INTRODUCTION

Caveolae are vesicular organelles located near or attached to the plasma membrane (1, 2). They represent an appendage of the plasma membrane. Caveolae are most abundant in endothelial cells, adipocytes, smooth muscle cells, and fibroblasts, although they are thought to exist in most cell types (reviewed in Refs. 3-5). The exact function of caveolae remains largely unknown; however, they are thought to function in both cellular transport processes and signal transduction (3, 5).

Caveolin, a 21-24-kDa integral membrane protein, is a principal component of caveolae membranes in vivo (6, 7). It has been proposed that caveolin functions as a scaffolding protein to organize and concentrate specific lipids (cholesterol and glyco-sphingolipids) (8, 9) and lipid-modified signaling molecules (Src-like kinases, H-Ras, and G-proteins) (10-12) within caveolae micro-domains (13). Recently, we and others have identified a family of caveolin-related proteins (caveolin-2 and caveolin-3) (14-17); caveolin has been retermed caveolin-1 (14).

Caveolin-1 appears to be an essential component of caveolae (18). For example, caveolin-1 protein expression directly parallels caveolae formation during adipocyte differentiation (14, 19, 20). Conversely, caveolin-1 mRNA and protein expression are lost or reduced during cell transformation, and caveolae are absent from these cell lines (21). Recombinant over-expression of caveolin-1 in caveolin-deficient cell lines results in: (i) the correct biochemical targeting of caveolin-1 to caveolae-enriched membrane fractions in FRT cells (22) and (ii) the formation of recombinant caveolae vesicles in lymphocytes (23) and Sf21 insect cells (18). These results provide direct evidence that caveolin family members participate in caveolae formation.

However, it remains unknown how caveolin-1 expression induces caveolae formation. This may be related to the self-assembly properties of caveolin-1. Caveolin-1 undergoes two stages of oligomerization. First, in the endoplasmic reticulum, caveolin-1 monomers assemble into discrete multivalent homo-oligomers, containing ~14-16 monomers per oligomer (13, 24). Subsequently, these individual caveolin-1 homo-oligomers (4-6 nm spherical particles) can interact with each other to form clusters of particles that are ~25-50 nm in diameter (13). Also caveolin-1 homo-oligomers interact specifically with glycosphingolipids (25) and cholesterol (8, 9) and require a high cholesterol content (30%) to insert into model lipid membranes (8, 9). Thus, we envisage that through the interaction of caveolin-1 with itself and the caveolin-mediated selection of endogenous lipid components, a caveolae-sized vesicle is generated (18).

The specialized lipid composition of caveolae is thought to convey resistance of this membrane domain to detergent solubilization by Triton X-100 (at low temperatures) (20, 26-31). This property appears to be unique to caveolae membranes. For example, when intact cells were fixed in paraformaldehyde, extracted with Triton X-100, and then examined by electron microscopy, the insoluble membranes that remained were found to be caveolae (32). However, it is not known whether caveolin-1 contributes to the detergent insolubility of caveolae membranes.

Caveolin proteins can be divided into three distinct regions: (i) a cytoplasmic N-terminal domain; (ii) an unusual 33-amino acid hydrophobic membrane spanning segment; and (iii) a cytoplasmic C-terminal domain (6, 14, 15, 17, 22, 33-35). Here, we have employed a deletion mutagenesis approach to dissect which regions of caveolin-1 are required for its detergent insolubility, homo-oligomerization, and targeting to low density Triton-insoluble membrane fractions that are enriched in caveolae-membranes. Our results suggest a novel role for the C-terminal domain in mediating homo-typic caveolin-caveolin interactions between individual caveolin-1 oligomers.

EXPERIMENTAL PROCEDURES

Monoclonal antibody 2297 directed against full-length caveolin-1 was the generous gift of Dr. John R. Glenney (22) (Transduction Laboratories, Lexington, KY). The monoclonal antibody 9E10 was provided by the Harvard Monoclonal Antibody Facility (Cambridge, MA). The cDNAs for caveolins-1, -2, and -3 were as we described previously (14, 15, 26). A variety of other reagents were purchased commercially: fetal bovine serum (JRH Biosciences); prestained protein markers (Life Technologies, Inc.); and Lab-Tek chamber slides (Nunc Inc., Naperville, IL).

MDCK¹ cells were propagated in t75 tissue culture flasks in DME supplemented with antibiotics and 10% fetal bovine serum (26). For experiments, cells were seeded at high density in 6-well dishes and harvested for experiments 2-3 days after reaching confluency. The expression levels of a given transfected antigen were increased by an overnight incubation with normal medium containing 10 mM sodium butyrate, as described previously (4, 36).

In order to recombinantly express epitope-tagged forms of caveolin-1 in MDCK cells, we incorporated the myc epitope tag into the extreme N terminus (MGG-caveolin) using the polymerase chain reaction (see Fig. 1). We placed GG as a spacer between the epitope and the caveolin coding sequences, as has been suggested previously (34, 37). Correct placement of the epitope tag and caveolin coding sequences were verified by double-stranded DNA sequencing in both directions. Epitope-tagged forms of caveolin were subcloned into the multiple cloning site (HindIII/BamHI) of the vector pCB7 (containing the hygro^R marker; gift of J. Casanova, Massachusetts General Hospital) for expression in MDCK cells. MDCK cells were stably transfected using a modification of the calcium phosphate precipitation procedure, as we described previously (4, 36). After selection in medium supplemented with 400 µg/ml hygromycin B, resistant colonies were picked by trypsinization using cloning rings. Individual clones were screened by immunofluorescence and immunoblotting for recombinant expression of caveolin-1. Epitope-tagged forms of caveolin-1 expressed in MDCK cells were detected using monoclonal antibody, 9E10, that recognizes the myc epitope (EQKLISEEDLN). Constructs were also transiently transfected into COS-7 cells by the DEAE-dextran method (38).

C

N

C

N

The Triton insolubility of a given protein was determined essentially as we described previously with minor modifications (27). Briefly, MDCK cells grown to confluence in 35-mm dishes were first extracted with 1 ml of Mes-buffered saline (25 mM Mes, pH 6.5, 0.15 M NaCl) containing 1% Triton X-100 and 1 mM phenylmethylsulfonyl fluoride. After 30 min on ice without agitation, the Triton-soluble extract (see Fig. 2, S) was gently decanted, and the remaining Triton-insoluble material (see Fig. 2, I) was solubilized in 1 ml of 1% SDS. Each extract was then concentrated by acetone precipitation, solubilized in 10% SDS, and diluted into 4 × sample buffer for analysis by SDS-PAGE and Western blotting.

C

N

C

N

The molecular mass of caveolin-1 deletion mutants was estimated as described previously for mammalian caveolins-1, -2, and -3 (13-15). Briefly, samples were dissociated in Mes-buffered saline containing 60 mM octyl-glucoside. Solubilized material is loaded atop a 5-40% linear sucrose gradient and centrifuged at 50,000 rpm (340,000 × g) for 10 h in an SW 60 rotor (Beckman). Gradient fractions were collected from above and subjected to immunoblot analysis. Molecular mass standards for velocity gradient centrifugation were as described (13-15). Note that caveolin-2 is a dimer of ~40 kDa, and it correctly migrates between the 29 and 66 kDa molecular mass standards using this velocity gradient system (14).

MDCK cells (recombinantly expressing caveolin-1 deletion mutants) were grown to confluence in 150-mm dishes and used to prepare caveolin-enriched membrane fractions, essentially as described (20, 26, 28, 39). Briefly, MDCK cells from a confluent 150-mm dish were scraped into 2 ml of Mes-buffered saline (25 mM Mes, pH 6.5, 0.15 M NaCl) containing 1% Triton X-100 and 1 mM phenylmethylsulfonyl fluoride. Homogenization was carried out initially with 10 strokes of a loose-fitting Dounce homogenizer, followed by a Polytron tissue grinder (three 10 s bursts; Brinkmann Instruments, Westbury, NY). The homogenate was adjusted to 40% sucrose by the addition of 2 ml of 80% sucrose prepared in Mes-buffered saline and placed at the bottom of an ultracentrifuge tube. A 5-30% linear sucrose gradient was formed above the homogenate and centrifuged at 39,000 rpm for 16-20 h in a SW41 rotor (Beckman Instruments, Palo Alto, CA). A light-scattering band confined to the 15-20% sucrose region is harvested, diluted 3-fold with Mes-buffered saline, and pelleted in the microfuge (14,000 × g; 15 min at 4 °C). The majority of protein remained within the 40% sucrose region of the gradient. Approximately 4-6 µg of caveolin-enriched domains were obtained from one 150-mm dish of MDCK cells representing 10 mg of protein, a yield of ~0.05% relative to the homogenate. We and other laboratories have demonstrated that these domains exclude a variety of organelle-specific membrane markers (for endoplasmic reticulum, Golgi, lysosomes, mitochondria, and noncaveolar plasma membrane) but are dramatically enriched ~2000-fold in caveolin-1, a caveolar marker protein (4, 20, 26, 28, 31, 39).

Gradient fractions were separated by SDS-PAGE (10% or 15% acrylamide) and transferred to nitrocellulose. After transfer, nitrocellulose sheets were stained with Ponceau S to visualize protein bands and subjected to immunoblotting with 9E10 ascites (1:500) to visualize epitope-tagged caveolin-1 deletion mutants. For immunoblotting, incubation conditions were as described by the manufacturer (Promega and Amersham Corp.), except we supplemented our blocking solution with both 1% bovine serum albumin and 1% nonfat dry milk (Carnation). Quantitation was performed with a Molecular Dynamics computing densitometer. To ensure that these estimates were made in the linear range, we used multiple autoradiographic exposures and monitored their linearity using the densitometer essentially as described (40).

The construction, expression, and purification of GST-caveolin-1 fusion proteins was as we described previously (10, 13, 22). Briefly, full-length caveolin (residues 1-178) and the C-terminal domain of caveolin-1 (residues 135-178) were subcloned into the vector pGEX-4T-1. After expression in Escherichia coli (BL21 strain, Novagen, Inc.), GST-caveolin fusion proteins were affinity purified by affinity chromatography on glutathione-agarose beads (41).

The interaction of GST-caveolin-1 fusion proteins with caveolins-1, -2, and -3 and caveolin-1 deletion mutants was evaluated essentially as we described for the interaction of caveolin-1 with heterotrimeric G-protein - subunits (10), H-Ras (11), and Src-tyrosine kinases (12). Briefly, GST or GST-caveolin-1 fusion proteins bound on glutathione agarose beads were extensively washed first with phosphate-buffered saline once and lysis buffer containing protease inhibitors three times. These beads contained ~100 pmol of a given fusion protein per 100 µl of packed volume. Approximately 100 µl of this material was incubated with 1 ml of precleared lysates by rotating overnight at 4 °C. After binding, the beads were extensively washed (6-8×) with wash buffer containing 50 mM Hepes, pH 7.5, 120 mM NaCl, 1 mM EDTA, 0.5% CHAPS, and protease inhibitors. Finally, associated proteins were eluted with 100 µl of elution buffer containing 50 mM Tris, pH 8.0, 1 mM EDTA, 1% Triton X-100, 10 mM reduced glutathione, and proteinase inhibitors. The eluate was mixed 1:1 with 2× sample buffer and subjected to SDS-PAGE (10 or 15% acrylamide). After transfer to nitrocellulose, Western blot analysis was performed with 9E10 ascites (1:500) to visualize bound epitope-tagged forms of caveolin. Horseradish peroxidase-conjugated secondary antibodies (1:5000 dilution, Amersham Corp.) were used to visualized bound primary antibodies by an enhanced chemiluminescence assay (ECL) (Amersham Corp.).

RESULTS AND DISCUSSION

Epitope-tagged forms of full-length caveolin-1 (Cav-1FL) and two caveolin-1 deletion mutants (Cav-1C and Cav-1N) were constructed as illustrated schematically in Fig. 1. Note that Cav-1C retains the complete membrane spanning region and N-terminal domain; and Cav-1N retains the complete membrane spanning region and C-terminal domain. A myc epitope tag was placed at the extreme N terminus of these constructions to distinguish these recombinantly expressed forms of caveolin-1 from endogenous caveolin-1. It is important to note that N-terminally myc-tagged caveolin-1 is transported to caveolae with the same efficiency as endogenous caveolin, as shown previously (22, 34, 42).

MDCK cells were stably transfected and clones expressing equivalent amounts of Cav-1FL, Cav-1C, or Cav-1N were selected for further analysis (see below).

Endogenous caveolin-1 is insoluble in nonionic detergents such as Triton X-100 at low temperatures (26, 27); however, it can be efficiently solubilzed by the mild detergent, octyl-glucoside (26, 27). It is thought that octyl-glucoside solubilization occurs through the displacement of endogenous lipid components (such as glycosphingolipids and cholesterol) that are concentrated within caveolae membranes and interact directly with caveolin (20, 26-31).

Fig. 2A shows that in MDCK cells greater than 85-90% of total cellular proteins are Triton-soluble; as expected, endogenous caveolin-1 remains Triton-insoluble under these conditions. Full-length epitope-tagged caveolin-1 (Cav-1FL) also remained ~80-90% Triton-insoluble, whereas both deletion mutants (Cav-1C or Cav-1N) were predominantly Triton-soluble (Fig. 2B). More specifically, Cav-1C was ~70-80% Triton-soluble, whereas Cav-1N was virtually 100% Triton-soluble. These results clearly indicate that an intact N-terminal domain and an intact C-terminal domain are both required to confer optimal Triton insolubility. This is perhaps surprising, because it would be predicted that the transmembrane domain would be sufficient to confer Triton insolubility through interaction with specific caveolar lipid components.

Caveolin-1 forms a ~350-kDa homo-oligomer containing ~14-16 caveolin monomers per oligomer (13, 24). These homo-oligomers are thought to function as building blocks in the construction of caveolae membranes. Similarly, caveolin-3 forms homo-oligomers of the same size as caveolin-1 (15). In contrast, caveolin-2 exists as a homo-dimeric complex (14).

Thus, we next investigated the oligomeric state of caveolin-1 deletion mutants. For this purpose, we employed an established velocity gradient system developed previously to study the oligomeric state of caveolins-1, -2, and -3 (13-15). Fig. 3 shows that Cav-1C behaved as a high molecular mass complex, migrating between the 200- and 443-kDa molecular mass standards (peak fractions 6, 7, and 8). In contrast, Cav-1N failed to form a high molecular mass oligomer. The migration of full-length epitope-tagged caveolin-1 (Cav-1FL) is shown for comparison. These studies directly implicate the N-terminal domain in the formation of caveolin-1 oligomers.

C

N

These results are consistent with our previous studies using GST-caveolin-1 bacterial fusion proteins that mapped this homo-oligomerization activity to a 41-amino acid membrane proximal region of the cytoplasmic N-terminal domain (13). These results also indicate that homo-oligomerization and an intact transmembrane domain are not sufficient to confer detergent insolubility, suggesting an unknown role for the C-terminal domain in this process.

To separate membranes enriched in caveolin-1 from the bulk of cellular membranes and cytosolic proteins, an established equilibrium sucrose density gradient system was utilized (4, 10, 20, 22, 26, 28, 31, 39, 43-45). In this fractionation scheme, immunoblotting with anti-caveolin-1 IgG can be used to track the position of caveolae-derived membranes within these bottom-loaded sucrose gradients. Using this procedure, caveolin-1 is purified ~2000-fold relative to total cell lysates as ~4-6 µg of caveolin-rich domains (containing ~ 90-95% of total cellular caveolin-1) are obtained from 10 mg of total MDCK proteins (4, 10). We and others have shown that these caveolin-rich fractions exclude >99.95% of total cellular proteins and also markers for noncaveolar plasma membrane, Golgi, lysosomes, mitochondria, and endoplasmic reticulum (20, 26, 28).

Fig. 4 illustrates that in this fractionation scheme ~85-90% of full-length epitope-tagged caveolin-1 (Cav-1FL) is targeted to these low density Triton-insoluble membrane fractions that are enriched in caveolae membranes (fractions 4 and 5). In contrast, both caveolin-1 deletion mutants (Cav-1C and Cav-1N) were quantitatively excluded from these caveolae-enriched fractions. These results indicate that both a complete N- and C-terminal domain are required for incorporation into caveolae membranes.

C

N

C

N

Caveolin-1 undergoes two stages of oligomerization. First, caveolin-1 monomers assemble into discrete multivalent homo-oligomers, containing ~14-16 monomers per oligomer (13, 24). Second, these individual caveolin-1 homo-oligomers (4-6 nm spherical particles) interact with each other to form caveolae-sized clusters of particles that are ~25-50 nm in diameter (13). These clusters exhibit side-by-side packing of individual homo-oligomer units, suggesting that a given caveolin-1 homo-oligomer interacts with several of its nearest neighbors simultaneously (13). In this regard, we and others have suggested that caveolin homo-oligomers represent the assembly units of caveolae (13, 24).

To understand how caveolin-1 homo-oligomers interact with each other, we first reconstituted this caveolin-caveolin interaction in vitro. Caveolin-1 homo-oligomers bound to glutathione-agarose beads (GST-FL-Cav-1) were incubated with lysates of MDCK cells expressing full-length epitope-tagged caveolin-1 (Cav-1FL). Fig. 5 shows that myc epitope-tagged caveolin-1 interacted specifically with GST-FL-Cav-1 but not with GST alone. Virtually identical results were obtained using two purified recombinant caveolin-1 fusion proteins produced in E. coli (GST-FL-Cav-1 and Cav-1-myc-H₇), indicating that this represents a direct interaction (not shown). This binding activity was localized to the C-terminal domain of caveolin using a variety of GST-caveolin fusion proteins (Fig. 6 and data not shown).

C

N

C

N

Two additional lines of evidence suggest that this caveolin-caveolin interaction is very specific. The C-terminal domain of caveolin-1 (GST-C-Cav-1) failed to interact with the two caveolin-1 deletion mutants (Cav-1C and Cav-1N; Fig. 6). These results indicate that both an intact N- and C-terminal domain are required for interaction with the C-terminal domain of an adjacent caveolin-1 molecule. Also this interaction appears to be homotypic because no interaction of caveolin-1 was observed with caveolin-2 or caveolin-3 under the same conditions (Fig. 7). This is despite the fact that caveolins-1, -2, and -3 are very closely related proteins. Caveolin-2 is 58% similar and 38% identical to caveolin-1, whereas caveolin-3 is 85% similar and 65% identical to caveolin-1 (14, 15). These and all other results are summarized in Table I.

Table I. Summary of the properties of caveolin-1 deletion mutants and full-length caveolins-1, -2, and -3  Detergent insolubility Homo-oligomer formation Targeting to caveolae-enriched fractions Interaction with the Cav-1 C-terminal domain Cav-1   FL + + HMWd + +   C  a + HMW       N         Cav-2 +b + Dimerb +b   Cav-3 +c + HMWc +c   a  20-30% detergent-insoluble. b  Scherer et al. (14). c  Tang et al. (15). d  HMW, high molecular mass oligomer of ~300-350 kDa.

Taken together, our results immediately suggest a possible mechanism for the side-by-side packing of caveolin homo-oligomers. The C-terminal domain of one homo-oligomer would interact with both the C-terminal domain and the N-terminal domain of an adjacent homo-oligomer (Fig. 8). This would allow for the construction of a network of caveolin-caveolin interactions and provide a simple explanation for why an intact C-terminal domain is required for Triton insolubility (Fig. 2) and targeting of caveolin-1 to low density Triton-insoluble membrane fractions (Fig. 4) but is not required for homo-oligomer formation (Fig. 3).

It has been previously demonstrated that an anti-peptide antibody generated against the extreme C terminus of caveolin-1 (residues 161-178) only recognizes caveolin-1 associated with the Golgi and trans-Golgi network (46). However, this antibody fails to recognize caveolin-1 associated with plasma membrane and caveolae membranes (46). This is despite the observation that >90% of total cellular caveolin is associated with caveolae membranes at steady state (7, 29, 46).

These results indicate that this C-terminal epitope is exposed when caveolin-1 is in the Golgi and becomes masked when caveolin-1 is integrated within caveolae membranes. Masking of this C-terminal epitope within caveolae membranes could be explained by our current findings that the C-terminal domain plays an important role in caveolin-caveolin interactions. Perhaps this interaction begins during or after transport of caveolin-1 from the Golgi to the plasma membrane, thereby facilitating caveolae formation. In this regard, it is important to note that Golgi-associated caveolin-1 is completely Triton-soluble (39) but that caveolin-1 associated with plasma membrane caveolae is Triton-insoluble (20, 26-31).

The extreme C terminus is one of the regions within caveolin family members that demonstrates the greatest protein sequence divergence (see Tang et al., 1996 for an alignment (15)). This may explain why the C-terminal domain of caveolin-1 only recognizes caveolin-1 homo-oligomers but not caveolin-2 or caveolin-3 under identical conditions.