Characterization of the Role of the Rab GTPase-activating Protein AS160 in Insulin-regulated GLUT4 Trafficking*

Insulin stimulates the translocation of the glucose transporter GLUT4 from intracellular vesicles to the plasma membrane. In the present study we have conducted a comprehensive proteomic analysis of affinity-purified GLUT4 vesicles from 3T3-L1 adipocytes to discover potential regulators of GLUT4 trafficking. In addition to previously identified components of GLUT4 storage vesicles including the insulin-regulated aminopeptidase insulin-regulated aminopeptidase and the vesicle soluble N-ethylmaleimide factor attachment protein (v-SNARE) VAMP2, we have identified three new Rab proteins, Rab10, Rab11, and Rab14, on GLUT4 vesicles. We have also found that the putative Rab GTPase-activating protein AS160 (Akt substrate of 160 kDa) is associated with GLUT4 vesicles in the basal state and dissociates in response to insulin. This association is likely to be mediated by the cytosolic tail of insulin-regulated aminopeptidase, which interacted both in vitro and in vivo with AS160. Consistent with an inhibitory role of AS160 in the basal state, reduced expression of AS160 in adipocytes using short hairpin RNA increased plasma membrane levels of GLUT4 in an insulin-independent manner. These findings support an important role for AS160 in the insulin regulated trafficking of GLUT4.

Insulin stimulates the translocation of the glucose transporter GLUT4 from intracellular vesicles to the plasma membrane. In the present study we have conducted a comprehensive proteomic analysis of affinity-purified GLUT4 vesicles from 3T3-L1 adipocytes to discover potential regulators of GLUT4 trafficking. In addition to previously identified components of GLUT4 storage vesicles including the insulin-regulated aminopeptidase insulin-regulated aminopeptidase and the vesicle soluble N-ethylmaleimide factor attachment protein (v-SNARE) VAMP2, we have identified three new Rab proteins, Rab10, Rab11, and Rab14, on GLUT4 vesicles. We have also found that the putative Rab GTPase-activating protein AS160 (Akt substrate of 160 kDa) is associated with GLUT4 vesicles in the basal state and dissociates in response to insulin. This association is likely to be mediated by the cytosolic tail of insulinregulated aminopeptidase, which interacted both in vitro and in vivo with AS160. Consistent with an inhibitory role of AS160 in the basal state, reduced expression of AS160 in adipocytes using short hairpin RNA increased plasma membrane levels of GLUT4 in an insulin-independent manner. These findings support an important role for AS160 in the insulin regulated trafficking of GLUT4.
Glucose transport into mammalian muscle and fat cells is an important step in insulin action and is critical for the maintenance of glucose homeostasis within the body (1). In mammalian muscle and fat cells, insulin stimulation activates a phosphorylation cascade, which in turn causes intracellular vesicles that contain the glucose transporter GLUT4, 4 to translocate to the plasma membrane (PM) and fuse (2,3). In the basal state GLUT4 is distributed between the endosomal system, the trans-Golgi network (TGN), and a GLUT4 storage vesicle (GSV) compartment that is highly insulin-responsive (4 -6).
The protein kinase Akt is activated in response to insulin and plays a critical role in GLUT4 translocation (1,7). However, the link between the insulin signaling pathway and GLUT4 translocation is not fully understood. The insulin-dependent movement of GLUT4 vesicles to the PM is an Akt-independent process, and this is followed by an Aktdependent step likely involving the docking and fusion of vesicles with the PM (7)(8)(9). The mechanism by which Akt controls the docking and fusion of GLUT4 vesicles with the PM is not known. However, it was previously shown that a Rab GTPase-activating protein (RabGAP) known as AS160 is phosphorylated by Akt in response to insulin (10). How AS160 functions in GLUT4 trafficking and its cognate Rab proteins are not known. The role of a variety of Rab proteins in GLUT4 trafficking including Rab3d, Rab4, Rab5, and Rab11 has been examined (11)(12)(13)(14)(15)(16). However, although these Rab proteins may participate in some aspects of GLUT4 trafficking, no compelling evidence for specific involvement in the insulin-regulated trafficking of GLUT4 has been found.
In this study we describe four key findings that add to our understanding of GLUT4 trafficking. Using mass spectrometry we have identified three Rab proteins on GLUT4 vesicles that could potentially be substrates of AS160 and, thus, play an important role in insulin-regulated glucose transport. In addition, we have shown that AS160 is present on GLUT4 vesicles in the basal state and dissociates with insulin. We have gone on to show that the association of AS160 with GLUT4 vesicles is mediated at least in part via a direct interaction with the cytosolic tail of IRAP, and this interaction appears to be insulin-dependent. Furthermore, a decrease in AS160 expression by shRNA leads to increased levels of GLUT4 at the PM in the basal state. Together, these results provide novel insights into the regulatory mechanism of insulin-stimulated GLUT4 translocation.

EXPERIMENTAL PROCEDURES
Materials-3T3-L1 murine fibroblasts, C2C12 fibroblasts, and Chinese hamster ovary (CHO) cells were purchased from the American Type Culture Collection (ATCC, Manassas, VA). Dulbecco's modified Eagle's medium and newborn calf serum were obtained from Invitrogen, Myoclone-Plus fetal calf serum from Trace Scientific (Melbourne, Australia), and antibiotics were from Invitrogen. Paraformaldehyde was from ProSciTech (Thuringowa Central, Australia  (19), and VAMP3 (18) have been described previously. ALEXA 488 and Cy3-conjugated secondary antibodies were obtained from Molecular Probes (Leiden, The Netherlands) or Jackson ImmunoResearch (West Grove, PA), respectively. Horseradish peroxidase-conjugated secondary antibodies were from Amersham Biosciences.
Production of a Rabbit AS160 Antibody-Rabbit polyclonal antibodies against human AS160 (TBC1D4) were produced as previously described (18) using a region of human AS160 from amino acids 621-766 fused with GST. Serum from these rabbits was affinity-purified using the AS160 antigen coupled to a GF-2000 column according to manufacturer's instructions.
Cell Culture-3T3-L1 fibroblasts, CHO, and C2C12 fibroblasts were cultured as described previously (5). Briefly, cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% newborn calf serum or fetal calf serum, 2 mM L-glutamine, 100 units/liter penicillin, and 100 g/liter streptomycin at 37°C in 10% CO 2 and passaged at ϳ60% confluence. In the case of 3T3-L1 fibroblasts, confluent cells were differentiated into adipocytes and used between days 8 and 10 postdifferentiation and between passages 10 and 20. To establish basal conditions before use, cells were incubated in serum-free Dulbecco's modified Eagle's medium for 2 h at 37°C in 10% CO 2 .
Subcellular Fractionation of Adipocytes-3T3-L1 adipocyte fractionation was carried out as described previously (5). Briefly, cells were incubated for 2 h in Dulbecco's modified Eagle's medium containing 25 mM glucose at 37°C and lysed using 12 passes through a 22-gauge needle followed by 6 passes through a 27-gauge needle in HES buffer (20 mM HEPES, 10 mM EDTA, 250 mM sucrose pH 7.4) containing Complete protease inhibitor mixture and phosphatase inhibitors (2 mM sodium orthovanadate, 1 mM pyrophosphate, 1 mM ammonium molybdate, 10 mM sodium fluoride) at 4°C. The lysate was then centrifuged at 500 ϫ g for 10 min to remove unbroken cells and at 10,080 ϫ g for 12 min, 15,750 ϫ g for 17 min, and 175,000 ϫ g for 75 min at 4°C to obtain the PM, mitochondria and nuclei high density microsomal pellet containing the endoplasmic reticulum and large endosomes, and the low density microsomal (LDM) pellet which contained intracellular transport vesicles and the majority of GLUT4 vesicles, respectively. The LDM was then resuspended in PBS-containing inhibitors for use in immunoprecipitations.
Cationic Colloidal Silica Plasma Membrane Isolation-Plasma membranes were purified as per Chaney and Jacobson (21) with modifications. Briefly, basal or insulin-stimulated 3T3-L1 adipocytes were washed twice with ice-cold PBS and twice in ice-cold coating buffer (20 mM MES, 150 mM NaCl, 280 mM sorbitol, pH 5.0 -5.5). Cationic silica 1% (stored as a 30% stock and diluted to 1% in coating buffer) was added to the cells in coating buffer for 2 min on ice. Excess silica was removed by washing once with ice-cold coating buffer. Sodium polyacrylate (1 mg/ml, pH 6 -6.5) was added to the cells in coating buffer and incubated at 4°C for 2 min. Cells were washed once in ice-cold coating buffer and then washed with modified HES (20 mM HEPES, 250 mM sucrose, 1 mM dithiothreitol, 1 mM magnesium acetate, 100 mM potassium acetate, 0.5 mM zinc chloride, pH 7.4) at 4°C and lysed as described above. Nycodenz (100%) in modified HES buffer was added to the lysate to a final concentration of 50%. The lysate was layered onto 0.5 ml of 70% Nycodenz in modified HES and centrifuged in a swing-out rotor at 41,545 ϫ g for 20 min at 4°C. The supernatant was discarded, and the pellet was resuspended in 0.5 ml of modified HES buffer and centrifuged at 500 ϫ g for 5 min at 4°C. The pellet was resuspended in SDS-PAGE sample buffer and heated to 65°C for 10 min. The sample was then centrifuged at 10,000 ϫ g at 24°C, and the supernatant was retained for analysis by immunoblotting.
Immuno-isolation of GLUT4 Vesicles and Mass Spectrometry Analysis-GLUT4 vesicles were immuno-isolated using the protocol of Hashiramoto et al. (22) with the exception that either control mouse IgG or the anti-GLUT4 monoclonal 1F8 was covalently coupled to GF-2000 beads according to manufacturer's instructions. LDM (2.5 mg) was resuspended in 100 l of PBS and added to 50 l of both control mouse IgG and 1F8 beads and incubated overnight with rotation at 4°C in the presence of 0.1% bovine serum albumin and PBS in a final volume of 250 l. The beads were then washed 5 times with PBS, and the proteins were eluted with 100% formic acid. The eluate was removed and vacuum-dried. Ammonium bicarbonate (500 mM) was added to adjust the pH to 8 and vacuum-dried again. Urea (10 M) and 5 mM dithiothreitol were then added, and samples were incubated at 37°C for 1 h. Cysteines were subsequently alkylated using a 10-fold molar excess of iodoacetamide and incubated at 37°C for 1 h. Proteins were digested by the addition of modified trypsin (12.5 ng/l) in 100 mM NH 4 HCO 3 and incubated overnight at 37°C. The digestion was stopped by the addition of 5% formic acid, and peptides were desalted using C18 Stagetips. Peptides were analyzed by two-dimensional liquid chromatography tandem mass spectrometry using a CapLC high performance liquid chromatography system (Waters, Milford, MA) by binding to a strong cation exchange cartridge and sequential elution of the peptides using salt steps of 5,10,15,20,25,30,40,50,75,150, 300, 1000 mM ammonium acetate. After each salt step eluted peptides were desalted on a capillary C18 cartridge and resolved on a 100-mm ϫ 75-m C18 Magic reverse phase analytical column with a flow rate of 200 nl/min. Peptides were ionized by nanoelectrospray at 2.8 kV from the end of the column, which was pulled to an inner diameter of 5 m by a P-2000 laser puller (Sutter Instruments Co). Tandem mass spectral analysis was carried out on a Waters (Milford, MA) quadrupole time-of-flight Ultima mass spectrometer. A data-dependent acquisition method was used for all experiments where precursor ions needed to have an intensity higher than 10 counts and be in the ϩ2, ϩ3, or ϩ4 charge state. MS/MS spectra were searched against a metazoan data base generated from Swiss-Prot and Trembl containing 574734 sequences using Sequest (Thermo Electron Corporation, Waltham, MA). Peptides were counted as valid if in the ϩ2 charge state they had a Xcorr value Ͼ2 or in the ϩ3 charge state, an Xcorr Ͼ3.5. All peptides also needed a ⌬CN value greater than 0.08. One missed cleavage per peptide was tolerated, and peptides could be partially tryptic. Proteins identified by less than two peptides were validated manually.
Immunoblotting-All samples were subjected to SDS-PAGE analysis on 10% resolving gels according to Laemmli (23). Equal amounts of protein were loaded for each sample in a single experiment, with 1-10 g per lane unless otherwise stated. Separated proteins were electrophoretically transferred to polyvinylidene difluoride membrane, blocked with BB (2% nonfat skim milk in 0.1% Tween 20 in PBS), and incubated with primary antibody in BB. After incubation, membranes were washed 3 times in BB and incubated with horseradish peroxidaselabeled secondary antibodies in BB. Proteins were visualized using Supersignal West Pico chemiluminescent substrate and imaged using a Versadoc 5000 imager (Bio-Rad).
Sucrose Gradient Flotation Experiments-LDM (165 l) obtained from 3T3-L1 adipocytes as described above was mixed with 835 l of 70% sucrose solution in HES to a final concentration of 60% sucrose (24) and placed in an 11 ϫ 60-mm centrifuge tube overlaid with 1 ml of 50, 30, and 10% of sucrose and 0.4 ml of 5% sucrose. The sample was then centrifuged for 18 h at 111,132 ϫ g in a Beckman SW61 swing-out rotor. Fractions (0.4 ml) were collected from the bottom of the tube by gravitational flow.
Confocal Laser Scanning Microscopy-3T3-L1 adipocytes were cultured as described above on glass coverslips. The cells were serumdepleted for 2 h at 37°C, after which they were incubated in the absence or presence of 200 nM insulin for 20 min. Cells were then fixed with 3% paraformaldehyde in PBS. Fixed cells were washed with PBS, and free aldehyde groups were quenched with 50 mM glycine in PBS. The cells were then processed for immunolabeling by permeabilization and labeling in PBS containing 0.1% saponin and 2% bovine serum albumin using standard procedures. Primary antibodies were detected with ALEXA-488 or Cy3-conjugated secondary antibodies. Optical sections were analyzed by confocal laser scanning microscopy using a Leica TCS SP system. For double labeling, fluorophores were scanned separately and overlaid using Adobe Photoshop software. Images were generated by the maximum projection of a stack sections from the middle of each cell.
GLUT4 Recycling Experiments-3T3-L1 adipocytes expressing a control shRNA or the AS160 shRNA and HA-GLUT4 were incubated in the presence or absence of 100 nM insulin for 20 min. Anti-HA antibody (60 g/ml) was then added for 10 and 60 min. Subsequently the cells were fixed, permeabilized, and incubated with a secondary Cy3conjugated anti-mouse antibody. Confocal laser scanning microscopy was carried out as described above.
Transfection of C2C12 Cells and CHO Cells-C2C12 and CHO cells were transiently transfected with DNA constructs for expression of FLAG-tagged AS160 and the AS160 and control shRNA constructs using Lipofectamine 2000 according to manufacturer's instructions.
Immunoprecipitation of Endogenous AS160 from 3T3-L1 Adipocytes-The LDM fraction from basal or insulin-stimulated cells prepared as described above was resuspended in IP buffer (60 mM ␤-octylglucoside, 1% Triton X-100, 137 mM sodium chloride, 10% glycerol, 25 mM Tris, pH 7.4), incubated for 30 min at 4°C, and centrifuged at 200,000 ϫ g for 15 min. An aliquot of the supernatant (200 g) was incubated with either IgG or anti-AS160 or anti IRAP antibodies coupled to CNBr-activated Sepharose 4B beads overnight at 4°C. Beads were washed four times in IP buffer without ␤-octylglucoside and twice in PBS and boiled in sample buffer.
Retroviral Transfection of shRNA-The shRNA for AS160 was created using standard procedures with the target sequence 5Ј-TAAC-GAGGATGCCTTCTAC-3Ј as described previously (25). 3T3-L1 fibroblasts were infected with pBabe-puro-HA-GLUT4 retrovirus as previously described (5). After 24 h cells were subsequently infected using a similar protocol with pBabe-hygro-shRNA-AS160. Doubly infected cells were selected in 2 g/ml puromyocin and 200 g/ml hygromyocin. The surface HA-GLUT4 was measured in non-permeabilized adipocytes with anti-HA antibody as described previously (26). Subsequently the cells were permeabilized and incubated with anti-AS160 antibody. The quantification of fluorescence of single cells infected with retrovirus expressing an shRNA against AS160 and HA-GLUT4 was performed using the Leica confocal software. The confocal images of the different conditions and cells were scanned with exactly the same settings. A region of interest was set around each cell, and the amount of fluorescence per unit area was determined for each channel using the Leica confocal software.

Identification of GLUT4 Vesicle-associated Proteins in 3T3-L1
Adipocytes-To identify proteins that regulate GLUT4 trafficking, we used two-dimensional liquid chromatography tandem mass spectrometry to identify all of the proteins associated with immuno-isolated GLUT4 vesicles. The immuno-isolation of GLUT4 vesicles was performed using antibodies covalently attached to acrylamide beads (27). To ensure that we ignored proteins that were isolated non-specifically, we used a subtractive experimental design where we compared the protein composition of eluates from the control IgG versus that obtained with the GLUT4 antibody. Only peptides identified using the GLUT4 antibody and not the control IgG in at least three separate experiments were classified as valid.
The consensus of these experiments in 3T3-L1 adipocytes yielded 48 proteins (TABLE ONE), many of which are known to be involved in vesicle trafficking. Details of peptides identified from each protein are given in Supplemental Table 2. The proteins identified fall into five classes; they are previously known GLUT4 vesicle proteins (e.g. sortilin), vesicle transport regulatory proteins (e.g. SNAREs), proteins from the secretory pathway (e.g. cystatin c), recycling membrane proteins (e.g. low density lipoprotein receptor), and proteins with unknown GLUT4 vesicle transport functions (e.g. ubiquitin, V-ATPase). Samples were isolated from basal and insulin-stimulated adipocytes to determine whether we could identify additional proteins that associated with GLUT4 vesicles in an insulin-specific manner. However, no additional proteins were identified in samples isolated from insulin-treated cells. NOVEMBER 11, 2005 • VOLUME 280 • NUMBER 45

Summary of GLUT4 vesicle proteins identified by mass spectrometry from 3T3-L1 adipocytes
Proteins shown were the consensus of three separate experiments. Ϫ, no significant translocation; ϩ, Ͻ2-fold translocation; ϩϩ, Ͼ2-fold translocation. Summary of localization of protein identified in GLUT4 vesicles was based on either the present study ( †) or other published data.

AS160 and GLUT4 Trafficking
We next wanted to determine which of these proteins might play a functional regulatory role in GLUT4 trafficking or indeed determine which proteins may be targeted with GLUT4 into insulin-responsive vesicles. To accomplish this we divided these proteins into integral versus peripheral membrane proteins.
Characterization of Integral Membrane Proteins Found in GLUT4 Vesicles-It has previously been shown that GLUT4 is found in GSVs, endosomes, and the TGN (28 -30). Therefore, it was not surprising to find a large number of endosomal and TGN-derived proteins in immuno-isolated GLUT4 vesicles. To determine which of these were localized to GSVs, we used two separate criteria. First, we examined insulin-responsive movement to the PM, as this is a distinguishing feature of GSV constituents (31). Second, we performed co-localization experiments to compare the intracellular distribution of each integral membrane protein with GLUT4.
To quantify the insulin-dependent movement of proteins to the PM, we developed a method to isolate highly purified plasma membranes (21). Electron microscopy studies of these membranes revealed that they contained large PM sheets with very little contamination from other organelles (data not shown). The only other organelle we could detect by electron microscopy was mitochondria. We were unable to detect Golgi or endoplasmic reticulum membranes proteins in the purified PM fraction by immunoblotting (data not shown), which suggests that this is a highly purified fraction. In contrast PM markers such as syntaxin 4 or caveolin 1 were highly enriched in this fraction (data not shown). We observed very little GLUT4 in the PM in the absence of insulin, whereas after insulin stimulation there was a time-dependent movement of GLUT4 to the PM, and this effect was inhibited by the phosphatidylinositol 3-kinase inhibitor wortmannin (Fig. 1A). The level of the t-SNARE syntaxin 4 in the PM remained constant under these conditions (Fig. 1A). Among the integral membrane proteins examined GLUT4, IRAP and VAMP2 exhibited the largest insulin-dependent increase at the PM, showing 10-, 6-, and 2.5-fold increases, respectively (Fig. 1B) (n ϭ 4). The protein that exhibited the next most significant change with insulin was the TfR, showing a 1.5-fold increase in cell surface levels (Fig. 1B). In contrast, other endosomal proteins such as VAMP3 did not undergo any significant change in PM localization with insulin. Consistent with previous studies (22,(32)(33)(34)(35) we observed a slight effect of insulin on PM levels of CI-MPR, CD-MPR, and sortilin, but this was much lower than that observed for GLUT4, IRAP, or VAMP2 (Fig. 1b). Intriguingly, the copper-transporting ATPase (ATP7A), which was identified in GLUT4 vesicles and can translocate to the plasma membrane with copper-stimulation (36), did not translocate with insulin-stimulation.
We next compared the intracellular localization of these proteins with GLUT4. A defining feature of the intracellular localization of GLUT4 is its concentration in peripheral vesicles that are thought to represent the insulin-responsive vesicles (37). Our laboratory has previously described quantitative immunoelectron microscopic labeling of basal versus insulin-treated 3T3-L1 and primary rat adipocytes (4, 37). We observed a 53% reduction in GLUT4 labeling of cytosolic vesicles in response to insulin, whereas only an 18% reduction in GLUT4 labeling was seen for the TGN or perinuclear region. These data indicate that the peripheral cytosolic vesicles correspond to the insulin-responsive GSVs. In agreement with this conclusion, we observed significant colocalization between GLUT4, IRAP, and VAMP2 in cytosolic vesicles ( Fig. 2A). However, this was not the case for any of the other integral membrane proteins examined. For example some proteins such as sortilin ( Fig. 2A) and syntaxin 16 (data not shown) were highly concentrated in the perinuclear area, presumably corresponding to TGN labeling, but were not concentrated in the peripheral cytosolic vesicles. Some proteins like the copper-transporting ATPase ATP7A ( Fig. 2A) or VAMP3 (data not shown) exhibited both perinuclear and peripheral cytosolic vesicular staining ( Fig. 2A). However, the peripheral cytosolic vesicular staining did not correspond to that containing GLUT4, indicating that these proteins reside in separate vesicles. Therefore, these studies indicate that the major integral membrane proteins in GSVs are GLUT4, IRAP, and VAMP2.
Characterization of Peripheral Membrane Proteins Found in GLUT4 Vesicles-Several novel peripheral membrane proteins were discovered associated with GLUT4 vesicles in this study by mass spectrometry. Perhaps the most interesting among these were 3 Rab proteins Rab10, Rab11, Rab14, and the RabGAP AS160. Rab11 has previously been studied in the context of GLUT4 trafficking, but there is little evidence to indicate a role for this protein in the insulin-regulated trafficking of GSVs to the PM (11). In contrast, Rab10 and Rab14 have not previously been described in GLUT4 vesicles. To verify our mass spectrometry data, we obtained antibodies against Rab11 and Rab14; however, no Rab10-specific antibodies are currently available. We immunoblotted GLUT4 vesicles isolated from adipocytes that had been incubated in the absence or presence of insulin and found that both Rab11 and Rab14 remained associated with the GLUT4 vesicles with and without insulin stimulation (Fig. 3). Rab7, which had been found to non-specifically associate with GLUT4 vesicles in the mass spectrometry analysis, is also shown (Fig. 3). Rab5, which has previously been shown not to associate with GLUT4 vesicles (38), was neither detected in the control nor in GLUT4 immuno-isolations (Fig. 3). We next wanted to examine the intracellular localization of the Rabs using confocal immunofluores-   NOVEMBER 11, 2005 • VOLUME 280 • NUMBER 45 cence microscopy. To accomplish this we attempted to localize the endogenous Rab proteins because overexpression of Rab proteins has been shown to impair endogenous trafficking pathways (39). Only Rab14 antibodies gave specific labeling in adipocytes. Rab14 displayed punctate staining in the periphery of the cell and perinuclear staining as previously described in other cell types (19) (Fig. 2B). When the colocalization of Rab14 with GLUT4 was studied we found some overlap in peripheral GLUT4 vesicles and in the perinuclear region corresponding to the Golgi/TGN (Fig. 2B). However, co-localization in peripheral GLUT4-containing vesicles was not as significant as that seen for IRAP and VAMP2.

AS160 and GLUT4 Trafficking
The identification of AS160 on the same membranes as the three Rab proteins was particularly interesting and raises the possibility that they may be in vivo substrates for the GAP activity of AS160. To examine the function of AS160 in GLUT4 trafficking, we generated a rabbit polyclonal antibody against a central region of AS160. Consistent with previous studies by Sano et al. (40), our antibody against AS160 specifically recognizes immunoprecipitated FLAG-tagged AS160 (Fig. 4A). This antibody also immunolabeled a protein of 160 kDa in adipocytes that were enriched in the LDM fraction (Fig. 4B), which contains the majority of GLUT4 vesicles (22). We have previously shown that the LDM fraction is comprised of both membrane vesicles and large protein complexes (24). To clarify which structures AS160 is associated with in the LDM, we subjected this fraction to sucrose density flotation analysis. Intriguingly, although the majority of AS160 was found at the bottom of the gradient, a significant amount was found in the vesicle-enriched fraction, migrating at a similar position to GLUT4. To confirm the mass spectrometry data showing the association of AS160 with GLUT4 vesicles, we immunoblotted purified GLUT4 vesicles with the AS160 antibody. Consistent with the mass spectrometry data, AS160 was detected in the GLUT4 vesicles but was not detected in control IgG samples. In addition, we found that insulin stimulation induced a marked decrease in the amount of AS160 associated with GLUT4 vesicles (Fig. 4D). The dissociation of AS160 from GLUT4 vesicles was blocked by treatment with the phosphatidylinositol 3-kinase inhibitor wortmannin before insulin stimulation (Fig. 4E).
Identification of IRAP as the Binding Partner for AS160 on GLUT4 Vesicles-The above data are consistent with a model whereby the interaction of AS160 with GLUT4 vesicles plays a key role in their intracellular sequestration in the basal state. Therefore, we next wanted to identify the molecular basis for this interaction. In view of our data showing that GLUT4, IRAP, and VAMP2 are the major integral membrane proteins in GLUT4 vesicles, we next explored the possibility that AS160 may interact with one or more of these proteins. To test this we performed in vitro GST pull-down experiments using lysates from CHO IR/IRS-1 cells that had been transiently transfected with FLAG-tagged AS160. As shown in Fig. 5A, whereas we were unable to observe any specific binding of AS160 to GST alone, GST fused to the cytosolic tail of VAMP2, or GST fused to the carboxyl tail of GLUT4, there was a significant interaction between AS160 and the GST-IRAP1-109 fusion protein (data not shown) and a GST-IRAP1-58 fusion protein (Fig. 5A). In contrast, we were unable to observe an interaction between AS160 and a truncated IRAP cytosolic tail comprising the first 27 amino acids (Fig. 5A). Moreover, the interaction between AS160 and the IRAP tail was significantly higher using basal compared with insulin-treated cell lysate. To further confirm this interaction, we next endeavored to show an interaction between AS160 and IRAP in vivo. LDM was prepared from basal or insulin-stimulated 3T3-L1 adipocytes and solubilized in 60 mM ␤-octylglucoside, 1% Triton X-100, and the soluble fraction was incubated with either control antibodies or antibodies against AS160 or IRAP coupled to CNBr beads. As shown in Fig. 5B, IRAP co-immunoprecipitated with AS160, and this interaction was only observed in basal but not in insulin-stimulated cells. Similarly, AS160 co-immunoprecipitated with IRAP from adipocyte lysate, but in contrast to that observed in the AS160 immunoprecipitate, this interaction was only slightly reduced by insulin (Fig. 5C). The explanation for this difference is not clear; however, these data suggest that the effect of insulin on the interaction between AS160 and IRAP may be more complex than a simple dissociation event.
Co-localization of AS160 with GLUT4 and Characterization of AS160 shRNA Knockdown-To further characterize AS160 in 3T3-L1 adipocytes, we determined the co-localization of AS160 with GLUT4 by confocal immunofluorescence microscopy. AS160 showed a predominantly peripheral vesicular staining pattern (Fig. 2B). Co-localization of AS160 with GLUT4 indicated some overlap of the two proteins in peripheral vesicles and the perinuclear area (Fig. 2B). The insulin-dependent release of AS160 from vesicles suggests a model whereby the targeting of AS160 to GLUT4 vesicles in the absence of insulin may play an important role in the intracellular sequestration of this compartment. To test this we used RNA interference to suppress the expression of AS160. We constructed a hairpin vector specifically targeting AS160 that was able to suppress AS160 levels by greater than 80% after transient transfection in C2C12 cells (Fig. 6A). To express the shRNA in insulin-responsive adipocytes we employed the pBabe retrovirus. As indicated in Fig. 6A adipocytes infected with this retrovirus demonstrated a 40% reduction in AS160 protein levels. To measure GLUT4 translocation in cells in which AS160 levels were suppressed, we coinfected cells with retroviruses expressing HA-GLUT4 and AS160 shRNA. Because of the variable reduction in AS160 levels, we used a single cell analysis to study the relationship between total AS160 expression and cell surface levels of HA-GLUT4. This analysis revealed that overall, cells infected with the AS160 shRNA had an 8-fold increase in surface HA-GLUT4 labeling compared with control cells (p Ͻ 0.005). Analysis of the individual data points clearly indicates that there is a threshold level of AS160 expression, below which surface levels of HA-GLUT4 are significantly elevated. Importantly, the increase in PM levels of HA-GLUT4 upon reduced AS160 expression were similar to that observed in control cells treated with insulin (Fig. 6B). It has previously been shown that overexpression of the AS160 4P mutant in adipocytes principally affects GLUT4 translocation by perturbing the exocytosis of GLUT4 (41). We wanted to determine whether the AS160 shRNA was functioning in a similar manner. To determine this, recycling experiments were performed in which either control cells or cells expressing the AS160 shRNA were incubated with anti-HA antibody at 37°C for 10 min or 60 min. As shown in Fig. 6C in non-stimulated control cells there was little antibody uptake over the course of the 60-min incubation, consistent with previous studies showing that the basal recycling rate of HA-GLUT4 is very slow (6,42). In contrast, in cells expressing the AS160 shRNA there was a significant increase in antibody uptake even in the absence of insulin, and labeling was detected at an intracellular localization consistent with increased exocytosis of GLUT4 upon AS160 reduction.

DISCUSSION
In this study we have made several novel observations about insulinregulated GLUT4 trafficking in adipocytes. First, using a comprehensive proteomic and biochemical analysis of affinity-purified GLUT4 vesicles we have found that the major membrane protein constituents of GSVs are GLUT4, IRAP, and VAMP2. The remaining constituents are probably endosomal or TGN-derived proteins. Second, we have identified three Rab proteins associated with GLUT4 vesicles, Rab10, Rab11, and Rab14, two of which have not previously been reported to associate with GLUT4 vesicles. Third, we have shown that the putative RabGAP AS160 is associated with GLUT4 vesicles in 3T3-L1 adipocytes and dissociates from GLUT4 vesicles with insulin stimulation. Fourth, we have demonstrated that the binding of AS160 to GLUT4 vesicles in the basal state is mediated at least in part via an interaction with the integral membrane protein IRAP and that this interaction is insulin-regulated. Finally, we show that knockdown of AS160 in adipocytes with shRNA causes insulin-independent movement of GLUT4 to the PM.
Previous studies attempting to identify proteins associated with GLUT4 vesicles used analytical techniques much less sensitive than that used in the present study and have, therefore, only identified the most abundant proteins. The proteins identified previously include GLUT4, IRAP (43), VAMP2 (44), TfR (45), CD-MPR (37), CI-MPR (46), sortilin (35), copper amine oxidase (34), SCAMPs (47), and VAMP3 (for review, see Bryant et al. (1)). We have identified all of these proteins and many other novel proteins not previously identified in GLUT4 vesicles. The large number and the diverse localizations of the proteins identified demonstrate that we have achieved a comprehensive analysis of the majority of proteins associated with GLUT4 vesicles. Based on our morphological and biochemical analysis of a large number of these proteins, we have concluded that GLUT4, IRAP, and VAMP2 are the major proteins that are uniquely targeted to peripheral GLUT4 vesicles in adipocytes, and these studies likely explain the unique responsiveness of these FIGURE 3. Rab proteins associate with immuno-isolated GLUT4 vesicles. The LDM fraction and a control and GLUT4 vesicle immuno-isolation eluate purified from the LDM fraction of 3T3-L1 adipocytes were immunoblotted for the presence of Rab proteins. These fractions were purified from 3T3-L1 adipocytes that had been incubated in the absence (Ϫ) or presence (ϩ) of 100 nM insulin for 30 min. Protein levels were detected by immunoblotting with specific antibodies and visualized by chemiluminescence. Immunoblots are from a representative experiment (n ϭ 4). IP, immunoprecipitated. FIGURE 4. AS160 associates with GLUT4 vesicles in the basal state and dissociates with insulin stimulation in 3T3-L1 adipocytes. A, FLAG-tagged AS160 was overexpressed in CHO cells and immunoprecipitated (IP) using a monoclonal anti-FLAG antibody. Control and FLAG-AS160 lysates and FLAG immunoprecipitates were immunoblotted with the AS160 antibody. B, 3T3-L1 adipocytes incubated in the absence (Ϫ) or presence (ϩ) of 100 nm insulin for 30 min were fractionated to generate cytosol (Cyto), PM, LDM, high density microsome, mitochondria, and nuclei (M/N). AS160 levels were detected by immunoblotting with the AS160 polyclonal antibody. C, 3T3-L1 adipocytes in the basal state were fractionated to obtain the LDM fraction, which was then analyzed by sucrose density flotation. The pellet and fractions 1-11, which represent the bottom and the top of the sucrose gradient, respectively, were then analyzed by immunoblotting. D, the LDM fraction and a control and GLUT4 vesicle immuno-isolation eluate purified from the LDM fraction of 3T3-L1 adipocytes were immunoblotted for the presence of GLUT4, IRAP, VAMP2, and AS160. These fractions were purified from 3T3-L1 adipocytes that had been incubated in the absence (Ϫ) or presence (ϩ) of 100 nm insulin for 30 min. E, GLUT4 vesicle immuno-isolation eluate purified from the LDM fraction of 3T3-L1 adipocytes were immunoblotted for the presence of GLUT4 and AS160. These fractions were purified from 3T3-L1 adipocytes that had been incubated in the absence (Ϫ) or presence (ϩ) of 100 nM insulin for 30 min or incubated with 100 nM wortmannin for 30 min before incubation with 100 nM insulin for 30 min. Protein levels were detected by immunoblotting with specific antibodies and visualized by chemiluminescence. Immunoblots are from a representative experiment (n ϭ 4).
proteins. Notably, insulin also has a smaller effect on the translocation of other proteins including the TfR (Fig. 1b) and GLUT1 (48). There are at least three possible explanations for this effect (1). First, insulin may inhibit general endocytosis. This is unlikely because it has been shown that insulin has no effect on TfR endocytosis in adipocytes (49). Second, a small proportion of the TfR may be targeted to GSVs and undergo exocytosis together with GLUT4, VAMP2, and IRAP. However, Akt is not involved in TfR exocytosis, which suggests that the mechanism for its regulated release is different compared with these other proteins (50). Third, insulin may increase endosomal recycling. This seems likely because a number of endosomal recycling proteins including MPR, GLUT1, and TfR have been shown to undergo a relatively small increase in translocation to the PM with insulin (1). We cannot rule out that some of the other proteins identified for which we do not have antibodies may be in GSVs. However, studies of synaptic vesicle proteins have also shown that synaptic vesicles are quite simple in composition, as their sole role is the stimulated release of neurotransmitter (51). However, similar to this study, low levels of other proteins not crucial for synaptic vesicle exocytosis, such as Vti1A and SCAMPs, are also found in purified synaptic vesicles. This suggests the idea that such factors arise from the recycling of small synaptic vesicles through endosomes or the TGN (52,53). Two proteins that we have identified that may be worthy of future study are the marvel domain-containing proteins proteolipid protein 2 and the myeloid-associated differentiation marker. Proteins containing this domain have been shown to regulate vesicle trafficking, but their exact role is not clear (54). Most of the other proteins found in the purified GLUT4 vesicles are involved or associated with GLUT4 recycling within endosomes or the TGN. This minimal GSV translocation system is consistent with previous electron micros-copy studies of GLUT4 vesicles in adipocytes (37), which showed the majority of GLUT4 is recruited to the PM from vesicles that do not contain recycling membrane proteins such as CD-MPR. It also supports a growing number of studies that suggest the majority of regulatory factors involved in vesicle transport, aside from cargo and SNAREs, are peripherally associated cytosolic factors such as coat proteins or Rab proteins.
Previous studies have tried to examine which Rab proteins play a role in GLUT4 trafficking, but only Rab4 and Rab11 have been implicated in insulin-stimulated GLUT4 vesicle translocation (11,14,15). Expression of a dominant negative mutant of Rab4 in adipocytes inhibits GLUT4 translocation by ϳ50%, and the only direct evidence of a role for Rab4 in GLUT4 translocation is its binding to syntaxin 4 (14). However, the significance of this observation is not clear, as Rab proteins have not been found to bind directly to t-SNAREs in other vesicle transport pathways. Moreover, we failed to observe Rab4 in our GLUT4 vesicle fraction, suggesting that the majority of Rab4 probably does not co-localize with GLUT4 in adipocytes. Rab11 is known to be involved in traffic through recycling endosomes, and it has previously been shown to be associated with GLUT4 vesicles (11). Rab11 translocates to the PM slightly with insulin stimulation (11). However, Rab11 does not translocate to the PM with GTP␥S stimulation, indicating that the PM may not be the Rab11 target organelle (13). Rab11 function has been implicated in insulin-stimulated GLUT4 translocation through interactions with Rab11bp (55). However, the mechanism of action of Rab11 is unknown. The discovery in this study that Rab10, Rab11, and Rab14 are associated with GLUT4 vesicles is exciting, because one or more of these three Rab proteins may be substrates for the RabGAP AS160 and play an integral role in insulin action. Rab14 has been shown to be involved in TGN to endosomal trafficking (19); however, significant amounts of Rab14 are found at the PM. Therefore, Rab14 may have other more specialized roles in exocytosis such as insulin-stimulated GLUT4 translocation. The role of Rab10 has yet to be elucidated. Intriguingly, it is one of the closest mammalian homologues of the yeast sec4 protein (56) and, therefore, may function in the post-Golgi secretory pathway. Sec4 has been shown to regulate assembly of the tethering complex known as the exocyst (57). This is of interest because the exocyst has been implicated in insulin-regulated GLUT4 trafficking in adipocytes (58). Amino acid sequence alignment of the three Rab proteins identified in this study shows that Rab14 and Rab11 are much more closely related to each other than to Rab10, which may indicate that Rab10 has a more unique role in GLUT4 exocytosis. Further work will be required to examine the role of each Rab in GSV translocation and their association with AS160. Intriguingly, Lienhard and co-workers (59) have recently shown in an in vitro GAP assay that the GAP domain of AS160 shows GAP activity toward Rab10 and Rab14.
Recently, we and others have suggested that the movement of GLUT4 vesicles toward the PM is Akt-independent, whereas a step close to the PM, likely involving docking and/or fusion of GLUT4 vesicles, is the major Akt-dependent step (7,9,40,41,60). In this study we have shown that AS160 dissociates from GLUT4 vesicles with insulin stimulation. Furthermore, we have shown that there is a specific association between AS160 and the cytosolic tail of IRAP. This observation is intriguing for a number of reasons. First, it has previously been shown that microinjection of a fusion protein comprising the cytosolic N terminus of IRAP into adipocytes was sufficient to trigger GLUT4 translocation to the PM (61). A similar effect was observed using a fusion protein encompassing the first 58 amino acids of the IRAP tail, and we have also observed that this peptide interacts with AS160 in vitro. Hence, it is tempting to speculate that overexpression of the IRAP cytosolic tail in adipocytes blocks FIGURE 5. Identification of IRAP as the binding partner for AS160 on GLUT4 vesicles. A, CHO cells overexpressing FLAG-AS160 were incubated in the absence or presence of insulin (100 nM) for 15 min. Cells were lysed and incubated with GST alone, GST-GLUT4 CT , GST-VAMP2 1-94 , GST-IRAP 1-27 , and GST-IRAP 1-58 coupled to CNBr-activated Sepharose 4B Beads. AS160 in the GST pull-down was detected with anti-FLAG antibody. B and C, 3T3-L1 adipocytes were incubated in the absence or presence of insulin for 15 min, and a LDM fraction was prepared. LDM was dissolved in 1% Triton X-100 and 60 mM ␤-octylglucoside, centrifuged at 200,000 ϫ g for 15 min, and used for immunoprecipitation (IP) with IgG control, IRAP, or AS160 antibodies. Immunoprecipitates were then immunoblotted with antibodies specific for both AS160 and IRAP. Ø indicates beads-only control. Immunoblots are from a representative experiment (n ϭ 3). NOVEMBER 11, 2005 • VOLUME 280 • NUMBER 45 the interaction of AS160 with the GLUT4 vesicles, thus allowing their constitutive translocation to the PM. Second, it has been shown that GLUT4 trafficking is unaffected in adipocytes from IRAP knock out mice, suggesting that IRAP is not necessary for the intracellular sequestration of the vesicles (62). A likely explanation for this is that AS160 may be capable of interacting with other proteins associated with GLUT4 vesicles, possibly even its putative Rab substrate. Cargo binding may be a general function of RabGAPs in view of the large number of RabGAPs in the human genome, and one could envisage that if each member of this family encoded unique cargo binding specificity, this could represent a novel mechanism for regulating protein trafficking at different locations in eukaryotic cells. Each cargo-RabGAP interaction could be controlled by discrete signaling inputs. In the case of AS160, this appears to involve phosphorylation by Akt, and as described above, this would place the action of AS160 close to the PM. In this way the dissociation of AS160 from GSVs may trigger GTP loading of a relevant Rab protein. Although Rabs have been implicated at several steps in vesicle transport, one of their major functions appears to involve regulation of vesicle docking at the target membrane (63). The fact that the down-regulation of AS160 led to insulin-independent movement of GLUT4 to the PM indicates the involvement of AS160 before fusion. A recent study by Zeigerer et al. (41) has also shown that a step before GLUT4 vesicle fusion is AS160-dependent. These studies also suggest that AS160 plays an important role in the intracellular sequestration of GLUT4 in the absence of insulin. As reported previously (6,26), GLUT4 has a very low rate of cell surface recycling in the absence of insulin as indicated in Fig. 6C. Suppression of AS160 levels enhance this basal recycling rate (Fig. 6C). This is unlikely due to an effect of AS160 on endocytosis because we observed rapid internalization of GLUT4 in cells expressing the AS160 shRNA. This is consistent with experiments performed by Zeigerer and co-workers (41) showing that overexpression of the AS160 4P mutant inhibited GLUT4 exocytosis with no effects on GLUT4 endocytosis. These studies are consistent with a role of AS160 in GLUT4 exocytosis. This demonstrates that AS160 is a specific regulator for the exocytosis of insulin-responsive GLUT4 vesicles in adipocytes. We propose a model of GLUT4 translocation such that in the basal state AS160 is associated with GSVs by interacting with IRAP and possibly other vesicle cargo, and its GAP activity maintains a key GTPase(s) in an inactive GDP-bound form. Upon stimulation with insulin the GSVs move toward the PM, and AS160 is phosphorylated by Akt, leading to its dissociation from GSVs. This allows GTP loading of one or more Rab proteins, enabling the docking and fusion of the GSVs with the PM. . Knockdown of AS160 expression by shRNA induces basal GLUT4 translocation. A, C2C12 cells and 3T3-L1 adipocytes transiently transfected with a control shRNA or the AS160 shRNA were lysed, and protein levels of AS160 and syntaxin 4 were determined by immunoblotting. B, 3T3-L1 fibroblasts were infected with HA-GLUT4 and AS160 shRNA and differentiated into adipocytes. Cell surface HA-GLUT4 levels were determined by incubation of fixed non-permeabilized cells with a mouse anti-HA antibody followed by incubation with a Cy3-conjugated secondary antibody. AS160 was then labeled by permeabilizing the cells and incubating with the AS160 antibody followed by ALEXA 488-conjugated secondary antibody. The confocal images of the different conditions and cells were scanned with exactly the same settings. A region of interest was set around each cell, and the total fluorescence per unit area was determined for Cy3 and ALEXA 488 for each cell shown as arbitrary units. The fluorescence values for each cell type are shown as control shRNA basal (ϫ), control shRNA insulin (Ϫ), AS160 shRNA basal (ࡗ), AS160 shRNA insulin (E). C, 3T3-L1 fibroblasts were infected with a retrovirus expressing HA-GLUT4 and the AS160 shRNA and differentiated into adipocytes. Cells were incubated in the presence or absence of 100 nM insulin for 20 min followed by the addition of HA antibody for 10 or 60 min. Cells were subsequently fixed, permeabilized, incubated with Cy3-conjugated secondary antibody, and processed for immunofluorescence. Images are from a representative experiment.