Cellular and Molecular Mechanisms of Vascular Lumen Formation

Abstract

The formation of vascular lumens by endothelial cells is a critical step in the angiogenic process that occurs during invasion and growth of the incipient vascular sprout. Once a lumen is established, capillaries are rapidly exposed to the physical forces associated with the flow of blood which, together with genetic information, regulate the ultimate size of inner vessel diameter. Here we review the recent literature on vascular lumen formation and compare it to lumen formation in other epithelial systems. We also discuss the regulation of lumen diameter after vascular morphogenesis has been completed.

Introduction

The organization of the vascular tree requires coordinated interplay of genetic, microenvironmental, and epigenic factors. The final outcome is a network of hierarchical interconnected tubules that efficiently perfuses tissues, with the strength to absorb the flow and pressure produced by cardiac contractions and yet the flexibility to adapt to specific homeostatic needs. As our understanding of the processes that regulate the emergence and differentiation of blood vessels improves, the attention of the field is shifting to questions associated with more complex aspects of the assembly of 3D vascular structures. For example, how do blood vessels remodel into hierarchical structures that are able to regulate blood pressure and control efficient perfusion of tissues? How do cell-cell interactions convey organ-specific differences to the endothelium and impact the structure of capillaries? How is the vascular tree able to adapt and respond to the distinct physiological needs of individual organs, and how is this affected in disease states? Many of these questions require a sophisticated understanding of how the lumens of endothelial cell (EC) tubular networks are formed and maintained and how they are regulated in embryonic versus adult tissues.

Our current understanding of how lumens originate in blood vessels is modest, but evolving rapidly. A major challenge is the difficulty of establishing appropriate models, systems, and technology in vitro and in vivo to enable molecular dissection of these events. A number of in vitro models have now been used to elucidate molecular requirements for EC lumen and tubule formation in 3D extracellular matrix (ECM) environments (Aplin et al., 2008; Davis et al., 2007; Davis and Senger, 2005; Holderfield and Hughes, 2008; Koh et al., 2008b; Nakatsu and Hughes, 2008). One of the challenges we face in this analysis is dissociating the early sprouting process from lumen formation, as they occur concurrently. Unlike epithelial cells, ECs form lumens as they invade tissues. In addition, their squamous shape and their relative lack of specific apical/basolateral markers have made it extremely difficult to address mechanistic questions of lumen formation in vivo. Like epithelial lumen formation (Bryant and Mostov, 2008; Levi et al., 2006; Lubarsky and Krasnow, 2003; Martin-Belmonte and Mostov, 2007), the morphogenesis of tubes in vascular structures requires the coordinated participation of multiple molecules including small GTPases, cell surface receptors, and cell-matrix and cell-cell adhesion proteins (Adams and Alitalo, 2007; Avraamides et al., 2008; Bazzoni and Dejana, 2004; Davis et al., 2007; Dejana, 2004; Dejana et al., 2009; Horowitz and Simons, 2008; Koh et al., 2008a) (Figure 1). However, in contrast to epithelial biology, progress in understanding vascular lumen formation has not been able to benefit significantly from genetic screens in Drosophila. Finally, although some of the aspects of lumen formation seem to be conserved between epithelium and endothelium, the emerging details about molecular players and regulatory mechanisms suggest that there are also important differences.

Figure 1. Cellular and Molecular Regulation of Vascular Lumen Formation.

The morphogenesis of lumens in ECs occurs concomitantly with the invasion of vascular sprouts. Either as single or multiple cellular aggregates, the first cellular indication of lumen formation is the presence of large intracellular vesicles (left panel). Upon fusion, the vesicles form an incipient lumen (central panel) that aligns with the patent circulation. Molecular requirements for these events include the activity of integrins, Cdc42 and Rac, in addition to Pak2, Pak4, and the polarity complex Par3/Par6/PKCζ. The activity of proteinases, in particular MT1-MMP, is also required.

The process of lumen formation is extremely efficient and it enables ECs to immediately cope with physical forces imposed by shear stress, as well as a complex combination of intermittent, laminar, and turbulent flow patterns. During this process, homotypic cell-cell interactions are ready to withstand these physical forces initially in the absence of additional support from mural cells. Although the diameter of vessels appears to be determined genetically, it has been demonstrated that flow is essential for maintenance and regulation of lumen diameter (Jones et al., 2006). Some of the genetic regulators of vascular lumen size are emerging through serendipitous findings from mammalian genetic models, but the interplay between genetics and mechanotransduction has yet to be explored in detail.

The inner diameter of vessels impacts blood pressure and therefore perfusion of both nutrients and the blood cells that control oxygen delivery and immunological surveillance. Many of the factors that influence EC luminal integrity, or patency, have a broad functional impact. Thus, understanding the control and maintenance of lumen size and the factors that regulate lumen diameter is of high pathophysiological relevance. Here, we have compiled current knowledge on the molecular process of lumen formation by ECs and the regulatory mechanisms that establish, and later maintain, the inner diameter of blood vessels.

Cellular Mechanisms Underlying Vascular Lumen Formation

In an important conceptual review a few years ago, Lubarsky and Krasnow described five potential mechanisms by which epithelial cells could form lumens and tubular structures during morphogenetic processes (Lubarsky and Krasnow, 2003). These were (1) wrapping, whereby a planar cell sheet wraps to form a tube; (2) budding, where a sprout emanates from a pre-existing tube; (3) cavitation, where a space is created by elimination of centrally placed cells in a cell-cell aggregated sphere or cylinder; (4) cord hollowing, in which a cord or cylinder of packed cells creates a central space by cell shape changes, such as flattening of cells along the wall of the cylinder; and (5) cell hollowing, whereby individual cells create intracellular spaces within them to generate a lumenal structure. Considerable work by a number of laboratories suggests that three of these mechanisms, budding, cord hollowing, and cell hollowing, operate in vascular ECs during developmental or postnatal angiogenic events (Adams and Alitalo, 2007; Davis et al., 2002, 2007; Davis and Senger, 2005; Egginton and Gerritsen, 2003; Holderfield and Hughes, 2008). In our view, the budding and cord hollowing mechanisms may be essentially synonymous in the context of angiogenic sprouting. Leading EC tip cells (Gerhardt et al., 2003) invade matrices, creating spaces that can be occupied transiently by a cord of trailing cells without an apparent lumen, but in other instances the lumen becomes more immediately apparent in the trailing trunk cells (Bayless and Davis, 2003; Davis et al., 2007; Gerhardt et al., 2003; Holderfield and Hughes, 2008; Saunders et al., 2006). Despite these temporal distinctions, there is a common underlying process in which trunk ECs flatten onto the wall of a matrix space created by the lead invasive cell, initiating a lumenal area and leading to a tube that develops from the trailing cells. To illustrate these events, we have included EC tube morphogenic movies in 3D collagen matrices in which invading EC tip cells lead invasion followed by the appearance of EC lumenal structures (see Movies S1–S3 available online). These movies illustrate that vessel morphogenesis is a highly dynamic process in which invasion, motility, and lumenogenesis occur concurrently in different vessels or different regions of the developing tube.

By contrast, cell hollowing, or intracellular vacuolation, is a mechanism by which individual cells generate vesicles that, after exocytic events, enable the cells to interconnect with neighbors to form multicellular lumens and tubes (Figure 1, Movie S4) (Bayless and Davis, 2002; Bayless et al., 2000; Davis and Bayless, 2003; Davis et al., 2007; Egginton and Gerritsen, 2003; Folkman and Haudenschild, 1980; Kamei et al., 2006; Koh et al., 2008a). The intracellular vacuolation mechanism has been most frequently observed when single ECs are initiating morphogenesis in a situation where they do not have contact with adjacent cells, as occurs during vasculogenesis in a 3D environment (Adams and Alitalo, 2007; Drake, 2003; Risau and Flamme, 1995). The intracellular vacuolation process is also extremely dynamic and it is a rapid way to create EC luminal spaces, as observed in vitro and in vivo (Bayless and Davis, 2002; Bayless et al., 2000; Davis and Camarillo, 1996; Kamei et al., 2006; Koh et al., 2008a). Intracellular vacuolation is also noted in matrix invasion assays from monolayers (Movies S2 and S3) in which ECs are associated with neighboring cells, so it is not restricted to individual ECs. As intracellular vacuoles have been demonstrated in vivo in many cell types that make lumenal structures (in both epithelial and endothelial cells), and since single cells can make lumens in vivo and in vitro (Bagnat et al., 2007; Davis and Bayless, 2003; Lubarsky and Krasnow, 2003), it is clear that this mechanism plays a role during tubulogenesis. In Movie S5, EC lumens appear to form primarily by EC flattening against a pre-existing matrix space, and lumen expansion is most likely propelled by continuous membrane-bound matrix metalloproteinase (MMP)-dependent proteolysis (Davis et al., 2007; Saunders et al., 2006). In this case, intracellular vacuoles are much less prominent. We believe it likely that the different mechanisms discussed above occur in a context-dependent manner, but in many instances they might also occur concurrently (see Movies S1–S5).

A common feature of all these different mechanisms is that the ECs create physical spaces within the 3D ECM via surface-located proteolysis as they form lumen and tube networks (Davis et al., 2007). ECs can flatten and assume the characteristic cobblestone appearance along the wall of these spaces through EC-matrix contacts and then connect with adjacent cells through intercellular junctional adhesions (Bazzoni and Dejana, 2004). Thus, in many respects these potentially distinct initiating mechanisms are likely to have common molecular and signal transduction requirements. For example, ECs seeded as single cells or as aggregates require MMPs to form lumenal structures in 3D collagen matrices (Davis and Saunders, 2006). Similarly, membrane MMPs such as MT1-MMP are required for ECs to sprout and form tube structures from either an EC monolayer surface or an intact tissue such as an aortic ring (Chun et al., 2004; Lafleur et al., 2002; Saunders et al., 2006). In some instances invading cells appear to organize cord-like structures first and only subsequently lumenize, while on other occasions the lumen is observed earlier just following the advancing tip cell.

Signaling Mechanisms Controlling Formation of EC Lumens and Tubes

Most of our current understanding of signaling processes during EC lumen formation has come from analyses in culture using 3D ECM environments, with some supporting evidence from zebrafish and knockout mice. These analyses have shown that integrin-ECM interactions play a critical functional role during vascular morphogenesis to regulate sprouting, lumenogenesis, and tube stabilization (Avraamides et al., 2008; Davis and Senger, 2005; Mahabeleshwar and Byzova, 2008; Rupp and Little, 2001; Stupack and Cheresh, 2004). Substantial evidence suggests that several integrins and ECM proteins play unique roles during these events. For example, collagen (Davis and Senger, 2005), fibrin-fibronectin (Holderfield and Hughes, 2008), and fibronectin-rich embryonic matrices (Astrof et al., 2007; Hynes, 2007) represent promorphogenic environments, whereas basement membrane proteins appear to be inhibitory or nonpermissive for vascular morphogenesis (Davis and Senger, 2005, 2008). Pharmacological blockade of β1 integrin using antibodies resulted in complete absence of lumens in the aorta of chicken embryos (Drake et al., 1992). In addition to the contribution of the β1 family of integrins (α2β1, α1β1, α5β1, and α4β1), the αv integrins, αvβ3 and αvβ5, which recognize promorphogenic matrices, have also been shown to influence lumen formation in vitro (Figure 1) (Bayless et al., 2000; Calzada et al., 2004; Davis and Camarillo, 1996; Davis and Senger, 2005; Hood et al., 2003; Senger et al., 1997).

Downstream of integrin signaling, activation of Src and FAK kinases contributes to the process of lumen formation. A number of studies demonstrate essential roles for these kinases in early tube morphogenic events and tube stabilization (Eliceiri et al., 2002; Ilic et al., 2003; Im and Kazlauskas, 2007; Liu and Senger, 2004). EC-specific knockouts of FAK show embryonic lethality, and global FAK knockouts in mice reveal vascularization defects as well (Shen et al., 2005). Interestingly, early vasculogenic events were not disrupted in EC-specific FAK knockouts, while later events associated with tube stabilization appeared to be defective. A recent study shows that Pyk2, a FAK-related kinase, can compensate for EC-specific FAK knockout in adult mice (Weis et al., 2008). Thus, it appears that a combination of FAK and Pyk2 downstream of integrin signaling can regulate EC morphogenesis, and probably stabilization.

Integrin signaling also activates Rho GTPases, which regulate endothelial cytoskeleton and tube morphogenic responses (Bayless and Davis, 2002; Bryan and D’Amore, 2007; Connolly et al., 2002; Davis and Bayless, 2003; Hoang et al., 2004; Kiosses et al., 2002; Koh et al., 2008a). The first reports indicating a role for Rho GTPases in EC morphogenesis revealed the involvement of Cdc42 and Rac1 during these events (Bayless and Davis, 2002; Connolly et al., 2002; Kiosses et al., 2002). Cdc42 was shown to be a critical regulator of EC lumen formation along with Rac1 by expressing dominant-negative mutants of these GTPases (Bayless and Davis, 2002), and also, more recently by an siRNA approach (Koh et al., 2008a). EC-specific knockout of Rac1 or one of its downstream effectors, Wave2, leads to defects in vascular development, suggesting their involvement during these events in vivo (Tan et al., 2008; Yamazaki et al., 2003). Previous studies have shown that intracellular vacuole formation and coalescence occurs within ECs in vitro and that the intracellular vacuole compartment can be labeled by extracellular membrane-impermeant dyes, demonstrating that it arises via pinocytic events (Davis and Camarillo, 1996). To some extent, the EC intracellular vacuole formation and coalescence process resembles the process of macropinocytosis, which is also Cdc42 and Rac1 dependent. Both processes also require functional actin and microtubule cytoskeletons (Bayless and Davis, 2002; Davis and Bayless, 2003; Davis et al., 2007).

GFP-Rac1 and GFP-Cdc42 target vesicles to the intracellular vacuole compartment that controls EC lumen formation in vitro (Bayless and Davis, 2002). Similarly, GFP-Cdc42 labels intracellular vacuoles in ECs in vivo as shown by time-lapse imaging in zebrafish embryos. These vacuoles were found to uptake dye delivered by intravascular injection (Kamei et al., 2006). The findings suggest that this initially enclosed membrane space fuses with the developing luminal surface in individual cells through exocytosis, creating a lumen compartment. When this occurs in adjacent ECs, multicellular tubular structures form and assemble (see Movie S4) (Kamei et al., 2006).

An alternative view is that lumen formation initially requires cell-cell interactions mediated via junctional contacts between ECs (Blum et al., 2008). It is certainly possible, and actually quite likely, that multiple mechanisms contribute to lumen formation and expansion, as we discuss above. In fact, and as it will be discussed below in detail, the formation of lumens by the Drosophila endocardium suggests yet another distinct form of lumen formation that requires the redistribution of cadherins under the regulatory control of Slit-Robo (Medioni et al., 2008; Santiago-Martinez et al., 2008).

Downstream Effectors of Small GTPases in EC Lumen and Tube Morphogenesis

Rho GTPases such as Cdc42 and Rac1, in their GTP-bound form, bind effectors to influence signal transduction and modulate the actin and microtubule cytoskeletons. For example, Rac1 is known to activate p21-activated kinases (Paks), which influences EC tube morphogenesis (Kiosses et al., 2002; Koh et al., 2008a). In a recent study, both Pak2 (activated by both Rac1 and Cdc42) and Pak4 (selectively activated by Cdc42) were found to be required for ECs to form lumens and tubes in 3D collagen matrices (Koh et al., 2008a). These data were obtained using dominant-negative inhibitors of both kinases plus siRNA suppression of both genes, which strongly blocks EC lumenogenesis. The activation of these two kinases parallels the activation of Cdc42 and Rac1 during these events and directly correlates with the lumen formation process (Figure 1). Interestingly, their activation is strongly stimulated by protein kinase C (PKC) activation (which also stimulates EC lumen formation), particularly the novel PKC isoform PKCε (Koh et al., 2008a). Blockade of PKCε by either siRNA suppression or the use of pharmacologic inhibitors markedly attenuates the Pak2- and Pak4-dependent EC lumen formation response. The study also showed that Cdc42 forms a complex with activated Pak2 and Pak4 during the acquisition of patency by ECs, and that the polarity regulator Par3 was also part of the multiprotein complex (Koh et al., 2008a). Thus, in this signaling pathway downstream of integrin-ECM interactions, Cdc42 and Rac1 activate Pak kinases which, in conjunction with polarity proteins, controls lumen formation in incipient endothelial tubes (Koh et al., 2008a). It is likely that this signaling also controls the positioning of intracellular vacuoles and the organization of a lumen compartment within individual ECs to facilitate proper orientation and coordinate a multicellular tubular structure.

Recent studies also implicate cerebral cavernous malformation signaling proteins (e.g., CCM2) in the regulation of lumen formation and maintenance during mouse and zebrafish embryogenesis (Kleaveland et al., 2009; Whitehead et al., 2009). Underlying mechanisms controlling these phenomena include the ability of CCM2 to affect small GTPase signaling involving Rho and Rap GTPases. siRNA suppression of CCM2 in human ECs blocks lumen formation, as well as EC sprouting in vitro, which correlates with the inability of particular vessels to become patent following EC-specific knockout of CCM2 in vivo (Kleaveland et al., 2009; Whitehead et al., 2009).

Contribution of MMPs to Lumen Formation

ECM degradation is necessary to form lumen and tube structures in either 3D collagen or fibrin matrices (Davis et al., 2007; Davis and Senger, 2005; Lafleur et al., 2002). EC sprouting from either surface monolayers or pre-existing vascular walls into a 3D matrix also requires proteolysis. In particular, MMPs are necessary for these events, and the membrane-anchored MMP MT1-MMP is a major proteinase involved in this process (Chun et al., 2004; Saunders et al., 2006) (Figure 1). Recently, it has been shown that MT1-MMP is required for EC lumenogenesis in 3D collagen matrices and that it represents the major target of tissue inhibitor of metalloproteinases (TIMPs) such as TIMP-2 and TIMP-3, known to block EC morphogenic events (Saunders et al., 2006). Interestingly, TIMP-2 from ECs and TIMP-3 from pericytes represent an inhibitory pair of TIMPs that facilitate tube stabilization in maturing vessels by suppressing both promorphogenic and separate proregression stimuli that depend on different subsets of MMPs (Saunders et al., 2006). An interesting question that emerges from this work is how MT1-MMP interfaces with integrin signaling through Cdc42 and Rac1 to control the EC lumen and tube formation cascade.

A final point concerning the involvement of ECM degradation in EC lumen and tube formation is that physical spaces within the matrix are created during these events. These physical spaces have recently been termed vascular guidance tunnels and influence both EC motility and vascular remodeling in 3D matrix spaces (Davis et al., 2007). When perturbations occur in the patterning of vessels as a result of alterations in signaling or deficiency of a given required molecule (e.g., changes in Notch signaling), the malformed vessels will also have a coincident alteration in the pattern and organization of the matrix tunnel spaces that are generated during the abnormal tube morphogenic response. These abnormalities would also be expected in the context of the tumor vasculature. The consequences of such matrix alterations are not fully appreciated at the moment, but may have profound influences in later steps of vascular remodeling such as pericyte and/or vascular smooth muscle recruitment, which are necessary for vascular stabilization.

Cell Polarity in EC Lumen Formation

A key characteristic of both epithelial and endothelial cell lumens is that they possess apical/basolateral polarity with respect to a fluid-filled lumenal compartment and ECM that is associated with the basal surface. Maintenance of apical-basal polarity requires cell-cell junctional contacts and cell-ECM contacts on the basal surface (Bryant and Mostov, 2008). A fundamental question is how cell polarity mechanisms affect the process of EC lumenogenesis and how the ensuing events lead to the development of patent vessels with polarized apical and basolateral surfaces. These issues are just now being addressed by investigators in the endothelial field, and there are therefore many unanswered questions. We feel that it is important to point out that cell polarization in epithelial cells and ECs appears to be distinct and there is much less evidence for apical and basolateral sorting of proteins in ECs. The reasons for these differences are unclear, but they likely reflect the distinct molecular components within the different cell types and the critical functional differences between them, such as the exposure of ECs to blood flow, high shear stresses, and pressures, which are initiated early during development.

In broad terms, cell polarity is controlled by a complex series of proteins, including the Par proteins Par6 and Par3 and atypical PKC isoforms (such as PKCζ) (Etienne-Manneville and Hall, 2003a; Joberty et al., 2000; Lin et al., 2000; Macara, 2004). Interestingly, Par6 binds directly to Cdc42, a Rho GTPase required for centrosome reorientation during directed cell migration events (Etienne-Manneville and Hall, 2003a, 2003b). This Cdc42/Par6/Par3/atypical PKC pathway is now thought to interact with the tight junctional apparatus (Macara, 2004). As discussed above, Cdc42 and the Par6/Par3/atypical PKC complex are required for lumen formation by ECs in 3D ECMs (Bayless and Davis, 2002; Kamei et al., 2006; Koh et al., 2008a). Cadherins play an integral role in maintaining polarity in epithelial cells, and an association between Par3/Par6 and VE-cadherin in ECs has been reported recently, although the functional role has not yet been elucidated (Iden et al., 2006). Cell polarity in epithelial cells is also controlled by the Discs large/Scribble/Lgl proteins and Crumbs/PALS1/PATJ complexes, which facilitate epithelial junctional contacts to further regulate apical/basolateral polarity (Bryant and Mostov, 2008; Macara, 2004). At this point, there is little to no information on whether these latter proteins play any role in EC function and polarity regulation.

Lumen Formation and the Mechanosensory Function of the Endothelium

From an early point, a developing vascular lumen has to withstand a large variety of physical forces generated by circulating plasma and blood cells. In all vertebrates, the initiation of heart beating precedes the completion of vascular remodeling events; thus, developing endothelial tubes must be sufficiently stable to sustain pressure, shear stress, and flow forces. These forces generate remarkable changes in cell morphology and cytoskeletal organization, as well as alterations in cell-substrate and cell-cell junctional complexes (Alon and Ley, 2008; Garin and Berk, 2006; Heil and Schaper, 2004). It is well known that vessels regress when not constantly perfused and that they enlarge when exposed to increased flow and pressure. Although the media layer of the vessel is the most reactive in such situations, the endothelium is also significantly activated by flow (Davies, 2008; Gimbrone, 1999) and it contributes to the alterations in lumen size. Changes in fluid shear stress mediate EC activation by first promoting cell swelling, which partially contributes to hyperplasia. In an effort to antagonize this effect, ECs open ion channels to mediate efflux of osmolytes and enable the return to its normal volume (Nilius et al., 1996). Interestingly, blockade of these channels inhibits collateral growth and facilitates vascular dilation instead (Manolopoulos et al., 2000). While much remains to be understood, the mechanosensory function of the endothelium provides an important adaptative response for many organs and tissues. It also underlines the conclusion that lumen formation and its maintenance is highly responsive to vascular flow.

Dynamic Control of Vascular Lumen Diameter

In addition to characterizing mechanisms of EC lumen and tube formation, it is also critical to understand how vascular lumen size is established and how this diameter is maintained. In the case of blood vessels this is particularly important because the inner diameter of vessels impacts blood pressure, blood flow, and perfusion of tissues.

In the blood vasculature, maintenance of patent tubular structures is also necessary for vessel maturation, which involves the recruitment of pericytes and vascular smooth muscle cells (Adams and Alitalo, 2007; Armulik et al., 2005; Holderfield and Hughes, 2008; Hughes, 2008). These cells catalyze further remodeling events needed to form the characteristic hierarchical artery-capillary-vein vascular circuits. The process of mural cell recruitment is interesting in a number of respects. For example, some data suggest that lumen and tube formation is also accompanied by the activation of genes and molecules that actually promote tube regression events (Davis and Saunders, 2006). Thus, ECs, as they progress to form tubes, may establish the groundwork for their own subsequent regression, which could occur unless other signals prevent it. Interestingly, recruitment of pericytes and vascular smooth muscle cells can actually inhibit the intrinsic tendency of tubes composed only of ECs to regress (Armulik et al., 2005; Benjamin et al., 1998; Davis and Senger, 2008). Molecules that are supplied by pericytes to facilitate vessel stabilization include angiopoietin-1, which counteracts EC angiopoietin-2 (Thurston, 2003; Ward and Dumont, 2002), and TIMP-3, which prevents proteolysis by MMPs, ADAM, and ADAMTS proteases (Davis and Saunders, 2006; Saunders et al., 2006). ADAMs and ADAMTS have multiple functions, including shedding of growth factors, receptors, and cell adhesion molecules that affect EC function (Blobel, 2005), (Rocks et al., 2008). They have direct inhibitory effects on vascular morphogenesis through the regulation of angiogenesis inhibitors (Lee et al., 2005a; Vazquez et al., 1999). As pericytes express abundant levels of TIMP-3, an endogenous inhibitor of these proteases, pericyte recruitment along EC tubes can suppress such activities and facilitate tube stabilization (Janssen et al., 2008; Saunders et al., 2006). TIMP-2 and TIMP-3 have been reported to antagonize VEGFR2 as well (Qi et al., 2003; Seo et al., 2003; Stetler-Stevenson, 2008), meaning that they can also contribute to EC tube maintenance and stabilization by preventing VEGF-induced sprouting to initiate new tube morphogenic events.

VEGF, the master regulator of angiogenesis, has been also shown to regulate lumen size directly. Using a knockin strategy to investigate the specific effect of VEGFA isoforms, Ruhrberg and colleagues found that the soluble VEGF isoform VEGF120 generated vessels of larger diameter than VEGF188. In contrast, mice expressing only the VEGF164 isoform had capillaries of intermediate caliber, similar to those in wild-type animals (Ruhrberg et al., 2002). The distinction between the isoforms depends on their relative abilities to interact with elements of the ECM: the longer the isoform, the greater its affinity for matrix proteins (Harper and Bates, 2008). It is unclear how these differences impose molecular changes that lead to a distinct lumen diameter. It has been postulated that bound VEGF provides a gradient that elicits endothelial chemotactic responses, while soluble VEGF does not facilitate the formation of gradients and consequently does not generate directional migratory stimuli (Grunstein et al., 2000). Along these lines, while exploring the relevance of VEGF processing by MMPs, Lee and colleagues provided further support for the concept of gradients and its relation to lumen formation (Lee et al., 2005b). They found that tumors expressing a mutant form of VEGF164 that could not be cleaved by MMPs remained bound to the matrix, but it was able to activate VEGFR2 and induce a significant angiogenic response (Lee et al., 2005b). The resulting capillaries displayed extremely thin lumens and were highly branched compared with those in tumors expressing wild-type VEGF164 (Lee et al., 2005b). In contrast, tumors expressing the MMP-cleaved and highly soluble form of VEGF resulted in enlarged capillaries with poor branching patterns (Lee et al., 2005b) (Figure 2). Interestingly, MMP-cleaved VEGF had far-reaching effects, promoting hyperplasia in preexistent capillaries of adjacent normal tissues (Lee et al., 2005b). These findings indicate that soluble VEGF is highly effective at inducing endothelial proliferation, and promoting hyperplasia, tortuosity, and increased lumen diameter (Figure 2). Past and recent work has indicated that VEGF can promote fusion of endothelial tubes and thereby facilitate the expansion of lumens (Bohman et al., 2005; Drake and Little, 1995; Gentile et al., 2008; Nakatsu et al., 2003). Although the molecular mechanisms that precipitate these effects are unclear, the combined evidence seems to indicate that matrix-tethered and soluble VEGF both significantly impact lumen size of incipient and existent capillaries (Grunstein et al., 2000; Helm et al., 2005; Lee et al., 2005b; Wirzenius et al., 2007). Interestingly, lumen size in lymphatic vessels is also under the regulatory control of the VEGF family of growth factors. VEGF-E, through the stimulation of VEGFR2, mediates hyperplasia of lymphatics in the absence of lymphatic sprouts (Wirzenius et al., 2007).

Figure 2. Genetic Determinants of Vascular Lumen Size.

(A) Gain- and loss-of-function studies in mice have indicated that Notch regulates the diameter of vascular lumens. Specifically, loss-of-function analysis of Notch1 null E9.5 mouse embryos showed narrow vascular tubes. In contrast, overexpression of an activated form of Notch results in significant vascular lumen expansion. The diagram illustrates a transverse section of mouse embryos at E9.5 to indicate the relative differences between wild-type, loss of Notch 1 expression, and gain of Notch 1 expression. NT, neural tube.

(B) Genetic studies with exclusive expression of specific VEGF isoforms, as well as tumor studies, have indicated that matrix-bound and soluble VEGF mediates distinct modes of vascular expansion in a manner that affects the lumenal compartment. Thus, soluble VEGF (right panel) mediates expansion of existent vessels, resulting in enlarged, hyperplastic structures. In contrast, matrix-bound VEGF elicits rapid capillary sprouts and results in increased vascular density (middle panel) in comparison with wild-type control (left panel).

Notch is another signaling pathway shown to impact lumen diameter (Figure 2). In particular, gain- and loss-of-function studies have demonstrated complementary effects on lumen size. Inactivation of both Notch1 and Notch4 resulted in embryonic lethality by E9.5 with reduced vascular lumens (Uyttendaele et al., 2001). In contrast, a constitutively active form of Notch under the regulatory control of the VEGFR2 promoter resulted in enlarged and dilated vessels (Uyttendaele et al., 2001). More recent findings from overexpression studies using active Notch under the control of a different promoter also supported these initial conclusions (Carlson et al., 2005). Together the work suggests that, at least to some extent, lumen size appears to be under genetic control.

To maintain a tubular structure, ECs must remain adherent to the matrix tunnel surface created during morphogenesis, and maintain cell-cell contacts initially mediated by adherens junctions, which contain VE-cadherin and tight junctions. Interestingly, recent data suggest a direct relationship between VE-cadherin cell-cell contacts and TGF-beta receptor signaling (Rudini et al., 2008), a known regulator of both vascular development and tube stabilization. Furthermore, the establishment of adhesion contacts, specifically via VE-cadherin, leads to the transcriptional changes needed to increase expression of claudin-5, an EC-specific tight junction component, which further strengthens cell-to-cell contacts between ECs and thus tube stability (Taddei et al., 2008). Interestingly, this upregulation is controlled by inhibition of Foxo1 (Dejana et al., 2007), which normally suppresses claudin-5 expression (Taddei et al., 2008). Therefore, establishment of initial cell-cell contacts via cadherins provides a positive feedback loop to reinforce the stability of the tube through the upregulation of tight junction proteins.

The dynamic expression and distribution of cadherins is a central component of lumen formation by epithelial cells. Recent and unexpected contributions of the Slit-Robo signaling system have revealed its functional interaction with cadherins during the process of heart lumen formation in Drosophila (Medioni et al., 2008; Santiago-Martinez et al., 2008). Taken together, these studies demonstrate that Slit and Robo, through what appears to be an autocrine mechanism, regulate cell shape changes and coordinate subdomains of E-cadherin expression that ultimately result in the formation of a lumen. Loss of either Slit or Robo either ameliorates or completely blocks lumen formation (Medioni et al., 2008; Santiago-Martinez et al., 2008). Absence of Robo results in an increased level and distribution of E-cadherin (Medioni et al., 2008). Similarly, loss of Slit leads to an expansion in the association of beta-catenin with the adhesion complexes (Santiago-Martinez et al., 2008). By contrast, overexpression of either Robo or Slit results in ectopic lumens, most likely as a result of their mislocalization rather than their higher expression, and deregulation of the E-cadherin-β-catenin complex. The work was very important because it defined new molecular mechanisms regulating polarity and formation of lumens between juxtaposed epithelial/endothelial layers. Taking into account recent information on the evolutionary origins of both Drosophila and vertebrate hearts (Hartenstein and Mandal, 2006), it might well be interesting to investigate further whether the Slit-Robo signaling axis also contributes to lumen formation in the development of the vertebrate cardiovascular system.

Mechanisms that Regulate Lumen Diameter

Because of its multiple clinical implications, questions relating to the maintenance and stabilization of vascular lumens have largely been explored in the context of atherosclerosis, restenosis, and hypertension. However, it is likely that some of these mechanisms are also operative during development and in normal adult tissues. In fully differentiated multilayered vessels, lumen diameter is largely regulated by the contractile status of the arterial medial layer in response to pressure and flow. Physiological changes in lumen diameter can be achieved by alterations in eNOS or in NO levels with distinct consequences for blood pressure (Gregg et al., 1998; Vita et al., 2008). Pharmacologically, the same can be achieved by agents that regulate the status of smooth muscle cell contraction (Katsumi et al., 2007; Yu et al., 2005). Indeed, the endothelial layer displays a great ability to adapt to these alterations in diameter, probably through net changes in plasma membrane surface area via exocytosis and endocytosis, although this is a subject that has not been studied in great depth. In fact, many basic questions remain, such as: how does the endothelium perceive alterations in contractile status of the underlying arterial media to regulate its plasma membrane surface? And/or how are physical changes on the basal side of the endothelium transduced to achieve vesicular fusion on the luminal side (i.e., to add luminal plasma membrane) or vice-versa to remove luminal plasma membrane through vesicle internalization? These questions have yet to be addressed at a molecular level.

Sustained pathological changes in blood pressure can result in alterations of lumen size; although the tendency of the vessel is to always retain its original diameter, a concept better known as “Glagov’s phenomenon” (Glagov et al., 1987; Korshunov et al., 2007). The vascular wall of arteries is able to perceive pressure and adapt by remodeling all components of the vessel wall. Thus, increases in blood pressure result in endothelial, smooth muscle, and fibroblast hyperplasia and hypertrophy in a manner that compensates for the increase in pressure to retain the original lumen size. This remarkable ability to retain lumen diameter has been validated by a large number of atherosclerosis studies and by models of vessel remodeling in several species (Boutouyrie et al., 1999; Chironi et al., 2003; Leidenfrost et al., 2003; Lim et al., 1997). However, Glagov’s phenomenon only applies for pathological alterations that impose up to 40% vessel remodeling; more sustained or drastic changes result in reduction of lumen and chronic hypertension (Korshunov and Berk, 2003; Miyashiro et al., 1997; Ward et al., 2001).

Clearly, flow-dependent lumenal changes require the input of multiple cell types and processes. The molecular read-outs of sensing changes in pressure are global alterations in gene expression that are likely coordinated by multiple transcription factors, some of which are currently being identified (Parmar et al., 2006). In addition, it has been recognized that additional genetic factors are important modulators, as revealed by the contribution of genetic background in carotid remodeling in partial ligation models (Korshunov and Berk, 2008).

Concluding Remarks

Our understanding of vascular lumen formation is still in its infancy. Although some of the molecules and signaling pathways regulating these events appear to be similar to those in epithelial cells, it is likely that there are important features unique to the endothelial system. While development of an extensive vascular plexus can occur in the absence of flow both in vitro and in vivo, indicating that the formation of lumens is genetically programmed, lumen diameter also incorporates the input of physical forces. Flow imposes unique signaling cues, and there is a remarkable level of homeostatic regulation that is initiated during development and that requires the combined coordinated response of all layers of the blood vessel wall. The dynamic nature of EC lumen and tube formation coupled with coincident sprouting events suggest additional important differences between epithelium and endothelium. Critical distinctions are likely to exist in the nature of the junctional contacts between ECs compared with those present between epithelial cells. Although it is a teleological argument at this point, one would also anticipate differential junctional regulation along the hierarchical vascular tree within arteries, capillaries, and veins and within different tissue-specific vascular beds. Upcoming challenges include delineating the molecular sequence that triggers and completes lumen formation, identifying the processes that regulate lumen size, and understanding how the different layers of the vessel wall crosstalk to control vessel diameter.