Gα12 Interaction with αSNAP Induces VE-cadherin Localization at Endothelial Junctions and Regulates Barrier Function*

The involvement of heterotrimeric G proteins in the regulation of adherens junction function is unclear. We identified αSNAP as an interactive partner of Gα12 using yeast two-hybrid screening. glutathione S-transferase pull-down assays showed the selective interaction of αSNAP with Gα12 in COS-7 as well as in human umbilical vein endothelial cells. Using domain swapping experiments, we demonstrated that the N-terminal region of Gα12 (1–37 amino acids) was necessary and sufficient for its interaction with αSNAP. Gα13 with its N-terminal extension replaced by that of Gα12 acquired the ability to bind to αSNAP, whereas Gα12 with its N terminus replaced by that of Gα13 lost this ability. Using four point mutants of αSNAP, which alter its ability to bind to the SNARE complex, we determined that the convex rather than the concave surface of αSNAP was involved in its interaction with Gα12. Co-transfection of human umbilical vein endothelial cells with Gα12 and αSNAP stabilized VE-cadherin at the plasma membrane, whereas down-regulation of αSNAP with siRNA resulted in the loss of VE-cadherin from the cell surface and, when used in conjunction with Gα12 overexpression, decreased endothelial barrier function. Our results demonstrate a direct link between the α subunit of G12 and αSNAP, an essential component of the membrane fusion machinery, and implicate a role for this interaction in regulating the membrane localization of VE-cadherin and endothelial barrier function.

Heterotrimeric G proteins, consisting of G␣ and G␤␥ subunits, are crucial regulators of multiple signaling pathways. Activation of G protein-coupled receptors promotes the dissociation of G␣ subunit in a GTP-bound form, and both G␣ and G␤␥ subunits interact with downstream effectors (see Ref. 1 for review). G␣ 12 and G␣ 13 belong to one of the four major families of G␣ subunits (G s , G i , G q , and G 12 ) based on their sequence similarities (2). G␣ 12 and G␣ 13 thus far are implicated in mediating several functionally relevant responses and often have overlapping functions, such as involvement in actin stress fiber formation and focal adhesion assembly (3,4), neurite retraction (5), fibroblast transformation (6 -9), activation of phospho-lipase D (10), activation of c-Jun N-terminal kinase (11), and stimulation of Na ϩ -H ϩ exchanger (8). The monomeric G proteins Rho, Rac, and Cdc42 are effectors of both G␣ 12 and G␣ 13 , and their activation is mediated by several guanine nucleotide exchange factors (GEFs), 1 which bind directly to G␣ 12 and G␣ 13 . Despite overlapping functions and a similar set of effectors, G␣ 12 and G␣ 13 likely couple to distinct sets of receptors (4,5,12,13) and RhoGEFs (14). Moreover, although the absence of G␣ 13 leads to a defect in angiogenesis and is embryonically lethal (15), G␣ 12 knock-out mice have no apparent phenotype and survive (16).
The complex roles of G␣ 12 and G␣ 13 may reflect their ability to interact with many protein partners, such as several Rho-GEFs (14,17,18), radixin (19), the protein Ser/Thr phosphatase PP5 (20), the scaffolding subunit of another protein Ser/ Thr phosphatase PP2A (21), and the chaperone Hsp90 (22), which can mediate a variety of responses (23). Some of these interacting partners bind both G␣ 12 and G␣ 13 , whereas others are specific for one of them.
In vascular endothelial cells, the endothelial-specific cadherin, VE-cadherin, plays fundamental roles in the control of microvascular permeability and angiogenesis (see Ref. 24 and references therein). Newly synthesized VE-cadherin is transported to the plasma membrane and forms a homotypic interaction with VE-cadherin from other endothelial cells, resulting in the assembly of adherens junctions. An important aspect of the monolayer confluence is that cadherin junctions stabilize as cells grow to confluence and become "contact-inhibited" (25). It is possible that the cell contact-mediated inhibition of proliferation is the result of "outside-in" signaling events generated in the trans-interacting cadherins such as the activation of Rac and Cdc42 (26). G␣ 12 has recently been found to bind directly to the C-terminal region of E-cadherin, a transmembrane glycoprotein with a single plasma membrane-spanning region near the C terminus (27), and thus it is possible that it may also be involved in promoting junctional assembly and endothelial cell confluence.
Regulation of cadherin targeting to the plasma membrane is poorly understood. It presumably involves a balance between the delivery of the protein to the membrane and its endocytosis followed by either targeting for degradation or recycling back to the plasma membrane (24,25,28,29). The final step in any intracellular membrane trafficking event is membrane fusion, which requires soluble N-ethylmaleimide-sensitive factor (NSF)-associated protein (SNAP) to associate with SNAREs (SNAP receptors) and NSF (see Ref. 30 and references therein). ␣SNAP (and its yeast homolog, Sec17) couples the energy produced by NSF-mediated ATP hydrolysis to induce conformational changes of SNAREs and leads to the disassembly of the NSF⅐SNAP⅐SNARE complex. Sec17 is a 14-helix ␣/␣ protein with nine N-terminal antiparallel ␣-helices forming a twisted sheetlike structure (with two faces, concave and convex, and two ridges) and five C-terminal ␣-helices forming a globular bundle (31). Depletion of ␣SNAP affects cytokinesis, leading to the disruption of intercellular bridges (32). Also decreased ␣SNAP levels may lead to disruption of apical transport and mislocalization of apical proteins, including cadherin (33).
In this study, we have demonstrated that ␣SNAP specifically interacts with G␣ 12 in yeast two-hybrid and GST pull-down assays. We show herein that ␣SNAP binds to the N-terminal region of G␣ 12 and that the ␣SNAP⅐G␣ 12 interaction is involved in VE-cadherin trafficking to the plasma membrane of endothelial cells, in which it contributes to the formation of adherens junctions and hence the control of cell confluence and endothelial barrier function.
Yeast Two-hybrid Screening and Plasmid Constructs-The yeast two-hybrid MATCHMAKER LexA system (Clontech) was employed to detect specific protein-protein interactions. The G␣ 12 two-hybrid bait was constructed by subcloning the cDNA for human mutationally activated G␣ 12 Q229L into the pLexA polylinker region. A human testis library in the vector pB42AD was screened. All of the colonies were assayed for lacZ and LEU2 reporter gene activities; double-positive clones were picked, and plasmids were isolated and sequenced. (EcoRI-XhoI) insert of ϳ900 bp encoding a fragment of ␣SNAP was recloned into pCMV-Tag3B (Stratagene) or pGEX-4T (Amersham Biosciences) for expression in mammalian cells or in bacteria, respectively.
GST Pull-down Assay and Immunoblotting-Cells were lysed in a buffer containing 50 mM Hepes (pH 7.5), 1 mM dithiothreitol, 50 mM NaCl, 5 mM MgCl 2 , 1% Lubrol, 0.1 mM phenylmethylsulfonyl fluoride, and 1:200 dilution of a protease inhibitor mixture (Sigma). When indicated, AlF 4 Ϫ (5 mM NaF and 50 M AlCl 3 ) was included. Lysates were cleared by centrifugation and incubated for 3 h at 4°C with purified GST or GST⅐␣SNAP (or its various mutants) immobilized on GSHagarose beads. The sorbent was washed three times with lysis buffer. Bound proteins were separated on 5-20% gradient SDS gels and transferred onto a polyvinylidene difluoride membrane (Osmonics). Membranes were probed with appropriate antibodies and developed using ECL Plus reagents (Amersham Biosciences). Densitometry of protein bands was performed on scanned images using NIH Image 1.63 software. Data were normalized as described in the legend to Fig. 3. For GST pull-down assays from glycerol gradient fractions, HUVEC extracts were supplemented with GST⅐␣SNAP and incubated for 1 h at 4°C prior to loading on top of the gradient. Gradient fractions were combined with GSH-agarose beads and processed as described above.
Velocity Sedimentation Assay-Velocity sedimentation was performed essentially as described by Wilson et al. (34) with some modifications. HUVEC monolayers (2 days after confluence) were washed three times with phosphate-buffered saline, and cells were collected by centrifugation, resuspended in a buffer (0.5 ml/100-mm plate) containing 20 mM Hepes/KOH (pH 7.4), 100 mM KCl, 2 mm dithiothreitol, 2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and a 1:200 dilution of a protease inhibitor mixture (Sigma) in the presence of 1% Triton X-100 (w/v) or 0.5% Lubrol (decaethylene glycol monododecyl ether, Sigma), solubilized by 20 passages through a 25-gauge needle, and kept on ice for 30 min with occasional agitation. Detergent insoluble material was removed by two consecutive centrifugations, at 17,000 ϫ g for 10 min and at 100,000 ϫ g for 1 h, in a TLX ultracentrifuge (Beckman Instruments). The supernatant was carefully recovered, and protein concentration was measured by Bradford assay. Typically, 0.5 mg of protein was layered on the top of a stepwise 10 -35% (w/v) glycerol gradient. After centrifugation at 40,000 rpm (SW40 rotor, Beckman Instruments) for 20 h at 4°C, 1-ml fractions were recovered, precipitated with trichloroacetic acid, washed with acetone, dried, and dissolved in an SDS gel loading buffer. Fractions were analyzed by immunoblotting as described above.
Confocal Microscopy-HUVECs cultured on gelatin-coated coverslips were fixed with 3.7% paraformaldehyde followed by permeabilization in 0.5% Triton X-100. Cells were incubated with primary antibodies followed by incubation with appropriate secondary antibodies, using Trisbuffered saline containing bovine serum albumin as a blocking buffer.
siRNA Experiments-Inhibition of ␣SNAP expression was performed using ␣SNAP siRNA reagent (Santa Cruz Biotechnology). COS-7 cells were transfected using siRNA transfection reagent (Santa Cruz Biotechnology) or Lipofectamine 2000 (Invitrogen). HUVECs were transfected with siRNA transfection reagent (Santa Cruz Biotechnology) or CytoPure (Q-Biogene) according to the manufacturer's instructions. Santa-Cruz Control siRNA was purchased from Santa Cruz Biotechnology. Cells were harvested 24 or 48 h after transfection.
Assessment of Endothelial Barrier Function-Endothelial barrier function was assessed using an electric cell substrate impedance sensor (ECIS) to measure real-time changes in electrical resistance across endothelial cell monolayers. The measurement with ECIS was performed as described (35)(36)(37) with some modifications. Briefly, HUVECs were plated in a well containing a small gold electrode and transfected as described above. Culture medium was used as the electrolyte. The small electrode and the larger counter electrode were connected to a phase-sensitive lock-in amplifier. A 1-V, 4,000-Hz AC signal was supplied through a resistance of 1 megohm to approximate a constant current of 1 A. The voltage change between the small electrode and the larger counter electrode was continuously monitored, stored, and processed on a computer. The data are presented as the change in the resistive (in-phase) portion of the impedance normalized to its initial value at time zero.

RESULTS
The Yeast Two-hybrid Screening Identifies ␣SNAP as G␣ 12 binding Protein-Full-length constitutively activated G␣ 12 (G␣ 12 Q229L) was used as the "bait" in a yeast two-hybrid screening of a human testis cDNA library (3.5 ϫ l0 6 clones). Among the positive clones (22), one had an insert corresponding to a partial sequence of ␣SNAP (residues 35-262). The interaction between G␣ 12 and the ␣SNAP fragment appeared to be specific because we did not observe the transcription of LEU2 and lacZ in yeast cells transfected with G␣ 12 or ␣SNAP alone (data not shown).
G␣ 12 Interacts with ␣SNAP Independently of Its Activation State-To address whether endogenous G␣ 12 from mammalian cells interacts with ␣SNAP, we employed the GST pull-down assay using GST⅐␣SNAP fusion protein expressed in Escherichia coli. A protein of ϳ46 kDa (i.e. expected molecular mass of endogenous G␣ 12 ) recognized by G␣ 12 -specific antibody was pulled down from COS-7 cells and HUVECs with GST⅐␣SNAP but not with GST alone (Fig. 1A). These results are consistent with the yeast two-hybrid screening data. Interaction of G␣ 12 with its functional partners may occur when it is in its activated form (e.g. interaction with p115 RhoGEF (14)) or independent of its activation state (e.g. interaction with Hsp90 (22) or PP2A (21)). To determine whether the binding to ␣SNAP de-pends on the activation state of G␣ 12 , we performed GST pulldown experiments in the absence or presence of AlF 4 , an activator of G␣ subunits that promotes a conformation similar to that of the transition state for GTP hydrolysis (38). ␣SNAP was shown to bind G␣ 12 both in the absence and in the presence of AlF 4 Ϫ (Fig. 1A), demonstrating that G␣ 12 interacts with ␣SNAP independently of the activation state of G␣ 12 .
␣SNAP Binds Selectively to G␣ 12 -We next examined whether ␣SNAP can interact with other G␣ subunits as well as with G␤␥. In these GST⅐␣SNAP pull-down experiments, we used HA-tagged G␣ 12 , G␣ 13 , and G␣ s , EE-tagged G␣ q and G␣ z , and G␤␥ without tag. Although we detected some specific in-teraction with GST⅐␣SNAP (as compared with GST alone) for all G␣ subunits except G␣ q , the relative amount of protein co-precipitated with GST⅐␣SNAP was considerably higher in the case of G␣ 12 as compared with other G␣ subunits tested (Fig. 1B). No binding of GST⅐␣SNAP to G␤␥ could be detected (Fig. 1C), and no G␤␥ was present in the GST⅐␣SNAP⅐G␣ 12 complex (data not shown). Thus, ␣SNAP selectively interacts with G␣ 12 , whereas interactions with other G␣ subunits are much weaker or not detectable.
N Terminus of G␣ 12 Is Required for Interaction with ␣SNAP-To delineate the structural determinants of the specificity of ␣SNAP binding to G␣ 12 , we compared the primary structures of several G␣ subunits. Comparison of G␣ 12 with G␣ 13 was of particular interest because these two proteins are relatively closely related (67% amino acid identity) and have overlapping sets of binding partners (reviewed in Ref. 23). The most conspicuous difference between G␣ 12 and G␣ 13 is in their N-terminal regions (which are extended in comparison with the other G␣ subunits ( Fig. 2A)): (i) the N-terminal region of G␣ 12 is longer and is characterized by a considerably higher content of charged amino acid residues (eight positive and five negative charges) in comparison with that of G␣ 13 (one positive and three negative charges); (ii) the N-terminal region of G␣ 13 is more hydrophobic than that of G␣ 12 ; and (iii) G␣ 12 and G␣ 13 share virtually no sequence similarity in their N-terminal extensions. Therefore, we addressed whether the difference in the primary structures of the N termini might contribute to the differences in the ability of the two G␣ subunits to interact with ␣SNAP. We took advantage of the chimeric proteins G␣ 12N / G␣ 13C and G␣ 13N /G␣ 12C (13) (see Fig. 2A). Strikingly, an exchange of 37 and 30 N-terminal amino acids between G␣ 12 and G␣ 13 , respectively, rendered G␣ 13 fully competent for binding to ␣SNAP and abolished this ability in G␣ 12 (Fig. 2B). These data show that the N-terminal region of G␣ 12 is the major determinant of its specific interaction with ␣SNAP.  (46) and marked with asterisks. Regions swapped between G␣ 12 and G␣ 13 (13) are underlined. B, interaction of ␣SNAP with HA-G␣ 12 , G␣ 12/13 , G␣ 13/12 , and HA-G␣ 13 was assessed by GST pull-down assay as described in Fig. 1. G␣ subunits were detected by Western blotting (WB) using anti-HA, anti-G␣ 12 , or anti-G␣ 13 antibodies as indicated.
␣SNAP Mutations Interfere with Its Interaction with G␣ 12 -Interaction of ␣SNAP with SNAREs, its major functional partners, has been characterized in detail, and a number of mutations are known that affect assembly and disassembly of ␣SNAP⅐SNARE complexes (39). ␣SNAP consists of a twisted sheet domain (nine N-terminal ␣-helices) with convex and concave surfaces and a globular bundle domain (five C-terminal ␣-helices) (31). The concave surface of the twisted sheet domain interacts with SNAREs (39). To assess whether sites for binding SNAREs and G␣ 12 overlap, we examined the effects of several point mutations in ␣SNAP known to affect its interaction with SNAREs. We also tested interaction of G␣ 12 with the double deletion mutant lacking N-and C-terminal regions, initially identified by yeast two-hybrid screening. We examined four ␣SNAP point mutants with an altered ability to bind and/or disassemble SNARE complexes (39). K53E (␣2 helix) and K122A (␣6 helix) are located on the concave face, and the K140A mutation (␣7 helix) is on the convex face (39) (Fig. 3A). Tyr-200 (see Fig. 3A) is a highly conserved residue located on the edge of the globular bundle and the twisted sheet (39). K53E, K122A, and Y200K mutations dramatically reduce the capacity of ␣SNAP to bind SNARE complexes (39). The K140A mutation enhances SNARE complex disassembly without affecting its binding (39). The K140A mutation considerably increased G␣ 12 binding (Fig. 3, B and C). This effect was similar in the absence or presence of AIF 4 Ϫ used to activate G␣ 12 (data not shown). However, K53E, Y200K (not shown), and K122A (Fig. 3, B and C) mutations had no significant effect on the ␣SNAP⅐G␣ 12 interaction. These results indicate that the ability of ␣SNAP to bind SNAREs does not affect its ability to bind G␣ 12 and suggest that the convex rather than the concave surface of ␣SNAP is important for its binding to G␣ 12 . The double deletion ␣SNAP mutant (lacking ␣1 helix, ␣1-␣2 loop, part of ␣13 helix, and the entire ␣14 helix) was found to interact with G␣ 12 to a greater extent than did wild type ␣SNAP and ␣SNAP (K140A) (Fig. 3, B and C), suggesting that the N-and/or C-terminal regions of ␣SNAP are negative regulator(s) of its interaction with G␣ 12 .
␣SNAP and G␣ 12 Have Overlapping Profiles in a Velocity Sedimentation Assay-Because G␣ 12 is able to bind ␣SNAP, we analyzed the distribution of these proteins in a velocity sedimentation assay to examine whether they are found in the same fractions. Extractability of different membrane-associated proteins may vary considerably depending on the detergent used (40). We tested two different detergents, Triton X-100 (data not shown) and Lubrol (Fig. 4). HUVECs were lysed with each of the detergents, and detergent-soluble material was cleared by sequential low speed and high speed centrifugations and loaded on the top of a 10 -35% (w/v) glycerol gradient. Proteins in different fractions were analyzed by immunoblotting ( Fig. 4 and data not shown). Upon cell solubilization, G␣ 12 and ␣SNAP were detected preferentially in the insoluble fraction. The G␣ 12 distribution profile showed several peaks, probably reflecting its involvement in a broad range of molecular interactions. In particular, the major peak of G␣ 12 overlapped with the peak of ␣SNAP (Fig. 4), which is compatible with the interaction of these proteins in vivo. It is also possible that G␣ 12 and ␣SNAP form a complex that is detergent-insoluble as Ͼ80% of these proteins are recovered in the detergent-resistant material.
␣SNAP and G␣ 12 Stabilize VE-cadherin at the Plasma Membrane-Cadherin, the transmembrane glycoprotein critically involved in cell-cell contact via adherens junctions, binds to G␣ 12 (27). We also observed that VE-cadherin was detected along with G␣ 12 in the material pulled down from glycerol gradient fractions with GST⅐␣SNAP (data not shown). We tested the hypothesis that the G␣ 12 ⅐␣SNAP interaction plays a role in the regulation of VE-cadherin translocation to membrane, thus regulating adherens junctional integrity and endothelial barrier function. To assess the role of G␣ 12 ⅐␣SNAP interaction in VE-cadherin localization to the membrane, we examined the effects of expression of G␣ 12 and/or ␣SNAP on VE-cadherin localization in the HUVEC plasma membrane using confocal microscopy. In subconfluent HUVEC cultures, expression of G␣ 12 as well as co-expression of G␣ 12 and ␣SNAP promoted the localization of VE-cadherin at the membrane FIG. 3. ␣SNAP mutations affecting its interaction with G␣ 12 . Wild type (WT) and four mutants of ␣SNAP were used. A, location of the mutations relative to convex and concave surfaces of ␣SNAP molecule is schematically depicted according to Chae et al. (33) and Marz et al. (39). Interaction between G␣ 12 and ␣SNAP was assessed using GST pulldown assays as described in Figs. 1 and 2. C, C terminus; N, N terminus. B, bound G␣ 12 was detected using anti-G␣ 12 antibody. Relative amounts of bound G␣ 12 were determined by densitometry of scanned images using NIH Image 1.63 software. Data were normalized to GST or GST⅐SNAP content (determined from Coomassie R250 binding) in respective samples. ⌬N⌬C, ␣SNAP double deletion mutant lacking N-and C-terminal regions. C, data shown are the means of two replicates with error bars indicating values from each replicate. Experiment shown is representative of two similar experiments. (Fig. 5, middle and top, respectively); however, expression of ␣SNAP alone had no apparent effect (Fig. 5, bottom). In contrast, G␣ 12 plus ␣SNAP or G␣ 12 alone did not have such a pronounced effect in confluent cultures, which already have a high level of VE-cadherin at the plasma membrane (data not shown).
To address the direct involvement of ␣SNAP in regulating VE-cadherin membrane localization, the effects of G␣ 12 were examined in confluent endothelial cells with ␣SNAP depleted using siRNA. To establish the efficiency of siRNA in decreasing the level of ␣SNAP, we first tested its effect in COS-7 cells. The protein level of ␣SNAP was significantly decreased 24 h after transfection, with depletion levels reaching ϳ 90% 48 h after transfection (Fig. 6A). Depletion of ␣SNAP did not considerably affect the levels of G␣ 12 (Fig. 6B).
When ␣SNAP siRNA was transfected into confluent HUVECs, most of the cells showed moderate or no decrease in ␣SNAP staining and no effect on cadherin. In the cells that showed a strong decrease in endogenous ␣SNAP staining, destabilization of cadherin was detectable by confocal microscopy (Fig. 6C,  upper panel).
When ␣SNAP siRNA and G␣ 12 were co-transfected in confluent HUVECs, the membrane localization of VE-cadherin was notably reduced in the cells overexpressing G␣ 12 (Fig. 6C,  lower panel). To determine whether co-transfection of ␣SNAP siRNA and G␣ 12 affects the barrier function of the HUVEC confluent monolayers, we measured changes in transendothe-lial electrical resistance using ECIS (35)(36)(37). As a point zero, we chose the 3-h period after transfection and continued the resistance measurement for 24 h at 10-min intervals. Fig. 6D shows tracings for a typical experiment in which the monolayers were transfected with ␣SNAP siRNA and/or G␣ 12 . Interestingly, neither ␣SNAP knock-down nor G␣ 12 expression alone produced any effect, whereas a combination of both induced the destabilization of barrier function as evident by decreased electrical resistance. The absence of the effect of ␣SNAP siRNA on monolayer permeability is probably due to a very low proportion of cells with severely depleted ␣SNAP and is consistent with the absence of ␣SNAP siRNA effect on cadherin in most cells.
These functional data show that both G␣ 12 and ␣SNAP are involved in the regulation of VE-cadherin localization to the junctions and the control of endothelial barrier function.

DISCUSSION
In this study we provide evidence that G␣ 12 interacts with ␣SNAP, as demonstrated by yeast two-hybrid assay and GST pull-down experiments. Interaction with ␣SNAP was detected for both overexpressed HA-tagged G␣ 12 and endogenous G␣ 12 in COS-7 cells and HUVECs, eliminating a possibility that it was the result of altered functionality or folding of G␣ 12 because of the introduced HA tag sequence. Both activated and nonactivated forms of G␣ 12 were found to be competent for the interaction with ␣SNAP. This is similar to the binding of G␣ 12 to Hsp90 (22) and to the scaffolding subunit of PP2A (21), which also occurs independently of the activation state of G␣ 12 . We found that among several G␣ subunits tested the interaction with ␣SNAP was specific for G␣ 12 . The major structural determinant(s) of this specificity lie within the 37-amino acid N-terminal region of G␣ 12 . Interestingly, the N-terminal regions of G␣ 12 and G␣ 13 also determine their selective coupling to receptors (13). Thus, the structural dissimilarity of the N-terminal regions of G␣ 12 and G␣ 13 may underlie a number of functional differences between these related G␣ subunits. It is worth noting that at least in three cases (Hsp90, PP2A, and ␣SNAP) the interaction is specific for G␣ 12 versus G␣ 13 , and it is activation-independent. For ␣SNAP and Hsp90, 2 the specificity is conferred by the N-terminal region of G␣ 12 . It is unclear whether this is also the case for the G␣ 12 ⅐ PP2A interaction.
The concave surface of ␣SNAP, which is known to interact with SNAREs (39), does not appear to be involved in the interaction of ␣SNAP with G␣ 12 . The interaction of ␣SNAP with G␣ 12 may involve its opposite, convex surface, because muta-tion of a residue located on the convex surface of ␣SNAP (K140A) increased its binding to G␣ 12 severalfold. The convex surface is negatively charged (31), whereas the N terminus of G␣ 12 is positively charged (see Fig. 2A), raising a possibility that electrostatic interaction is involved in the binding of ␣SNAP to G␣ 12 . The observed effect of the K140A mutation is in line with this possibility, as it would further increase the overall negative charge on the convex surface of ␣SNAP. N-or C-terminal regions of ␣SNAP may negatively regulate the interaction, because the ability to bind G␣ 12 was further increased in a truncated version of ␣SNAP lacking these regions. ␣SNAP is a critical determinant of membrane fusion. Because the ability to interact with ␣SNAP does not appear to be a general property of G␣ subunits but, as we have shown, is limited to G␣ 12 , the interaction of G␣ 12 with ␣SNAP may be important in regulating membrane trafficking and protein delivery to a correct location. VE-cadherin, a transmembrane glycoprotein required for calcium-dependent homotypic adhesion of endothelial cells (24), is localized at the plasma membrane on which it induces the formation of adherens junctions. Cadherin location at the adherens junctions may depend on the balance between its delivery to the plasma membrane and its endocytosis followed by its targeting for degradation (24,25,28,29). These processes involve membrane fusion events in which ␣SNAP would be expected to play a primary role. A decreased level of ␣SNAP mRNA due to a point mutation has been suggested to underlie an abnormal distribution of cadherin and several other proteins in neural cells, leading to defects in neural development and hydrocephaly (33). G␣ 12 has also been shown to interact with several cadherin isoforms, some of which (type I cadherins) bind to G␣ 12 in an activation-dependent manner, whereas other isoforms (type II) do not discriminate between activated and inactive forms of G␣ 12 (27). G␣ 12 binds to cadherin downstream of the ␤-catenin binding domain, induces the release of ␤-catenin, and thus is able to disrupt cadherin-mediated cell-cell adhesion (27,41). Notably, G␣ 12 also binds to the cadherin-interacting protein p120-catenin (42). p120-catenin is believed to serve as "gatekeeper" for cadherins, determining whether they are stably retained at the cell surface or endocytosed (43,44). G␣ 12 has been suggested to control the amount of p120-catenin that is bound to cadherin (27,41,42). In this regard, G␣ 12 appears to be involved in both the functioning and the trafficking of cadherins. Thus, we addressed the possible role of the interaction of G␣ 12 with ␣SNAP in promoting the localization of VE-cadherin at the plasma membrane and in regulating the endothelial barrier function.
The major question addressed by our findings is the functional role of the observed G␣ 12 ⅐␣SNAP interaction. Several lines of evidence demonstrated that G␣ 12 -␣SNAP interaction is critical in the regulation of VE-cadherin trafficking to the plasma membrane and thereby in establishing the integrity of the endothelial barrier. We observed that VE-cadherin was detected along with G␣ 12 in the material pulled down from glycerol gradient fractions with GST⅐␣SNAP, indicating that these proteins exist in a triple complex. Also VE-cadherin, ␣SNAP, and G␣ 12 were co-localized propitiously at the plasma membrane.
The available data are consistent with a model that postulates the following: (i) interaction between G␣ 12 and ␣SNAP stabilizes cadherin on the cell surface, and (ii) G␣ 12 also has an opposite destabilizing effect independent of ␣SNAP. The stabilizing effect is obvious in subconfluent cells, in which it results in a clear appearance of cadherin on the cell surface. The effects of G␣ 12 overexpressed alone or together with ␣SNAP are similar, suggesting that endogenous ␣SNAP levels are not limiting and are sufficient to accommodate not only endogenous but also overexpressed G␣ 12 .
Because in confluent cells cell contacts are formed and cadherin is already stabilized on the cell surface, the stabilizing effect of G␣ 12 ⅐␣SNAP interaction is not easily detectable. In these cells, moderate depletion of ␣SNAP has no effect on cadherin and on monolayer permeability, consistent with sufficient amounts of ␣SNAP still present in most cells to interact with G␣ 12 . As could be expected, severe depletion of ␣SNAP in some cells does lead to cadherin destabilization.
Overexpression of G␣ 12 together with ␣SNAP knock-down destabilizes cadherin because of an effect of G␣ 12 independent of its interaction with ␣SNAP. When G␣ 12 is expressed in the confluent cells without decreasing levels of ␣SNAP, its destabilizing effect on cadherin is masked by the stabilizing effect of G␣ 12 ⅐␣SNAP interaction. The above model is also consistent with the published findings that decreased levels of ␣SNAP lead to abnormal distribution of cadherin (33) and that the activated mutant of G␣ 12 disrupts adherens junctions (45).
Stabilization of VE-cadherin at the membrane in regions of cell contact may occur because of anterograde transport of the newly synthesized protein and preferential recycling of the endocytosed protein to the site of the contact, as opposed to targeting for degradation. Our results do not allow us to distinguish between these two possibilities. However, our results clearly identify a critical role for the G␣ 12 ⅐␣SNAP interaction in inducing VE-cadherin localization at the membrane, which is responsible for maintaining junctional stability in endothelial cells and a normal barrier function.