INTRODUCTION

Insulin resistance manifests itself in two pathophysiological disease states, non-insulin-dependent diabetes and obesity. In both of these conditions, adipocytes become desensitized to the biological effects of insulin, which is reflected in a reduction of the efficacy of insulin to stimulate glucose transport. This cell type contains two isoforms of a family of proteins that facilitate transport of glucose across the plasma membrane: GLUT1, the constitutive glucose transporter and GLUT4, the insulin-sensitive glucose transporter. Both of these proteins are integral membrane proteins, which share about 40% sequence identity and have a similar predicted secondary structure. 3T3-L1 adipocytes have been used extensively to study the regulation of these transporters. Derived from mouse embryonic tissue, these cells differentiate in culture from a cell which exhibits a fibroblast phenotype to that of an adipocyte phenotype under the appropriate conditions (1, 2). GLUT1 is present in both phenotypes while GLUT4 is expressed only in the adipocyte phenotype (3, 4). In the adipocyte, GLUT1 is distributed between the plasma membrane and an intracellular vesicular storage site (5-7). GLUT4 under basal conditions resides almost exclusively intracellularly but translocates to the plasma membrane when cells are acutely stimulated with insulin (6-8). With long term exposure to pharmacological doses of insulin, GLUT4 expression (both mRNA and protein) is reduced (9). Mechanistically, this has been ascribed to both down-regulation of transcription and enhanced turnover of mRNA. However, the concentration of insulin required to affect a 50% change in expression of message was reported as 23 nM (9). This level of insulin is at least 2 orders of magnitude higher than physiological (circulating) insulin in humans (10). No comparable dose-response studies have examined the development of insulin-resistant glucose transport activity or GLUT4 expression in these cells. We describe a procedure that has allowed us to measure glucose transport after chronic exposure to physiological concentrations of insulin. Under these conditions, we show that the concentration of insulin required for 50% desensitization of glucose transport activity was 100 pM. Maximal desensitization in the presence of 10 nM insulin was complete within 8 h. In agreement with Marshall and colleagues (11-13), we show that glucose (or glucosamine) is required to establish desensitization. Correlated to the loss in activity was the loss in GLUT4 expression. A 50% decrease in GLUT4 led to a 50% decrease in translocation. GLUT4 mRNA also declined, but the ED₅₀ was approximately 5 nM. These data suggest that post-translational loss of GLUT4 protein likely underlies the cause of desensitization.

EXPERIMENTAL PROCEDURES

Dulbecco's modified Eagle's medium (DMEM)¹ (430-2100 EG) was obtained from Life Technologies, Inc. Fetal bovine serum (1020-75) and calf serum (1100-90) were obtained from Intergen. Bovine serum albumin (A-7030, lot 15H0100) was purchased from Sigma. Polyclonal antibodies were generated against C-terminal peptides of GLUT1 and GLUT4 in our laboratory. Specificity of these antibodies has been previously demonstrated (14).

Cells were grown and differentiated following the procedure of Frost and Lane (15). 35-mm plates (2.1 × 10⁶ cells) were used for glucose transport assays, and 10-cm plates (containing 1.2 × 10⁷ cells) were used for Western and Northern blotting. 24 h prior to the start of the experiment, cells were given fresh DMEM, 10% fetal bovine serum. For kinetic experiments, cells were refed at time 0 with DMEM, 10% fetal bovine serum containing the appropriate insulin concentration as indicated in the figure legends. To define the insulin-dependent dose response, the medium was changed every 2 h to maintain extracellular insulin levels, as 3T3-L1 adipocytes exhibit substantial "insulinase" activity.² To remove insulin prior to the transport assay or translocation analysis, cells were washed with Krebs-Ringer phosphate buffer (KRP) containing 5 mM glucose and 0.1% bovine serum albumin as depicted in Fig. 1. Glucose transport activity was measured as described previously (15). To determine the difference in the dose response to acute insulin after chronic insulin treatment (Fig. 1C), 0.1% defatted bovine serum albumin was included in the assay buffer to maintain the lower concentrations of insulin in solution.

Plasma membrane (PM), high density membrane (HDM), and low density membrane (LDM) fractions were isolated by a modification (16) of a technique described by Weber et al. (17). Briefly, control or insulin-treated cells were scraped into TES (10 mM Tris-HCl, pH 7.4, 1 mM EDTA, and 250 mM sucrose) at 18 °C. The cells were passed over a tungsten ball 10 times in a steel block homogenizer (at 4 °C) with a clearance of 0.0025 inch. A crude plasma membrane fraction was collected at 17,000 × g for 15 min at 4 °C. Purified membranes were collected from this fraction by sucrose gradient centrifugation (17). HDM and LDM fractions were collected by differential centrifugation (17). Alternatively, a total membrane fraction was collected by centrifugation of the homogenate at 212,000 × g for 70 min. Membrane fractions were stored in TES at 20 °C. Protein was determined by the method of Markwell et al. (18).

Equal amounts of membrane protein were mixed with Laemmli sample dilution buffer (19) containing 6 M urea and 10% -mercaptoethanol before separation by SDS-polyacrylamide gel electrophoresis and transfer to nitrocellulose as described previously (20). Immunoblot analysis of GLUT1 and GLUT4 utilized antibody dilutions of 1:500 using the enhanced chemiluminescence system as described previously (20). The bands were quantitated by video densitometry on a Visage bioscan (Millipore) in the linear range of the film and peroxidase reaction.

RNA isolation and Northern blotting were performed as described previously (20). Twenty µg of total RNA were loaded onto a 1% formaldehyde-agarose gel and transferred to nylon membranes. The membranes were probed with cDNAs for GLUT1 and GLUT4 (generously provided by Dr. Maureen Charron, Albert Einstein). The results were quantitated by video densitometry in the linear range of the film.

RESULTS

To validate 3T3-L1 adipocytes as a model for analyzing insulin resistance, it was imperative to reestablish basal transport after chronic insulin treatment. To this end, we first treated cells with either 10 nM or 1 µM insulin for 12 h, a time frame defined in isolated rat adipocytes as sufficient to complete the desensitization process (11). We then washed the cells in KRP containing 5 mM glucose and 0.1% defatted bovine serum albumin at 20-min intervals over a 140-min time course. At specific times during this washout period, cells were rinsed in glucose-free KRP to assess transport activity in a 2-min pulse. Fig. 1A shows the comparison between control cells (washed in an identical manner) and those treated chronically with either 1 µM or 10 nM insulin. Cells treated chronically with 1 µM insulin showed significantly elevated transport at the start of the washout (time 0) but never achieved basal values over time despite the extensive washing. In contrast, cells treated with 10 nM insulin returned to basal values within 60 min of the initial removal of insulin. These latter cells when subsequently rechallenged with 1 µM insulin (after the 60-min wash) showed a 50% reduction in the rate of glucose transport in comparison with control cells (Fig. 1B). In addition to the decreased rate, the cells were less sensitive to insulin in that the dose response to acute insulin challenge was shifted to the right by an order of magnitude (Fig. 1C). As far as we know, this is the first demonstration of insulin-resistant glucose transport activity in 3T3-L1 adipocytes because of the ability to reinitiate stimulation from a "true" basal state.

To determine if physiological insulin could establish the insulin-resistant state, we exposed cells to specific concentrations of insulin for 12 h. We refed cells every 2 h to maintain the extracellular insulin levels, particularly important at the lower concentrations of insulin. As shown in Fig. 2A, the concentration of insulin that elicits a 50% reduction in insulin-sensitive glucose transport was approximately 100 pM. This is extremely interesting because the fasting level of insulin in non-diabetic humans is about 40 pM while that in the obese individuals is about 70 pM and in individuals with non-insulin-dependent diabetes is about 200 pM (21). Insulin as low as 1 nM was sufficient to completely desensitize the transport system. At 10 nM, maximal desensitization was achieved whether the cells were refed every 2 h or not. Thus, we chose a concentration of 10 nM to examine the time required for the development of insulin resistance. As shown in Fig. 2B, the phenomenon of densensitization was completely established within 8 h of the initial exposure to insulin. This result is similar to that described in isolated adipocytes (12).

Marshall and his colleagues (11-13) have shown in a series of elegant experiments the requirement of glucose and glutamine, as well as insulin, for the expression of insulin resistance in isolated rat adipocytes implicating the N-acetylglucosamine biosynthetic pathway in this phenomenon. To test if the same is true in 3T3-L1 adipocytes, we performed similar experiments. One complication in our experiments that was not encountered in isolated adipocytes is the time-dependent activation of glucose transport activity in the absence of glucose (20, 22-24). We hypothesize that this difference between rat adipocytes and 3T3-L1 adipocytes results from higher glycogen stores in the former (25, 26) compared with 3T3-L1 adipocytes (14), which might provide a metabolic buffer from external glucose deprivation. We therefore minimized the time that cells were exposed to glucose-free medium but suffered in that only 75% of maximal desensitization was achieved in these experiments. Importantly, though, the basal rates of transport were not affected such that true resistance could be evaluated. Fig. 3 shows that in the absence of glucose, insulin was unable to induce desensitization. Either glucose (in the presence of glutamine, Fig. 3A) or glucosamine (in the absence of both glucose and glutamine, Fig. 3B) provided appropriate substrate for the development of the insulin-resistant state.

A total membrane fraction revealed that insulin-resistant cells (i.e. cells exposed to 10 nM insulin for 12 h) expressed 2.4-fold less GLUT4 than control cells while GLUT1 increased by 2.2 (data not shown). To examine the distribution of these changes, we took advantage of a membrane isolation technique recently developed in our laboratory (16). Fig. 4 shows the distribution of GLUT4 and GLUT1 among three membrane fractions (PM, LDM, and HDM) in control and insulin-resistant cells. Each set went through the washout procedure prior to membrane fractionation. Control cells, which were stimulated acutely with 1 µM insulin, showed redistribution of both GLUT4 and GLUT1; GLUT4 increased by about 6-fold in the PM (Fig. 4, A and B) while GLUT1 increased by about 2-fold (Fig. 4, C and D). These data are similar to those that analyzed translocation using the cell surface photolabel, ATB[2-³H]BMPA (2-N-4-(1-azi-2,2,2-trifluoroethyl)benzoyl-1,3-bis(D-mannos-4-yloxy)-2-propylamine) (27). Cells chronically exposed to insulin (followed by washout) showed levels of GLUT4 in the plasma membrane about equal to that of controls (Fig. 4, A and B). However, upon acute insulin challenge, translocation was reduced by about 50% compared with controls, which correlates with the loss of insulin-sensitive glucose transport activity. Cells chronically exposed to 10 nM insulin showed a 2-fold increase in the level of GLUT1 in the PM after washout compared with controls (Fig. 4, C and D), despite the equivalent rates of glucose transport. Acute insulin challenge stimulated translocation but to a much more limited degree than in control cells.

As the LDM fraction reflects the loss of GLUT4, we used this fraction to examine the dose-dependent loss in cells chronically treated with specific concentrations of insulin. As shown in Fig. 5A, the level of GLUT4 decreased over time in response to increasing insulin. The dose dependence of this down-regulation (Fig. 5B) was similar to that of insulin-resistant glucose transport (see Fig. 2A).

Based on the observation that glucose deprivation prevented the loss in insulin sensitivity (see Fig. 3A), we examined the expression of GLUT4 in the LDM fraction of cells exposed to glucose-free medium. Fig. 6 shows that glucose deprivation blocks the loss of GLUT4 in chronically treated cells. Thus we show for the first time that glucose is important in regulating the expression of GLUT4 in response to chronic insulin.

To evaluate the underlying mechanism of the reduction in GLUT4 protein, we measured the level of GLUT4 mRNA after exposure to specific concentrations of insulin. As shown in Fig. 7A, the level of GLUT4 decreases with increasing insulin concentration. However, the concentration of insulin required to elicit a 50% loss of GLUT4 mRNA was about 5 nM (Fig. 7B), which is 15 times greater than that required for equivalent loss of insulin-sensitive glucose transport activity or GLUT4 expression.

DISCUSSION

In this study, we have tested the hypothesis that 3T3-L1 adipocytes can serve as a model for studying the development of insulin resistance under conditions that might be realized in a physiological setting. Support for this hypothesis has been gained from the following observations. Chronic exposure to physiological levels of insulin decreased the ability of an insulin challenge to stimulate glucose transport. Interestingly, postprandial concentrations of insulin in normal, obese, and diabetic humans (21) plot along the inflection in the dose-response curve between no change in insulin responsiveness and that of maximal resistance. Thus, we have shown for the first time that 3T3-L1 adipocytes develop insulin resistance in response to physiologically relevant concentrations of insulin. We have extended previous work by demonstrating that insulin challenge of resistant cells promotes translocation, although the extent of recruitment is suppressed relative to controls due to the reduction in the total expression of GLUT4. Second, we have shown that glucose deprivation, which prevents the development of insulin-resistant glucose transport, also prevents the loss in GLUT4. Together, these data suggest that the loss of GLUT4 protein underlies the inability of 3T3-L1 adipocytes to respond to insulin after chronic exposure. This mimics the clinical manifestation of human obesity and non-insulin-dependent diabetes where loss of GLUT4 protein has been observed in adipose tissue (28), although not muscle (29).

It should be pointed out that transporter expression differs in adipose tissue relative to 3T3-L1 adipocytes. In isolated rat adipocytes, GLUT4 represents 97% of the GLUT transporter pool (30). In 3T3-L1 adipocytes, GLUT4 represents only 33% of the pool (6) indicating the substantially higher expression of GLUT1 relative to GLUT4 in this cell line. In control 3T3-L1 adipocytes, the PM fraction contains about 25% of the GLUT1 pool. Chronic insulin treatment increases the total pool of GLUT1, which in turn doubles the GLUT1 content of the PM fraction. Despite this 2-fold increase in GLUT1 in the PM of resistant cells, we observed no difference in "basal" transport activity (after washout) compared with controls. Resistant cells treated acutely with insulin show little additional change in GLUT1 in the PM. This argues that GLUT1 plays but a small role in insulin-resistant glucose transport. In contrast, only 3% of the GLUT4 pool resides in the PM of either control or resistant cells (again, after washout). When insulin is added acutely, GLUT4 content in the PM reveals significant translocation; resistant cells show 50% that of controls reflecting the difference in the total pool. To reiterate, this suggests that GLUT4 expression determines insulin resistance, even with the elevated levels of GLUT1 in the 3T3-L1 adipocyte cell line.

There are some differences between our data and those reported previously on 3T3-L1 adipocytes. Flores-Riveros et al. (9) reported that the concentration of insulin required to reduce GLUT4 mRNA was 23 nM in contrast to the value we calculated, which was about 5 nM, close to the K_d for the insulin receptor (31). This difference can be explained by our refeeding protocol during chronic insulin treatment to maintain the level of extracellular insulin, particularly important at low hormone concentration, in the face of extensive degradation by these cells. Our cells were exposed as well to insulin for only 12 h compared to the 24-h exposure in the Flores-Riveros study (9), which further lessens the impact of insulin degradation. This temporal difference (24 versus 12 h of insulin treatment) also accounts for the smaller magnitude of the increase in GLUT1 and decrease in GLUT4 in our study relative to previous studies (9, 32). Importantly, the down-regulation of GLUT4 mRNA occurs at insulin concentrations that are not likely to persist in the physiological state. These insulin concentrations also do not correlate with those required for the development of densensitization. Thus, GLUT4 expression appears to be regulated transcriptionally, but this regulation may not be relevant to insulin resistance.

Other studies in 3T3-L1 adipocytes have shown varying results in transporter protein expression. Tordjman et al. (33) and Kozka et al. (27) showed that chronic insulin treatment did not affect total GLUT4 protein expression while the studies of Flores-Riveros et al. (9) and Clancy and Czech (32) showed a marked decrease. These latter data along with ours are consistent with the accelerated turnover of GLUT4 in the presence of chronic insulin as measured by Sargeant and Paquet (34). Ricort et al. (35, 36) showed a small decrease in the expression of GLUT4 but also very little translocation to the PM with acute insulin stimulation. These authors interpreted their data to mean that GLUT4 translocation was blocked, which clearly differs from our studies. Kozka et al. (27) interpreted their cell surface ligand binding experiments similarly, even though they demonstrated a 50% reduction in cell surface GLUT4, which would agree with our studies. We can, of course, only speculate as to the cause for the different results. In both of these latter studies, the loss in the GLUT4 pool was determined by analyzing homogenate protein, revealing only modest changes in expression. As the translocatable GLUT4 resides in the LDM fraction, it may be that the loss was substantially underestimated as pharmacological concentrations of insulin were used to induce resistance. Neither study separated the LDM fraction from the HDM fraction; thus we cannot evaluate this possibility. Finally, it is important to point out that our experiments are the first to show that basal transport activity can be achieved after chronic insulin treatment, which allowed us to evaluate true insulin resistance. Data collected under these conditions are consistent with the hypothesis that the onset of insulin resistance (i.e. depressed insulin-sensitive glucose transport) is a reflection of the reduced GLUT4 pool, not a defect in translocation.