Immunopurification and Characterization of Rat Adipocyte Caveolae Suggest Their Dissociation from Insulin Signaling*

Adipocytes play an important role in the insulin-dependent regulation of organismal fuel metabolism and express caveolae at levels as high or higher than any other cell type. Recently, a link between insulin signaling and caveolae has been suggested; nevertheless, adipocyte caveolae have been the subject of relatively few studies, and their contents have been minimally characterized. With the aid of a new monoclonal antibody, we developed a rapid procedure for the immunoisolation of caveolae derived from the plasma membrane of adipocytes, and we characterized their protein content. We find that immunopurified adipocyte caveolae have a relatively limited protein composition, and they lack the raft protein, flotillin, and insulin receptors. Immunogold labeling and electron microscopy of the adipocyte plasma membrane confirmed the lack of insulin receptors in caveolae. In addition to caveolins, the structural components of caveolae, their major protein constituents, are the semicarbazide-sensitive amine oxidase and the scavenger lipoprotein receptor CD36. The results are consistent with a role for caveolae in lipid flux in and of adipocytes.

Caveolae are 50 -100-nm invaginations of the plasma membrane (PM) 1 which are formed by the expression of one or more isoforms of caveolin, the protein that produces their distinct structure (1). The membrane lipid composition of caveolae is enriched in sphingolipids and cholesterol, and caveolae represent a subtype of membrane lipid rafts, that is, subdomains of the PM with specific lipid and protein compositions (2). The physiological roles of caveolae remain uncertain (3), but they have been suggested to participate in a large number of important cellular functions. These include the formation of transcytotic/endocytic vesicles in endothelial cells (4,5), the organization/localization of numerous transmembrane signaling complexes in many cell types (6 -8), and the regulation of cellular cholesterol homeostasis (9 -12). Given all of the important roles for caveolae postulated above, it is somewhat surprising that caveolin-1 knockout animals survive and are relatively normal (13,14). On the other hand, these animals do have vascular abnormalities, particularly in the lung, and consequently, they have a reduced ability to exercise. Interestingly, with age they show abnormalities in lipid metabolism as a result of apparent adipocyte pathology (15). Indeed, normal adipocytes have perhaps the highest content of caveolae in any cell type. Estimates have been made that from 15 to 30% of the adipocyte PM are caveolae (15)(16)(17), although why these structures are so abundant in adipocytes is unknown.
It is now recognized that adipocytes play a complex and pivotal role in organismal fuel metabolism as both recipients and generators of endocrine/cytokine signals (18,19). Indeed, Bergman (20) has postulated that the actions of insulin on adipocytes are rate-limiting for the regulation of overall fuel homeostasis by this hormone. For this and other reasons, the possible role of lipid rafts and caveolae on insulin signaling and regulated GLUT4 trafficking has been studied extensively (for a recent review, see Ref. 21). As summarized by Bickel (22), the published data are contradictory in that insulin receptors and GLUT4 are reported to be localized in adipocyte caveolae by some investigators and to be absent by others. Moreover and regardless of the presence of insulin receptors there, lipid rafts that may include caveolae have been suggested to be the locus of an important signaling complex downstream from the insulin receptor, linking it to GLUT4 translocation (22)(23)(24). Thus, the possible physiological role of caveolae in insulin action has generated considerable interest and activity.
As with the study of caveolae in any cell, technical issues of isolation and purity may lie at the heart of the uncertainty regarding the composition and physiological function of adipocyte caveolae. Because caveolae are integral structures of the PM, methods needed to be devised to separate them from the bulk PM. As reviewed in Ref. 6, these methods include mechanical disruption (sonication/shearing) and/or physicochemical treatment (brief extraction in Triton X-100 at 4°C) followed by flotation in density gradients. Typically, these protocols are lengthy, and their specificity and effectiveness are questionable. Noncaveolar, detergent-resistant membrane rafts and cytoskeleton aggregates may copurify with caveolae under some of these conditions. An alternative approach is to coat the cell surface (of endothelial cells) with cationized silica to stabilize the PM and facilitate the detachment of caveolae (25). This procedure was improved further with the introduction of an immunoisolation step using anti-caveolin antibody (26,27). As noted in the latter paper, the speed of caveolae isolation can be a critical parameter, and preparations obtained by immunoiso-lation show a more limited protein composition for caveolae than those involving detergent resistance.
Here we describe a monoclonal antibody specific to caveolin-1 (7C8) which can be immobilized on acrylic beads and used to immunoisolate caveolae rapidly. We find that homogenization of adipocytes results in a certain amount of caveolae being pinched off from the PM, and these caveolae can be immunoisolated and characterized rapidly. We find caveolae purified by this method to be devoid of GLUT4, the insulin receptor, and flotillin. Indeed, and in agreement with previous freeze fracture studies (28), adipocyte caveolae have a relatively limited protein composition. In addition to the caveolins only two major protein components of caveolae were identified: (a) the semicarbazide-sensitive amine oxidase (SSAO), a very abundant adipocyte protein (29) of unknown physiological significance (30); and (b) the scavenger receptor CD36, which plays an important role in mammalian fatty acid/lipid metabolism (31,32). These results plus those from the caveolin-1 knockout (15) are consistent with a major function of adipocyte caveolae in lipid trafficking into and out of these cells.

MATERIALS AND METHODS
Antibodies and Western Blotting-Monoclonal anti-caveolin-1 antibody (clone 7C8) was raised in our laboratory following our published procedures used to obtain monoclonal antibodies against intracellular membrane proteins (33,34). The specificity of this reagent was confirmed by its recognition in Western blots of proteins corresponding in mass to caveolin-1␣ and -1␤ (see Fig. 1A) and by immunoprecipitation of caveolin with a commercial antibody followed by Western blotting with 7C8 and by the reciprocal experiment of reversing the roles of the two antibodies. Monoclonal antibodies recognizing GLUT4 (33,34) and SCAMPs (33,34) have been described previously. Polyclonal rabbit anti-peptide antibodies against the insulin receptor were prepared as in Ref. 35. The following antibodies were commercially acquired: anticaveolin-1 (C13630), anti-caveolin-2, anti-flotillin, and anti-TGN38 (from Transduction Laboratories); anti-VAMP2 (from Synaptic Systems); anti-actin (from Developmental Studies Hybridoma Bank, University of Iowa); anti-transferrin receptor (from Zymed Laboratories); anti-CD36 (Cascade Bioscience). Various researchers kindly provided sera against other proteins: SSAO (Dr. Antonio Zorzano, University of Barcelona, Spain); VAMP3 (Dr. Ronald Corley, Boston University School of Medicine); insulin receptor (Dr. Ken Siddle, Cambridge University, UK). Primary antibodies were detected in Western blots using secondary antibodies conjugated to horseradish peroxidase (Sigma) diluted 1:3,000 and chemiluminescent substrate (PerkinElmer Life Sciences).
Subcellular Fractionation of Adipocytes-The protocol was adapted from Simpson et al. (36) as described previously (37). Briefly, epididymal fat pads were removed from male Sprague-Dawley rats (150 -175 g) and transferred to KRP (12.5 mM HEPES, 120 mM NaCl, 6 mM KCl, 1.2 mM MgSO 4 , 1 mM CaCl 2 , 0.6 mM Na 2 HPO 4 , 0.4 mM NaH 2 PO 4 , 2.5 mM glucose, and 2% BSA (pH 7.4)) at 37°C. Isolated adipocytes were obtained by collagenase B (Roche Applied Science) treatment at 37°C for 45 min (38). After recovery from digestion for 45 min, cells were stimulated or not with 20 nM insulin for 15 min. Hormonal action was stopped with 2 mM KCN. Cells were then transferred to HES (20 mM HEPES, 5 mM EDTA, 250 mM sucrose (pH 7.4)) and homogenized with a Teflon-glass tissue grinder. Subcellular fractions (PM, mitochondria and nuclei, heavy microsomes (HM), and light microsomes (LM)) were obtained by differential centrifugation and resuspended in HES. LM could be fractionated further (see Fig. 9) by sucrose velocity gradient (37). Microsomes (0.3 mg of total protein) were loaded over 4.6 ml of a 10 -35% (w/v) sucrose gradient in 20 mM HEPES, 5 mM EDTA and spun for 55 min at 280,000 ϫ g max . After centrifugation, fractions were collected from the bottom of the tube. All buffers used with subcellular fractions in this work contained a mixture of protease inhibitors consisting of 1 M aprotinin, 10 M leupeptin, 1 M pepstatin (American Bioanalytical), and 5 mM benzamidine (Sigma).
Immunofluorescence-The procedure described in Souza et al. (39) was followed. Briefly, 3T3-L1 adipocytes at day 8 or 9 of differentiation were fixed in 2% paraformaldehyde for 10 min at 25°C, washed, and treated with primary antibody (2-8 g/ml) and respective secondary antibody labeled with Cy-3 or Cy-5 (Jackson Immunoresearch) diluted 1:250. Staining of lipid droplets was achieved with 1 M Nile Red (Sigma). Fluorescence of dyes was assessed by confocal microscopy.
Immunoprecipitation-The PM fraction (50 -100 g of total protein) suspended in PBS was solubilized with 60 mM octyl glucoside for 2 h at 4°C with constant agitation. Insoluble material was removed by pelleting for 10 min in a microcentrifuge. Monoclonal and polyclonal anticaveolin antibodies or nonspecific mouse and rabbit IgGs (5 g) were incubated with the supernatant overnight at 4°C, then 20 l of protein A beads (Pierce) was added for 4 h. The supernatant with unbound proteins was collected, and the beads were washed four times with octyl glucoside in PBS buffer, rinsed once PBS, and eluted with SDS-PAGE loading buffer containing 2% SDS.
Immunoadsorption of LM-Protein A-purified 7C8 antibody as well as nonspecific mouse IgG (Sigma) were immobilized to acrylic beads (Reacti-gel GF 2000, Pierce) at ϳ1 mg of antibody/ml of resin, according to instructions from the manufacturer. Beads were blocked with 2% BSA in PBS for 2 h and washed with PBS. Microsomes resuspended in PBS (containing 0.1% BSA for biotinylated samples) were added at 5-20 g of total protein/l of resin for 16 h at 4°C. The supernatant was recovered, and beads were washed with PBS. Bound vesicular proteins were eluted sequentially with 1% Triton X-100 in PBS and sample buffer for PAGE containing 2% SDS.
Electron Microscopy-LM were precleared with nonspecific IgG beads for 16 h at 4°C. The unbound fraction was mixed with 7C8 beads for 16 h at 4°C. Beads were washed extensively with PBS, and immunoadsorbed material was eluted with 100 l of 0.2 M NaHCO 3 (pH 11.0) in the presence of 0.1% BSA for 30 min on ice. The supernatant was collected and the pH adjusted to 7.0 with 6 M HCl. The sample was fixed with 0.4% OsO 4 for 1 h on ice, applied to carbon-coated 300 mesh copper grid, and incubated 30 min at room temperature for adsorption. The grid was rinsed sequentially with water and 1% sodium phosphotungstate (pH 7.4). Samples were analyzed on a Philips CM12 transmission electron microscope.
Immunogold electron microscopy was performed in the following fashion. Adipocyte PM, adsorbed to EM grids with their cytoplasmic face exposed, were prepared and labeled as described previously (17). The primary monoclonal antibody against the C terminus of the insulin receptor ␤-subunit (CT-1) was kindly provided by Dr. K. Siddle, Cambridge, UK. The primary antibody against caveolin has been described earlier (40). The secondary antibody (rabbit anti-mouse immunoglobulins, DAKO, Denmark) was conjugated with 5-nm gold particles according to the protocol of Slot and Geuze (41). The peptide used for control of the specificity of insulin receptor labeling was comprised of the 15 C-terminal amino acids of the human insulin receptor and was obtained from a commercial source.
Vectorial Biotinylation of Membrane Proteins-For cell surface labeling, some modifications of the procedure described in Ref. 42 were made. Primary amines of BSA from KRP buffer were blocked with acetic acid N-hydroxysuccinimide ester (Sigma) for 1 h at 37°C and the reagent removed by dialysis at 4°C. Isolation of adipocytes proceeded using KRP with blocked BSA until just before biotinylation, when cells were washed with KRP without BSA. Sulfosuccinimidobiotin (sulfo-NHS-biotin, Pierce) was added to adipocytes at a concentration of 0.5 mg/ml and incubated for 2-15 min in the presence or absence of insulin, according to specific experimental design. After labeling, cells were treated with 50 mM Tris (pH 7.4) to quench unreacted biotin and 2 mM KCN, then the fractionation followed the regular protocol. For biotinylation of isolated vesicles (either total LM or vesicles enriched by velocity gradient (43), samples were treated with 0.5 mg/ml sulfo-NHS-biotin for 30 min at 37°C. The reaction was stopped with 100 mM Tris (pH 7.4) and vesicles recovered by centrifugation (230,000 ϫ g max , 120 min). Detection of biotinylated proteins transferred to Immun-Blot polyvinylidene difluoride membranes (Bio-Rad) was performed with 100 ng/ml streptavidin-horseradish peroxidase conjugate (Pierce) prepared in PBS-Tween with 2% BSA and chemiluminescent substrate.
Protein Microsequencing-Sequence analysis was performed under the supervision of Dr. William Lane at the Harvard Microchemistry Facility by microcapillary reverse-phase high performance liquid chromatography nanoelectrospray tandem mass spectrometry on a Finnigan LCQ quadrupole ion trap mass spectrometer. Fig. 1A shows that caveolin-1␣ and -1␤ are detected by Western blotting with 7C8, whereas a commercial polyclonal antibody (no longer available) recognizes only the ␣-form. These caveolin isoforms differ in the N terminus because of alternate initiation sites for translation, methionines 1 and 32, respectively, for ␣ and ␤ (44), and thus 7C8 must recognize an epitope between residue 32 and the C terminus of caveolin. Fig. 1B shows by immunofluorescence that caveolin detected by 7C8 is almost exclusively localized in the cell surface, as expected. Fig. 1C shows the labeling pattern in fractionated rat adipocytes of caveolin (by 7C8), as well as that for various other proteins of interest from resting and insulin-treated cells. As expected, insulin causes a redistribution of intracellular (LM and HM) GLUT4 to the cell surface (PM) (33). As we demonstrated previously by Western blotting (45), the insulin receptor undergoes ligand-dependent endocytosis, whereas caveolin shows a small insulin-dependent decrease in the LM as a result of apparent redistribution to the PM (37). The change in PM caveolin cannot be accurately measured because of the large amount of caveolin already at the cell surface. Interestingly, flotillin, reported to be associated with caveolin by coimmunoprecipitation (46), has a different distribution than caveolin in the LM and HM membrane fractions (Fig. 1C), does not coimmunoprecipitate with anti-caveolin antibodies (Fig. 2), and has a different sedimentation pattern than caveolin (see Fig. 9).

Characterization of Anti-caveolin Antibody, 7C8 -
Immunoprecipitated Caveolin Complexes from the Adipocyte PM Exclude the Insulin Receptor-Given that several groups have shown association of the insulin receptor with caveolin both in cell-free pulldowns (47) and by biochemical and morphological methods in adipocytes (48 -50), we wished to deter-mine whether this was the case in rat adipocyte PM. We used monoclonal 7C8 and polyclonal anti-caveolin-1 antibodies as shown in Fig. 2 to immunoprecipitate proteins from the rat adipocyte PM and then Western blotted with the indicated antibodies. As has been shown in other cells (51), caveolins-1 and -2 form a stable complex that can be immunoprecipitated with both antibodies. None of these antibodies coimmunoprecipitate the insulin receptor under the described conditions. We verified this result with 7C8 using similar immunoprecipitation conditions except that Triton X-100 was the detergent used (data not shown). Interestingly, the caveolin complex immunoprecipitated with 7C8 or caveolin-1 antibodies excluded flotillin (also, see Figs. 7 and 9). Note that the lanes designated 7C8, IP, show nonspecific bands that bracket the location of flotillin.
Microsomal Caveolin-rich Membranes Contain "Pinched Off " Caveolae-Although most of caveolin is in the PM fraction, ϳ10% of the protein is found in internal membrane fractions ( Fig. 1C and Refs. 37 and 52). Our original assumption was that this represented caveolin-rich vesicles, perhaps Golgi-derived (53), and we decided to immunoisolate and characterize these membranes. Antibody 7C8 was immobilized on acrylic beads and was used to immunoadsorb membranes from the LM fraction as shown in Fig. 3. After binding of membranes, proteins were eluted sequentially in nondenaturing (1% Triton X-100) and denaturing (2% SDS) conditions. Surprisingly, of the panel of eight proteins tested that are postulated to be involved in aspects of endocytosis/exocytosis, only caveolins-1 and -2 were immunoadsorbed under these conditions. Proteins known to be markers of vesicular trafficking in adipocytes such as SCAMPs and VAMPs (54) were absent as was GLUT4. The reciprocal experiment of adsorbing GLUT4 was repeated and confirmed earlier results (37) that caveolin is absent from GLUT4 vesicles (data not shown). A trans-Golgi network marker (TGN38) and a recycling endosomal marker (transferrin receptor) also did not colocalize with immunoadsorbed caveolin-rich membranes. In more than 10 experiments, an average of 60% of caveolin-1 could be immunoadsorbed, and increasing the amount of immobilized antibody resulted in a maximum of 75% adsorption, presumably because of the topography of the caveolin complexes which prevented their complete immunoisolation. In any case, these results and those of the next two figures support the notion that caveolin-rich ves- The immunoadsorbed material obtained as in the previous figure was treated with high pH buffer (0.2 M sodium bicarbonate (pH 11)) for elution of intact membranes. Analysis by electron microscopy of the structures recovered from 7C8 beads showed them to be vesicular (Fig. 4) and identical to caveolae isolated by other investigators using independent methodology (28). These structures have a thickening of the membrane which may correspond to caveolae coat, although the characteristic striations of caveolae which can be revealed by rapid freeze etching studies (55) were absent. Moreover, it is possible to identify regions that might correspond to sealed openings of caveolae which would occur upon homogenization/disruption of cells (Fig. 4). The range of vesicle diameter, 30 -130 nm, is broader than observed for native caveolae (70 -80 nm) but corresponds well to that observed for isolated caveolae obtained by coating with cationized silica and/or Triton X-100 extraction (25,26,28).
If the structures shown in Fig. 4 represent pinched off caveolae, prior to pinching/vesiculation, their lumen should be accessible to cell surface labeling by vectorial reagents such sulfo-NHS-biotin, a negatively charged molecule that cannot cross the hydrophobic barrier of the lipid bilayer (42). Fig. 5A shows that labeling of adipocytes for periods of 2-15 min followed by immunoadsorption of the LM fraction with immobilized 7C8, elution, and detection with streptavidin conjugates results in labeling of a relatively small number of caveolar proteins. There is minimal if any change in labeling pattern as a function of labeling time in this protocol. As seen in Fig. 5B, exposure of adipocytes to insulin has no effect on labeling, indicating that this hormone does not dramatically affect the behavior and protein composition of caveolae. Fig. 5C compares the labeling pattern of the bulk PM after Triton X-100 solubilization. The pattern of soluble and insoluble PM differs substantially from immunoadsorbed caveolin, indicating that we are not simply immunadsorbing a vesiculated part of the bulk PM which happens to have some caveolin/caveolae present. Fig. 5D further shows that only cell surface proteins are being labeled by this procedure. Here, the LM fraction, including pinched off caveolae, is labeled from the cytoplasmic side with sulfo-NHS-biotin before immunoadsorption. The very heavily labeled band of 100 kDa in Fig. 5, A and B, which corresponds to SSAO (Fig. 6), is not labeled at all in this experiment because of its short cytoplasmic tail (see below). Indeed, only the caveolins are heavily labeled, indicating that the caveolae are sealed. Additional experiments where adipocytes were labeled with a thiol-cleavable, cell-impermeant biotinylation reagent, the PM isolated, solubilized, and adsorbed on immobilized avidin followed by reduction and Western blotting, revealed that 100% of cell surface SSAO could be labeled under these conditions. We observed no labeled cortical actin or G␣ s in this experiment (data not shown). These results and controls showing absence of cytoplasmic protein labeling (42) support the fact that the biotinylation reaction is vectorial and labels only cell surface-accessible proteins. Interestingly and as shown in Fig. 5, A and B, we always see significant cell surface labeling of caveolin in rat adipocytes (band designated p20). This result is in contradiction to accepted models for caveolin topology, which predict no exposure to the extracellular milieu for this protein (7). It is possible that this discrepancy may arise from reaction of caveolin with reagent after cell homogenization, a possible consequence of using rat adipocytes that require removal of fat before further analysis. The controls described above argue against this, and in any case, because SSAO cannot be labeled except from the cell surface, our interpretation remains unchanged by the caveolin labeling results.
Determination of Caveolae Protein Composition-The results of Fig. 5 led us to purify enough caveolae to determine what proteins in addition to caveolin are associated with this structure. Eluates from an immunoadsorption experiment were subjected to SDS-PAGE and silver stained (Fig. 6A). The major stained bands include caveolins at 22-24 kDa in both Triton and SDS eluates. Two additional bands were observed in the Triton X-100 eluate, with apparent molecular masses of 100 and 90 kDa. The two high molecular mass bands were cut from the gel, digested with trypsin, and the peptides obtained were sequenced by mass spectrometry. The sequence of 15 peptides derived from p100 and 7 from p90 revealed these proteins to be, respectively, SSAO and CD36. These are predicted to be transmembrane proteins with large extracellular domains and very small intracellular domains. Indeed, SSAO corresponds to the band heavily biotinylated by cell surface labeling detected at approximately 97 kDa (Fig. 5, A and B). However, there is no biotin-labeled band around the molecular mass expected for CD36 (88 kDa) (Fig. 5, A and B). Most likely, this is because the FIG. 3. Immunoadsorbed microsomal caveolin-containing membranes lack markers of membrane protein trafficking. LM from resting adipocytes (500 g) were immunoadsorbed with 50 l of 7C8-and nonspecific IgG-coupled acrylic beads. Proteins were sequentially eluted with Triton X-100 and SDS, and equivalent volumes of the eluates and unbound material (SN) were subjected to SDS-PAGE followed by Western blotting for the indicated proteins (TFR for transferrin receptor), which were visualized by chemiluminescence as in the previous figures. exceptionally heavy glycosylation of this protein (56) sterically hinders it from reaction with labeling reagents such as sulfo-NHS-biotin. The presence of SSAO in caveolae was confirmed by Western blot (Fig. 6B). In this experiment, approximately 30% of SSAO from LM was recovered along with 50% of caveo-lin. We were unable to detect CD36 by Western blotting with a variety of commercial antibodies, but others have also reported this protein to be localized in caveolae in other cell types (for review, see Ref. 56). It was expected that SSAO and caveolin will not completely colocalize in LM because some SSAO (18 -24% of total) is found in GLUT4-containing vesicles (29,57), and caveolin and GLUT4 define independent compartments (Ref. 37 and this paper). SSAO and GLUT4 may share early recycling pathways and then diverge because SSAO, differently from GLUT4, is not translocated to the PM in response to insulin stimulation (29,57).
Dissociation of Caveolae and Insulin Signaling-We could not demonstrate the association of several markers for vesicular transport with immunopurified caveolae (Fig. 3), nor could we detect the association of caveolin with insulin receptors by coimmunoprecipitation (Fig. 2) or by immunoisolation. However, flotillin has been implicated to associate with caveolin by coimmunoprecipitation (46), but we are unable to confirm this (Fig. 2). Flotillin has also been implicated in bridging insulin signaling to GLUT4 trafficking (22,24). Consequently, we examined the effects of insulin on the association of insulin receptors and flotillin with immunoadsorbed caveolae. As shown in Fig. 7 by Western blotting, the insulin receptor is absent from caveolae isolated from both insulin-stimulated and basal adipocytes (Fig. 7). As expected, GLUT4 is also absent from FIG. 5. Vectorial biotinylation reveals immunoadsorbed microsomal caveolin-containing membranes to be pinched off caveolae. Isolated adipocytes were labeled with 0.5 mg/ml sulfo-NHSbiotin for the times indicated in the absence of insulin. Labeling was stopped by the addition of 50 mM Tris (pH 7.4). The cells were homogenized and fractions isolated as described under "Materials and Methods." LM (500 g) were immunoadsorbed on 7C8-coupled acrylic beads and eluted with SDS. After SDS-PAGE and transfer, biotinylated proteins were detected by blotting with a streptavidin-horseradish peroxidase conjugate and a chemiluminescent substrate. B, adipocytes were exposed to 20 nM insulin or not for 15 min. The biotinylation reagent was added 5 min after the addition of insulin. The cells were processed and proteins analyzed as in A except that membranes were also exposed to beads coupled to a nonspecific IgG as a control. C, 50 g of PM from cells labeled for 15 min with sulfobiotin was treated with 1% Triton X-100 at 4°C for 5 min. The samples were spun at 16,000 ϫ g max for 30 min, and equivalent amounts of soluble and insoluble material were compared with 7C8-immunoadsorbed LM as in A and B. D, LM from nonstimulated adipocytes were fractionated by sucrose velocity gradient (see "Materials and Methods"). The caveolin-rich fractions were pooled and labeled with sulfo-NHS-biotin for 30 min at 37°C. Aliquots corresponding to 200 g of starting material (LM) were immunoadsorbed by 20 l of 7C8-and nonspecific IgG-coupled beads. Equivalent volumes of Triton X-100 and SDS eluates were subjected to SDS-PAGE, transferred to polyvinylidene difluoride membranes, blotted, and probed with streptavidin-horseradish peroxidase conjugate as in A.
FIG. 6. SSAO and CD36 are major components of adipocyte caveolae. A, LM from 3.4 mg of nonstimulated adipocytes were subjected to a sucrose velocity gradient, caveolin-rich fractions were pooled, divided into two equal aliquots, and immunoadsorbed with 100 l of beads coupled to either 7C8 or nonspecific mouse IgG. The total amount of proteins recovered with 1% Triton X-100 and half from SDS eluate were loaded onto a 6 -15% SDS-PAGE and silver stained. B, LM from 500 g of nonstimulated adipocytes were immunoadsorbed with 50 l of 7C8 and nonspecific IgG beads. Equivalent volumes of unbound material (SN), Triton X-100, and SDS eluates were analyzed in Western blot as in previous figures. these structures, and it changes its distribution upon insulin stimulation, as does the insulin receptor, each in the expected direction. Importantly, we see no association of flotillin with caveolae in basal or insulin stimulated adipocytes (Fig. 7) (see also Fig. 9) in contrast to previous results using gradients (58).
There are data supporting the presence of TC-10 in rafts/ caveolae (22,24). TC-10 is a rho family GTPase that regulates the actin-based cortical cytoskeleton (59), remodeling of which has been suggested to play a part in insulin-regulated GLUT4 translocation (60,61). However, we failed to observe caveolaeassociated actin by Western blotting (Fig. 7B).
The localization of the insulin receptor in the PM of adipocytes was also studied by electron microscopy of membranes from untreated cells. PM were adhered to EM grids with the cytoplasmic face exposed and labeled with mouse monoclonal antibodies against the C terminus of the insulin receptor (Fig.  8A) or caveolin (Fig. 8B), followed by gold-conjugated rabbit anti mouse immunoglobulins. Labeling of the insulin receptor (long arrows) was exclusively localized in the planar part of the PM and not in caveolae (short arrows). The labeling density was 3.9 (Ϯ0.4) (mean Ϯ S.E.) gold particles/m 2 of membrane. Controls were prepared by competitive blockade of the primary antibody by the addition of a C-terminal peptide of the receptor before incubation with the membranes (data not shown), and the resultant labeling density was 0.9 (Ϯ0.3) particles/m 2 . Similarly prepared membranes labeled for caveolin showed dense and specific labeling of all caveolae (Fig. 8B).
The dissociation of caveolin-containing rafts from those with flotillin is consistent with the results from fractionation of the LM by sucrose velocity gradient (Fig. 9). Following this procedure, the distribution of caveolin (caveolae) overlapped only to a small degree with GLUT4, flotillin, and the insulin receptor. Caveolin is enriched in fractions close to the top of the gradient where membranes containing flotillin and the insulin receptor are minimally present. On the other hand, these latter two proteins show similar but distinct distribution profiles in insulin-stimulated adipocytes. Thus, they may indeed colocalize to some degree to mediate insulin-dependent signal transduction as proposed by Saltiel, Pessin and colleagues (22,24). On the other hand, we see only traces of the insulin receptor from resting cells in flotillin-containing fractions of the gradient, whereas, after insulin stimulation there is a marked increase in the receptor with no apparent changes in flotillin (Fig. 9). These results are not consistent with the existence of a major insulin receptor-flotillin complex, although we cannot rule out a transient association between these two proteins.
FIG. 8. Immunogold electron microscopy shows insulin receptors to be in the planar regions of adipocyte PM. A, the electron micrograph (see "Materials and Methods") shows a view of the cytoplasmic face of the adipocyte PM immunogold-labeled for the C terminus of the insulin receptor ␤-subunit. All labeling (long arrows) is located in the planar part of the membrane. Caveolae, clustered or single (short arrows), are devoid of label. B, analogously prepared adipocyte membrane immunogold-labeled for caveolin. Caveolae (short arrows) are densely labeled, whereas the planar part of the membrane is unlabeled. Specimens were negatively stained with 2% silicotungstate. Scale bars indicate 200 nm. The experiment is representative of three independent studies FIG. 7. Immunoadsorbed adipocyte caveolae lack insulin receptors, flotillin, and GLUT4. LM from insulin-treated (ϩ) and -untreated (Ϫ) adipocytes (500 g) were immunoadsorbed with 50 l of beads coupled to 7C8 or nonspecific IgG beads. Equivalent volumes of unbound material (SN), Triton X-100, and SDS eluates were subjected to SDS-PAGE and analyzed by Western blotting as in the other figures. FIG. 9. Caveolin-and flotillin-containing membranes do not colocalize in velocity gradient. LM (250 g) from insulin-treated (ϩ) and -untreated (Ϫ) adipocytes were fractionated by sucrose velocity gradient, and equal volumes of the fractions were analyzed by Western blot with the antibodies indicated.

DISCUSSION
We describe the simple and rapid isolation of caveolin-containing membranes/caveolae from rat adipocyte LM using an antibody specific to caveolin immobilized to acrylic beads. This procedure has advantages over other methods (for review, see Ref. 6) in that it requires a minimum number of manipulations and can be completed rapidly, speed being a consideration to avoid possible artifacts such as the presence of noncaveolar proteins (27). We present several lines of evidence indicating that these immunologically purified membranes correspond to pinched off caveolae from the PM. These are: (a) their morphology by electron microscopy; (b) the ready labeling of their proteins from the cell surface; (c) the presence of a caveolin coat; and (d) the absence of vesicle trafficking markers (Figs. [3][4][5]. Morphological studies of adipocyte caveolae (17) reveal that many of them have very narrow necks, and it is easy to imagine that some of them can pinch off from the PM and vesiculate during homogenization of adipocytes. Although caveolin has been localized in the endocytic pathway (62,63) and the Golgi network (53,64) in some cell types, we see no evidence for this in rat adipocytes. The rapid time course of cell surface labeling (Fig. 5) did not support the possibility that caveolin-containing vesicles originate by endocytosis. The prevalence of detached caveolae over recycling endosomes/Golgi vesicles containing caveolin may reflect the proportion of these two subpopulations in adipocyte microsomes, the former being much more abundant than the latter. In any case, we also see a minimal effect of insulin on caveolae ( Fig. 1C and Ref. 37), suggesting that in the time course of our experiments, 15-30 min, rat adipocyte caveolae are relatively static cell surface structures. This conclusion is identical to that reported recently for caveolae in several cultured cell lines as studied by microscopic procedures (65).
A possible caveat that may apply to our study of adipocyte caveolae is that our experimental approach yields from 5 to 10% of the total cellular caveolin/caveolae. Thus, it cannot be excluded that what we isolate is a subpopulation of caveolae with a particular protein composition that is distinct from "average" PM caveolae. This itself would be an interesting finding in any case, and it would still support a role for these structures in lipid flux into and out of adipocytes as discussed below. However, we favor the idea that what we get is representative of the average caveolar environment because numerous morphological studies do not support the concept of heterogeneous caveolae populations. In any case, we are exploring methods to increase the yield of caveolae from adipocytes to prove this point.
We were not able to find any association with immunopurified adipocyte caveolae of some key molecules involved in insulin-sensitive glucose transport in adipocytes, namely GLUT4, the insulin receptor, and flotillin. The caveat noted above does not apply in the case of the insulin receptor as studied by electron microscopy (Fig. 8). As noted previously, there are published data both for and against the presence of GLUT4 and insulin receptors in caveolae, and we will summarize this evidence for each protein in turn and in light of our current results.
There is evidence at the electron microscopic level that GLUT4 is de-enriched (66) or absent from caveolae (17,55), or in contrast, it is present and abundant in these structures (67,68). It is known that the presence of proteins in caveolae as determined morphologically can be an artifact of fixation techniques and antibody clustering (69), but nevertheless, these contradictory data are not possible to resolve at face value. On the other hand, trafficking of GLUT4 in adipocytes represents the movement of large amounts of membrane to and from the cell interior to the cell surface and back again. The fact that we (this study) and others (65) find caveolae to be quite static structures in the time frame of vesicular traffic, 15-30 min, argues against their playing a major role in this process. On the other hand, there is evidence that clathrin-coated vesicles can serve this role for GLUT4 in adipocytes (55,70). GLUT4 association with caveolae has also been suggested on the basis of detergent insolubility (44,68), but detergent insolubility implies association with lipid rafts, not necessarily caveolae (2). Our conclusion is that the relatively rapid movement of GLUT4 to and from the cell surface of adipocytes which requires 20 -30 min for a full cycle (71) is not compatible with dynamics and composition of caveolae in these cells.
The insulin receptor has been detected in adipocyte caveolae through immunostaining of adipocytes with gold particles (49). Furthermore, it has been proposed that the receptor itself binds caveolin, based on in vitro interaction data (47,48). On the other hand, Mastick et al. (72) found no evidence for insulin receptors in caveolae from cultured adipocytes, although more recent study by some of the same authors has reached the opposite conclusion. However, we find no evidence for association of these proteins by coimmunoprecipitation (Fig. 2), immunoadsorption (Fig. 7), and immunogold electron microscopy (Fig. 8A), despite the abundance of caveolin in rat adipocytes (Fig. 8B). It has been well established that the insulin receptor undergoes ligand-dependent endocytosis (73) (shown here in Figs. 1 and 7) via clathrin-coated vesicles (74). Thus, the kinetics of insulin receptor trafficking are similar to those of GLUT4, and the same arguments put forward for GLUT4 also apply against a role for caveolae in insulin receptor dynamics. A direct role for caveolae in insulin signaling would therefore also seem unlikely, and in any case, it is unnecessary in principle because the insulin receptor signals normally in cells lacking caveolae (75).
Concerning flotillin, we cannot detect its interaction with caveolin by coimmunoprecipitation, nor was it found in immunoisolated caveolae (Fig. 7). These proteins also have a minimal colocalization in velocity gradient centrifugation (Fig. 9). Thus, we suggest that caveolin and flotillin are not ordinarily associated in the cell. The relationship of the insulin receptor to flotillin and insulin signaling remains unclear. Data supporting a role for flotillin in insulin signaling do not include experiments showing a direct interaction of these two proteins (22,24). Our data only say that flotillin does not appear to be a component of caveolae, in contradiction to what was implied in these recent studies (22,24) where immunofluorescence indicated some degree of colocalization for these proteins. However, very recent studies have suggested that this colocalization may be a result of the formation in cultured adipocytes of large PM caves that are representative of the bulk membrane composition and not isolated caveolae (76). Moreover, insulin signaling occurs in hepatocytes where caveolae are rare or absent but where association of the insulin receptor with lipid rafts was suggested (75). Further studies will be needed to address the role of flotillin in insulin signaling.
What we did find in adipocyte caveolae are the SSAO and the lipoprotein scavenger receptor CD36. SSAO is found in many tissues, but it is particularly abundant in caveolae-rich tissues including fat, lungs, and muscle (29,77). SSAO is estimated to represent 2% of total protein in the PM of adipocytes (29), and accordingly it seems highly likely that it must have an important function for those cells, but what this function might be is not clear for any cell type (30). The endogenous substrates for SSAO have not been determined, although it is quite possible that oxidation of catecholamines by SSAO modulates lipolysis in adipocytes. It was proposed that the generation of hydrogen peroxide during amine oxidation by SSAO may originate insulin-like effects such as GLUT4 translocation (57), although these studies employed high concentrations of a nonphysiological amine. It also remains unclear as to what percentage of SSAO is targeted to caveolae, although this is likely to be a high percentage based on Fig. 6.
CD36 belongs to the class B scavenger receptor family and plays an important role in lipid metabolism in mammals (31,32). Previously, this protein was proposed to localize in caveolae in lung endothelium, first based on resistance to Triton solubilization (78) and later confirmed by other methods (56).
Here we confirm the presence of CD36 in adipocyte caveolae both by immunoisolation (Fig. 6) and immunocytochemistry (data not shown). Together with CD36, another member of the same family of scavenger receptors, the HDL receptor SR-BI has also been proposed to localize in caveolae (56). The presence of both proteins in caveolae and the putative trafficking of intracellular lipoproteins containing caveolin (79) highlight the possible importance of caveolae and caveolin in the traffic of lipids in and out of the adipocytes.
We think it is of particular interest that SSAO and CD36 share the structural feature that both are transmembrane proteins with very short cytoplasmic tails comprising 5-8 amino acid residues. CD36 has two predicted transmembrane domains located at the N and C termini of the protein (80), and SSAO has a single predicted transmembrane domain at the N terminus (29). This topology is in agreement with the absence of biotin labeling for those proteins when performed from the cytoplasmic face of caveolae (Fig. 5C). Caveolins were the proteins almost exclusively labeled by such approach, and we interpret this as indicating that caveolae lack transmembrane proteins with large cytoplasmic domains. This observation is consistent with a model where a type of net originating from caveolin oligomerization (Fig. 10) covers the cytoplasmic surface of caveolae. In that model, membrane proteins with small cytoplasmic domains such as SSAO and CD36 would have free access to caveolae, whereas proteins with large cytoplasmic domains (e.g. transferrin receptor) would be excluded by steric limitations. This model can be tested by transfection of the appropriate chimeric proteins, and we are in the process of doing so.
In summary, isolation and characterization of rat adipocyte caveolae support their dissociation from insulin signaling in these cells. Furthermore, we did not find other signaling molecules among the major protein components of caveolae. Instead we found abundant CD36, a surface receptor implicated in lipid metabolism, and SSAO of unknown physiological role. Our results, together with other published data, favor the hypothesis of adipocyte caveolae as cellular nodes for control of lipid flux. Indeed, release of lipids from adipocytes has to be tightly regulated because both an excess and a shortage of these molecules can be harmful to the organism.