytic activity in 3D collagen matrices. Blockade of MT1-MMP with inhibitors such as TIMP-2, TIMP-3 or GM6001, using siRNA suppression or expression of a dominant negative MT1-MMP mutant completely attenuates EC lumen and tube formation in 3D collagen matrices (Sacharidou et al., 2010; Saunders et al., 2006; Stratman et al., 2009b).

3.4. Mechanisms that control remodeling and maintenance of EC-lined tube networks

The EC tubulogenic process is highly dynamic as revealed by real-time morphogenesis movies observed both in vitro and in vivo (Herbert et al., 2009; Iruela-Arispe and Davis, 2009; Kamei et al., 2006; Sato et al., 2010; Saunders et al., 2006; Stratman et al., 2009a; Stratman et al., 2009b; Wang et al., 2010; Yaniv et al., 2006). Not only are tube formation and sprouting events very dynamic, subsequent remodeling events are equally dynamic. This is dramatically illustrated during mouse development where flow-induced changes through the initial vascular plexus (with very uniform vessel diameters) lead to dramatic remodeling where both large and small vessels are then observed (Lucitti et al., 2007). Thus, marked changes in lumen diameters and tube lengths occur to develop the characteristic features of the functional microcirculation including artery, arterioles, capillary, venules and veins. In addition to flow-induced remodeling, a second major regulator of vascular tube remodeling is recruitment of mural cells (at later stages of vascular development). Mural cell recruitment, which involves both pericytes and vascular smooth muscle cells, is responsible for facilitating tube maturation and stabilization (Adams and Alitalo, 2007; Gaengel et al., 2009; Senger and Davis, 2010; Stratman et al., 2009a). Furthermore, recruitment of vascular smooth muscle cells in particular, to vascular tubes allows for lumen diameters of EC tubes to be regulated through smooth muscle contractility (i.e. via molecules such as nitric oxide to relax vessels and angiotensin II to constrict them) (Hill et al., 2007; Martinez-Lemus et al., 2003; Segal, 2005). This ability of vascular smooth muscle to dynamically control vessel lumen diameter is critical to the molecular control of blood pressure and blood flow to tissues (Hill et al., 2007; Segal, 2005). Interestingly, pericytes have also been shown to play a role in regulating vessel diameters in capillary beds, particularly, in the central nervous system (Peppiatt et al., 2006). Alterations in such controls underlie diseases such as hypertension and diabetes where abnormalities in EC-mural cell interactions occur (Hammes, 2005). In hypertension or diabetes, vessel wall remodeling occurs which involves both cellular and ECM changes and that make treatment options more difficult, particularly with longstanding disease (Arroyo and Iruela-Arispe, 2010; Hill et al., 2007; Raffetto and Khalil, 2008; ten Dijke and Arthur, 2007; van den Akker et al., 2010).

Another issue that is not typically considered with respect to lumen and tube formation is how lumen and tube networks are maintained during the initial phases of the process (i.e. without regard to flow forces or contributions of other cell types such as mural cells). Our laboratory has shown for example that maintenance of EC tube structures depends on both the microtubule and actin cytoskeletons (Bayless and Davis, 2004). Microtubule depolymerizing factors (i.e. vinblastine, colchicine) dramatically cause collapse of EC-lined tubes in a manner dependent on the RhoA GTPase (Bayless and Davis, 2004) (Figure 10). Interestingly, disruption of the actin cytoskeleton with cytochalasin B does not cause tube collapse. However, addition of cytochalasin B prior to addition of vinblastine or colchicine to diassemble microtubules blocks tube collapse suggesting that actin-based cell contractility is necessary for this tube collapse mechanism (Bayless and Davis, 2004). What is interesting is that there must be an important signaling pathway within ECs that maintains tube structures in a 3D matrix environment. When disrupted, EC tubes rapidly collapse and disassemble. This mechanism must involve a combination of cell-matrix adhesions as well as cell-cell adhesions that actively prevent tube collapse. The fact that tubes can rapidly collapse (within minutes) suggests that the tube networks may be under mechanical tension because this rapid collapse leads to clustering of groups of ECs together (Bayless and Davis, 2004; Stratman et al., 2009b). Following tube collapse, it becomes obvious that vascular guidance tunnels are present as physical spaces in which the tube structures were embedded (Stratman et al., 2009b). Furthermore, the tunnel spaces represent matrix conduits in which tube remodeling events can occur since ECs freely migrate within them. Interestingly, we have shown that EC motility responses within vascular guidance tunnels occurs in an MMP-independent manner (Stratman et al., 2009b), and can lead to tube remodeling events. It becomes easier to interpret important vascular responses such as flow-induced remodeling when it is realized that EC-lined tubes are present within vascular guidance tunnel networks that are created as a result of the tube morphogenic process. We also presented evidence that more tunnels are generated than are occupied by tubes at any given time (Stratman et al., 2009b) so this could account for rapid changes in tube structure and patterns in response to an applied stimulus such as flow or recruitment of mural cells. Since stimuli such as flow, fluid pressure, ECM remodeling can regulate EC tube remodeling and stabilization events, tunnels may be important for such responses. A detailed molecular analysis of how tubes are maintained in a 3D matrix environment is particularly important from the standpoint of ECs, so that the influence of external stimuli can be examined and understood. Clearly, much further investigation is necessary in this area.

4.1. Evidence for a fundamental role of Cdc42 and Rac1 in lumen formation

Both Cdc42 and Rac1 have been shown to be required for EC lumen and tube formation and this has been demonstrated utilizing dominant negative mutants and siRNA suppression experiments (Figures 5, 6, and 10) (Bayless and Davis, 2002; Davis et al., 2007; Koh et al., 2008a). In addition, Cdc42 and Rac1 control intracellular vacuole formation which plays a fundamental role in EC lumen formation, particularly when ECs are seeded as single cells in 3D matrices (Bayless and Davis, 2002). Other investigators have obtained similar results in vitro and interestingly, recent experiments demonstrate requirements for both Cdc42 and Rac1 in EC lumen formation in vivo (Hoang et al., 2010a; Hoang et al., 2010b; Tan et al., 2008). Also, we have observed intracellular vacuole labeling with GFP-Rac1, GFP-Cdc42 and GFP-RalA (Bayless and Davis, 2002; Davis et al., 2007). GFP-Cdc42, when expressed from an EC-specific promoter in developing Zebrafish revealed intracellular vacuole formation and fusion during the lumen formation process of intersegmental vessels (Kamei et al., 2006). In support of these findings, a recent study showed that a GFP-moesin construct also revealed EC intracellular vacuoles that form and fuse to control intersegmental vessel formation in Zebrafish embryos (Wang et al., 2010).

Over the past 15 years, many molecular requirements and signaling pathways have been revealed that control how human ECs form lumen and tube structures in 3D extracellular matrices (Figures 6 and 10). Most of our studies have utilized collagen matrices, and, thus, a major requirement for these events is the α2β1 integrin, a collagen-binding integrin (Davis and Camarillo, 1996). Blocking antibodies directed to α2β1 markedly block lumen formation (Figure 12) as well as α2 integrin subunit siRNAs (Sacharidou et al., 2010). Blocking antibodies directed to other integrin subunits expressed by ECs, including α5β1, a fibronectin receptor, have no blocking influence in 3D collagen matrices with EC only cultures (Figure 12). Furthermore, considerable work has shown that α2β1 integrin is important in vivo for both developmental and postnatal vascularization responses (San Antonio et al., 2009; Senger et al., 1997). In contrast, both αvβ3 and α5β1 control EC lumen and tube formation in 3D fibrin matrices (Bayless et al., 2000), while α2β1 was not shown to be involved. Importantly, these integrins have been shown to be involved during vessel formation during development and in postnatal mice (Astrof et al., 2007; Hynes, 2007; Senger et al., 1997; Somanath et al., 2009; Stupack and Cheresh, 2004; van der Flier et al., 2010). An important point is that data obtained in vitro models accurately predicted integrin requirements controlling vascular morphogenesis in vivo. Another key point is that in vitro model systems demonstrated first that multiple integrin chains can be utilized by ECs to regulate tube morphogenesis in pro-morphogenic ECM environments such as collagen and fibrin. The role of particular integrins in morphogenesis appears to be directly linked to the ECM environment (and the predominant ECM components) that are in contact with the ECs.

Integrins are known to activate Rho GTPases (Hall, 2005) and our laboratory first reported that Cdc42 was a critical GTPase controlling EC lumen formation (Bayless and Davis, 2002). Subsequent studies revealed that Cdc42 is a critical regulator of lumen formation for both ECs and epithelial cells (Bryant et al., 2010; Davis and Bayless, 2003; Jaffe et al., 2008; Koh et al., 2008a; Koh et al., 2009; Martin-Belmonte et al., 2007). We also reported a role for Rac1 in EC tubulogenesis (Bayless and Davis, 2002; Koh et al., 2008a), while RhoA had no apparent influence on these events. In contrast, we made the key observation that RhoA controlled EC tube collapse in response to depolymerization of microtubules in 3D matrices (Bayless and Davis, 2004) (Figure 10). Both Cdc42 and Rac1 were shown to be activated during the morphogenic cascade in 3D collagen matrices (Koh et al., 2008a; Koh et al., 2008b; Sacharidou et al., 2010). In a recent study, we demonstrated that Cdc42 activation in 3D matrices was dependent on junction adhesion molecules (Jam)-B and -C as well as MT1-MMP (Sacharidou et al., 2010) (Figure 6). Disruption of these three molecules led to marked decreases in EC Cdc42 activation in 3D collagen matrices (Sacharidou et al., 2010).

4.2 Role of the Cdc42 effectors, Pak-2 and Pak-4, in EC lumen formation

To identify downstream effectors that are responsible for the influence of Cdc42 and Rac1, a series of known effectors were screened using siRNA treatment of ECs (Koh et al., 2008a). Strong blocking phenotypes were observed using siRNAs to both p21-activated kinase (Pak)-2 and Pak-4 (Koh et al., 2008a). Previous work suggested a potential role for Pak-1 in EC tube formation (Kiosses et al., 2002) and recent studies have shown that both Pak-2 and Pak-4 are involved in developmental vascularization events, which corroborate the in vitro findings (Galan Moya et al., 2009; Liu et al., 2007; Tian et al., 2009). Also, Pak-2 and Pak-4 siRNAs block EC sprouting into 3D collagen matrices (Koh et al., 2008a). Importantly, Pak-2 and Pak-4 activation (detected by phosphorylation) directly correlated with both Cdc42 and Rac1 activation and the EC lumen formation process (Koh et al., 2008a). Activated Pak-2 and Pak-4 could also be demonstrated to be associated with activated Cdc42 (i.e. Cdc42-GTP) during these events. Expression of dominant negative mutants of either Pak-2 or Pak-4 in ECs completely inhibited lumen and tube formation (Koh et al., 2008a). Interestingly, both Cdc42 and Rac1 are capable of activating Pak-2 while Cdc42 selectively activates Pak-4 (Bokoch, 2003; Galan Moya et al., 2009). Interestingly, Pak-2 can be activated in a Rho GTPase-independent manner by binding to GIT-1 (Loo et al., 2004), a protein that also binds to a known Cdc42 GEF and Cdc42 and Rac effector, β-PIX. Pak-2 is also known to be a substrate of caspase-3 and is cleaved during vascular tube collapse and regression when microtubules are disassembled (Bayless and Davis, 2004). Interestingly, Pak phosphorylation can lead to EC junctional disassembly (Orr et al., 2007; Stockton et al., 2004) and increased vascular permeability in the context of atherosclerotic lesions (Orr et al., 2007). Pak-2 is of particular interest in that it appears to be involved in both EC lumen and tube assembly but possibly also in tube disassembly. Thus, Pak-2 may represent a bimolecular switch that can participate in either the processes of tube assembly or disassembly and its regulation may be fundamental to controlling these events. Further studies will need to assess how Pak-2 can perform such distinct functions and determine which molecules work in conjunction with Pak-2 to affect these different events.

4.3. Role for PKCε and Src family kinases in EC tube morphogenesis and subsequent Pak-2 and Pak-4 activation events

Kinases that are activated by cell-ECM interactions include protein kinase C isoforms (i.e. conventional, novel and atypical isoforms) and Src family kinases in addition to others (Figure 6). In our studies of EC lumen formation in 3D collagen matrices, PKCɛ was shown to be involved but not PKCα or PKCδ (Koh et al., 2008a; Koh et al., 2009). Inhibition of PKCα or PKCβ using the chemical inhibitor, Go6976, fails to block EC lumen formation and in fact, accelerates the process (Koh et al., 2009). Interestingly, this inhibitor shows great selectivity for these conventional isoforms and does not block PKCε or atypical PKC isoforms. In contrast, Go6983, which does block PKCε and atypical PKC isoforms, strongly inhibits EC lumen formation (Koh et al., 2009). In further support for these findings, expression of dominant negative mutants of either PKCε or siRNA suppression of PKCε blocks lumen formation as well as downstream Src and Pak activation (Koh et al., 2009). Also, increased expression of PKCε markedly enhanced EC lumen and tube formation in 3D collagen matrices. This treatment further increased Src, Pak-2 and Pak-4 phosphorylation events that directly correlated with its morphogenic influence (Koh et al., 2009). Overall, these studies indicate a fundamental role for PKCε in EC lumen formation and furthermore, that it appears to signal upstream of Src activation, while Src activation is upstream of Pak activation (Figures 6 and 10) (Koh et al., 2009). Key future studies here include identifying PKCε phosphorylation targets that underlie its biological influence during these morphogenic processes.

Blockade of Src family kinases by siRNA suppression, increased expression of the Src inhibitor, Csk (i.e. C-terminal Src kinase), or treatment with chemical inhibitors with specificity for Src (e.g. PP2) markedly interferes with EC tube formation (Koh et al., 2009). Expression of a dominant negative Csk construct increased lumen formation, which further supports a key role for Src during EC lumen formation. Furthermore, Src and Pak kinases are known to activate Raf kinases to affect processes such as cell survival which have previously been demonstrated to effect angiogenesis in vivo (Alavi et al., 2003) and a B-Raf mouse knockout shows an embryonic lethal phenotype that is due to vascular abnormalities (Galabova-Kovacs et al., 2006). Our laboratory demonstrated that Raf kinase activation (i.e. both B-Raf and C-Raf) occurs downstream of Src and Pak activation to control EC tube formation in 3D collagen matrices (Koh et al., 2009). This is accompanied by downstream Erk1/2 activation which is also necessary for ECs to form tube networks. MKP-3 (also called DUSP-6), a serine/threonine phosphatase with selectivity for phospho-Erk1/2, markedly decreased Erk phosphorylation and completely abrogated lumen formation (Koh et al., 2009). A dominant negative MEK kinase inhibitor also blocks lumen formation as well as Erk1/2 phosphorylation events. Overall, this morphogenic signaling pathway is coupled to cytoskeletal rearrangements, cell surface-mediated proteolysis of ECM, survival and transcriptional events that are necessary to coordinately regulate EC lumen formation (Figure 6).

4.4. Cdc42-mediated control of cell polarity pathways mediates EC lumen and tube formation in 3D collagen matrices

A key function of Cdc42 is its ability to affect cell polarity signaling by interfacing with the polarity proteins, Par6, Par3 and atypical PKC isoforms (Etienne-Manneville, 2004; Etienne-Manneville and Hall, 2003; Macara, 2004). Cell polarity signaling controls directional cell motility that involves Cdc42 (Etienne-Manneville and Hall, 2003) and reorientation of the centrosome and positioning of the nucleus during these events (Gomes et al., 2005). In fact, active Cdc42 (i.e. Cdc42-GTP) binds directly to Par6 which then couples to Par3 (Macara, 2004). Rac1 also appears to interface with this pathway through its ability to interact with Par6 (Brazil and Hemmings, 2000). Par3 is known to interact with a variety of important regulatory proteins including Jams A, B, and C and VE-cadherin through their cytoplasmic tails and through the PDZ domains of Par3 (Ebnet et al., 2003; Ebnet et al., 2004). Par6 is known to interact directly with atypical protein kinase C isoforms, such as PKCζ. Par6 is also a phosphorylation target of these kinases (Macara, 2004). We recently reported that Cdc42-dependent EC lumen and tube formation was dependent on Par6b, Par3 and atypical PKC (Koh et al., 2008a). Thus, this work reveals a fundamental role for Cdc42-dependent polarity signaling during EC tubulogenesis. Furthermore, we have recently reported that JamB and JamC associate with Par3 in ECs to control EC lumen formation in 3D collagen matrices (Sacharidou et al., 2010). Furthermore, blocking antibodies directed to Jams B and C markedly block EC lumen formation particularly when they are combined (Figure 12). These Jam proteins co-assemble into a defined EC lumen signaling complex consisting of α2β1, MT1-MMP, Jam-C, Jam-B, Par3, Par6b and Cdc42-GTP that together control the ability of ECs to form tubes in 3D collagen matrices (Sacharidou et al., 2010). These lumen signaling complexes are also directly coupled to the kinase cascade discussed above including PKCε, Src, Pak, Raf and Erk 1/2 since disruption of these complexes completely interferes with the downstream signaling necessary to regulate vascular tube morphogenesis (Sacharidou et al., 2010). Cdc42 and Par3 have also been shown to control lumen formation in epithelial cells and a recent study shows that Par3 and beta1 integrins co-regulate arteriolar lumen formation in vivo using a conditional beta1 integrin subunit knockout mouse system (Zovein et al., 2010). Thus, this latter work again confirms our prior conclusions obtained in vitro showing that beta1 integrins (i.e. α2β1), Cdc42 and polarity proteins control the lumen and tube formation process in 3D matrix environments (Davis et al., 2007; Koh et al., 2008a; Koh et al., 2009; Sacharidou et al., 2010).

4.5. Functional interdependence of Cdc42 and MT1-MMP-mediated signaling events control EC tube assembly in 3D collagen matrices

Several new findings are beginning to elucidate how Cdc42-dependent signaling events, which activate kinase cascades and interact with the cell polarity molecules (i.e. Par3, Par6, atypical PKC), intersect with MT1-MMP-dependent proteolytic events to create EC lumens and tubes in 3D extracellular matrices (Figure 6) (Bayless and Davis, 2002; Ispanovic et al., 2008; Koh et al., 2008a; Koh et al., 2009; Sacharidou et al., 2010; Stratman et al., 2009b). It is clear that EC lumen and tube formation is a 3D matrix-specific process (Davis et al., 2002), and furthermore, that tubulogenesis does not occur on 2D matrix surfaces. Interestingly, EC motility can occur on 2D matrix surfaces and also in 3D matrices, but in the latter case, the motility response depends on MT1-MMP proteolytic events (Stratman et al., 2009b). These EC movement and proteolytic events are directly coupled during tube morphogenesis.

Since the lumen formation process in 3D collagen matrices requires both Cdc42-dependent and MT1-MMP-dependent activities, we addressed how these might be functionally coupled. Blockade of MT1-MMP activity using siRNA suppression or MT1-MMP inhibitors leads to marked interference with Cdc42 activation (a critical step necessary for activation of effectors such as Pak2, Pak4 and Par6 that lead to EC tubulogenesis) in 3D collagen matrices (Sacharidou et al., 2010). However, MT1-MMP blockade does not affect Cdc42 activation of ECs when they are seeded on 2D collagen surfaces, and random motility is also not affected (Sacharidou et al., 2010). Expression of a dominant negative MT1-MMP construct also markedly blocks lumen formation and Cdc42 activation (Sacharidou et al. 2010) in 3D collagen matrices. In contrast, activation of RhoA, is not affected by blockade of Cdc42 or MT1-MMP. Overall, this data shows for the first time that MT1-MMP activity is directly coupled to Cdc42 activation in 3D, but not 2D matrices, to control the tube formation process (Sacharidou et al., 2010). The opposite is also true, in that blockade of Cdc42 using siRNA suppression or dominant negative mutants leads to marked decreases in both tube and vascular guidance tunnel formation, a consequence of inactivation of MT1-MMP-dependent proteolysis (Sacharidou et al., 2010). Thus, Cdc42 and MT1-MMP represent interdependent signaling molecules that control vascular tube morphogenic events specifically in 3D matrix environments.

4.6. Definition of an EC lumen signaling complex that controls vascular tube morphogenesis in 3D extracellular matrices

In order to explain how Cdc42 and MT1-MMP regulate each other during vascular tube morphogenesis, we wondered if they might be co-associated in a multimolecular signaling complex to control this process. During our analysis of Cdc42 activation from ECs undergoing morphogenesis in 3D matrices, we utilized Pak beads, a specific adsorbant for Cdc42-GTP and Rac1-GTP (the beads do not bind the GDP bound GTPases or RhoA-GTP). Eluates from these beads revealed Cdc42-GTP, but also contained MT1-MMP, the α2β1 integrin, JamC, JamB, Par3, and Par6b (Sacharidou et al., 2010) suggesting the presence of a multimolecular protein complex from ECs undergoing tube morphogenesis (Figure 6). Furthermore, Cdc42, MT1-MMP and JamB were each epitope tagged with an S-tag and when these proteins were expressed in ECs as recombinant proteins, we were able to demonstrate the co-association of the same molecules which now defines an EC lumen signaling complex that controls EC tubulogenesis in 3D collagen matrices (Sacharidou et al., 2010). Each of the molecules is required for lumen formation and disruption of any of these proteins leads to blockade of this process. Our evidence suggests that MT1-MMP and α2β1 integrin are tightly associated and that they interact together with JamC (Figure 6). SiRNA suppression of either component eliminates their binding to JamC and dissociates them from the signaling complex suggesting that MT1-MMP and α2β1 appear to be interdependent for their association with the lumen signaling complex. This makes considerable sense since they both bind collagen triple helices to perform their functions. It is also possible that binding of collagen by MT1-MMP and α2β1 play a role in the preparation of collagen type I for proteolysis that is critical for the lumen formation process. Collagenolytic MMPs do appear to play a role in facilitating unwinding of the collagen triple helix so that specific cleavage of collagen type I occurs (Chung et al., 2004; Lauer-Fields et al., 2002). JamC, which appears to associate with the MT1-MMP/α2β1 complex, also shows affinity for JamB, a known binding partner for JamC. Both JamB and JamC bind through their cytoplasmic tails to Par3, a scaffolding protein with affinity for both Par6 and atypical PKC isoforms (Sacharidou et al., 2010). Par6 binds Cdc42-GTP to complete the signaling complex (Figure 6). Interestingly, new data suggests that Rac1-GTP is also part of the EC lumen signaling complex, perhaps through its binding affinity for Par6 (Sacharidou et al., unpublished observations).

Expression of cytoplasmic tail deleted forms from either JamB or JamC completely inhibits EC lumen formation and Cdc42 activation (Sacharidou et al., 2010). Interestingly, these tail deleted mutants selectively block Cdc42 activation in 3D collagen matrices and not on 2D collagen surfaces. Also, the Jam tail mutants do not affect RhoA activation in 3D matrices or on 2D collagen surfaces. Interestingly, we observed that JamA appears to block EC lumen formation in that siRNA suppression increases lumen formation while increased expression of JamA blocks lumen formation and Cdc42 activation (Sacharidou et al., 2010). One obvious possibility is that increased expression of JamA blocks this process by competing for Par3 binding and thus interferes with the ability of JamB and JamC to bind. Further investigation is necessary to address this issue. Overall, our data indicates that interactions between JamB and JamC with Par3 are necessary for the lumen formation process. One question of great interest to us is to determine the identity of the Cdc42 guanine exchange factors (GEF) that are responsible for Cdc42 activation during these events. Also, we wish to determine how this Cdc42 GEF interacts with the EC lumen signaling complex since we hypothesize that it will show affinity for one of the components within the complex. Considerable future work will be necessary to identify other components of this signaling complex as well as to identify both positive and negative regulators of the complex that modulate these processes.

5.1. MT1-MMP controls EC lumen formation and EC sprouting events in 3D matrices

Our work and that of others has demonstrated a role for MT1-MMP (i.e. MMP-14) in EC morphogenic events in 3D matrix environments (Chun et al., 2004; Haas et al., 1999; Lafleur et al., 2002; Saunders et al., 2006; Stratman et al., 2009b). MT1-MMP is a transmembrane protein and its cell surface expression is required for it to perform the localized ECM degradation necessary to control cell movement and morphogenic responses in 3D matrices (Fisher et al., 2006; Fisher et al., 2009; Itoh and Seiki, 2004; Ota et al., 2009; Sabeh et al., 2009). This proteinase has been shown to play a crucial role in cell invasive behavior in cross-linked ECM environments particularly in adult tissues (Sabeh et al., 2009). Its role during embryonic development is less clear although it could participate and work with other enzymes to facilitate invasive behavior in this context. More work here is necessary to understand these events by comparing invasion responses in embryos versus postnatal animals. MT2-MMP appears to have similar abilities compared to MT1-MMP (Ota et al., 2009; Sabeh et al., 2009) while MT3-MMP appears to be distinct in its functions. However, it may be capable of acting in conjunction with either MT1-MMP, MT2-MMP or both to control cell invasion. The degree of functional redundancy among MT-MMPs, other cell surface MMPs such as ADAMs or other proteinases remains an open question that needs to be investigated during developmental vascularization.

Considerable information suggests that membrane anchorage of the MMP is crucial to its ability to stimulate cell invasive behavior (also, a necessary function for EC tubulogenesis). For example, soluble MMP-1 is not able to regulate cell invasive behavior but when its catalytic domain is swapped in to replace the catalytic domain of MT1-MMP, it can promote cell invasion only if it is membrane tethered (Li et al., 2008). Knockout of MT1-MMP in mice is compatible with embryogenesis but the mice are very small and sickly and typically die within a month or two after birth (Zhou et al., 2000). It appears that there is a marked reduction of vessel formation within developing bones even in embryonic life (perhaps due to increased ECM density within this developing tissue). Attempts to induce angiogenesis in postnatal MT1-MMP knockout mice reveal that these responses do not occur (Zhou et al., 2000) and aortic ring assays show no sprouting in either 3D collagen or fibrin matrices compared to controls (Chun et al., 2004). Thus, MT1-MMP plays an important role in controlling vascular morphogenic events in 3D matrices and during in vivo angiogenic responses.

To elucidate molecular mechanisms underlying how MT1-MMP controls vascular tube morphogenesis and cellular invasive events, our laboratory has examined this question using both vasculogenesis and angiogenesis models (Figure 13) (Bayless and Davis, 2003; Sacharidou et al., 2010; Saunders et al., 2006; Stratman et al., 2009b). EC sprouting in response to sphingosine-1-phosphate (incorporated into the collagen matrices) was blocked by the broad spectrum inhibitor, GM6001, as well as tissue inhibitor of metalloproteinase (TIMP)-2, TIMP-3, and TIMP-4, but not TIMP-1 (Bayless and Davis, 2003). MT-MMPs are insensitive to TIMP-1, but the other inhibitors utilized block their activity. GM6001, TIMP-2, TIMP-3 and TIMP-4, all block EC lumen and tube formation completely, while TIMP-1 had no influence (Saunders et al., 2006; Stratman et al., 2009b) (Figure 13). One important distinction among the TIMPs is that TIMP-3 is able to block the activity of MT-MMPs but also many members of the ADAM family of metalloproteinases including ADAM-17 and ADAM-10 (Baker et al., 2002; Brew and Nagase, 2010). To functionally dissect which EC surface expressed metalloproteinases were relevant during lumen formation or EC sprouting, siRNA suppression analysis was performed. These results suggested that the dominant metalloproteinase controlling these events is MT1-MMP (Figure 13), with a lesser influence of MT2-MMP during both sprouting and lumen formation (Sacharidou et al., 2010; Saunders et al., 2006; Stratman et al., 2009b). Also, a partial blocking effect of ADAM-15 siRNA knockdown in EC sprouting assays was observed using stromal-derived factor-1α as the invasion stimulus (Saunders et al., 2006), in support of previous work showing a role for ADAM-15 in regulating angiogenic responses (Horiuchi et al., 2003). There was no influence of either MT3-MMP or ADAM-17 siRNAs in these invasion responses (Saunders et al., 2006). Further important support for a role for MT1-MMP during EC tubulogenesis is that increased expression of MT1-MMP using viral vectors leads to marked increases in lumen formation that depends on its MMP catalytic domain and activity since GM6001 addition completely inhibits the stimulatory influence of the recombinant protein and a catalytically inactive recombinant protein fails to stimulate lumen formation (Stratman et al., 2009b). In very recent experiments, we have further shown that increasing the expression of wild-type, full length MT1-MMP increases both the rate and extent of EC lumen formation in 3D collagen matrices (Sacharidou et al., 2010). Furthermore, we have generated a construct (an MT1-MMP active site mutant without its cytoplasmic tail) that behaves like a dominant negative mutant of MT1-MMP since its expression in the presence of wild-type MT1-MMP completely blocks the ability of ECs to form lumens and tubes (Sacharidou et al., 2010). It also completely inhibits the formation of vascular guidance tunnels, the carved out spaces of degraded ECM which characterize the lumen formation process. An additional finding is that expression of wild-type MT1-MMP without its cytoplasmic tail markedly stimulates the rate and extent of EC lumen formation even compared to increased expression of full-length wild-type MT1-MMP (Sacharidou et al., 2010). A number of studies suggest that the cytoplasmic tail of MT1-MMP plays a role in increased endocytic recycling (and possible turnover of MT1-MMP) and thus, deleting the tail increases cell surface expression. Overall, these data demonstrate that MT1-MMP is a major regulator of EC lumen and tube formation and that it works closely in conjunction with the collagen-binding integrin, α2β1, as well as Cdc42- and Rac1-dependent signaling to control this process (Figure 6).

5.2. MT1-MMP-dependent EC tubulogenesis leads to the formation of a network of physical spaces within the ECM termed vascular guidance tunnels

A critical role for MT1-MMP during EC lumen and tube formation is to create networks of physical spaces within the ECM that we term vascular guidance tunnels (Stratman et al., 2009b). These form as a result of MT1-MMP-mediated proteolysis of collagen matrices and can be demonstrated to be physical spaces by microinjection of silicone oil. In every instance that our laboratory has examined, there is a direct relationship between EC tube formation and the formation of vascular guidance tunnels (Stratman et al., 2009b). Thus, the lumen and tube formation mechanism must be integrated directly with cell surface proteolysis to create these physical spaces. For example, inhibitors that interfere with EC lumen and tube formation including anti-α2 and anti-β1 integrin blocking antibodies, chemical inhibitors of PKCε and Src, as well as MT1-MMP inhibitors, completely abrogate the formation of vascular guidance tunnels (Stratman et al., 2009b). The presence of these tunnels raised the interesting possibility that vessel remodeling could occur through these pre-formed physical tunnel spaces which are generated as part of the EC morphogenic cascade in 3D matrices. Although MT1-MMP was necessary for the formation of vascular guidance tunnels and initial EC migratory events that are required to assemble tubes, once tunnels were formed, MT1-MMP blockade did not affect the ability of ECs to migrate within the spaces (Stratman et al. 2009b). Thus, vascular guidance tunnel spaces are similar to 2D matrix surfaces where EC motility is insensitive to MT1-MMP inhibition (Stratman et al., 2009b). Interestingly, ECs have been shown to rapidly migrate in vivo during vascular development to regulate both tube assembly and vascular remodeling (Lucitti et al., 2007; Sato et al., 2010). Clearly, EC tube assembly is an obligate step for the formation of vascular guidance tunnels and thus, these processes are directly linked. Several critical questions arise from these insights including how EC lumen and tube formation are functionally connected with the cell surface proteolytic machinery to create vascular guidance tunnels. Also, what is the functional purpose of vascular guidance tunnels during vascular tube assembly, remodeling and maturation events?

5.3. Critical functional role for EC-generated vascular guidance tunnels during vascular tube assembly and remodeling in 3D matrices

ECs utilize a lumen signaling complex to form tubes in 3D matrices while at the same time generate networks of vascular guidance tunnel spaces. During early vascular development, there is evidence for marked tube remodeling that occurs following the onset of flow (Lucitti et al., 2007), and we hypothesize that this occurs due to the presence of vascular guidance tunnels which allows for ECs to rapidly rearrange to accommodate flow and pressure forces that are applied. Interestingly, addition of MMP inhibitors that block lumen formation (i.e. TIMP-3 and GM6001), but not TIMP-1, interfere with vascular remodeling events in rotating mouse embryo cultures suggesting that this flow-induced remodeling may also depend on MMP activity (Saunders et al., 2006). Increased vessel width, which occurs during this remodeling step in vivo, depends on MT1-MMP activity in our 3D collagen matrix cultures (Stratman et al., 2009b). It remains to be determined if MT1-MMP, other MMPs, ADAMs, or ADAMTS enzymes control these events during this developmental vascular remodeling event. In addition, at this stage of development, the ECM is likely to be more mechanically pliable and, thus, the applied forces generated could expand lumen or tunnel widths. We have also shown that groups of cells comprising a tube structure can migrate together through vascular guidance tunnels to connect with adjacent EC tubes to affect vascular remodeling (Stratman et al., 2009b). Interestingly, the EC lumen and tube formation process generates more vascular guidance tunnels than are utilized in 3D collagen matrices (Stratman et al., 2009b). If this occurs in vivo, the extra tunnels could allow for vascular remodeling events to rapidly occur when external forces (i.e. flow) are applied or other cell types such as pericytes are recruited so that a proper microcirculatory network matches the flow patterns and oxygenation requirements of the developing tissues being perfused. Similar principles should apply in adult tissues following tissue injury or other pathologic events that lead to vascularization responses.

5.4. Vascular guidance tunnels are matrix conduits that can be utilized for vessel regrowth following vascular tube regression

Vascular guidance tunnels are also important to consider in the context of vascular tube regression and regrowth of vessels. We have previously reported that thrombin addition to networks of EC-lined tubes can induce rapid tube collapse, but when thrombin activity is reversed by addition of the thrombin inhibitor, hirudin, vessels can rapidly reform within the vascular guidance tunnel space that was previously lined by an intact vessel tube (Stratman et al., 2009b). Thus, vascular regrowth clearly can occur along pre-formed matrix conduits (i.e. vascular guidance tunnels) and if these were eliminated along with loss of the tubes, this would strongly decrease the possibility of vessel regrowth. We have previously reported that both EC produced MMP-1 and MMP-10 play a role in vascular tube regression and together they are capable of collapsing both tubes and vascular guidance tunnel networks due to the fact that the collagen matrix scaffold is degraded during this process (Davis et al., 2001; Davis and Saunders, 2006; Saunders et al., 2005). Thus, such a regression mechanism would eliminate the possibility of vessel regrowth following a tissue injury response. The presence of pericytes, which blocks MMP-1 and MMP-10-dependent tube regression event in our in vitro model (Saunders et al., 2006), can thus protect the vascular tube structure but also the integrity of the vascular guidance tunnels through MMP inhibition (i.e. through EC TIMP-2 and pericyte TIMP-3) (Figure 10).

Of related interest here is that tumor vessels are highly resistant to vascular regression. It is possible that this lack of regression is due to their production of regression inhibitors such as TIMP-1 and PAI-1 (Davis and Saunders, 2006). TIMP-1 is capable of protecting both the vessels and the tunnel spaces by blockade of MMP-1 and MMP-10 activity (Davis and Saunders, 2006; Saunders et al., 2005). Interestingly, when tumors are treated with vascular regression agents such as VEGFR2 antagonists, vessels regress, but they can rapidly regrow (following withdrawal of the regression agent) in a manner that appears to recapitulate the original vessel pattern (Mancuso et al., 2006). This may occur through vascular guidance tunnels that were previously generated during initial tumor vessel formation. So an important therapeutic consideration here in the context of tumor treatment would be to devise approaches to induce both vessel as well as vascular guidance tunnel regression within tumors. In this manner, vessel regrowth would be less likely to occur, allowing for a better therapeutic opportunity to treat tumors, as well as its associated vasculature and matrix conduits (that may also facilitate local tumor invasive behavior) to affect vascular and tumor regrowth.

5.5. Potential role for vascular guidance tunnels in arteriovenous differentiation events

It is also important to consider how developmental events such as arteriovenous identity and sorting (Herbert et al., 2009; Kim et al., 2008; Swift and Weinstein, 2009; Wang et al., 1998) could be regulated by vascular guidance tunnels since they represent a 2D matrix surface in a 3D matrix environment (Stratman et al., 2009b). EphrinB2 (an arterial marker) and EphB4 (a venous marker) represent a repulsive signaling pair that appear to distinguish ECs that differentially express these arterial versus venous markers (Wang et al., 1998). These repulsive interactions allow for differential cell sorting when cells are in contact on a 2D matrix surface. Early in development, ECs expressing these different markers are intermixed, but with time they segregate to either the arterial or venous side (Herbert et al., 2009; Kim et al., 2008; Wang et al., 1998). Differential sorting is observed very early in development even at the level of initial cardinal vein formation due to sprouting from the aorta (Herbert et al., 2009). This process appears analogous to that described for lymphatic sprouting and development from the cardinal vein (Yaniv et al., 2006). Also, Notch signaling controls this phenomenon and when constitutively active Notch-4 is expressed in ECs, venous ECs inappropriately express ephrinB2 which contributes to the development of arteriovenous malformations (Kim et al., 2008). Interestingly, many of these lesions regress following withdrawal of activated Notch4. Repulsive ephrinB2-EphB4 interactions should occur within vascular guidance tunnels and the ability of ECs to segregate following such interactions requires their ability to migrate on 2D matrix surfaces within tunnels to assemble together with other arterial or venous ECs. Disruption of either arterial or venous differentiation leads to major vascular developmental defects or vascular malformations (Kim et al., 2008), so it is likely that EC-EC interactions within vascular guidance tunnels play a major role in controlling appropriate cell sorting and differentiation functions.

The concept of vascular guidance tunnels also extends to recruitment and sorting of mural cells within the vessel wall. For example, ephrinB2 is also expressed on vascular smooth muscle cells with selectivity in arteries so similar repulsive interactions are likely to also control differential mural cell sorting in the arterial versus venous vasculature (Foo et al., 2006). Interestingly, mural cell recruitment to EC-lined tubes and within vascular tunnel spaces occurs in a polarized fashion exclusively on the EC ablumenal surface (Stratman et al., 2009a), which would be required in order for such mural cell-mural cell repulsive interactions to allow them to be sorted similarly to ECs. Thus, mural cells could sort through repulsive interactions with each other but may also occur through repulsive interactions with ECs. The vascular guidance tunnel matrix conduit is a critical ECM structure that is necessary in order for these EC-EC, mural cell-mural cell, and EC-mural cell interactions to occur and to also allow for motility events required for proper sorting along vessels in a 3D matrix environment. In support of these possibilities are that ECs and mural cells are highly migratory during vascular tube assembly within vascular guidance tunnels during tube co-assembly and maturation events (Stratman et al., 2009a; Stratman et al., 2010).

5.6. Vascular guidance tunnels are matrix conduits that allow for dynamic EC-pericyte interactions necessary for vascular tube maturation and basement membrane matrix assembly events

Although pericytes were known to accumulate along the EC ablumenal tube surface, how this recruitment and how such interactions occurred in 3D tissue spaces were not previously well understood. Our new findings show that EC tubulogenesis leads to the generation of vascular guidance tunnel networks within 3D ECM and that these matrix conduits are physical spaces where ECs and pericytes interact to regulate EC tube remodeling, maturation and stabilization events (Stratman et al., 2009a). Thus, pericytes that recruit to EC-lined tubes are actually within tunnel spaces (Figure 8). These matrix spaces allow for rapid EC and pericyte motility within them to control these events (Stratman et al., 2009a). Interestingly, both EC and pericyte motility within these pre-formed spaces is MMP-independent. Real time videos show for the first time the ability of pericytes to actively recruit to tubes but also to move along EC tube ablumenal surfaces to regulate tube maturation and vascular basement membrane matrix assembly (Stratman et al., 2009a). Pericyte recruitment to EC tubes in our defined serum-free system depends on EC-derived PDGF-BB and HB-EGF which controls directed pericyte motility and proliferation in 3D collagen matrices (Stratman et al., 2010). Also, this recruitment process leads to vascular basement membrane matrix assembly in between ECs and pericytes and within vascular guidance tunnels (Figure 7). Blockade of PDGF-BB and HB-EGF interferes with pericyte accumulation within vascular guidance tunnels and along EC tubes and importantly, vascular basement membrane assembly is blocked despite the close proximity of ECs and pericytes (Stratman et al., 2010). Thus, direct EC and pericyte contact within vascular guidance tunnels is necessary to affect basement membrane formation around EC tubes.

6.1. Pericyte recruitment to EC-lined tubes regulates vascular tube maturation and stabilization

Many studies indicate that microvessels covered with pericytes are more stable to pro-regressive stimuli and they show reduced vascular permeability indicative of tube stabilization. Interestingly, angiopoietin-1 is a growth factor that acts through EC Tie-2 to facilitate this tube stabilization process (Thomas and Augustin, 2009). Pericytes are believed to be a major source of angiopoietin-1 to act on ECs to facilitate tube stabilization in part through EC junction stability (Thomas and Augustin, 2009) (Figure 14). Delivery of recombinant angiopoietin-1 in vivo has been shown to decrease vascular permeability to angiogenic vessels (Thurston et al., 2000). Different vascular beds have varying pericyte numbers covering capillary networks although many have approximately 20–25% coverage of pericytes relative to ECs. Individual pericytes can span to touch multiple ECs which resemble other types of supporting cells such as glia in the nervous system interacting with multiple neurons. Tissues such as the central nervous system (including the retina) have a very high pericyte to EC ratio which approaches 1:1 (Armulik et al., 2005). Recent work suggests that pericyte recruitment to EC tubes plays a role in the development of the blood-brain barrier which shows strongly increased barrier functions with reduced vascular permeability relative to other vascular beds (Armulik et al., 2010; Daneman et al., 2010). Considerable work suggests that the high expression of VEGF within tumors is one reason why pericyte coverage is decreased compared to normal vascular beds (Greenberg et al., 2008; Jain, 2005) and why the vessels within the tumor microenvironment are highly permeable. Interestingly, treatment with VEGF antagonists leads to increased pericyte coverage which results in improved microcirculatory function (i.e. vascular normalization) with decreased permeability (Greenberg et al., 2008; Jain, 2005). This work has lead to a new therapeutic strategy to improve drug delivery into tumors.

6.2. Molecular mechanisms underlying why pericytes are able to stabilize EC-lined tube networks in 3D extracellular matrices

A major question that has not been sufficiently addressed is why pericyte coverage stabilizes vessels and what their functions are when they arrive at the EC ablumenal surface. To address this question, we established novel EC-pericyte coculture models in 3D collagen matrices and have developed systems using either bovine retinal pericytes or human brain pericytes mixed with human ECs (typically HUVECs). Both pericyte populations express the pericyte markers, NG2 proteoglycan, PDGFRβ, 3G5 ganglioside, smooth muscle actin and desmin. Perhaps the most important known function of pericytes is to recruit to microvascular capillary beds which then regulates vascular tube maturation and stabilization. Using a model of EC vasculogenic tube assembly, we have developed a system whereby ECs and pericytes are randomly mixed at a 5:1 or 5:1.25 ratio (i.e. 20–25% pericytes compared to 100% of ECs) under serum-free defined conditions (Stratman et al., 2009a). Remarkably, ECs form tube networks which are then followed by pericyte recruitment to these tubes (Figure 8). This ratio of ECs to pericytes is particularly optimal for tube assembly and subsequent pericyte recruitment but the reasons for this are currently not clear. Perhaps too many pericytes (through their production of TIMP-3) relative to ECs (Saunders et al., 2006) can interfere with morphogenesis, by inhibiting MT1-MMP-dependent signaling, or by counteracting the ability of ECs to find other ECs to properly form multicellular tubes.

Furthermore, EC tube diameters from EC-only cultures eventually become much wider than EC tubes from EC-pericyte cocultures. Vascular tube diameters reached a range of 20–25 µm in EC-pericyte cocultures which are vessel diameters observed in vivo during vasculogenesis, while in EC-only cultures, diameters can reach 80–100 µm over a 5 day period (Stratman et al., 2009b). Thus, pericytes have the ability to negatively regulate vascular tube diameters which may have to do with the induction of TIMP-2 and TIMP-3 from ECs and pericytes, respectively. As discussed earlier, these TIMPs are induced in EC-pericyte cocultures (Figure 14) (Saunders et al., 2006) and can restrict EC lumen diameters. Furthermore, pericyte recruitment leads to basement membrane assembly and laminins can inhibit tube morphogenesis and may contribute to vascular diameter as reported using laminin knockout embryoid body cultures where vessel formation occurs but with wider lumens than controls (Jakobsson et al., 2008). A number of studies indicate that vascular diameters are greater when pericyte recruitment is reduced or when ECM components such as fibronectin are knocked out of ECs during vasculogenesis in vivo (Astrof et al., 2007; Francis et al., 2002). We recently demonstrated the same phenomenon when PDGF-BB and HB-EGF are inhibited together which blocks pericyte recruitment to EC tubes, a condition leading to increased EC tube widths in vitro and in vivo (Stratman et al., 2010).

Previous work indicated that pericytes could be attracted to ECs on 2D plastic surfaces through PDGF-BB (Bjarnegard et al., 2004; Gaengel et al., 2009; Hirschi et al., 1998). Interestingly, EC-specific knockout of PDGF-BB leads to about a 50% decrease in pericyte coverage of vessels (Bjarnegard et al., 2004), although it is unclear whether this is due to effects on pericyte proliferation, recruitment or both. Of interest is that these mice show primary defects in microvessel beds (where pericyte coverage is present) while larger vessels are much less affected. This microvascular deficiency phenotype manifests particularly in the kidney and central nervous system which strongly resembles that observed in diabetic microangiopathy (Bjarnegard et al., 2004). Loss of pericytes is a major pathogenic cause of this type of microvascular disease (Hammes, 2005). It is important to further understand the signaling mechanisms which underlie pericyte recruitment to EC-lined tubes to both identify other factors that regulate this recruitment as well as understand how pericytes invade 3D matrices to recruit to these tubes.

A new study by our group addresses these questions using our EC-pericyte tube coassembly model under defined serum-free conditions. The work reveals that EC-derived PDGF-BB and HB-EGF control pericyte recruitment to EC-lined tubes in 3D collagen matrices (Stratman et al., 2010). Under these conditions, pericyte motility in 3D collagen matrices depends on the co-presence of ECs which produce both PDGF-BB and HB-EGF (Stratman et al., 2010). Without ECs, pericyte motility is minimal and shows no directional component. This was demonstrated using blocking antibodies to the two growth factors or using soluble receptor traps with affinity for these factors. To confirm these findings, chemical inhibitors directed to PDGF and EGF receptors or blocking antibodies to PDGF-BB and HB-EGF interfered with pericyte recruitment to EC tubes when administered in vivo to developing quail embryos (Stratman et al., 2010).

6.3. Pericyte recruitment to EC-lined tubes stimulates ECM remodeling events including vascular basement membrane matrix assembly

Also, we sought to elucidate how pericytes, through their recruitment to tubes, affect vascular tube maturation and stabilization events. At different time points of tube co-assembly, immunofluorescence microscopy and transmission electron microscopy were performed to examine if basement membrane matrix assembly occurred (Stratman et al., 2009a). In Figure 7, we demonstrate immunostaining for the EC-specific marker, CD31, while the pericytes express green fluorescent protein (GFP), and are shown to be recruited to the EC tube ablumenal surface. This is further demonstrated in cross-sections which show both EC lumens and associated pericytes selectively on this ablumenal surface (Figure 8). In addition, there is marked deposition of collagen type IV, a critical basement membrane matrix component along this ablumenal EC surface (Figure 7). We recently reported that fibronectin, laminin, nidogen-1, nidogen-2 and perlecan were also strongly deposited around EC-lined tubes only when ECs and pericytes were cocultured (Stratman et al., 2009a). Furthermore, we demonstrated that basement membrane matrices were observed by transmission electron microscopy only when EC-pericytes were cocultured (Stratman et al., 2009a) (Figure 7). In our work over many years of study, we have never observed basement membrane deposition around EC tubes in the absence of pericytes using electron microscopic analysis. Importantly, these results were confirmed in vivo showing that pericyte recruitment to developing quail EC tubes directly correlates with vascular basement membrane assembly at day 7 of embryonic development and this was also documented using both transmission electron microscopy and immunofluorescent staining (Stratman et al., 2009a). Prior to pericyte recruitment, no vascular basement membranes around EC tubes were observed in vivo. Also, we recently reported that blockade of pericyte recruitment in vitro or in vivo blocks vascular basement membrane assembly, despite the close proximity of both cell types (Stratman et al., 2010). This result is very intriguing and suggests that physical heterotypic cell-cell contacts and pericyte motility along tubes are necessary events for basement membranes to form.

The described immunostaining experiments were performed utilizing detergent-free conditions so that only ECM that is deposited extracellularly was examined (i.e. excluding intracellular ECM molecules) (Stratman et al., 2009a). We utilized this strategy for our in vitro 3D cultures, but also stained an in vivo tissue, the quail chorioallantoic membrane, in an identical manner (Stratman et al., 2009a). This is a central point because we demonstrated that extracellular deposition of vascular basement membrane matrix is markedly stimulated by pericyte recruitment both in vitro and in vivo (Stratman et al., 2009a).

Human EC and bovine pericyte cocultures were utilized to determine which cell types produce particular ECM components over time to regulate basement membrane matrix assembly (using species-specific RT-PCR primer sets). ECs were shown to increase their production of fibronectin selectively in the presence of pericytes (and not in their absence) and nidogen-1 was induced in pericytes selectively when ECs were present (Stratman et al., 2009a). Also, we observed induction of particular laminin isoforms as well as perlecan at the mRNA level that occurred in response to EC-pericyte interactions. Thus, EC-pericyte contacts during tube coassembly events affected mRNA and protein levels for key basement membrane matrix molecules (Stratman et al., 2009a). Interestingly, both fibronectin and nidogen-1 are known to bridge basement membrane matrix proteins together which are essential for the assembly of this specialized matrix (Davis and Senger, 2005; Miner and Yurchenco, 2004; Senger and Davis, 2010). For example, fibronectin binds collagen type IV and perlecan, while nidogen-1 binds collagen type IV and laminin isoforms. Thus, it is possible that fibronectin and nidogen-1 create a nidus which leads to assembly of the insoluble basement membrane matrix around EC-lined tubes. Most ECM proteins possess self-assembly functions but it is also known that they need to interact with one another to create the complex meshwork that characterizes an assembled basement membrane matrix (Miner and Yurchenco, 2004). Collagen type IV, a fundamental basement membrane component that is greatly responsible for its structural integrity, shows affinity for both fibronectin and nidogen-1, and this may be one reason why EC-pericyte interactions are necessary for this essential ECM remodeling event during vascular wall assembly (Stratman et al., 2009a). Interestingly, pericyte-induced fibronectin assembly around EC tubes appears also to directly affect collagen type IV assembly (Stratman et al., 2009a). To assess which cell types contributed to collagen type IV assembly during this process, siRNA suppression experiments were performed showing that ECs were the major source of deposited collagen type IV. Knockdown of collagen type IV in ECs strongly reduced collagen type IV deposition and this resulted in increased vascular tube width (Stratman et al., 2009a). Knockdown of collagen type IV in pericytes had a lesser influence, but nonetheless an inhibitory effect was detected. Thus, ECs and pericytes both contribute collagen type IV during vascular basement membrane assembly in 3D matrices (Figure 7).

We further addressed how continuous basement membrane assembly is accomplished along EC-lined tubes despite the fact that pericytes represent only one-fifth to one-fourth of the total number of ECs. Our data suggests that this occurs due to the dramatic motility of pericytes along the ablumenal EC tube surface (demonstrated by real-time videos of these events) which scan along the tubes to stimulate the deposition of the basement membrane in a continuous manner (Stratman et al., 2009a) (Figure 7). We speculate that the movement of both pericytes and ECs along each other within vascular guidance tunnels exerts mechanical forces on newly deposited ECM components to facilitate basement membrane assembly. Thus, it is intriguing that fibronectin, a mechanosensitive ECM component and whose assembly is facilitated by cell-exerted tensional forces (Smith et al., 2007; Vogel, 2006; Zhong et al., 1998; Zhou et al., 2008), is a basement membrane component that strongly deposits along EC-lined tubes when pericytes are present in vitro and also in vivo (Stratman et al., 2009a). Disruption of pericyte recruitment in vivo leads to markedly decreased fibronectin deposition around EC tubes and fibronectin is known to play a fundamental role during cardiovascular development in mouse knockout studies (Francis et al., 2002; Hynes, 2007). Another interesting possibility is that the polarized presence of pericytes along the EC ablumenal surface stimulates the directional secretion and selective deposition of basement membrane components from both cell types toward the other. Thus, both mechanical forces and vectorial secretion mechanisms may account for how pericyte recruitment to EC-lined tubes leads to vascular basement membrane matrix assembly. Further work needs to define these issues in more detail and also examine whether production of particular ECM components or isoforms made by either ECs or pericytes might play crucial roles during this fundamental process.

6.4. Pericyte recruitment and vascular basement membrane matrix assembly alter integrin requirements in both ECs and pericytes to regulate tube maturation events

Again using a co-culture system with human ECs and bovine pericytes, EC versus pericyte integrin expression was assessed during this process. mRNA levels were determined for both cell types and function blocking experiments with anti-integrin monoclonal antibodies were performed at various times during the EC-pericyte tube coassembly process. Importantly, blocking antibodies to the fibronectin-binding α5β1 integrin had function blocking effects that selectively occurred in the EC-pericyte co-cultures and not in EC-only cultures (Stratman et al., 2009a). Also, the EC α5 integrin subunit mRNA was induced in EC-pericyte co-cultures, and not in EC-only cultures where it was downregulated. Overall, an important theme that was evident from these studies is that integrins with affinity for basement membrane components were upregulated in both ECs and pericytes in cocultures, but not in EC- or pericyte-only cultures (Stratman et al., 2009a). In contrast, the α2 integrin, which recognizes collagen type I, is downregulated in cocultures but is upregulated in EC-only cultures. Thus, as basement membranes assemble around EC tubes in EC-pericyte cocultures, the direct interaction of ECs with collagen type I decreases, while their contact with basement membrane matrix components increases. Concomitantly, increases in the expression of integrin α5, α3, and α6 from ECs were observed, which recognize fibronectin, nidogens and laminin isoforms, while α5, α3, α6 and α1 integrin subunits were increased from pericytes which recognize fibronectin, nidogens, laminin isoforms and collagen type IV (Stratman et al., 2009a). Also, functional deficits were observed when blocking antibodies to these integrins were administered to cocultures (but not EC-only cultures) (Stratman et al., 2009a). This data indicates that a major purpose of the multiple β1 integrins on the EC cell surface is to recognize key ECM components that they encounter at different stages of the tube morphogenic and later maturation process.

When exposed to collagen type I matrices, which serves as a strong agonist for tubulogenesis, ECs utilize collagen-binding integrins such as α2β1. However, when EC-lined tubes attract pericytes, vascular basement membranes form that are recognized by different sets of integrins such as α5β1 (a fibronectin receptor), α3β1 (a nidogen, and laminin isoform receptor), α6β1 (a laminin isoform receptor) and α1β1 (a collagen type IV, collagen type I and laminin receptor). Interestingly, α1β1 appears to be predominantly pericyte-derived during this tube coassembly process and thus, the effects of blocking antibodies that have been observed may be due to a specific inhibitory influence on pericyte recognition of basement membrane matrices (Stratman et al., 2009a). Also, EC-dependence on α2β1 which is observed over time in EC-only cultures is lost with time in EC-pericyte cocultures as basement membrane matrix assembly occurs. Thus, exposure of ECs to collagen type I is strongly diminished while EC exposure to basement membrane proteins increases. These studies reveal that EC-pericyte interactions control vascular basement membrane matrix assembly and that critical changes occur in EC and pericyte integrin expression to recognize this newly remodeled matrix to facilitate further tube maturation events.

6.5. Pericyte-derived TIMP-3 controls vascular basement membrane matrix assembly by facilitating collagen type IV deposition and stability

Another important role of pericytes during these tube assembly processes is the delivery of TIMP-3, a basement membrane- and ECM-binding protein (Figures 10 and 14). TIMP-3 has been previously shown to play a critical role in pericyte-induced tube stabilization by blocking MMP-1, MMP-10, and ADAM-15, which together promote vascular regression events (Saunders et al., 2006) (Figures 10 and 14). Also, TIMP-3 can inhibit further EC morphogenic events by blocking MT1-MMP function (Saunders et al., 2006). New data shows that TIMP-3 plays as yet another role by facilitating collagen type IV assembly in EC-pericyte cocultures. siRNA suppression of pericyte TIMP-3 results in markedly decreased collagen type IV deposition around developing tubes in 3D collagen matrices (Stratman et al., 2009a). The mechanism by which TIMP-3 exerts its effects is not entirely clear but a major effect likely relates to its inhibition of MT1-MMP which can degrade type IV collagen. With decreased collagen type IV assembly around EC tubes when pericyte TIMP-3 expression is suppressed, there was a significant increase in vessel diameter (Stratman et al., 2009a). Thus, collagen type IV assembly around developing tubes may be a primary determinant of vessel diameter. It is intriguing to consider the fact that EC-only tubes, which are not surrounded by basement membrane matrices, become very wide during morphogenic events (secondary to continued morphogenesis) suggesting a lack of inhibitory signals. Those inhibitory signals are likely to be TIMP-3 and specific ECM components including laminin isoforms and collagen type IV that are delivered to ECs following interactions with pericytes (and which results in much narrower tube widths) (Stratman et al., 2009a). The marked differences in vessel diameter in these two situations demonstrate functional evidence for basement membrane assembly but also demonstrate that ECs recognize the deposited proteins and respond by restricting tube diameter. Thus, EC tube diameter is an important indicator of functional EC-pericyte interactions that lead to vascular stabilization.

7.1. MMP-1, MMP-10 and ADAM-15 control vascular tube regression responses in 3D collagen matrices

A variety of past and recent studies have identified MMPs which control vascular tube regression events (Davis et al., 2001; Saunders et al., 2005; Zhu et al., 2000). The secreted MMPs, MMP-1 and MMP-10, were shown to be highly induced during EC tube morphogenesis but interestingly, neither enzyme could be shown to be involved in tube formation in 3D collagen matrices (Saunders et al., 2005). In contrast, when pro-MMP-1 and pro-MMP-10 are activated by plasmin, plasma kallikrein or other serine proteinases, these MMPs fail to enhance tube formation, but dramatically cause EC tube network regression and collapse (Davis et al., 2001; Davis and Saunders, 2006; Saunders et al., 2005) (Figure 14). Importantly, the tube morphogenic systems that our laboratory has developed are performed in the absence of serum (Koh et al., 2008b) and thus, there is no source of plasminogen or plasma pre-kallikrein which are typically supplied in serum and also produced by the liver in vivo. Interestingly, this indicates that EC tube formation as well as pericyte recruitment and tube stabilization in 3D collagen matrices does not require plasminogen or plasma pre-kallikrein addition. When plasminogen is added, it is converted to plasmin through plasminogen activators and this leads to activation of both pro-MMP-1 and pro-MMP-10. The addition of plasminogen does not affect tube formation, but interestingly, it induces dramatic tube regression phenomena following MMP-1 and MMP-10 activation. As expected, addition of MMP inhibitors such as TIMP-1 which can block MMP-1 and MMP-10, but not MT1-MMP (and thus does not block EC lumen and tube formation), can interfere with plasmin- and plasma kallikrein-dependent tube regression since the process is mediated through these two MMPs (Davis et al., 2001; Saunders et al., 2005). In support of these findings, siRNA suppression of either MMP-1 or MMP-10 had no affect on tube formation but markedly blocked tube regression following addition of plasminogen or plasma kallikrein (Saunders et al., 2005). Interestingly, a mouse knockout of histone deacetylase 7 (HDAC7) caused a vascular hemorrhage phenotype in vivo during vascular development leading to embryonic lethality (Chang et al., 2006). siRNA suppression of HDAC7 resulted in marked increases in MMP-10 and marked decreases in TIMP-1 expression which appear to control this vascular developmental regression response (Chang et al., 2006).

We also demonstrated that ADAM-15 plays a role during these events in vitro, although its specific role was not elucidated in terms of its molecular targets. ADAM-15 is known to be expressed in EC junctional contacts and is also capable of inducing shedding of cell surface molecules (Ham et al., 2002). Recently, ADAM proteinases and MMPs have been shown to induce shedding of multiple EC surface proteins including VE-cadherin and VEGFR2, which could play a role in vessel regression events (Schulz et al., 2008; Swendeman et al., 2008). Also, MMPs and plasmin are known to cleave VEGF that controls its release from ECM (Lee et al., 2005) which might make it more susceptible to inhibition by soluble VEGF traps such as secreted forms of VEGFR1.

7.2. EC-pericyte interactions induce the expression of TIMP-2 and TIMP-3 to block MMP-1-, MMP-10- and ADAM-15-dependent regression leading to vascular tube stabilization

Importantly, we have demonstrated that pericytes, when added to EC cultures can interfere with these plasminogen-induced regression events (Figures 10 and 14). Both EC-derived TIMP-2 and pericyte-derived TIMP-3 were shown to be involved in controlling the ability of EC-pericyte cocultures to resist this mechanism of tube regression (Saunders et al., 2006). This work was an important demonstration of how pericytes contribute to EC tube stabilization by preventing EC tube regression mechanisms. EC tube regression leads to EC tube collapse and EC apoptosis where caspase-dependent cleavage of molecules such as procaspase-3, gelsolin and Pak-2 are observed (Bayless and Davis, 2004; Davis et al., 2001) (Figure 14). It is well known that pericytes protect EC tubes from pro-regression stimuli (Benjamin et al., 1999; Benjamin et al., 1998), but little information previously existed to explain the molecular basis for such phenomena. Multiple factors appear to contribute to pericyte-induced EC tube stability including the induction of vascular basement membranes, presentation of cytokines such as angiopoietin-1 to oppose EC-derived angiopoietin-2 (Thomas and Augustin, 2009), and delivery of TIMP-2 and TIMP-3 to block pro-regressive MMPs and ADAMs (Saunders et al., 2006) (Figures 10 and 14). Interestingly, both TIMP-2 and TIMP-3 are able to block VEGFR2 mediated signaling (Qi et al., 2003; Seo et al., 2003) (Figure 14). Thus, EC-pericyte interactions can lead to vascular stabilization through inhibition of both vessel formation and regression mechanisms.