The Glucose Transporter 4-regulating Protein TUG Is Essential for Highly Insulin-responsive Glucose Uptake in 3T3-L1 Adipocytes*

Insulin stimulates glucose uptake in fat and muscle by redistributing GLUT4 glucose transporters from intracellular membranes to the cell surface. We previously proposed that, in 3T3-L1 adipocytes, TUG retains GLUT4 within unstimulated cells and insulin mobilizes this retained GLUT4 by stimulating its dissociation from TUG. Yet the relative importance of this action in the overall control of glucose uptake remains uncertain. Here we report that transient, small interfering RNA-mediated depletion of TUG causes GLUT4 translocation and enhances glucose uptake in unstimulated 3T3-L1 adipocytes, similar to insulin. Stable TUG depletion or expression of a dominant negative fragment likewise stimulates GLUT4 redistribution and glucose uptake, and insulin causes a 2-fold further increase. Microscopy shows that TUG governs the accumulation of GLUT4 in perinuclear membranes distinct from endosomes and indicates that it is this pool of GLUT4 that is mobilized by TUG disruption. Interestingly, in addition to translocating GLUT4 and enhancing glucose uptake, TUG disruption appears to accelerate the degradation of GLUT4 in lysosomes. Finally, we find that TUG binds directly and specifically to a large intracellular loop in GLUT4. Together, these findings demonstrate that TUG is required to retain GLUT4 intracellularly in 3T3-L1 adipocytes in the absence of insulin and further implicate the insulin-stimulated dissociation of TUG and GLUT4 as an important action by which insulin stimulates glucose uptake.

Insulin controls glucose utilization by adipose and muscle tissues by modulating the subcellular distribution of GLUT4 2 glucose transporters (1,2). In unstimulated cells, GLUT4 is retained in intracellular membranes, and its predominant exclusion from the plasma membrane minimizes glucose uptake from the extracellular space. Within minutes of insulin stimulation, GLUT4 is translocated so that the number of transporters in the plasma membrane is increased and glucose uptake is accelerated. This insulin action is defective in muscle and adipose from individuals with type 2 diabetes and other insulin-resistant states (3)(4)(5). The molecular mechanisms governing GLUT4 distribution and its regulation by insulin remain poorly characterized.
To redistribute GLUT4, insulin acts primarily on the exocytic arm of the GLUT4 recycling pathway. The signaling proteins involved include Akt/protein kinase B and its substrate AS160; the latter protein was recently shown also to contribute to intracellular retention of GLUT4 in unstimulated cells (6,7). Studies of GLUT4 vesicle traffic have been complicated by the fact that unstimulated cells, particularly the widely used 3T3-L1 adipocyte cell line, contain multiple pools of intracellular GLUT4. These include GLUT4 in endosomes, the trans-Golgi network, and poorly characterized membranes sometimes termed as the "insulin-responsive compartment" or "GLUT4 storage vesicles" (GSVs). Insulin stimulates endosome recycling, which may contribute a 2-fold increase in cell surface GLUT4 and other proteins such as the transferrin receptor (TfnR). The more dramatic, insulin-stimulated increase in cell surface GLUT4, and not in other recycling proteins, is due to mobilization of GSVs. These membranes comprise small, preformed vesicles that are depleted of TfnR and other endosomal markers (8,9). Data indicate that GSVs are derived by budding from the trans-Golgi network, yet they may also exchange with endosomes (10 -12). In adipocytes, GSVs are distinct from secretory vesicles containing adiponectin (13). The biogenesis of these vesicles is both cell type-specific and developmentally regulated (14).
How GSVs are retained intracellularly in the absence of insulin, and how these GLUT4-containing membranes are mobilized by insulin stimulation, remains uncertain. We recently proposed that in unstimulated cells, GLUT4 that is present in GSVs may be tethered to intracellular structures to restrict its movement to the plasma membrane (15). According to this model, insulin "untethers" this GLUT4 to accelerate its movement to the cell surface and subsequent fusion at the plasma membrane. We identified TUG as a putative tethering protein, which acts in concert with other proteins to retain GLUT4 within cells in the absence of insulin. Supporting this model, TUG and GLUT4 form a complex in unstimulated 3T3-L1 adipocytes, and colocalize on TfnR-negative intracellular membranes that sediment as light microsomes (15). These are properties shared by GSVs. The TUG-GLUT4 complex is insulin-responsive, because insulin acts rapidly to stimulate the dissociation of TUG and GLUT4 (15). Overexpression of TUG both increases the size of the insulin-responsive GLUT4 pool, assessed using kinetics, and increases the number of insulin-responsive TUG-GLUT4 protein complexes (15). Conversely, expression of a dominant negative TUG UBX-Cter fragment decreases the size of the insulin-responsive pool and decreases the number of TUG-GLUT4 complexes. Together, these data suggested that TUG is part of an insulin-responsive mechanism governing the sequestration of GLUT4 in GSVs within unstimulated 3T3-L1 cells.
Here we use RNAi-mediated depletion of TUG to further characterize its role in GLUT4 trafficking and glucose uptake, using 3T3-L1 adipocytes. These data show that TUG is required for intracellular retention of GLUT4 in nonendosomal, perinuclear membranes within unstimulated cells, consistent with the model described above. The magnitude of the effects we observe suggests that the TUG-GLUT4 complex may be an important site of insulin action to regulate glucose uptake in 3T3-L1 adipocytes.

EXPERIMENTAL PROCEDURES
Cell Lines and Reagents-3T3-L1 adipocytes were cultured as described previously, except that the fully supplemented, initial differentiation medium was used for 72 h, and differentiation medium containing insulin alone was omitted (13). siRNAs were electroporated into mature 3T3-L1 adipocytes as described previously (16). In some experiments siPORT electroporation buffer (Ambion Inc., Austin, TX) was used. 3T3-L1 cells stably expressing a GLUT4 -7Myc-GFP reporter protein, with or without the TUG UBX-Cter fragment or overexpressed TUG protein, have been described previously (13,15). For infection of 3T3-L1 cells with a retrovirus containing a TUG shRNA, the pMSCV-U3-H1 vector was used (a kind gift from Dr. Biao Luo, Broad Institute, Cambridge, MA) (17). Replication-defective, murine ecotropic retrovirus was produced as described previously, and infected preadipocytes were purified using flow cytometry of a GFP marker (13,15,18). To reintroduce TUG into the shRNA cells, a TUG cDNA containing silent mutations in the shRNA target site was expressed using the pBICD2 retrovirus vector, and infected cells were isolated by fluorescence-activated cell sorting of a CD2 marker (15,19).
Polyclonal antisera directed to TUG were described previously (15). Antisera to the insulin receptor ␤-chain were purchased from Upstate Biotechnology (Charlottesville, VA), and antisera to GLUT4 were a kind gift from Dr. Maureen Charron and were also raised to a C-terminal peptide as described (20) or purchased from Chemicon Inc. (Temecula, CA). Antisera to GLUT1 were purchased from Chemicon Inc. An anti-Myc (9E10) monoclonal antibody was purchased from Covance Inc. (Berkeley, CA), anti-␤-actin antibody from Abcam (Cambridge, MA), anti-transferrin receptor antibody from BD Pharmingen (San Diego, CA), and anti-Hsc70 antibody from Stressgen Bioreagents (Victoria, British Columbia, Canada).
Subcellular Fractionation and Immunoblotting-Subcellular fractionation was performed as described previously (13). Pellets previously termed "LDM" and "HDM" are here more accurately called light microsomes (LMs) and heavy microsomes (HMs). Pellets were resuspended in radioimmune precipitation assay buffer containing Complete protease inhibitors (Roche Applied Science), as described (13). Total protein concentrations were measured in triplicate using micro-BCA (Pierce) or EZQ protein quantitation kits (Molecular Probes, Eugene, OR) and a PerkinElmer Life Sciences Victor 3 plate reader. For SDS-PAGE, equal protein amounts were loaded in each well, and Invitrogen Novex NuPAGE gels and buffers were used. Proteins were transferred to nitrocellulose membranes using a semi-dry apparatus, and immunoblotting was performed using enhanced chemiluminescent detection on film as described previously (13). For densitometry, film exposures in the linear range were scanned using transillumination at 16-bit pixel depth, and quantification was done using ImageJ. Statistical significance was assessed using a paired, two-tailed Student's t test.
2-Deoxyglucose Uptake-Assays were done as described, except that 12-or 6-well plates were used and mannitol was omitted (15). Briefly, 3T3-L1 adipocytes were differentiated (for stable cell lines) or reseeded (for cells electroporated with siRNA duplexes) in multiwell plates. Electroporated cells were assayed 2 days after siRNA transfection. Cells were starved, washed, then stimulated using 80 or 160 nM insulin for 20 min at 37°C. 1 Ci/ml 2-deoxy-D-[2,6-3 H]glucose was added so that the final concentration of 2-deoxyglucose was 50 mM, and uptake was allowed to proceed at 37°C for 6 min or at 25°C for 8 min. Cells were washed using cold phosphate-buffered saline and lysed in radioimmune precipitation assay buffer. A portion of each lysate was used for a BCA protein assay (Pierce), so that total protein could be used to correct for variation in the number of cells in each well. Nonspecific uptake was measured in the presence of 20 M cytochalasin B and was subtracted from 3 H counts obtained from other wells. Data points were obtained in triplicate or quadruplicate, and statistical significance was assessed using a paired, two-tailed Student's t test.
Recombinant Proteins and Binding Experiments-Plasmids containing the large intracellular loop of rat GLUT4 (residues 223-288) or GLUT1 (residues 207-271) in a modified pGEX-2T vector were a kind gift of Drs. Amira Klip and Philip Bilan. These proteins and control GST protein were expressed in the BL21(DE3) Escherichia coli strain or its derivative, Rosetta(DE3) (Novagen, Madison, WI). Expression was induced using isopropyl-␤-D-thiogalactopyranoside, then cells were pelleted and lysed in ice-cold NETN buffer (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40). Proteins were purified from the soluble phase using glutathione-Sepharose beads (Amersham Biosciences). To use equal amounts of GST or GST fusion proteins in binding experiments, aliquots of the beads were removed and the amount of protein bound was assessed by elution in sample buffer, SDS-PAGE, and Gelcode Coomassie staining (Pierce). Recombinant, full-length TUG protein was subcloned in the pTYB3 vector (New England Biolabs, Ipswich, MA) and produced using the Roche Applied Science RTS system according to the manufacturer's instructions. TUG fragments containing residues 1-164 or 165-550 were produced and labeled with [ 35 S]methionine using a TNT coupled reticulocyte lysate (Promega Corp., Madison, WI). Intact or truncated TUG proteins were incubated overnight at 4°C with bound GST, GST-GLUT4loop, or GST-GLUT1loop in NETN buffer. The following day, the beads were washed four times with NETN buffer, and proteins were eluted in SDS-PAGE sample buffer. Bound proteins were analyzed by SDS-PAGE and immunoblotting (for intact TUG) or autoradiography (for the 35 S-labeled fragments). Duplicate gels were stained with Gelcode Coomassie to demonstrate equivalent loading of GST or GST-loop proteins.
Confocal Microscopy-3T3-L1 adipocytes expressing a GLUT4 reporter were reseeded on coverslips, starved, stimulated, and prepared for imaging as previously described (13,15,21). To detect externalized Myc epitope tag, staining with 9E10 monoclonal antibody was done on live cells at 4°C prior to fixation. For intracellular antigens (e.g. TfnR), cells were fixed and permeabilized before staining for indirect immunofluorescence. Alexa Fluor 594-conjugated secondary antibodies (Molecular Probes) were used for simultaneous detection with GFP. Images were acquired on a Zeiss LSM 510 microscope using a 63ϫ/1.20 water immersion objective and pinhole settings for an Airy unit of 1.0. Images were acquired at 12-bit pixel depth with a scan head speed of 6, and signal averaging of 8 or 16 total acquisitions. Postprocessing was done at 12-bit pixel depth and linear gamma using ImageJ and Adobe Photoshop.
Other Methods-To assess the effect of lysosomal inhibition on GLUT4 protein amounts, 3T3-L1 adipocytes were placed in Dulbecco's modified Eagle media containing 0.5% fetal bovine serum with or without 0.4 mM chloroquine (Sigma), essentially as described (14). Cells were incubated for various amounts of time, and then lysed in radioimmune precipitation assay buffer. The protein concentration of each sample was measured as described above, and equal amounts of protein were analyzed by SDS-PAGE and immunoblotting.

RESULTS
To evaluate the role of TUG in the control of glucose uptake, we transfected 3T3-L1 adipocytes with synthetic siRNA duplexes. Fig. 1a demonstrates that two different siRNA duplexes effectively deplete TUG protein. In control, mock transfected cells, and in those transfected with irrelevant siRNA duplexes, there was no effect on TUG protein amount. In contrast, the two siRNAs that target TUG each depleted the protein by ϳ85%, assessed 48 h after transfection. Amounts of insulin receptor ␤ chain were unaffected. Thus, the two different siRNAs effectively and specifically depleted TUG protein in 3T3-L1 adipocytes.
The effect of siRNA-mediated TUG depletion on glucose uptake in 3T3-L1 adipocytes is shown in Fig. 1b. In control mock or luciferase siRNA-transfected cells, insulin stimulated a ϳ4-fold increase in glucose uptake. In cells transfected with either of the two TUG siRNAs, basal glucose uptake was markedly increased, and no further increase was observed after insulin stimulation. In several experiments, siRNA duplex B was the slightly more effective of the two duplexes, and it may have enhanced glucose uptake more than siRNA duplex A, although this did not reach statistical significance. We conclude that transient depletion of TUG markedly augments glucose uptake, similar to insulin, in unstimulated 3T3-L1 adipocytes.
To determine if the enhanced glucose uptake we observed after TUG siRNA transfection is attributable to increased GLUT4 or GLUT1 protein amounts, these proteins were assayed on immunoblots of total cell lysates. As shown in Fig.  1c, neither GLUT4 nor GLUT1 was increased in 3T3-L1 adipocytes transfected with TUG siRNA B, compared with mock transfected control cells. Indeed, GLUT4 was actually decreased, raising the possibility that TUG may coordinately regulate GLUT4 targeting and protein levels, as discussed further below. No consistent effects on GLUT1 protein amount were noted. Thus, the enhanced glucose uptake we observed could not be attributed to increased GLUT4 or GLUT1 protein amounts. These data are consistent with the possibility that TUG depletion caused GLUT4 targeting to the plasma membrane to enhance glucose uptake in cells not treated with insulin.
To further study the effect of siRNA-mediated TUG depletion on GLUT4 subcellular localization, we performed confocal microscopy of 3T3-L1 adipocytes stably expressing a GFP-and Myc-tagged GLUT4 reporter (13). As shown in Fig. 2, cells were transfected using a control siRNA and not treated with insulin targeted GLUT4 to intracellular membranes, and the typical perinuclear distribution was observed (top row). Minimal surface-exposed Myc epitope tag was detected in these unstimulated cells. After insulin exposure, GLUT4 was seen in both perinuclear and plasma membrane distributions, and detection of externalized Myc epitope tag demonstrated that GLUT4 had been incorporated into the plasma membrane (second row). In cells in which TUG was depleted by transfection of siRNA B, GLUT4 was observed at the cell surface even in the absence of insulin (third row). Externalized Myc epitope was abundant, showing that this GLUT4 had been incorporated into the plasma membrane. The surface GLUT4 may be more punctate than in the insulin-treated control cells, which have a more homogeneous pattern, yet the overall pattern was clearly distinct from that in unstimulated control cells. After insulin stimulation of the TUG siRNA-treated cells, GLUT4 was again observed at the cell surface (bottom row). Externalized Myc tag was readily detected, and again demonstrated a punctate distribution. Thus, the main result was that transient, siRNA-mediated TUG depletion caused marked incorporation of GLUT4 into the plasma membrane in unstimulated 3T3-L1 adipocytes.
To perform more detailed analyses, we constructed 3T3-L1 cells in which TUG was stably depleted by retroviral expression of a shRNA. This shRNA targets the same sequence as TUG FIGURE 1. Effect of transient, siRNA-mediated TUG depletion on glucose uptake. a, 3T3-L1 adipocytes were electroporated with synthetic siRNA duplexes as indicated. 24 or 48 h after transfection, cells were lysed and analyzed by SDS-PAGE and immunoblotting. Control transfections include a mock electroporation with buffer only, an siRNA duplex targeting luciferase (which is not present in the cells), and a scrambled siRNA duplex targeting no known gene. Immunoblots were done to detect TUG and, as a control, insulin receptor ␤ chain. The experiment was performed twice with similar results. b, 3T3-L1 adipocytes were transfected with the indicated siRNA duplexes, then seeded to multiwell plates. 48 h after transfection, glucose uptake assays were performed. Counts of [2-3 H]deoxyglucose were normalized to the amounts of protein present in the wells and are plotted relative to uptake in unstimulated control cells. Error bars show standard deviation (*, p ϭ 0.02). The experiment was performed four times with similar results. c, 3T3-L1 adipocytes were transfected with or without TUG siRNA duplex B, then immunoblots were performed to assess the amounts of GLUT4 or GLUT1 present. Lysates were made 48 h after transfection of the siRNA, and basal and insulin stimulated cells (similar to those used in glucose uptake experiments) were assayed. Lysates were immunoblotted to detect Hsc70, a heat shock protein, as a loading control. The experiment was repeated twice using cells with (as here) or without a tagged GLUT4 reporter, with no effect on the result. FIGURE 2. Effect of siRNA-mediated TUG depletion on GLUT4 distribution. 3T3-L1 adipocytes stably expressing a Myc-and GFP-tagged GLUT4 protein (13) were electroporated with luciferase siRNA (control) or with siRNA duplex B (TUG siRNA), then seeded to coverslips. 48 h after transfection, cells were starved, treated with or without 160 nM insulin for 15 min, then chilled to 4°C and stained to detect the externalized Myc epitope. Images were acquired by confocal microscopy of GFP (total GLUT4) and Myc epitope (surface GLUT4). GLUT4 present at the plasma membrane is highlighted by arrows; n indicates the position of the nuclei. Scale bar, 10 m. Similar results were obtained in two independent experiments. siRNA duplex B, described above. We isolated a homogeneous population of infected 3T3-L1 preadipocytes by flow cytometry. These "shRNA" cells undergo normal adipose differentiation, as judged by lipid accumulation and expression of GLUT4, compared with control, uninfected 3T3-L1 cells. Fig. 3a demonstrates that TUG was depleted by Ͼ95% in adipocytes containing the shRNA. To further control for the specificity of TUG depletion, the shRNA cells were superinfected with a retrovirus containing a shRNA-resistant form of TUG, to make "shRNA ϩ TUG" cells. These cells were also isolated by flow sorting and underwent normal adipose conversion. Fig. 2a shows that the reintroduced TUG protein was present in amounts that were ϳ5-fold higher than that of endogenous TUG in control, uninfected 3T3-L1 adipocytes. As well, amounts of GLUT4 were similar or slightly decreased in shRNA cells, and markedly increased in shRNAϩTUG cells, suggesting that GLUT4 abundance may be regulated by TUG protein lev-els. No large effect on GLUT1 protein amounts was observed. Thus, data from both stable and transiently transfected 3T3-L1 adipocytes suggest that TUG abundance may positively regulate GLUT4 protein amount, as examined further below.
We first tested the effect of stable TUG depletion on glucose uptake. Fig. 3b shows that unstimulated 3T3-L1 adipocytes containing the shRNA had increased glucose uptake compared with control cells. This increased basal glucose uptake was nearly equal to that caused by insulin treatment of the control cells. Insulin further enhanced glucose uptake in cells depleted of TUG, but only by ϳ2-fold because of the increased basal uptake. The enhanced glucose uptake in the shRNA cells was not attributable to increased GLUT1 or GLUT4 protein amounts, which were similar or slightly decreased in shRNA cells compared with control cells, as noted above (Fig. 3a). Reintroduction of shRNA-resistant TUG rescued the effect on glucose uptake, indicating that it resulted from TUG depletion and was not a nonspecific effect of the shRNA. Remarkably, glucose uptake in shRNAϩTUG cells was similar to that in control cells even despite the markedly increased GLUT4 abundance in shRNAϩTUG cells (Fig. 3a). This may reflect the ability of overexpressed TUG to retain GLUT4 within cells, sequestering it away from the plasma membrane ( Fig. 4a) (15). Finally, it was notable that the variability in glucose uptake measurements was much smaller in stable cells (Fig. 3b) than in electroporated cells (Fig. 1b), permitting the observation of a significant effect of insulin to enhance glucose uptake in the shRNA cells. Thus, both stable and transient depletion of TUG resulted in increased glucose uptake in unstimulated 3T3-L1 adipocytes, and in neither case was this attributable to increased amounts of GLUT4 or GLUT1 proteins.
To study the effect of the TUG shRNA on glucose transporter distribution, we performed subcellular fractionation of basal and insulin-stimulated control, shRNA, and shRNA ϩ TUG 3T3-L1 adipocytes. As shown in Fig. 4a, insulin increased GLUT4 in plasma membranes of control cells, and there was a corresponding decrease of GLUT4 in light microsomes (top left panel). In shRNA cells, GLUT4 was abundant in plasma membranes, and was relatively decreased in light microsomes, both in basal and insulin-treated cells (middle left panel). Some GLUT4 remained in basal light microsomes and was decreased by insulin, but this effect was less marked than in control cells and may have been due to endosome recycling. In the shRNA ϩ TUG cells, reintroduction of TUG rescued both basal GLUT4 retention in light microsomes and insulin-stimulated GLUT4 movement to the plasma membrane (bottom left panel). Therefore, these effects were due to TUG depletion and not nonspecific effects of the shRNA. As predicted by data in Fig. 3, GLUT4 was very well excluded from plasma membranes of basal shRNA ϩ TUG cells. Immunoblotting of the fractions to detect GLUT1 demonstrated that, in control cells, insulin caused some movement of GLUT1 out of light microsomes and to the plasma membrane (top right panel), consistent with previous data (22). TUG depletion slightly enhanced GLUT1 targeting to plasma membranes in basal 3T3-L1 adipocytes. This effect was rescued by reintroduction of TUG and was thus due to TUG depletion, yet it was much less marked than the effect on GLUT4 targeting (middle and bottom right panels). Insulin . Effects of stable shRNA-mediated TUG depletion on glucose uptake. a, TUG was depleted stably in 3T3-L1 cells using a shRNA-producing retrovirus to make "shRNA cells." Wild-type, shRNA-resistant TUG was expressed stably in the shRNA cells using a second retrovirus to make "shRNA ϩ TUG cells." Homogeneous populations of cells were isolated by flow sorting, and adipose differentiation was induced. Control, shRNA, and shRNA ϩ TUG 3T3-L1 adipocytes were analyzed by immunoblotting to demonstrate depletion and reintroduction of TUG, and to assess relative amounts of GLUT4 and GLUT1. Insulin receptor ␤ chain was immunoblotted as a loading control. b, control, shRNA, and shRNA ϩ TUG 3T3-L1 adipocytes were differentiated on multiwell plates, and glucose uptake assays were performed. Counts of [2-3 H]deoxyglucose were normalized to the amounts of protein present in the wells and are plotted relative to the uptake in unstimulated control cells. Error bars show standard deviation (*, p Ͻ 0.002; **, p Ͻ 0.01). For each part of the figure, consistent results were obtained in three independent experiments. stimulated a ϳ2-fold further increase in plasma membrane GLUT1 in shRNA cells and also appeared to increase GLUT1 in light microsomes. Overall, depletion of TUG by a stably expressed shRNA had a large effect on GLUT4 targeting, and a much smaller effect on GLUT1 targeting, in 3T3-L1 adipocytes. The main effect was marked redistribution of GLUT4 out of light microsomes and to the plasma membrane in the absence of insulin.
To quantify the effect of the TUG shRNA on GLUT4 distribution, immunoblots from several experiments were analyzed by densitometry. No large effects of TUG depletion on heavy microsomal GLUT4 were observed, thus only plasma membrane and light microsome fractions from basal and insulinstimulated cells were included in this analysis. Fig. 4b shows that TUG depletion significantly increased GLUT4 in plasma membranes and decreased GLUT4 in light microsomes, in unstimulated 3T3-L1 adipocytes. No significant effects were observed in insulin-stimulated cells. Together with the results above, the data are consistent with the notion that TUG deple-tion redistributes GLUT4 from light microsomes to the plasma membrane, and that this is sufficient to increase glucose uptake in unstimulated 3T3-L1 adipocytes.
We studied the effect of a dominant inhibitory fragment of TUG, UBX-Cter, using 3T3-L1 adipocytes stably expressing this protein. Previous data suggest that this fragment prevents GLUT4 retention in an intracellular, insulin-responsive pool of GSVs within unstimulated 3T3-L1 adipocytes (15). Thus it was anticipated that expression of UBX-Cter and RNAi-mediated TUG depletion might have similar effects. Results from glucose uptake experiments, shown in Fig. 5a, supported this prediction. The data showed that UBX-Cter expression substantially increased glucose uptake in unstimulated 3T3-L1 adipocytes, such that it was indistinguishable from that of insulin-treated control cells. Insulin further increased glucose uptake in cells  b, immunoblots of cell lysates from the glucose uptake assay were done to assess the relative amounts of GLUT4 and GLUT1. Hsc70 was immunoblotted as a loading control. The experiment was repeated twice with consistent results.
containing UBX-Cter but only by ϳ2-fold because of the enhanced basal uptake. As shown in Fig. 5b, there was at most a minor increase in GLUT1 amount, and no change in GLUT4 amount, in UBX-Cter cells compared with control cells. Thus, altered glucose transporter abundance cannot account for the markedly enhanced glucose uptake observed in UBX-Cter cells, consistent with the notion that UBX-Cter may enhance glucose uptake by targeting GLUT4 to the plasma membrane.
We further characterized the effect of the UBX-Cter fragment on GLUT4 and GLUT1 distribution by performing subcellular fractionation of 3T3-L1 adipocytes. Fig. 6a shows that GLUT4 was markedly increased in plasma membranes prepared from unstimulated cells containing UBX-Cter (top panel). There was some further increase after insulin treatment, however this was minimal because of the abundant basal amount. In unstimulated cells, UBX-Cter decreased the amount of GLUT4 in light microsomes, suggesting that some of the increased plasma membrane GLUT4 originated from these membranes. Insulin further reduced GLUT4 in light microsomes of UBX-Cter-expressing cells, yet this was a small effect that may be due to endosome recycling. Similar findings were noted for GLUT1, although the effect on plasma membrane GLUT1 was much less marked because of the abundance of GLUT1 in plasma membranes from unstimulated cells (middle panel). Expression of the UBX-Cter fragment had no large effect on heavy microsome GLUT4 or GLUT1. In Fig. 6b, the effects on GLUT4 were quantified, based on densitometry of several experiments. These data show that UBX-Cter significantly increased GLUT4 in plasma membranes and decreased GLUT4 in light microsomes, from unstimulated 3T3-L1 adipocytes. These results were consistent with previous microscopy data suggesting that GLUT4 and TUG colocalize in GSVs, as well as with previous kinetic data suggesting that the TUG UBX-Cter fragment prevents the basal accumulation of GLUT4 in GSVs (15). Indeed, data presented here demonstrated a more marked effect on basal GLUT4 distribution to the plasma membrane. The flow cytometric assay used previously is better suited to kinetic measurements, and the increase in cell surface GLUT4 that was observed was likely underestimated because of the significant background fluorescence inherent in the assay. Data shown here suggest that the fragment not only blocked the retention of GSVs but also further indicate that this was sufficient to target GLUT4 to the plasma membrane and to enhance glucose uptake in unstimulated 3T3-L1 adipocytes.
Confocal microscopy results were consistent with this interpretation. Fig. 7 shows the distribution of stably expressed Mycand GFP-tagged GLUT4 in control 3T3-L1 adipocytes and in cells expressing the dominant negative UBX-Cter fragment. In  control cells, insulin caused the appearance of a homogeneous rim of GLUT4 at the plasma membrane. The incorporation of GLUT4 into the plasma membrane was demonstrated by staining of the externalized Myc epitope tag in the intact cells (second row). In cells containing UBX-Cter, there was clear targeting of GLUT4 to the plasma membrane of unstimulated cells (third row). In these cells, the GLUT4 was incorporated into the plasma membrane, because extracellular Myc tag could be detected. The plasma membrane GLUT4 was more punctate in appearance than in insulin-stimulated control cells, particularly as assessed by GFP, similar to findings with TUG siRNA (Fig. 2). After insulin stimulation, the UBX-Cter cells had a slightly more homogeneous rim of GLUT4 at the cell surface (bottom row). As discussed below, this finding was consistent with the notion that insulin stimulates fusion of vesicles containing GLUT4 with the plasma membrane, which may also account for the 2-fold further increase in glucose uptake upon insulin stimulation of cells containing UBX-Cter. Overall, the data showed that UBX-Cter caused targeting and incorporation of substantial GLUT4 into the plasma membrane. Together with results above, we conclude that the effect of the dominant negative TUG UBX-Cter fragment on GLUT4 was equivalent to that of RNAi-induced TUG depletion. Both caused GLUT4 targeting to the plasma membrane, and this was sufficient to augment glucose uptake in 3T3-L1 adipocytes.
To further test if TUG controls the basal accumulation of GLUT4 in a nonendosomal, perinuclear compartment, we performed confocal microscopy of unstimulated 3T3-L1 adipocytes. Our previous data show that TUG colocalizes with GLUT4, and not with the endosomal marker TfnR, on intracellular membranes of unstimulated 3T3-L1 cells (15). Here, we examined the distributions of GLUT4 and TfnR in the perinuclear region of cells overexpressing TUG or expressing the UBX-Cter fragment. Fig. 8 shows that, in control cells, despite the substantial overlap of GLUT4 and TfnR, membranes that contained GLUT4 and lacked appreciable TfnR were readily visualized (arrowheads, first and second rows). These membranes correspond to nonendosomal vesicles containing GLUT4. When TUG was overexpressed, these GLUT4-positive, TfnR-negative membranes were much more prominent (arrowheads, third and fourth rows). This suggests expansion of a nonendosomal pool of GLUT4 in cells overexpressing TUG. By contrast, when UBX-Cter was expressed, the GLUT4-positive, TfnR-negative vesicles were essentially undetectable (fifth and sixth rows). In these cells, GLUT4 was targeted to the plasma membrane, as noted above ( Fig. 7 and bottom left panel of Fig. 8). Strikingly, the remaining intracellular GLUT4 overlapped completely with TfnR. These findings suggest that UBX-Cter-mediated disruption of TUG function resulted in targeting of GLUT4 to endosomes. Together with data shown above, we conclude that TUG overexpression and TUG UBX-Cter expression had opposite effects. TUG overexpression enhanced the normal accumulation of GLUT4 in intracellular, nonendosomal membranes, concurrent with its sequestration away from the plasma membrane. UBX-Cter expression disrupted this action, so that GLUT4 distribution was similar to that of TfnR-containing endosomes and GLUT4 was targeted to the plasma membrane.
We hypothesized that the effects on GLUT4 targeting we observed may have had secondary consequences for GLUT4 Unstimulated 3T3-L1 adipocytes containing GFP-tagged GLUT4 were starved, fixed, permeabilized, and stained by indirect immunofluorescence to detect transferrin receptor (TfnR), an endosomal marker. Control cells, cells overexpressing intact TUG, and cells expressing TUG UBX-Cter were used as indicated. At the left, the distribution of GLUT4-GFP in representative cells is shown, and arrowheads indicate GLUT4 at the plasma membrane. The indicated perinuclear regions, as well as similar regions from an additional example from each group of cells, are shown at the right. Images of GLUT4 (green), TfnR (red), and merged images are presented. The overlap of GLUT4 and TfnR appears as yellow in the merged images. Arrowheads indicate perinuclear GLUT4 that does not colocalize with TfnR. protein stability. Previous work indicates that GLUT4 is degraded mainly by lysosomes (14). Therefore we considered that TUG-mediated retention of GLUT4 in nonendosomal, intracellular membranes may sequester GLUT4 away from lysosomes, as well as from the plasma membrane, thus prolonging GLUT4 half-life. This possibility was suggested by the decreased GLUT4 we observed after siRNA-mediated TUG depletion in 3T3-L1 adipocytes (Fig. 1c). In this transient experiment, done using fully mature 3T3-L1 adipocytes, potential effects of TUG disruption on 3T3-L1 differentiation would not be expected to play a role. The proposed mechanism may also contribute to the increased GLUT4 observed in stable 3T3-L1 cells overexpressing TUG, as in shRNA ϩ TUG cells (Fig. 3a).
To test this hypothesis, we first sought to determine more definitively if siRNA-mediated TUG depletion affects GLUT4 protein abundance, independent of GLUT4 mRNA abundance. We reasoned that an effect on GLUT4 protein stability might be most dramatic in cells not exposed to insulin, because GLUT4 targeting was most affected by TUG disruption in unstimulated cells. Accordingly, cells were transfected with TUG siRNA B or with a control siRNA, then cultured in the absence of insulin or serum for 24 h before analysis of GLUT4 protein (by immunoblotting) and mRNA (by quantitative, real-time PCR). As shown in Fig. 9a, TUG knockdown caused a dramatic decrease in GLUT4 protein abundance. This was not accompanied by a decreased GLUT4 mRNA amount, which was unchanged as shown in Fig. 9b. The overall decrease in GLUT4 protein amount was ϳ60%, as assessed by densitometry of immunoblots (Fig. 9c). Finally, control experiments showed that the TUG siRNA depleted TUG mRNA and protein similarly, by ϳ80% (Fig. 9, d and e). To facilitate measurement of GLUT4, the experiment shown was done using cells that stably express the GLUT4 reporter, which was ϳ5-fold more abundant than native GLUT4 and contained a Myc epitope tag. Similar results were obtained for endogenous GLUT4 mRNA and protein in 3T3-L1 adipocytes not expressing the reporter. Finally, the decrease in GLUT4 protein was slightly more marked in Fig. 9 than in Fig. 1c, possibly because cells in Fig. 9 were starved for a longer period of time. In conclusion, TUG depletion can cause decreased GLUT4 protein abundance, independent of any change in GLUT4 mRNA amount.
The decreased abundance of GLUT4 protein after TUG depletion could result from decreased GLUT4 translation, increased GLUT4 degradation, or both. As described above, we hypothesized that loss of TUG-mediated retention may enhance GLUT4 trafficking to lysosomes, and consequently increase the rate of GLUT4 degradation. The rate of GLUT4 degradation in lysosomes can be revealed by the increase in GLUT4 abundance after chloroquine addition, because this drug inhibits GLUT4 lysosomal degradation (14). Therefore, to test if dominant negative UBX-Cter fragment accelerates GLUT4 lysosomal degradation, 3T3-L1 adipocytes containing this fragment were treated with chloroquine for various amounts of time. As in Fig. 9, we reasoned that effects might be most dramatic in unstimulated cells. Therefore, to minimize insulin stimulation, cells were cultured in 0.5% fetal bovine serum before and during chloroquine addition. As shown in Fig. 10a, chloroquine treatment modestly increased GLUT4 in control cells, and this was detected after long (14 h) incubation with this drug (top left panel). This finding was consistent with the relatively slow rate of GLUT4 turnover in unstimulated 3T3-L1 adipocytes, as described previously (23). In cells containing UBX-Cter, chloroquine treatment much more markedly increased GLUT4 protein, and this effect was observed as soon as 4 h after its addition (top right panel). Control immunoblots demonstrated that chloroquine had no large effect on TUG itself (bottom panels). Of note, TUG is likely degraded primarily by the proteasome, because it is mostly cytosolic and contains ubiquitin-like domains. Thus, these data support the FIGURE 9. Effects of siRNA-mediated TUG depletion on GLUT4 mRNA and protein abundance. 3T3-L1 adipocytes stably expressing the GLUT4 reporter were electroporated with siRNA duplexes targeting TUG (duplex B) or, as a control, luciferase (which is not present in the cells). Cells were serumstarved beginning 24 h after transfection, and then duplicate plates were lysed in sample buffer (for protein analysis by SDS-PAGE and immunoblotting) or TRIzol reagent (for mRNA analysis by quantitative, real-time PCR). a, triplicate samples were immunoblotted to detect GLUT4 reporter (using anti-Myc antibody), endogenous TUG, and, as a control, ␤-actin, as indicated. Similar results were obtained for endogenous GLUT4 in cells not expressing the reporter. b, samples treated in parallel to those shown in part a were subjected to quantitative real-time PCR to measure the abundance of GLUT4 reporter mRNA, plotted relative to that in control cells. Similar results were obtained for endogenous GLUT4. c, immunoblots of GLUT4 in part a were subjected to densitometry, and the band intensity is plotted relative to control cells. d, quantitative real-time PCR was done to measure the abundance of TUG mRNA and is plotted relative to control cells. e, immunoblots of TUG in part a were analyzed by densitometry, and band intensity is plotted relative to that in control cells. Error bars show standard error, and p values were calculated using a paired, two-tailed t test. The experiment shown was repeated three times with similar results.
To further test if lysosomal GLUT4 degradation is accelerated by shRNA-mediated TUG depletion, and decelerated by TUG overexpression, similar experiments were done with chloroquine treatment of control cells, shRNA cells, and shRNA ϩ TUG cells . Fig 10b shows that, in control cells, chloroquine addition caused a gradual increase in GLUT4 abundance, most marked at 16 h, consistent with previously described rates of GLUT4 turnover (top left panel) (14,23). In shRNA cells, chloroquine induced a larger and more rapid increase in GLUT4, suggesting that lysosomal GLUT4 degradation was more rapid in these cells (top center panel). Conversely, when TUG was overexpressed, in shRNA ϩ TUG cells, the change in GLUT4 abundance was minimal, indicating a reduced rate of GLUT4 lysosomal degradation (top right panel). Control immunoblots showed no effect of TUG depletion or overexpression on Hsc70 abundance, which was slightly decreased at the 16-h time point in all cases (bottom panels). Together, these data are consistent with a model in which TUG normally acts to sequester GLUT4 away from the plasma membrane in the absence of insulin, and that, in doing so, it also prevents GLUT4 trafficking to lysosomes and consequent degradation. TUG depletion or dominant negative fragment disrupted these actions to cause both enhanced glucose uptake and accelerated lysosomal GLUT4 degradation. Reintroduction of TUG in the shRNA cells not only restored highly insulin-responsive glucose uptake but also (because the reintroduced protein was overexpressed) prolonged the half-life of GLUT4 proteins.
To study the interaction of TUG and GLUT4, we produced recombinant TUG and GLUT4 proteins. For GLUT4, the large intracellular loop between transmembrane segments 6 and 7 (TM6 -7) was produced as a GST fusion, and GST alone and GST fused to the TM6 -7 loop of GLUT1 were used as controls. Proteins were incubated together, and then bound TUG was assayed by immunoblotting. As shown in Fig. 11a, intact TUG bound specifically to the loop of GLUT4 and interacted minimally or not at all with that of GLUT1 or with GST alone. On darker exposures of the film, some interaction with the GLUT1 loop was detectable over the GST alone control, yet this was clearly lower affinity than the interaction with the GLUT4 loop. These data are consistent with previous results showing that TUG binds GLUT4, and not GLUT1, in coimmunoprecipitations using transfected cells (15). Similar results using recombinant proteins were also obtained by X. Huang and coworkers (39).
To identify an interacting region on TUG, residues 1-164 or 165-550 were produced in vitro, and incubated with GST alone or fused to the TM6 -7 loop of GLUT4. Fig. 11b demonstrates that a TUG fragment containing residues 1-164, but not 165-550, bound directly and specifically to the GLUT4 loop. These data are consistent with previous findings that deletion of residues 1-164 inhibits the ability of TUG to efficiently retain GLUT4 within cotransfected 293 cells (see Fig. 4b of Ref. 15). Previous data also show that deletion of residues 1-77 has no effect on the ability of TUG to retain GLUT4 intracellularly or to coimmunoprecipitate GLUT4 in transfected 293 cells (Figs. 2a and 4b of Ref. 15). Thus, the GLUT4 loop likely interacts directly with TUG residues 78 -164.
These data strengthen and expand upon previous results, which suggest that two independent interactions may be required to mediate high affinity binding of TUG and GLUT4. Previous data show that forms of TUG containing truncations of the amino terminus to residue 270 are still able to partially FIGURE 10. Effect of chloroquine on GLUT4 accumulation in 3T3-L1 cells. a, control cells and cells containing TUG UBX-Cter were treated with chloroquine for the indicated amounts of time. Cells were lysed, and equal amounts of total protein were analyzed by immunoblotting to detect GLUT4 and TUG, as indicated. The experiment was repeated twice, with consistent results. b, control, shRNA, and shRNA ϩ TUG 3T3-L1 adipocytes were treated with chloroquine as indicated. Cells were lysed, and equal amounts of total protein were analyzed by immunoblotting to detect GLUT4 and Hsc70, as indicated. Similar results were obtained in two independent experiments. retain GLUT4 within 293 cells and are also able to coimmunoprecipitate GLUT4 when both proteins are overexpressed (Fig.  4, b and c, of ref. 15). Therefore one interaction between TUG and GLUT4 was predicted to involve TUG residues 270 -376. This interaction may be direct or indirect, and the required residues in GLUT4 are not known. The second interaction, characterized here, was direct and involved TUG residues 1-164 (and probably 78 -164, as noted above) together with the large intracellular loop of GLUT4. Together, the data are consistent with a model in which both interactions are required for high affinity binding and efficient intracellular retention of GLUT4 by TUG.

DISCUSSION
The data presented here show that TUG is an essential regulator of GLUT4 targeting and of glucose uptake in 3T3-L1 adipocytes. In unstimulated cells, TUG was required to sequester GLUT4 glucose transporters in intracellular membranes. Disruption of TUG function, by RNAi or by a dominant negative fragment, led to marked GLUT4 redistribution out of light microsomes, where both TUG and insulin-responsive GLUT4 reside in unstimulated control cells. The GLUT4 was distributed to the cell surface and enhanced glucose uptake. Insulin had some additional effect to increase glucose uptake in cells with stable disruption of TUG, but this was only ϳ2-fold because of the large effect on basal glucose uptake. Previous data show that insulin stimulates the rapid dissociation of TUG and GLUT4 and that this precedes marked GLUT4 movement out of light microsomes (13,15). Thus, it may be that the insulin-stimulated dissociation of TUG and GLUT4 accounts for a large proportion of the overall action of insulin to enhance glucose uptake.
The precise point at which TUG acts to control GLUT4 distribution remains poorly defined. Previous data are consistent with a role in targeting GLUT4 to GSVs, or in retaining GSVs intracellularly in the absence of insulin, or in both of these processes (15). The data do not exclude the possibility that TUG functions in GLUT4 endocytosis or at some other step in the recycling pathway. Yet, based on previous data, these possibilities seem less likely because: 1) the amount of GLUT4 that was rapidly translocated by insulin correlates with the amount of GLUT4 that dissociated from TUG. Both amounts were increased by TUG overexpression and decreased by TUG UBX-Cter expression. 2) TUG colocalized with GLUT4 in intracellular, light microsomal membranes that are distinct from endosomes. These properties are shared by GSVs. 3) TUG binds with higher affinity to GLUT4 than to GLUT1. This specificity is similar to that described for insulin-responsive translocation (22). Thus, the simplest model to account for previous data is that TUG participates in trafficking of GLUT4 through GSVs and controls the accumulation of a pool of GSVs in unstimulated cells. The new data presented here fully support this model. Other, more complicated models for TUG action are not excluded but are much less likely. Therefore, the present data are discussed in the context of a model in which TUG regulates an insulin-responsive, intracellular GLUT4 pool.
GLUT4 translocation to the cell surface comprises several steps, and insulin may act at one or more points to govern GSV production and disposition. Insulin enhances the production of GSVs from donor membranes by ϳ50%, and this effect may be particularly important for sustained insulin action (9). In unstimulated cells, GLUT4 in these GSVs is retained intracellularly, and resides in a perinuclear location in 3T3-L1 adipocytes. A major action of insulin is to mobilize this pool of GLUT4. The mechanism by which GLUT4 is retained within unstimulated cells is not understood. One possibility is that TUG and other proteins tether GLUT4 in GSVs to some intracellular structure (15). Candidate anchoring proteins to which TUG may link GLUT4 include p115 and Golgin-160 (24,25). Such anchoring proteins are predicted to bind the TUG C terminus, a region distinct from that which binds GLUT4 (15). A second possibility is that, in unstimulated cells, TUG may constrain an intracellular cycle in which GSVs fuse with, and rebud from, endosome or trans-Golgi network membranes (1,11,12,26). Insulin may redirect these vesicles into a plasma membrane recycling pathway by disrupting the TUG-GLUT4 interaction (15,27). A third possibility is that, in the absence of insulin, TUG restricts the association of GLUT4 with kinesin motors, which upon insulin addition carry the transporters toward the plasma membrane (28,29). These possibilities are not mutually exclusive, and present data do not distinguish among them. Present data do show that TUG disruption and insulin stimulation have effects that are similar in magnitude. Thus if TUG mediates the intracellular retention of GSVs, by any of the above mechanisms, then dissociation of TUG and GLUT4 may be an important action by which insulin releases these vesicles to the plasma membrane.
We do not know if the vesicles mobilized to the cell surface by TUG disruption are identical to those mobilized by insulin stimulation. It could be that TUG functions primarily as a sorting adaptor to target GLUT4 to GSVs. Then, in unstimulated cells in which TUG is disrupted, GLUT4 may move to the cell surface in the membranes from which GSVs are normally derived. These could be endosomes or trans-Golgi network membranes. The possibility that TUG is a sorting adaptor is fully compatible with the possibility that TUG more directly mediates basal intracellular retention of GLUT4, perhaps by one of the mechanisms described above. The selective incorporation of GLUT4 into GSVs and the basal, intracellular retention of these vesicles may be processes that are mechanistically coupled.
In addition to acting at an intracellular site to mobilize GSVs, insulin acts at the plasma membrane to accelerate the fusion of vesicles containing GLUT4 (30). Indeed, insulin signals to regulators of SNARE complex formation that are present at the cell surface (31)(32)(33). Disruption of TUG likely does not impact directly on mechanisms controlling vesicle fusion at the plasma membrane, because TUG is not abundant in this location (15). Yet, by preventing the intracellular retention of GLUT4, TUG disruption is expected to increase the flux of GLUT4-containing vesicles arriving at the plasma membrane of unstimulated cells. The data show that this is largely sufficient to place GLUT4 in the plasma membrane and to enhance glucose uptake. Therefore the regulated fusion of GLUT4-containing vesicles at the plasma membrane is one point of control, which may be overcome to a large degree by mobilization of intracellular GLUT4 stores.
The ability of insulin to accelerate the fusion of GLUT4-containing vesicles at the plasma membrane may account for two other observations made here. First, in 3T3-L1 adipocytes in which TUG is stably disrupted by shRNA or UBX-Cter, insulin has a 2-fold further effect to enhance glucose uptake. This effect was not observed after transient, siRNA-mediated depletion of TUG, possibly because there is more variability in this measurement. The residual effect of insulin in stable shRNA or UBX-Cter cells may result from insulin-regulated fusion of vesicles containing GLUT4 (or GLUT1) at the plasma membrane. Effects of insulin on GLUT4 endocytosis may also contribute to this 2-fold effect. Second, a subtle, but reproducible, finding is that TUG disruption results in a more punctate, less homogeneous appearance of plasma membrane GLUT4-GFP compared with that observed in insulin stimulated control cells. Although several explanations for this effect are possible, it seems most likely that this pattern results from an accumulation of GLUT4-containing vesicles that are docked, but not fused, at the plasma membrane. The docked vesicles would mark sites of exocytosis, creating a more punctate appearance along the plasma membrane. In contrast, fusion of the vesicles after insulin stimulation of control cells permits lateral diffusion of GLUT4-GFP in the plasma membrane, so that a more homogeneous rim is observed by confocal microscopy.
The data support that UBX-Cter expression and TUG depletion are functionally equivalent, with respect to both GLUT4 trafficking and glucose uptake, in 3T3-L1 adipocytes. The UBX-Cter fragment is hypothesized to bind an (as yet unidentified) intracellular anchoring site to which intact TUG normally links GLUT4 in unstimulated cells and to exclude intact TUG from this site. It may also interfere with the selective sorting of GLUT4 to GSVs, as described above. Previous data show that UBX-Cter expression blocks the formation of a protein complex containing endogenous TUG and GLUT4 (15). RNAimediated depletion of TUG would also prevent the formation of this complex. Data here show that TUG depletion increases GLUT4 at the plasma membrane, similar to UBX-Cter. Both siRNA transfection and UBX-Cter expression cause incorporation of GLUT4 into the plasma membrane, because the Myc epitope tag is externalized in both cases. Light microsomal GLUT4 is similarly decreased in unstimulated shRNA cells and cells containing UBX-Cter. Finally, both TUG depletion and UBX-Cter expression markedly enhance basal glucose uptake. Thus, UBX-Cter expression and TUG depletion have indistinguishable effects on GLUT4 distribution and glucose uptake, implying that the dominant negative effect of UBX-Cter is equivalent to a null phenotype.
In the presence of insulin, glucose uptake is greater in 3T3-L1 adipocytes in which TUG is disrupted than in control cells in which TUG function is intact. In addition, GLUT4 is more dramatically redistributed by the combined effects of insulin and TUG shRNA or UBX-Cter, as compared with the effect of insulin alone. These findings are consistent with the proposal that insulin acts both to mobilize GSVs and to stimulate fusion of GLUT4 at the plasma membrane and that TUG is required only at the former step. It has previously been shown that maximal insulin stimulation fails to mobilize the entire pool of sequestered GLUT4 in 3T3-L1 adipocytes (27,34). There remains a subset of the GSV pool, corresponding to about a third of the entire population of vesicles, that is not discharged to the plasma membrane. These GSVs continue to be retained intracellularly, possibly by the action of TUG and other proteins. Intriguingly, insulin stimulates the dissociation of two-thirds of the TUG-GLUT4 protein complexes present in basal 3T3-L1 adipocytes and leaves the remaining third of these complexes intact (15). The remaining complexes may retain the GSVs that are not mobilized by insulin. If so, then TUG disruption may have an effect greater than that of insulin, because it may liberate nearly the entire pool of GSVs and not just two-thirds of this pool. The present data show that RNAi-mediated TUG depletion reduces TUG protein by Ն85%, and previous work indicates nearly complete (Ն90%) disruption of TUG-GLUT4 complexes by UBX-Cter expression (15). Consistent with the suggestion that UBX-Cter blocks the sequestration of GLUT4 in GSVs of unstimulated cells, confocal microscopy showed that GLUT4-positive, TfnR-negative membranes were depleted compared with control cells or cells overexpressing TUG (Fig. 8). Thus, it may be that retained GSVs are more quantitatively released by TUG disruption than by insulin and that a larger fraction of total cellular GLUT4 is then available for insertion into the plasma membrane.
The data show that TUG regulates GLUT4 abundance as well as its distribution. Indeed, particularly after transient, siRNA-mediated TUG depletion, the increased glucose uptake was more striking because total GLUT4 was actually decreased. Quantification of GLUT4 mRNA and protein amounts is consistent with the hypothesis that decreased GLUT4 abundance may result from decreased translation or increased degradation, and experiments using chloroquine suggest the latter effect. Overall, the data suggest that by retarding the movement of GLUT4 to the plasma membrane, TUG may also limit the flux of GLUT4 to lysosomes and consequently inhibit its degradation. Other proteins have been suggested to coordinately regulate GLUT4 trafficking and degradation, including sortilin and Ubc9 (14,35). Of note, GLUT4 protein amounts are substantially decreased in IRAP Ϫ/Ϫ mice, although it is not clear if this results from decreased GLUT4 production or enhanced degradation (36). The GLUT4 that remains is distributed normally. It may be that GLUT4 that escapes a sequestration mechanism is rapidly degraded by lysosomes, and not detected in other locations. Such an effect on GSV retention would likely be more prominent in primary cells than in 3T3-L1 adipocytes, because a much smaller proportion of total cellular GLUT4 resides in GSVs in 3T3-L1 cells. Finally, it has been known for many years that GLUT4 degradation is accelerated by insulin stimulation (23,37). Thus, the data shown here suggest that TUG disruption not only mimics much of the effect of insulin on GLUT4 trafficking and glucose uptake but may also mimic the effect of insulin to accelerate GLUT4 degradation.
The results demonstrate that TUG binds directly and specifically to the large intracellular TM6 -7 loop of GLUT4. This region has been implicated in controlling isoform-specific, insulin-responsive GLUT4 trafficking (38). The N-terminal region of TUG that binds this loop in vitro was shown to be important for the ability of TUG to retain GLUT4 intracellularly in transfected cells (15). In addition, the data suggest that TUG and GLUT4 participate in a second interaction, as described above. This involves a more central region of TUG and may be direct or indirect. Both regions of TUG are required for efficient intracellular retention of GLUT4 in transfected cells, and for high affinity binding of TUG and GLUT4 in coimmunoprecipitation experiments. It will be interesting to see if this second interaction involves the GLUT4 N terminus, which is also implicated in insulin responsive trafficking.
The data show that TUG disruption has a small effect to redistribute GLUT1 to the plasma membrane. This effect is much less dramatic compared with GLUT4 redistribution. Yet, because GLUT1 is so abundant in cultured 3T3-L1 adipocytes, the effect on GLUT1 distribution likely contributes to the increased basal glucose uptake we observed. Recent data show that insulin does cause GLUT1 translocation to the plasma membrane of 3T3-L1 adipocytes, although this effect is less dramatic than that of GLUT4 (22). It is notable that the in vitro binding data we present suggest some interaction of TUG and the GLUT1 TM6 -7 loop, however this is clearly much lower affinity than the interaction of TUG and the GLUT4 TM6 -7 loop. Thus, the data are consistent and show a large role of TUG in controlling GLUT4 distribution and, potentially, a much smaller role in controlling GLUT1 distribution in 3T3-L1 adipocytes.
It is not known what other proteins may be present in a TUG-GLUT4 complex or how this complex functions to retain GLUT4 within unstimulated cells. As well, it is not known how insulin acts on the TUG-GLUT4 complex to cause dissociation of TUG and GLUT4 and to mobilize GLUT4. Finally, it is not known if TUG function is altered in insulin-resistant states such as type 2 diabetes. Nonetheless, the data here demonstrate that TUG is an important determinant of GLUT4 trafficking and glucose uptake and suggest that this protein may play a prominent role in insulin action.