Characterization of a Cytosolic Heat-shock Protein-Caveolin Chaperone Complex

Caveolin is a 22-kDa protein that appears to play a critical role in regulating the cholesterol concentration of caveolae. Even though caveolin is thought to be a membrane protein, several reports suggest that this peculiar protein can traffic independently of membrane vesicles. We now present evidence that a cytosolic pool of caveolin is part of a heat-shock protein-immunophilin chaperone complex consisting of caveolin, heat-shock protein 56, cyclophilin 40, cyclophilin A, and cholesterol. Treatment of NIH 3T3 cells with 1 μm cyclosporin A or 100 nmrapamycin disrupted the putative transport complex and prevented rapid (10–20 min) transport of cholesterol to caveolae. The lymphoid cell line, L1210-JF, does not express caveolin, does not form an immunophilin-caveolin complex, and does not transport newly synthesized cholesterol to caveolae. Transfection of caveolin cDNA into L1210-JF cells allowed the assembly of a transport complex identical to that found in NIH 3T3 cells. In addition, newly synthesized cholesterol in transfected cells was rapidly (10–20 min) and specifically transported to caveolae. These data strongly suggest that a caveolin-chaperone complex is a mechanism by which newly synthesized cholesterol is transported from the endoplasmic reticulum through the cytoplasm to caveolae.

Caveolae are cholesterol/sphingomyelin-rich plasma membrane microdomains that can be found in most cells (1,2). Historically, caveolae were defined solely based on an invaginated morphology, however, this definition has recently been expanded to include flat membranes (3)(4)(5), as well as a subclass of vesicles that have a similar lipid composition and similar biophysical properties (6). Cholesterol is central to the structure and function of caveolae (7)(8)(9). Proteins linked to the extracellular side of the plasma membrane by glycosylphosphatidylinositol associate with caveolae in a cholesterol-dependent manner (10,11). In addition, the cytoplasmic surface of caveolae is covered with a characteristic coat structure that is disrupted and partially disassembled by cholesterol-binding drugs such as filipin and nystatin (10,12). Cholesterol-binding drugs also cause invaginated caveolae to flatten within the plane of the membrane (10). Studies with cholesterol-depleted cells showed that the number of invaginated caveolae were dramatically reduced. The number of invaginated caveolae returned to control levels when the cells were cholesterol replete (10). In some cells, caveolae cycle between invaginated and flat morphologies suggesting that both structural states have essential roles (13).
Caveolin is a membrane protein that seems to be involved in regulating caveolae cholesterol (8,14,15). However, other functions for caveolin have been proposed such as vesicle sorting (16,17) and a molecular "scaffold" interacting with heterotrimeric G-proteins (18). Several isoforms of caveolin have been cloned (19,20) and shown to be acylated, phosphorylated, and localized in proximity of the caveolae coat material (21)(22)(23). Currently the relationship between the function(s) of caveolin, caveolin isoforms, and cell type is unclear.
We have generated three lines of evidence demonstrating that caveolin is involved in cholesterol trafficking. First, transfection of caveolin-1 cDNA into a lymphocyte cell line lacking morphological caveolae or an enrichment of cholesterol in isolated caveolae (sometimes called GEMs, glycolipid-enriched membranes) domains induce characteristic invaginations and a 4-fold enrichment of cholesterol with respect to bulk plasma membrane (14). Second, oxidation of cholesterol in normal human fibroblasts with cholesterol oxidase caused caveolin to translocate from caveolae to the Golgi region of cells. Caveolin did not move directly to the Golgi, rather it migrated to the lumen of the ER, 1 moved through a Golgi-ER intermediate compartment, and then accumulated in the Golgi (8). Once cholesterol oxidase was removed, non-oxidized cholesterol and caveolin returned to caveolae by a microtubule-dependent process (24). This appears to be a normal trafficking pattern for the processing and transport of caveolin. Third, we recently used [ 3 H]acetate and pulse-chase assays to demonstrate the active involvement of caveolin in the rapid movement of cholesterol from the ER to the cell surface (14). Murata et al. (9) have demonstrated that caveolin binds cholesterol in vitro and Fra et al. (25) showed that caveolin plays a role in forming invaginated caveolae. Recently, Fielding and Fielding (26,27) showed that low density lipoproteinderived free cholesterol traffics to caveolae and that caveolin is regulated by oxysterols. In addition, Babitt et al. (28) have demonstrated that SR-B1, a high density lipoprotein receptor, which mediates the selective uptake of cholesterol esters resides in caveolae. Taken as a whole these data suggest that caveolae act as a central regulator of intracellular cholesterol flux. However, past studies (8,10,23) were unable to find evidence that caveolae membranes directly transported cholesterol. It is difficult to envision a mechanism whereby caveolin, a membrane-associated protein, could translocate cholesterol through the cytoplasm. Chaperone-protein complexes are often found associated with proteins that contain hydrophobic domains. We now present evidence that caveolin is part of a cytosolic HSP-immunophilin chaperone complex and that this cytosolic complex binds to and translocates newly synthesized cholesterol to caveolae membranes.

Materials
Cell Culture-NIH 3T3 cells were cultured in a monolayer and set up according to a standard format. On day 0, 10,000 cells were seeded into 100-mm dishes with 5 ml of DMEM supplemented with 100 units/ml penicillin/streptomycin, 0.5% (v/v) glutamine, and 10% (v/v) calf serum. The cholesterol pool was radiolabeled by changing the medium on day 3 to DMEM plus 20 mM HEPES, pH 7.4, adding [ 3 H]acetate (30 Ci/ dish), and incubating for the indicated times. All experiments were carried out on day 4. L1210-JF cells (kindly provided by Dr. Bart Kamen) are a murine lymphocyte cell line that does not express caveolin (14). On day 0, 1 ϫ 10 5 cells were seeded onto 100-mm dishes in RPMI medium 1640 plus 0.5% (v/v) glutamine, 25 mM HEPES, pH 7.4, and 10% (v/v) calf serum. Transfected cell medium also contained 300 g/ml Geneticin. The cholesterol pool was labeled the same as NIH 3T3 cells and used on day 4.
Isolation of Caveolae-Caveolae membranes were isolated as described previously (29) with the modifications described below. This procedure generates a highly purified plasma membrane microdomain that is free from intracellular markers (data not shown) and bulk plasma membrane markers. This method has been used extensively to characterize caveolae membrane signaling events (11, 14, 29 -32).
The top 5 ml of the first OptiPrep gradient was collected, placed in a fresh SW41 centrifuge tube, and mixed with 4 ml of Buffer B. The sample was overlaid with 1 ml of 15% (v/v) OptiPrep and 0.5 ml of 5% (v/v) OptiPrep (prepared by diluting Buffer B with Buffer A) and centrifuged at 52,000 ϫ g for 90 min at 4°C. A distinct opaque band was present at both interfaces. The band at the 5% interface was collected and designated caveolae membranes. We typically obtained 10 -20 g of protein in this band. The cytosol fraction is obtained by collecting the overlaid material on the Percoll gradient, centrifuging at 350,000 ϫ g for 1 h, and collecting the resulting supernatant. The intracellular membranes are obtained by pooling all of the Percoll fractions below the plasma membrane and then taking the pellet after centrifuging at 100,000 ϫ g for 1 h. The bulk plasma membranes (BPM) are obtained by pooling the bottoms of both OptiPrep gradients. The BPM were originally called non-caveolae membranes but because they still contain a significant amount of caveolae we are now calling this fraction the BPM to more accurately reflect its contents.
Radiolabeled Cholesterol Determination-Thin-layer chromatography and liquid scintillation counting was used to measure the amount of [ 3 H]sterol in each sample (14). Each sample was adjusted to a volume of 1 ml with distilled water. Methanol (1.2 ml) containing 2% (v/v) acetic acid was added to the sample before vortexing two times, 30 s each. Chloroform (1.2 ml) was then added and the sample vortexed two times, 30 s each. The organic and aqueous phases were separated in a Beckman Clinical Centrifuge at 3750 rpm, 15 min, room temperature. The organic phase was dried under nitrogen and then suspended in 50 l of the solvent system (80:20:1; petroleum ether:ethyl ether:acetic acid). Pure cholesterol was dissolved in the solvent system and used as a standard (5 g/spot). Lipids were visualized by charring with sulfuric acid:ethanol and heating at 180°C for 10 min. Unlabeled cholesterol was added to each fraction to facilitate visualization. The appropriate spots were scraped and the amount of radiation quantified by liquid scintillation counting.
Immunoprecipitations-Protein A-Sepharose beads were first blocked by incubating them for 4 h at 4°C with the appropriate cell lysate (200 g/ml) plus 30 mg/ml bovine serum albumin in Buffer E or RIPA buffer (150 mM NaCl, 1.0% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 50 mM Tris, pH 8.0). Blocked beads were then used to pre-clear the experimental fractions that had been adjusted to 1% (v/v) Triton X-100, 60 mM octyl glucoside, and 0.1% (w/v) SDS. Pre-cleared fractions were then incubated for 18 h at 4°C with the appropriate antibody before adding blocked, Protein A-Sepharose beads and incubating an additional 2 h at 4°C. The beads were collected by centrifugation, washed five times in high salt RIPA buffer (500 mM NaCl), and then placed in sample buffer or lipids extracted and analyzed by TLC (33). Immunoprecipitated proteins were detected by immunoblots or silver staining. Slight alterations in the protein to detergent ratio dramatically effected the solubility of caveolae components, in particular, caveolin (data not shown). Consequently, SDS was used to ensure consistent and complete solubilization of membranes. Without the addition of 0.1% (w/v) SDS, Ras often co-immunoprecipitated with caveolin which most likely represents incomplete solubilization of caveolae membranes (data not shown). In addition, for the silver stain experiment (Fig. 1B), 30 mg/ml L-lysine was substituted for the bovine serum albumin.
Transfection with Caveolin cDNA-A full-length caveolin-1 cDNA was subcloned into a pCI-NEO vector (Promega) using EcoRI sites. Twenty-four hours before transfection about one million cells were plated per 100-mm dish. On the day of transfection, 5 g of plasmid DNA was diluted in 200 l of serum-free RPMI media. In a separate tube, 20 l of LipofectAMINE was diluted in 200 l of serum-free RPMI media. The diluted DNA and LipofectAMINE were then gently mixed and incubated at 25°C for 30 min. After incubation, 6 ml of serum-free RPMI media was added to the DNA:LipofectAMINE mixture, mixed, then placed onto cells rinsed with serum-free RPMI media. The cells were incubated for 5 h at 37°C. Without washing, 3.6 ml of RPMI media containing 20% calf serum (v/v) was added. The cells were grown for 24 h. The media was removed and RPMI media containing 10% calf serum (v/v) and 1.5 mg/ml Geneticin was added. The cells were grown under constant selection in medium containing 300 g/ml Geneticin.
Electrophoresis and Immunoblots-Samples were concentrated by trichloroacetic acid precipitation and washed in acetone (29). Pellets were suspended in Buffer F that contained 1.2% (v/v) ␤-mercaptoethanol and heated at 95°C for 3 min before being loaded onto gels. Proteins were separated in a 12.5% SDS-polyacrylamide gel using the method of Laemmli (33). The separated proteins were then transferred to PVDF. The PVDF was blocked in Buffer D that contained 5% dry milk for 1 h at room temperature. Primary antibodies were diluted in Buffer D that contained 1% dry milk and incubated with the PVDF for 1 h at room temperature. The PVDF was washed four times, 10 min each in Buffer D ϩ 1% dry milk. The secondary antibodies (goat anti-mouse IgG, goat anti-rabbit IgG, or goat anti-rat IgG all conjugated to horseradish peroxidase) were diluted 1:20,000 in Buffer D ϩ 1% dry milk and incubated with the PVDF for 1 h at room temperature. The PVDF was then washed and the bands visualized using enhanced chemiluminescence (Pierce).
Enzyme Assays-Alkaline phosphatase was measured by the method of Engstrom (34). Galactosyl transferase and NADPH cytochrome c reductase were assayed using methods adapted from Graham and Higgins (35). Lactate dehydrogenase was measured with a standard kit from Sigma.
Immunoelectron Microscopy-Immunogold localization of cyclophilin A and caveolin was carried out using whole mount plasma membrane preparations as described previously (36). Membranes attached to grids were incubated sequentially for 30 min each with the indicated primary antibodies (anti-cyclophilin A IgG diluted 1:500 or anti-caveolin IgG-2234 diluted 1:100), followed by 50 g/ml goat anti-mouse IgG, and then a 1:30 dilution of gold-conjugated rabbit anti-goat IgG. Phosphatebuffered saline containing 0.15% bovine serum was used to dilute all of the antibodies. After each antibody incubation, the grids were washed three times for 30 min each in phosphate-buffered saline containing 0.15% bovine serum albumin. The membranes were then fixed with 2.5% glutaraldehyde in phosphate-buffered saline for 1 min followed by 1% osmium tetroxide in phosphate-buffered saline for 10 min. Each sample was stained sequentially for 10 min with 1% tannic acid and 1% uranyl acetate. The grids were examined and photographed with an electron microscope (100 CX; JEOL USA Inc., Peabody, MA).
Other Methods-Protein concentrations were determined using the Bio-Rad Bradford kit (37). OptiPrep interfered with other standard protein determination methods including the micro-Bradford assay.
Each [ 3 H]acetate labeling experiment was conducted at least six times. The mean values with standard error are shown as error bars. Because the specific activity of the synthesized cholesterol was not determined for each experiment the data are presented as disintegrations per minute of [ 3 H]sterol per sample.

RESULTS
Caveolin-Chaperone Complex-Caveolin is intimately involved in the rapid transport of newly synthesized cholesterol from the ER to caveolae membranes (14). Two possible models can explain the mechanism of caveolin-dependent cholesterol transport: vesicle-dependent and vesicle-independent. We have been unable to find evidence of caveolin-containing transport vesicles (14). Previous work by Kaplan and Simoni (38) and others (39,40) have suggested that the direct transport of cholesterol to plasma membrane is dependent on a low-density transfer intermediate. However, the identity of the transfer intermediate is still unknown. One possibility is protein-lipid chaperone complexes. Protein-chaperone complexes can provide a membrane independent mechanism to transport hydrophobic molecules through the cytoplasm (41).
To test if a cytosolic pool of caveolin exists we centrifuged NIH 3T3 cell lysates at 350,000 ϫ g for 1 h. The supernatant (cytosol) contained all of the lactate dehydrogenase activity and the pellet (membrane) contained all of the alkaline phosphatase activity (data not shown). Fifty micrograms from the cytosol, the membrane, and the lysate were resolved by SDS-PAGE and immunoblotted for caveolin (Fig. 1A). Approximately 85-90% of the caveolin (compared with lysate) was associated with the membranes. However, 10 -15% (compared with lysate) of the caveolin consistently separated with the cytosol fraction.
To test if this cytosolic pool of caveolin is part of a proteinchaperone complex we separated NIH 3T3 cell lysates into cytosol and membrane fractions and used the total of each fraction to immunoprecipitate caveolin. The immunoprecipitation conditions were stringent to ensure solubilization of the membranes and to limit artifactual associations (see "Experimental Procedures"). The total precipitated material was resolved by SDS-PAGE and silver stained (Fig. 1B). The gels were developed in the silver stain reagent for a short time (ϳ4 min) so that only the most prominent bands were visualized. Four distinct bands of approximately 18, 22, 41, and 58 kDa were detected in the cytosol fraction (Fig. 1B, Cytosol). The membrane fraction contained an extremely prominent band at 22 kDa and a band at 18 kDa but lacked the 41-and 58-kDa bands (Fig. 1B, Membrane). Overexposure of the membrane lane still did not allow the detection of the 41-and 58-kDa bands (data not shown). The 22-kDa band most likely represents caveolin, consequently the silver stain and immunoblot data are qualitatively in agreement with regards to the relative distribution of the protein.
To further characterize these proteins we subfractionated NIH 3T3 cells into cytosol and membrane fractions and then immunoprecipitated caveolin from each fraction (total fraction). The precipitated material was divided into four equal volume aliquots and resolved by SDS-PAGE, transferred to nylon, and probed for known chaperone proteins. Each protein seen in the silver stain was subsequently identified by immunoblotting ( Fig. 2A) to be cyclophilin A (18 kDa), caveolin (22 kDa), cyclophilin 40 (41 kDa), and HSP56 (58 kDa). We also probed the nylons for protein kinase C␣, HSP90, HSP100, and cHSP70 all of which have been shown to be associated with chaperone complexes (41,42). These proteins were not detected in any of the precipitates (data not shown). In addition, an irrelevant antibody, 2001, did not precipitate any of the proteins (data not shown). Importantly, the putative chaperone complex could only be precipitated from the cytosol fraction ( Fig. 2A) which suggests that the complex is cytosolic.
To confirm that caveolin IgG was not inducing artifactual protein aggregation, IgGs for HSP56, cyclophilin 40, caveolin, and cyclophilin A were incubated with NIH 3T3 cytosol then precipitated with protein A-Sepharose. IgGs for HSP56, cyclophilin 40, caveolin, and cyclophilin A all precipitated the same set of four proteins (Fig. 2B). Semi-quantification of a series of immunoblots (near-linear range) allowed us to estimate the percent of each protein co-immunoprecipitated by caveolin with respect to the total cytosolic pool: caveolin, Ͼ95%; cyclophilin A, 3-6%; cyclophilin 40, 18 -24%; HSP56, 6 -9%.
Cyclophilin A also co-precipitated with caveolin from the membrane fraction ( Fig. 2A, CypA). Although cyclophilin A is predominately found in the cytosol the protein also localized to caveolae on the plasma membrane by immunoelectron microscopy (Fig. 3).
Reconstitution of the Caveolin-Chaperone Complex-We previously showed that L1210-JF cells do not have invaginated FIG. 1. A cytosolic pool of caveolin. A, NIH 3T3 cells lysed in Buffer A by a Dounce homogenizer were centrifuged at 350,000 ϫ g for 60 min at 4°C to separate cytosol (supernatant) and membrane (pellet). Fifty micrograms from the cytosol, the membrane, and the lysate were resolved by SDS-PAGE and immunoblotted for caveolin. A series of immunoblots (near-linear range) detecting differing amounts of protein were used to generate an estimate of the percent caveolin in the cytosol fraction (data not shown). B, NIH 3T3 cells were fractionated as described above. The total amount of protein in each fraction was immunoprecipitated with caveolin IgG in Buffer E. The total amount of each precipitate was resolved by SDS-PAGE and silver stained. Representative data from four separate experiments are shown. caveolae, do not rapidly transport cholesterol to the plasma membrane, and do not contain detectable levels of caveolin in whole cell immunoblots (14). Transfection of caveolin into these cells facilitated the rapid (10 min) transport of cholesterol to caveolae and the formation of invaginated caveolae (14). We concluded that caveolin was essential for rapid and caveolaespecific cholesterol trafficking (14).
We next used L1210-JF cells to determine if the formation of a cytosolic caveolin-chaperone complex was a general phenomenon or specific to NIH 3T3 cells. First, we determined if the stable transfection of caveolin into L1210-JF cells altered the relative levels of chaperone protein expression. L1210-JF cell lysate was centrifuged at 350,000 ϫ g for 1 h to isolate a cytosol and membrane fraction. Fifty micrograms of cytosolic protein from cells lacking caveolin and from cells expressing caveolin were resolved by SDS-PAGE and immunoblotted for HSP56, cyclophilin 40, caveolin, and cyclophilin A. Fig. 4A demonstrates that both caveolin non-expressing and caveolin expressing cells had qualitatively the same amount of HSP56, cyclophilin 40, and cyclophilin A.
To determine if expression of caveolin induced the formation of the chaperone complex, we incubated HSP56 IgG with cytosol isolated from L1210-JF cells that did or did not express caveolin. Only HSP56 was precipitated from cells not expressing caveolin (Fig. 4B, Ϫ). The entire chaperone complex was co-immunoprecipitated with HSP56 IgG from cells expressing caveolin (Fig. 4B, ϩ). Similar data were obtained when cyclophilin 40 IgG or cyclophilin A IgG were used (data not shown).
Although the data presented thus far clearly demonstrates that HSP56, cyclophilin 40, caveolin, and cyclophilin A directly interact it is possible that the interaction is a post-lysis artifact. Because the complex is a soluble cytosolic component we were unable to use cross-linkers without first disrupting the cells. Consequently, we performed a radiolabeled mixing experiment to determine if the complex forms after cell lysis. Briefly, L1210-JF cells not expressing caveolin were metabolically labeled with [ 35 S]methionine (0.5 mCi/ml) for 24 h. Control immunoprecipitations with IgGs for HSP56, cyclophilin 40, and cyclophilin A showed that essentially the entire cytosolic pool of these proteins was radiolabeled (data not shown). A 10-fold excess of radiolabeled cells not expressing caveolin was mixed with unlabeled cells that were expressing caveolin. The cells were then lysed and cytosol isolated exactly as described above. Caveolin IgG was added to the cytosol to precipitate caveolin and associated proteins. The precipitate was resolved by SDS-PAGE, the gel was dried and exposed to film to detect radiolabeled proteins. Caveolin IgG did not co-immunoprecipitate any radiolabeled bands (Fig. 5, lane 1) (exposures up to 16 days)

FIG. 2. Immunoprecipitation of a cytosolic caveolin-chaperone complex. A, NIH 3T3 cells lysed in Buffer
A by a Dounce homogenizer were centrifuged at 350,000 ϫ g for 60 min at 4°C to separate cytosol (supernatant) and membrane (pellet). The total amount of protein in each fraction was immunoprecipitated with caveolin IgG in Buffer E. The precipitates were divided into four equal volume aliquots, resolved by SDS-PAGE, and immunoblotted with IgGs for HSP56, cyclophilin 40, caveolin, and cyclophilin A. The nylons immunoblotted with caveolin (Caveolin) and cyclophilin A (CypA) were exposed for 15 s, whereas the nylons immunoblotted with cyclophilin 40 (Cyp40) and heat-shock protein 56 (HSP56) were exposed for 60 s. C, cytosol; M, membranes. B, NIH 3T3 cells lysed in Buffer A by a Dounce homogenizer were centrifuged at 350,000 ϫ g for 60 min at 4°C to separate cytosol (supernatant) and membrane (pellet). The total amount of protein in the cytosol fraction was separated into four equal volume aliquots and immunoprecipitated with IgGs for HSP56, cyclophilin 40 (Cyp40), caveolin (Caveolin), or cyclophilin A (CypA) in Buffer E. Each immunoprecipitate was resolved by SDS-PAGE and immunoblotted concurrently with a mixture of IgGs for HSP56, Cyp40, caveolin, and CypA. Each nylon was exposed for 35 s. A series of immunoblots (near-linear range) detecting differing amounts of protein were used to generate an estimate of the percent caveolin in the cytosol fraction (data not shown). Representative data from four separate experiments are shown.

FIG. 3. Localization of caveolin and cyclophilin A in plasma membranes by immunoelectron microscopy.
Immunogold labeling was performed using monoclonal antibody anti-caveolin (A) or using monoclonal antibody anti-cyclophilin A (B). Gold particles were found over small invaginations on the membrane surface that had the characteristic appearance of caveolae, whereas coated pits and smooth membrane areas were not decorated with gold. Controls with an irrelevant antibody (mAb 2001) were negative (data not shown). Caveolae (arrowheads) and coated pits (marked with an X) are evident in both images. Bar, 0.45 m.
which strongly suggests that the complex is not the result of a post-lysis artifact. As a control, we radiolabeled the same amount of cells expressing caveolin as used in the mixing experiment. Caveolin IgG precipitated all four radiolabeled proteins from control cells (Fig. 5, lane 2

) (exposure time 48 h).
Disruption of the Caveolin-Chaperone Complex-Cyclophilin A, cyclophilin 40, and HSP56 are immunophilins, a class of proteins that have proline isomerase activity, form chaperone complexes, and are involved in intracellular transport (41). Cyclosporin A and rapamycin are pharmacological reagents that disrupt immunophilin-chaperone complexes and alter intracellular trafficking (41). Treatment of live, intact cells with these compounds should disrupt the caveolin-chaperone complex. To test this hypothesis, NIH 3T3 cells were treated with 1 M cyclosporin A or 100 nM rapamycin for 1 h at 37°C before whole cell lysates were prepared and cytosols isolated. In one set of cells cyclosporin A was washed out and the cells allowed to recover for 1 h prior to lysis. Caveolin IgG was used to immunoprecipitate caveolin-associated proteins from isolated cytosols. Treatment with cyclosporin A (Fig. 6, CysA) prevented the association of cyclophilin A and cyclophilin 40 with caveolin, whereas rapamycin (Fig. 6, Rap) prevented association with HSP56 and cyclophilin 40. Removal of cyclosporin A for 1 h (Fig. 6, Wash-out) allowed the putative complex to reform. The concentration of each drug and time of incubation were optimized for maximal effects (data not shown). Cyclosporin A and rapamycin also inhibited complex formation in L1210-JF cells expressing caveolin (data not shown).
Caveolin Co-immunoprecipitates Sterol-We previously showed that caveolin plays a role in the direct transport of newly synthesized cholesterol from the ER to caveolae (14). We have now provided evidence for a cytosolic caveolin-chaperone complex that may provide a mechanism for the transport of cholesterol through the cytoplasm. To determine if the caveolin-chaperone complex and cholesterol directly interact in intact cells, cellular cholesterol pools were radiolabeled by incubating NIH 3T3 cells for 16 h in the presence of [ 3 H]acetate at 37°C (14). The cells were then washed and subfractionated using an established, detergent-free method to isolate caveolae (29).
Caveolin IgG was used to immunoprecipitate (see "Experimental Procedures") caveolin from each subcellular fraction using an equal amount of protein (relative specific activity) or

FIG. 4. L1210-JF cells expressing caveolin contain a chaperone complex. A, L1210-JF cells not expressing caveolin (Ϫ) and cells ex-
pressing caveolin (ϩ) were lysed in Buffer A by a Dounce homogenizer and centrifuged at 350,000 ϫ g for 60 min at 4°C to separate cytosol (supernatant) and membrane (pellet). Fifty micrograms from the cytosol was resolved by SDS-PAGE and immunoblotted concurrently with a mixture of IgGs for HSP56, Cyp40, caveolin, and CypA. Each nylon was exposed for 60 s. B, cytosol (200 g) from L1210-JF cells not expressing caveolin (Ϫ) and cells expressing caveolin (ϩ) were immunoprecipitated with HSP56 IgG, resolved by SDS-PAGE, and immunoblotted concurrently with a mixture of IgGs for HSP56, Cyp40, caveolin, and CypA. Each nylon was exposed for 60 s. Representative data from five separate experiments are shown.
FIG. 5. The caveolin-chaperone complex is not a post-lysis artifact. L1210-JF cells not expressing caveolin were metabolically labeled with [ 35 S]methionine (0.5 mCi/ml) for 24 h. A 10-fold excess of radiolabeled cells not expressing caveolin was mixed with unlabeled cells that were expressing caveolin. The cells were then lysed and cytosol isolated exactly as described in the legend to Fig. 1. Caveolin IgG was added to the cytosol to precipitate caveolin and associated proteins. The precipitate was resolved by SDS-PAGE, the gel dried and exposed to film to detect radiolabeled proteins (lane 1). The same amount of cells expressing caveolin as used in the mixing experiment was radiolabeled and processed identically (lane 2). Representative data from four separate experiments are shown.
FIG. 6. Cyclosporin A and rapamycin disrupt the caveolinchaperone complex. NIH 3T3 cells were treated for 1 h, 37°C in DMEM plus 20 mM HEPES, pH 7.4, containing: buffer only (Control), 1 M cyclosporin A (CYSA), or 100 nM rapamycin (RAP). One set of cells were treated with 1 M cyclosporin A for 1 h at 37°C before they were washed and incubated an additional hour at 37°C in buffer only (Washout). After incubation, cytosols were prepared in Buffer A and incubated with caveolin IgG. The total amount of each precipitate was resolved by SDS-PAGE, transferred to PVDF, and immunoblotted concurrently with: heat-shock protein 56 (HSP56), cyclophilin 40 (CYP40), caveolin (Caveolin), and cyclophilin A (CYPA). All exposures were for 30 s. Representative data from five separate experiments are shown. the same number of cell equivalents (relative abundance in each fraction). Half of the precipitated material was resolved by SDS-PAGE and immunoblotted for caveolin and the small GTP-binding protein Ras. Fig. 7, anti-caveolin, equal protein, shows that caveolin was most enriched in caveolae. However, a significant amount of caveolin was detected in BPM (Fig. 7, anti-caveolin, cell equivalents) which represents incomplete extraction of caveolae from the plasma membranes. We have already documented that caveolae are not contaminated with bulk plasma membrane markers (11, 14, 28 -32). When standard curves and experiments were conducted in the near-linear range of the immunoblots, it was estimated that approximately 10 -15% of the caveolin found in the post-nuclear supernatant was in the cytosol (defined as the supernatant of a 350,000 ϫ g, 1 h centrifugation). Because caveolin is associated with caveolae membranes it was important to ensure that entire membrane fragments were not being precipitated. Several researchers, using different methods, have shown Ras to be highly enriched in caveolae (30,43). Ras was not co-immunoprecipitated (Fig. 7, Anti-Ras) with caveolin which suggests that entire caveolae membranes were not precipitated. In addition, the irrelevant antibody, 2001, did not interact with caveolin or cholesterol (Fig. 7, CM-2001).
Lipids were extracted from the other half of the precipitated material and resolved by thin-layer chromatography (TLC). Sterol was only precipitated when caveolin was precipitated (Fig. 7, compare all lanes to BPM). The total amount of labeled sterol in the cytosol, intracellular membranes, and plasma membranes was approximately equal to that in the post-nuclear supernatant (starting material). Furthermore, caveolin IgG also co-precipitated similar amounts of cholesterol from L1210-JF cells expressing caveolin but not from 1210-JF cells not expressing caveolin (data not shown). Because caveolin contains a 33-amino acid hydrophobic region it was possible that lipids were nonspecifically associating with the protein. However, this is not likely because we did not detect any other lipids on the TLC plates (data not shown). The lack of other lipids also supports the supposition that entire membrane fragments were not precipitated.
Cholesterol Transport to Caveolae-We have presented evidence to show the existence of a cytosolic caveolin-chaperone complex and that the caveolin-chaperone complex can bind cholesterol in cells. If the cytosolic caveolin-chaperone complex transports cholesterol to caveolae then cyclosporin A and rapamycin should inhibit the accumulation of cholesterol in caveolae. To test this NIH 3T3 cells were grown in [ 3 H]acetate for 16 h, 37°C, to label total cellular cholesterol pools. In the continuous presence of labeled acetate, cyclosporin A (1 M) or rapamycin (100 nM) were incubated with the cells for 1 h, 37°C. The cells were then subfractionated and the total amount of cholesterol-specific disintegrations per minute in each fraction determined by thin-layer chromatography and scintillation counting. We found that both cyclosporin A and rapamycin caused a 100-fold decrease in the amount of cholesterol-disintegrations per minute associated with caveolae (Table I). Removing the drugs and incubating the cells for 1 h in normal medium reversed the effect (Table I). These reagents did not affect the purity or yield of caveolae membranes (data not shown). Importantly, neither drug affected the total amount of radiolabeled cholesterol in the cells (Table I, compare control-PNS to rapamycin-PNS and cyclosporin A-PNS). In addition, small (2-3-fold) but reproducible increases in intracellular membrane cholesterol levels were observed.
We next used a temperature shift assay (14,44) to look at the flux of cholesterol from the ER to caveolae membranes. Cholesterol can be synthesized at 14°C in the ER but very little of the sterol can traffic through the cell (14,44). Shifting the temperature to 37°C permits a bolus of labeled cholesterol to move through intracellular trafficking routes. We preincubated NIH 3T3 cells in the presence of 1 M cyclosporin A or 100 nM rapamycin for 1 h at 37°C. Without removing the drugs, the cells were chilled to 14°C and incubated with [ 3 H]acetate for 1 h before adding excess unlabeled acetate and shifting the temperature to 37°C for various times. The cells were then subfractionated and the amount of cholesterol in each fraction determined by TLC and scintillation counting. Initially, in control cells, all of the labeled cholesterol was in intracellular membranes (Fig. 8A, f). As we showed previously (14), the labeled cholesterol translocates to caveolae within 10 -20 min (Fig. 8A, E) and by 30 -40 min is dispersed within the plasma membrane (Fig. 8A, Ⅺ). Pretreatment of cells with either cyclosporin A or rapamycin inhibited translocation of newly synthesized cholesterol to caveolae ( Fig. 8B and C, E) although a small amount of label was detected in bulk plasma membranes ( Fig. 8B and C, Ⅺ). Importantly, the reagents and temperature did not affect overall cholesterol synthesis because the amount of labeled cholesterol associated with intracellular membranes was comparable to control cells (compare f in B and c to f in A). Furthermore, removal of the drugs for 1 h generated a cholesterol trafficking pattern indistinguishable from that of control cells (Fig. 8D, rapamycin data not shown). The effects of cyclosporin A and rapamycin were reversible even when protein synthesis was inhibited (cycloheximide, 30 g/ml) which indicates that all of the necessary machinery for cholesterol trafficking was constitutively present (data not shown). L1210-JF cells not expressing caveolin did not rapidly transport cholesterol to caveolae (14), furthermore, cholesterol trafficking in these cells were unaffected by cyclosporin A or rapamycin (data not shown). However, L1210-JF cells expressing caveolin did rapidly transport cholesterol to caveolae (14), and this transport was disrupted by cyclosporin A and rapamycin (data not shown).

DISCUSSION
We have established the existence of a cytosolic caveolinchaperone complex and that this complex can bind to and transport sterol to caveolae in intact cells. Overall the data suggest that a caveolin-chaperone complex transports newly synthesized cholesterol through the cytosol to caveolae membranes by a vesicle independent mechanism. This putative function would not be expected of caveolin because caveolin is a membrane-associated protein, albeit with an unusual hairpin structure (15). A caveolin-chaperone complex may provide a membrane independent mechanism to transport sterols through the cytoplasm (41).
Previous work by many laboratories have shown that newly synthesized cholesterol by-passes the Golgi and trafficks directly to the plasma membrane in an energy requiring process that is completed in 10 -20 min (39,40,44). We recently demonstrated that newly synthesized cholesterol initially appears in cell surface caveolae (14). Surprisingly, cholesterol tran- buffer only (control), 1 M cyclosporin A (CysA), or 100 nM rapamycin (rapamycin). Two sets of cells were treated with cyclosporin A or rapamycin for 1 h at 37°C before they were washed and incubated an additional hour at 37°C in buffer containing radiolabeled acetate but not the drugs (CysA washout and rapamycin washout). The cells were washed with Buffer A and processed to isolate caveolae. After organic extraction, lipids were resolved by thin-layer chromatography. Spots corresponding to cholesterol were scraped and quantified by liquid scintillation counting. Mean Ϯ S.E., n ϭ 7. Values are disintegrations per minute/mg of protein. siently associates with caveolae membranes then flows into the bulk plasma membrane (14). We have hypothesized that vectorial movement of cholesterol through caveolae may contribute to the overall organization and function of this membrane domain as well as to total cellular cholesterol homeostasis. Multiple lines of research support this hypothesis. Fielding and Fielding (45) have shown that low density lipoprotein-associated free cholesterol trafficks to caveolae after release from the lipoprotein. Babitt et al. (28) have shown that the high density lipoprotein receptor, SR-B1, is localized to caveolae membranes. Finally, the invaginated morphology of caveolae is dependent on cholesterol (10,14,25).
Several studies have implicated caveolin in intracellular cholesterol trafficking. We showed that oxidation of caveolae cholesterol caused caveolin to leave plasma membrane caveolae and accumulate in the lumen of the ER (8,24). Removal of cholesterol oxidase allowed caveolin to traffic to the Golgi and return to caveolae in a microtubule-dependent process (24). Coincident with the return of caveolin was the accumulation of non-oxidized cholesterol in caveolae membranes (8,24). Surprisingly, the total number of invaginated caveolae at the cell surface did not change nor could we detect caveolin in any other kind of vesicle (8). Trafficking through the Golgi was expected because the work of Simons and colleagues (15)(16)(17) have shown that newly synthesized caveolin is packaged in apically targeted Golgi vesicles. To date, the trafficking of caveolin to basolateral plasma membranes has not been addressed. In addition, work by Murata et al. (9) and Sargiacomo et al. (46) have shown that caveolin can specifically bind to cholesterol in vitro. Finally, recent work by Fielding et al. (26) show that the expression of caveolin is regulated by oxysterols.
We have presented evidence to demonstrate that cholesterol, caveolin, cyclophilin A, cyclophilin 40, and HSP56 form a cytosolic lipoprotein-chaperone complex. Furthermore, this complex appears to be a mechanism by which newly synthesized cholesterol traffics directly between the ER and caveolae. Additional work is required to determine if cholesterol can traffic from caveolae to the ER and if the other caveolin isoforms have a similar function.