Aldolase Mediates the Association of F-actin with the Insulin-responsive Glucose Transporter GLUT4*

To identify potential proteins interacting with the insulin-responsive glucose transporter (GLUT4), we generated fusion proteins of glutathione S-transferase (GST) and the final 30 amino acids from GLUT4 (GST-G4) or GLUT1 (GST-G1). Incubation of these carboxyl-terminal fusion proteins with adipocyte cell extracts revealed a specific interaction of GLUT4 with fructose 1,6-bisphosphate aldolase. In the presence of aldolase, GST-G4 but not GST-G1 was able to co-pellet with filamentous (F)-actin. This interaction was prevented by incubation with the aldolase substrates, fructose 1,6-bisphosphate or glyceraldehyde 3-phosphate. Immunofluorescence confocal microscopy demonstrated a significant co-localization of aldolase and GLUT4 in intact 3T3L1 adipocytes, which decreased following insulin stimulation. Introduction into permeabilized 3T3L1 adipocytes of fructose 1,6-bisphosphate or the metabolic inhibitor 2-deoxyglucose, two agents that disrupt the interaction between aldolase and actin, inhibited insulin-stimulated GLUT4 exocytosis without affecting GLUT4 endocytosis. Furthermore, microinjection of an aldolase-specific antibody also inhibited insulin-stimulated GLUT4 translocation. These data suggest that aldolase functions as a scaffolding protein for GLUT4 and that glucose metabolism may provide a negative feedback signal for the regulation of glucose transport by insulin.

The insulin-responsive glucose transporter GLUT4 is expressed primarily in adipose tissue, skeletal, and cardiac muscle (1)(2)(3)(4). Under basal conditions, GLUT4 slowly recycles between poorly defined intracellular compartments and the plasma membrane with the vast majority sequestered in these intracellular storage sites. Insulin stimulates a large increase in the rate of GLUT4 exocytosis concomitant with a smaller decrease in the rate of GLUT4 endocytosis (5)(6)(7). The overall insulin-induced changes in GLUT4 trafficking kinetics result in a 10 -20-fold increase in the number of cell surface GLUT4 proteins that accounts for the majority of insulin-stimulated increases in glucose transport activity (8,9).
Recently, several laboratories have begun to examine the subcellular distribution of GLUT4 to identify the mechanism responsible for the intracellular sequestration of the GLUT4 protein. Steady-state and kinetic analysis of various expressed GLUT4 chimeric proteins have indicated that both the aminoand carboxyl-terminal domains are important in GLUT4 internalization from the plasma membrane (10 -15). In particular, the carboxyl-terminal dileucine motif (SLL) was found to substantially alter GLUT4 trafficking kinetics and steady-state localization (16,17). Although a specific GLUT4 sequence responsible for intracellular localization has yet to be identified, the presence of a carboxyl-terminal retention signal is consistent with the observation that expression of the GLUT4 carboxyl-terminal domain results in the translocation of the endogenous GLUT4 protein to the plasma membrane (18). In addition, the presence of a GLUT4 carboxyl-terminal binding protein can also account for the apparent increase in GLUT4 carboxylterminal antibody immunoreactivity following insulin stimulation (19,20).
To address these issues, we have used GST fusion proteins to identify and characterize GLUT4 carboxyl-terminal specific binding proteins. In this manuscript, we demonstrate that the carboxyl-terminal domain of GLUT4 can specifically associate with the bifunctional glycolytic enzyme fructose-1,6-bisphosphate aldolase. This in vitro interaction appears to be physiologically important as glycolytic intermediates that disrupt aldolase-actin interactions also inhibit insulin-stimulated GLUT4 translocation.
GST Fusion Protein Pull-down Experiments-3T3L1 fibroblasts were differentiated into adipocytes as described (21). In labeling experiments, 3T3L1 adipocytes were incubated with 10 Ci/ml Tran 35 S label (ICN) for 24 h before isolation of cell extracts in 50 mM Hepes, pH 7.4, 1% Triton X-100, 10% glycerol. Insoluble material was separated by microcentrifugation and extracts precleared by incubation with glutathione-Sepharose beads for 1 h at 48°C followed by a second 1 h incubation with the GST fusion protein. The beads were then washed three times with 50 mM Hepes, pH 7.4, 0.1% Triton-100, 10% glycerol, bound proteins eluted using Laemmli sample buffer and samples subjected to SDS-PAGE, and autoradiography or aldolase immunoblotting. For large scale preparation of the GST-GLUT4 binding proteins, rat epididymal fat pads were isolated from 50 rats. Cell extracts were incubated with the GST fusion proteins, eluants transferred onto polyvinylidene difluoride membrane and N-terminal sequences determined using an automated protein sequencer at the University of Michigan * 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.
Protein Core Facility.
F-actin/Aldolase/GLUT4 Co-pelleting Experiments-Aldolase and GST fusion protein solutions were dialyzed against 10 mM imidazole, 10 mM KCl, 1 mM MgCl 2 , 0.1 mM dithiothreitol, pH 6.8, overnight at 48°C and precleared by centrifugation at 45,000 rpm (TLA100 rotor, Beckman). 200 g of rabbit skeletal muscle F-actin (Sigma) purified as described (23) and 100 l of the 20-l aldolase solution were mixed for 30 min at room temperature following 5-10 strokes in a pestle homogenizer. The mixture was centrifuged and 300 g/ml GST fusion protein was added to the pellet and suspended with a small plastic homogenizer rod. After incubation for 30 min at room temperature, the mixture was again centrifuged, and the supernatants and pellets subjected to SDS-PAGE. Proteins were detected by Coomassie Brilliant Blue staining. Fructose 1,6-bisphosphate (FBP) or glyceraldehyde 3-phosphate (G3P) were added to both the aldolase and the GST fusion proteins at a concentration of 50 M as indicated.
Whole Cell Immunofluorescence Double Labeling-3T3L1 adipocytes on glass coverslips were serum-starved for 3 h then left untreated or stimulated for 30 min with 100 nM insulin. Cells were fixed and permeabilized with 2% paraformaldehyde, 0.2% Triton X-100 in phosphate-buffered saline for 10 min and quenched in 100 mM glutamine/ phosphate-buffered saline for 15 min. Cells were then washed and sequentially incubated in 5% donkey serum, 1% bovine serum albumin/ phosphate-buffered saline blocking buffer, 1:200 IA02 anti-GLUT4 and 1:100 anti-aldolase (Biogenesis), and 12.5 g/ml FITC-conjugated donkey anti-rabbit and 12.5 g/ml Texas Red-conjugated donkey anti-goat antibodies (Jackson ImmunoResearch Laboratories). Coverslips were mounted onto glass slides with Vectashield mounting medium (Vector Labs) and viewed with a Bio-Rad laser confocal.
Plasma Membrane Sheet Preparation-Plasma membrane sheets were prepared from 3T3L1 adipocytes as described (24). For 2-deoxyglucose (2DG) treatment of intact 3T3L1 adipocytes, cells were serum starved and placed into glucose-free Dulbecco's modified Eagle's medium containing 5 mM pyruvate and 5 mM glutamine with or without 5 mM 2DG for 5 min before insulin stimulation. For fluorescence microscopy, the membrane sheets were fixed with 2% formaldehyde and sequentially incubated with 5% donkey serum, a 1:100 dilution of a rabbit anti-insulin regulatable glucose transporter antibody (East Acres) and 6.25 g/ml of LRSC-conjugated donkey anti-rabbit antibody (Jackson ImmunoResearch Laboratories). For immunoblotting experiments, cells were stimulated with 20 nM insulin for 20 min, membrane sheets were isolated and solubilized in a modified lysis buffer (22) with 0.05% Triton X-100. A protein assay was performed, and lysates were placed into Laemmli sample buffer without boiling.
Microinjection-3T3L1 adipocytes were placed into L-15 medium and microinjected using an Eppendorf Micromanipulator 5171 and Transjector 5246. Injection buffer consisted of 100 mM KCl and 5 mM sodium phosphate, pH 7.2, with 1.0 mg/ml FITC-dextran (Sigma), 1.2 mg/ml maltose-binding protein (MBP)-Ras CAAX, and 2.5 mg/ml either aldolase antibody (Biogenesis) or pre-immune serum (Research Genetics). Microinjected cells were placed in serum-free Dulbecco's modified Eagle's medium containing 0.1% bovine serum albumin for 1.5 h, treated with or without 20 nM insulin for 20 min, and plasma membrane sheets isolated. Sheets were incubated with a 1:1,000 dilution of sheep anti-MBP antibody (generously provided by Dr. Morris Birnbaum) and 1:400 dilution of IAO2 GLUT4 antibody, followed by incubation in 12.5 mg/ml FITC-conjugated donkey anti-sheep and Cy5-conjugated donkey anti-rabbit antibodies (Jackson ImmunoResearch Laboratories). In each experiment, 20 -40 injected cells positive for FITC-MBP were scored for GLUT4-specific immunofluorescence by two individuals blinded to treatment. The results of two or three independent experiments were analyzed using the analysis program InStat 2.0.

RESULTS AND DISCUSSION
To identify proteins that interact with the GLUT4 carboxylterminal domain, we expressed the GLUT1 and GLUT4 carboxyl-terminal 30 amino acids as GST fusions, GST-G1 and GST-G4, respectively. An additional fusion construct of the GLUT4 carboxyl terminus was generated containing a substitution of two residues shown to be important in GLUT4 endocytosis, leucines 488 and 489, with alanine and serine (GST-G4*), respectively (15)(16)(17). Incubation of GST and GST-G4 with extracts from [ 35 S]methionine-labeled 3T3L1 adipocytes resulted in the binding of multiple proteins. However, there were two faint bands at 40 and 43 kDa that appeared to specifically precipitate with GST-G4 and GST-G4* but not with GST or GST-G1 (Fig. 1A). Pretreatment of 3T3L1 adipocytes with insulin before isolation of the cell extracts had no significant effect on the in vitro association of the 40 and 43 kDa protein with GST-G4 and GST-G4* (Fig. 1A).
Cell extracts were also prepared from primary isolated rat adipocytes and incubated with the GST fusion proteins (Fig.  1B). Similar to 3T3L1 adipocytes, Coomassie Blue staining of these precipitates demonstrated that the 40-and 43-kDa proteins were specifically precipitated with GST-G4 and GST-G4* but not by GST or GST-G1. Although other more intense bands were detected on the Coomassie Blue stained gel, these were also seen in the absence of lysate and were therefore nonspecific. We could not obtain a defined amino acid sequence following purification of the 43-kDa protein; however, the 40-kDa protein was partially sequenced and identified as fructose 1,6bisphosphate aldolase. To confirm that the 40-kDa protein that bound GST-G4 and GST-G4* was aldolase, immunoblotting with an aldolase-specific antibody was performed (Fig. 1C). Incubation of cell extracts with GST-G4 and GST-G4* resulted in the co-precipitation of a 40-kDa protein that demonstrated specific immunoreactivity with an aldolase antibody. In contrast, no immunoreactive proteins were detected in the GST or GST-G1 precipitates. Furthermore, as observed in the [ 35 S]methionine-labeled 3T3L1 adipocytes, cell extracts from insulinstimulated cells also precipitated aldolase immunoreactivity in the presence of GST-G4 and GST-G4* but not GST or GST-G1. These data demonstrate that the 40-kDa protein that specifi- cally binds the carboxyl-terminal domain of GLUT4 is fructose 1,6-bisphosphate aldolase.
Aldolase is an unusual protein in that it not only possesses enzymatic activity but also plays a structural role in the assembly of the actin cytoskeleton (26). The interaction of aldolase with actin can be modulated by aldolase substrates and products (27,28). We therefore sought to determine the allosteric effect of these and related molecules on the binding of aldolase to GST-G4 (Fig. 2). Dose-response relationships demonstrated that 1 mM FBP, a concentration similar to the K m of the enzyme for this substrate, completely inhibited the binding of aldolase to GST-G4 in vitro ( Fig. 2A). The aldolase reaction product, G3P, also inhibited GST-G4 binding to aldolase, although at a significantly higher concentration, 10 mM (Fig. 2B). In contrast, the aldolase product dihydroxyacetone phosphate, as well as two other structurally related sugars, fructose 1-phosphate and fructose 6-phosphate, had no effect on the interaction between GST-G4 and aldolase (Fig. 2, C, D, and E).
Previous studies have demonstrated that aldolase is associated with the actin cytoskeleton and can cross-link actin fibers (27,29). Because the actin cytoskeleton has also been implicated in the regulation of GLUT4 trafficking (30, 31), we next examined the ability of aldolase to function as a molecular scaffold linking GLUT4 to F-actin (Fig. 3). F-actin was preincubated with aldolase, followed by a second incubation with GST-G1 or GST-G4. When the actin was pelleted by centrifugation, GST-G4 was found to be specifically associated with actin in the presence of aldolase (Fig. 3, lane 1). GST-G4 did not co-pellet with actin in the absence of aldolase, suggesting that the GLUT4 protein binds to actin indirectly (Fig. 3, lane 5). Likewise, the aldolase⅐GST-G4 complex was not pelleted by centrifugation in the absence of actin (Fig. 3, lane 7). GST-G1 did not associate with actin in either the absence or presence of aldolase (Fig. 3, lanes 4 and 6). As previously observed in Fig.  2, FBP and G3P can inhibit the association of aldolase with GST-G4. FBP and G3P were also able to prevent the formation of the F-actin⅐aldolase⅐GST-G4 ternary complex (Fig. 3, lanes 2  and 3). These data indicate that in vitro, aldolase can function as a molecular scaffold linking the carboxyl-terminal domain of GLUT4 with filamentous actin.
To determine whether aldolase and GLUT4 were co-localized in vivo, double immunofluorescence labeling was performed. 3T3L1 adipocytes were treated with or without insulin, fixed, and incubated with aldolase and GLUT4-specific antibodies followed by labeling with complimentary fluorescent-conjugated secondary antibodies (Fig. 4). Under basal conditions, the membrane bound GLUT4 protein was found primarily in a peri-nuclear distribution and in small vesicles scattered throughout the cytoplasm (Fig. 4, panel A). Following insulin stimulation, GLUT4 moved to the cell surface generating a "rim"-like fluorescence (Fig. 4, panel D). Conversely, the soluble protein aldolase stained in a diffusely cytosolic pattern although it was also concentrated in the perinuclear region of the cell (Fig. 4, panels B and E). When their respective labels are superimposed, GLUT4 and aldolase exhibited a high degree of co-localization (Fig. 4, panels C and F), suggesting an in vivo association between aldolase and the GLUT4-containing compartments.
To further investigate the in vivo significance of these findings, we took advantage of the glycolytic inhibitor 2DG, which induces the dissociation of perinuclear aldolase from the actin cytoskeleton (32). In adipocytes, the insulin-induced translocation of GLUT4 to the plasma membrane can be detected by the fluorescent detection of a GLUT4 antibody on the cytoplasmic face of isolated plasma membrane sheets. Pretreatment of 3T3L1 adipocytes with 5 mM 2DG slightly reduced the amount of basal plasma membrane GLUT4 immunofluorescence (Fig.  5A, panels 1 and 3). In contrast, pretreatment with 2DG markedly attenuated the insulin-stimulated translocation of GLUT4 to the cell surface (Fig. 5A, panels 2 and 4). The 2DG inhibition of insulin-stimulated GLUT4 translocation occurred in a dosedependent manner with an EC 50 of approximately 100 mM (data not shown). This inhibition of GLUT4 translocation was also observed in GLUT4 immunoblots of isolated plasma membranes without any effect on the plasma membrane association of the clathrin heavy chain (Fig. 5B).
The blockade of insulin-stimulated GLUT4 translocation by 2DG could be due to either an inhibition of exocytosis or an increase in endocytosis. To determine whether 2DG affected the rate of GLUT4 endocytosis, we stimulated cells with insulin, removed the insulin by washing with an acidic buffer, and then assayed plasma membrane GLUT4 at various times after insulin removal in the presence of 2DG. Disappearance of GLUT4 from the cell surface after insulin removal was not inhibited by 2DG treatment but rather seemed to be slightly enhanced (data not shown). This small increase in the rate of GLUT4 endocytosis in the presence of 2DG likely reflected the inhibition of further GLUT4 exocytosis following insulin removal. This effect was similar to that reported for wortmannin,  3. Actin, aldolase, and GLUT4 carboxyl terminus form a ternary complex that is disrupted by FBP or G3P. F-actin was resuspended in a precleared aldolase solution and then pelleted by centrifugation. The pellet was then resuspended in GST-G4 or GST-G1 containing buffer. FBP or G3P were included in both aldolase and fusion protein solutions as indicated. Polymerized actin was isolated by microcentrifugation, the pellets resolved by SDS-PAGE, and proteins visualized by Coomassie Blue staining. which also inhibits GLUT4 exocytosis but appears to enhance endocytosis due to the inhibition of residual exocytosis (33). These data demonstrate that the rate of GLUT4 endocytosis was not significantly affected by 2DG pretreatment and is consistent with an inhibition of GLUT4 exocytosis.
Although cells were incubated with glutamine and pyruvate as alternative energy sources during 2DG treatment, cellular ATP levels were found to decrease by 40% (data not shown). Because endocytosis and exocytosis may have different sensitivities to ATP depletion, the inhibition of insulin-stimulated GLUT4 exocytosis could have been due to this decrease in ATP.
To ensure adequate cellular ATP levels while directly assessing the uncoupling of aldolase from GLUT4 in vivo, 3T3L1 adipocytes were permeabilized with SLO in the presence of 10 mM MgATP (Fig. 5, C and D). Under these experimental conditions, cellular metabolism did not deplete local concentrations of ATP (data not shown). In the SLO-permeabilized adipocytes, insulin induced an increase in the translocation of GLUT4 to the plasma membrane (Fig. 5C, panels 1 and 2). As in the intact cells, preincubation with 2DG prevented GLUT4 translocation (data not shown). In addition, although preincubation with FBP slightly enhanced the basal amount of GLUT4 associated FIG. 4. Aldolase and GLUT4 are partially co-localized in 3T3L1 adipocytes. Serum-starved 3T3L1 adipocytes were stimulated without (panels A-C) or with (panels D-F) 100 nM insulin. Cells were then fixed and incubated with GLUT4 and aldolase-specific primary antibodies and FITC or Texas Red (TR)-conjugated secondary antibodies as indicated. Cells were visualized by laser confocal microscopy. Bar, 110 m.

FIG. 5. 2DG and FBP inhibit insulin-stimulated GLUT4 translocation.
Panel A, 3T3L1 adipocytes were pre-incubated with or without 5 mM 2DG. The cells were then treated with insulin followed by isolation of plasma membrane sheets and immunofluorescence microscopy using a GLUT4-specific antibody. Panel B, 3T3L1 adipocytes were incubated as above and the samples immunoblotted with a GLUT4 or clathrin antibody. C, 3T3L1 adipocytes were permeabilized with SLO, pre-incubated in the absence or presence of 25 mM FBP and then treated as in panel A. Panel D, permeabilized 3T3L1 adipocytes were pre-incubated with FBP and then treated as in panel B. Bar, 110 m.
with the plasma membrane, FBP pretreatment markedly inhibited the insulin-stimulated translocation of GLUT4 (Fig.  5C, panels 3 and 4). It should be noted that the small increase in plasma membrane-associated GLUT4 following SLO permeabilization has also been previously observed by others (25). In agreement with the immunofluorescence analysis, immunoblotting of isolated plasma membranes demonstrated that FBP preincubation partially increased the basal plasma membraneassociated GLUT4 and inhibited the insulin-stimulated increase without affecting the plasma membrane association of the clathrin heavy chain (Fig. 5D).
To examine the metabolite specificity for the inhibition of insulin-stimulated GLUT4 translocation, we compared the effects of several aldolase substrates, products and related sugars on GLUT4 translocation (data not shown). Consistent with their effect on aldolase-actin binding, pretreatment of SLOpermeabilized adipocytes with dihydroxyacetone phosphate, fructose 1-phosphate, fructose 6-phosphate, or glucose had no significant effect on insulin-stimulated GLUT4 translocation compared with control cells. To further confirm a specific requirement for aldolase in insulin-stimulated GLUT4 translocation, we next microinjected cells with control or aldolase antibodies and assessed single cell GLUT4 translocation (Fig.  6). Specific identification of the microinjected cells was accomplished by co-injection with a carboxyl-terminal domain of Ras fused to the MBP as a plasma membrane marker. In unstimulated cells, microinjection of either pre-immune IgG or the aldolase IgG had no effect on GLUT4 translocation, which remained at low basal levels (Fig. 6A, panels 1-4). Microinjection of pre-immune IgG did not inhibit the insulin-stimulated translocation of GLUT4 (Fig. 6B, panels 1 and 2). In contrast, the aldolase-specific IgG reduced the extent of GLUT4 translocation only in the cells that were microinjected but not in the surrounding nonmicroinjected cells (Fig. 6B, panels 3 and 4). Quantitation of these data demonstrated that insulin stimulated GLUT4 translocation in 49% of cells microinjected with pre-immune IgG and 29% of cells microinjected with aldolasespecific IgG (Fig. 6C). Together, these data provide compelling evidence for a specific functional role of fructose 1,6-bisphosphate aldolase in the insulin stimulation of GLUT4 translocation in adipocytes.
Previous studies have demonstrated that aldolase exhibits functional duality. In addition to its enzymatic activity, this protein plays a structural role in the binding and polymerization of actin (28,29,34). Depolymerization of the actin cytoskeleton with cytochalasin D or latrunculin B has been shown to attenuate insulin-stimulated GLUT4 translocation, suggesting that an intact actin cytoskeleton is required for insulin-stimulated GLUT4 translocation (30,31). Consistent with this model, our data demonstrate that aldolase, a functional tetramer, may also serve as a scaffolding protein linking GLUT4 and hence GLUT4 vesicles, to the actin cytoskeleton. In in vitro experiments, aldolase not only specifically interacted with the carboxyl-terminal domain of GLUT4, but linked GLUT4 to FIG. 6. Microinjection of an aldolase antibody inhibits insulin-stimulated GLUT4 translocation. 3T3L1 adipocytes were microinjected with MBP-Ras and either pre-immune serum or an aldolase antibody. Cells were left unstimulated (A) or stimulated (B) with insulin. Isolated plasma membrane sheets were incubated with MBP or GLUT4 antibodies followed by FITC or Cy5-conjugated secondary antibodies. Microinjected cells (arrowheads) were identified by FITC anti-MBP-specific immunofluorescence, whereas the membrane-localized GLUT4 was assessed by Cy5 anti-GLUT4-specific immunofluorescence. Bar, 30 m. C, microinjected 3T3L1 adipocytes were scored for GLUT4-specific immunofluorescence. The percentage of injected cells demonstrating GLUT4 at the cell surface (GLUT4 positive sheets) are shown for control (Ⅺ) and insulinstimulated (f) cells. Asterisks (*) indicate that the percentage of GLUT4 positive cells following insulin stimulation was significantly different between those injected with anti-aldolase antibody compared with pre-immune serum (p Ͻ 0.05). Results are the average of two (control) or three (insulin-stimulated) independent experiments. polymerized actin. This interaction was disrupted by specific substrates of aldolase that inhibit GLUT4-aldolase or aldolaseactin interactions. These specific substrates not only prevented aldolase-actin interaction but also prevented GLUT4-aldolase binding in vitro and in a manner consistent with their ability to interfere with insulin-stimulated GLUT4 translocation in vivo. Furthermore, aldolase and GLUT4 exhibited an overlapping subcellular distribution, and microinjection of an aldolase-specific antibody reduced the insulin-stimulated translocation of GLUT4. Together, these data strongly suggest that aldolase plays a critical role in the dynamic association of GLUT4 vesicles with the actin cytoskeleton. The ability of aldolase substrates to disrupt the actin⅐aldolase⅐GLUT4 complex, as well as GLUT4 vesicle trafficking, suggests that the metabolism of glucose may provide a negative feedback signal to prevent further exocytosis of GLUT4 vesicles and potentially contribute to the mechanism of glucotoxicity.