Baculovirus-based Expression of Mammalian Caveolin in Sf21 Insect Cells A MODEL SYSTEM FOR THE BIOCHEMICAL AND MORPHOLOGICAL STUDY OF CAVEOLAE BIOGENESIS

Caveolae were originally defined morphologically as 50–100 nm noncoated vesicular organelles located at or near the plasma membrane. Caveolin, a vesicular inte- gral membrane protein of 21 kDa, is a principal protein component of caveolae membranes in vivo . Caveolin in- teracts with itself to form high molecular mass oligomers, suggesting that it might play a structural role in the formation of caveolae membranes. However, it remains controversial whether recombinant expression of caveolin is necessary or sufficient to generate caveolae membranes in vivo . To directly address this issue, we have taken a different experimental approach by ex-ploiting a heterologous expression system. Here, we have recombinantly expressed mammalian caveolin in Sf21 insect cells using baculovirus-based vectors. Two isoforms of caveolin have been identified that differ at their extreme N terminus; (cid:97) -caveolin contains residues 1–178, and (cid:98) -caveolin contains residues 32–178. After recombinant expression in Sf21 insect cells, both (cid:97) - and (cid:98) -caveolin formed SDS-resistant high molecular mass oligomers of the same size as native caveolin. Morpho- logically, expression of either caveolin isoform resulted in the intracellular accumulation of a homogeneous population of caveolae-sized vesicles with a diameter between 50 and 120 nm (80.3 (cid:54) 14.8 nm). This indicates that each caveolin isoform can independently generate these structures and that caveolin residues 1–31


ologous system is sufficient to drive the formation of caveolae-like vesicles. Further functional analysis demonstrated that caveolin was capable of interacting with a known caveolin-interacting protein, Ha-Ras, when coexpressed in insect cells by co-infection with two recombinant baculoviruses. Taken together, our results demonstrate that baculovirus-based expression of caveolin in insect cells provides an attractive experimental system for studying the biogenesis of caveolae.
Caveolae are vesicular organelles located at or near the plasma membrane (1,2). They possess a characteristic uniform diameter in the range of 50 -100 nm and are most abundant in endothelial cells, which contain ϳ5,000 -10,000 caveolae/cell (3). As such, they have also been termed plasmalemmal vesicles or endothelial vesicles (4,5), although they are present in most cell types. Depending on the cell type, caveolae are found attached to the plasma membrane via a short caveolar neck, or they may reside intracellularly (2). Their exact function remains unknown, although they have been implicated in both cellular transport processes and signal transduction-related events (6 -8).
Caveolin, a 21-24-kDa integral membrane protein (9 -12), is a principal component of caveolae membranes in vivo (13). Caveolin may function within caveolae membranes as a scaffolding protein to organize and concentrate specific lipid classes and lipid-modified signaling molecules (14,15).
(i) Both the N-and C-terminal domains of caveolin face the cytosol and are thus accessible for the interaction with cytosolic molecules (16 -18). In accordance with this topology, caveolin is inaccessible to biotinylation probes that efficiently label other plasma membrane proteins within caveolae (15).
(iv) A membrane proximal region of the cytosolic N-terminal domain of caveolin, the caveolin-scaffolding domain, interacts directly with lipid modified signaling molecules such as G ␣ subunits and Ha-Ras (23,24). This protein-protein interaction is sufficient to recruit nonlipid-modified forms of G ␣ and Ha-Ras onto membranes both in vitro and in vivo (20,23,24). These self-assembly properties of caveolin also suggest that caveolin may be critical to the formation or the organization of caveolae membranes, acting as a plasma membrane "platform or scaffold" (15).
Several independent lines of evidence suggest that caveolin expression correlates with the formation of caveolae membranes. Caveolin is most abundantly expressed in cell types that contain numerous caveolae: endothelial cells, adipocytes, type I pneumocytes, smooth muscle cells, and fibroblasts (reviewed in Refs. 7 and 8). Both caveolin protein and caveolae organelles are induced 10 -25-fold during the process of adipocyte differentiation (25)(26)(27). In contrast, transformation of NIH 3T3 cells by activated oncogenes such as Ha-ras (G12V) or v-abl leads to the down-regulation of caveolin protein expression and the disappearance of caveolae (28). Interestingly, caveolin expression levels correlated inversely with the ability of these cells to grow in soft agar (28). This is consistent with the idea that caveolin may play a negative regulatory role in signal transduction as an allosteric effector of several distinct classes of signaling molecules (20,23,24). Caveolin preferentially recognizes the inactive GDP-liganded conformation of both G ␣ subunits and Ha-Ras but fails to interact with mutationally activated copies of these molecules (20,23,24).
Because caveolin protein expression parallels caveolae formation, this suggests that overexpression of caveolin can drive caveolae formation and that caveolin may be required for the formation of caveolae membranes. In accordance with this idea, recombinant expression of caveolin in a caveolin-negative cell line results in the correct biochemical targeting of caveolin to caveolae-enriched membrane fractions (18). However, a study by Fra et al. (29) demonstrating the morphological formation of caveolae in lymphocytes by caveolin expression resulted in the production of only a limited number of caveolae, a maximum of ϳ15 caveolae/cell; these caveolin-induced vesicles remained predominantly intracellular, although this was not noted by the authors. These studies remain inconclusive because two recent reports by Smart et al. (30) and Conrad et al. (31) suggest that caveolin can cycle between the Golgi and the plasma membrane in response to extracellular stimuli, that plasma membrane caveolae remain unaffected by this process, and that these caveolae do not contain caveolin.
Thus, it remains an open question whether caveolin is required or directs the formation of caveolae membranes. This is complicated by the fact that it is now recognized that there are two distinct isoforms of caveolin, which differ in their extreme N-terminal sequence; ␣-caveolin contains residues 1-178, and ␤-caveolin contains residues 32-178 (18). In the above study demonstrating that caveolin is not required for caveolae formation, these authors used mAb 1 2234 (30,31). This antibody epitope maps to caveolin residues 1-21 and is absolutely ␣-isoform-specific; it does not recognize ␤-caveolin, which lacks residues 1-31 (18). Thus, these authors cannot formally exclude the possibility that the "unaffected" caveolae that "lack caveolin" still contain the ␤-isoform of caveolin.
Also, some reports show the presence of caveolae in caveolinnegative cell lines (based on a lack of immunoreactivity with anti-caveolin IgG) (14). How can caveolae exist without caveolin expression? This apparent contradiction may be explained by the recent observation that caveolin is the first member of a gene family of related molecules; caveolin has been renamed caveolin-1 (27). There are now three known caveolin genes (caveolin-1, caveolin-2, and caveolin-3) and four corresponding caveolin protein products (caveolin-1␣, caveolin-1␤, caveolin-2, and caveolin-3) (18,27,32,33).
Caveolins 1, 2, and 3 are structurally homologous proteins but are immunologically distinct molecules; they have different but overlapping tissue distributions (18,27,32,33). For exam-ple, the expression of caveolin-3 is absolutely muscle-specific (skeletal and cardiac muscle cells) (32,34). Caveolin-1 is not expressed in these striated muscle tissues, but smooth muscle cells coexpress caveolins 1 and 3 (32,34). Furthermore, caveolin-1 and caveolin-2 are coexpressed in adipocytes and share the same overlapping tissue distribution (27). Thus, in a given mammalian cell, such as smooth muscle cells or fibroblasts, up to three or four immunologically distinct caveolin protein products may be coexpressed. Also, as yet unidentified caveolin genes are likely to exist. 2 Due to this unexpected complexity, we have chosen to simplify the system by using a heterologous expression system: baculovirus-based expression in insect cells (Sf21). As a first step, we have overexpressed caveolin-1␣ and caveolin-1␤ individually in insect cells. This system allows the production of milligram quantities of mammalian proteins for structural studies, and these proteins have been shown to be functional (35,36). Another advantage of the system is that insect cells correctly acylate exogenously expressed proteins, although they have a different system of glycosylation (35,36). This difference in glycosylation is not an issue for caveolin expression because the N-and C-terminal domains of caveolin remain entirely cytoplasmic and, as such, do not undergo luminal modifications (N-and O-linked glycosylation).
Our results indicate that heterologous expression of either caveolin-1␣ or caveolin-1␤ in Sf21 insect cells induces the intracellular accumulation of caveolae-like vesicles. The cytoplasm of these insect cells is literally filled with hundreds of homogeneous caveolae-sized vesicles. Quantitative morphological analysis indicates that these caveolin-induced vesicles have the correct uniform size range expected for caveolae vesicles (50 -120 nm in diameter; 80.3 Ϯ 14.8 nm). This system will allow the mass production and purification of a uniform population of recombinant caveolae-sized vesicles that contain a given caveolin isoform, greatly facilitating future in vitro biochemical studies.
It is important to note that we have systematically tested numerous expression systems to reconstitute caveolae formation by recombinant expression of caveolin-1 (18,27,32,34). These include transient expression of caveolin-1 in COS-7 and 293-T cells and stable expression of caveolin-1 in MDCK cells and FRT cells (a caveolin-1-negative cell line). However, none of these systems produces a recognizable increase in caveolae by electron microscopy, despite the fact that caveolin-1 is well expressed by Western blotting. 2 Thus, not all expression systems will be amenable to the type of analysis we describe here. One major difference between these systems and the baculovirus-based system is the higher levels of caveolin overexpression that can be achieved. This may account for the dramatic demonstration of caveolin functioning that the baculovirus system provides.

Construction of Recombinant Baculovirus-expressing Caveolin-1␣ and Caveolin-1␤
Caveolin-1␣ (starting with methionine at position 1) and caveolin-1␤ (starting with methionine at position 32) were generated by polymerase chain reaction using the canine caveolin cDNA, as described previously (14). The cDNA encoding either caveolin-1␣ or caveolin-1␤ were subcloned into a transfer plasmid vector, pBacPAK 9 (Clontech), using the BamHI and XbaI restriction sites. A mixture of 2 g of recombinant plasmid pBacPAK 9-Caveolin-1␣ or caveolin-1␤ DNA and 1 g of purified engineered baculoviral vector DNA BacPak 6 (Bsu36I digest) (Clontech) were transfected into insect Sf21 cells, as suggested by the manufacturer (35). Four days later, culture supernatants were removed and centrifuged at 1000 rpm for 10 min. Clarified supernatants containing wild-type and recombinant baculoviruses were plaque assayed on a monolayer of Sf21 cells. Occlusion-negative plaques were picked and seeded onto 2.5 ϫ 10 6 cells. After 3 days of incubation, cells and culture supernatants were removed and centrifuged at 1000 rpm for 10 min. The cell pellets were analyzed by immunoblotting analysis using anticaveolin-1 mAb 2297 or anti-Myc tag mAb 9E10. Those plaques testing positive for the presence of caveolin-1 were selected for three rounds of plaque purification. The selected plaques with highest yield of caveolin-1 expression were used as recombinant baculovirus stock for producing caveolin-1 protein by infecting insect Sf21 cells.

Characterization of Caveolin-1␣ and Caveolin-1␤ Expression in Sf21 Insect Cells Using the Baculovirus Expression System
The expression of caveolin-1␣ and caveolin-1␤ in the baculovirus expression system was evaluated by SDS-PAGE and Western blot analysis. For detection, samples were separated using SDS-PAGE (10% or 15% acrylamide) and transferred to nitrocellulose membrane for primary antibody binding (anti-caveolin mAb 2297 or anti-Myc mAb 9E10). HRP-conjugated secondary antibodies (1:5000 dilution, Amersham) were used to visualize bound primary antibodies by an enhanced chemiluminescence assay (ECL; Amersham). Total proteins from Sf21 cells infected with recombinant caveolin-1␣ and caveolin-1␤ baculovirus or purified recombinant caveolin-1 were quantitated by the BCA assay (Pierce).

Association of Recombinant Baculovirus-expressed Caveolin-1 with GST-Ha-Ras Fusion Proteins Using Coexpression of Ras and Caveolin in Sf21 Cells
The association of caveolin-1 and Ha-Ras in insect cells was evaluated essentially as we described for the association of caveolin-1 with Ha-Ras in mammalian cells using affinity purification on Ni 2ϩ -NTAagarose beads (24). Briefly, Sf21 cells (representing one 150-mm dish) were co-infected with recombinant caveolin-1 and GST-Ha-Ras baculoviruses. Infected insect cells were lysed in 6 ml of lysis buffer containing 50 mM Hepes, pH 7.5, 1 mM EDTA, 0.5% CHAPS, and protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 20 g/ml aprotinin, 1 g/ml pepstatin, and 1 g/ml leupeptin). After 30 min of extraction on ice, lysates were clarified by centrifugation for 15 min in a microcentrifuge at top speed (14,000 ϫ g) and adjusted to pH 8.0. Ni 2ϩ -NTA-agarose beads (200 l) were preequilibrated with Tris-buffered saline (TBS; 10 mM Tris, pH 8.0, and 0.15 M NaCl) and precleared using Sf21 cell lysate for 30 min at 4°C. The lysates were then incubated with the resin and rotated for 6 h at 4°C. After binding, beads were allowed to gently settle by gravity (5 min on ice) and washed extensively up to four times, rotating for 5 min each at 4°C as the following: twice with TBS and twice with TBS plus 30 mM imidazole. After washing, bound proteins were specifically eluted with TBS containing 200 mM imidazole, as per the manufacturer's instructions (Qiagen). These eluates were then subjected to immunoblot analysis with mAbs directed against caveolin-1 or Ras.

Detergent-free Purification of Caveolin-enriched Membrane Fractions
Caveolin-enriched membrane fractions were prepared as described previously for mammalian cells (24,34). Sf21 insect cells infected with recombinant caveolin baculovirus (150-mm dishes) were scraped into 2 ml of 500 mM sodium carbonate, pH 11.0. Cell suspensions were ho-mogenized using a loose-fitting Dounce homogenizer (10 strokes), sheared in a Polytron tissue grinder (three 10-s bursts; Kinematica GmbH, Brinkmann Instruments, Westbury, NY), and subjected to sonication using an ultrasonicator (three 20-s bursts; Branson Sonifier 250, Branson Ultrasonic Corp., Danbury, CT). The homogenate was then adjusted to 45% sucrose by the addition of 2 ml of 90% sucrose prepared in MBS (25 mM Mes, pH 6.5, and 0.15 M NaCl) and placed at the bottom of an ultracentrifuge tube. A 5-35% discontinuous sucrose gradient was formed above (4 ml of 5% sucrose/4 ml of 35% sucrose; both in MBS containing 250 mM sodium carbonate) and centrifuged at 39,000 rpm for 16 -20 h in an SW41 rotor (Beckman Instruments). A light-scattering band confined to the 5-35% sucrose interface was observed that contained ϳ95% of recombinantly expressed caveolin-1 but excluded most other cellular proteins.

Electron Microscopy
Both transmission electron microscopy and immunogold labeling were performed as described previously by our laboratory (14,38). Samples were examined under the Philips 410 transmission electron microscope.
Transmission Electron Microscopy-Samples were fixed with glutaraldehyde, postfixed with osmium tetroxide, and stained with uranyl acetate and lead citrate, as detailed by Sargiacomo et al. (14) and Lisanti et al. (38).
Whole-mount Electron Microscopy-Samples were prepared as described previously by Chang et al. (39) for the visualization of purified smooth muscle cell caveolae. Samples were fixed in 3% paraformaldehyde and air-dried onto carbon-coated grids. After quenching with NH 4 Cl, the grids were washed and postfixed with osmium tetroxide. The samples were then sequentially stained with tannic acid, uranyl acetate, and lead citrate.
Immunogold Labeling of Sections-Samples were fixed in 2.5% paraformaldehyde in PBS (pH 7.2) for 2 h on ice. Note that to preserve immunoreactivity with anti-caveolin IgG (mAb 2297) or anti-Myc IgG (9E10), it was necessary to omit fixation with glutaraldehyde and osmium tetroxide, although these conditions are not optimal for structural preservation. After washing with PBS, samples were dehydrated in graded dimethyl formamide and embedded in Lowicryl K4M resin. Thin sections were cut and incubated with anti-caveolin IgG (1:50 dilution); bound antibodies were visualized with 10 nm of gold-conjugated secondary IgG (Zymed Laboratories, South San Francisco, CA). Control experiments omitting anti-caveolin IgG indicated that gold labeling was specifically dependent on incubation with anti-caveolin IgG (38).

Morphometric Analysis
Random fields of insect cells expressing caveolin-1 were examined for vesicular profiles. The sizes of over 100 vesicles were measured and tabulated. Their diameters fell into six size ranges designated as a-f (Fig. 3F). The mean and S.D. were calculated using the program Excel.

Scanning Densitometry
Quantitation was performed as detailed previously (40). Protein bands were digitized by high resolution optical scanning; volumetric integration of signal intensities was carried out using ImageQuant Software (Fast Scan; Molecular Dynamics).

Baculovirus-based Expression of Caveolin-1 Isoforms in Sf21
Insect Cells-A strategy was developed for heterologous expression and purification of mammalian caveolin-1 isoforms using Sf21 insect cells. The entire coding region for canine caveolin-1␣ or caveolin-1␤ was integrated into an engineered baculovirus genome via a recombinant transfer plasmid, as detailed under "Experimental Procedures." For these constructions, a Myc epitope tag was placed at the C terminus of both caveolin isoforms with a polyhistidine tag following the Myc tag (Fig. 1A). The polyhistidine tag provides an anchor for purification on the affinity matrix, Ni 2ϩ -NTAagarose beads.
Both caveolin-1␣ and caveolin-1␤ were expressed very well in Sf21 cells using the baculovirus system. We estimate that expressed caveolin-1␣ or caveolin-1␤ represent about 15-20% of total proteins in Sf21 cells. Western blotting analysis shows that caveolin-1␣ migrated at 29 kDa, whereas caveolin-1␤ migrated at 26 kDa. This slightly slower mobility than endogenous caveolin-1 isoforms reflects the recombinant addition of Myc and polyhistidine tags (Fig. 1B). Previous biochemical and morphological studies have established that these tags do not interfere with the targeting of recombinant caveolin-1 to caveolae membranes (12, 17, 18, 27, 32).
These homo-oligomers of caveolin-1 are resistant to dissociation by SDS and 2-mercaptoethanol, except at elevated temperatures (boiling at 100°C) (15,19,27). Thus, the same conditions were used to determine whether caveolin-1 expressed in Sf21 cells forms similar high molecular mass oligomers. As expected, Fig. 2 shows that recombinant caveolin-1␣ or caveolin-1␤ expressed using the baculovirus system: (i) migrated as an SDS-resistant high molecular mass complex of more than 200 kDa; and (ii) that this complex is sensitive to heat treatment. These results suggest that caveolin-1 isoforms expressed using the baculovirus system are correctly folded, because they form high molecular mass oligomers of the same relative size as native caveolin-1.

Expression of Caveolin-1 Isoforms in Sf21 Cells Induces the Formation of a Large Homogeneous Population of Caveolae-like
Vesicles-Previous studies show that bacterially expressed caveolin-1 can function as endogenous caveolin in terms of its homo-oligomerization and its in vitro recruitment of signaling molecules onto model lipid membranes (20). However, caveolin-1 expression fails to drive the formation of any morphological structures in Escherichia coli that resemble caveolae. 2 Further functional analysis demonstrated that incorporation of recombinant caveolin-1␣ into model lipid membranes is cholesterol-dependent (20,21). Because it is well known that bacteria do not synthesize cholesterol, this may explain why caveolae did not form in E. coli cells that expressed caveolin-1.
Insect cells use plasma membrane sterols, including cholesterol (41). Thus, we have now used baculovirus-based expression of caveolin in insect cells with the the aim of reconstituting caveolae formation in vivo on a large scale. Electron microscopic analysis of uninfected insect cells or insect cells infected with other control proteins did not show the morphological appearance of caveolae ( Fig. 3A and data not shown). However, insect cells infected with a recombinant baculovirus carrying caveolin-1␣ accumulated a uniform population of caveolae-like structures within their cytoplasm (Fig. 3, B-D). Greater than 100 vesicles can be counted in the field shown in Fig. 3B (lower magnification). This population of vesicles appeared homogeneous in size and were the same size expected for mammalian caveolae, ϳ50 -120 nm in diameter. Quantitative morphological analysis indicates that ϳ68% of these caveolin-induced vesicles have a diameter of 65-95 nm (80.3 Ϯ 14.8 nm; Fig. 3F). Thus, in many ways these vesicles are morphologically indis-

FIG. 1. Construction of baculovirus-based vectors for expression of caveolin-1␣ and -1␤ in Sf21 insect cells.
A, schematic diagram summarizing the construction of recombinant caveolin baculoviruses for expression in insect cells. The caveolin-1 molecule can be divided into three sections: an N-terminal domain; a membrane spanning region; and a C-terminal domain. A myc epitope tag was placed at its C terminus, and a polyhistidine tag was placed following the myc tag for affinity purification by Ni 2ϩ -NTA-agarose chromatography. Caveolin-1␣ (Cav-1␣) is the long isoform that contains residues 1-178, whereas caveolin-1␤ (Cav-1␤) is the short isoform that contains residues 32-178. These isoforms are generated by two distinct translation initiation sites (methionine residues at positions 1 and 32, respectively). Note that mAb 2297 recognizes both caveolin-1 isoforms via an epitope within caveolin-1 residues 61-71, whereas mAb 9E10 recognizes the C-terminal Myc epitope (EQKLISEEDLN). B, recombinant expression of caveolin-1 isoforms in insect Sf21 cells using baculovirus-based expression vectors (see "Experimental Procedures"). Cell extracts were prepared from: (i) uninfected Sf21 cells (mock); (ii) cells infected with the engineered baculovirus vector (vector alone); (iii) cells infected with recombinant caveolin-1␣ baculovirus (Cav-1␣); or (iv) cells infected with recombinant caveolin-1␤ baculovirus (Cav-1␤). Cell lysates were then subjected to SDS-PAGE (15% acrylamide) and immunoblot analysis with mAb 2297 that recognizes both caveolin-1 isoforms. Equivalent amounts of total proteins were loaded in each lane. Identical results were obtained by immunoblotting with mAb 9E10 that recognizes the myc epitope. The differential mobility of ␣ and ␤ isoforms of caveolin-1 is indicated at right; positions of molecular mass markers (kDa) are indicated at left.

FIG. 2. Caveolin-1␣ and caveolin-1␤ expressed in Sf21 cells form high molecular mass SDS-resistant homo-oligomers.
Cell lysates were prepared from Sf21 cells infected with recombinant caveolin-1␣ or caveolin-1␤ baculoviruses. Samples with (ϩ) and without (Ϫ) heat treatment were separated by SDS-PAGE (8% acrylamide) and subjected to immunoblot analysis with mAb 2297 to detect SDS-resistant ϳ350-kDa caveolin homo-oligomers. Equivalent amounts of protein were loaded in each lane. Note that both caveolin-1␣ and caveolin-1␤ form a discrete high molecular mass homo-oligomer, as shown previously for native caveolin (15,19) and recombinant caveolin expressed in E. coli (20). tinguishable from their native counterparts in mammalian cells. However, none of these caveolin-induced vesicles were seen attached to the plasma membrane. It is important to note that similar results were obtained with insect cells expressing caveolin-1␤ (residues 32-178) (Fig. 3E). These results indicate that: (i) caveolin residues 1-31 are not functionally required for this process; and (ii) both caveolin isoforms are independently competent to drive the formation of these caveolae-sized structures.
Because these two caveolin isoforms show a predominantly nonoverlapping distribution within a single mammalian cell, we have speculated previously that coexpression of these two isoforms gives rise to at least two distinct populations of caveolae (18). Thus, our current findings with caveolin-1␣ and caveolin-1␤ expression generally support the idea that independent populations of caveolae can be generated by exclusively using either isoform.
Purification and Characterization of Recombinant Caveolinrich Vesicles Produced in Sf21 Insect Cells-To purify these caveolin-induced vesicles from the bulk of cellular membranes and cytosolic proteins, we used an established equilibrium sucrose density gradient system (24,34). In this detergent-free fractionation scheme, immunoblotting with anti-caveolin-1 IgG has been used previously to track the position of native mammalian caveolae (24,34). Fig. 4 illustrates that in this fractionation scheme, ϳ90 -95% of both caveolin-1 isoforms are confined to fractions 4 -6 and is separated from the bulk of Sf21 cellular proteins. These vesicles (fractions 4 -6) were collected and examined by wholemount electron microscopy, a technique used to visualize mam-malian caveolae. Using this technique, which preserves membrane structure and stains membranes black, these vesicles appeared predominantly as 50 -100-nm membranous structures, as expected (Fig. 5, upper panel, arrowheads). These vesicles were also sedimented, fixed with paraformaldehyde, and subjected to immunogold labeling with anti-caveolin IgG. Fig. 5 (lower panel) directly shows that these vesicles contain caveolin-1, although their morphology is not well preserved by this technique.
Thus, these purified recombinant vesicles biochemically and morphologically behave as native mammalian caveolae purified from cultured cells and whole tissue because they: (i) have the same buoyant density as mammalian caveolae; (ii) appear as 50 -100-nm structures by whole-mount electron microscopy; and (iii) contain ϳ95% of the recombinantly expressed caveolin-1 protein by Western blotting .   FIG. 4. Distribution of total cellular proteins, caveolin-1␣, and caveolin-1␤ in infected Sf21 cells using a carbonate-based fractionation scheme. Sf21 cells were infected with baculoviruses encoding either caveolin-1␣ or caveolin-1␤ and subjected to subcellular fractionation after homogenization in a buffer containing sodium carbonate (see "Experimental Procedures"). An aliquot from each of 13 1-ml sucrose gradient fractions was subjected to SDS-PAGE and transferred to nitrocellulose. The distribution of total proteins was revealed by staining with Ponceau S (upper panel). The distribution of recombinant caveolin-1␣ or caveolin-1␤ was detected by immunoblot analysis with anti-caveolin IgG (mAb 2297). Note that recombinantly expressed caveolin-1 is enriched in fractions 4, 5, and 6, essentially the same pattern as for endogenous caveolin in mammalian cells (24,34). This indicates that: (i) that ϳ90 -95% of recombinant caveolin has been incorporated into Sf21 membranes with the same buoyant density as native caveolae membranes in mammalian cells; and (ii) caveolin-1 is well separated from most cellular proteins. This is despite that fact that caveolin-1 now represents a significant fraction of total cellular protein within these lysates. Caveolin-1␣ or caveolin-1␤ are indicated at left (Cav-1␣ and Cav-1␤, respectively); positions of molecular mass markers (kDa) are indicated at right.

FIG. 5. Electron microscopic analysis of purified recombinant caveolin-rich membrane vesicles. Morphology is shown in the upper panel.
Purified caveolin-rich vesicles (prepared as detailed in Fig. 4) were analyzed for their overall morphology by whole-mount electron microscopy, as described previously for the visualization of purified smooth muscle cell caveolae (39). As with preparations of native caveolae visualized by the same technique, these purified recombinant caveolin-rich vesicles appeared as ϳ50 -100-nm vesicular structures (arrowheads). Identical results were obtained with both caveolin-1␣ and caveolin-1␤. Bar, 100 nm. Immuno-labeling is shown in the lower panel.
Purified caveolin-rich vesicles were also analyzed by immunogold cytochemistry. These vesicles were sedimented and processed as a packed pellet for fixation in paraformaldehyde and immunogold labeling with anti-caveolin IgG (mAb 2297; detailed under "Experimental Procedures"). Bound primary antibodies were visualized with a 10-nm goldconjugated secondary antibody. Control experiments indicated that gold labeling was specifically dependent on incubation with anti-caveolin IgG (data not shown). Note that to preserve immunoreactivity with anti-caveolin IgG (mAb 2297), it was necessary to omit fixation with glutaraldehyde and OsO 4 . This allows immuno-labeling but does not preserve the morphology of the sample. This is common pitfall of immunoelectron microscopy, as we have noted previously (38). Identical results were obtained with both caveolin-1␣ and caveolin-1␤. Bar, 100 nm. These vesicles must contain a very high concentration of caveolin because it has been estimated that immunogold labeling of caveolin with anti-caveolin IgGs has an efficiency of less than 5% (29).

Reconstituting the Interaction of Ha-Ras with Caveolin-1 by Co-infection of Sf21 Cells with Two Recombinant
Baculoviruses-Ha-Ras is highly concentrated in purified preparations of caveolae membranes (24,42). In addition, caveolin interacts directly with wild-type Ha-Ras, but not a mutationally activated copy of the same molecule, Ha-Ras (G12V) (24). The caveolin-Ras interaction does not strictly require lipid modification of Ha-Ras (24). Farnesylation of Ras is normally required or greatly facilitates its caveolar localization (24,42). However, recombinant overexpression of caveolin in intact mammalian cells is sufficient to recruit a nonfarnesylated mutant of Ras (C186S) onto membranes, overcoming the normal lipid requirement (24). Thus, we have proposed that caveolin may act as a membrane-anchored scaffolding protein to sequester caveolin-interacting signaling molecules, such as Ha-Ras, within caveolin-rich membrane microdomains (15,20,24).
Here, we used this established interaction to examine whether caveolin-1, expressed in insect cells, is functionally competent to interact with Ha-Ras. For this purpose, we coexpressed a form of Ha-Ras (GST-Ha-Ras) (43) with polyhistidine-tagged caveolin-1 by co-infection with two distinct recombinant baculoviruses. As a control for nonspecific association, we also expressed GST-Ha-Ras alone. Lysates from infected insect cells were then prepared and subjected to affinity chromatography on Ni 2ϩ -NTA-agarose. After binding, extensive washing and elution with imidazole, the eluates were analyzed by Western blotting with mAbs to Ha-Ras and caveolin-1. Fig.  6A shows that polyhistidine-tagged caveolin was specifically retained and eluted from the Ni 2ϩ -NTA resin. In addition, little or no Ha-Ras was retained and eluted when cells were infected with GST-H-Ras alone. Relative to this control, ϳ4 -5 times more Ha-Ras was recovered and eluted when coexpressed with polyhistidine-tagged caveolin-1 (Fig. 6B). These findings directly support the previous observation that Ha-Ras associates and interacts directly with caveolin in mammalian cells (24). Thus, this system should allow us and others to systematically evaluate the interaction of caveolin-1 with candidate caveolininteracting proteins. DISCUSSION In this report, we have developed an experimental system in which caveolin-1 has been overproduced in Sf21 insect cells by using baculovirus-based vectors. Recombinant caveolin-1 was expressed at high levels in insect cells and was indistinguishable from the native caveolin protein in terms of its apparent molecular weight in SDS-PAGE gels, homo-oligomeric state, and interaction with a known caveolin-interacting signaling molecule, Ha-Ras. Most importantly, the ectopic expression of caveolin-1 in insect cells induced the intracellular accumulation of a plethora of caveolae-like structures, which contain caveolin-1. These vesicles have the same uniform diameter (50 -100 nm) as endogenous caveolae in mammalian cells.
What is the mechanism by which the ectopic expression of caveolin-1 induces the formation of a homogeneous population of caveolae-like structures in insect cells? It is known that caveolin undergoes two stages of oligomerization (15,19). In the endoplasmic reticulum, caveolin monomers assemble into discrete multivalent homo-oligomers, containing ϳ14 -16 monomers per oligomer (15,19). Subsequently, these individual caveolin homo-oligomers (4 -6-nm spherical particles) can interact with each other to form clusters of particles that are ϳ25-50 nm in diameter (15). Also caveolin homo-oligomers interact specifically with glycosphingolipids (22) and cholesterol (21) and require a high cholesterol content to insert into model lipid membranes (20,21). Thus, we envisage that through the interaction of caveolin with itself and the caveolinmediated selection of endogenous lipid components, a caveolae-sized vesicle is generated.
There are numerous advantages to this heterologous expression system for studying caveolae vesicle formation: (i) vectors and host cells are commercially available; (ii) hundreds of uniform caveolae-sized vesicles are produced per cell (Fig. 3); (iii) the simple purification of these structures offers large quantities of recombinant caveolae for in vitro biochemical studies (Figs. 4 and 5); (iv) it allows efficient coexpression of caveolin-1 with other proteins by co-infection with two or more recombinant baculoviruses encoding different proteins; and (v) affinity purification of caveolin-1 and caveolin-1-associated proteins is possible using the polyhistidine tag and Ni 2ϩ -NTA affinity chromatography, as demonstrated for Ha-Ras (Fig. 6).
The system also provides a clear functional assay for caveolin-1 in driving the formation of caveolae-sized vesicles. This will be useful in future mutagenesis studies aimed at defining which regions of caveolin-1 are critical for caveolin-induced vesicle formation. For example, using this approach, we have

FIG. 6. Association of Ha-Ras with caveolin-1 in Sf21 cells coinfected with polyhistidine-tagged caveolin-1 and GST-Ha-Ras baculoviruses.
A, cell lysates were prepared from Sf21 cells either infected with GST-Ha-Ras baculovirus alone or co-infected with GST-H-Ras and caveolin-1␣ (Ϫ and ϩ signs indicate without and with co-infection with caveolin-1). These lysates were then incubated with Ni 2ϩ -NTA beads, washed (four times), and specifically eluted with imidazole. The eluted proteins were then subjected to SDS-PAGE and Western blotting analysis with mAbs directed against caveolin and Ha-Ras. Lysates were adjusted to contain equivalent amounts of GST-Ha-Ras protein. GST-Ha-Ras migrates at ϳ45-46 kDa, as expected, since it should have the cumulative molecular mass of GST (26)(27) and Ha-Ras (20 kDa). Note that GST-Ha-Ras is better retained and eluted from Ni 2ϩ -NTA beads in cells co-infected with polyhistidinetagged caveolin-1␣. Quantitation reveals that coexpression with polyhistidine-tagged caveolin-1␣ results in a ϳ4-fold enrichment in specific GST-H-Ras binding relative to nonspecific binding. Caveolin-1␣ and GST-H-Ras are indicated at right; positions of molecular mass markers (kDa) are indicated at left. B, quantitation of the amount of GST-Ha-Ras protein bound and eluted from Ni-NTA beads in the absence (Ϫ) or presence (ϩ) of polyhistidine-tagged caveolin-1. Densitometry is expressed in arbitrary units; our results indicate that the specific binding of Ha-Ras to caveolin-1 is ϳ4-fold the nonspecific binding observed without caveolin coexpression. already determined that residues 1-31 of caveolin-1 are not required for this process, because expression of caveolin-1␤ (residues 32-178) also drives the formation of these caveolaesized vesicles (see "Results" ; Fig. 3E).
This system represents an important first step toward reconstituting caveolae biogenesis in vitro. Caveolin-induced vesicles most likely represent transient vesicular intermediates on the way toward fusion with the plasma membrane to form plasma membrane-attached caveolae. In this regard, caveolin-induced vesicles may represent the precursor to the caveolae bulb that is seen attached to a discrete caveolar "neck or annulus." Other factors may exist to construct caveolar necks that connect or fuse caveolae bulbs to the plasma membrane. Purified caveolininduced vesicles (donor) could be used along with purified plasma membranes (acceptor) to reconstitute this fusion event in vitro by the addition of cytosol and an ATP-regenerating system. This type of approach has been extremely useful in defining the factors that contribute to the formation of other vesicular organelles.
Fra et al. (29) have also attempted to reconstitute caveolae formation in vivo (29). They expressed a myc-tagged version of caveolin (now known as caveolin-1␣) using a different transient expression system: SFV (Semliki Forest virus)-based vectors and a transformed lymphocytic cell line (29). However, they did not evaluate the effects of caveolin-1␤. In contrast to our present results, they observed that only a few caveolae or caveolaelike vesicles were induced, a maximum of ϳ15 per cell versus Ͼ200 per field observed here. The caveolae they observed were also predominantly intracellular, although this was not noted by the authors. In addition, these vesicles had a smaller, more restricted diameter (54 Ϯ 6.4 nm) than endogenous caveolae (50 -100 nm) and the caveolin-induced vesicles we observed here (80.3 Ϯ 14.8 nm). Thus, the baculovirus-based system more closely approximates the state of caveolae within endothelial cells: (i) where there are ϳ5,000 -10,000 caveolae/endothelial vesicles per cell; and (ii) these caveolae/endothelial vesicles are mostly intracellular, but some are attached to the plasma membrane (3).
In summary, we conclude that high levels of caveolin expression can drive the formation of a homogeneous population of intracellular vesicles that resemble caveolae in size and that contain caveolin-1, but that caveolin-1 alone is not responsible for the formation of the caveolar neck. Other protein cofactors may be required to construct the caveolar neck or annulus for joining the caveolae bulb to the plasma membrane proper.