Use of Bismannose Photolabel to Elucidate Insulin-regulated GLUT4 Subcellular Trafficking Kinetics in Rat Adipose Cells EVIDENCE THAT EXOCYTOSIS IS A CRITICAL SITE OF HORMONE ACTION*

The subcellular trafficking of tracer-tagged GLUT4 between the plasma membranes and low-density microsomes of rat adipose cells has been studied. Cell-sur-face GLUT4 have been initially tracer-tagged in the insulin-stimulated state with the [SH]bismannose photolabel 2-N-4-(1-azi-2,2,2-trifIuoroethyl)benzoyl-1,3-bis-(~-mannos-4-yloxy)-2-propylamine. The half-time for internalization of tracer-tagged GLUT4 when insulin is removed by collagenase treatment is similar to that observed for the decrease in immunodetectable GLUT4 in the plasma membranes and the decrease in glucose transport activity in the intact cells. In contrast, inter- nalization of tracer-tagged GLUT4 also occurs when cells are maintained in the continuous presence of insulin even though the plasma membrane level of immuno- detectable GLUT4 and glucose transport of internalization. In our studies, we have used KCN to prevent cycling during the processing of the cells and have observed only 5% of labeled GLUT4 in the low-density microsomal fraction at zero time.

The subcellular trafficking of tracer-tagged GLUT4 between the plasma membranes and low-density microsomes of rat adipose cells has been studied. Cell-surface GLUT4 have been initially tracer-tagged in the insulin-stimulated state with the [SH]bismannose photolabel 2-N-4-(1-azi-2,2,2-trifIuoroethyl)benzoyl-1,3-bis-(~-mannos-4-yloxy)-2-propylamine. The half-time for internalization of tracer-tagged GLUT4 when insulin is removed by collagenase treatment is similar to that observed for the decrease in immunodetectable GLUT4 in the plasma membranes and the decrease in glucose transport activity in the intact cells. In contrast, internalization of tracer-tagged GLUT4 also occurs when cells are maintained in the continuous presence of insulin even though the plasma membrane level of immunodetectable GLUT4 and glucose transport activity in the intact cells are unaltered. These data show, for the first time, that insulin has little, if any, effect on the rate constant for GLUT4 endocytosis, but instead, primarily increases the rate constant for exocytosis. 'Ikacer-tagged GLUT4 that is returned to the low-density microsomes can be restimulated with fresh insulin to recycle to the plasma membranes and to a steady-state distribution level that is the same as that observed in cells that are maintained in the continuous presence of insulin. These data suggest that the cells' entire complement of GLUT4 is invorved in the recycling process. Following insulin stimulation of adipose cells initially in the basal state, the increase in immunodetectable GLUT4 in the plasma membranes precedes the increase in accessibility of GLUT4 to exofacial2-N-4-fl-azi-2,2,2-trifluoroethyl)ben-zoy~-1,3-bis(~-mannos-4-y~oxy)-~-propy~amine photolabeiing, and this in turn precedes the increase in cellular glucose transport activity. Such time course data sug gest that there may be plasma membrane intermediate states in the GLUT4 trafficking pathway. The kinetic properties of GLUT4 translocation and its recycling Insulin is now known to stimulate glucose transport in isolated rat adipose cells through a mechanism involving the translocation of glucose transporters, primarily the GLUT4 isoform, from a Iarge intracellular pool to the plasma membrane ( 1 4 . In rat adipose cells, insulin has been shown to produce increases and corresponding decreases in the concentrations of ghcose transporters in the plasma membranes and low-density microsomes, respectively. This translocation has been clearly shown to be rapid, reversible, insulin concentration-dependent  (9) and heart muscle (10). Slot's group has shown that in brown adipose tissue in the basal state, GLUT4 are primarily localized to "the trans-GoIgi reticulum and tubdovesicular structures elsewhere in the cytoplasm," with ~1 % associated with the plasma membrane; after insulin stimulation, -40% of the glucose transporters are located at the cell surface, and additional GLUT4 are enriched in the early endosomes (9). However, Smith et al. (11) found a smaller, -12-fold, change in the redistribution of GLUT4 to the plasma membrane of white adipose cells.
Immunocytochemical studies suggest that glucose transporter trafficking is a n insulin-regulated process similar to the receptor-mediated endocytosis that occurs in the processing of receptors such as the transferrin, asialoglycoprotein, and insulin receptors (12-15). However, measurements of the steadystate distribution of glucose transporters (1-4,8-11) leave several mechanistic questions unanswered. For example, only kinetic studies and not steady-state distribution measurements can be used to determine whether glucose transporters continually recycle in the basaI and insulin-stimulated states.
It is not known whether the whole intracellular pool of glucose transporters is available for cycling to the plasma membrane. In addition, kinetic studies are needed to determine whether insulin regulates endocytosis or exocytosis.
To address these issues, it is necessary to track individualglucose transporter molecules as they recycle between subcellular compartments. Recently, we have developed a procedure for specifically labeling the plasma membrane pool of glucose transporters with the membrane-impermeant bismannose photolabel ATB-BMPA.l Using this technique, we have shown that insulin produces a 15-20-fold increase in cell-surface GLUT4 and a 3-5-fold increase in cell-surface GLUT1 in both rat adipose cells (4,16) and 3T3-Ll cells (17)(18)(19). We have previously used this photolabel to examine the kinetics of glucose transporter appearance and loss from the cell surface of rat adipose cells (16), but in the study described here, we have used the probe to examine the kinetics of glucose transporter recycling between subcellular membrane fractions and have kinetically analyzed the possible sites of insulin action within this cycle. The determination of the rates of internalization of photolabeled glucose transporters in the presence and absence of insulin, when combined with an analysis of time courses for stimulation of cycling by insulin, has allowed the construction of a kinetic scheme in which docking and fusion steps in membrane vesicle exocytbsis are identified as important intermediate steps in the trafficking pathway and critical sites of hormone action.
Preparation of Rat Adipose Cells-Male rats (CD strain, Charles River Breeding Laboratories, Inc.) were used. They were fed standard NIH chow ad libitum for at least 5 days prior to study. T h e rats were anesthetized with a gas mixture of 70% CO,, 30% 0, and killed by decapitation between 7 and 8 a.m. The epididymal fat pads were removed, minced, and digested with collagenase as previously described (21). All incubations were carried out at 37 "C in a Krebs-Ringer bicarbonate HEPES buffer, pH 7.4, containing 10 m~ sodium bicarbonate, 30 m M HEPES, 200 n~ adenosine (KRBH buffer), and 1% (w/w) bovine serum albumin.
Photolabeling with ATB-BMPA-Adipose cells were suspended at a 40% cytocrit in 1% albuminlKRBH buffer at 37 "C and stimulated with 0.7 n~ insulin for 30 min. 3-ml samples of cells were added to 500 pl of KRBH buffer without albumin and containing 250 pCi of ATB-[2-3H]B-MPAin 55-mm polystyrene dishes. The dishes were irradiated (with the polystyrene lids on the dishes) for 1 min using a Rayonet photochemical reactor with 300 nm lamps. After photolabeling, either the cells were maintained in the continuous presence of insulin for 60 min, or insulin was removed by incubation with 0.2 mdml crude collagenase for 60 min (5). In studies of restimulation, insulin was readded to the collagenasetreated cells to give a final concentration of 0.7 PM, and then this incubation was continued for an additional 30 min. At the times indicated in the figure legends, 2 m M KCN was added to arrest glucose transporter recycling (5), and the cells were transferred to a 9-ml centrifuge tube with a small volume of TrialEDTAlsucrose buffer (20 m M Tris-HC1, 1 m M EDTA, 255 m M sucrose, pH 7.4), spun through 1 ml of dinonylphthalate oil, resuspended in TrialEDTNsucrose buffer, and immediately homogenized. Plasma membrane, high-density microsomal, and low-density microsomal fractions were then prepared as previously described (22).
were photolabeled with 3H prior to assay, we have used the isotopedilution oil flotation method (231, which systematically gave a 2-3-fold lower estimated transport rate irrespective of whether cells had been irradiated. Half-times for stimulation or reversal of glucose transport activity were calculated from linear regression analyses of -ln(lf ) uersus time or In f uersus time, respectively, where f i s the fraction of the maximum change.
Quantitation of Glucose Dansporters-The amount of GLUT4 in each membrane fraction was assessed by SDS-PAGE, gel transfer, and Western blotting using an anti-GLUT4 COOH-terminal peptide antiserum and 1251-protein A (24). The amount of 3H-photolabeled GLUT4 in each membrane fraction was determined by immunoprecipitation using the same anti-GLUT4 antiserum. To obtain the immunoprecipitates, membrane fractions were first solubilized in 600 pl of 2% C12Ee detergent buffer containing 1 m M phenylmethylsulfonyl fluoride, 5 m~ EDTA, 1 pg/ml pepstatin A, and 2 m M N-ethylmaleimide in phosphatebuffered saline. Following incubation for 30 min at room temperature, the detergent-solubilized material was centrifuged at 20,000 x g, , and the supernatants were mixed with a pre-equilibrated conjugate of 20 pl of antiserum and 60 pl of protein A-Sepharose. The samples were then mixed at 4 "C for 90 min. The pellets were subsequently washed twice with 600 pl of the same detergent buffer as described above but containing 0.2% C12Ee and then once with phosphate-buffered saline. Finally, conjugate was released from the gel matrix with electrophoresis sample buffer. SDS-PAGE analysis of the immunoprecipitated material was followed by gel slicing and counting (4). Western blotting and photolabeling results in each experiment were expressed per cell by first determining the respective lZ5I and 3H disintegrationdminutdunit weight of membrane protein loaded in each gel lane and then by multiplying by the total amount of membrane protein recovered in each respective membrane fraction from the starting samples. This was then divided by the number of starting cells in each sample determined from weighed samples of extracted triglyceride and by using previously determined typical triglyceride weights per adipose cell (25). No correction was made for immunoprecipitation efficiency, although it was checked regularly and found to be reproducibly 435%. Protein was determined by the bicinchoninic acid assay (Sigma) using crystalline bovine serum albumin as the standard. Half-times for the changes in both immunodetectable GLUT4 and ATB-[2-3HlBMPA-tagged GLUT4 were calculated as described above for changes in glucose transport activity.
Analysis of Endocytic and Exocytic Rate Constants-To obtain initial estimates of the endocytic and exocytic rate contents, we assumed the existence of only two pools of glucose transporters, one in the low-ATB-BMPA-tagged GLUT4 equilibration would then be as follows density microsomes and one in the plasma membranes. The rate of (Equation 1): where T, is the fraction of the glucose transporters in the plasma membrane, and ken and k. , are the endocytosis and exocytosis rate constants, respectively (14,26). Integration with T,, = T,,, at t = 0 gives Equation 2.

(Eq. 3)
Equations 2 and 3 can be used to analyze label redistribution under steady-state and non-steady-state conditions and is independent of the specific activity of the labeled glucose transporters; all that is required is that the total pool of glucose transporters is conserved. Under nonsteady-state conditions, such as that occurring following insulin removal, the use of these equations assumes that k,, and k,, are instantaneously changed. In the insulin restimulation experiments, T, , in Equation 2 is obtained directly as the fraction of label remaining in the plasma membranes following reversal. Equation 2 can also be used to analyze the non-steady-state redistribution of glucose transporters as detected by Western blotting. In the analysis of these data, T, is the initial fraction of total immunodetectable GLUT4 in the plasma membranes. The rate constants were calculated by least-squares fitting (weighted for proportional error) to Equations 2 and 3 using the Fig P  software (Biosoft).

RESULTS
Reversal of Effect of Insulin on Glucose Dansport Activity "Insulin at a concentration of 0.7 nM is just sufficient to stimulate glucose transport activity to a maximum steady-state level in 30 min, with a tlI2 of 3-5 min as previously reported (data not shown) (27). To reverse this effect, we have added collagenase to digest the residual insulin as originally described by Kono et al. (5). Fig. l shows the results of an experiment in which the decrease in glucose transport activity over 60 min has been studied at collagenase concentrations from 0.1 to 5 mg/ml so that we could find a n optimum concentration that would allow rapid reversal of the insulin response but would not itself stimulate this activity. 0.2 mg/ml collagenase restores the glucose transport activity to the basal state in -60 min, consistent with the reversal obtained using alternative methods for reversing the insulin response including successive washes (20) and a low pH washing buffer (281, but somewhat faster than anti-insulin antiserum (27). Fig. 1 also shows that collagenase concentrations >0.5 mg/ml tend to raise basal glucose transport activities. Fig. 2 illustrates not only the detailed time course of the reversal of insulin-stimulated glucose transport activity with 0.2 mg/ml collagenase and the stability of the maximum glucose transport response in adipose cells continuously exposed to 0.7 nM insulin, but also the time course of restimulation of glucose transport activity with readdition of insulin to cells orginally stimulated by insulin and then reversed with collagenase. The decrease in insulin-stimulated glucose transport activity with collagenase occurs with a tuz of 11.8 * 0.8 min (mean * S.E., n = 3), while the stimulated activity observed in the continuous presence of insulin remains highly stable for at least 60 min. Addition of 0.7 p~ insulin to cells that have been reversed with collagenase results in a restimulation of glucose transport activity to the maximum level observed in cells maintained in the continuous presence of insulin throughout the entire 120-min incubation. The t l / 2 of the latter restimulation with 0.7 PM insulin (4.6 * 1.0 min) is virtually identical to that of the initial stimulation with 0.7 nM insulin noted above. Addition of 0.7 insulin to cells continuously exposed to 0.7 nM insulin has no detectable further effect.
Because the data in Fig  irradiation on the time courses of the initial stimulation of glucose transport activity by insulin, the reversal of insulinstimulated glucose transport activity by collagenase, and the restimulation of glucose transport activity by readdition of insulin after collagenase reversal were examined. Indeed, insulin stimulation was added specifically to the experimental protocol to determine if the rate of translocation of internalized photolabeled GLUT4 from the low-density microsomes to the plasma membranes in response to insulin would be influenced by the presence of covalent photolabel (see below). UV irradiation under the conditions employed here does not detectably influence the tU2 for either the reversal by collagenase or the restimulation by insulin readdition, but does increase the tlI2 for the initial stimulation by insulin by -30% when the latter immediately follows irradiation (data not shown).
Recycling of Glucose Dansporters between Subcellular Membrane Fractions- Fig. 3 illustrates an SDS-PAGE analysis of the subcellular distribution of cell-surface ATB-BMPA-labeled GLUT4 in rat adipose cells under various steady-state conditions. For cells in the basal state, such an analysis shows that photolabeled and immunoprecipitated GLUT4 in the low-density microsomal fraction (Fig. 3B) is only -2% of that observed in the plasma membrane fraction of fully insulin-stimulated cells (Fig. 3A). Similarly, the insulin-treated cells that have been immediately treated with 2 mM KCN to arrest the recycling process show low-density microsomal levels of ATB-BM-PA-tagged GLUT4 that are only 5% of those observed in the plasma membranes. The exclusion of the photolabel from the low-density microsomal pool in the basal state is due to the impermeable property of these photolabeling reagents (29). The low level observed in the insulin-stimulated state reflects the extent to which plasma membranes are excluded from the lowdensity microsomes during the subcellular fractionation procedure (22). GLUT4 that occurs in the plasma membrane fraction when insulin-stimulated cells are treated with collagenase is accompanied by an increase in ATB-BMPA-tagged GLUT4 in the low-density microsomal fraction. Fig. 3B shows that the amount of photolabeled GLUT4 recovered in the low-density microsomal fraction increases over 60 min to a level that is 15-fold higher than that observed in the sample from fully insulin-stimulated cells that have been immediately treated with KCN. In the low-density microsomal fraction isolated from cells maintained in the continuous presence of insulin, the ATB-BMPA-tagged GLUT4 recovered is 9-fold higher than the initial value. The distribution of photolabeled GLUT4 among all three membrane fractions, determined from six experiments, is shown quantitatively in Table I. When the results are expressed in this quantitative fashion (see "Experimental Procedures"), any changes in the amount of ATB-BMPA-tagged GLUT4 in the plasma membrane fraction are fully accounted for by reciprocal changes in the low-density microsomal fraction, thus demonstrating a true transfer process. Table I also shows that only small amounts of cell-surface ATB-BMPA-labeled glucose transporter cofractionate with the high-density microsomal membrane fraction (from two experiments). Fig. 4A demonstrates the results obtained by Western blotting plasma membrane fractions isolated from adipose cells that were incubated according to the experimental protocol described in the legend to Fig. 2. The plasma membrane fractions from cells maintained in the continuous presence of insulin retain high levels of immunodetectable GLUT4, as would be expected from the observed maintenance of a constant cellular glucose transport activity throughout this time (Fig. 2 6.9 f 0.6 (6.6, 6.8) genase to remove insulin lose immunodetectable GLUT4 with a tY2 of 13.1 * 1.3 min and reach a level that is 17 * 2% of the initial value observed in the fully insulin-stimulated cells. The rate of decrease in immunodetectable GLUT4 in the plasma membrane fraction is similar to the rate of decrease in glucose transport activity noted above (tllz = 11.8 min) (Fig. 2). Insulin readdition restores the level of immunodetectable GLUT4 to the normal, fully stimulated level with a tug of -2.7 * 0.2 min (Fig. 4A), which is slightly faster than the increase in glucose transport activity that occurs with a tll2 of 4.6 min as noted above (Fig. 2). In contrast to the results obtained by immunoblotting GLUT4, the loss ofATB-BMPA-photolabeled GLUT4 occurs not only during insulin removal with collagenase, but also in adipose cells maintained in the continuous presence of insulin (Fig. 4B). In the plasma membrane fraction from the collagenase-treated cells, tracer-tagged GLUT4 decreases with a tu2 of 9.4 * 0.8 min to reach a fully reversed level at 17 -+ 2% of its initial value, which is similar to that observed by Western blotting. ATB-BMPA-tagged GLUT4 found in the plasma membrane fraction isolated from cells maintained in the continuous presence of insulin decreases to a level -45% of the initial value. The tu2 for this decrease is 10.6 * 1.5 min and is very similar to that obtained in the cells treated with collagenase to fully reverse insulin action. Also in contrast to the Western blotting results, addition of fresh insulin restores ATB-BMPAtagged GLUT4 to a new steady-state level, also only -45% of the initial, fully insulin-stimulated level. However, the latter occurs with a tu2 of 2.7 * 0.3 min, which is identical to that for the restoration of immunodetectable GLUT4 and slightly faster than that for the restimulation of glucose transport activity (Fig. 2). Fig. 5A shows that the time course for the increase in immunodetectable GLUT4 observed in the low-density microsomal fraction is consistent with the corresponding decreases in immunodetectable GLUT4 in the plasma membrane fraction. The low-density microsomal membrane fractions from adipose cells maintained in the continuous presence of insulin show a constant level of immunodetectable GLUT4 throughout the time course. The time courses for the increases in ATB-BMPA-labeled glucose transporters that return to the low-density microsomes are shown in Fig. 5B. The increases occur both in the cells treated with collagenase and in the cells maintained in the continuous presence of insulin. Restimulation of the collagenase-reversed cells with insulin results in a decrease in lowdensity microsomal tracer-tagged GLUT4 with the same halftime as is observed for the increase in plasma membrane ATB-BMPA-tagged GLUT4. Rate constants for the endocytosis (ken) and exocytosis (k,,) of GLUT4 were estimated from the ATB-BMPA-tagged and immunodetectable GLUT4 internalization and restimulation data shown in Fig. 4 using Equations 2 and 3 (see "Experimental Procedures"). This analysis assumes that adipose cells comprise only two pools of glucose transporters, a plasma membrane pool and an intracellular pool. However, this may be a marked oversimplification (see below). The results are outlined in Table 11. The estimate of ken for the internalization of ATB-BMPA-tagged GLUT4 in the continuous presence of insulin is not significantly different from that following insulin removal. In contrast, the exocytosis rate constant is -%fold lower following insulin removal than in the presence of continuous insulin, but -9-fold lower than with insulin restimulation. Comparing the rates of redistribution of immunodetectable GLUT4 during insulin removal and insulin restimulation shows that k,, is increased -25-fold by insulin, again without a significant change in ken. Combining all the estimates of k , under conditions of insulin removal and comparing these with the combined values obtained from insulin stimulation and continuous insulin indicates an -10-fold increase in k,, in response to insulin without a significant alteration in using a similar approach based on photolabeling rat adipose cells with While this paper was in revision, Jhun et al. (30) reported a study a [3H]bisglucose derivative. In this instance, however, these investigators labeled cells only during the steady state, either basal or maximally insulin-stimulated. They subsequently analyzed their results using the same simple two-pool model as used here and concluded that insulin stimulates glucose transport by decreasing ken by 2.8-fold and increasdecrease (perhaps 3650%) in ken in response to insulin, we believe that  Endocytosis (ked and exocytosis (ked rate constants for GLUT4 glucose transporter subcellular recycling as assessed by ATB-BMPA photolabeling and Western blotting in rat adipose cells in the basal a n d insulin-stimulated states The plasma membrane ATB-BMPA-tagged and immunodetectable GLUT4 data obtained from the three separate experiments described in the legends to Figs. 2 and 3 and shown in Fig. 4 as mean values were combined and analyzed as individual data points. The parameter values and their standard errors were determined for insulin removal (16 data points), continuous insulin (19 data points), and insulin restimulation ( (16) have shown that the appearance of GLUT4 at the adipose cell surface that can be photolabeled with ATB-BMPA precedes the increase in glucose transport activity. These results are consistent with those of Karnieli et al. (27), who showed that the rate of increase in the number of cytochalasin B-binding sites in the plasma membrane fraction isolated from insulin-stimulated cells is greater than the rate of increase in glucose transport activity observed in intact cells. Similarly, Gibbs et al. (321, using a technique involving borohydride labeling of GLUT1, showed a fast rate of appearance of this isoform, while Yang et al. (33) showed that the rates of appearance of both GLUTl and GLUT4, detected by using the ATB-BMPA photolabeling technique, are greater than the rate of onset of fully activated glucose transport activity in 3T3-Ll cells. As shown in Fig. 6, we have extended these observations here by comparing the rate of appearance of GLUT4 detectable by Western blotting with that of GLUT4 detectable by ATB-BMPA labeling. At 37 "C in cells maintained in the presence of adenosine, the rates of appearance of GLUT4 as detected by these methods show only a small difference. However, the rates of appearance of cellsurface GLUT4 detected by both methods precede the rate of increase in glucose transport activity (Fig. 6A). The rate of increase in immunodetectable GLUT4 under these conditions has a tl,z of -1.5 min, and any difference between the rates of appearance of immunodetectable GLUT4 and GLUT4 accessible to the ATB-BMPA photolabel is difficult to resolve.
To improve the resolution of precursor intermediate states in the insulin-stimulated subcellular trafficking, we have carried rats by 20-40-fold (Figs. 1,2, and 6) (3,4), whereas Jhunet al. report an 11-fold response at most; a n increased basal rate of glucose transport activity due to the bovine serum albumin and/or collagenase preparations used and/or some other technical problem in cell treatment is the usual explanation for a low -fold response to insulin. In addition, a 5-10-fold increase in immunodetectable GLUT4 and a 15-30-fold increase in photolabeled GLUT4 in the plasma membranes in response to insulin are typical (Figs. 3-5) (4, 7, 16, 181, compared to the -3-and -6-fold increases, respectively, reported by Jhun et al. Indeed, the extremely low photolabeling of GLUT4 in basal cells reported here (Figs. [3][4][5] discouraged us from even attempting an experiment in the basal steady state. Finally, Yang and Holman (31) very recently reported a study almost identical to the one described by Jhun et al., but using 3T3-Ll cells instead of rat adipose cells. Their results show -9-and -3-fold increases in the rates of GLUT4 and GLUTl exocytosis, respectively, in response to insulin and -30% decreases in the rates of endocytosis of both isoforms. out the comparisons at 20 "C ( Fig. 6B). At 20 "C, the half-times for the rates of appearance of GLUT4 detected by ATB-BMPA photolabeling (tilz -9 min) and by Western blotting (tila = 5 min) are both faster than the rate of increase in glucose transport activity, which occurs with a tlpz of -12 min. The difference in the half-time for appearance of GLUT4 that could be photolabeled by ATB-BMPA and the half-time for the rate of increase in glucose transport activity is similar to that observed by Yang et al. (331,who showed that in 3T3-Ll cells at 27 "C, the til2 values for these increases are 5.7 and 8.6 min, respectively.

DISCUSSION
In 1980, Suzuki and Kono (2) and Cushman and Wardzala (1) independently proposed that glucose transporters in rat adipose cells were mainly localized in an intracellular pool in the basal state and that insulin produced a shift in this distribution such that glucose transporters localized in the plasma membrane were markedly increased in the insulin-stimulated state. Since this proposal, the hypothesis has been extensively supported by measurements of steady-state distributions of glucose transporters (1)(2)(3)(4)(5)(8)(9)(10)(11). However, the direct demonstration that tracer-tagged glucose transporters redistribute to the plasma membrane in response to insulin has previously not been studied in detail. Oka and Czech (34) photolabeled intact cells in the basal state with cytochalasin B in the presence of 4,6-O-ethylidene-~-glucose to inhibit labeling of the cell-surface glucose transporters such that labeling was restricted to the low-density microsomal glucose transporter pool. They were then able to show that the cytochalasin B tracer-tagged glucose transporters moved to the cell surface in response to insulin stimulation and were then associated with the plasma membrane fraction of the cells.
Our own studies involving the use of the impermeant photolabel ATB-BMPA have shown that insulin increases the cellsurface availability of GLUT4 in rat adipose cells by 15-20-fold (4,16,35). An advantage of the bismannose photolabel ATB-BMPA is that it is an impermeable reagent and does not have access to the glucose transporters located in the intracellular pool. Thus, the discrete plasma membrane pool of glucose transporters can be selectively labeled, and the transfer of tracer-tagged glucose transporters t o the low-density microsomes can be followed. By using benzophenone derivatives of bisman-nose and bisglucose, we have shown that glucose transporters in rat adipose cells that were initially labeled in the plasma membranes are internalized to the low-density microsomes following 40 min a t 37 "c even in the presence of insulin (35). A similar internalization ofATB-BMPA-tagged glucose transporters was observed following incubation of 3T3-Ll cells with insulin for 60 min at 37 "C (19). Our previous studies have therefore already suggested that glucose transporters recycle in the presence of insulin. We now show here that by using the hishexose photolabeling approach, the time courses for internalization and recycling between subcellular membrane fractions can be determined in rat adipose cells.
While our previous studies suggested that glucose transporters recycle in the presence of insulin, this study documents this phenomenon in detail. GLUT4 tagged with the bismannose tracer on the surface of the adipose cell in the insulin-stimulated state moves from the plasma membranes to the lowdensity microsomes with a t l / 2 of 10.6 min in the continuous presence of insulin. At the same time, a constant insulin-stimulated steady-state distribution of immunodetectable GLUT4 is maintained between these two membrane fractions. If the movement of tracer-tagged GLUT4 is indicative of the movement of unlabeled GLUT4, then GLUT4 must be rapidly and continuously recycling in the continuous presence of insulin. The best evidence that tracer-tagged GLUT4 and unlabeled GLUT4 move similarly is the similarity in tlI2 values for their internalization when insulin is removed by collagenase treatment (9.4 and 13.1 min, respectively) and for their restimulation back to the plasma membrane when insulin is readded to collagenase-treated cells (2.7 and 2.7 min, respectively). The latter two experimental circumstances are the only ones in which both tracer-tagged GLUT4 and unlabeled GLUT4 are expected to move together.
The internalization of tracer-tagged GLUT4 in the continuous presence of insulin ultimately results in the achievement of an apparent new steady-state distribution of labeled GLUT4, with the plasma membranes retaining -45% of the initial value and the low-density microsomes accounting for the other =55%. It is highly significant that exactly this same steadystate distribution is achieved when adipose cells labeled in the insulin-stimulated state are reversed with collagenase and then restimulated with the readdition of insulin. These observations are consistent with the possibility that ATB-BMPAtagged GLUT4 fully equilibrates with the entire intracellular pool of glucose transporters. The specific activity of the ATB-BMPA-tagged glucose transporters thus decreases, and only an -30% proportion of these return to the plasma membrane on restimulation. The level of ATB-BMPA-tagged GLUT4 distributed between the plasma membranes and the low-density microsomes in the restimulated state is therefore equal to the steady-state distribution of GLUT4 between these fractions detected by Western blotting (24) or by cytochalasin B binding (27) with isolated membranes obtained by subcellular fractionation of cells in the fully insulin-stimulated state.
These data also provide for the first time direct evidence that the site of insulin action in stimulating glucose transporter translocation lies in the exocytosis leg of the recycling process. By simple inspection alone, the close similarity of the rates of internalization of GLUT4 in the continuous presence of insulin (till = 10.6 min for tracer-tagged GLUT4) and with insulin removal (t,,, = 13.1 min for immunodetectable GLUT4,9.4 min for tracer-tagged GLUT4, and 11.8 min for glucose transport activity) shows that insulin does not significantly decrease the rate of endocytosis of glucose transporters. Thus, insulin must markedly increase the rate of their exocytosis. In a more formal manner, application of the simple two-pool analytical procedure also suggests that insulin increases the rate of exocytosis of GLUT4 without significantly changing the rate of endocytosis (see "Results").
One of the assumptions of the analysis we have used is that following insulin removal, ken and k,, are immediately changed. However, this may not be the case if insulin is not instantaneously removed by the collagenase treatment. Indeed, Quon and Campfield (26), also using a two-pool model, suggested that the rate of insulin dissociation from its receptor determines the rate of net reversal of insulin-stimulated glucose transport. This may be the case in the early experiments described by Karnieli et al. (271, which were computer-simulated by Quon and Campfield. However, we have previously reported that insulin is removed from rat adipose cells with a t l l z of <5 min under conditions of collagenase treatment (3). In addition, we have now shown that ifthe reversal rate is delayed because of residual insulin or residual stimulus, then one would expect to see a more rapid internalization of label than of net loss of immunodetectable plasma membrane GLUT4 since the recycling of the former would be reduced by dilution within the intracellular vesicle ~0 0 1 .~ Because we have observed similar rates of internalization of ATB-BMPA-tagged GLUT4 and of net loss of immunodetectable GLUT4 following insulin removal, we conclude that both GLUT4 net endocytosis and ATB-BMPA-tagged GLUT4 exchange with unlabeled intracellular GLUT4 are both dependent on a slow endocytosis rate constant and are not rate-limited by insulin dissociation from its receptor.
The experimental approach we have taken may also underestimate the magnitude of insulin's stimulation of Kex. First, as discussed above with reference to ken, the change in K,, following insulin removal may not be instantaneous. Second, the reversal of insulin action may not have been complete, and the cells may not have returned to a true basal state. This may be due to a small but significant collagenase stimulation of GLUT4 translocation (see "Results") and possible UV radiation damage of the cells (although the cells were protected by irradiation through polystyrene lids on the dishes, and as noted below, various control experiments appear to rule out radiation damage as a significant variable). Third, the estimate of photolabel remaining in the plasma membrane is dependent on good subcellular fractionation techniques, and the plasma membrane fraction may be partly contaminated by photolabel that had transferred to the low-density microsomal fraction. Other evidence also suggests that insulin produces a much larger stimulation of k,, than estimated here. In experiments in which glucose transporter endocytosis in basal and insulintreated rat adipose cells has been blocked by incubation with a major histocompatibility complex class I peptide, the increases in glucose transport activity can be used to estimate exocytosis rate constants. These experiments show that insulin increases k,, by -20-fold over basal levels.* An apparent anomaly arises in comparing the rate of internalization of tracer-tagged GLUT4 in the continuous presence of insulin with the rate of reappearance of photolabeled GLUT4 in the plasma membranes stimulated by readdition of insulin to collagenase-treated cells, however, in so far as the latter (t1,2 = 2.7 min) is much faster than the former ( t l l z = 10.6 min). Examination of Equations 2 and 3 (see "Experimental Procedures") shows that in a simple system with just two pools of glucose transporters, the tIl2 values for these experiments would be expected to depend only on the exponential terms. Therefore, tu2 = In 2/(k,, + ken) for both the recycling time course and the insulin stimulation time course. Because k , and ken have insulin-stimulated values in both these experiments, it follows that the tU2 values would be expected to be the same.
However, since the tIl2 values in these experiments are not equal, we have examined several possibilities that may account for the discrepancy. First, the possibility that irradiation may have damaged the cells can be considered. Indeed, we have found that if adipose cells are not protected by irradiating them through the plastic lids of the Petri dishes, then irradiation markedly slows the reversal of the insulin response. Nevertheless, in this regard, we note that the recycling tl,2 (10.6 min) and the insulin stimulation tuz (2.7 min) which are compared were both carried out in irradiated cells. Although it could be argued that the latter tu2 is determined after cells have recovered from irradiation, control experiments (see "Results") demonstrate that acute irradiation slows the tllz for the initial insulin stimulation in fresh cells by no more than 30%. We have further considered the possibility that irradiation may have slowed the endocytosis rate constant (ken). A direct comparison of changes in glucose transport activity when insulin is removed by collagenase treatment (when one would expect that internalization would be mainly dependent on ken) in irradiated and nonirradiated cells shows no significant difference. An irradiation effect on k,, is possible. However, this would have to be transitory because if k , were reduced but ken were not, then we would have observed an abnormally low steady-state distribution of label and not one that corresponds to the distribution of glucose transporters as detected in cytochalasin B binding experiments (27) and in Western blotting experiments (24) in nonirradiated cells. 5 While we cannot totally exclude the possibility that the discrepancy between the insulin stimulation tl12 and the recycling tu2 is due to cell damage, we think it is important to also consider the possibility that the effect is due to the recycling of glucose transporters through intermediate states in the cytosol as illustrated on the right of Fig. 7. In their analysis of immunocytochemical studies on the subcellular distribution of GLUT4 in brown adipose tissue, Slot et al. (9) identified some 11 separate locations of immunodetectable GLUT4, eight cytoplasmic and three associated with the plasma membrane. In their description of the translocation model for insulin stimulation of glucose transport, Karnieli et al. (27) postulated that there may be four plasma membrane intermediate states in the subcellular trafficking pathway. These intermediate precursor states are also shown in Fig. 7. It is most unlikely that recycling through all these intermediates will be a simple kinetic process. Therefore, the method we have used to determine the apparent ken and kex, although useful, almost certainly oversimplifies the recycling process. We have examined the kinetic Jhun et al. (30) reported a tl,z of 3.24 f 0.51 min for the internalization of tracer-tagged GLUT4 in the continuous presence of insulin, compared to the value reported here of 10.6 t 1.5 min. Their tu, for GLUT4 recycling is thus more consistent with the tm for insulin stimulation than is ours. Possible sources of this difference include differences in the photolabels and/or W lamps. However, as discussed above, our W irradiation conditions do not appear to significantly influence the GLUT4 cycling process as measured here. Another potential source of difference is the possibility that the apparent label internalization rates are influenced by the processing of the samples after labeling. Jhun et al. report in their Fig. 5 that at their zero time, the label recovered in their low-density microsomal fraction is approximately stimulated state. This suggests that their subcellular fractionation pro-one-third of that recovered in the plasma membranes in the insulincedure and/or their failure to arrest cycling leads to an overestimate of internalization. In our studies, we have used KCN to prevent cycling during the processing of the cells and have observed only 5% of labeled GLUT4 in the low-density microsomal fraction at zero time. predictions of a system (Fig. 7) in which two intracellular compartments, the endosomes and the specialized "secretory" tubulovesicular compartment, are involved. It is clear from computer simulations of this system that the involvement of a second intracelIular compartment allows a rapid initial stimulation of translocation by insulin, but a slower recycling of glucose transporters in the continuous presence of insulin. This is because the tubulovesicular compartment could rapidly dock and fuse with the plasma membrane without endocytosis, contributing significantly to the stimulation tu2 (as it does when there is only one intracellular pool) (Equations 2 and 3). HOWever, with two intracellular pools, the recycling in the continuous presence of insulin would depend on endosome processing and recycling, which could markedly slow the turnover tu2 that maintains a steady-state distribution of glucose transporters.
The model shown in Fig. 7 would also account for the observation of intermediate and partially occluded forms of the glucose transporter that may be precursor states in the stimulation leg of the pathway. At present, it is not entirely clear how abundant the occluded forms of the glucose transporter are. Their apparent presence and high concentration in the plasma membrane at early times following insulin stimulation (Fig. 6) may be, to some extent, influenced by the use of KCN to arrest recycling (36). However, several observations suggest that consideration of occluded and partially occluded glucose transporters ( Fig. 7) is required. First, because insulin stimulation is rapid but recycling of glucose transporters is relatively slow, one would expect the bulk of the glucose transporters to be localized to the plasma membrane unIess some glucose transporters translocate to the plasma membrane and back again without participating in glucose transport. Second, a rapid onset of residence of glucose transporters in the occluded and partially occluded states is required to account for the observation that the tUP values for the appearance of GLUT4, as detected by Western blotting of plasma membrane fractions and by ATB-BMPA photolabeling, are faster than those for glucose transport stimulation. ATB-BMPA might label partially occluded glucose transporters at the point where occluded vesicles or tubulovesicular elements fuse with the plasma membrane. At this point, glucose transporters might be exposed to photolabel as well as transport substrate, but not fully participate in glucose transport because of association with trafficking proteins (33). The small (-1.5-2-fold) discrepancy that we have noted in comparing glucose transport activities and photolabeled GLUT4 levels in the basal and insulin-stimulated states (4,18,37) is likely to be due to a proportion of the glucose transporters being present in this catalytically inactive state in basal adipose cells. Third, the observation that the steady-state distribution of glucose transporters at 20 " C , when endocytosis is markedly reduced (16); is similar to that which occurs at 37 "C when glucose transporters recycle through the endosomes suggests that glucose transporters may emerge at the cell surface and be returned t o the low-density microsomes via a mechanism that does not involve the normal endocytosis route.
The model that we have described in this study extends our previous model of glucose transporter translocation and has identified exocytosis steps in subcellular trafficking that will be the focus for further studies on the mechanisms for insulin stimulation of glucose transport. In addition, the identified intermediate states provide a mechanistic basis for the glucose transporter intrinsic activity changes previously proposed to have been observed in rat adipose cells in response to adenosine and isoproterenol (38-40) or during specific altered metabolic states (41)  Occluded, partially occluded, and fully functional glucose transporters are defined by their detectability by Western blotting in plasma membranes, their susceptibility to photolabeling with ATB-BMPA, and their ability to transport 3-0-methylglucose in intact adipose cells. Occluded transporters (including those in endosomal vesicles) are detectable by immunoblotting, but neither are labeled by ATB-BMPA nor participate in glucose transport activity. Partially occluded transporters are detectable by immunoblotting and are labeled by ATB-BMPA, but do not participate in glucose transport activity. Fully functional transporters are detectable by immunoblotting, are labeled by ATB-BMPA, and do transport 3-0methylglucose.