Expression of a Dominant Interfering Dynamin Mutant in 3T3L1 Adipocytes Inhibits GLUT4 Endocytosis without Affecting Insulin Signaling*

To examine the role of clathrin-coated vesicle endocytosis in insulin receptor signaling and GLUT4 trafficking, we used recombinant adenovirus to express a dominant interfering mutant of dynamin (K44A/dynamin) in 3T3L1 adipocytes. Functional expression of K44A/dynamin, as measured by inhibition of transferrin receptor internalization, did not affect insulin-stimulated insulin receptor autophosphorylation, Shc tyrosine phosphorylation, or mitogen-activated protein kinase activation. Although the tyrosine phosphorylation of insulin receptor substrate-1 was slightly reduced, correlating with a 25% decrease in insulin receptor substrate-1-associated phosphatidylinositol 3-kinase activity, insulin-stimulated Akt kinase activation was unaffected. In contrast, expression of K44A/dynamin resulted in the cell-surface accumulation of GLUT4 under basal conditions and an inhibition of GLUT4 endocytosis without affecting insulin-stimulated GLUT4 exocytosis. These data demonstrate that disruption of clathrin-mediated endocytosis does not significantly perturb insulin receptor signal transduction pathways. Furthermore, K44A/dynamin expression causes an accumulation of GLUT4 at the cell surface, suggesting that GLUT4 vesicles exist in at least two distinct intracellular compartments, one that undergoes continuous recycling and a second that is responsive to insulin.

The insulin-responsive glucose transporter, GLUT4, 1 is expressed at high levels in adipose tissue, skeletal muscle, and cardiac myocytes (1)(2)(3)(4). Unlike other glucose transporter isoforms such as GLUT1, GLUT4 is primarily distributed to poorly defined intracellular compartments under basal conditions (5)(6)(7). GLUT4 slowly recycles between these intracellular compartments and the plasma membrane, with the majority of steady state GLUT4 residing in a vesicle population distinct from the endosomal compartments (8 -11). In response to insulin, plasma membrane GLUT4 is enhanced 10 -20-fold, an amount sufficient to account for the majority of insulin-stimulated glucose uptake, with a concomitant decrease in the low density microsome-associated GLUT4 protein (5,7,12,13).
Concurrent with the insulin-dependent increase in GLUT4 vesicle exocytosis, insulin also reduces the rate of GLUT4 endocytosis up to 3-fold (8 -10). Although the exact mechanism of this regulation is unknown, several lines of evidence suggest that GLUT4 internalization occurs via clathrin-coated vesicles. For example, disassembly of clathrin lattices by potassium depletion results in the accumulation of GLUT4 at the cell surface (14). Immunofluorescence and electron microscopy studies have localized GLUT4 to coated pits both in the basal and insulin-stimulated states (6,13,15). Furthermore, internalization of a chimeric transferrin receptor/GLUT4 protein was inhibited by acidification of the cytosol, a technique for blocking clathrin-coated vesicle endocytosis (16).
One protein important in the formation of clathrin-coated vesicles is dynamin, a 100-kDa GTPase implicated in the membrane scission step of vesicle formation (17,18). Three different dynamin genes have been identified with each having multiple splice variants (19). These isoforms share a high degree of homology in the amino-terminal GTPase domain but have distinct proline-rich carboxyl-terminal tails that have been shown in vitro to bind various SH3-containing proteins (20 -25). Expression of a dominant interfering form of dynamin I with a mutation in the conserved GTP-binding motif has been shown to prevent the constriction and budding of clathrin-coated vesicles (26,27). Under these conditions, the inhibition of clathrincoated vesicle formation interferes with the recycling of cellsurface receptors such as the transferrin receptor, the ␤-adrenergic receptor, and the epidermal growth factor receptor (26 -29).
In this study, we examined the requirement of dynaminmediated endocytosis for insulin-dependent signal transduction events in 3T3L1 adipocytes using adenoviral expression of a dominant interfering K44A/dynamin mutant. Our data demonstrate that inhibition of clathrin-mediated endocytosis does not impair either insulin receptor signaling or insulin-stimulated GLUT4 translocation. However, it does inhibit the clearance of GLUT4 from the cell surface, both in the basal state and following insulin stimulation, suggesting that dynamin may function in the regulation of GLUT4 endocytosis.

EXPERIMENTAL PROCEDURES
Materials-Chemical reagents were purchased from Sigma except protease inhibitors aprotinin, pepstatin, leupeptin, and Pefabloc purchased from Boehringer Mannheim, and as indicated. Antibodies were purchased from Transduction Laboratories (Lexington, KY) except polyclonal Akt1 (C20) antibody from Santa Cruz Biotechnology (Santa Cruz, CA), Akt2 antibody generously provided by Dr. Morris Birnbaum (Philadelphia, PA), phospho-Akt and phospho-p44/42 MAP kinase antibodies from New England Biolabs (Beverly, MA), and polyclonal antibody against rat GLUT4 (IA02) which was obtained by immunizing a rabbit with GLUT4 carboxyl-terminal peptide, SATFRRTPSLLEQ-* This work was supported in part by Research Grants DK33823 and DK25295 from the National Institutes of Health. 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 U.S.C. Section 1734 solely to indicate this fact.
‡ Recipient of a postdoctoral fellowship award from the Juvenile Diabetes Foundation.
Cell Culture and Reagents-3T3L1 adipocytes (American Type Tissue Culture, Rockville, MD) were grown and differentiated as described previously (30). Cells were maintained in Dulbecco's modified essential media (DMEM) with 10% fetal bovine serum for 1-2 days post-differentiation and then infected with recombinant adenovirus for 48 h, unless otherwise indicated. Prior to use, cells were washed 2 times with phosphate-buffered saline (PBS) and serum-starved in DMEM with 0.1% bovine serum albumin for at least 2 h.
Generation of Recombinant Adenovirus-The K44A/dynamin expressing recombinant adenovirus was constructed using the method of Becker et al. (31). The shuttle vector pACCMVpLpA and the modified adenovirus type 5 (Ad5) genome plasmid pJM17 were generously provided by Dr. Christopher Newgard (University of Texas, Southwestern, Dallas, TX). The cDNA for the K44A/dynamin (generously provided by Dr. Sandra Schmid, Scripps Institute, La Jolla, CA) was subcloned into pACCMVpLpA using 5Ј BamHI and 3Ј HindIII sites. The pACCMV-K44A/dynamin I and pJM17 were co-transfected into 80% confluent human embryo kidney (HEK) 293 cells by the calcium-phosphate method. Successful homologous recombination and production of Ad5-K44A/dynamin resulted in cell lysis within 14 -21 days. Lysates were collected, subjected to several freeze-thaw cycles, and single clones of recombinant adenovirus isolated through agarose overlay of serial dilutions. Midscale stocks of virus were generated by infecting 10-cm plates of subconfluent HEK293 cells with 15 l of the single plaque isolate lysate. Midscale stocks were used to infect subconfluent HEK293 cells for 1 h at 37°C. Virus was removed, and fresh media were replaced and the cells allowed to incubate for an additional 36 -48 h. Prior to complete lysis, the cells were harvested, subjected to several freeze-thaw cycles, and these concentrated lysates used for adipocyte infection as described previously (32).
Receptor Internalization-Transferrin receptor internalization was determined by measuring the amount of acid-dissociable prebound ligand as described by Lamb et al. (33). Thirty-five-mm dishes of infected 3T3L1 adipocytes were incubated with 125 I-transferrin (3 nM; 1 Ci/g, Amersham Pharmacia Biotech) at 4°C for 3 h. The media were changed to remove unbound ligand, and internalization was initiated by the addition of 37°C DMEM. At the indicated times, endocytosis was terminated by washing cells 3 times with ice-cold PBS. Acid-dissociable cell-surface ligand was determined by pooling consecutive 8-min washes with 1.5 ml and 1.0 ml of stripping buffer (0.5 M NaCl, 0.2 M acetic acid). Acid-washed cells were solubilized in 1% SDS, and the amount of internalized ligand was determined by quantitation of the amount of radioactivity using a Packard Auto-gamma 5000 Series gamma counter (Meriden, CT).
Shc was immunoprecipitated from whole cell lysates by incubation with 4 g of polyclonal Shc antibody for 2 h at 4°C. The resulting immune complexes were precipitated by incubation with protein A-Sepharose for 1 h at 4°C. Pellets were washed three times in wash buffer (25 mM Tris, pH 7.4, 150 mM NaCl, 0.1% Triton X-100, 50 mM NaF, 1 mM Na 3 VO 4 ), two times in Tris-buffered saline (25 mM Tris, pH 7.4, 150 mM NaCl), resuspended in gel loading buffer, and boiled at 100°C for 5 min.
Whole cell lysates or immunoprecipitates were separated on 10% SDS-polyacrylamide gels (SDS-PAGE), transferred to polyvinylidene fluoride membrane (Millipore Corp., Bedford, MA) or nitrocellulose membrane (Schleicher & Schuell, Dassel, Germany), and Western blotted as indicated. Dynamin II and IRS1 levels were quantitated using Adobe Photoshop and NIH Image software.
Phosphoinositol (PI) 3-Kinase and Akt Kinase Assays-PI 3-kinase activity was determined as described by Turinsky et al. (34). Briefly, cell lysates were immunoprecipitated with the IRS1 antibody followed by incubation with protein Gϩ-agarose. The immunoprecipitated lipid kinase activity was assessed by incubation with 40 Ci of [␥-32 P]ATP (Amersham Pharmacia Biotech) plus 20 g of phosphatidylinositol (Avanti Polar Lipids, Birmingham, AL) for 15 min at room temperature.
The phosphorylated lipids were separated by thin layer chromatography, visualized by autoradiography, and quantitated by scraping and counting the radiolabeled phosphoinositides.
Akt protein kinase activity was determined as described by Moule and Denton (35). Cell lysates were prepared and subjected to immunoprecipitation with a polyclonal Akt antibody. Protein kinase activity was assessed by incubation of the immunoprecipitates with 20 Ci of [␥-32 P]ATP and 0.5 mg/ml histone 2B (Boehringer Mannheim) for 20 min at 30°C. The extent of histone 2B phosphorylation was determined by separation on a 16% SDS-polyacrylamide gel, autoradiography, and quantitation by excision of the radiolabeled band followed by scintillation counting. The amount of 32 P incorporation was normalized for the amount of Akt immunoprecipitated as determined by densitometric scanning of Akt immunoblots.
GLUT4 Internalization Assays-Uninfected and adenovirus-infected 3T3L1 adipocytes were serum-starved in DMEM with 0.1% bovine serum albumin for 2 h and stimulated with 100 nM insulin for 30 min at 37°C. Insulin was removed, and cells were washed twice with acidic buffer (5 mM sodium acetate, pH 5.0, 140 mM sodium chloride), twice with PBS, and fresh starvation media were replaced. GLUT4 was allowed to internalize at 37°C for the indicated times, and then plasma membrane sheets were harvested by the method of Robinson et al. (13). Briefly, the cells were washed in ice-cold PBS and incubated for 30 s in ice-cold 0.5 mg/ml poly-L-lysine/PBS. The cells were then rinsed 3 times with 1/3ϫ KHMgE buffer (1ϫ concentration, 70 mM KCl, 30 mM HEPES, pH 7.5, 5 mM MgCl 2 , 3 mM EGTA), placed in 1ϫ KHMgE buffer with 1 mM DTT and 0.1 mM PMSF, and sonicated for 5-10 s using a microtip at setting 4.8 on a 550 Sonic Dismembrator (Fisher). For immunofluorescence, plasma membrane sheets were fixed in 1% paraformaldehyde (Electron Microscopy Science, Ft. Washington, PA) for 20 min at room temperature, placed into quench buffer (100 mM glycine in PBS), and washed 3 times in PBS. The membrane sheets were blocked for 30 min at room temperature in 5% donkey serum/PBS, incubated for 1 h with the GLUT4 antibody, followed by a second 1-h incubation with a LRSC-donkey anti-rabbit antibody. The plasma membrane sheets were washed, coverslipped with Vectashield (Vector Labs, Burlington, CA), and viewed on a Bio-Rad laser confocal microscope.
For immunoblotting, unfixed plasma membrane sheets were washed several times in KHMgE buffer and collected with a cell scraper into detergent-free Lysis Buffer (see above). Lysates were homogenized for 10 s on ice with a pellet pestle motor (Kontes, Vineland, NJ). The resultant membrane lysates were subjected to protein determination, placed into Laemmli sample buffer without heating, and immunoblotted as described above.

Dynamin Expression Is Increased during 3T3L1 Adipocyte
Differentiation-Differentiated 3T3L1 adipocytes represent an excellent model system to study GLUT4 trafficking since they share many of the characteristics of primary adipocytes, including movement of GLUT4 from intracellular stores to the plasma membrane in response to insulin (36,37). To examine the potential role of dynamin in insulin signaling and GLUT4 translocation, we initially examined the expression of dynamin during 3T3L1 adipocyte differentiation. By using a dynamin II-specific antibody, we observed a 6 -9-fold increase in dynamin expression during differentiation of 3T3L1 fibroblasts into adipocytes (Fig. 1A, lanes 1-4). The expression of clathrin heavy chain was not significantly affected by adipocyte differentiation (Fig. 1B, lanes 1-4). In contrast, GLUT4 expression was highly induced following 3T3L1 adipocyte differentiation (Fig. 1C, lanes 1-4).
Expression of K44A/Dynamin Inhibits Transferrin Receptor Endocytosis-Previous studies have demonstrated that expression of a GTPase-deficient dynamin such as K44A/dynamin I in mammalian cells inhibits clathrin-mediated endocytosis by functioning in a dominant interfering manner over endogenous dynamin I or dynamin II proteins, whereas wild-type dynamin has no effect (26,27,38). Due to the difficulty of quantitatively transfecting 3T3L1 adipocytes by standard methods, we generated a recombinant adenovirus encoding K44A/dynamin I (K44A) and infected cells for 48 h. Dynamin expression, as assessed by a dynamin I/II antibody, was markedly increased in K44A/dynamin adenovirus-infected 3T3L1 adipocytes compared with uninfected cells or cells infected with a LacZ-encoding adenovirus (LacZ) ( Fig. 2A, top panel, lanes 1-3). Immunoblotting of these extracts with a dynamin II-specific antibody demonstrated no significant change in endogenous dynamin expression ( Fig. 2A, middle panel, lanes 1-3). Since dynamin II expression was unchanged, the increase in dynamin I/II protein levels was due to the expression of K44A/dynamin I. Insulin receptor ␤ (IR␤) expression was also unchanged in K44A/ dynamin expressing cells compared with uninfected or LacZ ( Fig. 2A, bottom panel, lanes 1-3). Under these conditions, greater than 95% of the cell population was infected with adenovirus with no observable changes in cell morphology by light microscopy (data not shown).
Having demonstrated increased expression of K44A/dynamin, we next determined its effect on clathrin-mediated endocytosis. This was assessed by examining the internalization of the transferrin receptor in uninfected, LacZ-, and K44A/ dynamin-infected 3T3L1 adipocytes (Fig. 2B). In control adipocytes approximately 70% of the cell-surface transferrin recep-tor was internalized over a 60-min period (squares). Adenovirus infection slightly reduced the extent of transferrin receptor endocytosis, with approximately 60% internalized over the same time course (circles). In contrast, expression of K44A/ dynamin resulted in only 30% of the cell-surface transferrin receptors being internalized with a significant decrease in initial rate of internalization compared with uninfected or LacZinfected cells (triangles). Thus, increased expression of K44A/ dynamin results in an inhibition of transferrin receptor internalization and therefore functions in a dominant interfering manner to reduce coated vesicle endocytosis in 3T3L1 adipocytes.
Relationship between K44A/Dynamin Expression and Insulin Receptor Downstream Signaling-Several studies have suggested that endosome localized insulin receptor provides the major pathway leading to insulin signal transduction and biological responsiveness (39 -42). Since expression of K44A/dynamin in other cells lines has been shown to interfere with signaling events downstream of the epidermal growth factor and ␤-adrenergic receptors (28,29), we investigated the insulin receptor signaling characteristics in 3T3L1 adipocytes expressing K44A/dynamin. The dose response for insulin-stimulated insulin receptor ␤ subunit tyrosine autophosphorylation was similar in uninfected, LacZ-, and K44A/dynamin-expressing cells (Fig. 3A). In each case, autophosphorylation of the insulin receptor was not observed in the absence of or following stimulation with 0.1 nM insulin (Fig. 3A, lanes 1, 2, 6, 7, 11 and 12). Insulin receptor autophosphorylation in response to 1 nM insulin remained low (Fig. 3A, lanes 3, 8 and 13) but became more readily observable at 10 and 100 nM insulin (Fig. 3A, lanes 4, 5, 9, 10, 14 and 15). In contrast, tyrosine phosphorylation of IRS1 was significantly more sensitive to insulin, reflecting downstream amplification of insulin receptor kinase activity (Fig.  3B). Tyrosine phosphorylation of IRS1 was clearly discernible at 0.1 nM insulin (Fig. 3B, lanes 1, 2, 6, 7, 11 and 12) and increased in a dose-dependent manner up to 100 nM insulin (Fig. 3B, lanes 3-5, 8 -10 and 13-15). Relative to uninfected and LacZ-expressing cells, K44A/dynamin overexpression had only a minor effect on IRS1 tyrosine phosphorylation in response to varying doses of insulin (Fig. 3B, compare lanes 1-5 with lanes 6 -10 and lanes [11][12][13][14][15]. Tyrosine phosphorylation of the 52-kDa Shc adapter protein in response to insulin was also unaffected by K44A/dynamin expression (data not shown).
To examine downstream insulin receptor signaling pathways, we next determined the effect of K44A/dynamin expression on the ERK family of mitogen-activated protein (MAP) kinases (Fig. 4). Uninfected, LacZ-, and K44A/dynamin-expressing cells exhibited essentially identical insulin dose-response patterns when immunoblotted with a phospho-specific ERK1/ERK2 antibody (Fig. 4A, lanes 1-15). Similarly, the time dependence of insulin-stimulated ERK activation was unaffected by adenovirus-mediated expression of K44A/dynamin (Fig. 4B, lanes 1-18). These data are consistent with Shc tyrosine phosphorylation providing the predominant pathway for the insulin activation of ERK1 and ERK2.
In contrast to Shc, the IRS family of docking proteins is thought to provide the major pathways leading to the metabolic actions of insulin (43). As observed in Fig. 3, insulin stimulation of IRS1 tyrosine phosphorylation was slightly attenuated following expression of K44A/dynamin compared with uninfected and LacZ-expressing cells. To examine further the potential effect of decreased IRS1 tyrosine phosphorylation, we performed PI 3-kinase assays in IRS1 immunoprecipitates from LacZ-and K44A/dynamin-expressing cells. Following insulin stimulation, IRS1-associated PI-3 kinase was decreased 25% in K44A/dynamin-expressing cells relative to the LacZexpressing cells (Fig. 5, A and B).
Since maximal activation of the PI-3 kinase may not be required for full biological responsiveness to insulin, we further examined the ability of insulin to stimulate Akt kinase activity. Akt is a serine-threonine kinase involved in the metabolic actions of insulin whose activity appears to be dependent upon PI-3 kinase (44 -46). Akt activation is associated with phosphorylation on serine and threonine residues resulting in a characteristic decrease in SDS-polyacrylamide gel electrophoretic mobility. Relative to unstimulated cells, insulin caused an Akt gel shift in the uninfected and both LacZ-and K44A/dynamin adenovirus-infected cells (Fig. 6A). When insulin-stimulated Akt protein kinase activity was directly measured, no significant difference was found between the LacZ-and K44A/dynamin-expressing adipocytes (Fig. 6B). To confirm these findings, we also examined the insulin dose and time dependence of Akt activation by using a phosphoserine-specific Akt antibody and by Akt gel shift (Fig. 7). Under these conditions, uninfected, LacZ-, and K44A/dynamin-expressing cells exhibited similar patterns of Akt activation in response to increasing doses or times of insulin stimulation. Thus, expression of K44A/ dynamin does not have a significant effect on the major prox-imal insulin receptor signaling pathways examined.
K44A/Dynamin Expression Results in the Accumulation of GLUT4 at the Plasma Membrane-Morphological studies have demonstrated that in the basal state, the small amount of GLUT4 present at the cell surface is localized to coated pits (5,6). We therefore examined the effect of K44A/dynamin on the plasma membrane localization of GLUT4 by immunofluorescence microscopy of isolated plasma membrane sheets (Fig. 8).
In the absence of insulin, membrane sheets exhibited a low level of background GLUT4 immunofluorescence which markedly increased following insulin stimulation (Fig. 8, panels  1-3). Similarly, following 48 h of infection with the LacZ-encoding adenovirus, there was also a low level of plasma membrane-associated GLUT4 immunofluorescence in the basal  11-15). The cells were then serum-starved and incubated in the absence (lanes 1, 6, and 11) or presence of 0.1 (lanes 2, 7,  and 12), 1 (lanes 3, 8, and 13), 10 (lanes 4, 9, and 14) and 100 nM (lanes 5, 10, and 15) insulin for 5 min at 37°C. Whole cells lysates were generated, subjected to SDS-PAGE, transferred to nitrocellulose, and immunoblotted using an anti-phosphotyrosine antibody to visualize the autophosphorylation of the insulin receptor ␤ subunit (A) and IRS1 (B).  2 and 4) of 100 nM insulin for 5 min at 37°C, followed by cell lysis and IRS1 immunoprecipitation (IP). IRS1 immunoprecipitates were incubated with [␥-32 P]ATP plus 20 g of phosphatidylinositol. Reaction products were subjected to thin layer chromatography and visualized by autoradiography. B, chromatography plates were scraped, and radioactivity was determined by gamma counting. PIP, phosphatidylinositol phosphate; ori, origin; C, control; I, insulin. state that dramatically increased in response to insulin (Fig. 8,  panels 4 -6). Surprisingly, expression of K44A/dynamin resulted in a significant increase in basal plasma membraneassociated GLUT4 immunofluorescence, although the cells remained responsive to insulin stimulation (Fig. 8, panels 7-9).
To confirm these findings, 3T3L1 adipocytes were allowed to express K44A/dynamin for various times, and the relative amount of plasma membrane-associated GLUT4 was determined ( Fig. 9). K44A/dynamin adenovirus infection resulted in a time-dependent accumulation of GLUT4 at the plasma membrane (Fig. 9, panels 1-3). After 24 and 48 h of K44A/dynamin expression, insulin was still capable of inducing GLUT4 translocation to the plasma membrane (Fig. 9, panels 4 and 5). Following 72 h of K44A/dynamin expression, there was an additional elevation of basal plasma membrane-associated GLUT4. Although at 72 h insulin stimulation was able to increase further cell-surface GLUT4, the fold increase was lower due to the elevated basal plasma membrane GLUT4 (Fig.  9, panels 3 and 6).
Expression of K44A/Dynamin Inhibits GLUT4 Endocytosis-Since expression of K44A/dynamin did not appear to alter insulin-stimulated GLUT4 translocation (Figs. 8 and 9), the increased basal amount of GLUT4 at the plasma membrane was most likely due to an inhibition of GLUT4 endocytosis. To address this issue, we examined the rate of GLUT4 internalization in uninfected, LacZ, and K44A/dynamin-infected 3T3L1 adipocytes (Fig. 10). Insulin stimulation resulted in a marked translocation of GLUT4 to the plasma membrane in uninfected cells (Fig. 10, panels 1 and 2). Following removal of insulin, there was a time-dependent decrease in plasma membraneassociated GLUT4 indicative of endocytosis (Fig. 10, panels 3  and 4). Similarly, insulin stimulation of LacZ-expressing adi-pocytes caused an increase in cell-surface GLUT4 which decreased subsequent to removal of insulin (Fig. 10, panels 5-8).

FIG. 8. Overexpression of K44A/dynamin does not affect insulin-stimulated GLUT4 exocytosis. Uninfected (Uninf) cells (panels 1-3) or cells infected for 48 h with adenovirus encoding LacZ (panels
K44A/dynamin inhibits GLUT4 endocytosis and are consistent with the continuous recycling of GLUT4 to and from the plasma membrane even in the absence of insulin. To assess further the relative levels of cell-surface GLUT4, isolated plasma membrane sheets were subjected to GLUT4 immunoblotting (Fig. 11). In uninfected, control 3T3L1 adipocytes, little GLUT4-immunoreactive protein was detected in isolated plasma membrane sheets (Fig. 11, lane 1). Thirty min of insulin stimulation resulted in a dramatic increase in the amount of plasma membrane-associated GLUT4 which recovered to nearly basal levels 2 h after insulin removal (Fig. 11,  lanes 2 and 3). Similarly, plasma membrane sheets from cells infected with the LacZ-encoding virus contained low levels of basal GLUT4 that increased following insulin stimulation (Fig.  11, lanes 4 and 5). The amount of plasma membrane GLUT4 also significantly decreased 2 h after insulin removal, although the effect was not as dramatic as in the uninfected cells (Fig.  11, lane 6). Even though expression of K44A/dynamin did not affect the insulin-stimulated translocation of GLUT4 to the cell surface, we observed an inhibition of GLUT4 endocytosis from the plasma membrane subsequent to insulin removal (Fig. 11,  lanes 7-9).
K44A/Dynamin Does Not Affect GLUT4 Expression or Prevent Insulin Receptor Inactivation-In addition to the specific inhibition of dynamin-dependent endocytosis, it remained possible that expression of K44A/dynamin resulted in increased expression of GLUT4 protein and/or persistent insulin signaling following insulin removal. To examine the effect of K44A/ dynamin expression on GLUT4 protein levels, whole cell extracts were immunoblotted for GLUT4 (Fig. 12A). Expression of K44A/dynamin had no significant effect on the cellular content of GLUT4 protein compared with uninfected and LacZexpressing cells (Fig. 12A, lanes 1-3). To assess the possibility of persistent insulin signaling, whole cell lysates from basal, insulin-stimulated, and insulin-stimulated cells in which the insulin was removed were subjected to phosphotyrosine immunoblotting (Fig. 12B). In both the LacZ and K44A/dynamin cells, insulin stimulation resulted in equal amounts of insulin receptor (IR␤) and IRS1 tyrosine phosphorylation (Fig. 12B,  compare lanes 1 and 2 with lanes 4 and 5). Following insulin removal, the LacZ-and K44A/dynamin-expressing adipocytes demonstrated a parallel loss of insulin receptor and IRS1 tyrosine phosphorylation (Fig. 12B, lanes 3 and 6). Thus, the retention of GLUT4 at the plasma membrane of K44A/dynamin-expressing cells was not due either to increased expression of the GLUT4 protein or prolonged insulin receptor activation. DISCUSSION Since its identification as the insulin-responsive glucose transporter, the mechanism and regulation of GLUT4 trafficking have been areas of intense investigation. GLUT4 expression is predominantly limited to striated muscle and adipose tissue (1)(2)(3)(4)47). In addition, GLUT4 vesicles undergo regulated exocytosis in response to a number of stimuli including insulin, exercise, and osmotic shock (7,48,49). Although the precise biochemical pathway by which GLUT4 is recruited from intracellular vesicles has not been fully delineated, this vesicle pool must be replenished following each round of insulin stimulation and withdrawal. In order to replenish intracellular GLUT4 stores after post-prandially elevated circulating insulin has declined, GLUT4 must be removed from the plasma membrane and sorted back into the insulin-responsive vesicle pool.
Specific sequence motifs in the amino-and carboxyl-terminal domains of GLUT4 have been shown to be responsible for directing appropriate trafficking of the GLUT4 protein. The carboxyl-terminal domain appears to function as a sequestration signal directing and/or inducing retention of GLUT4 to the insulin-responsive vesicle pool (50 -53). In contrast, the aminoterminal FQQI motif appears to be necessary for the internalization of the plasma membrane-localized GLUT4 although the carboxyl-terminal dileucine motif may also contribute to GLUT4 endocytosis (53)(54)(55)(56). Although the major effect of insulin is to increase the rate of GLUT4 exocytosis, insulin also reduces the rate of GLUT4 endocytosis suggesting that these two pathways are coordinately regulated (8 -10).
One molecule implicated in endocytic coated-vesicle formation is the GTPase dynamin (for reviews, see Refs. 57 and 58). In addition to its highly conserved amino-terminal GTPase domain, dynamin contains a central pleckstrin homology domain and a carboxyl-terminal proline-rich region which negatively and positively regulate the dynamin self-assembly and GTPase activities, respectively (59). These domains can functionally interact with several signaling intermediates, including Grb2 (21)(22)(23)60), Shc (61), inositol phospholipids (24), and amphiphysin (25). In this regard, insulin has been observed to induce the association of a dynamin-Grb2 complex with tyrosine-phosphorylated IRS1, the binding of which can modulate the rate of dynamin GTPase activity in vitro (21). Together, these data suggest insulin may regulate the endocytic rate of constitutively recycling cell-surface proteins such as GLUT4 through a dynamin-dependent mechanism.
Previous studies have demonstrated that expression of the GTPase-defective K44A/dynamin I mutant in non-neuronal cells blocks coated vesicle formation and budding by functioning in a dominant, interfering manner over endogenous dynamin II (27,38). These same studies have shown that expression of wild-type dynamin I has no effect on steady state rates of endocytosis. Since dynamin assembles into tetrameric complexes (59), overexpressed K44A/dynamin I likely interferes with endogenous dynamin II function by formation of heterotetrameric complexes. Consistent with previous studies, we have observed in 3T3L1 adipocytes that overexpressing K44A/ dynamin I interferes with endogenous dynamin II functions by assessing transferrin receptor endocytosis.
Based upon the kinetics and degree of insulin receptor kinase activation and subcellular localization, others (40 -42, 62) have suggested that endosome-localized insulin receptor is the major site of insulin receptor signal transduction and biological responsiveness. Likewise, inhibition of epidermal growth factor receptor endocytosis by expression of K44A/dynamin I has been reported to interfere with several specific signaling pathways, whereas wild-type dynamin I had no effect (29). In contrast, insulin receptor autophosphorylation and IRS1 tyrosine phosphorylation still occur when insulin receptor endocytosis was inhibited by incubation of cells at 4°C, suggesting that plasma membrane-localized insulin receptor was sufficient for these proximal signaling events (62,63).
We therefore directly assessed the ability of insulin to activate several signaling pathways in K44A/dynamin-expressing 3T3L1 adipocytes. Our data demonstrate that with the exception of a small but probably physiologically irrelevant decrease in IRS1-associated PI 3-kinase activity, inhibition of dynamindependent endocytosis does not affect insulin-stimulated signal transduction pathways leading to either Akt or MAP kinase activation. In addition, we have demonstrated that expression of a dominant interfering dynamin mutant does not affect the dephosphorylation and inactivation of the insulin receptor kinase or IRS1 following insulin removal. These data are consistent with a recent study demonstrating that insulin receptor signaling in the rat hepatoma H4IIE cell line is largely independent of insulin receptor internalization (64). However, these findings are in disagreement with epidermal growth factor receptor signal transduction events where expression of K44A/dynamin was observed to enhance DNA synthesis and tyrosine phosphorylation of Shc and phospholipase C␥ but to inhibit ERK1 and phosphatidylinositol 3-kinase activation (29). We are currently unable to explain this discrepancy, but it may reflect a fundamental difference in the specificity of insulin receptor versus epidermal growth factor receptor signaling mechanisms.
Nevertheless, consistent with the presence of functional signal transduction pathways, inhibition of clathrin-mediated endocytosis did not affect the ability of insulin to induce GLUT4 translocation to the plasma membrane. Expression of K44A/ dynamin did, however, result in the accumulation of GLUT4 at the plasma membrane in the basal state. This gradual accumulation of GLUT4 at the cell surface suggests that the population of continuously recycling GLUT4 was unable to internalize from the plasma membrane. While this study was in progress, it has also been reported that expression of a mutant dynamin in Chinese hamster ovary cells co-expressing both the insulin receptor and GLUT4 results in a redistribution of GLUT4 from an intracellular location to the cell surface (65). Similarly, microinjection of a dynamin inhibitory peptide also FIG. 10. K44A/dynamin expression delays GLUT4 endocytosis following removal of insulin. Uninfected (Uninf) cells (panels 1-4), or cells infected for 48 h with adenovirus encoding LacZ (panels [5][6][7][8] or K44A/dynamin (panels 9 -12) were serum-starved and incubated in the absence (panels 1, 5, and 9) or presence (panels 2, 6, and 10) of 100 nM insulin (Ins) for 30 min (30Љ) at 37°C. The insulin was removed, and the cells were allowed to internalize GLUT4 for 30 min (panels 3, 7, and 11) or 2 h (panels 4, 8, and 12). Plasma membrane sheets were prepared and processed for GLUT4 immunofluorescence. Bar, 50 M.  1, 4, and 7) or presence (I, panels 2, 5, and 8) of 100 nM insulin for 30 min at 37°C. The bound insulin was removed, and cells allowed to internalize GLUT4 for 2 h (W, lanes 3, 6, and 9). Plasma membrane sheets were collected, subjected to SDS-PAGE, transferred to nitrocellulose, and immunoblotted (IB) with a GLUT4-specific antibody.  1 and 4) or presence (I, lanes 2 and 5) of 100 nM insulin for 30 min at 37°C. Insulin was removed by washing in acidic buffer, and cells were incubated for an additional 2 h at 37°C (W, lanes 3 and 6). Whole cell lysates were prepared, subjected to SDS-PAGE, transferred to nitrocellulose, and immunoblotted with an anti-phosphotyrosine-specific antibody. resulted in appearance of GLUT4 at the cell surface concomitant with a reduction in GLUT4 internalization (66). Together these data are consistent with the presence of at least two intracellular GLUT4 vesicle populations, one which is undergoing continuous recycling to and from the plasma membrane and a separate pool that is responsive to insulin stimulation.
In summary, our data demonstrate that insulin receptor signal transduction events in 3T3L1 adipocytes, particularly the activation of GLUT4 translocation, can occur in the absence of clathrin-mediated endocytosis. We conclude that GLUT4 endocytosis is dependent upon dynamin, and when its function is inhibited, GLUT4 accumulates at the plasma membrane. The ability of insulin to fully induce the translocation of the remaining intracellular GLUT4 vesicles strongly suggests the presence of two independent GLUT4 vesicle populations, one which undergoes constitutive recycling and a second which is uniquely responsive to insulin action. Although these two populations probably equilibrate with each other at a slow rate, the physical nature of these vesicles remains to be determined. These findings underscore the importance of dynamin function in GLUT4 endocytosis and suggest a potential mechanism by which insulin is able to regulate the rate of GLUT4 endocytosis.