Temporal and Spatial Modulation of Rho GTPases during in Vitro Formation of Capillary Vascular Network

Endothelial cells (ECs) self-organize into capillary networks when plated on extracellular matrix. In this process, Rho GTPases-mediated cytoskeletal dynamics control cell movement and organization of cell-to-matrix and cell-to-cell contacts. Time course analysis of RhoA and Rac1 activation matches specific morphological aspects of nascent pattern. RhoA-GTP increases early during EC adhesion and accumulates at sites of membrane ruffling. Rac1 is activated later and localizes in lamellipodia and at cell-to-cell contacts of organized cell chains. When ECs stretch and remodel to form capillary structures, RhoA-GTP increases again and associates with stress fibers running along the major cell axis. N17Rac1 and N19RhoA mutants impair pattern formation. Cell-to-cell contacts and myosin light chains (MLC) are targets of Rac1 and RhoA, respectively. N17Rac1 reduces the shift of β-catenin and vascular endothelial cadherin to Triton X-100-insoluble fraction and impairs β-catenin distribution at adherens junctions, suggesting that Rac1 controls the dynamics of cadherin-catenin complex with F-actin. During the remodeling phase of network formation, ECs show an intense staining for phosphorylated MLC along the plasma membrane; in contrast, MLC is less phosphorylated and widely diffused in N19RhoA ECs. Both N17Rac1 and N19RhoA have been used to investigate the role of wild type molecules in the main steps characterizing in vitro angiogenesis: (i) cell adhesion to the substrate, (ii) cell movement, and (iii) mechanical remodeling of matrix. N17Rac1 has a striking inhibitory effect on haptotaxis, whereas N19RhoA slightly inhibits EC adhesion and motility but more markedly Matrigel contraction. We conclude that different Rho GTPases control distinct morphogenetic aspects of vascular morphogenesis.

To distribute nutrients throughout the body, vertebrates have evolved a branching blood vascular system that terminates in a network of size-invariant units, namely capillaries. The development of capillary networks characterized by typical intercapillary distances ranging from 50 to 300 m is instrumental for optimal metabolic exchange (1). In embryo, homogeneous vascular networks form by vasculogenesis and are remodeled through angiogenesis (2). Both processes require that ECs 1 or their mesodermic precursors, angioblasts, move, proliferate, and then change their shape folding into capillaries. The ability to form networking capillary tubes is a cell autonomous property of ECs and requires permissive extracellular cues such as growth factors, autacoids, signals coming from the extracellular matrix, and physical forces (2)(3)(4).
It is well known that culturing ECs on a tridimensional scaffold of extracellular matrix protein markedly accelerates their morphological differentiation in geometric tubular networks (5)(6)(7)(8), which are almost identical to vascular beds formed by vasculogenesis in vivo (9) or to some anatomical districts by angiogenesis (10 -12). When this process occurs on Matrigel, a natural basal membrane matrix, it is rapid (6 -16 h), does not require cell proliferation, and is mainly characterized by cell motility and shape rearrangement, which imply dramatic changes in EC cytoskeletal dynamics and adhesive behavior (7,13).
In the last few years it has been widely demonstrated that Rho GTPases are major regulators of cell polarization and motility. In fibroblasts, Cdc42 mediates filopodia extension and Rac1 both lamellipodia formation and membrane ruffling, and RhoA controls stress fiber formation (14). Both Rac1 and RhoA also participate in the ruffling phenomena of ECs (15). However, GTPase activities exert antagonistic effects in the regulation of different steps of cell motility (14,16) and show discrete subcellular localizations (17)(18)(19). Recent evidence (23)(24)(25)(26) suggests that Rho GTPases are part of the signaling pathways triggered by angiogenic inducers (20 -22) and regulate the shape of ECs.
We analyzed the temporal and spatial activation of Rac1 and RhoA during capillary vascular network formation by ECs and show that their spatially and temporally regulated activation was required for proper formation of such a network. By employing dominant negative molecules, we found that Rac1 and RhoA, respectively, regulate the early formation of adherens junction, among ECs that at the beginning self-organize into clusters and chains, and the phosphorylation level of myosin light chain (MCL) when ECs start stretching to remodel the vascular network.
Cell Culture-Human dermal capillary ECs (Biotech, BD Biosciences) were cultured according to the manufacturer's instructions. EC infection was performed as described previously (38). Briefly, N17Rac1, N17Cdc42, N19RhoA, RhoAQL, and Rac1QL cDNAs subcloned into Pinco vector were transiently transfected into Phoenix packaging cells, and the supernatants were used to infect ECs. As control of infection, GFP analysis was performed both through fluorescence microscope and fluorescence-activated cell sorting (FACS Advantage SE, Biotech, BD Biosciences, San Jose, CA) giving 85-90% of positive ECs. The expression of transgene molecules was evaluated by Western blot analysis and always gave an increase of proteins recognized by specific mAbs anti-Rac1, anti-RhoA, and Ab anti-Cdc42 or as compared with ECs infected with vector alone (not shown). Infection conditions per se were without effect on EC morphology and on cell cycle analyzed by propidium iodide fluorescence by FACS flow cytometer.
In Vitro Angiogenesis-To analyze in vitro angiogenesis, glass coverslips (2 cm 2 growth area) were coated with 0.25 ml of Matrigel (8.8 mg/ml; Biotech, BD Biosciences), which was immediately aspirated, and the remaining film was allowed to solidify. ECs (125 cells/mm 2 ) were incubated at 37°C in a 5% CO 2 humidified atmosphere for 14 h on glass coverslips and observed with an inverted photomicroscope (model DM IRB HC; Leica Microsystems, Solms, Germany). Phase contrast snap photographs and 12-14-h long movies (one frame every 5 min) were taken with a cooled digital CCD Hamamatsu ORCA camera (Hamamatsu Photonics Italy, Milano, Italy), recorded, and analyzed with ImageProPlus 4.0 imaging software to count the number of capillaries/mm 2 (Media Cybernetics, Carlsbad, CA) (39).
In the sprouting assay capillary ECs carrying vector alone or N17Rac1 or N19RhoA were suspended at a density of 4 cells/l in culture medium containing 20% of Methocel stock (6 g of carboxymethylcellulose in 500 ml of M199) in M199 containing 20% FCS, 0.1 mg/ml porcine heparin. Eight hundred cells were seeded into non-adherent round-bottomed 96-well plates (Falcon, BD Biosciences), and cultured overnight at 37°C (5% CO 2 , 100% humidity). The spheroids were harvested by gently pipetting them and were centrifuged at 300 ϫ g for 15 min. The EC spheroids were suspended in 200 l of M199 medium containing 40% FCS, 1.2% (v/w) methylcellulose with or without 30 ng/ml vascular endothelial growth factor (VEGF)-A 165 (R&D Systems, Minneapolis, MN), and mixed with an equal volume of diluted collagen solution (7 volumes of collagen from rat tail, 1 volume of 10ϫ M199, 1 volume of 0.1 N NaOH, and 1 volume of 0.2 M Hepes, pH 7.3) (39). After 24 h spheroid photographs were taken with digital CCD Hamamatsu ORCA camera linked to an inverted photomicroscope (model DM IRB HC, Leica).
Biochemical Assays during in Vitro Angiogenesis-Capillary ECs, ECs carrying vector alone, N17Rac1, or N19RhoA (9 ϫ 10 6 ) were seeded on 100-mm Petri dishes coated with 9 ml of Matrigel, which was immediately aspirated. Cells were incubated at 37°C for the indicated times, and after washing with cold PBS, pH 7.4, cells were treated with 8 ml of MatriSperse Cell Recovery Solution (Biotech, BD Biosciences) at 4°C for 2 h. After complete release from the gel, ECs were centrifuged at 500 ϫ g for 5 min at 4°C.
To analyze the GTP-bound form of Rho GTPases, cells were lysed for 20 min on ice in 50 mM Tris, pH 7.4, 2 mM MgCl 2 , 100 mM NaCl, 1% Nonidet P-40, 10% glycerol supplemented with protease inhibitors. Lysates were centrifuged for 5 min at 13,000 ϫ g, and equal amounts of proteins (500 g) were incubated with 30 g of GST-RBD, GST-PBD, or GST-WBD G-Sepharose beads coupled for 60 min at 4°C. At the end of incubation, beads were washed with the same buffer, and solubilized proteins were resolved on SDS-PAGE (12%), transferred onto polyvinylidene difluoride membranes (Immobilon, Millipore Corp., Bedford, MA), probed with mAbs anti-RhoA, anti-Rac1, or Ab anti-Cdc42, and detected by enhanced chemiluminescence techniques (Amersham Biosciences). Total cell lysates were treated in the same conditions to evaluate the total amounts of RhoA, Rac1, and Cdc42.
Adhesion Assay-96-Well plates were coated with 1.28 g/l Matrigel for 3 h at 37°C, washed, and blocked with 3% BSA for 2 h at 37°C. 1 ϫ 10 4 capillary ECs carrying vector alone or N17Rac1 or N19RhoA were plated to adhere for 30 min. Attached cells were fixed and stained by crystal violet, and the absorbance was read at 540 nm in microtiter plate spectrophotometer (HT6 7000 Bio Assay Reader, PerkinElmer Life Sciences).
Motility Assay-24-Well transwell chambers (Falcon, BD Bioscience Biotech) containing polycarbonate filters with 8-mm pores were used. The underside of the filters were coated with 1.28 g/l Matrigel for 1 h at 37°C and saturated in 3% BSA for 1 h at 37°C. Two hundred microliters of cell suspension containing 3 ϫ 10 4 capillary ECs carrying vector alone or N17Rac1 or N19RhoA were plated in the upper cham-bers of transwell apparatus, and M199 supplemented with 10% FCS was added to the lower chamber. Cells were incubated for 3 h at 37°C in a 95% air, 5% CO 2 atmosphere. After washing with PBS, the cells migrated to the bottom side of the transwell membrane were fixed with 3.7% glutaraldehyde, and cells on the upper side of the filter were carefully removed by scraping with the edge of a cotton swab. Migrated cells were stained with crystal violet, and the absorbance was read at 540 nm in microtiter plate spectrophotometer. Absorbance values were linear from 1 ϫ 10 2 to 2 ϫ 10 4 cells.
Gel Contraction-Gel contraction assay was performed as described previously (41). Capillary ECs (200 cells/mm 2 ) carrying vector alone or N17Rac1 or N19RhoA were plated as detailed above. Cells were incubated at 37°C in a 5% CO 2 humidified atmosphere until they spread. In some experiments 100 nM cytochalasin D or 5 M Y-27632 (Calbiochem) were added from the beginning of incubation. To control the numbers of cell spread on Matrigel, ECs were recovered by using of MatriSperse Cell Recovery Solution and counted. To initiate Matrigel contraction, polymerized gels were gently released from the underlying culture dish. The degree of Matrigel contraction was determined after 14 h by recording pictures with a digital camera ( 1 ⁄2 inch CCD, JVC) connected to a Stereomicroscope Olympus SZX9 (Olympus Europe, Hamburg, Germany). The gel area was calculated by Image Pro Plus 4x software.
Cell Dissociation Assay-Confluent cultured ECs were treated with 0.01% trypsin in Hepes-buffered saline (1 mM CaCl 2 , 1.3 M NaCl, 50 mM KCl, 3.3 mM Na 2 HPO 4 -7H 2 O, 55 mM glucose, 100 mM Hepes, pH 7.4) (TC treatment) or in Hepes-buffered saline without Ca 2ϩ supplemented with 1 mM EGTA (TE treatment) for 15 min at 37°C and dissociated through 10 times pipetting. The extent of dissociation cells was represented by the index N TC /N TE , where N TC and N TE are the number of clusters and single cells, respectively (42).
Immunofluorescence-Cells cultured on glass coverslips coated with Matrigel were fixed for 30 min in 4% paraformaldehyde in PBS, pH 7.5, quenched with 20 mM NH 4 Cl for 30 min, permeabilized with 0.1% saponin in PBS for 30 min on ice, and rinsed twice in PBS. Blocking incubations were performed in PBS containing 10% goat serum and 0.25% fish skin gelatin (1 h, at room temperature). GST-PBD, GST-RBD, or GST-WBD (30 g) were diluted in 10% of blocking solution and incubated for 1 h at room temperature or overnight at 4°C. Cells were extensively washed with PBS and then incubated for 30 min at 37°C with a mAb anti-GST. After washes they were incubated with the secondary FITC-conjugated Ab. In order to exclude interferences of cell fixation on the GTP-bound state of Rho GTPases, pull-down assay was performed on cells fixed, permeabilized, and isolated from Matrigel as detailed above.
Fixed and permeabilized cells were also incubated with mAbs anti-Ve-cadherin, anti-␤-catenin, anti-MLC-P, and anti-GST or Abs anti-RhoA or anti-Rac1, and then visualized with secondary TRITC-conjugated and FITC-conjugated Abs.
For F-actin staining, cells were incubated for 30 min at 37°C with FITC-conjugated phalloidin. ECs infected with a modified Pinco vector without GFP were used for double staining experiments.

Morphological Events of Capillary EC Morphogenesis-
Time-lapse video microscopy over a period of 14 h shows that in vitro vascular network formation by capillary ECs follows specific and peculiar morphological phases (Fig. 1). During the first 30 min, ECs randomly plated on Matrigel spread, moved, and started to form small and seldom interconnected clusters. After 1 h, ECs were completely spread and clusters increased in size and highly connecting among them. These changes became more evident after 2 h, when discrete Matrigel areas were empty and surrounded by EC islets or chains. Four hours after plating, endothelial chains begun stretching and ECs appeared very elongated. This process continued for another 2 h and was accompanied by a reduction in size of cell clusters and the delineation of polygonal spaces whose dimensions were similar to those observed at the end of the process. During the last 8 h EC chains became progressively thinner, and by 14 h these chains gave rise to EC cords interconnected with nodes that originated from the reduction in size of clusters and their compression.
Activation and Cellular Localization of Rho GTPases during Capillary EC Morphogenesis-Because cell motility largely de-pends on the control of the actin cytoskeletal dynamics by the Rho GTPase family (14), we studied the time courses of RhoA, Rac1, and Cdc42 activation and their intracellular localization in ECs during vascular network formation Matrigel. To analyze the enzyme GTP-bound state of Rac1, RhoA, and Cdc42 in capillary EC lysates (27), pull-down experiments were performed by using fusion proteins between GST and the Cdc42/ Rac interactive binding domain of human PAK1B (GST-PBD), or the Rho binding domain of the Rho effector rhotekin (GST-RBD), or the Cdc42/Rac interactive binding domain of the Wiskott-Aldrich syndrome protein (GST-WBD), which binds specifically to active Cdc42 (31).
When ECs were put in suspension the amount of RhoA-GTP was very low. Thirty min after plating ECs on Matrigel, RhoA-GTP markedly increased and remained elevated up to 1 h ( Fig.  2A). After 2 and 4 h, RhoA-GTP decreased, but it was again detectable at 6 h, reaching its maximum at 10 h and subsequently declining again ( Fig. 2A). Rac1-GTP was detected in suspended cells and increased 2 h after plating, remaining elevated up to 4 h (Fig. 3A). Then Rac-GTP progressively decreased and was lower than the basal level after 14 h. The amount of Cdc42-GTP slightly increased in adhering cells as compared with suspended cells but did not change over the time of network formation, including the early step when Rac1 was activated (Fig. 3C). This observation suggested that modulation of Cdc42 activation was not essential in our experimental model.
To gain insights into the subcellular localization of Rho GT-Pases, permeabilized cells were incubated with either GST-PBD, GST-RBD, or GST-WBD, which were then detected by indirect immunofluorescence with an anti-GST Ab. Permeabilization and fixation procedures in the presence of paraformaldehyde did not interfere in the association of Rho GTPase GTP-bound forms with their effectors, as inferred by pull-down experiments performed after capillary EC fixation and permeabilization (Fig. 4). However, this is only a qualitative analysis,  B, Cdc42-GTP staining in ECs after 2 h from the beginning of the morphogenetic process is shown. Cdc42-GTP was localized by incubating the cells with GST-WBD and stained as detailed in the legend to Fig. 2. C, subcellular localization of Rac1-GTP after 2 h from the beginning of the morphogenetic process. ECs were fixed and permeabilized, and F-actin cytoskeleton was stained with TRITC-phalloidin (a). Total Rac1 was detected by Ab anti-Rac1, which was stained by a TRITC anti-rabbit IgG (g). Rac1-GTP (b, d, f, and h) was localized by incubating the cells with GST-PBD and stained as detailed in the legend to Fig. 2. d, arrow and arrowhead show Rac1-GTP localization in ruffles and at cell-to-cell contact, respectively. Images a and b and g and h were respectively merged in c and i showing (arrows in c) the sites of co-localization of Rac1-GTP with F-actin and i the distribution of Rac1-GTP compared with total Rac-1. e and f, fixed and permeabilized cells were co-incubated with mAb anti-Ve-cadherin (e), which was stained by anti-mouse TRITC-IgG, and GST-PBD (f), which was stained by as above described. as paraformaldehyde is a cross-linking agent that may interfere in the stoichiometry of antigen-Ab interactions (43). The validity of this technique has been demonstrated previously in COS-7 cells carrying RhoQ63L, RacQ61L, and Cdc42Q61L (44), which respectively represent the constitutively active forms of RhoA, Rac1, and Cdc42 (36,37).
To test the binding specificity of RhoA-GTP, Rac1-GTP to GST-RBD and GST-PBD, ECs cells were infected with Pinco vector carrying the constitutively active forms RhoAQL, Rac1QL, and the dominant negative mutants N17Rac1 and N19RhoA. RhoAQL (Fig. 5c) and Rac1QL (Fig. 5f) but not N19RhoA (Fig. 5b) or N17Rac1 (Fig. 5e), respectively, increased the staining of GST-RBD and GST-PBD. Anti-GST Ab did not detect any subcellular compartment (data not shown and see Ref. 44). No staining was detected following cell incubation with GST protein (data not shown and see Ref. 44). These data suggest that in our experimental conditions the intracellular immunostaining of GST-RBD and GST-PBD specifically tracks the localization of GTP-bound form of RhoA, Rac1 in ECs, as already reported in the nervous system (45).
In spreading ECs, 30 min after plating GST-RBD immunostaining showed predominant plasma membrane and perinuclear localization (Fig. 2B, b and e). At the plasma membrane level, RhoA-GTP concentrated at sites of membrane spreading and ruffling co-localizing with cortical F-actin (Fig. 2B, c). The GST-RBD immunofluorescence signal then became negative paralleling the low amount of RhoA-GTP detected in the pulldown assay (not shown). At 10 h EC-forming cords were characterized by few prominent stress fibers and by F-actin accumulation at cell-cell contacts (Fig. 2C, a). GST-RBD was found highly concentrated in areas near cell-cell contact and along the elongated body of the cell (Fig. 2C, arrows in b, d, and e). In these areas, it co-localized with F-actin (Fig. 2C, arrows in c). Fig. 2, B, d, and C, f, shows the distribution of RhoA, which appeared diffused in the cytosol and in the plasma membrane. Fig. 2, B, f and C, h, shows the merge between total RhoA and RhoA-GTP demonstrating that the active form was mainly concentrated in the plasma membrane 30 min after the beginning of morphogenetic process and in specific areas along the longitudinal cell axis after 10 h. In general this double staining suggests that only a fraction of RhoA becomes active and is recruited in discrete cell sites.
When RhoA was in active form (Fig. 2), Rac1-GTP was undetectable in agreement with the low levels of the active Rac1 measured in pull-down assay (data not shown and Fig. 3A). At 2 h, when Rac1 reached its activation peak, GST-PBD staining showed cytoplasm and plasma membrane positivity (Fig. 3C,  b). Fig. 3C shows that Rac1-GTP concentrated in ruffles (d), at the edge of lamellipodia (b), and co-localized with cortical Factin (c). Moreover, in early cell-cell contact events Rac1-GTP was near Ve-cadherin at cell-cell junctions (Fig. 3C, e and f,   arrows). Active Rac1 versus Rac1 staining highlights the specific Rac1-GTP localization (Fig. 2C, f).
Subcellular localization analysis of Cdc42-GTP confirmed the data obtained by pull-down experiments, showing that GST-WBD staining was negligible in ECs (Fig. 3B) and did not co-localize with F-actin during vascular pattern formation (not shown).
Distinct Roles of RhoA and Rac1 during Capillary EC Morphogenesis-The observed differences in temporal activation profiles of distinct Rho GTPases as well as the different distribution of their GTP-bound forms suggested that they could play different functional roles in capillary network assembly. To gain insights into their functions, dominant negative mutants of Rac1 (N17Rac1), RhoA (N19RhoA), and Cdc42 (N17Cdc42) were expressed in capillary ECs through a GFPretroviral vector (Fig. 6B, insets). To verify N17Rac1 and N19RhoA dominant negative activity, pull-down assays were performed when Rac1 and RhoA, respectively, reached its maximal activation level ( Figs. 2A and 3A). Indeed, the amounts of Rac1-GTP at 2 h (Fig. 6A) and of RhoA-GTP at 30 min (not shown) and 10 h (Fig. 6A) were clearly reduced by the expression of the specific dominant negative mutants. Dominant activity of N17Cdc42 was not evaluated because the level of Cdc42-GTP did not change over time in our model. As shown in Fig. 6B, expression of both N19RhoA and N17Rac1 (Fig. 6B, b  and c) impaired EC morphogenesis. The effect of N17Rac1 (Fig.  6B, c) was more dramatic with a relevant impairment of the initial phase when ECs migrate and establish cell-to-cell contacts generating cellular chains. The network formed by N19RhoA-ECs was less mature with evident interruption of cords (Fig. 6B, b). Quantification of capillary-like structures showed that the number of tubes were 123 Ϯ 15/mm 2 , 28 Ϯ 7/mm 2 , and 76 Ϯ 8/mm 2 (mean Ϯ S.D. of five experiments) in cells infected with the vector alone, N17Rac1, and N19RhoA, respectively. Capillary ECs expressing N17Cdc42 self-assembled into a mature vascular network without any morphological alteration (110 Ϯ 5/mm 2 tubes; Fig. 6, d). The role of RhoA and Rac1 was further evaluated in an in vitro angiogenesis assay, in which ECs were aggregated into spheroids, embedded in type I collagen, and stimulated by VEGF-A 165 . Fig. 6C shows that VEGF-A 165 elicited sprouting of capillary ECs from spheroids and that both N17Rac1 and N19RhoA impaired this phenomenon.
To exclude any possible interference of these GTPase mutants in the cell cycle that could influence the morphogenetic process per se, the apoptotic rate was evaluated 10 h after the beginning of the Matrigel assay. As also observed with ECs grown on plastic dishes (not shown), the expression of these mutants did not alter the number of apoptotic cells, which normally appear during in vitro angiogenesis (46,47)  RhoA and Rac1 Regulate Key Biological Functions Required for Capillary EC Morphogenesis-We evaluated how Rac1 and RhoA could regulate key biological function required for the execution of a proper vascular morphogenetic process (7,48) such as EC adhesion, migration toward Matrigel, and gel contraction (Fig. 7). We found that N17Rac1 did not affect EC adhesion to Matrigel (Fig. 7A, a and d), and expression of N19RhoA caused only a little reduction in the number of adhering ECs (Fig. 7A, a) even if it severely impaired the spreading of adherent ECs plated on Matrigel (Fig. 7A, c). Both N17Rac1 and N19RhoA reduced EC haptotaxis toward Matrigel, the former being more active (Fig. 7B).
ECs transduced with the vector alone contracted the Matrigel of 35%, and this effect was strongly reduced by cytochalasin D, an inhibitor of actin polymerization, and Y-27632, an inhibitor of the RhoA effector Rho kinase. These data suggest that Matrigel contraction requires tension generated by actin-myosin contractility (Fig. 7C). These inhibitors blocked also the formation of the tubule-like structures of the ECs (data not shown). Both N17Rac1 and N19RhoA reduced the capacity of ECs to contract the Matrigel, N19RhoA being more active (Fig.  7C). The effect of N19RhoA on gel contraction did not depend on the number of adherent cells because an increase of gel contraction was not observed even plating ECs carrying N19RhoA at higher concentration (not shown).
N17Rac1 Alters ␤-Catenin Localization-It is well established that Rac1 regulates homotypic cell contacts mediated by the cadherin-catenin complex (49). To evaluate the effect of N17Rac1 on adherens junction, we analyzed ␤-catenin and Ve-cadherin localization at cell-to-cell adhesion sites. In ECs infected with vector alone and plated for 2 h on Matrigel, ␤-catenin (Fig. 8A, b) and Ve-cadherin (data not shown) accumulated at cell-to-cell junctions. Capillary ECs expressing N17Rac1 showed a diffuse staining of ␤-catenin, which was not localized at intercellular contacts (Fig. 8A, d), whereas Vecadherin did (Fig. 8A, f). Small amounts of both ␤-catenin and Ve-cadherin was also seen in a punctate, vesicular pattern in the cytosol (Fig. 8A, d and f, insets).
Ve-cadherin forms intracellular complexes with ␤-catenin that recruits ␣-catenin, being responsible for actin cytoskeleton anchorage (50). The dynamics of the complex Ve-cadherincatenins during the early steps of EC morphogenesis (30 min, 2 h) were evaluated by Western blotting analysis of proteins co-immunoprecipitated with ␤-catenin (Fig. 9A). In ECs carrying vector alone, Ve-cadherin co-immunoprecipitated with ␤-catenin heavily increased at 2 h compared with 30 min. ␣-Catenin was already complexed with ␤-catenin at 30 min, and the amount recruited to the complex slightly increased after 2 h. The expression of N17Rac1 did not modify the dynamics of the complex, suggesting that Rac1 inhibition did not impair cadherin-catenins interactions (Fig. 9A). Furthermore, we looked at the presence of IQGAP1 in the complex. This is a Rac1 effector that in the absence of active Rho GTPases may induce the dissociation of ␣-catenin from ␤-catenin (51, 52). In vector-ECs or N17Rac1-ECs, IQGAP1 is barely detectable in total cell lysate but was completely absent in ␤-catenin immu- noprecipitates (Fig. 9A) suggesting that IQGAP1 is not involved at these stages of EC contact formation.
Association of cadherin-catenin complexes with actin cytoskeleton provides strength to adhesion (53). We then investigated the association of the Ve-cadherin complex to actin cytoskeleton analyzing the Triton X-100-insoluble fraction, as detergent insolubility is an indicator for cytoskeletal association (54). In ECs carrying vector alone, we observed an increase of Ve-cadherin and ␤-catenin amounts in the Triton X-100insoluble fraction of cell lysates at 2 h compared with 30 min (Fig. 9B). N17Rac1 partially blocked this shift suggesting that in N17Rac1-ECs cadherin/catenin association to actin cytoskeleton was impaired. To test if the inhibitor effect of N17Rac1 may result from the perturbation of actin dynamics during new ECs contact formation, we looked at actin reorganization at sites of initial cell-cell contacts of ECs carrying N17Rac1. Noninfected ECs showed actin cables perpendicular to the cell-cell contacts, whereas poor actin reorganization and less cable protrusions were observed in N17Rac1 ECs (Fig. 9C). Morphomet-ric analysis performed on 50 cell clusters containing at least one GFP-positive cell carrying the N17Rac1 and one GFPnegative cell indicates that 38.6 Ϯ 5.3% (mean Ϯ S.D. of four experiments) of the N17-Rac1 showed the absence of F-actin cables perpendicular to the nascent adherens junctions.
Because N17Rac1 affected the Ve-cadherin compartmentalization from soluble to insoluble Triton X-100 fraction, we performed a cell dissociation assay to determine cell adhesion activity of Ve-cadherin under conditions of high confluence when cell-cell junctions become mature. Confluent ECs carrying vector alone or N17Rac1 were treated with trypsin in the presence of Ca 2ϩ or EGTA (TC or TE treatment), which respectively favor dissociation of EC monolayer in cluster or single cells (42). The ratio between the number of clusters (N TC ) and that of single cells (N TE ) was calculated (Fig. 9C). Because cadherin-mediated adhesion is reserved after TC but not TE treatment, smaller index reflects stronger adhesion activity. Both ECs carrying vector alone and N17Rac1 showed a similar N TC /N TE index. These results suggest that N17Rac1 effects could be rescued in mature cell-cell contacts of long confluence grown ECs.
MLC Is an Effector of RhoA at Specific Time Points of Capillary EC Morphogenesis-Because RhoA controls cell contraction by regulating the phosphorylation state of MLC (55, 56), we analyzed in N19RhoA-ECs the level of MLC-P and its cell localization by biochemical and immunofluorescence approaches.
At late stages of the vascular morphogenetic process (10 h), ECs showed an intense staining for MLC-P with focal accumulation at the periphery of the cell in the proximity of cell-to-cell contacts where it co-localized with F-actin (Fig. 10A, a and b). In contrast, in N19RhoA expressing ECs, the intensity of MLC-P staining was dramatically decreased and exhibited a diffuse cytoplasmic pattern (Fig. 10A, c and d). Because ECs expressing N17Rac1 did not form stretched cords (Fig. 6B, c), the pattern of MLC-P distribution was not compared with that of ECs carrying vector or N19RhoA. The effect of N19RhoA expression on the phosphorylation status of MLC was further validated by Western blot analysis on lysates from ECs plated on Matrigel for 30 min and 10 h, i.e. the time points when RhoA reached its maximal activation during in vitro vascular morphogenesis. MLC-P levels were clearly reduced in N19RhoA ECs when compared with N17Rac1 or vector alone infected cells (Fig. 10B). DISCUSSION We investigated the role of Rho GTPases during vascular network formation by human capillary ECs plated on a Matrigel scaffold. For the first time, we show here that Rac1 and RhoA are activated at specific time points during this morphogenetic process; we demonstrate that Rac1 is implicated in the early phases when EC migration and cell-to-cell contact formation occur, whereas RhoA is required for generating contractile forces during the remodeling phase.
Time-lapse analysis of the vascular morphogenetic process shows that a spread bidimensional multicellular network, whose geometry is not substantially modified in the following steps, forms during the first 2 h and undergoes complete maturation at later time points. Establishment of the primitive network is characterized by cell migration and maturation of cell-to-cell and cell-to-matrix interactions (7,13,48,57). During this period two peaks of Rho GTPase activation are evident. RhoA-GTP is maximally activated at a very early stage, when ECs start adhering to Matrigel. The increase in Rac1-GTP levels occurs later, when ECs migrate and form Ve-cadherinbased cell-to-cell adhesions. During the early phases of ECs morphogenesis, total RhoA is localized both in the perinuclear region and in ruffling plasma membrane of spreading ECs, but the GTP-bound form concentrates at membrane sites where it co-localizes with cortical F-actin. Total Rac1 is in cytosol, plasma membrane, and lamellipodia, whereas Rac1-GTP mainly co-localizes with F-actin in ruffles and lamellipodia plasma membrane and at cell-to-cell contacts of both ECs chains and clusters. The second stage of the morphogenetic process is primarily characterized by a remodeling of the network in which cells change shape to form capillary-like structures. During this latter phase, RhoA-GTP is augmented and co-localizes with F-actin containing structures, i.e. cell-to-cell junctions and stress fibers running along the longitudinal axis of ECs. Mobilization of Rho GTPases from cytosol to membranes has been described previously in polarized cells (17,58) or after stimulation with platelet-derived growth factor, lysophosphatidic acid, and endothelia (59). However, this is the first demonstration that the GTP-bound form of Rho-GTPases may translocate in specific cell areas, including plasma membrane, during a morphogenetic process.
The use of dominant negative mutants N17Rac1, N17Cdc42, and N19RhoA allowed us to confirm the relevant role of Rac1 and RhoA in capillary network formation. The morphological features of the capillary network at the end of the morphogenetic process obtained with capillary ECs carrying dominant negative mutants indicate that inhibition of Rac1 exerts the earliest and most dramatic effect with a profound alteration of the network. RhoA inhibition slightly compromises cell alignment along the geometric pattern. Cdc42 does not seem to play any role. These data are in contrast with a previous report (57) using a morphogenetic model in which human ECs from cord veins were plated in a tridimensional gel of collagen or fibrin. In this model Cdc42 is required as inferred by the use of a mutant molecule. However, it is well documented that the process of morphogenetic differentiation of ECs on collagen is completely dissimilar from that of ECs on Matrigel both in terms of temporal sequences of the process and types of integrin engaged (5,6,8,60).
In vitro vascular network formation is a dynamic process characterized by three main stages: (i) adhesion to the extracellular matrix, (ii) movement on the extracellular matrix and formation of cell-to-cell contacts, and (iii) mechanical remodeling of the extracellular matrix by cell-generated tension force (7,48,61). N17Rac1 has a striking inhibitory effect on EC haptotaxis, whereas N19RhoA induces a slight inhibition of EC adhesion and motility but a more marked reduction of network remodeling and Matrigel contraction. Our observation that Matrigel contraction is impaired by inhibitors of Rho kinase (Y27632) and actin polymerization (cytochalasin D) suggests that it depends on actin-myosin contractility. Because RhoA is upstream RhoA kinase, we speculate that RhoA-GTP regulates vascular network remodeling through the control of actin-myosin contractility. Pharmacologically inhibiting Rho and RhoA kinase by using C3 exoenzyme and Y27632, respectively, Connolly et al. (13) did not notice any role of RhoA and actinmyosin contractility in the assembly of capillary-like structures. These discrepancies may be due to the different model they employed. Actually, the results by Connolly et al. (13) were obtained in an assay in which large vein ECs grown on glass surface were overlaid with Matrigel. On the contrary, our model focused on the ability of ECs adhering on Matrigel to undergo morphogenesis.
We found that when ECs are re-aligning to form chains, expression of N17Rac1 impairs ␤-catenin but not Ve-cadherin localization at intercellular junctions. By biochemical analysis of Ve-cadherin/catenins compositions, we have demonstrated that N17Rac1 does not modify the features of the complex during the early phases of vascular network formation. Similarly, Ve-cadherin ability to promote calcium-dependent cellto-cell aggregation is not influenced by N17Rac1. However, the expression of this mutant does not allow the shift of Ve-cadherin-␤-catenin complex to the Triton X-100-insoluble cellular fraction that contains cytoskeletal proteins (53). It has been reported that actin binding to the cadherin-␤-catenin complex through ␣-catenin accounts for the shift of the junctional complex to cytoskeleton-associated detergent-insoluble structures (50,54). Therefore, Rac1 seems to be crucial in allowing the association of Ve-cadherin-␤-catenin complex to F-actin at adherens junctions in the early phases of vascular morphogenesis. This conclusion is also supported by immunofluorescence evidence that F-actin is absent at cell-to-cell contact in ECs expressing N17Rac1, and by previous observations (62) that a Ve-cadherin mutant lacking the last cytoplasmic amino acids was still able to support homotypic cell aggregation but was not able to provide cytoskeletal mediated strength to the junction. Therefore, ECs carrying N17Rac1 could arrest their morphogenetic process for the reduction of contractile forces generated during the formation of cell-cell contacts.
The role of Rac1 in controlling established Ve-cadherin/␤catenin-based cell-to-cell contacts in confluent ECs has been reported previously (24, 49, 63, 64) and seems to be largely dependent on the indirect regulation of the phosphorylation state of the complex. Actually, it has been reported that a constitutively active form of Rac1 induces a rapid loss of mature Ve-cadherin-mediated cell-to-cell contacts in confluent ECs concomitantly with an increase in tyrosine phosphorylation of ␣-catenin (64). Our data further extend the role of this GTPase to the early steps of cell junctions formation during in vitro angiogenesis in which Rac1 would regulate the connections between adherens junction and cytoskeleton.
Expression of N19RhoA leads to a dramatic reduction in MLC phosphorylation during the remodeling phase, when ECs assembled in stretched cords and stress fibers undergo orientation along the major cell axis. In control cells, MLC-P is mainly localized at cell-to-cell contact sites and along the major cell axis, whereas it is diffused in the cytosol of cells expressing the Rho mutant. Such MLC-P distribution is likely to be crucial for the control of tension driving forces within ECs during the remodeling phases. The residual MLC-P detected could be due to the activation of MLC kinase by the Ca 2ϩ /calmodulin system (65). In muscle and non-muscle cells, inhibition of myosin phosphatase and the activation of Rho kinase, both regulated by RhoA, play a central role in MLC phosphorylation as well as in the assembly of stress fibers, focal adhesions, and in cell contraction and locomotion (19,55,56,66,67). Alteration of MLC-P in ECs carrying N19RhoA may in part explain their reduced ability to migrate on Matrigel and to contract the gel. Indeed actomyosin filaments provide the driving force for cell rear retraction and for contractility (58). RhoA has been implicated in several contractile events (59) and participates in modulation of the endothelial barrier function through the regulation of cytoskeletal tension (60). Nevertheless, examples of the generation of isometric tension in physiological morphogenetic processes are still poorly described. Our data shed light on the role of RhoA and MLC as key players in the execution of the vascular morphogenetic program.
The role of RhoA and Rac1 in regulating the formation and the dynamics of cell-to-matrix and cell-to-cell adhesive interactions is well established (14,49,68). Here we have shown that the different phases of the morphogenetic process, according to which human capillary ECs plated on Matrigel self-assemble into a vascular network, require distinct spatial and temporal activation of Rac1 and RhoA. Moreover, we have found that, during the different phases of this vascular morphogenetic process, the dynamics of nascent cell-to-cell adhesive contacts and MLC phosphorylation are major targets of Rac1 and RhoA.