Cdc42- and Rac1-mediated endothelial lumen formation requires Pak2, Pak4 and Par3, and PKC-dependent signaling

New vessel formation is regulated by two major processes, vasculogenesis and angiogenesis, where endothelial cells (ECs) undergo morphogenic processes including cell proliferation, migration, lumen formation, sprouting/branching and remodeling in response to various stimuli. These events are controlled by key molecules, including integrins, extracellular matrices, proteases and growth factors (Carmeliet and Jain, 2000; Davis and Bayless, 2003; Davis et al., 2002; Hanahan, 1997; Kuwano et al., 2001; Saunders et al., 2006). Lumenization of vessels is required for establishing blood flow and stabilizing vessel networks during vascular development and thus represents a crucial step during vascular EC tube morphogenesis (Adams and Alitalo, 2007; Conway et al., 2001; Davis et al., 2002; Davis and Senger, 2005).

Over the years, many studies have implicated that EC lumen and tube development involves formation of intracellular vacuoles both in vivo and in vitro (Davis et al., 2002; Davis and Camarillo, 1996; Folkman and Haudenschild, 1980; Guldner and Wolff, 1973; Meyer et al., 1997; Montesano and Orci, 1988; Montesano et al., 1987; Nicosia et al., 1982; Yang et al., 1999). Recently, our laboratory has shown that EC lumen formation in a three-dimensional (3D) extracellular matrix (ECM) is mediated by the formation and coalescence of pinocytic intracellular vacuoles (Bayless et al., 2000; Davis and Camarillo, 1996), in a process regulated by the Cdc42 and Rac1 GTPases in response to integrin and ECM interactions (Bayless and Davis, 2002; Davis and Bayless, 2003). High-resolution time-lapse analysis of vascular development in zebrafish or in vitro using human endothelial cells has shown that EC lumen development is regulated by pinocytic intracellular vacuole formation and fusion. These vacuoles can be labeled and thus can be visualized in vitro and in vivo using GFP-Cdc42 (Bayless and Davis, 2002; Kamei et al., 2006). These data together reveal that Rho GTPases represent crucial regulators of EC lumen and tube formation and strongly encourage the importance of further investigations elucidating how they control these events.

A major known function of Rho GTPases is the regulation of cytoskeletal organization; therefore, Rho GTPases influence various cellular functions required for EC morphogenesis such as cell shape, differentiation, migration, polarity and vesicular trafficking events (Etienne-Manneville and Hall, 2002; Hall, 1998; Hall, 2005; Kaibuchi et al., 1999; Nobes and Hall, 1995; Ridley, 2001). Moreover, Rho GTPases act downstream in signaling pathways stimulated by integrin-ECM interactions as well as growth factor receptors, which are relevant to EC morphogenesis (Davis and Senger, 2005; Ridley et al., 1992; Schwartz and Shattil, 2000; Shattil and Ginsberg, 1997). Regulation of biological activities by Rho GTPases is mediated by diverse downstream target proteins (Bishop and Hall, 2000; Hall, 2005). Therefore, it is necessary to identify Cdc42 and Rac1 downstream targets involved in EC lumen and tube formation in order to further understand the molecular signaling pathways controlling Cdc42- and Rac1-mediated EC vascular morphogenesis.

There has been growing evidence that phorbol esters strongly stimulate EC morphogenesis both in vivo and in vitro (Davis and Camarillo, 1996; Montesano and Orci, 1985; Morris et al., 1988). Phorbol esters are biological compounds that are known to promote tumors as well as to induce inflammation. Studies have shown that phorbol esters induce tumor promotion or inflammation in part by the stimulation of angiogenesis (Gross et al., 1983; Montesano and Orci, 1985; Moscatelli et al., 1980). A molecular mechanism underlying angiogenesis stimulation by phorbol esters is activation of protein kinase C (PKC) (Davis and Camarillo, 1996; Montesano and Orci, 1985). PKC, a well-known molecular target of phorbol esters, is a serine/threonine kinase involved in signal transduction pathways that regulate a wide range of cellular functions (Newton, 2001; Niedel et al., 1983; Nishizuka, 1995). Many laboratories have indicated the involvement of PKC in angiogenesis (Flamme et al., 1997; Podar et al., 2007; Yoshiji et al., 1999; Zhang et al., 2003). However, how EC lumen and tube formation is controlled by PKC and its isoforms is not well understood.

In vitro 3D matrix angiogenic assay models have emerged as unique tools for the analysis of EC morphogenesis as they enable systemic molecular manipulations during EC morphogenic events (Davis et al., 2002; Egginton and Gerritsen, 2003; Ilan et al., 1998; Nicosia and Villaschi, 1999; Sainson et al., 2005). Here, we utilize our 3D collagen matrix microassay systems to examine EC lumen and tube formation signaling pathways regulated by Cdc42 and Rac1. Using RNA interference (RNAi) technology, we show that Cdc42 and Rac1 are required for EC morphogenesis as well as EC invasion in 3D collagen matrices. We further analyze the function of four Cdc42 and Rac1 downstream effectors, Pak2, Pak4, Par3 and Par6, as well as PKC and associated signaling pathways in EC lumen and tube formation in 3D collagen matrices.

A previous study from our laboratory, using recombinant adenoviruses expressing mutant forms of Rho GTPases, has shown that Cdc42 and Rac1 play a key role in EC vacuole and lumen formation (Bayless and Davis, 2002). By contrast, RhoA had no effect on EC lumen and tube formation. To examine further the functional role of Rho GTPases not only in EC lumen and tube formation but also in EC invasion in 3D collagen matrices, we utilized an RNAi suppression approach. ECs were treated with siRNAs targeting the small Rho GTPases Cdc42, Rac1 or RhoA and were suspended within or on the surface of collagen matrices based on our microassay systems in 3D collagen matrices representing vasculogenesis or angiogenesis (Bayless and Davis, 2003; Davis et al., 2002; Davis and Camarillo, 1996). We examined the role of Rho GTPases during either EC tube morphogenesis or invasion. When ECs are suspended within 3D collagen matrices, they undergo EC tube morphogenesis, whereby ECs form and coalesce intracellular vacuoles and lumens, followed by assembly of ECs, leading to formation of capillary-like tubular networks. Luciferase (control) or RhoA siRNA-treated ECs were able to undergo EC morphogenesis, where they form lumens and tubular networks over a period of 48 hours (Fig. 1A, arrows). However, siRNA suppression of Cdc42 or Rac1 abolished EC morphogenesis, as no lumen or tube formation was observed (Fig. 1A). The inhibitory effect of Cdc42 siRNA on EC morphogenesis was further illustrated using time-lapse analysis (Fig. 2). Cdc42 siRNA-treated ECs not only were unable to initiate intracellular vacuoles or lumen formation but also appeared to tumble in stationary areas within 3D matrices without any directional movement. Control luciferase siRNA-treated ECs showed a normal ability to form EC lumens (Fig. 2). Cdc42 is a well-known regulator of cell polarization (Etienne-Manneville, 2004; Johnson, 1999). Therefore, this disoriented behavior we observed in the Cdc42 siRNA-treated ECs could be due to the loss of cell polarity and suggests that establishment of polarity is important for EC morphogenesis.

To further evaluate the biological effect of Rho GTPase siRNAs, siRNA-treated ECs were seeded on the surface of collagen matrices and were allowed to invade collagen matrices in response to sphingosine 1-phosphate (S1P) (Bayless and Davis, 2003) or stromal-derived factor 1α (SDF-1α) (Saunders et al., 2006). ECs treated with control luciferase or RhoA siRNA invaded into 3D collagen matrices and formed tubular networks in response to SDF-1α (Fig. 1B, arrows). By contrast, minimal to no EC invasion was observed in ECs treated with Cdc42 or Rac1 siRNAs as ECs stayed on the monolayer on top of collagen (Fig. 1B, arrowhead). EC invasion and tube formation stimulated by S1P were also significantly blocked with Cdc42 or Rac1 suppression. Quantification of the effect of siRNAs against Rho GTPases on either EC morphogenesis or EC invasion showed significant blockade of both responses with Cdc42 or Rac1 siRNA (Fig. 3A-C). Western blot analysis indicated significant reduction in protein levels and specificity for each Rho GTPase siRNA (Fig. 3D). These data show that Cdc42 and Rac1 are required for EC lumen formation and tube morphogenesis as well as EC invasion in 3D collagen matrices, whereas RhoA does not appear to play a role in either process. Importantly, the siRNA experimental approach shown here corroborates previously published work on the requirement for Cdc42 and Rac1, but not RhoA, for EC lumen formation in 3D collagen matrices (Bayless and Davis, 2002).

We then extended our analysis of the role of Cdc42 and Rac1 on EC lumen and tube formation by examining the activation of Cdc42 and Rac1 during this process. ECs were suspended in 3D collagen matrices, and extracts were made 0, 4, 16 and 24 hours later. Lysates were incubated with GST-PAK-PBD protein agarose beads and assayed for GTP-bound Cdc42 or Rac1. Pull-down assays indicate that there is a strong activation of both Cdc42 and Rac1 during EC lumen and tube formation, with a 2-2.5 fold increase in the level of activated Cdc42 or Rac1 protein over total Cdc42 or Rac1 protein (Fig. 3E,F). A pull-down assay using control blank beads did not show any capture of GTP-bound Cdc42 or Rac1, indicating the specificity of the GST-PAK-PBD protein agarose bead interaction with activated Cdc42 or Rac1 protein (Fig. 3E). Rac1 activation appears to slightly precede that of Cdc42 and peaks earlier than Cdc42 activation, which might also explain why Pak2 activation appears to occur earlier than Pak4 activation (see below). This might be indicative of differential roles for Rac1 and Cdc42 during EC lumen formation.

Rho GTPases influence diverse cellular functions through their interactions with downstream target effectors (Bishop and Hall, 2000). To identify downstream targets involved in Cdc42- and Rac1-mediated EC tube morphogenesis and invasion signaling pathways, we examined the functions of 11 known downstream effectors of Cdc42 and Rac1 using RNAi suppression technology. The eleven effectors screened were Par3, Par6, Pak2, Pak4, MRCK beta, N-WASP, IQGAP1, ACK1, CIP4, SPEC1 and SPEC2. Using this approach, we assessed the biological effects of a large number of effectors simultaneously. ECs transfected with siRNAs targeting Cdc42 or Rac1 downstream effectors were then allowed to undergo EC morphogenesis or invasion in 3D collagen matrices. A quantification of their effects on EC morphogenesis and invasion is shown in Fig. 4. Our screening reveals that certain effectors such as Par6, Pak2, Pak4, MRCK beta and SPEC1 appeared to play a role in both EC tube morphogenesis and invasion, whereas other effectors appeared to be more selective. RNAi-mediated suppression of Par3, ACK1 and CIP4 inhibited EC lumen formation in vasculogenesis-like assays. N-WASP and SPEC2 siRNAs selectively blocked S1P-induced invasion and tube formation, whereas IQGAP1 siRNA suppression inhibited invasion and tube formation in response to SDF-1α only. Our screening results illustrate the complexities underlying two different molecular mechanisms of blood vessel formation, vasculogenesis and angiogenesis, whose processes appear to be regulated by distinct signaling pathways. They further suggest that EC angiogenic responses stimulated by either S1P or SDF-1α might also be mediated through different signaling molecules. Although each effector identified in our screening might have a distinct or important function in signaling pathways regulating EC morphogenesis and invasion, in this study, we focus our attention on the roles of four Cdc42 and Rac1 effectors as well as the regulatory role of protein kinase C isoforms. The four effectors are Pak2, Pak4, Par3 and Par6, which are known to regulate a number of signaling pathways affecting crucial cell behaviors such as motility and morphogenesis (Bokoch, 2003; Etienne-Manneville and Hall, 2003b; Fryer and Field, 2005).

EC cytoskeletal function is regulated by Rho GTPases, and these molecules play a crucial role in EC tube morphogenesis (Bayless and Davis, 2002; Hall, 2005; Hoang et al., 2004; Nobes and Hall, 1995). Here, we investigate the functional importance of Pak2 and Pak4 in this process as p21-activated kinases are major Cdc42 and Rac1 downstream effectors in cytoskeletal regulation (Bokoch, 2003; Daniels and Bokoch, 1999; Kiosses et al., 1999; Lim et al., 1996). When ECs were suspended within 3D collagen matrices for 48 hours, RNAi-mediated suppression of Pak2 or Pak4 significantly blocked EC lumen and tube formation compared with the effect of luciferase control siRNA (Fig. 4A and Fig. 5A). They also had a strong inhibitory effect on EC invasion responses, whereas control luciferase siRNA-treated ECs invaded and sprouted into collagen matrices (Fig. 4B,C and Fig. 5B). The siRNA treatments in ECs specifically reduced the protein expression of Pak2 or Pak4 (Fig. 5C).

To further establish that Pak2 and Pak4 are required for EC lumen and tube formation, we generated recombinant adenoviruses expressing dominant-negative (DN) Pak2 and Pak4. In both Pak2 and Pak4, mutation of lysine residues in the kinase domain disrupts ATP binding, thereby making mutants Pak2 K278R and Pak4 K350M kinase inactive (Abo et al., 1998; Huang et al., 2003). The mutant Pak2 T402A also has little kinase activity as full activation of Pak2 requires phosphorylation of a threonine residue located in the activation loop (Chong et al., 2001; Renkema et al., 2002). We used western blot analysis to confirm that there is a selective induction of Pak2 and Pak4 proteins by Pak2 and Pak4 mutant viruses (data not shown). ECs were then infected with control GFP virus or mutant forms of DN Pak2 K278R, DN Pak2 T402A or DN Pak4 K350M virus. After 24 hrs of infection, ECs were suspended in 3D collagen matrices and were allowed to undergo EC tube morphogenesis for 24 hrs. As shown in Fig. 6, expression of DN Pak2 and DN Pak4 viruses markedly blocked EC lumen formation compared with a GFP control, suggesting that EC lumen and tube formation in 3D collagen matrices requires Pak2 and Pak4.

As reduced activity of Pak2 or Pak4 significantly blocks EC lumen and tube formation in 3D collagen matrices, we examined the activation levels of these proteins downstream of Cdc42 or Rac1 during this process. We measured EC lumenal area over time during EC morphogenesis (quantitatively assessed by tracing EC lumenal areas using the Metamorph software program) to analyze systemically EC lumen and tube formation events. As shown in Fig. 7A, the majority of EC lumenization and morphogenesis occurs within 24 hours of ECs being suspended in 3D collagen matrices. When Pak2 and Pak4 are activated downstream of Cdc42 or Rac1, they undergo phosphorylation or autophosphorylation at various sites that increase their kinase activity (Bokoch, 2003; Gatti et al., 1999; Lim et al., 1996). Therefore, we used antibodies that recognize phosphorylation at one of these sites (S141 for Pak2 and S474 for Pak4) to examine activation of Pak2 or Pak4 during a time-course of EC lumen and tube formation. Both Pak2 and Pak4 phosphorylation were induced over a period of 24 hours during EC morphogenesis (Fig. 7B,C). Furthermore, activation of Pak2 and Pak4 directly correlates with the timing and progression of EC lumen and tube formation in 3D collagen matrices.

We next asked whether there is a direct association between Cdc42 and its effectors during EC lumen and tube formation in 3D collagen matrices. To examine interactions between Cdc42 and its downstream effectors during EC morphogenesis, we generated a recombinant virus of Cdc42 tagged with both GFP and an S-tag, S-GFP-Cdc42. ECs were infected with S-GFP-Cdc42 or control GFP-Cdc42 virus 24 hours before they were suspended within 3D collagen matrices. Treatment of ECs with S-GFP-Cdc42 or control GFP-Cdc42 adenovirus allowed the ECs to form vacuoles and lumens in an equivalent manner. EC culture extracts were prepared at 4, 8, 16 and 24 hours into EC morphogenesis, and lysates were incubated with S-protein agarose beads to capture the recombinant Cdc42 protein. We examined binding partners of Cdc42 during EC morphogenesis by western blot analysis. Identified effectors include phosphorylated Pak2 and Pak4 as well as Par3, whereas no binding to the S-protein column was detected from control GFP-Cdc42 cultures, indicating the specificity of these interactions (Fig. 8). Their binding to Cdc42 increased over time, suggesting that associations between Cdc42 and its effectors are strongly induced during EC lumen and tube formation. The associations of Cdc42 with its effectors were strongly induced when ECs were allowed to undergo EC morphogenesis in 3D collagen matrices, with a minimal interaction occurring when ECs were plated on 2D plastic dishes (data not shown). These data suggest that suspension of ECs in 3D collagen matrices activates Cdc42, which thereby triggers activation or recruitment and binding of different downstream effectors to Cdc42, including Par3, to promote EC lumen and tube formation.

With suppression of Cdc42, ECs appeared to be disoriented in 3D collagen matrices and were not able to undergo EC morphogenesis. This suggests that polarity might play an important role in EC morphogenesis. Moreover, we have observed that EC vacuoles accumulate and fuse in a perinuclear location directly adjacent to the microtubule organizing center when ECs undergoing morphogenesis were immunostained with an antibody against gamma tubulin (Davis et al., 2007).

Furthermore, Cdc42 is known to regulate diverse cell polarization events by forming a quaternary protein complex with its two effectors, Par3 and Par6, and atypical protein kinase C (aPKC) (Etienne-Manneville, 2004; Etienne-Manneville and Hall, 2003a; Etienne-Manneville and Hall, 2003b; Lin et al., 2000; Noda et al., 2001). By targeting of Par6 or aPKC to the apical plasma membrane, Cdc42 regulates lumen formation in epithelial morphogenesis (Martin-Belmonte et al., 2007). However, the role of Cdc42-mediated polarity is not well understood in the context of EC lumen formation. Our screening of effectors showed that suppression of Par3 or Par6 inhibited EC lumen and tube formation (Fig. 4A). Par6 in mammalian cells is known to have diverse isoforms with distinct cellular functions and localization (Gao and Macara, 2004). Therefore, we examined the function of two different Par6 isoforms, Par6a and Par6b, in EC lumen and tube formation. Analysis of the number of EC lumens formed showed that siRNA against Par6b was more effective in blocking EC lumen formation than siRNA against Par6a (Fig. 4A). The latter, by contrast, significantly inhibited EC invasion (Fig. 4B,C), suggesting that different Par6 isoforms might regulate distinct steps of the morphogenic process. To further distinguish the different roles of Par6 isoforms in EC tube and lumen formation, the areas of EC lumens were measured using Metamorph software to study the progression of EC lumen development, in addition to EC lumen formation. Par6b siRNA blocked both EC lumen formation and progression, suggesting that Par6b is a dominant isoform involved in EC morphogenesis in 3D collagen matrices (Fig. 4A and Fig. 9A,B). By contrast, siRNA suppression of Par6a revealed only a modest blocking effect in both EC lumen formation and development (Fig. 4A and Fig. 9A,B). Par3 siRNA was strongly effective in blocking both the number of EC lumens formed and the lumen development process (Fig. 4A and Fig. 9A,B). The specificity of each Par siRNA and its ability to reduce the protein level was confirmed by western blot analysis or semi-quantitative reverse transcriptase PCR (Fig. 9C,D). These data suggest that regulation of cell polarity is crucial for EC morphogenesis, whereby disruption of the Par3 and Par6 polarity complex inhibits EC lumen and tube formation in 3D collagen matrices.

Studies have shown that phorbol esters induce tube formation and angiogenesis in vivo and in vitro (Davis and Camarillo, 1996; Montesano and Orci, 1985; Morris et al., 1988; Taylor et al., 2006). In our assay system, the phorbol ester 12-O-tetradecanoyl-phorbol-13-acetate (TPA) markedly stimulates EC tubular morphogenesis and invasion in 3D collagen matrices. In the absence of TPA, EC lumen formation and morphogenesis are significantly reduced compared with those in the presence of TPA (Fig. 10A,B). With TPA, ECs develop lumen and tube structures through the formation and coalescence of vacuoles in 3D collagen matrices (Fig. 10C, arrow). In the absence of TPA, however, the progression of lumen and tube development is much less well developed, where ECs form cord-like structures with few lumenal compartments (Fig. 10D, arrows).

Phorbol ester is well known to activate PKC strongly (Castagna et al., 1982; Newton, 2001; Pears and Parker, 1991). Therefore, we further investigated TPA-stimulated EC lumen and tube formation in 3D collagen matrices through activation of PKC. The PKC family can be divided into three broad groups: novel, conventional and atypical. Activation of novel and conventional PKC isoforms is dependent on lipid or phorbol ester, whereas that of atypical isoforms is not (Deacon et al., 1997; Newton, 2001). To identify the relevant PKC isoform group involved in our signaling pathway, we added different PKC chemical inhibitors to our EC vasculogenesis assay. As shown in Fig. 11, addition of a PKC inhibitor that targets all isoforms (GF109203X) or PKC inhibitors that target novel or atypical groups (Go6983 and Ro-32-0432) significantly blocked EC lumen and tube formation, whereas addition of a PKC inhibitor specific to conventional group isoforms (Go6976) did not have any effect on EC morphogenesis. These data suggest that TPA is likely to induce EC lumen and tube formation by activating novel PKC isoforms. Atypical PKC isoforms, owing to their known association with the Cdc42-Par6-Par3 polarity complex, might also play a role in EC lumen and tube formation.

To dissect further EC morphogenesis signaling pathways mediated by PKC, we examined molecular targets downstream of PKC. TPA strongly induces EC lumen and tube formation, and we have shown that Pak2 and Pak4 are highly activated during EC morphogenesis. Therefore, we investigated whether TPA promotes EC lumen and tube formation by activating Pak2 and Pak4. Phosphorylation of Pak2 and Pak4 was highly induced in the presence of TPA in 3D EC cultures, suggesting that TPA induces EC morphogenesis in part by activating Pak2 and Pak4 (Fig. 12A). Activation of Pak2 and Pak4 also corresponded to PKC activation in EC morphogenesis. Phosphorylation of Pak2 and Pak4 was markedly diminished when EC morphogenesis was blocked by the addition of the PKC inhibitors GF109203X, Go6983 or Ro-32-0432. Addition of the selective PKCα and β inhibitor Go6976 did not block morphogenesis or Pak phosphorylation (Fig. 12B). These data suggest that TPA activates PKC, which in turn regulates Pak2 and Pak4 phosphorylation, promoting EC lumen and tube formation in 3D collagen matrices.

Our experiment using PKC chemical inhibitors suggests that TPA most likely activates novel PKC isoforms to stimulate EC lumen and tube formation in 3D collagen matrices and that atypical PKC isoforms, even though they are not activated by TPA, might also be involved in EC lumen and tube formation. To identify more precisely the relevant PKC isoforms in this pathway, we utilized an RNAi-based approach in which we suppressed the conventional isoform PKCα, novel isoforms PKCδ and PKCϵ and atypical isoform PKCζ. PKCα, PKCδ and PKCϵ have been shown to be involved in angiogenesis, EC permeability, tumor metastasis and inflammation, whereas PKCζ is a key member of the Cdc42-mediated polarity complex (Cheng et al., 2001; Griner and Kazanietz, 2007; Macara, 2004b; Orr et al., 2006; Saha et al., 2003; Taylor et al., 2006). RNAi-mediated suppression of PKCα did not have any effect on EC lumen and tube formation, as assessed by the number and area of EC lumens compared with the luciferase control (Fig. 13A-C). Western blot analysis showed the specificity and effectiveness of each PKC siRNA treatment (Fig. 13D). PKCδ siRNA also did not exhibit any inhibitory effect on EC lumen and tube formation, whereas suppression of PKCϵ significantly blocked EC lumen and tube formation, suggesting that TPA activates PKCϵ to induce EC lumen and tube formation in 3D collagen matrices (Fig. 13A-C). Suppression of atypical PKCζ, whose activity is independent of TPA but important for cell polarity, also had a significant inhibitory effect on EC lumen and tube formation. These latter data, coupled with the results shown in Fig. 9, which revealed a role for both Par3 and Par6b in lumen formation, suggest the involvement of Cdc42–Par3–Par6–atypical-PKC polarity complex signaling in EC lumen and tube formation. Importantly, the RNAi suppression data identifying a role for both PKCϵ and PKCζ in EC lumen formation are also consistent with the marked blocking ability of the PKC inhibitors GF109203X, Go6983 and Ro-32-0432 (Fig. 11), which are known to inhibit these PKC isoforms. Further support for this conclusion is that the Go6976 PKC inhibitor, which has little ability to block these isoforms, failed to block EC lumen formation.

We have found that Cdc42 and Rac1 play a crucial role in EC lumen and tube formation as well as EC invasion in 3D collagen matrices using RNAi-mediated suppression technology. We further identified some downstream molecular targets and examined their role during EC morphogenesis and invasion. In this study, we investigated four Cdc42 or Rac1 effectors, Pak2, Pak4, Par3 and Par6, as well as the regulatory role of protein kinase C isoforms during EC morphogenic events. Previous studies have revealed an important role for Rho GTPases in EC morphogenesis (Bayless and Davis, 2002; Davis and Senger, 2005; Hoang et al., 2004) as well as important functions such as the control of vascular permeability (Stockton et al., 2004). Furthermore, Rho GTPases such as Cdc42 and Rac1 are known to play a crucial role in epithelial lumen formation (Martin-Belmonte et al., 2007; Pirraglia et al., 2006), similar to that observed for EC lumen formation (Bayless and Davis, 2002; Davis and Bayless, 2003). Our previous studies utilized dominant-negative mutants as well as GFP-Rho GTPase fusion proteins to analyze EC lumen formation (Bayless and Davis, 2002).

A detailed understanding of how Rho GTPases control EC morphogenesis and invasion is important as they represent major downstream effectors of integrin-ECM signaling pathways and are regulators of the actin and microtubule cytoskeletons, which are essential for EC morphogenesis (Davis et al., 2002; Davis and Senger, 2005; Hall, 1998; Schwartz and Shattil, 2000). We used an RNAi-mediated suppression approach and confirmed that Cdc42 and Rac1 are required for EC lumen and tube formation as well as for EC invasion in 3D collagen matrices, whereas RhoA appears to play a minimal role in both processes. Previous studies have suggested that RhoA plays a role in EC tube collapse and regression (Bayless and Davis, 2004) and initial EC cord assembly steps during morphogenesis (Hoang et al., 2004). We extended our analysis to study Cdc42- and Rac1-mediated EC lumen and tube formation in more detail by identifying relevant downstream molecular targets, including the four key effectors investigated in this paper – Pak2, Pak4, Par3 and Par6.

The p21-activated kinase proteins, which are serine/threonine kinases, were the first Cdc42 and Rac1 effectors identified. Based on their structure, they can be divided into two groups (Bokoch, 2003; Jaffer and Chernoff, 2002). Group I includes Pak1, Pak2 and Pak3, whose members have been investigated in many studies. Previous work has implicated Pak1 signaling in various EC behaviors such as migration, permeability and morphogenesis (Fryer and Field, 2005; Kiosses et al., 1999; Kiosses et al., 2002; Stockton et al., 2004). Recently, a study has shown that Pak2 regulates cerebrovascular stabilization and integrity during vascular development in zebrafish (Buchner et al., 2007; Liu et al., 2007). Later studies identified group II members, including Pak4, Pak5 and Pak6. Pak4 was the first member of this group to be identified and has been shown to be involved in microtubule and actin cytoskeleton regulation and to be selectively activated by Cdc42 (Callow et al., 2005; Dan et al., 2001). Pak4 is also known to regulate morphological development of the heart and neural tubes as well as cell differentiation (Qu et al., 2003). However, no direct role for Pak2 or Pak4 in EC lumen and tube formation has been presented previously. Here, disruption of Pak2 and Pak4 either by RNAi-mediated suppression or by expression of DN mutants markedly impaired EC lumen and tube formation as well as EC invasion of 3D collagen matrices. Moreover, Pak2 and Pak4 were highly activated, as indicated by phosphorylation mediated downstream of Cdc42 during EC lumen and tube formation. The ability of Cdc42 to associate with its activated effectors was observed when ECs were suspended within 3D collagen matrices (Fig. 8), suggesting that EC interactions with the ECM play a key role in regulating signaling cascades during EC morphogenesis. ECM signaling is crucial during vascular development and angiogenic responses (Davis and Senger, 2005).

The Par proteins Par3 and Par6 are Cdc42 and Rac1 downstream effectors that are involved in the cell polarity signaling pathway (Etienne-Manneville, 2004; Etienne-Manneville and Hall, 2003b; Macara, 2004b). Our data show that disruption of any member of the Cdc42-Par3-Par6b-aPKC polarity complex impaired EC lumen and tube formation in 3D collagen matrices. This suggests that establishment of polarity is important for EC morphogenesis. We have observed that formation of EC vacuoles and lumens occurs in a perinuclear and polarized fashion within ECs (Davis et al., 2007). Therefore, it is likely that Cdc42-mediated cell polarity controls the position of intracellular vacuoles and lumens within ECs. It has been found that, during epithelial morphogenesis, the Cdc42-mediated polarity complex also plays a key role in tight junction formation and that cell polarity and tight junction formation are regulated coordinately (Joberty et al., 2000; Martin-Belmonte et al., 2007; Nelson, 2003; Schnittler, 1998). Interestingly, Cdc42 has been shown to influence Rac1 activation through a Par3- and Tiam1-dependent process (Chen and Macara, 2005). Our work showing the crucial role of Cdc42, Rac1 and Par3 during EC lumen formation suggests that this novel signaling pathway could be operative during EC lumen formation as well. Therefore, a more detailed analysis is necessary to identify the signaling molecules involved in cell polarity control and their subsequent functions in EC lumen and tube formation.

Pak2 and Pak4 were also found to act downstream of PKC to stimulate EC lumen and tube formation. Phorbol esters are well known to stimulate angiogenesis in vivo and in vitro, possibly by activating PKC (Davis and Camarillo, 1996; Montesano and Orci, 1985; Morris et al., 1988). The signaling events that regulate their influence have remained unclear. Using either chemical PKC inhibitors or siRNAs against PKCα, PKCδ, PKCϵ and PKCζ, we found that TPA-induced EC lumen formation appears to involve the novel PKC isoform PKCϵ in 3D collagen matrices. Conventional PKCα or novel PKCδ did not appear to play any role in EC lumen and tube formation in 3D collagen matrices. RNAi-mediated suppression of PKCζ also led to inhibition of EC lumen and tube formation. As an atypical PKC, PKCζ is known to be insensitive to phorbol esters. However, PKCζ is a key member of the previously characterized Cdc42-mediated cell polarity complex signaling pathway (Etienne-Manneville and Hall, 2003b; Macara, 2004a), and its inhibitory effect on EC lumen and tube formation probably arises through disruption of this EC polarity. This concept is supported by the clear involvement of two additional polarity proteins, Par3 and Par6b, in this process (Fig. 8 and Fig. 9).

Our study shows two distinct, but also possibly correlated, signaling pathways that regulate EC lumen and tube formation in 3D collagen matrices. As shown in Fig. 14, one signaling pathway is mediated by Cdc42 and Rac1, and subsequent activations of downstream effectors including Pak2, Pak4, Par3 and Par6. The other signaling pathway regulating EC lumen and tube formation is mediated by PKC isoforms. We show that TPA activates PKC isoforms, including PKCϵ, to induce EC lumen and tube formation, followed by activation of Pak2 and Pak4. In both cases, Pak2 and Pak4 serve as common downstream targets. These raise the possibility that these two pathways might be interrelated. For example, it has been shown that PKC can activate Rho GTPases by phosphorylating Rho GDP dissociation inhibitors (RhoGDIs), which then dissociate from GDP-Rho GTPases (Mehta et al., 2001; Pan et al., 2005). Therefore, we speculate that TPA activates PKC leading to Cdc42 or Rac1 activation to regulate EC lumen and tube formation. The complexity underlying EC vascular morphogenesis calls for further molecular examination of the interactions between previously identified molecules and for continued investigation of additional signaling molecules. The data presented in this paper disclose several key molecules controlling EC lumen and tube formation and their interactions in 3D collagen matrices and provide us with new molecular details underlying this important morphogenic event.

GF109203X, Go6983, Ro-32-0432 and Go6976 were purchased from Calbiochem (La Jolla, CA). TPA and a polyclonal antibody against phospho-Pak2 (Ser141) were obtained from Sigma-Aldrich (St Louis, MO). Polyclonal antibodies against Pak2, Pak4, phospho-Pak4 (Ser474) and PKCζ were obtained from Cell Signaling Technology (Danvers, MA). Monoclonal antibodies specific for Cdc42, PKCα, PKCδ and PKCϵ were purchased from BD Biosciences (San Jose, CA). A monoclonal antibody against actin (CP01) was obtained from Calbiochem. A polyclonal antibody against Par3 was obtained from Upstate (Charlottesville, VA). A monoclonal antibody specific for RhoA and a polyclonal antibody specific for Rac1 were purchased from Cytoskeleton (Denver, CO).

Human umbilical vein ECs (HUVECs) were purchased from Clonetics (San Diego, CA) and were cultured (passage 2-5) as described previously (Davis and Camarillo, 1996). For vasculogenesis assays, ECs were suspended within 3.75 mg/ml or 2.5 mg/ml of collagen type I matrices and allowed to undergo EC morphogenesis as described previously (Davis and Camarillo, 1996). As for invasion assays, ECs were seeded on top of 3.75 mg/ml collagen type I matrices containing 1 μM S1P or 200 ng/ml SDF-1α, as described previously (Bayless and Davis, 2003; Saunders et al., 2006). Cultures were fixed at the indicated time-points with 3% glutaraldehyde for 30 minutes. In some cases, cultures were stained with 0.1% Toluidine Blue in 30% methanol and destained before photography and visualization. Some 3D collagen gels were also extracted to examine protein expression. Extracts were run on SDS-PAGE gels, transferred to polyvinylidene membranes, probed and developed.

Adenovirus infection of ECs was carried out as described previously (Bayless and Davis, 2002). Infected ECs were suspended within 3.75 mg/ml collagen type I matrices for 24 hours. Cultures were fixed with 2% paraformaldehyde for 30 minutes before photography and visualization.

siGENOME human Cdc42 duplex siRNA (D-005057-02) and SMARTpool human Rac1, RhoA, Par3, Par6, Pak2, Pak4, MRCK beta, N-WASP, IQGAP1, ACK1, CIP4, SPEC1, SPEC2, PKCα, PKCδ, PKCϵ and PKCζ siRNAs were obtained from Dharmacon (Lafayette, CO) and prepared as described previously (Saunders et al., 2005). Luciferase GL2 duplex was used as a control. EC transfection with siRNAs was carried out in growth media with 1% serum. Details of our siRNA transfection protocol have been described previously (Saunders et al., 2005).

Pak2 and Pak4 were amplified from human skeletal muscle cDNA (Clonetech) and human cDNA clone (Origene), respectively using the primers: Pak2 up (5′-AGCTCGAGGCCACCATGTCTGATAACGGAGAACTGGAAG-3′), Pak2 dn (5′-AGTCTAGATTAACGGTTACTCTTCATTGCTTC-3′), Pak4 up (5′-AGCTCGAGGCCACCATGTTTGGGAAGAGGAAGAAG-3′) and Pak4 dn (5′-AGTCTAGATCATCTGGTGCGGTTCTGGCG-3′) (Sigma Genosys, The Woodlands, TX). Standard restriction digestion cloning was performed to clone Pak2 and Pak4 into pAdTrack-CMV with XhoI and XbaI enzymes (Invitrogen). Positive clones were confirmed by restriction digest, sequence analysis and western blot analysis using transfected HEK293 cells. Pak2K278R, Pak2T402A and Pak4K350M were then generated by site-directed mutagenesis according to the manufacturer's protocol (Stratagene). Mutated clones were verified by restriction digest, sequence analysis and western blot analysis.

Generation of a GFP-Cdc42 fusion construct was achieved as described previously (Bayless and Davis, 2002). Standard restriction digestion cloning was performed to clone a GFP-Cdc42 fusion construct into the vector pTriEx-2neo (Novagen) using primers 5′ BamHI AGGGATCCGATGGTGAGCAAGGGCGAGGAG and 3′ BamHI AGGGATCCTTAGAATATACAGCACTTCC. Positive clones were confirmed by restriction digest and PCR. The S-GFP-Cdc42 construct was then cloned into pShuttle-CMV using primers 5′ KpnI AGGGTACCGCCACCATGAAAGAAACCGCTGCTGCG and 3′ XbaI AGTCTAGATTAGAATATACAGCACTTCCTTTTGGG. Positive clones were confirmed by restriction digest, sequence analysis and western blot analysis using transfected HEK293 cells.

Recombination and virus production were carried out as described previously (He et al., 1998). Our adenoviruses propagation protocol has been described previously (Bayless and Davis, 2002).

An EC vasculogenesis assay was set up as described above and EC cultures were extracted at the indicated time-points using cold detergent lysis buffer of 1% Triton X-100 in 1× TBS, pH 8.0, containing complete EDTA-free protease inhibitor cocktail tablets (Roche Diagnostics), 150 μg/μl high-purity collagenase (Sigma Aldrich) and 100 nM GTPγS (Calbiochem). Lysates were incubated at 4°C for 30 minutes to dissolve collagen and were clarified by centrifugation at 16,000 g for 20 minutes at 4°C. Supernatants were pre-cleared by incubation with Glycine sepharose 4B beads (Sigma-Aldrich) for 2 hours at 4°C. After removal of Glycine sepharose 4B beads by brief centrifugation, the supernatants were incubated with S-protein agarose beads (Novagen) for 45 minutes at 4°C. The beads were washed four times with washing buffer (1:10 dilution of lysis buffer in 1× TBS, pH 8.0). Bound Cdc42-associated proteins were detected by western blot analysis.

An EC vasculogenesis assay was set up as described above and EC cultures were extracted at the indicated time-points using cold detergent lysis buffer of 1% Triton X-100 in 1× TBS, pH 8.0, containing complete EDTA-free protease inhibitor cocktail tablets (Roche Diagnostics), 150 μg/μl high-purity collagenase (Sigma Aldrich),and 100 nM GTPγS (Calbiochem). Lysates were incubated at 4°C for 30 minutes to dissolve collagen and were clarified by centrifugation at 16,000 g for 20 minutes at 4°C. Supernatants were incubated with GST-PAK-PBD protein agarose beads (Cytoskeleton) for 45 minutes at 4°C. The beads were washed four times with washing buffer. Bound active Cdc42 proteins were detected by western blot analysis.

Total RNA was extracted from siRNA-treated (Luciferase, Par6 and Par6b) ECs using the Totally RNA Isolation kit obtained from Ambion (Austin, TX) according to the manufacturer's instructions. RNA (1 μg) was reverse-transcribed using an AccuScript High Fidelity 1^st strand cDNA synthesis kit (Stratagene). RT-PCR amplification was performed using the primers: Par6a up (5′-ATGGCCCGGCCGCAGAGGACTC-3′), Par6a dn (5′-TAGGTCCACGTCTATGACTGAGG-3′), Par6b up (5′-TCTACTGATAACAGCCTTCTTGGC-3′), Par6b dn (5′-TCATAATGTTATGATTGTTCCATC-3′), G3PDH-1 up (5′-GCCAAAAGGGTCATCATCTC-3′) and G3PDH-1 dn (5′-GTAGAGGCAGGGATGATGTTC-3′).

Visualization and image acquisition of EC vasculogenesis and invasion assays were performed using an inverted microscope (CKX41; Olympus), as described previously (Saunders et al., 2006). Real time-lapse imaging of living cells was done using a Nikon TE2000-E system with a temperature-controlled chamber set to 37°C with continuous flow of 5% CO₂. Image analysis was performed using MetaMorph software.

Statistical analysis of EC vasculogenesis and invasion was performed using SPSS 11.0 software. Statistical significances were accessed by paired-samples t test.

We thank Ronald Korthuis for helpful discussions. Our work was supported by NIH grants HL59373 and HL79460 to G.E.D.