Cell-specific targeting of caveolin-1 to caveolae, secretory vesicles, cytoplasm or mitochondria

Caveolin-1 was originally identified as a novel tyrosine kinase substrate in Rous sarcoma transformed cells (Glenney, 1989). Immunogold cytochemistry localized the protein to the distinctive striated coat structure that decorates the inner membrane surface of fibroblast caveolae (Rothberg et al., 1992) and to elements of the Golgi apparatus (Kurzchalia et al., 1992). Studies carried out concurrently implicated caveolin-1 in the sorting of molecules during vesicular trafficking to the apical surface of polarized epithelial cells (Kurzchalia et al., 1992). Subsequently, several laboratories used caveolin-1 as an integral protein marker to prepare cell fractions enriched in caveolae (reviewed in Anderson, 1998). Several methods of purification yield vesicles that are approximately the size of caveolae (Smart et al., 1995; Westermann et al., 1999), sometimes with an apparent striated coat (Chang et al., 1994). These and other studies suggest that caveolin-1 plays a role in the traffic of caveolae and caveolae-related membranes within the cell (Anderson, 1993; Anderson, 1998). Caveolin-2 and caveolin-3 are two homologous proteins that may have similar functions to caveolin-1 (Smart et al., 1999).

Studies on the function of caveolin-1 follow two main themes. One idea is that caveolin-1 is a scaffolding protein involved in organizing the activity of multiple signaling molecules in caveolae (Okamoto et al., 1998). The other is based on the cholesterol-(Murata et al., 1995) and fatty acid-(Trigatti et al., 1999) binding properties of caveolin-1 and the possibility that it mediates the intracellular transport of lipids such as cholesterol (Smart et al., 1996). The exact function of a protein is often reflected in how the molecule is used by various tissue cells. If, for example, caveolin-1 were a scaffolding protein, then the molecule should be in caveolae-related structures whenever it is expressed. In the current study, we tested the hypothesis that all cells expressing caveolin-1 will target the molecule to caveolae by determining with immunocytochemistry if caveolin-1 in different tissue cells is always in this membrane domain. Unexpectedly, we found examples of cells that preferentially target caveolin-1 to either the cytoplasm, mitochondria or elements of the secretory pathway but had little caveolin-1 in invaginated caveolae. The behavior of caveolin-1 in these cells suggests the existence of a novel pathway of intracellular and extracellular molecular trafficking that may be specialized for delivering lipids to multiple cellular compartments.

Cyclophilin A pAb was purchased from ABR (Golden, CO, USA). ApoA 1 pAb was a gift from Dr Helen Hobbs (University of Texas Southwestern Medical Center, USA). Purified human HDL was prepared by standard methods. OptiPrep was purchased from Accurate Chemical & Scientific Corporation (Westbury, NY, USA). Caveolin-1 pAb and mAb were either raised in our laboratory or obtained from Transduction Laboratories (Lexington, KY, USA). Caveolin-2 mAb and caveolin-3 pAb were obtained from Transduction Laboratories (Lexington, KY, USA). FITC-goat anti-rabbit IgG was from Zymed Laboratories (San Francisco, CA, USA). Protein A gold was from Dr J. W. Slot (Utrecht University, The Netherlands). BSA, proline, leupeptin, soybean trypsin inhibitor, pepstatin A, Dulbecco’s modified Eagle’s Medium (DMEM), benzamidine, adenine-free base, human apo-transferrin, aprotinin, L-ascorbic acid, bovine pancreatic insulin, ethylenediaminetetraacetic acid (EDTA), ethyleneglycoltetraacetic acid (EGTA), hydrocortisone, ATP, secretin, dexamethasone, cholecystokinin, mitomycin C, sodium chloride, sodium deoxycholate, sodium fluoride, sodium vanadate and 3,3′,5-triiodo-L-thyronine sodium salt were from Sigma (St Louis, MO, USA). A23187 was from Biomol (Plymouth Meeting, PA, USA). Bio-Rad Protein Assay was from Bio-Rad Laboratories (Hercules, CA, USA). F-12K medium was from GibcoBRL (Grand Island, NY, USA). Fetal bovine serum (FBS) and fetal calf serum (FCS) were from HyClone (Logan, UT, USA). C57Bl/6J mice were from Jackson Laboratory (Bar Harbor, MI, USA). Sprague Dawley rats were from Harlan Sprague Dawley (Indianapolis, IN, USA). Bovine type I collagen (Vitrogen-100) was purchased from Celltrix Laboratories (Palo Alto, CA, USA) and rat tail type I collagen was obtained from Upstate Biotechnology (Lake Placid, NY, USA). DMEM/F-12 medium, recombinant human epidermal growth factor, and penicillin/streptomycin were obtained from Life Technologies, Inc. (Rockville, MD, USA). Recombinant Cholera toxin B subunit was obtained from Calbiochem-Novabiochem corporation (Cambridge, MA, USA). Borosilicate cloning cylinders (8 mm inner diameter) were purchased form BellCo Glass, Inc. (Vineland, NJ, USA). Transwell™ polycarbonate cell culture inserts (4.0 μm pore size) were obtained from Corning Costar Corp. (Cambridge, MA, USA). Hybond N+ blotting membrane, Rediprime II random priming label and ECL immunoblotting systems were purchased from Amersham Pharmacia Biotech, Inc. (Piscataway, NJ, USA). [γ32P]-dCTP was obtained from NEN Life Science Products, Inc. (Boston, MA, USA). Immobilon-P polyvinylidene difluoride (PVDF) membrane was purchased from Millipore Corp. (Bedford, MA, USA).

Buffer A: 20 mM Tricine, pH 7.8, 250 mM sucrose, 1 mM EDTA.

Buffer B: 20 mM Tris-HCl, pH 7.6, 130 mM NaCl, 0.2% Tween-

20.

Buffer C: 100 mM sodium phosphate, pH 7.4, 0.15 M NaCl, 4 mM KCl, 2 mM MgCl_2, and 0.02% (wt/vol) sodium azide.

Buffer D: Buffer C containing 1% BSA, 0.05% Tween 20, 0.05% Triton X-100.

Buffer E: Buffer C containing 1% BSA, 0.01% Tween 20, 0.01% Triton X-100.

Buffer F: 10 mM Tris-HCl pH 7.5, 250 mM sucrose, 0.1 mM EDTA.

Normal human fibroblasts (Goldstein et al., 1983) grown to confluence in 150 mm plates were cultured in DMEM supplemented with penicillin and 10% FBS. 20 hours before each experiment, the medium was changed to MEM plus 200 μg/ml of BSA. The pituitary cell line GH3 was cultured (100 mm plates) in F-12K medium supplemented with 2.5% FBS and 15% horse serum (HS) for 4 days before each experiment.

Human primary keratinocytes were harvested from healthy adult skin obtained as surgical discard tissue. Subcutaneous fat and deep dermis were removed, and the remaining tissue was incubated in 0.25% trypsin in PBS at 25°C. After 16 hours, the epidermis was separated from the dermis with forceps, and the keratinocytes were scraped into DMEM. The keratinocyte suspension was added to fresh DMEM supplemented with 5% FCS and 0.1% penicillin/streptomycin and a sample of keratinocyte suspension was then plated onto tissue culture dishes coated with 1 mg/ml type I collagen. Under these culture conditions, keratinocytes proliferate, migrate, differentiate and cornify (Pentland and Needleman, 1986). Cultures of dermal fibroblasts, obtained from surgical discard tissue, were outgrown from dermal explant cultures and maintained in DMEM supplemented with 10% FCS and 0.1% penicillin/streptomycin. Cells used for the cultured skin equivalents were passaged 5-10 times. Fibroblast feeder layers were prepared using mouse 3T3 cells treated with mitomycin C (10 μg/ml) for 2 hours at 37°C. Normal epidermal keratinocytes were plated onto feeder layers in FAD medium (DMEM:Hams F12, 3:1) supplemented with 0.1% penicillin/streptomycin, 5% FCS, 5 μg/ml insulin, 20 ng/ml recombinant human epidermal growth factor, 5 μg/ml transferrin, 10⁻¹⁰ M cholera toxin, 180 μM adenine, 0.4 μg/ml hydrocortisone and 0.02 nM triiodothyronine. Cultured skin equivalents were generated using a technique modified from Stark et al. (Stark et al., 1999). Briefly, normal epidermal keratinocytes (1×10⁶/cm², passages 2-5) were seeded within 8 mm borosilicate cloning cylinders onto precontracted

fibroblast-populated collagen gels (2×10⁴ cells/ml, rat tail type I collagen) and placed into Transwell™ polycarbonate cell culture inserts (4.0 μm pore size). Following keratinocyte attachment (approx. 4 hours), the cloning cylinders were removed and cultures were immersed in FAD medium containing 50 μg/ml L-ascorbic acid. Following a 6 day culture period to allow the keratinocytes to reach confluence, the cultured skin equivalents were raised to the air-liquid interface to stimulate keratinocyte differentiation and cornification. FAD medium with additives was replaced every 2 days until cultured skin equivalents were harvested.

Caveolae were isolated by the method of Smart et al. (Smart et al., 1995). Briefly, confluent normal human fibroblasts were collected in hypotonic buffer and dounced 20 times on ice. Plasma membrane (PM) was isolated on a 30% Percoll gradient from the post-nuclear supernatant (PNS) and then sonicated. The sonicated sample was mixed with OptiPrep (final OptiPrep concentration, 23%) in a TH641 tube. A linear 20%-10% OptiPrep gradient was overlaid on the sample and centrifuged at 52,000 g for 90 minutes at 4°C. The bottom 1 ml was designated non-caveolae membranes (NCM). The top 5 ml were collected and mixed with 4 ml of 50% OptiPrep in a second TH641 tube. 2 ml of 5% OptiPrep were overlaid and the sample centrifuged at 52,000 g for 90 minutes at 4°C. The gradient was fractionated in 0.7 ml fractions. Fraction 3 was designated caveolae membranes (CM).

Equal amounts of the indicated mouse tissue (40 mg wet mass) was sonicated in 200 μl 2× SDS-sample buffer (100 mM Tris-HCL, pH 6.8, 20% glycerol, 4% SDS, 4% 2-mercaptoethanol). The 50 μl sample was diluted with 950 μl of water before adding 100 μl of 72% TCA to precipitate the protein. The precipitate was redissolved in 1000 μl 2× SDS-sample buffer. Equal fractions (30 μl, approx. 300 μg wet mass tissue) were heated at 95°C for 5 minutes in SDS-sample buffer (Laemmli, 1970) before being separated by electrophoresis at 25 mA per gel. The proteins were transferred to PVDF membranes. After blocking with buffer B containing 5% nonfat dry milk, the membranes were incubated with the first antibody followed by the second antibody conjugated with HRP in buffer B containing 1% nonfat dry milk. Bound antibody was detected using an ECL detection system. To isolate protein from keratinocyte cultures, cell layers were washed with PBS and treated for 10 minutes at 4°C with cell lysis buffer (1.0% NP-40, 20 mM Tris, pH 7.5, 2.0 mM sodium vanadate, 1.0 mM NaF, 100 mM NaCl, 5.0 μg/ml sodium deoxycholate, 2.0 mM EDTA, 2.0 mM EGTA, and 25 μg/ml each of leupeptin, aprotinin and pepstatin). Cell lysates were then dounce-homogenized and the lysate cleared by centrifugation (14,000 g) at 4°C. Proteins were separated by gel electrophoresis and immunoblotted using the same procedure as for the mouse tissue.

Total RNA was isolated by phenol-chloroform extraction (Chomczynski and Sacchi, 1987). RNA (10 μg) was denatured and resolved by electrophoresis through a 1% formaldehyde-agarose gel, transferred overnight to Hybond N+ (Amersham Pharmacia Biotech, Inc., Piscataway, NJ, USA), and hybridized with a radiolabeled caveolin-1 cDNA probe. The cDNA probe was labeled by random priming with [α32P]-dCTP. Following hybridization, the membranes were washed and visualized using the Typhoon 8600 phosphorimager system (Molecular Dynamics, Sunnyvale, CA, USA).

Adult C57Bl/6J mice or adult Sprague Dawley rats were fixed by perfusing euthanized animals via the left cardiac ventricle with 60 ml of 3% (w/v) paraformaldehyde in buffer C containing 3 mM trinitrophenol for 5 minutes. Tissues were removed and fixed further in a mixture of 60% (v/v) methanol, 10% (v/v) glacial acetic acid, 30% (v/v) inhibisol (1,1,1-trichloroethane) for 24 hours. Lung was perfusion-fixed via the right cardiac ventricle. Tissues were then embedded in paraffin. Paraffin sections (5 μm thick) were dewaxed in three changes of xylene (10 minutes each), rehydrated into buffer C. The sections were washed with 50 mM NH4Cl in buffer C for 30 minutes, rinsed in buffer C followed by buffer C containing 1% BSA and 1.5% normal goat serum for 60 minutes at room temperature. This was followed by an overnight incubation in the presence of either pAb (10 μg/ml) or mAb (30 μg/ml) caveolin-1 IgG, pAb (10 μg/ml) caveolin-3 IgG, non-immune rabbit IgG (10 μg/ml) or non-immune mouse IgG (30 μg/ml). Primary antibodies were localized by incubating sections for 2 hours in 15 μg/ml affinity-purified, goat anti-rabbit IgG or goat anti-mouse IgG both conjugated to FITC in buffer C containing 0.5% BSA. The sections were rinsed in buffer C containing 0.2% BSA, mounted on slides and photographed with a Zeiss Photomicroscope III.

Normal human foreskin was fixed by immersion in 3% (w/v) paraformaldehyde/0.1% glutaraldehyde in buffer C. Following fixation and a brief rinse with buffer C, the sample was embedded in paraffin and sectioned. For immunolabeling, deparaffinized and rehydrated tissue sections were rinsed with buffer D for 30 minutes at room temperature followed by incubation with pAb (10 μg/ml) caveolin-1 IgG, or non-immune rabbit IgG (10 μg/ml) in buffer D for 18 hours at 4°C. The sections were warmed to 25°C, and processed for streptavidin/biotin immunogold-silver labeling (Roth et al., 1992).

Euthanized mice were perfused with 60 ml of 3% (w/v) paraformaldehyde-0.1% glutaraldehyde in buffer C for 5 minutes. Animal tissues were removed, cut into small pieces and fixed for an additional 30 minutes in the same fixative. Some samples were embedded in Lowicryl K4M at low temperature as previously described (Roth, 1989). Ultrathin K4M sections were incubated for 17 hours in the presence of 20 μg/ml of either pAb anti-caveolin-1 or non-immune rabbit IgG in buffer E. The IgG was localized by incubating each set of sections for 1 hour in protein A-gold (10 nm diameter) diluted 1:65 in buffer E. Electron micrographs were taken with a JEOL 1200 electron microscope operating at 80 kV. To localize caveolin-1 in isolated liver mitochondria by immunogold labeling, mitochondria were fixed, immersed in 2.3 M sucrose containing 15% polyvinylpyrrolidone (10 kDa) and processed for immunogold localization in ultrathin cryosections using pAb Caveolin-1 IgG (Tokuyasu, 1980).

Cytosol of human fibroblasts was fractionated on CsCl gradients by standard methods (Kongshaug et al., 1989). A step gradient was prepared in a Ti60 rotor tube consisting of 2 ml 1.35 g/ml CsCl at the bottom followed by 9 ml 1.21 g/ml CsCl and 7 ml 1.063 g/ml CsCl. The gradient was overlaid with 7 ml cytosol of human fibroblasts and centrifuged at 214,000 g for 23 hours before being separated into 14 or 24 fractions. From each fraction, 1 ml was precipitated by TCA precipitation and the proteins separated by gel electrophoresis.

Mitochondria were isolated using the method of Weinbach (Weinbach, 1961). 1 g rat liver was homogenized with a dounce homogenizer (30 strokes) in 10 ml buffer F. The homogenate was centrifuged at 610 g for 10 minutes. The supernatant was collected and centrifuged at 8600 g for 10 minutes. The mitochondria (pellet) was washed with buffer F three times and processed either for immunoblotting, immunoprecipitation or immunogold labeling of ultrathin frozen sections.

Immunoprecipitation of caveolin-1 from mitochondria fractions was carried out as previously described (Liu et al., 1996).

We used immunoblotting to compare the caveolin-1 expression level in various mouse tissues (Fig. 1A). Equal weights (∼300 μg) of each tissue was separated by gel electrophoresis and immunoblotted with caveolin-1 pAb (caveolin-1). On a tissue weight basis, the lung had the highest expression level of all the tissues, while the lowest amount of caveolin-1 was found in the liver. Samples of tissue rich in smooth muscle (e.g., intestine and uterus) and adipocytes expressed high levels of caveolin-1, consistent with the large numbers of invaginated caveolae present in these cells (Carpentier et al., 1977). Interestingly, however, we found high expression in tissues such as adrenal, thymus and kidney, which are composed of cells not usually thought to have large numbers of caveolae. Heart also expressed high levels of this protein and skeletal muscle appeared to contain as much caveolin-1 as the adrenal.

The tissue distribution of caveolin-2 in mice (Fig. 1A) was similar to that of caveolin-1, although the staining intensity of the immunoblot for each tissue was generally lower. Caveolin-3 was most prominent in skeletal and heart tissue, although small intestine and lung also contained significant amounts of this isoform. We were unable to detect any caveolin-3 in uterus, even though this tissue is rich in smooth muscle cells.

We explored further the expression of caveolin-1 in skeletal and heart tissue (Fig. 1B,C) because other laboratories have reported that caveolin-1 is not expressed in muscle (Parton et al., 1997). The caveolin-1 pAb did not react with caveolin-3 because the caveolin-3 pAb and the caveolin-1 pAb recognized protein bands with distinctly different molecular masses that corresponded to the size of the respective proteins (compare left and right panels, Fig. 1B). Antibodies that recognize only α-caveolin-1 (mAb 2) did not react with skeletal muscle caveolin-1 (Fig. 1C) while those that recognized the α and p isoforms (pAb 1-3, mAb 1) demonstrated prominent staining, suggesting that only p-caveolin-1 is expressed in skeletal muscle cells.

Cultured fibroblasts, endothelial cells and polarized epithelial cells are commonly used to study caveolin-1 targeting to invaginated caveolae. Previous studies suggest that blood vessel endothelium (Esser et al., 1998; Feng et al., 1999) and gizzard smooth muscle (Chang et al., 1994) are tissue cells where caveolin-1 is localized in invaginated caveolae in situ. We also found that the caveolin-1 in mouse endothelial and alveolar type I cells was concentrated in typical, flask-shaped caveolae (data not shown). We conclude that in situ a variety of tissue cells have flask-shaped caveolae rich in caveolin-1 and that the antibodies used in the current study recognize caveolin-1 when it is in this location.

Previously we reported that most of the caveolin-1 in pancreatic acinar cells is located in the lumen of secretory vesicles (Liu et al., 1999). In these cells, caveolin-1 is cosecreted with amylase through a regulated pathway. The secreted caveolin-1 is in a lipid complex that has the properties of a high-density lipoprotein (HDL) particle. We examined other exocrine secretory cells to determine if they also target caveolin-1 to the secretory pathway (Fig. 2). Caveolin-1 was found in secretory vesicles of mouse serous secreting (Fig. 2A), but not mucous secreting (Fig. 2B), cells of the salivary gland. In addition, caveolin-1 was targeted to secretory vesicles of chief cells (Fig. 2C) and the apical cytoplasm of mammary epithelial cells (compare Fig. D with E).

HDL purified from plasma does not contain detectable caveolin-1 (Liu et al., 1999), which may mean that secretion of caveolin-1 is restricted to exocrine cells. Indeed, we found very little caveolin-1 protein in hepatocytes, the major source of lipoproteins in the body (Fig. 1). A survey of various endocrine tissues, however, revealed that anterior pituitary cells with the morphologic characteristics of somatotropes had caveolin-1 concentrated in secretory vesicles (Fig. 3A). By contrast, the secretory vesicles of neighboring pituitary endocrine cells with the morphology of corticotropes and mammatropes (Fig. 3B,C, respectively) did not contain caveolin-1. We confirmed that pituitary cells are able to secret caveolin-1 (Fig. 3D) by culturing the pituitary cell line GH3 overnight and immunoblotting either the medium (lanes 7-12) or the cells (lane 1-6) with caveolin-1 pAb. The culture medium was positive for caveolin-1, indicating that the endogenous caveolin-1 synthesized by these cells was secreted into the medium. Unlike pancreatic acinar cells (Liu et al., 1999), caveolin-1 secretion was not regulated by any of the agents we tested (Fig. 3D, lanes 8-12) and little caveolin-1 was detected in the cells (lanes 1-6). We examined the subcellular distribution of caveolin-1 in these cells by immunoblotting either the postnuclear supernatant (lane 1), cytosol (lane 2), plasma membrane (lane 3), non-caveolae membrane (lane 4) or caveolae membrane (lane 5) with caveolin-1 pAb (Fig. 3E). The only place caveolin-1 was detected was in the cytosol fraction (lane 2). An EM examination of these cells showed that they contained few secretory granules (data not shown). These results suggest that most of the newly synthesized caveolin-1 in GH3 cells is shunted into an unregulated secretory pathway.

A portion of the caveolin-1 expressed in cultured fibroblasts appears to be soluble in the cytoplasm of the cell (Uittenbogaard et al., 1998). We searched for tissue cells that might selectively target caveolin-1 to the cytosol (Fig. 4). Initially, we focused on skeletal muscle because caveolin-3 appears to substitute for caveolin-1 in invaginated caveolae (Parton et al., 1997). Immunofluorescence staining of mouse skeletal muscle with a pAb that recognizes both α and β-caveolin-1 showed a cytoplasmic distribution of the protein with a marked striated pattern (Fig. 4A). The caveolin-3 pAb, by contrast, exclusively stained the cell surface (Fig. 4B). Immunogold localization of p-caveolin-1 showed gold particles scattered in the cytoplasm of the cell with accumulations along the Z-line (arrowheads, Fig. 4C). No immunogold staining was seen with an mAb that only recognizes α-caveolin-1 (Fig. 4D). Thus, the striated pattern seen by immunofluorescence appears to be due to the accumulation of p-caveolin-1 at Z lines.

The skin keratinocyte is another tissue cell that targets caveolin-1 to the cytoplasm. Light microscopic, immunogold staining of human skin (Fig. 4E) showed that keratinocytes stained positively with caveolin-1 pAb. EM immunogold labeling of cells in the stratum spinosum (Fig. 4F) with caveolin-1 pAb showed that gold particles were primarily associated with intermediate filaments that appeared to emanate from desmosomes. The cytoplasm of fully differentiated, cornified keratinocytes was densely labeled with caveolin-1 pAb gold particles (Fig. 4G). No gold labeling of the cornified keratinocyte was seen with a non-immune IgG (Fig. 4H). The same result was obtained with five separate caveolin-1 antibodies that are directed against epitopes found in both α and p caveolin-1 (data not shown). Immunoblots indicated that mouse skin has as much caveolin-1 on a tissue mass basis as fat (Fig. 5A).

Undifferentiated human keratinocytes growing on a collagen matrix in culture can be induced to differentiate (Pentland and Needleman, 1986). We used this system to determine if caveolin-1 expression is regulated during keratinocyte differentiation (Fig. 5B-E). Undifferentiated cells did not contain any detectable caveolin-1 mRNA (Fig. 5B, 0 hours). Caveolin-1 mRNA was detected after 4 hours of induction, and continued to increase for the next 44 hours. Likewise, we saw a progressive increase in the amount of caveolin-1 protein during differentiation, which matched the increase of the two differentiation marker proteins, involucrin and fillagrin (Fig. 5C). Compared to day 1 cells, we estimated from gel scans that there was an 8.3-fold increase in the amount of caveolin-1 expression in day 4 cells. Immunofluorescence staining of day 5 skin equivalent tissue showed abundant caveolin-1 staining of keratinocytes (between arrows, Fig. 5D) with little staining of fibroblasts in the underlying collagen matrix (asterisk, Fig. 5D). No staining was observed in this cultured skin model when we used a non-immune pAb (Fig. 5E). The dramatic upregulation of caveolin-1 gene expression during keratinocyte differentiation confirms our immunohistochemistry findings that caveolin-1 is strongly expressed in suprabasal epidermal cells.

In fibroblasts, soluble caveolin-1 appears to be in a complex with three heat shock proteins plus cholesterol (Uittenbogaard et al., 1998). Since caveolin-1 secreted by pancreatic acinar cells behaves like an HDL particle (Liu et al., 1999), we reasoned that the cytoplasmic caveolin-1 might also be in a lipoprotein complex (Fig. 6). We used samples of cytoplasm from normal human fibroblasts for this analysis. We found that 1-2% of the total caveolin-1 in these cells was soluble. The cytosol was loaded on the top of a CsCl gradient and centrifuged for 23 hours. 14 fractions were collected and immunoblotted with a caveolin-1 pAb (Fig. 6A, Caveolin-1). Most of the caveolin-1 was in fractions 8-12 and clearly separated from the bulk protein at the bottom of the gradient (Fig. 6B). A sample of plasma HDL was loaded on a companion gradient, fractionated and immunoblotted with an ApoA1 pAb (Fig. 6A, ApoA1). The HDL sample migrated on the gradient exactly in the same position as soluble cytoplasmic caveolin-1.

One of the heat shock proteins associated with cytoplasmic caveolin-1 is cyclophilin A, which is also a prominent protein in caveolae (Uittenbogaard et al., 1998). We ran a second sample of the cytosol, this time taking 24 fractions, and immunoblotted each fraction with either caveolin-1 (Fig. 6C, Cav-1) or cyclophilin A (Fig. 6C, Cyclo-A) pAb. Remarkably, all the cyclophilin A floated in the light fractions rich in caveolin-1 (fractions 9-22). Most of the cytosolic protein, by contrast, was in the bottom six fractions of the gradient (Fig. 6D). We conclude that cytosolic caveolin-1, like secreted caveolin-1, behaves as if it is associated with a lipoprotein particle that has the buoyant density of HDL.

Tissue immunoblots (Fig. 1) indicated that the tissue with the highest caveolin-1 expression level in mice was the lung. Much of the caveolin-1 may be synthesized by endothelial, smooth muscle and alveolar cells, but a routine immunofluorescence survey indicated that mouse terminal airway epithelial cells expressed high levels of caveolin-1 as well (data not shown). We used immunogold labeling of Lowikryl K4M embedded material to obtain high-resolution images of the caveolin-1 pAb labeling pattern in these cells (Fig. 7). The gold was concentrated in numerous large, membrane-bound structures that had the appearance of secretory vesicles (Fig. 7A). No labeling was evident at the cell surface or in other compartments of these cells. The same result was obtained with several different caveolin-1 mAbs and pAbs (data not shown). The location of these cells in the airway suggested they might be Clara cells. Previous comparative morphology studies have shown that mouse Clara cells contain numerous large mitochondria with attenuated cristae (Smith et al., 1979; Widdicombe and Pack, 1982). Indeed, large mitochondria with attenuated cristae were easily identified in sections of Epon-embedded Clara cells (Fig. 7D, arrowheads). Examination of the Lowikryl K4M-embedded Clara cells at high magnification showed that a double membrane was around the caveolin-1 positive structures and that an occasional crista could be seen protruding into the matrix (arrowheads, Fig. 7C) that was similar to the cristae seen in the Epon sections (arrowheads, Fig. 7D). The same Clara cell also typically contained secretory granules (Fig. 7B) that had few gold particles. These results suggest that in Clara cells the majority of the caveolin-1 is targeted to modified mitochondria, not to secretory granules.

Immunogold particles were found associated with mitochondria in many cell types labeled with caveolin-1 pAb but the number of gold particles was very low (data not shown). We confirmed that mitochondria contain caveolin-1 by isolating liver mitochondria and processing them for either immunoblotting (Fig. 7E), immunoprecipitation (Fig. 7F) or EM immunogold detection of caveolin-1 (Fig. 7G,H). Caveolin-1 was detected in immunoblots of mitochondria fractions (Fig. 7E, Cav-1), although the amount was low. The caveolin-1 pAb also stained a band of approx. 60 kDa in these fractions. Most likely this band corresponds to oligomeric caveolin-1 that is sometimes seen in immunoblots because immunoprecipitation of caveolin-1 from the mitochondria fraction with a pAb caveolin-1 followed by immunoblotting with a mAb caveolin-1 (Fig. 7F, compare lane 1 with 2) showed both α and p caveolin-1 (Cav-1) and the 60 kDa band. Immunogold labeling of these fractions (Fig. 7G) detected caveolin-1 pAb reactivity in the matrix of the mitochondria, just as was observed for Clara cell mitochondria (Fig. 7A). No staining was seen with a non-immune IgG (Fig. 7H). The same result was obtained with two different polyclonal antibodies directed against opposite ends of the molecule. Thus, low levels of caveolin-1 appear to be present in the mitochondria of many cell types but in Clara cells large amounts of the protein are targeted to modified mitochondria that presumably are carrying out special functions for this cell.

The distribution of caveolin-1 in the various tissues we have examined suggests an unprecedented behavior for a cellular protein. The 178 amino acid long sequence of caveolin-1 predicts that it is an integral membrane protein (Kurzchalia et al., 1992) with both the COOH- and NH₂-termini in the cytoplasm. As expected of an integral membrane protein, the caveolin-1 in caveolae is resistant to salt extraction (Rothberg et al., 1992). Yet we found that in several different cellular compartments caveolin-1 is a soluble protein. For caveolin-1 to be both a soluble and a membrane protein during its lifetime, mechanisms must exist for converting it from one form to the other and for moving it to specific locations in the cell.

The various antibodies we used in this study recognize in tissue blots a single species of protein with the same apparent molecular mass as fibroblast caveolin-1 (Fig. 1). This suggests that the immunogold and immunofluorescence staining is not due to a cross-reacting epitope on another protein or that the caveolin-1 in these cells has undergone a post-translational modification. There appear to be just two mRNAs for caveolin-1, one for the alpha and the other for the beta isoform (Ko et al., 1998), which is consistent with the known genomic structure of caveolin-1 (Engelman et al., 1998). Pancreatic acinar cells in culture transfected with a caveolin-1 cDNA are able to faithfully target the molecule to the secretory pathway (Liu et al., 1999). We conclude that the molecule being detected in these various compartments is authentic caveolin-1 and that it has not been grossly altered by post-translational modification.

We have identified four locations in the cell where caveolin-1 can reside; caveolae, cytoplasm, mitochondria and elements of the secretory pathway. In some cells, caveolin-1 is in multiple locations (Kurzchalia et al., 1992), or moves between compartments in response to specific stimuli (Smart et al., 1994). The caveolin-1 that is not in caveolae or Golgi apparatus membranes behaves as if it were soluble (Liu et al., 1999; Uittenbogaard et al., 1998), which would require that the hydrophobic portions of the molecule be sequestered away from the aqueous environment. We propose that the hydrophobic regions of soluble caveolin-1 are embedded in a lipid particle surrounded by a phospholipid shell that has the buoyant density of an HDL particle.

The presence of soluble caveolin-1 in a lipid complex implies that it has an important function in intracellular lipid transport. Previous work has shown that caveolin-1 can move between caveolae and the ER as a soluble protein in a complex with several heat shock proteins and cholesterol (Smart et al., 1994; Uittenbogaard et al., 1998). Our results indicate that this mobile complex is a lipid particle. Thus, the soluble caveolin-

1 that travels to ER, mitochondria and various cytosolic locations may be embedded in a lipid particle. These particles may also be able to utilize novel machinery to move across membrane barriers into the interior of the ER and mitochondria compartments. The lipoprotein form of caveolin-1 should be capable of delivering lipids to these intracellular compartments, just as plasma lipoproteins carry lipids between tissues. The targeting of the particles to specific compartments may depend on the binding of caveolin-1, or associated proteins, to receptors at these locations.

Caveolin-rich lipid particles may belong to a family of intracellular lipid particles that have essential functions in the delivery and storage of intracellular lipids. All cells contain lipid particles. For example, yeast cells contain a uniform-sized population of lipid particles coated with specific sets of proteins that are essential for cell viability (Athenstaedt et al., 1999; Leber et al., 1994). Most metazoan cells are populated with numerous large lipid droplets that function in lipid storage. Many of these droplets are coated with proteins of unknown function such as vimentin, perilipin and ADRP (Londos et al., 1999). Recently, MAP kinase and phospholipase A₂, two molecules involved in signal transduction, were localized to these coats (Yu et al., 1998). We suggest that caveolin-1 is a cholesterol binding apolipoprotein for a smaller size of lipid particle (Murata et al., 1995) that shuttles cholesterol between intracellular compartments, including lipid storage droplets. Interestingly, a mutant form of caveolin-3 has been identified that associates with a cholesterol-rich, non-endocytic compartment that resembles a lipid droplet (Roy et al., 1999).

In vitro studies clearly indicate that caveolin-1 is cotranslationally inserted into ER membranes (Dupree et al., 1993). Specific regions of the protein have been identified that control its passage through ER and Golgi compartments on its way to the cell surface, all the while remaining a membrane protein (Machleidt et al., 2000). Therefore, caveolae are the most likely site where caveolin-1 leaves the membrane and enters the cytoplasm. While the mechanism of membrane release is not known, it may be a reversible process that involves the release and reincorporation of the lipid particle into the caveolae membranes (Uittenbogaard et al., 1998). We postulate that all soluble caveolin-1, regardless of its eventual destination, originates from caveolae membranes.

The behavior of caveolin-1 blurs the distinction between a molecule that operates inside the cell and one that functions outside. In professional secretory cells such as the pancreatic acinar cell, virtually all the caveolin-1 is targeted to the secretory pathway. Detailed morphological and biochemical studies have shown that in fibroblasts caveolin-1 constitutively cycles between caveolae and the Golgi apparatus, spending a portion of its time as a luminal protein of the ER (Conrad et al., 1995; Smart et al., 1994). Undoubtedly some of the cycling caveolin-1 in these cells gets diverted into secretory vesicles but most is reincorporated into membranes at the Golgi apparatus and moves back to the cell surface. The ability of caveolin-1 to travel this route in most cells raises the possibility that in doing so it ferries cytosolic molecules to the extracellular space. Soluble caveolin-1 probably interacts with multiple proteins in addition to the ones that have been identified and could potentially deliver them to the secretory pathway. For example, cyclophilin A is secreted both in culture (Sherry et al., 1992) and in tissues (Billich et al., 1997). HSP90-α may be another protein that is piggybacked into the secretory pathway by soluble caveolin-1 (Liao et al., 2000).

The abundance of caveolin-1 in the cytoplasm of keratinocytes, although unexpected, is consistent with a role for this molecule in lipid traffic. The skin is a major cholesterol producing organ (Spady and Dietschy, 1983). This sterol, together with ceramide and free fatty acids, is a critical lipid component of the extracellular water permeability barrier that is essential for survival in a desiccating environment (Downing, 1992). The permeability barrier is assembled extracellularly as keratinocytes differentiate from basal cells into corneocytes (Downing, 1992), and inhibition of cholesterol synthesis interferes with its assembly (Feingold et al., 1990). Caveolin-1 may play a critical role in the delivery of cholesterol to the extracellular site of barrier assembly.

A systematic immunocytochemical examination of different tissues using various antibodies to caveolin-1 has uncovered a curious and unexpected complexity to the behavior of this molecule. According to its localization in these tissue cells, caveolin-1 can either be a membrane or a soluble protein. The soluble form appears to be embedded in a lipid particle that may have critical functions in intracellular lipid transport. These functions may include the delivery of lipids to caveolae, lipid droplets, mitochondria, ER and other cellular compartments. The exact mechanism of caveolin-1 movement between soluble and membrane states remains to be determined.

We would like to thank Meifang Zhu and George Lawton for valuable technical assistance and Sue Knight for administrative assistance. This work was supported by grants from the National Institutes of Health, HL 20948, GM 52016, AR02153 and the Perot Family Foundation.