A direct interaction between Cdc42 and vesicle-associated membrane protein 2 regulates SNARE-dependent insulin exocytosis.

In pancreatic beta cells, insulin granule exocytosis is regulated by SNARE (soluble N-ethylmaleimide-sensitive factor attachment protein receptor protein (SNAP) receptor) proteins, and this is coupled to cortical F-actin reorganization via the Rho family GTPase Cdc42 by an unknown mechanism. We investigated interactions among the target SNARE protein Syntaxin 1A and the vesicle-associated membrane SNARE protein (VAMP2) with Cdc42 and compared these structural interactions with their functional importance to glucose-stimulated insulin secretion in MIN6 beta cells. Subcellular fractionation analyses revealed a parallel redistribution of Cdc42 and VAMP2 from the granule fraction to the plasma membrane in response to glucose that temporally corresponded with the glucose-induced activation of Cdc42. Moreover, within these fractions Cdc42 and VAMP2 were found to co-immunoprecipitate under basal and glucose-stimulated conditions, suggesting that they moved as a complex. Furthermore, VAMP2 bound both GST-Cdc42-GTPgammaS and GST-Cdc42-GDP, indicating that the Cdc42-VAMP2 complex could form under both cytosolic GDP-bound Cdc42 and plasma membrane GTP-bound Cdc42 conformational conditions. In vitro binding analyses showed that VAMP2 bound directly to Cdc42 and that a heterotrimeric complex with Syntaxin 1A could also be formed. Deletion analyses of VAMP2 revealed that only the N-terminal 28 residues were required for Cdc42 binding. Expression of this 28-residue VAMP2 peptide in MIN6 beta cells resulted in the specific impairment of glucose-stimulated insulin secretion, indicating a functional importance for the Cdc42-VAMP2 interaction. Taken together, these data suggest a mechanism whereby glucose activates Cdc42 to induce the targeting of intracellular Cdc42-VAMP2-insulin granule complexes to Syntaxin 1A at the plasma membrane.


In pancreatic beta cells, insulin granule exocytosis is regulated by SNARE (soluble N-ethylmaleimide-sensi-
tive factor attachment protein receptor protein (SNAP) receptor) proteins, and this is coupled to cortical F-actin reorganization via the Rho family GTPase Cdc42 by an unknown mechanism. We investigated interactions among the target SNARE protein Syntaxin 1A and the vesicle-associated membrane SNARE protein (VAMP2) with Cdc42 and compared these structural interactions with their functional importance to glucose-stimulated insulin secretion in MIN6 beta cells. Subcellular fractionation analyses revealed a parallel redistribution of Cdc42 and VAMP2 from the granule fraction to the plasma membrane in response to glucose that temporally corresponded with the glucose-induced activation of Cdc42. Moreover, within these fractions Cdc42 and VAMP2 were found to co-immunoprecipitate under basal and glucose-stimulated conditions, suggesting that they moved as a complex. Furthermore, VAMP2 bound both GST-Cdc42-GTP␥S and GST-Cdc42-GDP, indicating that the Cdc42-VAMP2 complex could form under both cytosolic GDP-bound Cdc42 and plasma membrane GTP-bound Cdc42 conformational conditions. In vitro binding analyses showed that VAMP2 bound directly to Cdc42 and that a heterotrimeric complex with Syntaxin 1A could also be formed. Deletion analyses of VAMP2 revealed that only the N-terminal 28 residues were required for Cdc42 binding. Expression of this 28-residue VAMP2 peptide in MIN6 beta cells resulted in the specific impairment of glucose-stimulated insulin secretion, indicating a functional importance for the Cdc42-VAMP2 interaction. Taken together, these data suggest a mechanism whereby glucose activates Cdc42 to induce the targeting of intracellular Cdc42-VAMP2-insulin granule complexes to Syntaxin 1A at the plasma membrane.
Glucose stimulates insulin exocytosis by inducing fusion of a readily releasable pool of plasma membrane-localized insulin granules (primed) and through the mobilization and trafficking of insulin granules to the cell surface from intracellular storage pools (1,2). Increased glucose flux into pancreatic beta cells results in increased glycolysis and elevation of the intracellular ATP/ADP ratio, which in turn induces closure of the ATPsensitive K ϩ channels and cell depolarization (3,4). This triggers the opening of the voltage-dependent calcium channels and increases intracellular cytoplasmic calcium concentration (5), which culminates in exocytosis of insulin secretory granules from the multiple intracellular pools (6,7). However, the detailed molecular mechanisms governing the targeting of granules to fusion sites at the plasma membrane are not fully understood (8,9).
Like many other regulated vesicle exocytosis events, fusion of granules with the plasma membrane is achieved by the specific pairing of the target membrane soluble N-ethylmaleimide-sensitive factor attachment protein (SNAP) 1 receptor (t-SNARE) proteins with the vesicle membrane SNAP receptor (v-SNARE) (10 -14). The plasma membrane t-SNARE isoforms Syntaxin 1A and SNAP25 associate with the insulin granule v-SNARE protein VAMP2 upon stimulation with glucose to mediate insulin release (15)(16)(17)(18). While this provides a mechanism for fusion in cellular pathways using SNARE proteins to mediate vesicular exocytosis, the pathway involved in the regulated targeting of granules/vesicles to their cognate t-SNAREs at the plasma membrane fusion sites remains unclear.
Filamentous actin (F-actin) is known to be important to the process of insulin secretion (19 -21). Our recent studies in MIN6 beta cells and isolated rat islets demonstrate that glucose transiently modulates cortical actin organization and disrupts the interaction of F-actin with the t-SNARE complex at the plasma membrane to facilitate glucose-stimulated insulin secretion (22). Further evidence indicates that the cortical actin reorganization induced by glucose occurs at a proximal step in the stimulus-secretion pathway perhaps at a step concurrent with the ATP-sensitive K ϩ channel closure (23). However, our understanding of the coupling of SNARE-mediated exocytosis to F-actin reorganization remains incomplete.
Multiple lines of evidence suggest that the F-actin effector protein Cdc42, a Rho family small GTPase, is a downstream target of glucose signaling in beta cells. For example, we have recently shown that stimulation with glucose results in alterations in the cycling of Cdc42 between the GTP-bound activated and GDP-bound inactivated states (23). Cdc42 has also been demonstrated to co-localize with VAMP2-containing insulin secretory granules in pancreatic beta cells (24,25). Moreover Cdc42 has been shown to interact indirectly with the t-SNARE Syntaxin 1A, linking Cdc42 and the actin cytoskeleton to the plasma membrane exocytotic machinery (26). Therefore, given that Cdc42 and VAMP2 are both present on insulin secretory granules and that both interact with Syntaxin 1A, we hypothesized that VAMP2 bridged the interaction between Cdc42 and Syntaxin 1A and that the interactions were functionally important for SNARE-dependent insulin exocytosis.
In this study, we describe a novel mechanism by which Cdc42 regulates SNARE-mediated exocytosis through vesicle targeting. We show that Cdc42 directly interacts with VAMP2 in the insulin granule compartment and that both proteins traffic in parallel to the plasma membrane in response to glucose stimulation at an early step in the stimulus-secretion pathway. This interaction is mediated through the N terminus of VAMP2 and is important for glucose-stimulated insulin secretion. Taken together, these data suggest a mechanism whereby glucose induces the activation of Cdc42 to facilitate the localization of intracellular Cdc42-VAMP2-bound vesicles to the Syntaxin 1A-SNAP25 complexes at the plasma membrane for subsequent fusion. Importantly the Cdc42-VAMP2 interaction may represent a more general mechanism by which vesicles are targeted to the SNARE machinery in other cell types as well.

EXPERIMENTAL PROCEDURES
Materials-Radioimmunoassay grade bovine serum albumin and Dglucose were obtained from Sigma. The rabbit polyclonal anti-Cdc42 antibody, mouse monoclonal anti-VAMP2 antibody, and rabbit polyclonal anti-VAMP2 were purchased from Santa Cruz Biotechnology (Santa Cruz, CA), Synaptic Systems (Gottingen, Germany), and Chemicon (Temecula, CA), respectively. The Syntaxin 1 antibody was obtained from Upstate Biotechnology (Lake Placid, NY). Monoclonal anti-Cdc42 antibody was purchased from BD Transduction Laboratories (Lexington, KY). A thrombin cleavage/capture kit by Novagen (Madison, WI) was used to cleave GST-VAMP2 and GST-Cdc42. Recombinant Syntaxin 1A and Cdc42-His proteins were purchased from Synaptic Systems and Cytoskeleton Inc. (Denver, CO), respectively. The MIN6 cells were a gift from Dr. John Hutton (University of Colorado Health Sciences Center). The ECL kit and Hyperfilm-MP were obtained from Amersham Biosciences.
Plasmids-The pcDNA3-Cdc42 wild-type construct, pcDNA3myc-Cdc42(Q61L), and pcDNA3-myc-Cdc42(T17N) constructs were a gift from Ian Macara (University of Virginia). The pGEX-Cdc42 construct was a gift from Lawrence Quilliam (Indiana University School of Medicine). The full-length pcDNA3-VAMP2-(1-116) construct was obtained as reported previously (27). Nucleotide sequences encoding VAMP2 amino acids 1-94, deletion of the transmembrane domain, were amplified by PCR using oligonucleotides that were designed to contain EcoRI and XhoI restriction sites at the 5Ј-and 3Ј-ends, respectively. This PCR product was subcloned into the pcDNA3 vector. Nucleotide sequences encoding VAMP2 amino acids 1-56 and amino acids 1-28, deletions of the C-terminal half and the N-terminal half of the SNARE motif, respectively, were amplified by PCR using oligonucleotides that were designed to contain BamHI and EcoRI restriction sites at the 5Јand 3Ј-ends, respectively, and subcloned into pcDNA3. PCR products were also designed with EcoRI and SalI restriction sites for insertion into the enhanced green fluorescent protein (pEGFP-N3) expression vector from BD Biosciences to generate VAMP2-(1-56)-EGFP and VAMP2-(1-28)-EGFP.
Cell Culture and Transient Transfection-CHO-K1 cells were purchased from the American Type Culture Collection (Manassas, VA) and cultured in Ham's F-12 medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, and 292 g/ml L-glutamine. At 80 -90% confluency cells were electroporated with 40 g of DNA as described previously (27). Under these conditions ϳ70 -80% of cells were transfected. After 48 h of incubation in medium, cells were harvested in Nonidet P-40 lysis buffer (25 mM Tris, pH 7.4, 1% Nonidet P-40, 10% glycerol, 50 M sodium fluoride, 10 mM sodium pyrophosphate, 137 mM sodium chloride, 1 mM sodium vanadate, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 1 g/ml pepstatin, and 5 g/ml leupeptin), and lysates were cleared by microcentrifugation for 10 min at 4°C for subsequent use in co-immunoprecipitation experiments. MIN6 cells were cultured in Dulbecco's modified Eagle's medium (25 mM glucose) supplemented with 15% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, 292 g/ml L-glutamine, and 50 M ␤-mercaptoethanol as described previously (22,23,28). MIN6 cells plated in 10-cm tissue culture dishes at 40 -60% confluency were electroporated with 300 g of plasmid DNA/cuvette (one 10-cm dish/cuvette) to obtain ϳ50 -70% transfection efficiency using a procedure described previously (29). After 48 h of incubation, cells were washed twice with and incubated for 2 h in 1 ml of modified Krebs-Ringer bicarbonate buffer (MKRBB) (22,23) and stimulated with glucose (20 mM) or KCl (50 mM). Insulin secreted into the MKRBB was quantitated by an insulin immunoassay kit (Linco Research, St. Charles, MO). Cells were subsequently lysed in Nonidet P-40 lysis buffer to generate whole-cell detergent homogenates for quantitation of insulin content by insulin radioimmunoassay and for co-immunoprecipitation assays. For measurement of human C-peptide release, MIN6 cells were transiently co-transfected with the human proinsulin expression vector (pCB6/INS), a kind gift from Chris Newgard (Duke University), using Tfx-50 (Promega) with 2.5 g of DNA/construct/35-mm dish. Forty-eight hours following transfection, cells were stimulated with the secretagogue (20 mM glucose or 50 mM KCl), and MKRBB was collected for quantitation of human C-peptide released using a human C-peptide immunoassay kit (Linco Research).
Recombinant Proteins and Interaction Assays-GST-Cdc42 and GST-VAMP2-(1-94) fusion proteins were expressed in Escherichia coli and purified by glutathione-agarose affinity chromatography as described previously (30). Recombinant VAMP2 was obtained following thrombin cleavage of GST-VAMP2 (Novagen kit). GST-Cdc42 linked to glutathione-Sepharose beads was loaded with GTP␥S or GDP as described previously (31). Briefly 10 g of GST-Cdc42 linked to Sepharose beads was incubated in buffer (0.1 M Tris, pH 7.4, 1 mM EDTA, 2 mM dithiothreitol, 0.2 M NaCl) at a final concentration of 0.1 mM GTP␥S or GDP for 10 min at 30°C and then combined with 50 mM MgCl 2 . GTP␥S-or GDP-loaded GST-Cdc42-Sepharose was incubated with cleared detergent lysates (1 mg) for 2 h at 37°C. Following three washes with phosphate-buffered saline supplemented with 2.5 mM MgCl 2 , proteins were eluted from the Sepharose beads and subjected to 12% SDS-PAGE followed by transfer to PVDF membrane for immunoblotting.
Co-immunoprecipitation and Immunoblotting-CHO-K1 and MIN6 cell cleared detergent homogenates (1-2 mg) were combined with 5 g of mouse VAMP2 (Sy Sy) antibody for 2 h at 4°C followed by a second incubation with protein G Plus-agarose for 2 h. The resultant immunoprecipitates were subjected to 12% SDS-PAGE followed by transfer to PVDF membranes for immunoblotting. Primary antibodies were used at 1:1000 dilutions, and secondary antibodies conjugated to horseradish peroxidase were diluted at 1:10,000 for visualization by enhanced chemiluminescence. Immunoprecipitation experiments using recombinant proteins were performed similarly using 10 g of the VAMP2 (Sy Sy) antibody using a mixture of Syntaxin 1A (10 M), Cdc42-His (10 M), and VAMP2-(1-94) (50 M) proteins in 0.25% Nonidet P-40 lysis buffer.
Subcellular Fractionation-Subcellular fractions were isolated as described previously (32). Briefly MIN6 cells at 70 -80% confluence were washed with cold phosphate-buffered saline and harvested into 1 ml of homogenization buffer (20 mM Tris-HCl, pH 7.4, 0.5 mM EDTA, 0.5 mM EGTA, 250 mM sucrose, and 1 mM dithiothreitol containing the following protease inhibitors: leupeptin (10 g/ml), aprotinin (4 g/ml), pepstatin (2 g/ml), and phenylmethylsulfonyl fluoride (100 M)). Cells were disrupted by 10 strokes through a 27-gauge needle, and homogenates were centrifuged at 900 ϫ g for 10 min. Postnuclear supernatants were centrifuged at 5,500 ϫ g for 15 min, and the subsequent supernatant was centrifuged at 25,000 ϫ g for 20 min to obtain the secretory granule fraction in the pellet. The supernatant was further centrifuged at 100,000 ϫ g for 1 h to obtain the cytosolic fraction. All steps were performed at 4°C. Plasma membrane fractions were obtained using a protocol by Hubbard et al. (33). Briefly the postnuclear pellet was mixed with 1 ml of Buffer A (0.25 M sucrose, 1 mM MgCl 2 , and 10 mM Tris-HCl, pH 7.4) and 2 volumes of Buffer B (2 M sucrose, 1 mM MgCl 2 , and 10 mM Tris-HCl, pH 7.4). The mixture was overlaid with Buffer A and centrifuged at 113,000 ϫ g for 1 h to obtain an interface containing the plasma membrane. The interface was collected and diluted to 2 ml with homogenization buffer for centrifugation at 3,000 ϫ g for 10 min, and the resulting pellet was collected as the plasma membrane fraction. Fractions were assayed for soluble protein content as described previously (34).

Cdc42 and VAMP2 Cycle to the Plasma Membrane upon
Stimulation with Glucose-We have recently reported that glucose-stimulated insulin secretion requires the cycling of Cdc42 activation states and that glucose induces peak activation of Cdc42 within 3 min of stimulation (23). To determine whether the alterations in Cdc42 activation state also affected the cellular localization of Cdc42 during stimulation by glucose, MIN6 beta cell extracts were partitioned into fractions enriched for plasma membrane, cytosolic, and granule components using differential centrifugation. Consistent with previous studies (35,36), the granule fraction had the highest insulin content with lesser amounts present in the plasma membrane (PM) and cytosolic fractions (Fig. 1A). Immunoblot analysis demonstrated the presence of Cdc42 in the PM, cytosol, and granule fractions (Fig. 1B). VAMP2 was localized to the particulate PM and granule fractions and was absent from the cytosol (Fig. 1B) as was the other transmembrane-containing SNARE protein Syntaxin 1A (data not shown). Glucose stimulation for 3 min resulted in an increase in the amount of Cdc42 protein in the plasma membrane fraction (Fig. 1B, lane 2) with a concomitant reduction of Cdc42 in the cytosol and granule fractions when compared with basal levels (Fig. 1B, lanes 3-6). Similarly glucose stimulated an increase in the amount of VAMP2 protein in FIG. 1. Subcellular redistribution of Cdc42 and VAMP2 in response to glucose stimulation. A, MIN6 cells were incubated in glucose-free MKRBB for 2 h followed by homogenization and fractionation into PM, cytosol, and granule fractions for quantitation of insulin content by radioimmunoassay. B, MIN6 cells were incubated in glucose-free MKRBB for 2 h and left unstimulated or stimulated with glucose (20 mM) for 3 min followed by fractionation as described above. Fractions (4 g of PM, 15 g of cytosol, and 2 g of granule protein/lane) were resolved by 12% SDS-PAGE and transferred to PVDF for immunoblotting (IB) using rabbit anti-VAMP2 and mouse anti-Cdc42 antibodies. Coomassie Blue gel staining shows similar protein levels within each set of non-and glucose-stimulated fractions. C, optical density scanning quantitation of three independent sets of fractions immunoblotted for VAMP2. D, optical density scanning quantitation of three independent sets of PM, granule, and cytosolic fractions immunoblotted for Cdc42. Data were normalized to unstimulated ϭ 1 for each fraction per experiment and are shown as means Ϯ S.E. *, p Ͻ 0.05 versus unstimulated using unpaired Student's t test.
the PM (Fig. 1B, lanes 1 and 2) with a parallel reduction in the granule fraction (Fig. 1B, lanes 5 and 6). Quantitation of immunoblots using optical density scanning revealed a 30% increase of VAMP2 in the plasma membrane following 3 min of glucose stimulation and a parallel loss in the granule fraction (Fig. 1C). Cdc42 abundance in the plasma membrane was increased by 2.4-fold with a simultaneous reduction in both the granule and cytosolic fractions after glucose stimulation (Fig.  1D). An independent analysis of Cdc42 and VAMP2 localization using immunofluorescent confocal microscopy showed a similar redistribution to the cell periphery upon glucose stimulation for 3 min (data not shown). Therefore, these data indicated that Cdc42 and VAMP2 co-localized in the insulin-enriched granule fraction and that glucose stimulated recruitment of Cdc42 from the cytosolic and granule compartment to the plasma membrane along with the mobilization of VAMP2-containing gran-ules to the plasma membrane. In addition, these data implicate a coordinated spatial and temporal movement of Cdc42 with the VAMP2-bound insulin secretory granules in response to glucose stimulation.
Cdc42 Associates with VAMP2-bound Insulin Granules-Since both Cdc42 and VAMP2 were found to co-localize in a similar spatial and temporal manner in response to glucose, we next asked whether Cdc42 and VAMP2 interacted and whether such an interaction could be modulated in a glucose-dependent manner. To determine this, VAMP2 was immunoprecipitated from detergent lysates made from MIN6 cells either unstimulated or stimulated with 20 mM glucose for 3 or 5 min ( Fig. 2A), the same time points in which we have previously shown that Cdc42 becomes activated and deactivated, respectively, in MIN6 beta cells (23). Immunoprecipitation of VAMP2 from unstimulated cells resulted in the co-immunoprecipitation of Cdc42, indicating that the two proteins associate under basal conditions. Furthermore Cdc42 was similarly co-immunoprecipitated from lysates prepared from cells stimulated with glucose for 3 and 5 min, suggesting that VAMP2 and Cdc42 complexes formed both on the granules and at the plasma membrane in both basal and glucose-stimulated states.
To determine whether VAMP2-Cdc42 complexes redistributed in parallel from the granule to the plasma membrane fraction, VAMP2 was immunoprecipitated from each fraction FIG. 2. Association of VAMP2 and Cdc42 in MIN6 cells. MIN6 cells were incubated in glucose-free MKRBB for 2 h followed by stimulation with 20 mM glucose for 0, 3, or 5 min, and cleared whole-cell lysates (A) or PM and granule fractions (B) were prepared for immunoprecipitation (IP) using mouse anti-VAMP2 antibody (20 -50 g of PM protein and 80 -200 g of granule protein). Immunoprecipitated proteins were separated by 12% SDS-PAGE, transferred to PVDF, and immunoblotted (IB) for VAMP2 and Cdc42. Shown is the optical density scanning quantitation of Cdc42 abundance normalized to the amount of VAMP2 immunoprecipitated from each fraction. Data are shown as means Ϯ S.E. for three independent experiments.

FIG. 3. VAMP2 interacts with either GDP-or GTP-bound
Cdc42. GST-Cdc42 linked to Sepharose beads was preloaded with GTP␥S or GDP and incubated with MIN6 detergent cell lysates. GST protein alone linked to Sepharose beads was used a control for specificity of binding. Beads were washed, and proteins were eluted for resolution by 12% SDS-PAGE and transferred to PVDF for immunoblotting (IB) with VAMP2, p21-activated kinase 1, and GST antibodies. Data shown are representative of greater than three independent experiments. isolated in the subcellular fractionation experiments for immunoblot analysis (Fig. 2B). In each fraction, VAMP2 and Cdc42 co-immunoprecipitated under both the basal and glucose-stimulated conditions. Moreover the amount of Cdc42 interacting with VAMP2 in each fraction under basal conditions remained relatively unchanged after stimulation with glucose for 3 min. Taken together, these data suggested that Cdc42 and VAMP2 interacted under basal conditions on intracellular pools of insulin granules and were recruited as a complex to the plasma membrane upon glucose-induced activation of Cdc42.
We have previously demonstrated that glucose activates Cdc42 within 3 min of stimulation, but by 5 min Cdc42 cycles to a deactivated state (23). Since we observed similar amounts of Cdc42 interacting with VAMP2 throughout this time course, we questioned whether the activation state of Cdc42 would affect its interaction with VAMP2. GST-Cdc42 bound to Sepharose beads was loaded with either GDP or GTP␥S, a nonhydrolyzable analogue of GTP, and immediately incubated with MIN6 cell lysates (Fig. 3). Precipitation of GST-Cdc42-GTP␥S or GST-Cdc42-GDP beads resulted in the co-precipitation of similar amounts of VAMP2 (Fig. 3, lanes 1 and 2). p21-activated kinase 1, a downstream effector of Cdc42, coprecipitated with GST-Cdc42 loaded with GTP␥S and not GDP, confirming the efficiency of GTP␥S and GDP loading (Fig. 3,  lanes 1 and 2). Similar results were obtained from either unstimulated or stimulated (3-min glucose) cell lysates (data not shown). In addition, GST alone failed to associate with VAMP2, indicating that the interaction between Cdc42 and VAMP2 was specific (Fig. 3, lane 3). Consistent with this, both the Cdc42-T17N (dominant negative mutation that confers predominantly GDP binding) and Cdc42-Q61L (constitutively active mutation that confers predominantly GTP binding) proteins co-immunoprecipitated with VAMP2 from MIN6 cell lysates prepared from T17N-or Q61L-expressing cells (data not shown). Thus, these data demonstrate that the interaction between Cdc42 and VAMP2 occurred independently of Cdc42 activation state and further suggested that Cdc42 could interact with VAMP2-bound insulin secretory granules in the intra-FIG. 5. Cdc42 interacts with the N terminus of VAMP2. A, CHO-K1 cells were co-electroporated with wild-type Cdc42 (Cdc42-WT) and full-length VAMP2 or VAMP2 deletion mutants, all in pcDNA3 vector. Forty-eight hours after electroporation detergent lysates were prepared and immunoprecipitated (IP) with anti-VAMP2 antibody. Proteins were resolved by 12% SDS-PAGE followed by transfer to PVDF membrane for immunoblotting (IB) with anti-Cdc42 antibody. B, immunoprecipitation of VAMP2-(1-56)-EGFP and VAMP2-(1-28)-EGFP fusion proteins from electroporated CHO-K1 cells using a mouse anti-VAMP2 antibody. Proteins were resolved by 12% SDS-PAGE followed by transfer to PVDF membrane for immunoblotting with rabbit anti-Cdc42 and rabbit anti-VAMP2 antibodies. Whole-cell lysates (WCL) were immunoblotted with anti-EGFP antibody to demonstrate the expression of all three proteins. Data shown are representative of three independent experiments. cellular storage pool(s) and at the plasma membrane.
Cdc42 Interacts Directly with the N Terminus of VAMP2-To first determine whether the binding interaction between Cdc42 and VAMP2 was direct or indirect, recombinant Cdc42 and recombinant soluble VAMP2-(1-94) were combined in detergent lysis buffer (0.25% Nonidet P-40) and immunoprecipitated with either anti-VAMP2 antibody or an IgG control (Fig. 4,  lanes 1 and 2). The anti-VAMP2 antibody specifically immunoprecipitated Cdc42, while the IgG control failed to immunoprecipitate either protein. These data indicated that Cdc42 bound directly to VAMP2. It should be noted that, unlike endogenous VAMP2, the recombinant VAMP2 migrated as a doublet likely due to partial proteolysis during purification and thrombinmediated cleavage from GST.
A recent study has demonstrated that Cdc42 interacts with Syntaxin 1A in the ␤HC9 line of pancreatic beta cells (26), although it was demonstrated that the interaction was not direct. However, since Syntaxin 1A does bind to VAMP2 directly (37), we investigated the possibility that interaction between Cdc42 and Syntaxin 1A could be mediated by VAMP2. To test this, we combined recombinant forms of Cdc42, soluble VAMP2-(1-94), and soluble Syntaxin 1A (⌬TM domain) in detergent lysis buffer for immunoprecipitation with anti-VAMP2 antibody or IgG control (Fig. 4, lanes 3 and 4). Anti-VAMP2 co-immunoprecipitated VAMP2 plus both Cdc42 and Syntaxin 1A, while parallel reactions using an IgG control failed to immunoprecipitate any proteins. These data suggest that a heterotrimeric complex between Cdc42, VAMP2, and Syntaxin 1A formed in vitro and that VAMP2 may mediate the observed indirect interaction between Syntaxin 1A and Cdc42 in beta cells.
To determine whether the VAMP2-Cdc42 interaction was specific to insulin-secreting cells, we co-expressed each protein in CHO-K1 cells for immunoprecipitation by monoclonal anti-VAMP2 antibodies raised against the N terminus of VAMP2 (Fig. 5A). CHO-K1 cells were chosen for heterologous expression because they can support regulated vesicle trafficking using VAMP3 (38) but have little to no immunodetectable endogenous VAMP2 or Cdc42. Expressed Cdc42 protein was co-immunoprecipitated only from cells that also co-expressed VAMP2 (Fig. 5A, lanes 1 and 2). These data were further confirmed by the reciprocal co-immunoprecipitation of VAMP2 by wild-type Cdc42 in CHO-K1 cells (data not shown) indicating that a factor specific to the MIN6 cells was not required for the interaction.
To identify which region of VAMP2 mediated its interaction with Cdc42, serial C-terminal deletion mutants of VAMP2 were constructed (Fig. 5A). The VAMP2 protein consists of a 'SNARE' coiled coil domain described as being composed of N-SNARE and C-SNARE domains followed by a C-terminal transmembrane (TM) domain (39). VAMP2-(1-94) lacked the TM domain, while VAMP2-(1-56) lacked the TM domain and the C-terminal SNARE domain required for interaction with Syntaxin 1A (39). VAMP2-(1-28) lacked the TM domain and entire SNARE domain. All three mutants were transiently co-expressed with wild-type Cdc42 into CHO-K1 cells and were subsequently found to interact with Cdc42, including the smallest composed of just the first 28 residues of VAMP2 (Fig.  5A). Visualization of the immunoprecipitation of the low molecular weight VAMP2-(1-56) and VAMP2-(1-28) proteins was achieved by fusing each just upstream of the EGFP cDNA for expression in CHO-K1 cells (Fig. 5B). Both VAMP2-(1-56)-EGFP and VAMP2-(1-28)-EGFP fusion proteins were detectable by immunoblotting with anti-VAMP2 and anti-EGFP antibodies and were immunoprecipitated by the anti-VAMP2 antibody in equivalent quantities, while no VAMP2 was immunoprecipitated from cells expressing EGFP alone (Fig. 5B, lane   1). Importantly both VAMP2-EGFP mutants were capable of interacting with Cdc42 (Fig. 5B). Taken together, these results indicated that Cdc42 interacted specifically with the far N terminus of VAMP2 within the first 28 amino acids and that the association did not require a MIN6 beta cell-specific factor.
The Cdc42-VAMP2 Complex Is Functionally Important for Glucose-stimulated Insulin Secretion-To evaluate the functional importance of the interaction between Cdc42 and VAMP2 in regulated insulin secretion, MIN6 cells were cotransfected with the VAMP2-(1-28) truncation mutant and human proinsulin (Fig. 6). The 28-residue fragment of VAMP2 lacks the SNARE domains needed for its interaction with SNARE proteins but does interact with Cdc42 and thus should specifically compete for and disrupt the interaction between endogenous full-length VAMP2 and Cdc42. Human C-peptide is immunologically distinct from the mouse C-peptide secreted from the MIN6 cells and is used to detect secretion specifically from transfected cells (40,41). Stimulation with glucose resulted in a significant increase in release of human C-peptide over basal levels, and this release was reduced by ϳ50% in cells expressing VAMP2-(1-28) (Fig. 6). In contrast, KCl-stimulated human C-peptide release was unaffected by the presence of VAMP2-(1-28). Using the electroporation method of transfection (29), MIN6 beta cells expressing the VAMP2-(1-28) construct showed a similar loss of glucose-stimulated insulin secretion (data not shown). Therefore, these data suggest that the VAMP2-Cdc42 interaction is functionally important for glucose-stimulated insulin secretion but perhaps not for KCl-stimulated secretion.

DISCUSSION
Regulated secretion involves the fusion of intracellular granules/vesicles with the plasma membrane and therefore must have a mechanism to specifically target vesicles to SNARE proteins participating in fusion. We have presented data here from MIN6 beta cells that support a mechanism whereby glucose-induced activation and translocation of Cdc42 from the insulin-enriched granule pools to the plasma membrane provides the mobilization and targeting mechanism required to deliver intracellular VAMP2-bound insulin granules to the appropriate t-SNARE proteins on the plasma membrane for sub- FIG. 6. The interaction between Cdc42 and VAMP2 is functionally important for glucose-stimulated insulin secretion. MIN6 cells were transiently co-transfected with pcDNA3-VAMP2-(1-28) or pcDNA3 alone plus human proinsulin DNA as a reporter of secretion specifically from transfectable cells. After 48 h of incubation cells were placed in glucose-free MKRBB for 2 h followed by stimulation for 1 h with 20 mM glucose (Gluc) or 50 mM KCl. Human C-peptide released into the MKRBB was measured by radioimmunoassay. Data represent the mean Ϯ S.E. from three independent experiments (normalized to basal ϭ 1 for each experiment). *, p Ͻ 0.05 versus glucose-stimulated with pcDNA3 control. V2, VAMP2. sequent docking and fusion (Fig. 7).
We report here a novel direct interaction between Cdc42 and the N terminus of VAMP2 and show this interaction to be functionally important for glucose-stimulated insulin secretion. This interaction occurred in vitro using purified recombinant proteins as well as in a non-insulin-secreting CHO-K1 cell heterologous expression system. Thus, this association may be used as a general mechanism for vesicle targeting to the SNARE machinery at the plasma membrane in other cell types and is not specific to the islet beta cell, although there may be post-translational modifications in other cell types that alter this interaction. Moreover in MIN6 beta cells we showed that the Cdc42-VAMP2 complex formed in the granule pool and that glucose induced a parallel redistribution of the complex to the PM. In addition, the interaction occurred independently of Cdc42 activation state: inactive (GDP-bound) Cdc42 localized on the granule and interacted with VAMP2, and the activation and conformational change of Cdc42 induced by GTP binding did not impact the ability of VAMP2 to interact with Cdc42. Furthermore the parallel recruitment of Cdc42 and VAMP2 to the plasma membrane fraction within 3 min of stimulation with glucose was temporally associated with the peak activation induced by glucose (23). Importantly the disruption of the Cdc42-VAMP2 complex using the N-terminal VAMP2 peptide specifically reduced glucose-stimulated insulin secretion. Thus, these data provide evidence for an expanded role for Cdc42 in vesicular trafficking to include targeting of granules such that its activation facilitates the mobilization of VAMP2-bound insulin granules to the plasma membrane for subsequent SNARE-mediated tethering, docking, and fusion events.
Our data support the proposal that a coupling between Syntaxin 1A and Cdc42 through an indirect mechanism could facilitate insulin secretion (26) and further expand upon this by demonstrating that VAMP2 can bridge the interaction between Syntaxin 1A and Cdc42. Truncation analyses revealed that only the N-terminal 28 or fewer residues of the VAMP2 protein were needed for the VAMP2 interaction with Cdc42. This Nterminal region of VAMP2 has been shown not to participate in the formation of the heterotrimeric complex with the t-SNAREs (42). Thus, this conformational availability of the N terminus of VAMP2 is consistent with the data showing that it is accessible to interact with Cdc42 both on the granule in the absence of the t-SNAREs and in their presence at the plasma membrane. It has been well established that VAMP2, Syntaxin 1, and SNAP25 assemble into a heterotrimeric complex with equimolar stoichiometry (43,44); however, the stoichiometry between Cdc42 and VAMP2 is not known. One interpretation of our data might suggest that two molecules of Cdc42 could interact with one molecule of VAMP2 based upon the degree of translocation of Cdc42 (2.4-fold) and VAMP2 (1.3-fold) to the plasma membrane upon glucose stimulation. However, further biophysical analyses are required to quantitatively determine the stoichiometry between Cdc42 and VAMP2.
We have previously shown that glucose induces the activation of Cdc42 (23), and in this report we demonstrate that Cdc42 translocation is coupled to glucose-induced activation of Cdc42 in MIN6 beta cells, although the mechanism by which glucose elicits these responses remains unknown. Glucose may affect Cdc42 activation and subcellular localization through GTPase regulatory proteins such as guanine nucleotide dissociation inhibitors (GDIs) and/or guanine nucleotide exchange factors (GEFs). GDIs play an essential role in the cycling of Cdc42 between the cytosol and membranes (45) and are thought to coordinate Cdc42-target interactions in different membrane locations (46). It has recently been shown that the phosphorylation state of Cdc42 can regulate its interaction with GDIs such that phosphorylation of Cdc42 increases the extraction of Cdc42 from the membrane to the cytosol by RhoGDI (47). In addition, phosphorylation and O-linked glucosylation have been proposed to occur in a dynamic and reciprocal manner on the same serine and/or threonine residues to synergistically control protein-protein interactions and/or protein functions such as subcellular localization, association with binding partners, and turnover (48,49). Moreover we have previously shown that glucose induces glucosylation of Cdc42 following activation (23). Thus, if Cdc42 becomes glucosylated at the phosphorylation site, then the RhoGDI may not be able to extract Cdc42 from the plasma membrane. However, whether glucose induces the phosphorylation of Cdc42 remains unknown. Alternatively glucose could alter the binding of GDIs to Cdc42 to sequester Cdc42 in the cytosol. The alteration in binding could result in the dissociation of GDIs from Cdc42, exposing the hydrophobic moiety at the C terminus of Cdc42, which is thought to promote the association with membranes (for a review, see Ref. 45). Eukaryotic Rho GEFs may also play a role in the ability of glucose to induce the activation and plasma membrane subcellular localization of Cdc42. There are more than 60 eukaryotic GEFs known to catalyze the exchange of GDP for GTP to activate Rho family GTPases, many of which are demonstrated to activate Cdc42 (for a review, see Ref. 50). For example, the Rho GEF known as DOCK9/zizimin 1 has recently been shown to activate Cdc42 through the membrane localization of DOCK9/zizmin 1 and/or removal of RhoGDI from Cdc42 (46,(51)(52)(53). However, at present it is not known whether pancreatic beta cells contain endogenous DOCK9/zizimin 1. Therefore, glucose may suppress negative regulators such as GDIs and concomitantly activate GEFs to synergistically activate Cdc42 in a spatial and temporal manner to mediate insulin secretion.
VAMP2 bound to both the GTP-and GDP-loaded GST-Cdc42 proteins, suggesting that VAMP2 binds to a region other than the effector domain of Cdc42. In the GTP-bound state, the effector domain of the GTPase is in an open conformation to enable interaction with potential effectors to induce downstream cellular responses (50,54,55). Since a downstream effector of Cdc42 would bind exclusively with the GTP-bound Cdc42, VAMP2 is unlikely to be an effector. In addition, VAMP2 does not contain a Cdc42 consensus Cdc42/Rac interactive binding domain (56). Data collected from cells expressing the predominantly GTP-bound (Q61L) or GDP-bound (T17N) forms of Cdc42 further confirmed our finding that VAMP2 interacts with both forms of Cdc42. VAMP2 is therefore unlikely to be a downstream target of Cdc42.
The inhibition of glucose-stimulated insulin secretion from MIN6 cells expressing the N-terminal 28-residue peptide indicates that the interaction between Cdc42 and VAMP2 is functionally important for regulated secretion. This 28-residue segment acts as a dominant negative in competing for the endogenous binding of Cdc42 with VAMP2 without disrupting the interaction between VAMP2 and the t-SNAREs. However, the N terminus of VAMP2 has been shown to mediate interaction with non-SNARE proteins such as synaptophysin and the Rab GTPase receptor protein PRA1 (57,58), and so it is formally possible that the 28-mer disrupted interactions between VAMP2 and other proteins in addition to Cdc42. However, the 28-mer did not universally inhibit critical interactions since KCl-stimulated insulin secretion was unaffected. The lack of full inhibition may be indicative of a more specific role for the Cdc42-VAMP2 interaction perhaps in the targeting of specific granule pools or during a particular phase of insulin secretion.
The temporal and spatial effects of glucose upon Cdc42 activation and translocation of Cdc42-VAMP2 complexes are consistent with a role for Cdc42 at a proximal step in the stimulussecretion pathway. The disruption of the Cdc42-VAMP2 complex using the N-terminal VAMP2 peptide specifically reduced glucose-stimulated insulin secretion but not KCl-stimulated insulin secretion, consistent with the notion that Cdc42 plays a role specifically in glucose-stimulated insulin secretion at a step proximal to or concurrent with KCl-mediated channel closure. Although KCl-stimulated secretion does require VAMP2 (59), it does not exhibit the requirement for the Cdc42-VAMP2 interaction. Thus, Cdc42 may function either upstream of KCl in the classical stimulus-secretion pathway or may function in a parallel pathway exclusive to glucose activation that regulates vesicle/granule mobilization and/or targeting to the plasma membrane. This proposed role for Cdc42 in insulin secretion would be consistent with the reports of others showing that Cdc42 regulates targeting of secretory vesicles (46,60) and enhances exocytosis by mast cells (61,62). Future investigations are aimed at better defining the pathway in which Cdc42 participates in vesicle targeting in glucose-stimulated insulin secretion.