Development of an Intracellular Pool of Glucose Transporters in 3T3-L 1 Cells*

The membrane-impermeant bis-mannose photolabel 2-N-4-(l-azi-2,2,2-trifluoroethyl)benzoyl-1,3-bis- (~-mannos-4-yloxy)-2-propylamine (ATB-BMPA) has been used to study the development of an intracellular pool of glucose transporters in 3T3-Ll cells. The subcellular distributions of the transporter isoforms GLUTl and GLUT4 were determined by comparing the cells in which the impermeant reagent only had access to the cell surface and the labeling obtained in digitonin-permeabilized cells. ATB-BMPA labeling showed that only GLUTl was present in preconfluent fibroblasts and that most of the transporters were distributed to the cell surface. In preconfluent fibroblasts, the 2-deoxy-~-glucose transport activity was e6 times higher than in confluent fibroblasts. ATB-BMPA labeling showed that the decrease in transport as cells reached was associated with a decrease in the proportion of GLUTl distributed to the cell surface. The sequestration of these transporters was associated with the development of an insulin-responsive transport activity which increased by ~2.6-fold compared with unstimulated confluent cells. ATB-BMPA

The membrane-impermeant bis-mannose photolabel 2-N-4-(l-azi-2,2,2-trifluoroethyl)benzoyl-1,3-bis-(~-mannos-4-yloxy)-2-propylamine (ATB-BMPA) has been used to study the development of an intracellular pool of glucose transporters in 3T3-Ll cells. The subcellular distributions of the transporter isoforms GLUTl and GLUT4 were determined by comparing the labeling obtained in cells in which the impermeant reagent only had access to the cell surface and the labeling obtained in digitonin-permeabilized cells. ATB-BMPA labeling showed that only GLUTl was present in preconfluent fibroblasts and that most of the transporters were distributed to the cell surface. In preconfluent fibroblasts, the 2-deoxy-~-glucose transport activity was e 6 times higher than in confluent fibroblasts. ATB-BMPA labeling showed that the decrease in transport as cells reached confluence was associated with a decrease in the proportion of GLUTl distributed to the cell surface. The sequestration of these transporters was associated with the development of an insulin-responsive transport activity which increased by ~2.6-fold compared with unstimulated confluent cells. ATB-BMPA labeling showed that insulin stimulation resulted in an e2-fold increase in surface GLUTl so that about one-half of the available transporters became recruited to the cell surface. Measurements of the changes in the distribution of both GLUTl and GLUT4 throughout the differentiation of confluent fibroblasts into adipocytes showed that both transporters were sequestered in parallel. Basal levels of transport and photolabeling remained low throughout the differentiation period when the total pool of transporters (GLUT1 plus GLUT4) was increased by &-fold. These results suggest that the sequestration process was present before new transporters were synthesized. Thus, the sequestration mechanism develops in confluent growth-arrested fibroblasts although the capacity to sequester additional transporters may increase as differentiation proceeds.
3T3-Ll cells have been used for studying insulin action on glucose transport because these cells can be differentiated to produce a reserve of internal glucose transporters which can be translocated to the cell surface in response to insulin. When the cells are grown as fibroblasts, they produce only * This work was supported by the Medical Research Council of the United Kingdom and the British Diabetic Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement'' in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. the glucose transporter isoform GLUTl (Kaestner et al., 1989;Garcia de Herreros and Birnbaum, 1989;Tordjman et al., 1990;Harrison et al., 1990;Reed et al., 1990;Weiland et al., 1990). The fibroblasts are induced to differentiate using isobutylmethylxanthine, dexamethasone, and insulin (Green and Kehinde, 1975;Frost and Lane, 1985). During the course of this differentiation, the cells produce the acutely insulinresponsive isoform GLUT4 (Kaestner et al., 1989;Garcia de Herreros and Birnbaum, 1989;Tordjman et al., 1990;Weiland et al., 1990). It is known from these studies that both GLUTl and GLUT4 mRNA and total cellular protein rise in concentration over 8 days, and, at 8-11 days after initiation of differentiation, the concentrations of these transporters within the cells reach a maximum.
To examine the proportion of newly synthesized transporters which become available at the cell surface, we have extended the use of our bis-mannose photolabel to measure both cell surface and the total cellular pool of transporters. We have previously used this photolabel to measure the cell surface availability of insulin-stimulated adipose cells  and differentiated 3T3-Ll cells (Calderhead et al., 1990;Kozka et al., 1991). We show here that permeabilization of 3T3-Ll cells with digitonin allows the normally impermeant photolabel access to those transporters that are sequestered within the cell. We have therefore compared the labeling at the cell surface and the labeling of the total cellular transporter pool both before differentiation, when the 3T3-Ll cells are confluent fibroblasts, and during the subsequent period of the differentiation regime. The results suggest that mechanisms for sequestration of transporters develop before new transporters are synthesized in the course of differentiation.
The method for examining the distribution of glucose transporters that we have developed here should be a useful addition to the range of techniques that are available for studying transporter processing. The technique has the advantage that it does not depend upon homogenization and separation of membrane fractions and should be useful in studying systems where homogenization procedures are technically very difficult. In addition, it may be useful in investigations of expression systems where transporter cDNA or mRNA are introduced and in which the proportion of the expressed protein that reaches the cell surface needs to be determined.
Immunoprecipitation and Electrophoresis-The detergent-solubilized samples were subjected to sequential immunoprecipitation with 30 pl of protein A-Sepharose coupled with 100 p l of anti-GLUT1 or 50 p1 of anti-GLUT4 antiserum. These antisera were raised against C-terminal peptides as described by . After incubation for 2 h at 0-4 "C and washing of the immunoprecipitates three times with 1.0% and once in 0.1% Cl,E, detergent buffer, the labeled glucose transporters were released from the antibody complexes with 10% SDS, 6 M urea, 10% mercaptoethanol electrophoresis sample buffer and subjected to electrophoresis on 10% acrylamide gels. The radioactivity on the gel was determined by cutting and counting gel slices. The radioactivity in transporter peaks was corrected for a background which was based on the average radioactivity of slices on either side of the peak (Calderhead et al., 1990).

RESULTS
We have examined the rats of 2-deoxy-~-glucose transport in fibroblasts at stages from initial seeding of dishes to the development of a confluent cell monolayer which occurs at about Day 5 after seeding (Fig. 1). One day after initial seeding of the dishes, the transport rate was 0.13 nmol/106 cells/min (with 0.026 X lo6 cells/35-mm dish). This gradually fell over 7 days to 0.025 nmol/106 cells/min (with 0.3 X lo6 cells/35mm dish) and stayed at this low level for at least another 6 days. We have also examined the stage at which the transport rate in fibroblasts becomes insulin-responsive. From about 7 days after seeding, the 2-deoxy-~-glucose transport rate was increased by acute insulin stimulation to ~2 . 5 times the basal level. However, even in the presence of insulin, the transport rate (0.06 nmol/106 cells/min) was approximately one-half the rate which was observed in preconfluent fibroblasts.
In order to determine the subcellular distribution of transporters, we have measured the cell surface availability of transporters with our impermeant photolabel ATB-BMPA and compared this with labeling that was obtained in digitonin-permeabilized cells. Fig. 2 shows the subcellular distribution of GLUTl in confluent cells. No GLUT4 was detected in the fibroblasts. There was an ~4-fold increase in labeling in permeabilized cells compared with untreated cells. Insulin only increased the cell surface labeling by -%fold and did not increase the total pool of cellular GLUT1. The data from three experiments are shown in Fig. 3. This figure also shows that in preconfluent cells there is no significant increase in labeling in the presence of digitonin indicating that most of the cell transporters were at the cell surface.
Together with the transport data in Fig. 1, the labeling data show that in preconfluent 3T3-Ll fibroblasts the high transport rate is associated with the presence of most of the transporters at the cell surface. As the cells reach confluence, transporters are sequestered from the cell surface, and an intracellular reserve pool of transporters develops. Associated with this sequestration is a fall in the transport activity which can then be partially restored by acute insulin treatment which redistributes approximately one-half of the available transporters to the cell surface.
The technique for comparing cell surface and total cellular transporters has also been used to examine the distribution of the acutely insulin-sensitive isoform GLUT4 in differentiated 3T3-Ll adipocytes. Fig. 4 shows an SDS-polyacrylamide gel of ATB-BMPA-labeled and immunoprecipitated GLUT4. As shown in previous studies by Calderhead et al. (1990), Kozka et al. (1991), and Yang et al. (1992), there was very little GLUT4 at the cell surface in the basal state. Upon insulin stimulation, the GLUT4 labeling increased by ~1 2fold in this experiment, However, as is the case with GLUTl in confluent fibroblasts, only about one-half of the total available transporters were returned to the cell surface upon insulin treatment. In the digitonin-permeabilized cells, there was only a slight difference between the amount of GLUT4 labeled in the basal compared with the insulin-treated condition. The small difference observed in this experiment and others (Fig. 5) was possibly due to a difference in efficiency of labeling of the intracellular pool compared with the cell surface GLUT4. This small difference was not observed in all experiments, and, over seven experiments, the mean was only slightly lower than that observed with insulin-treated cells.
However, Fig. 5 also shows that the S.E. was larger for the labeling of the total cellular pool of both GLUTl and GLUT4 in the basal cells. The greater variability of labeling of the intracellular pool (which constitutes a greater proportion of the total in the basal state) was probably a consequence of variations between cell batches in the susceptibility to digitonin permeabilization or the efficiency with which the uv light could penetrate to intracellular sites of transporter localization. Fig. 5 also shows that in the insulin-stimulated state approximately half of the total of both GLUTl and GLUT4 was recruited to the plasma membrane. However, as previously shown by Calderhead et al. (1990) and Yang et al. (1992), the GLUT4 has a greater tendency to be internalized in the absence of insulin, and the cell surface labeling of GLUT4 is about one-third of the GLUTl level in this basal state.
We next examined the time course for changes in the glucose transport activity and subcellular transporter distribution during the transition from confluent fibroblasts to fully differentiated adipocytes (Fig. 6). This figure shows that the basal level of 2-deoxy-D-glucose transport remained low throughout the differentiation period. The transport rate in basal cells was 4 pmol/dish/min in confluent fibroblasts (with 0.3 X lo6 cells/dish) increased to =25 pmol/dish/min in fully differentiated cells where the cell number had also in- creased to 0.6 X lo6 cells/dish. During the differentiation, the amount of protein per 35-mm dish increased enormously from 88 pg/dish to 1.1 mg/dish. The insulin-stimulated rate of transport was approximately double the basal rate for about 4 days after the initiation of differentiation. By 6 days, there was a very marked increase in the insulin-stimulated rate of transport to reach by 11 days, a rate which was "20 times the basal rate. Fig. 7a shows the distribution of GLUTl throughout this differentiation period. During the first 4 days, the cells behaved in a manner similar to that observed in confluent fibroblasts. The basal level of labeling was approximately 2.5- fold lower than the insulin-stimulated level. The total cellular levels were approximately double the level found at the cell surface in the insulin-stimulated state. Between 6 and 11 days, there was a marked rise in the total cellular GLUT1. The cell surface GLUTl in the basal cells rose slightly throughout this period, and this may have been a consequence of the increased cellular level of this transporter. The GLUT4 was produced in parallel with the rise in GLUTl (Fig. 76). Thus, there was little detectable GLUT4 over the first 2 days of differentiation and then this rose steeply between days 6 and 11.
In contrast with the rise in the basal level of labeled GLUT1, labeling showed that the GLUT4 that was newly synthesized was very effectively sequestered inside the cell. The basal GLUT4 labeling remained consistently low from days 4-11 after the initiation of differentiation while the total cellular GLUT4 rose by >lO-fold over this period.

DISCUSSION
The sequestration of glucose transporters in an intracellular pool is likely to be a general mechanism by which cells regulate the supply of glucose to metabolic enzymes in line with the growth demands and cellular metabolism requirements of the cell. The phenomenon of intracellular transporter sequestration was first demonstrated in insulin-sensitive rat adipose cells by Cushman and Wardzala (1980) and Suzuki and Kono (1980) and has subsequently been found in other insulinresponsive tissues such as brown adipose tissue (Slot et al., 1991a), heart muscle (Watanabe et al., 1984;Slot et al., 1991b), diaphragm muscle (Wardzala and Jeanrenauld, 1983)) and skeletal muscle (Klip et al., 1987;Hirshman et al., 1990;Klip et al., 1990). In addition, it has been shown that virus infection of BHK cells (Widnell et al., 1990) leads to a redistribution of internalized GLUTl transporters to the cell surface. Wid-ne11 et al. (1990) have also shown that in BHK cells the sequestered GLUTl transporters redistribute to the cell surface in response to stress stimuli such as arsenite and heat shock. Haspel et al., 1986 have shown that in fibroblasts GLUTl redistribution can occur in response to glucose starvation. Although the cellular redistributions of glucose transporters can be followed by subcellular fractionation and separation of plasma membrane from the light microsome membranes, this is not a technique that can be applied readily to cells and tissues that are difficult to successfully homogenize and fractionate such as skeletal muscle (Klip et al., 1987) and cultured cells (Calderhead et al., 1990). Immunochemical tech- niques (Blok et al., 1988;Slot et al., 1991aSlot et al., , 1991bSmith et al., 1991;Widnell et al., 1990;Tordjman et al., 1990) circumvent the need to obtain subcellular membrane fractions, but these methods are not easily adaptable to investigations of the kinetics of transporter regulation.
As an alternative to these methods, we have extended the use of our impermeant photolabel ATB-BMPA to measure both the cell surface and total cellular levels of transporters. We have shown here that the method when applied to differentiated 3T3-Ll adipocytes gives results which are consistent with those obtained by Gould et al. (1989a) who used Western blotting of isolated membrane fractions to determine that about one-half of the cell GLUTl was at the cell surface in the insulin-stimulated state. The new method has also been used successfully to measure the distribution of GLUTl in BHK cells2 and in CHO cells transfected with the GLUTl gene?
Here we have extended the use of this method to investigate at what stage during growth and differentiation of 3T3-Ll cells the transporter sequestration develops. We initially assumed that the cellular mechanisms necessary for transporter sequestration were most likely to develop during the period of differentiation from fibroblasts to adipocytes. We were surprised to find that GLUTl transporters were down-regulated from the cell surface as the cells reached confluence as fibroblasts. The sequestration mechanisms seem capable of internalizing over 75% of the total GLUTl transporter both in confluent fibroblasts and in differentiated adipocytes. At both these stages, insulin produced a redistribution of about one-half of the available GLUTl to the cell surface. Additional sequestration capability may have also developed as additional GLUTl transporters were synthesized during differentiation. Alternatively, there may have been spare capacity available in fibroblasts that could potentially sequester any new GLUTl or GLUT4 that was introduced into fibro- blasts or produced in subsequent differentiation. Gould et al. (1989b) have shown that 3T3-Ll cells transfected with human GLUT1, that ~5 0 % of both the transfected and endogenous murine GLUTl were sequestered in intracellular membranes, and that both human GLUTl and endogenous GLUTl responded equally to insulin stimulation of translocation.
Other investigators have shown that there is either no change in the basal transport activity throughout differentiation (Kaestner et al., 1989) or that basal activity falls in the first few days after the initiation of differentiation (Weiland et al., 1990;Harrison et al., 1990). Our results show that basal activity can fall markedly if cells are maintained at full confluence for several days without the initiation of differentiation. Thus, the studies cited above can be reconciled if there is a variable time after confluence was reached and before the differentiation regime was initiated. Thus, if the initiation program was started earlier than in our experiments, the basal rate would have fallen during the first few days of differentiation. However, our results suggest that the increase in transporter sequestration and the fall in the transport activity is not a direct consequence of the differentiation regime but develops because cells are in a growth-arrested phase of the cell cycle. Harrison et al. (l990,1991aHarrison et al. (l990, , 1991b and Clancy et al. (1991) have recently described evidence that the intrinsic activity of glucose transporters in 3T3-Ll cells and particularly that of GLUTl is suppressed during differentiation. Their evidence is based on an apparent higher insulin-stimulated glucose transport activity in fibroblasts where GLUTl is present, compared with differentiated 3T3-Ll cells where both GLUTl and GLUT4 are present. Our results, and those of other investigators (Kaestner et al., 1989;Garcia de Herreros and Birnbaum, 1989;Weiland et al., 1990), show, however, that insulin-stimulated transport activity in adipocytes is much higher than in fibroblasts. In addition, however, Harrison et al. (1990,1991a, 1991b) argue that as the binding of their exofacial GLUTl antibody, the &antibody, is increased 2.6-fold in adipocytes in the basal state compared with fibroblasts while the transport activity is reduced by -2.5-fold, the GLUTl intrinsic activity is suppressed by 90%. They suggest that inactivation is due to interaction with an inhibitory protein (Harrison et al., 1991b;Clancy et al., 1991). Our results also show a discrepancy between transport activity and labeled transporter level in comparing fibroblasts with differentiated adipocytes in the basal state but suggest that the discrepancy is smaller than the 90% (10-fold) discrepancy calculated by Harrison et al., 1990. The ratio of transport activity to GLUTl labeling in fibroblasts can be calculated as 130 pmol/1O6 cells/min divided by GLUTl labeling of 970 dpm/106 cells or 0.044 pmol/106 cells (the specific activity is 10 Ci/mmol). This intrinsic activity ratio is 41 pmol/106 cells/ min divided by GLUTl labeling of 590 dpm/106 cells or 0.026 pmol/106 cells in differentiated cells. Thus, during the transition from preconfluent fibroblasts to differentiated adipocytes, the intrinsic activity ratio falls from 2954 min" in fibroblasts to 1576 min-' in adipocytes in the basal state. Any GLUT4 present at the surface of basal adipocytes would lower this ratio. Thus these results suggest that the activity of GLUTl and probably GLUT4 is suppressed by 40-50% or by 1.5-2-fold and are consistent with estimates of the intrinsic activity of GLUTl and GLUT4 previously reported for the basal state Kozka et al., 1991;Palfreyman et al., 1992). Harrison et al. (1991b) have suggested that the exofacial probe that we have used may react only with transporters that are catalytically active. They suggest that in basal cells Time course for the increase in cell surface and total cellular pools of GLUTl and GLUT4 in 3T3-Ll cells. Cells in 35-mm dishes were treated with dexamethasone, isobutylmethylxanthine, and insulin in DMEM with 10% fetal bovine serum to initiate differentiation. On the indicated days, cells were incubated in serum-free medium for 2 h and were then maintained at 37 "C for 30 min either in the absence (sqwres) or presence (triangles) of 100 nM insulin in 1 ml of KRH buffer. In GLUTl ( a ) and GLUT4 (h), distribution was determined by labeling with 100 ICi (closed symbols) or total cellular levels in cells permeabilized with 0.025% digitonin (open symbols). Cells were washed twice in KRH buffer and then solubilized in C& detergent buffer except Day 0 cells which were immediately solubilized in detergent buffer. The isoforms were then immunoprecipitated with anti-C-terminal peptide antibodies and analyzed by electrophoresis to obtain the total disintegrations/ min under the gel peaks. Results are the mean from two experiments. a pool of transporters exists which has suppressed activity due to an inactivating modifier protein. However, several studies using the photolabel (as cited above) have shown an =%fold discrepancy between photolabeling and transport and have shown a lag between the appearance of transporters that can be photolabeled and the increase in transport activity following insulin stimulation of rat adipose cells (Clark et aZ., 1991; Satoh et aZ., 1991) and 3T3-Ll cells (Yang et al., 1992). These studies show that ATB-BMPA binds to transporters that are inactive in transport catalysis. We have suggested that it is this form of the transporter that may be associated with modifier proteins such as those involved in trafficking events in the normal translocation pathway (Yang et al., 1992). In addition we have suggested that where there is a further discrepancy between transporters that are detected by labeling compared with Western blotting, as in isoproterenoltreated insulin-stimulated rat adipocytes' and phenylarsine oxide-treated insulin-stimulated 3T3-Ll cells (Yang et al., 1992), that this is due to the formation of plasma membraneassociated but occluded vesicles. Thus, we, to some extent concur, with the hypothesis described by Harrison et al. (1990Harrison et al. ( , 1991aHarrison et al. ( , 1991b, but suggest that the transport-inactive form at the plasma membrane is a relatively small fraction of the total plasma membrane pool, that these forms represent intermediates in normal transporter trafficking and that translocation is the major mechanism by which glucose transport is stimulated in insulin-sensitive cells. Our experiments in which we have used digitonin suggest that the photolabel interacts with equal amounts of transporter in the basal and insulin-stimulated states and that this provides additional evidence, to that previously described (Calderhead et al., 1990;Holman et aZ., 1990), which suggests that basal and insulin-stimulated transporters do not have differing affinities for the photolabel. If an inhibitory protein were present in basal cells which produced a large increase in the proportion of transporters in a catalytically inactive state, then this association would be expected to reduce the apparent labeling of the total cellular transporter pool of basal cells. Thus, these experiments in which cells are digitonin-permeabilized also suggest that the proportion of transporters with suppressed activity is small and that the low surface levels of transporters detected by labeling of cells in the basal state is due to transporter sequestration within the intracellular pool.
During the process of 3T3-Ll cell differentiation, over 95% of the total cellular GLUT4 in the intracellular pool was sequestered. This is similar to the level of sequestration of GLUT4 seen in other insulin-responsive tissues (Slot et al., 1991a(Slot et al., , 1991b. The greater sequestration of GLUT4 may be related to the targeting of this isoform to separate intracellular vesicles (Zorzano et al., 1989), although Calderhead et al., 1990, have reported that GLUTl and GLUT4 were detected in the same vesicles. It seems likely that the two isoforms leave the cell surface by the same endocytosis route as we have shown that both isoforms are endocytosed with the same half-time when insulin is removed. This has been observed in rat adipocytes (Clark et aZ., 1991) and in 3T3-Ll cells (Yang et al., 1992). However, we have observed that, in insulinstimulated 3T3-Ll cells which are treated with the vicinal dithiol inactivating reagent phenylarsine oxide, there may be some separation of the transporter vesicle processing as GLUT4 re-exocytosis appeared to be blocked by this reagent, but GLUT1 re-exocytosis was not (Yang et aZ., 1992). Cellular mechanisms for the separate localization and processing of GLUT4 must be very dependent on the GLUT4 protein structure as GLUT4 transfected into expression systems tends to become localized inside the cell rather than at the plasma membrane (Gould et al., 1991).
The studies on the use of the photolabel in permeabilized 3T3-Ll cells that we have described here have shown that transporter sequestration can occur before differentiation is initiated and that transporters produced in differentiation probably enter a preformed sequestration process. Further studies will be required to determine whether growth arrest alone is sufficient to induce transporter sequestration or whether cell confluence and cell-cell contact phenomena result in morphological changes in the cell membrane system responsible for down-regulation of transporters and transport activity.