Insulin-mediated GLUT4 translocation is dependent on the microtubule network.

The GLUT4 facilitative glucose transporter is recruited to the plasma membrane by insulin. This process depends primarily on the exocytosis of a specialized pool of vesicles containing GLUT4 in their membranes. The mechanism of GLUT4 vesicle exocytosis in response to insulin is not understood. To determine whether GLUT4 exocytosis is dependent on intact microtubule network, we measured insulin-mediated GLUT4 exocytosis in 3T3-L1 adipocytes in which the microtubule network was depolymerized by pretreatment with nocodazole. Insulin-mediated GLUT4 translocation was inhibited by more than 80% in nocodazole-treated cells. Phosphorylation of insulin receptor substrate 1 (IRS-1), activation of IRS-1 associated phosphatidylinositide 3-kinase, and phosphorylation of protein kinase B/Akt-1 were not inhibited by nocodazole treatment indicating that the microtubule network was not required for proximal insulin signaling. An intact microtubule network is specifically required for insulin-mediated GLUT4 translocation since nocodazole treatment did not affect insulin-mediated GLUT1 translocation or adipsin secretion. By using in vitro microtubule binding, we demonstrated that both GLUT4 vesicles and IRS-1 bind specifically to microtubules, implicating microtubules in both insulin signaling and GLUT4 translocation. Vesicle binding to microtubules was not mediated through direct binding of GLUT4 or insulin-responsive aminopeptidase to microtubules. A model microtubule-dependent translocation of GLUT4 is proposed.

The GLUT4 facilitative glucose transporter is recruited to the plasma membrane by insulin. This process depends primarily on the exocytosis of a specialized pool of vesicles containing GLUT4 in their membranes. The mechanism of GLUT4 vesicle exocytosis in response to insulin is not understood. To determine whether GLUT4 exocytosis is dependent on intact microtubule network, we measured insulin-mediated GLUT4 exocytosis in 3T3-L1 adipocytes in which the microtubule network was depolymerized by pretreatment with nocodazole. Insulin-mediated GLUT4 translocation was inhibited by more than 80% in nocodazole-treated cells. Phosphorylation of insulin receptor substrate 1 (IRS-1), activation of IRS-1 associated phosphatidylinositide 3-kinase, and phosphorylation of protein kinase B/Akt-1 were not inhibited by nocodazole treatment indicating that the microtubule network was not required for proximal insulin signaling. An intact microtubule network is specifically required for insulin-mediated GLUT4 translocation since nocodazole treatment did not affect insulin-mediated GLUT1 translocation or adipsin secretion. By using in vitro microtubule binding, we demonstrated that both GLUT4 vesicles and IRS-1 bind specifically to microtubules, implicating microtubules in both insulin signaling and GLUT4 translocation. Vesicle binding to microtubules was not mediated through direct binding of GLUT4 or insulin-responsive aminopeptidase to microtubules. A model microtubule-dependent translocation of GLUT4 is proposed.
One of the many physiologic actions of insulin is to promote glucose uptake in skeletal muscle, heart, and adipose tissue. Several lines of evidence have shown that the effect of insulin on glucose uptake in these tissues is a direct result of the recruitment of the GLUT4 1 facilitative glucose transporter from an intracellular vesicle pool to the plasma membrane (for recent reviews see Refs. 1 and 2). GLUT4 was first implicated as the major insulin-responsive glucose transporter when it was shown to be the predominant isoform expressed in tissues exhibiting insulin-stimulated glucose uptake (3)(4)(5)(6). Prior to the cloning and identification of GLUT4, it was known that insulin stimulated the redistribution of a cytochalasin D-inhibitable glucose transport activity from an intracellular vesicle pool to the plasma membrane (7,8). It is generally accepted that the translocatable glucose transport activity is attributable to GLUT4 recruited from an intracellular location to the plasma membrane.
GLUT4 recycles between the plasma membrane and intracellular compartments in both the basal and insulin-stimulated states (9 -11). Insulin increases the amount of GLUT4 on the plasma membrane by stimulating exocytosis of the intracellular GLUT4 pool and by decreasing the endocytosis of the plasma membrane-associated GLUT4 protein (10,(12)(13)(14)(15)(16)(17). Increased exocytosis appears to be the most important event since complete inhibition of the GLUT4 endocytic pathway in the basal state does not result in a significant accumulation of GLUT4 at the plasma membrane (18). These data support a model in which a small portion of the total pool of GLUT4 recycles in the basal state, with the majority of GLUT4 sequestered in an insulin-responsive retention pool shown to be associated with tubulovesicular structures (19).
The mechanism by which GLUT4 vesicles move from their interior location to the plasma membrane in response to insulin is poorly understood. To date, research in this area has focused largely on docking and fusion of the vesicles at the plasma membrane. SNARE (for soluble NSF attachment protein receptor) proteins have been shown to play an important role in the fusion of GLUT4 vesicles with the plasma membrane. It has been clearly demonstrated that syntaxin 4, VAMP2, and Syndet/SNAP23 are the major SNARE proteins regulating GLUT4 membrane fusion (20 -26). SNARE protein assembly may be mediated by other syntaxin 4-binding proteins such as Munc18c and the newly cloned protein, Synip (27)(28)(29)(30)(31), but it is not clear that assembly of SNARE proteins is a rate-limiting step in insulin-mediated GLUT4 translocation. Measurement of GLUT4 translocation in the presence of a temperature-sensitive Munc18c mutant indicates that membrane fusion is not rate-limiting in GLUT4 translocation (32). These data suggest that mechanisms regulating the movement of vesicles from the retention pool to the plasma membrane need to be explored to determine the specific steps in GLUT4 translocation that are regulated by insulin.
Regulated insulin-dependent trafficking of GLUT4 vesicles from the retention pool to the plasma membrane may occur by several mechanisms. GLUT4 vesicles may pass through an endosomal intermediate en route to the plasma membrane. This is unlikely, based on data demonstrating that depletion of * This work was supported by Grant DK47894 from the National Institutes of Health and RA0064 from the American Diabetes Association The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  the endosomal compartment does not interfere with insulinmediated GLUT4 translocation (33,34). Alternatively, GLUT4 vesicle trafficking may be regulated by cytoskeletal elements. GLUT4 secretory vesicles may be tethered to cytoskeletal structures in a manner similar to endoplasmic reticulum and Golgi membrane structures (35). Insulin stimulation could induce a release of tethered vesicles, allowing movement to the plasma membrane by simple diffusion, or insulin may trigger the movement of vesicles along cytoskeletal tracks as has been observed for regulated exocytosis of secretory vesicles in other cell systems (36).
Studies examining the role of the actin cytoskeleton in GLUT4 translocation in L6 myotubes, 3T3-L1 adipocytes, and rat adipocytes show that disruption of the actin cytoskeleton by either cytochalasin D or latrunculin A inhibited insulin-mediated GLUT4 translocation (37)(38)(39). In L6 myotubes, disassembly of the actin network did not prevent GLUT4 translocation by stimuli other than insulin, suggesting that the actin cytoskeleton is required for transduction of the insulin signal to GLUT4 vesicles (40). Disassembly of the actin cytoskeleton in 3T3-L1 cells and L6 myotubes did not interfere with proximal insulin signaling events but did prevent the relocalization of PI 3-kinase to GLUT4 vesicles (38,41).
In this work, we provide evidence that the microtubule cytoskeleton is required for insulin-mediated GLUT4 translocation. We show that insulin-mediated GLUT4 translocation requires an intact microtubule network and that under basal conditions, GLUT4 vesicles associate with polymerized microtubules.

EXPERIMENTAL PROCEDURES
Cell Culture-3T3-L1 fibroblasts were obtained from the American Type Tissue Culture repository and were cultured at 37°C in 5% CO 2 and maintained in Dulbecco's modified Eagle's medium (DMEM) containing 25 mM glucose and 10% calf serum. Confluent cultures were induced to differentiate by incubation of the cells with DMEM plus 25 mM glucose, 10% fetal bovine serum, 175 nM insulin, 1 M dexamethasone, and 0.5 mM isobutyl-1-methylxanthine. After 4 days, the medium was changed to DMEM containing 25 mM glucose, 10% fetal bovine serum, and 175 nM insulin, with the incubation period continuing an additional 3 days. Under these conditions, greater than 95% of the cell population morphologically differentiated into adipocytes.
Treatment of 3T3-L1 adipocytes with drugs to modify the microtubule cytoskeleton were carried out for 3 h in serum-free F-12 Ham's media. Nocodazole was used at a final concentration of 33 M, and taxol was used at a final concentration of 12 M.
Plasma Membrane Sheet Assay-Preparation of plasma membrane sheets was carried out as described previously (23). Briefly, cultured adipocytes grown in a 6-well cluster dish were treated as described in the figure legends. Following experimental treatment, cells were washed with ice-cold PBS and attached to the plate with 0.5% poly-Dlysine. The cells were swollen in hypotonic buffer and then sonicated to release intracellular contents. Pure plasma membrane fragments remaining attached to the plastic dish were scraped into 120 l of solubilization buffer (1% SDS, 20 mM Hepes, pH 7.5, 150 mM NaCl, 1 mM EDTA). Protein content of the plasma membrane sheets was measured in a 10-l aliquot using a spectrofluorometric assay (42). For this assay, the solubilized protein sample was diluted to 60 l in solubilization buffer. The diluted sample was mixed with 0.25 ml of 0.2 M sodium borate buffer, pH 9.0, at room temperature for 5 min. A 20-l aliquot of fluorescamine solution (0.2 mg/ml in acetonitrile) was added with vigorous vortexing. Following a 20-min incubation at room temperature, fluorescence was measured using 395 nm excitation and 460 nm emission at high sensitivity using a Shimadzu RF5000U spectrofluorophotometer. A standard curve was generated using bovine serum albumin ranging from 0.1 to 15 g/ml.
Extraction of Monomeric and Polymeric Tubulin-Monomeric and polymeric tubulin were differentially extracted from cells plated on 35-mm dishes according to previous methods (43). Following treatments as indicated, cell monolayers were washed 2 times with PBS and then incubated with 0.3 ml of extraction buffer (2 M glycerol, 0.1 M Pipes, pH 7.1, 1 mM MgSO 4 , 1 mM EGTA, 10 g/ml leupeptin, 10 g/ml aprotinin, 0.5 mM benzamidine, 5 g/ml pepstatin A, 0.5 mM PMSF, 0.1% Triton X-100) for 30 min at 37°C. The Triton X-100-soluble extract contains monomeric tubulin. Polymeric tubulin was then extracted from the remaining Triton X-100-insoluble material by extracting in SDS lysis buffer (25 mM Tris, pH 7.4, 0.4 M NaCl, 0.5% SDS) for 5 min at 37°C. Aliquots from monomeric and polymeric fractions were fractionated by 10% SDS-PAGE and Western blotted for ␤-tubulin.
Whole Cell Detergent Lysates-100-mm plates of treated 3T3-L1 adipocytes were washed twice with ice-cold TBS followed by freezing in liquid nitrogen. The plates were thawed on ice and scraped into 1 ml of 1% Nonidet P-40 lysis buffer (1% Nonidet P-40, 20 mM Hepes, pH 7.4, 2 mM EDTA) containing phosphatase inhibitors (100 mM sodium fluoride, 10 mM sodium pyrophosphate, 1 mM sodium vanadate, and 1 mM molybdate) and protease inhibitors (10 M leupeptin, 10 g/ml aprotinin, 1.5 mM pepstatin A, and 1 mM PMSF). The cells were lysed on ice for 20 min, and insoluble material was removed by microcentrifugation for 10 min at 4°C. Protein concentrations of the detergent lysates were determined by a Bradford protein assay (Pierce) using manufacturer's specifications.
Electrophoresis and Immunoblotting-Samples were fractionated using SDS-PAGE and transferred to PVDF membranes (Millipore) in transfer buffer (25 mM Tris, 193 mM glycine, pH 8.5) for 3-4 A-h at 4°C. Membranes blotted for insulin receptor ␤-subunit, IRS-1, GLUT4, IRAP, tubulin, and dynein were blocked with 7-10% dried milk and 0.3% Tween 20 in TBS. Anti-phosphotyrosine blots were blocked with 2% bovine serum albumin and 0.1% Tween 20 in TBS. Anti-phosphotyrosine blots were carried out using PY99 monoclonal antibody (Santa Cruz Biotechnology). GLUT4 antiserum was provided by Dr. Gwyn Gould, and IRAP antiserum was provided by Dr. Kostantin Kandror. GLUT1 antibody was provided by Dr. Samuel W. Cushman. Tubulin antibody was obtained from Sigma (TUB 2.1). Phospho-PKB/Akt-1 antibody specific for Ser-473 was obtained from New England Biolabs, and antibody against total PKB/Akt-1 was purchased from Transduction Laboratories. Immunoblots were visualized using an enhanced chemiluminescence system (Pierce) and quantified using scanning laser densitometry.
2-Deoxyglucose Uptake-3T3-L1 adipocytes were placed in Ham's F-12 media containing 0.5% bovine serum albumin for 2 h at 37°C. Nocodazole was added to a final concentration of 33 M at designated times. Cell were then washed with KRPH buffer (5 mM Na 2 HPO 4 , 20 mM Hepes, pH 7.4, 1 mM MgSO 4 , 1 mM CaCl 2 , 136 mM NaCl, 4.7 mM KCl, and 1% bovine serum albumin) and treated with or without additional nocodazole to prevent repolymerization of microtubules and with or without 100 nM insulin added for 20 min. Glucose transport was determined at 37°C by incubation with 50 M 2-deoxyglucose uptake containing 0.5 Ci of 2-[ 3 H]deoxyglucose in the absence or presence of 10 M cytochalasin B. The reaction was stopped after 4 min by washing the cells 3 times with ice-cold PBS. The cells were solubilized in 1% Triton X-100 at 37°C for 30 min, and aliquots were subjected to scintillation counting or protein assay to calculate uptake as pmol/min/mg protein.
PI 3-Kinase Assay-Phosphatidylinositol 3-OH kinase activity was measured in IRS-1 immune complexes obtained from whole cell detergent lysates as described above. The detergent-soluble cell extract was incubated overnight with anti-IRS-1 pre-loaded on protein A-conjugated Sepharose. Immune complexes were washed and subjected to the in vitro PI 3-kinase assay as described previously using bovine phosphatidylinositol (Sigma) and [␥-32 P]ATP as substrates (44). Phosphorylated lipids were separated by thin layer chromatography, visualized by autoradiography, and quantified by collecting and counting radiolabeled phosphatidylinositol 3-phosphate.
Adipsin Secretion-Adipsin secretion into cell culture media was measured as described previously (45). Thirty five-mm dishes of 3T3-L1 adipocytes were treated without or with 33 mM nocodazole for 3 h in serum-free Ham's F-12 media. The cells were washed 2 times with fresh media, followed by replacement with 1 ml of Ham's F-12 media without or with 33 M nocodazole and/or 100 nM insulin for 30 min. The media were removed, and proteins were precipitated in the presence of 10% trichloroacetic acid and deoxycholate as carrier (46). Pelleted proteins were captured by microcentrifugation and resuspended in 40 l of 0.25 M Tris base. Protein concentration was determined using a Bradford assay. Equivalent protein aliquots were fractionated on 10% SDS-PAGE and transferred to PVDF. Membranes were immunoblotted using an adipsin antibody (provided by Dr. Jess Miner, University of Nebraska).
Microtubule Binding Assay-Microtubules and associated proteins were polymerized in vitro as described previously (47). Briefly, 3T3-L1 adipocytes (day 8 -10 post-differentiation) were serum-starved in Ham's F-12 media for 3 h and then treated without or with 100 nM insulin for specified times. Cells (from two 100-mm plates per treatment) were scraped into 500 l of microtubule stabilizing buffer (5 mM MgSO 4 , 5 mM EGTA, 35 mM PIPES, pH 7.1, and 142 mM sucrose) containing 1 mM dithiothreitol, protease inhibitors (10 g/ml leupeptin, 10 mg/ml aprotinin, 0.5 mM benzamidine, 5 g/ml pepstatin A, 0.5 mM PMSF), and phosphatase inhibitors (100 mM NaF, 1 mM sodium vanadate, and 1 mM molybdate). Cells were homogenized at 4°C by passage through a 26-guage needle. Homogenates were centrifuged at 100 ϫ g for 10 min to obtain the post-nuclear supernatant. The post-nuclear supernatant was subjected to a 40,000 ϫ g centrifugation using an SW60 rotor (Beckman) for 1 h. The supernatant was recovered, and protein concentration was adjusted to 4 mg/ml with homogenization buffer. Taxol (1 mM stock solution in Me 2 SO) was added to a final concentration of 20 M, and samples were incubated at 4°C for 1 h. Polymerized microtubules and associated proteins were pelleted by centrifugation at 16,000 ϫ g for 30 min using an SW60 rotor (Beckman). In some experiments 1% Triton X-100 was added to the 40,000 ϫ g supernatant prior to addition of taxol to solubilize membranes. The 16,000 ϫ g taxolstabilized pellets were resuspended in homogenization buffer and protein concentration measured by Bradford assay. The 16,000 ϫ g taxolstabilized pellets were analyzed by SDS-PAGE and transferred to PVDF membrane and Western blotting.

Nocodazole Treatment Inhibits Insulin-mediated GLUT4
Translocation-To determine whether an intact microtubule cytoskeleton is necessary for insulin-mediated GLUT4 translocation, we measured GLUT4 translocation in cells treated with 33 M nocodazole. This concentration depolymerizes greater than 95% of microtubules in other cell types (43). To confirm that depolymerization of the microtubule network occurred in 3T3-L1 adipocytes under this treatment, we compared monomeric and polymeric tubulin levels in cells treated without or with 33 M nocodazole for 3 h by Western blot analysis. Monomeric tubulin was extracted in 0.1% Triton X-100 extraction buffer plus protease inhibitors at 37°C for 15 min. After the Triton X-100-soluble fraction was removed, polymeric tubulin was extracted with buffer containing 0.4 M NaCl and 0.5% SDS. An equivalent portion of each extract was fractionated by SDS-PAGE, and samples were probed with an anti-␤-tubulin antibody (Fig. 1A). Nocodazole treatment shifted greater than 95% polymeric tubulin to the monomeric form (Fig. 1A, compare lanes 2 and 4). In control cells, approximately two-thirds of the total tubulin was polymeric, and the remaining third was monomeric (Fig. 1A, compare lanes 1 and 2).
After establishing the efficacy of nocodazole treatment for depolymerization of the microtubule network in 3T3-L1 adipocytes, we examined insulin-mediated GLUT4 translocation under identical treatment conditions. Insulin-dependent GLUT4 translocation to the plasma membrane was assessed by measuring GLUT4 levels in purified plasma membrane fractions prepared by sonication of monolayers of 3T3-L1 adipocytes (23,48). Quantification of protein levels in this fraction by Western blot analysis provides an estimate of the accumulation of GLUT4 in the plasma membrane of an entire plate of cells. Cells were treated for 3 h with 33 M nocodazole followed by treatment without or with 100 nM insulin for 30 min. Isolated plasma membrane fragments were solubilized and analyzed by Western blot. Basal levels of GLUT4 in plasma membrane sheets were not affected by nocodazole treatment (Fig. 1B,  lanes 3 and 4). In contrast, nocodazole inhibited recruitment of GLUT4 to the plasma membrane following insulin treatment (p Ͻ 0.005) (Fig. 1B, lanes 7 and 8). Quantification of four independent experiments by laser densitometry is shown in Fig. 1C.
In the previous experiment, insulin-mediated GLUT4 translocation was measured after 3 h of nocodazole treatment; however, a time course experiment revealed that complete depolymerization of microtubules in 3T3-L1 cells occurred between 15 and 30 min of nocodazole treatment (data not shown). To con-firm that the effects of nocodazole on insulin-mediated GLUT4 translocation occurred at early time points paralleling microtubule depolymerization, we measured GLUT4 translocation after 20 min of nocodazole treatment ( Fig. 2A). Cells were treated with nocodazole for 20 or 120 min prior to insulin treatment and plasma membrane sheets prepared as described above. Results of two independent time course experiments are shown in Fig. 2A. Densitometric analysis of the Western blot reveals a 4-fold increase in GLUT4 protein following insulin treatment in the absence of nocodazole ( Fig. 2A, lanes 1-4). In contrast to control conditions, the insulin-mediated GLUT4 translocation was increased 1.4-and 1.6-fold, respectively, in cells treated for 20 and 120 min with nocodazole ( Fig. 2A, lanes  5-12). Thus, the effects of nocodazole on insulin-mediated GLUT4 translocation do not likely results from progressive changes in the cell that develop overtime in the presence of nocodazole.
Nocodazole Treatment Does Not Affect Proximal Insulin Receptor Signaling-Nocodazole treatment of cells significantly reduces microtubule-based vesicle transport; however, it has also been shown to elicit effects on the cell including inhibition of protein synthesis, disruption of plasma membrane subdomains, and changes in some types of cellular signaling (49). Any of these perturbations could lead to changes in early insulin signaling events. To determine whether nocodazole interferes with GLUT4 translocation by decreasing steady state levels of relevant proteins, we measured the levels of insulin receptor, IRS-1, and GLUT4 protein in nocodazole-treated cells. 3T3-L1 adipocytes were incubated with 33 M nocodazole for 3 h and then treated without or with 100 nM insulin for 20 min at 37°C. Nocodazole did not alter steady state protein levels of IRS-1, insulin receptor ␤-subunit, or GLUT4 protein (data not shown).
The first event in insulin signaling is the activation of the intrinsic tyrosine kinase activity of the insulin receptor upon ligand binding. This leads to both mitogenic and metabolic effects of the cascade. Another early event in insulin signaling is the tyrosine phosphorylation of intracellular insulin receptor substrates of the IRS family (50). Phosphotyrosine residues on the IRS isoforms act as binding sites for specific SH2 domain molecules, including phosphatidylinositol 3-OH kinase (PI 3-kinase) (50,51). Activation of PI 3-kinase is required for most, if not all, intracellular actions of insulin, including GLUT4 translocation (52). In both rat adipocytes and 3T3-L1 adipocytes, PI 3-kinase binds predominantly to IRS-1 (53,54). The PI 3-kinase activity immunoprecipitated with either anti-IRS-1 or anti-phosphotyrosine antibodies is increased in both intracellular membranes and cytosol following insulin stimulation (53)(54)(55)(56).
Tyrosine-phosphorylated proteins were solubilized in 1% Nonidet P-40 and cell lysates and then analyzed by immunoblotting with anti-phosphotyrosine antibody. Insulin treatment increased tyrosine phosphorylation of proteins corresponding in size to the ␤-subunit of the insulin receptor and IRS-1. Insulin-mediated tyrosine phosphorylation of these proteins was not affected by nocodazole treatment (Fig. 3A).
To determine whether immediate downstream signaling events other than tyrosine phosphorylation were modified by nocodazole treatment, we assayed the IRS-1-associated PI 3-kinase activity. Control and nocodazole-treated cells were treated without or with 100 nM insulin for 30 min. IRS-1 complexes were immunoprecipitated from 1% Nonidet P-40 cell extracts and assayed for PI 3-kinase activity under standard conditions (48). IRS-1-associated PI 3-kinase activity increased 16-fold under insulin treatment (Fig. 3B). Nocodazole treatment did not significantly alter PI 3-kinase activity associated with IRS-1 (p ϭ 0.7) in either basal or insulin-treated cells. These data indicate that an intact microtubule system is not required for insulin-dependent association of PI 3-kinase with phosphorylated IRS-1.
Insulin treatment activates protein kinase B (PKB) activity through a PI 3-kinase-dependent mechanism (57). The pleckstrin homology domain on the N terminus of PKB interacts with the phospholipid product of PI 3-kinase (PI-3,4,5-P 3 ). This interaction induces a conformational change in PKB that permits phosphorylation of PKB at two regulatory sites (Thr-308 and Ser-473), activating the kinase (58). To determine whether nocodazole treatment interfered with activation of PKB, we measured phosphorylation of PKB at serine 473 in cell lysates using a phospho-PKB-specific antibody recognizing the phosphoserine. The ratio of phospho-PKB to total PKB was calculated from the densitometric analysis of the Western blots. The ratio in control cells for three independent experiments was 112 Ϯ 1.2 (mean Ϯ S.D.) compared with a ratio 92 Ϯ 6.9 in nocodazole-treated cells (Fig. 3C). This 18% reduction in the ratio of phospho-PKB to total PKB was statistically significant (p Ͻ 0.02); however, the physiologic significance of this decrease is not certain. Nocodazole treatment did not prevent insulin-mediated phosphorylation of PKB at Ser-473.
Nocodazole Does Not Effect Insulin-mediated Adipsin Secretion or GLUT1 Translocation-To determine whether depolymerization of the microtubule network interferes specifically with insulin-mediated GLUT4 translocation, we examined other insulin-mediated exocytosis pathways that are independent of the GLUT4 pathway. First, we examined the effects of nocodazole treatment on insulin-mediated secretion of adipsin. Recently, two groups have shown that insulin stimulates a 2-3-fold increase in adipsin secretion through an ARF6-dependent pathway that passes through an endosomal intermediate on its way to the plasma membrane (34,45). We measured adipsin secretion into culture media over a 30-min incubation period in control and nocodazole-treated cells with- out or with 100 nM insulin (Fig. 4). Quantification of three independent experiments is shown in Fig. 4B. Insulin treatment caused a 3-fold increase in adipsin secretion in both control and nocodazole-treated cells, indicating that an intact microtubule network is not required for this process.
GLUT1 is also recruited to the plasma membrane in response to insulin (59,60). Two lines of evidence suggest that the GLUT1 pathway is independent of the GLUT4 pathway. First, intracellular vesicles containing GLUT1 are largely distinct from vesicles containing GLUT4 (59,60). Second, accumulation of GLUT1 in the plasma membrane is not dependent on the syntaxin 4 SNARE protein, which is required for GLUT4 translocation (23). To determine whether insulin-dependent GLUT1 translocation was inhibited by nocodazole treatment, the plasma membrane sheets probed for GLUT4 (see Fig. 1B) were next probed for GLUT1 (Fig. 5A). In contrast to its effect on GLUT4, nocodazole treatment had no effect on insulinmediated GLUT1 translocation.
Taxol Treatment Inhibits Insulin-mediated GLUT4 Translocation-It is possible that the effect of nocodazole on GLUT4 translocation was independent of its effects on the microtubule cytoskeleton, but rather due to some unrelated cytotoxic effect. To test this, we treated the cells with 12 M taxol to stabilize the microtubule network. In vivo polymerization of microtubules by taxol has been shown to inhibit vesicular exocytosis in FIG. 3. Nocodazole treatment does not affect phosphorylation of the IRS-1 or insulin receptor ␤-subunit. Cells were stimulated with 100 nM insulin for 30 min and extracted with 1% Nonidet P-40 lysis buffer. Lysates were fractionated on SDS-PAGE, transferred to PVDF membranes, and probed using an anti-phosphotyrosine antibody (A). Nocodazole does not inhibit PI 3-kinase activity associated with IRS-1. PI 3-kinase activity was assayed in IRS-1 immunoprecipitated complexes following 30 min of insulin treatment in control (Con) and nocodazole (Noc)-treated cells. PI 3-kinase products were separated by thin layer chromatography. The band corresponding to PI 3-phosphate was cut and quantitated by scintillation counting. An autoradiogram experiment is shown, and the histogram shows quantitation of 4 experiments (mean and S.D.) (B). Phosphorylation of PKB was measured using a phosphoserine-specific antibody against phosphorylated Ser-473. Total PKB protein was blotted in the same lysate. Cells lysates were prepared as described above. Results of 3 independent cell lysates are shown. IB, immunoblot (C). other systems (36,49). We reasoned that if a dynamic microtubule network is required for GLUT4 exocytosis, taxol treatment should also inhibit this process. To show that taxol treatment polymerized the microtubule network in 3T3-L1 cells, we compared monomeric and polymeric tubulin in cells treated without or with 12 M taxol (Fig. 6A). Taxol treatment resulted in a complete conversion of Triton X-100 extractable tubulin (monomeric) to polymeric tubulin (Fig. 6A, compare lanes 1 and  2). Plasma membrane sheets were prepared from control and taxol-treated cells that were stimulated without or with 100 nM insulin for 20 min. Accumulation of GLUT4 in response to insulin was inhibited by 40% in plasma membrane sheets prepared from taxol-treated cells (p Ͻ 0.01) (Fig. 6B, compare  lanes 5 and 6 with lanes 7 and 8). Quantitation of three independent experiments by laser densitometry is shown in Fig.  6C.
GLUT4 Vesicles Associate with Microtubules in Vitro-Although our examination of insulin-signaling pathways was not exhaustive, it was sufficient to conclude that nocodazole treatment did not affect proximal insulin signaling events at the cell surface or decrease steady state levels of proteins known to be involved in proximal insulin signaling. A plausible explanation consistent with our observations is that destabilization of the microtubule network interferes with vesicle translocation and/or with later events in insulin signaling that lead to translocation. To begin to examine these possibilities, we performed an in vitro microtubule binding assay (47) to determine whether GLUT4 vesicles associate directly with microtubules.
Tubulin was polymerized in vitro by addition of taxol to a post-40,000 ϫ g supernatant from 3T3-L1 cells homogenized in microtubule stabilization buffer without detergent. The 40,000 ϫ g supernatant was obtained by centrifuging the postnuclear supernatant at 40,000 ϫ g, and this fraction consisted of low density vesicles including GLUT4 vesicles and cytosolic proteins including tubulin dimers (47,56). Addition of 20 M taxol at 4°C polymerized ϳ60% of tubulin (data not shown). Polymerized tubulin was pelleted from the taxol-treated 40,000 ϫ g supernatant by slower speed centrifugation at 16,000 ϫ g. Tubulin appeared in the 16,000 ϫ g pellets only in the presence of taxol (Fig. 7, lane 1). The 16,000 ϫ g supernatant (obtained from the 16,000 ϫ g spin of the 40,000 ϫ g supernatant) (Fig. 7, lane 5) depicts a proportional loading of the supernatant from samples not treated with taxol to indicate the level of each protein in the 40,000 ϫ g supernatant prior to taxol treatment. The 16,000 ϫ g pellet was probed for associated proteins by Western blot analysis. IRS-1, GLUT4, and the insulin-responsive aminopeptidase (IRAP), a protein that exclusively colocalizes with GLUT4 vesicles (61), were associated with the 16,000 ϫ g pellets, and this association was dependent on taxol (Fig. 7, lane 2). The ␤-subunit of the insulin receptor was not detected in this pellet even though insulin receptor was present in the 40,000 ϫ g supernatant (Fig. 7, compare lanes 2  and 5). Insulin receptor was not present the 16,000 ϫ g pellets, demonstrating that vesicles containing the insulin receptor do not bind microtubules. Together, these results indicate that our assay for vesicle binding to polymerized microtubules is specific.
Insulin treatment did not affect the polymerization of tubulin; however, IRS-1, GLUT4, and IRAP bindings to polymerized tubulin were decreased in response (Fig. 7, lanes 3 and 4). The decrease in GLUT4 and IRAP binding to polymerized microtubules in response to insulin is consistent with data indicating a flux from intracellular membrane compartments to the plasma membrane. In contrast to GLUT4, the abundance of IRS-1 in the 40,000 ϫ g supernatant was the same in basal and insulintreated cells (data not shown). Thus, the decreased association of IRS-1 with polymerized tubulin from the insulin-treated 40,000 ϫ g supernatant indicates that insulin treatment decreases the affinity of IRS-1 for microtubules.
To determine whether microtubule-associated IRS-1 was phosphorylated in response to insulin, the 16,000 ϫ g pellets and 16,000 ϫ g supernatant from cells were probed with antiphosphotyrosine antibody. IRS-1 appearing in the 16,000 ϫ g pellets was tyrosine-phosphorylated (Fig. 8, lanes 4 and 5) following insulin treatment. Phosphorylated IRS-1 levels were similar in the 16,000 ϫ g pellets and the 16,000 ϫ g supernatant (Fig. 8, compare lanes 4 and 5 with lanes 7 and 8), indicating that phosphorylation state of IRS-1 does not affect its binding to polymerized tubulin.
Characterization of GLUT4 Vesicle Binding to Microtubules in Vitro-In vitro microtubule binding was performed in a detergent-free cell homogenate; it was therefore possible that the association of GLUT4 vesicles with microtubules was mediated either directly through GLUT4, through IRAP, or through an undetermined vesicle-associated protein. To determine whether GLUT4 or IRAP bind to microtubules through direct protein interactions, we performed the microtubule bind- ing assay in the 40,000 ϫ g supernatant after adding 1% Triton X-100 to solubilized membranes. Addition of detergent to the 40,000 ϫ g supernatant had no effect on taxol-induced polymerization of tubulin (Fig. 9, lanes 2 and 4). Detergent also had no effect on IRS-1 binding to polymerized tubulin, indicating that IRS-1 binds microtubules through a protein-protein interaction. Contrasting with IRS-1, detergent treatment of the 40,000 ϫ g supernatant abolished binding of GLUT4 and IRAP to polymerized tubulin (Fig. 9, lanes 2 and 4) indicating that the association of both GLUT4 and IRAP with microtubules depends on the continuity of the GLUT4/IRAP vesicle membrane. This experiment demonstrates that GLUT4 vesicles bind to microtubules via a vesicle-associated protein, and this intermediate protein is not GLUT4 or IRAP. Specificity of protein binding to microtubules is also confirmed by this experiment since IRSl-1, but not GLUT4 or IRAP, binds microtubules in the presence of detergent. DISCUSSION Insulin stimulates the translocation of an intracellular storage pool of vesicles that contain the GLUT4 facilitative glucose transporter to the cell surface. Subcellular fractionation studies have biochemically defined the insulin-responsive GLUT4 storage pool (62,63). Immunocytochemical analysis of the GLUT4 storage vesicles in brown adipose tissue indicates that GLUT4 is localized to tubulo-vesicular structures near the trans-Golgi network, as well as discrete vesicles distributed throughout the cytoplasm (19). The observation that GLUT4 storage vesicles are located at a distance from the plasma membrane, rather than docked and awaiting fusion, implies that an insulin-induced signal triggers the movement of storage vesicles to the plasma membrane after which fusion occurs. Movement of intracellular vesicles to the plasma membrane, as opposed to regulation membrane fusion, has recently been shown to be a rate-limiting step in insulin-stimulated GLUT4 translocation (32). It is of considerable interest to understand movement of GLUT4 vesicles in response to insulin at the molecular level. Our studies indicate that the microtubule cytoskeleton plays a fundamental role in this process.
We demonstrated that destabilization of the microtubule cytoskeleton completely inhibits GLUT4 translocation to the plasma membrane. This observation strongly implies that recruitment of GLUT4 vesicles to the plasma membrane surface requires an intact microtubule network. A reasonable assumption, based on known functions of microtubules in other secretory pathways, is that microtubules act as a filament upon which the GLUT4 vesicle moves to its cell surface location (36). Movement of vesicles between cellular compartments can occur solely by membrane fusion and fission events, but an acute response to insulin utilizing regulated fusion would require that vesicles be predocked at the target membrane. This does not appear to be the case with GLUT4 vesicles, since insulin stimulates the recruitment of GLUT4 vesicles from varied locations within the cell.

FIG. 7. Polymerized microtubules bind GLUT4 vesicles and IRS-1 in vitro.
A 40,000 ϫ g supernatant was derived from the postnuclear supernatant of a cell homogenate from cells treated with 100 nM insulin for 0, 5, or 20 min. This cellular fraction, which contained low density microsomal membranes and cytosolic proteins, was treated without or with 20 M taxol to polymerize tubulin. The polymerized tubulin and associated proteins were pelleted by slow speed centrifugation (16,000 ϫ g). The pellet (16,000 ϫ g pellets (16 K P)) and a sample from the supernatant of the 16,000 ϫ g pellet that was not treated with taxol (16,000 ϫ g supernatant (16 K SN)) were analyzed by Western blot. The 16,000 ϫ g pellets were probed for GLUT4 and IRAP, insulin receptor ␤-subunit, IRS-1, and ␤-tubulin. IB, immunoblot.
FIG. 9. Detergent solubilization of the 40,000 ؋ g supernatant prevents GLUT4 and IRAP binding to polymerized microtubules in vitro. The microtubule binding assay was performed as described in Fig. 6 except that the 40,000 ϫ g supernatant was treated with or without 0.1% Triton X-100 to solubilize membranes prior to addition of taxol. The 16,000 ϫ g pellets (16 K P) were blotted for GLUT4, IRAP, IRS-1, and ␤-tubulin. IB, immunoblot.
Disruption of the microtubule network is a profound alteration of a major cellular structure that may have global effects on cellular processes and vesicular trafficking. To control for this possibility, we studied other insulin-mediated vesicular trafficking pathways independent of GLUT4 exocytosis. Neither the insulin-mediated secretion of adipsin nor translocation of GLUT1 were found to be affected by disruption of the microtubule network. Thus, the microtubule network is required for a subset of membrane trafficking events in these cells, and this subset includes insulin-mediated GLUT4 exocytosis.
We demonstrated that GLUT4 vesicles specifically bind polymerized tubulin in vitro, implying that this interaction occurs in vivo. In the in vitro microtubule binding assay, we demonstrated that GLUT4 vesicle binding to microtubules was specific, since insulin receptor-containing vesicles did not associate with polymerized tubulin under conditions in which GLUT4 vesicles associate with polymerized tubulin (Fig. 9). Thus the microtubule network may act as a scaffold upon which GLUT4 storage vesicles are tethered awaiting insulin stimulation and/or a track upon which they travel after a signal has been transduced to the vesicles.
Our data suggest that the microtubule network may provide a link between the insulin-signaling pathway and GLUT4 vesicle translocation. Although we demonstrated that proximal signaling events through activation of PKB were not affected by destabilization of microtubules in vivo (with nocodazole treatment), it is possible that the pathway(s) of the insulinsignaling cascade leading specifically to GLUT4 vesicle recruitment is localized to microtubules. In support of this hypothesis, we observed that IRS-1 specifically bound to polymerized tubulin. Interestingly, recent work indicates that IRS-1 is bound to cytoskeletal elements, but the specific cytoskeletal components have not yet been identified (56). Our data suggest that microtubules may be an in vivo binding site for IRS-1.
Other studies have investigated the role of the actin cytoskeleton in insulin-mediated GLUT4 translocation. Disruption of the actin cytoskeleton using either cytochalasin D or latrunculin A partially inhibited GLUT4 translocation in 3T3-L1 cells (38) and completely disrupted this process in rat adipocytes (39). Although a direct association between GLUT4 vesicles has not yet been demonstrated, experimental evidence suggests that an intact actin cytoskeleton is required for recruitment of PI 3-kinase activity to the GLUT4 vesicle (38). It has not yet been established that recruitment of PI 3-kinase to GLUT4 vesicles is required for insulin-mediated GLUT4 translocation. In a recent study, it was shown that targeting of constitutively active PI 3-kinase to GLUT4 vesicles is not sufficient on its own to cause redistribution of GLUT4 vesicles to the plasma membrane (64); however, that does not prove that it is not necessary for translocation. It is of considerable interest to determine whether the microtubule cytoskeleton in addition to the actin cytoskeleton is involved in the insulin-mediated recruitment of PI 3-kinase to GLUT4 vesicles.
There are numerous potential roles for the microtubule network in insulin-mediated GLUT4 translocation. One possibility is that the microtubule network may provide a supporting structure that tethers intracellular GLUT4 vesicles in the basal state. Insulin signaling would release the vesicles allowing them to move to the plasma membrane by simple diffusion. If this were correct, complete depolymerization of the microtubule network would be expected to result in accumulation of GLUT4 at the plasma membrane in the basal state. We did not observe a change in plasma membrane GLUT4 levels in basal cells treated with nocodazole; however, we cannot preclude the possibility that GLUT4 vesicles were docked at the plasma membrane, but not fused, in the nocodazole-treated cells. Al-ternatively, the microtubule network may provide a supporting structure upon which GLUT4 vesicles travel to the plasma membrane. It is possible that GLUT4 vesicles are tethered to microtubules in vivo in the basal state and that insulin promotes association of the GLUT4 vesicle with a microtubule motor protein. Support of this model requires the identification of a microtubule motor protein that is regulated by insulin and associates with GLUT4 vesicles. There are numerous examples of vesicular trafficking along microtubule filaments. For instance, the regulation of bile secretion is a regulated exocytotic process in liver cells that has been shown to require an intact microtubule system (65). This system resembles GLUT4 translocation in that the bile-containing vesicles are recruited to the plasma membrane upon stimulation of cells with an appropriate ligand.
In the present study, we provide functional and biochemical data demonstrating that insulin-mediated GLUT4 translocation is dependent on the microtubule cytoskeleton in 3T3-L1 adipocytes. The role of the microtubule network in this process is under investigation. We demonstrated the GLUT4 vesicles bind polymerized tubulin in vitro, and this suggests that the microtubule network may act to scaffold and compartmentalize GLUT4 vesicles. Our data also suggest that the microtubule network may provide a structure for the insulin-signaling cascade to intersect directly with GLUT4 vesicles and stimulate their exocytosis. Further studies focusing on the role(s) of the microtubule cytoskeleton in insulin-mediated GLUT4 translocation are currently in progress.