The glucose transporter GluT4 and secretory carrier membrane proteins (SCAMPs) colocalize in rat adipocytes and partially segregate during insulin stimulation.

Secretory carrier membrane proteins (SCAMPs) mark the recycling system for the insulin-responsive glucose transporter, GluT4, in rat adipocytes. Anti-GluT4 and anti-SCAMP antibodies each immunoadsorbed vesicles containing both antigens from a low density microsomal fraction that is enriched in both antigens. The immunoadsorbed vesicles also contain VAMPs (synaptobrevins), synaptic vesicle membrane proteins. All three antigens were colocalized in low density microsomal vesicles from both basal and insulin-stimulated adipocytes. The SCAMPs have the same electrophoretic mobility as a major polypeptides detected in GluT4 vesicles. During insulin stimulation, 40% each of GluT4 and VAMPs redistribute from low density microsomes to the plasma membrane fraction; however, < 10% of the SCAMPs redistribute. Immunocytochemical staining of adipose tissue shows almost complete coincidence of SCAMPs and GluT4 in the basal state and extensive redistribution of both antigens to the cell periphery during insulin stimulation. Segregation of antigens during stimulation is not as distinct as observed by fractionation, although there are regions at the cell border where the SCAMPs appear more concentrated than GluT4. These data suggest that during insulin stimulation, in contrast to the behaviour of GluT4, SCAMPs remain tightly associated with the recycling system.

in the region of the trans-Golgi network and also more peripherally. Upon insulin stimulation, these vesicles translocate to and fuse with the plasma membrane, redistributing -40% of the GluT4 to the cell surface and increasing glucose transport into the cells 20-30-fold (Holman et al., 1990;Slot et al., 1991aSlot et al., , 1991bSmith et al., 1991;Rodnick et al., 1992). The GluT4containing vesicles have been identified using immunoelectron microscopy (Slot et al., 1991a(Slot et al., , 1991bRodnick et al., 1992) and purified from low density microsomal fractions using vesicle immunoadsorption with antibodies against the cytoplasmic tail of the GluT4 molecule (Zorzano et al., 1989;Calderhead et al., 1990;Cain et al., 1992;Rodnick et al., 1992).
Translocation of GluT4 to the cell surface in response to insulin resembles regulated exocytosis in specialized secretory cells. Guided by this functional analogy, Cain et al. (1992) have shown recently that the synaptic vesicle protein VAMP(s) (vesicle associated gembrane protein, synaptobrevin) is enriched in the GluT4 vesicles isolated from adipocytes. Brand et al. (1991) have identified a family of secretory carrier membrane proteins (SCAMPS; Mr 35,000-40,000)2 that are found in all regulated secretory carrier membranes. More extensive examination of the SCAMPS has shown that they are present in a wide variety of cells regardless of their specialization for regulated secretion. The SCAMPs are localized in general to membranes (secretory and endocytic vesicles) that share the function of transport to and from the cell surface and thus constitute markers of a general recycling ~y s t e r n . ~ We now show that GluT4 and VAMPs colocalize with the SCAMPs in adipocytes. Using immunochemical and immunocytochemical approaches, the coincidence is practically complete in unstimulated cells, and the SCAMPS appear to be major polypeptides of GluT4 vesicles. Interestingly, immunocytochemical observations indicate that both SCAMPS and GluT4 relocate from the perinuclear region to the cell periphery during insulin stimulation. Although colocalization remains extensive upon reorganization of recycling membranes, we detect a partial segregation of GluT4 and SCAMPs that is especially clear using cell fractionation; only GluT4 redistributes appreciably to the plasma membrane fraction. Thus as GluT4 cycles to the cell surface in response to insulin, it may separate in part from the SCAMPs with which it is associated in vesicles. ered sodium chloride; PM, plasma membrane fraction; SCAMP, secre-TW, transferrin receptor; VAMP, (synaptic) vesicle-associated mem-tory carrier membrane protein; TBS, Tris-buffered sodium chloride; brane protein. 31,000 (Brand et al., 1991). According to more recent studies SCAMPs were originally stated to have apparent M, in the range including molecular cloning (S. H. Brand and J. D. Castle, manuscript in preparation), the estimated range should be revised to 35,000-40,000. Brand, S. H., and Castle, J. D. (1993) EMBO J., in press. MATERIALS AND METHODS Antibodies--The affinity-purified rabbit antibodies against the carboxyl-terminal peptide of the rat GluT4 (amino acid residues 491-509) have been previously described (Calderhead et al., 1990). The mouse monoclonal antibody (SC7C12) against the SCAMPs (Brand et al., 1991) was purified from hybridoma culture medium by concentration with a Centricon device, M, cut-off 30,000 (Amicon, Beverly, MA), or purified from ascites on a Protein A/G column (Pierce Chemical Co.). The anti-VAMP antibody was an affinity-purified rabbit antibody prepared against a peptide of rat brain VAMP 1 and 2 (Cain et al., 1992). A mouse monoclonal antibody, 1F8, against rat GluT4 was purchased from Genzyme (Cambridge, MA). A monoclonal antibody against the human transfemn receptor was a gift from Dr. I. Trowbridge (Salk Institute, San Diego, CA), and two polyclonal antibodies against the mannose-6-phosphaMGF-I1 receptor were g i b from Drs. D. Messner and S. Kornfeld (Washington University, St. Louis, MO) and Drs. C. Scott and R.
Electrophoresis and Irnrnunoblotting-Samples were reduced with 100 m M dithiothreitol and electrophoresed on 10, 11, or 13% polyacrylamide gels (Laemmli, 1970) and electrotransferred to 0.45-pm nitrocellulose (Schleicher & Schuell) for 3 h at 5 V/cm in 25 m~ Tris, 192 m~ glycine, 0.05% SDS, and 20% methanol (Towbin et al., 1979) for the SCAMPs; the SDS concentration was lowered to 0.005% for the GluT4 and the VAMPS. For immunoblotting the SCAMPs, the blots were blocked with 1.5% hemoglobin in phosphate-buffered saline (PBS), incubated overnight at 4 "C in anti-SCAMP (1-3 pg/ml), and then in rabbit anti-mouse I g G (1 pg/ml) and 1261-labeled goat anti-rabbit IgG for 1 h each at room temperature. All incubations and washes were done in PBS containing 1.5% hemoglobin. The GluT4 antibody was used at 6 pg/ml in 1% milk in PBS and detected using 1z61-labeled goat anti-rabbit antibody (Cain et al., 1992). As indicated in the figure legends, some blots were treated as described (Cain et al., 19921, i.e. blocked in milk, incubated in primary antibody, and goat anti-mouse or anti-rabbit antibody-conjugated to peroxidase and visualized using enhanced chemiluminescence (ECL reagent, Amersham Corp.). VAMPS were immunoblotted with anti-VAMP l/2 peptide (1 pg/ml), and visualized using ECL (Cain et al., 1992). For quantitative studies, computer-assisted video densitometry was used to measure the intensities of bands in the linear exposure range on autoradiograms.
Vesicle Immunoisolation with Staphlococcus Q U~U S Cells-This procedure was performed as described (Cain et al., 1992) with minor modifications as indicated. Formaldehyde-fixed Staphylococcus aureus cells (Immunoprecipitin, Life Technologies, Inc., S. Q U~~U S ) were washed twice with 10 mg/ml bovine serum albumin in PBS prior to use. For the isolation of GluT4 vesicles, S. aureus cells were incubated with anti-GluT4 antibody or a nonspecific rabbit IgG at 3 pg of antibody/pl of S. CLUIVUS for 2 h at 4 "C. To immunoprecipitate with anti-SCAMP antibody, S. QUWUS was precoated with goat anti-mouse IgG antibody (Fcspecific, 0.6 pdpl S. Q U W U S ) for 30 min at room temperature and then incubated with purified anti-SCAMP antibody at 1.1 pg/pl S. aureus for 1.5 h at 4 "C. S. QUWUS cells were washed twice with 250 m~ sucrose, 10 m M Hepes, 1 m~ EDTA (homogenization buffer) prior to use, to remove unbound antibody. Antibody-coated S. aureus was incubated (2 h at 4 "C) with a 48,000 x g supernatant prepared from adipocyte homogenates (Cain et d., 1992) that contains both low density microsomes and cytosol. The 48,000 x g supernatant was adjusted to 100 m M NaCl or 100 m M potassium phosphate, pH 7.4, and centrifuged for 6 min at 11,000 x g in a microcentrifuge prior to S. aureus addition. After incubation, vesicle-S. aureus complexes were washed twice with homogenization buffer with added salt, and all bound proteins were eluted in 133 pl of SDS sample buffer containing 2.5% SDS, 125 n m Tris-HC1, pH 6.8, 10% glycerol, and 8 M urea. Low density microsomes remaining in the 48,000 X g supernatant after immunoadsorption were recovered by centrifugation at 150,000 x g . , for 1 h. These membrane pellets were resuspended in 133 pl of SDS sample buffer. Equal volumes of samples were used for immunoblotting.
lb titrate immunoadsorption of vesicles from the LDM at limiting concentrations of antibody, 10 pl of S. Q U~~U S cells were coated with 140 pg of rabbit anti-mouse IgG and decreasing amounts of anti-SCAMP antibody (as indicated in Fig. 4) prior to combining with samples of the 48,000 x g supernatant and processing as above. The vesicles remaining in the 48,000 x g supernatant after immunoadsorption were concentrated as described above, and their content of GluT4 and SCAMP was quantitated using immunoblotting and the ECL method. A series of antigen-containing standards and duplicate samples were included on all blots.
To examine the polypeptide profile of the GluT4 vesicles, the vesicles were immunoprecipitated exactly as described (Cain et al., 1992) and eluted with the nonionic detergent Cl,E, (Calbiochem). Vesicle proteins were stained with colloidal gold (Bio-Rad) after separation by electrophoresis and transfer to nitrocellulose. Parallel lanes were used for immunoblotting with anti-SCAMP.
Vesicle Imrnunoisolatwn with Magnetic Bead-Electron microscopy was performed on samples of vesicles (from the 48,000 x g supernatant) that had been adsorbed to antibody-coated magnetic beads and immunostained for antigens using colloidal gold-conjugated secondary antibodies as described (Brand et al., 1991). Magnetic beads, 4.5 l.un in diameter, precoated with sheep antimouse antibody (Dynabeads, Dynal Inc, Great Neck, NY), were incubated with anti-SCAMPS or IF8 (10 pg of antibody/0.9 mg of beads) in 1 ml of 100 m~ potassium phosphate buffer, pH 7.4 (Kphos) with 5 mg/ml BSA. lbsyl-activated 450-pm Dynabeads were coupled to goat anti-rabbit Fc-specific antibody using the manufacturer's directions. These beads (1 mg) were incubated with 11 pg of anti-GluT4 in 1 ml of Kphos with 5 mg/ml BSA. Beads were washed in Kphos with 5 mg/ml BSA, and incubated with the 48,000 x g supernatant from rat adipocytes for 3 h rotating (-6 rpm) at 4°C. Supernatant from 0.5-1 rat was used in each sample after adjusting to 100 m M Kphos and a final volume of 1 ml. Beads were collected using a magnetic particle concentrator (also from Dynal Inc.) and washed and processed for electron microscopy as described previously (Brand et al., 1991). In most experiments, control beads were coated with an irrelevant rabbit or mouse immunoglobulin; as negligible vesicle binding was always observed, controls in some of the later experiments were samples with primary antibody omitted (again with negligible vesicle binding). Immunostaining of vesicles adsorbed to Dynabeads was performed using either anti-GluT4 antibody (15 pg/ml) or anti-VAMP antibody (15 pg/ml) in 0.3 M sucrose, 100 m M Kphos, 5 m g / d BSA followed, &r washing, by goat anti-rabbit I& conjugated to 10 nm gold (EY Laboratories, San Mateo, CA) used at a Y25 dilution (Brand et al., 1991).
Irnmuno~uorescence-Immunocytochemical localization of SCAMPs and GluT4 in brown adipose tissue was carried out according to Slot et al. (1991a) with minor modifications. Rats were starved overnight, chilled for 4 h at 4 "C to decrease endogenous stored lipid, and given an intraperitoneal injection of PBS or PBS containing insulin (8 unitskg) and glucose (1 gmkg). Rata were anesthetized with metaphane and between 20 and 30 min post-injection, they were perfused initially (for 30 8 ) with PBS containing 20 unitdml heparin and subsequently with periodate-lysine-paraformaldehyde fixative (McLean and Nakane, 1974) in which the formaldehyde concentration was lowered to 0.5%.
Tissues were infiltrated successively in 12,15, and 18% 8ucrose in PBS, frozen, and cryosectioned as described previously (Laurie et d . , 1992). The cryosections (stored on glass slides in absolute ethanol at -20 "C) were equilibrated with 10 lll~ Tris, 150 m~ NaCl, pH 7.4 (TBS) and quenched 1 h at room temperature with 16% heat-inactivated horse serum in TBS (HS-TBS). For double labeling, we used mixtures of primary antibody (anti-SCAMP (mouse monoclonal antibody SG7C12), 25 pg/mV and anti-GluT4 (affinity-purified rabbit antibody to the COOH-terminal 19 amino acids of GluT4, 12 pg/ml) diluted in HS-TBS and incubated for 2-3 h at room temperature. Following washing with TBS, specimens were incubated for 1 h at room temperature with mixtures of secondary antibodies (goat anti-mouse IgG coqiugated to Cy3 and goat anti-rabbit I g G conjugated to Cy5), both affinity-purified and preadsorbed against cross-reacting antibodies by the manufacturer (Jackson Immunochemicals, West Grove, PA) and both diluted to 5-7.5 pg/ml in HS-TBS. The slides were washed with TBS and mounted in 90% glycerol in PBS for visualization by immunofluorescence. Controls that were also examined included deletion of the primary antibody mixture, substitution of the primary antibody mixture with an irrelevant antibody at the same dilution, and combinations of each primary antibody with the wrong secondary antibody. In all controls, no staining above background was observed.
Immunofluorescent images were collected using a Laser Sharp MRC-600 Bio-Rad confocal box adapted to a Nikon Diaphot microscope equipped with a lOOx W apochromatic objective (N.A. 1.3) and a microcomputer outfitted with COMOS software (Bio-Rad). Illumination was provided by an argonkrypton laser. For detection of Cy3, excitation and emission filters were 568 DFlO and 585 EFLP, respectively, and for Cy5, excitation and emission filters were 647 DFlO and 680 EF32, respectively. Most images were recorded with the confocal detector aperture set near its maximum for the low intensity signals. The contrast of the recordings was enhanced using a standard adaptive histogram procedure. Illustrations were prepared using ANALYZE software 5.0.1 (Mayo Foundation, Rochester, M N ) using a Sun Sparc2 work station.

RESULTS
SCAMPs Are Components of Vesicles Containing GluT4 and VAMP. +-The SCAMPs are widely distributed among recycling cell surface carriers (Brand et al., 1991).3 Thus, we investigated whether the SCAMPs are components of the vesicles involved in the translocation of GluT4 to the plasma membrane. The LDM from rat adipocytes, which contains most of the insulinresponsive GluT4 (Cain et al., 1992), also contains the SCAMPs detected immunochemically in the LDM as bands of apparent M, 36,000-39,000.2 As the LDM is a heterogeneous fraction (Zorzano et al., 19891, we assessed the extent to which both SCAMPs and GluT4 (and VAMPs) are present in the same membranes by immunoblotting vesicle fractions that were immunoadsorbed from the LDM. As shown in Fig. IA, immunoadsorption with anti-GluT4 results in coadsorption of SCAMPs and VAMPS, and immunoadsorption with anti-SCAMPS results in the coadsorption of GluT4 and the VAMPs. There is no detectable immunoadsorption of any of these antigens when control nonspecific rabbit or goat anti-mouse IgG is substituted in the procedure. Under conditions in which GluT4 is completely depleted from the LDM by immunoadsorption with anti-GluT4 (Fig. IA, Super), the SCAMPs and VAMPs are also completely depleted from the LDM. In addition, immunoadsorption of all the SCAMPs using anti-SCAMP completely depletes the LDM of all GluT4 and the VAMPs. These results show that the distributions of SCAMPs, GluT4, and VAMPs in the LDM overlap. There are no sizable pools of SCAMPs or VAMPs in vesicles that do not contain enough GluT4 to allow immunoadsorption, and the converse applies to GluT4 and VAMPs with regard to immunoadsorption by the SCAMPs.
During insulin stimulation, GluT4 is redistributed to the cell surface yet continues to recycle through the endosomal system (Slot et al., 1991a;Jhun et al., 1992). We repeated the immunoadsorption experiments on the LDM prepared from insulintreated adipocytes (Fig. LB). Immunoadsorption of GluT4 still resulted in the complete depletion of the SCAMPs and VAMPs present in the LDM (Fig. 1B, Super). Similarly, immunoadsorption with anti-SCAMP still completely depleted the LDM of GluT4 and VAMPs. These data indicate that the SCAMPs, GluT4 and VAMPs that remain in the LDM during insulin stimulation are colocalized. Distribution of Antigens among Vesicles in the LDM "Immunoadsorption of vesicles is an efficient and sensitive procedure and may result from the interaction of only a few antigen and antibody molecules. It is possible that adsorbed vesicles are heterogeneous, containing subpopulations that are highly enriched in SCAMPs as compared to GluT4 and vice versa. Three different experiments were performed to check for subpopulations of SCAMP-rich, GluT4-poor vesicles in the LDM: vesicle immunoadsorption from the LDM using limiting amounts of antibody to bind the potential SCAMP-rich vesicles selectively, velocity gradient sedimentation of the LDM to compare antigen distributions among different-sized vesicles, and immunocytochemical labeling of immunoadsorbed vesicles. Titration of the Adsorption of Vesicles /?om the LDM Using Limiting Amounts of Antibody-Four experiments were conducted in which vesicles from a fmed amount of LDM were adsorbed with limiting amounts of anti-SCAMP. The progressive depletions of SCAMP and GluT4 from the supernatants were compared by immunoblotting (Fig. 2). The SCAMPs and GluT4 were depleted in parallel from the LDM prepared from both unstimulated (panel A ) and insulin-treated (panel B ) adipocytes. There are no SCAMP-rich, GluT4-poor subcompartments that bind preferentially to limiting amounts of anti-SCAMP, and there are no GluTCrich, SCAMP-poor subcompartments that cannot bind to limiting amounts of anti-SCAMP. These results suggest that vesicles especially enriched in one of the antigens are not present in the LDM.
Velocity Sedimentation on Glycerol Gradients-This procedure has been especially useful for resolving synaptic-like microvesicles of endocrine cells from endosomes (Clift-O'Grady et al., 1990;Cameron et al., 1991). As shown in Fig. 3, a population of small and probably heterogeneously sized, antigen-containing vesicles was resolved from the rest of the membranes in the adipocyte LDM on a glycerol gradient. GluT4, the SCAMPs and the VAMPs distributed approximately in parallel across the gradient as judged by immunoblotting of the separate frac- tions. In contrast, the transferrin receptor ('IYR) and the mannose 6-phosphate receptor (M6PR) were found almost entirely in the peak of larger vesicles at the bottom of the gradient that also contained GluT4, the SCAMPs, and the VAMPs. Thus there is no vesicle population selectively enriched in either GluT4 or the SCAMPs that can be resolved by size. The finding that there was little or no effect of insulin on the relative distributions of GluT4 and the VAMPs between the more slowly and more rapidly sedimenting vesicles is of some interest. As described below and elsewhere ( Fig. 6; Cain et al. (199211, insulin caused a 35% decrease in the amounts of GluT4 and VAMPS in the LDM as a result of translocation to the plasma membrane. The unchanged distributions on the gradients indicates that GluT4 and VAMPS from both vesicle populations translocate to the plasma membrane. Fig. 4 shows representative examples of vesicles that have been adsorbed from the LDM onto magnetic beads coated with either anti-SCAMP or anti-GluT4 and then labeled with anti-GluT4 or anti-VAMP and secondary antibodies conjugated to 10 nm colloidal gold.* In all cases, the beads were coated by many small vesicles, extended tubulovesicles, and a few larger, variably shaped vesicles and vesicle aggregates. Control beads (see "Materials and Methods") bound almost no vesicles (data not shown). Most of the vesicles isolated with anti-SCAMP (7040% based on counting the vesicles present on the profiles of six beads in each sample) are labeled with anti-GluT4 and anti-VAMP at a uniformly high density. Most of the vesicles isolated with anti-GluT4 are also labeled with anti-VAMP at a high density. These data indicate that if there are SCAMP-ricWGluT4-and VAMP-poor compartments in the LDM, they are the minority of the SCAMP-containing vesicles.

EM Immunocytochemical Analysis of Vesicles Bound to Antibody-coated Beads-
Immunocytochemical labeling with anti-SCAMP could not be performed because aldehyde fixation used to stabilize the immunoadsorbed vesicles prior to labeling interferes with the epitope for anti-SCAMP (Brand et uZ., 1991). Altogether, these experiments argue that GluT4, the SCAMPs, and the VAMPS all have very similar distributions within various membranes constituting the LDM, a fraction that contains most of the markers of cell surface recycling pathways.

SCAMPS Appear to Be Major Proteins of the G1uT4
Vesicles -With the knowledge that the SCAMPs and GluT4 are extensively, if not completely, colocalized within the LDM fraction, we were curious to examine the relative abundance of SCAMP polypeptides present in GluT4-containing membranes. Thus we eluted the proteins from vesicles adsorbed to anti-GluT4 coated S. aureus using a nonionic detergent, CI2Es. This procedure elutes no detectable GluT4 or antibody. As ,seen in Fig.   5, CI2E8 elutes polypeptides of a broad range of sizes, including a prominent bandts) at M, 37,000. This band comigrates with the major SCAMP bandts) immunoblotted in an adjacent lane of the gel, suggesting that this SCAMP is a major protein of GluT4 vesicles. Interestingly, the levels of the SCAMP bands (detected by both immunoblotting and protein staining) did not decrease with insulin stimulation (discussed below); the protein staining of certain other bands (e.g. 50 and 90-100 kDa) also did not decrease. In contrast, other polypeptides in the GluT4 vesicles (e.g. at 175 and 31 kDa) significantly decreased, as was the case for GluT4 and VAMP antigens (which are not visible in the protein staining pattern). Subcellular Distribution of GhT4 and SCAMPs-In view of the well established redistribution of both GluT4 and the VAMPs in response to insulin (Cain et al., 1992), we were quite .. . . identified by G4 on the bead) were incubated with LDM-containing 48,000 x g supernatant. Adsorbed vesicles were labeled with anti-GluT4 (c and d; identified by G4 above the bead surface) or anti-VAMP (e-h; identified by VA above the bead surface). Bound antibodies were detected using goat anti-rabbit IgG conjugated to 10 nm gold. Bars, 100 nm. interested by the observation that insulin did not appear to alter the levels of the SCAMPS in the LDM (Fig. 5). Consequently, we compared the distribution of the SCAMPs among all adipocyte fractions in the absence and presence of insulin stimulation. Examination of the same proportion of each fraction by immunoblotting showed that, as expected, the SCAMPs are mostly located in the LDM and are found at low levels in the HDM and PM fractions while being negligible in the mitochondriallnuclear and cytosol fractions (Fig. 6a). Notably, neither the distribution nor the total amount of detectable an- and SCAMPs (C) before and aRer insulin stimulation. Bands were visualized using lZ6I-secondary antibody and autoradiography. tigen appear to change significantly with insulin stimulation. SCAMP and GluT4 levels in the LDM and PM of control and insulin-stimulated samples are compared in Fig. 6 ( b and c ) on immunoblots of equal protein loads. The distributions of both antigens in these two fractions were quantitated for six different parallel fractionations of basal and insulin-treated adipocytes. The SCAMP distribution was affected only slightly by insulin stimulation as compared to the GluT4 distribution. The levels of SCAMPS in the LDM of insulin treated samples were 93 6% (S.E.; n = 6) those in the LDM from controls. In contrast, GluT4 levels in the LDM after insulin stimulation were 63 * 8% of control levels. For PM fractions, SCAMP levels increased 2.6 * 0.5-fold (S.E.; n = 6) with insulin stimulation while GluT4 levels in the PM increased 5.1 * 1.8-fold. Although -The distributions of SCAMPS and GluT4 were compared by immunocytochemistry in brown adipose tissue. Brown adipose tissue exhibits the same insulin-stimulated translocation of GluT4 to the plasma membrane as white adipose tissue (Slot et al., 1991a) and can be experimentally depleted of stored triglyceride, enabling examination of antigen distribution by indirect immunofluorescence. As shown in Figs. 7 and 8, the distributions of the two antigens in unstimulated tissue are nearly identical. Both are focally concentrated at intracellular, often perinuclear, sites having a reticular appearance. These sites are mostly located a t a visible distance from cell borders. Notably, occasional examples have been found where SCAMPs but not GluT4 are highly concentrated (Fig. 7), attesting to the specificity of antibody staining. The differences may stem from either the broader distribution of SCAMPs among different cell types (Slot et al., 1990;Brand et al., 199U3 or possibly the presence of occasional adipocytes with very attenuated expression of GluT4. In insulin-stimulated adipocytes, both GluT4 and SCAMPs undergo a striking and parallel redistribution to the cell periphery such that much of the staining is Concentrated at or near surfaces that frequently border the capillary endothelium (Fig. 8). This dramatic redistribution of GluT4 has been reported previously in brown fat (Slot et al., 1991a). In all specimens examined, the staining of SCAMPs and GluT4 showed the same overall pattern, although in certain cell profiles, SCAMP staining at the cell border was more focally concentrated than that of GluT4. This local difference may be significant; the more diffise GluT4 staining may be indicative of a broader cell surface distribution. This may be the basis for the partial segregation of GluT4 and the SCAMPs that is observed by fractionation of insulin-treated cells (Figs. 5 and 6; see "Discussion"). VAMPS) in adipocytes and their codistribution with GluT4 provide clear support for the idea that the insulin-regulated carriers of GluT4 share some of the same functional machinery as secretory vesicles in regulated secretory cells. Not only does this relationship extend to synaptic vesicles as recognized previously (Cain et al., 1992), it also now extends to protein storage granules of exocrine and endocrine cells. Moreover, the similarity to the latter organelles is increased by the observation of at least two sizes of SCAMPS in GluT4 vesicles rather than the single-sized M, 37,000 form(s) observed in synaptic vesicles (Brand et al., 1991).3 By drawing analogies between insulin-responsive GluT4 carriers and stimulus-regulated secretory vesicles, we are not implying that GluT4 in unstimulated adipocytes is entirely concentrated in discrete stimulus-regulated carriers in the same way that synaptic vesicle marker proteins are concentrated in synaptic vesicles in neurons. Instead, the, almost completely coincident localization of GluT4 with the SCAMPs suggests a different situation. The SCAMPs have recently been identified as markers for a general cell surface recycling system3 as broadly defined by Cameron et al. (1991). This system encompasses endocytic pathways and also includes regulated secretory pathways as derivatives in specialized cells. Accordingly, if SCAMPs delineate this system in adipocytes, then GluT4 also delineates this system, and its distribution may include not only regulated transport compartments but also components of constitutive recycling pathway(s). This view is supported by the findings presented in Fig. 3 showing that GluT4 and the SCAMPs (and the VAMPS) cosediment with the TfR and M6PR in rapidly sedimenting vesicles in addition to sedimenting in a slower peak in vesicles that appear to lack (or are relatively depleted of) these constitutively recycling receptors. Interestingly, the distribution on glycerol gradients of GluT4 and the SCAMPs (and the VAMPS) relative to that of the TfR (Fig 3) is quite similar to the distribution of synaptophysin relative to that of the TfR in PC12 cells (Cameron et al., 1991). Thus GluT4/SCAMPNAMP-rich, TfR-poor membranes in adipocytes may be analogous to synaptic-like microvesicles in neuroendocrine cells. Although regulated exocytosis of synaptic-like microvesicles has not been demonstrated as yet, insulin-stimulated export of GluT4 to the adipocyte cell surface may occur from such a compartment.

Implications from Colocalization
Constitutive export of TfR and MGPR may involve a different route within the general recycling system that also would contain some SCAMPs and GluT4. Recently, evidence for constitutive recycling of GluT4 in adipocytes has been reported (Jhun et al., 1992). It should be noted that there is evidence that in adipocytes, insulin also increases the rate constant for exocytosis of constitutively recycling cell surface proteins (Tanner and Lienhard, 1987) and at least one constitutively secreted protein (Kitagawa et al., 1989). This effect may account for the smaller increases in cell surface TfR, MGPR, and GluTl (relative to GluT4) caused by insulin (Davis et al., 1986;Lienhard, 1987, 1989;Appell et al., 1988;Zorzano et al., 1989) and suggests that regulated export extends beyond the GluT4 route.
An alternative, equally plausible proposal for the way in which GluT4 reaches the cell surface has been described by Robinson and James (1992). They suggest that in response to insulin, vesicles enriched in GluT4 (presumably corresponding to the less rapidly sedimenting ones lacking recycling receptors; see Fig. 3) may fuse with the endosomal compartment rather than the plasma membrane. In this scenario, the GluT4 then proceeds to the plasma membrane from the endosomes via the constitutive recycling pathway, which, as noted above, is also stimulated by insulin.

Insulin-stimulated 12.aficking of the SCAMPs and GluT4
-Given the codistribution of the SCAMPs and GluT4 in unstimulated adipocytes, their distinct distributions h m one another following insulin stimulation as revealed by cell fractionation (Fig. 6) are especially interesting. We believe that the different distributions arise by partial segregation of the SCAMPs from GluT4 when trafficking to the cell surface is increased. Several alternative explanations seem much less likely. First, segregation probably is not an artifact reflecting underdetection of SCAMPs in the plasma membrane fraction during stimulation since the total SCAMP level in all fractions is unchanged upon stimulation (Fig. 6). Second, SCAMP relocation from plasma membranes via dissociation (as observed for the GTP binding protein rab 3a during exocytosis of synaptic vesicles; Fischer von Mollard et al. (1991)) is very unlikely because sequencing predicts that the SCAMPs are transmembrane protein^.^ "hird, we have found no evidence for vesicle subpopulations that are separately highly enriched in GluT4 and SCAMP and could differ in their sensitivities to insulin, using either immunoprecipitation, velocity gradient sedimentation, EM immunogold labeling of isolated vesicles, or immunofluorescent staining of whole cells. These methods are unlikely to detect vesicles that only contain proportionately more SCAMPs than GluT4 as the result of partial segregation during insulin stimulation.
Although presently we cannot rule out the possibility that partial segregation of the SCAMPs and GluT4 during insulin stimulation occurs intracellularly, our prediction is that it arises from sorting at the cell surface involving more efficient sequestration and reinternalization of the SCAMPs than of GluT4. As a consequence, we expect that proportionately more SCAMPs than GluT4 will be concentrated in coated pits, coated vesicles, and early endosomal elements near the cell surface. Since the latter fractionate as part of the LDM fraction, the extent of translocation of SCAMPs is less than that of GluT4.
The co-redistribution of GluT4 and the SCAMPS to the cell border in insulin-treated brown fat, as assessed by immunofluorescence, strengthens our belief that both proteins cycle to the cell surface; however, the extensive colocalization appears in disagreement with the results from subcellular fractionation of white fat. Rather than attribute the difference to the type of fat used (because both brown and white adipocytes respond very similarly to insulin) (Zorzano et al., 1989;Calderhead et al., 1990;Slot et al., 1991a;Smith et al., 1991;Cain et al., 19921, we suspect that the difference is only an apparent one arising from the substantial differences between the two methodologies (fixation and immunofluorescence versus subcellular fractionation) that limit the direct comparison. For immunofluorescence, fixation of cells acutely interrupts membrane trafficking during insulin stimulation and largely preserves spatial relationships in situ, but for two proteins that are thought to be recycling at different rates, local concentration differences indicative of partial segregation may not be detected readily. Specifically, GluT4 is likely to be partially concentrated in coated pits (Robinson et al., 19921, and as described above, we suspect that the extent of concentration will be even greater for the SCAMPs. The GluT4 concentrated in coated pits will give a stronger signal in immunofluorescence than the more diffise GluT4 in the uncoated regions of the membrane. As a result, the co-localization of GluT4 and SCAMPs will appear greater than it really is. Additionally, our immunofluorescence method is not sufficiently resolving to distinguish clearly SCAMPS in coated vesicles and endosomal membranes close to the plasma membrane from SCAMPS actually in the plasma membrane. As noted with Fig. 8, SCAMP staining at the cell border sometimes appeared more concentrated than that of GluT4; this finding is consistent with the SCAMPS being proportionately more concentrated in vesicles bordering the plasma membrane (and thus partially segregated from GluT4). Although cell fractionation has enabled identification of partial segregation of SCAMP and GluT4, it can disrupt associations that exist in intact cells. Thus vesiculated microdomains of the plasma membrane containing co-concentrated SCAMPS and GluT4 may not fractionate with the PM fraction. Alternatively, such microdomains might bud from plasma membrane fragments during homogenization at ambient temperature (which is used to prevent congealing of triglyceride). Reduced relocation of antigens to the PM fraction would be the consequence in either case, and the effect would be greater on the SCAMPS if they are more efficiently concentrated within the cell surface. The finding that insulin-stimulated GluT4 translocation to the plasma membrane is higher when measured in intact cells than by cell fractionation (Zorzano et aZ., 1989;Calderhead et al., 1990;Holman et al., 1990;Cain et al., 1992) may be a relevant illustration of the limitation of the fractionation approach. We emphasize, however, that none of the procedural limitations we have discussed affeds our basic observation that the SCAMPs and GluT4 become partially segregated from one another during insulin stimulation. When we have obtained antibodies against the SCAMPs that are suitable for EM immunocytochemistry, it will be possible to compare the distributions of SCAMPS and GluT4 at the cell border and to search for a direct morphological correlate of the partial segregation.
In conclusion, the partial segregation of the SCAMPS from GluT4 and the VAMPdsynaptobrevin upon exposure to insulin suggests that the SCAMPs are more rapidly internalized. Because of this property and because of the widespread distribution of SCAMPs among different cell types and their association with the general recycling system, it may be that the SCAMPs perform an important function in the trafficking process itself. A major challenge is to elucidate this function.