Caveolin-1 Binding to Endoplasmic Reticulum Membranes and Entry into the Regulated Secretory Pathway Are Regulated by Serine Phosphorylation PROTEIN SORTING AT THE LEVEL OF THE ENDOPLASMIC RETICULUM*

Caveolin-1 serves as the main coat protein of caveolae membranes, as an intracellular cholesterol shuttle, and as a regulator of diverse signaling molecules. Of the 12 residues conserved across all caveolin isoforms from all species examined to date, only Ser 80 and Ser 168 could serve as phosphorylation sites. We show here that mimicking chronic phosphorylation of Ser 80 by mutation to Glu ( i.e. Cav-1(S80E)), blocks phosphate incorporation. However, Cav-1(S168E) is phosphorylated to the same extent as wild-type caveolin-1. Cav-1(S80E) targets to the endoplasmic reticulum membrane, remains oligomeric, and maintains normal membrane topology. In contrast, Cav-1(S80A), which cannot be phosphorylated, targets to caveolae membranes. Some exocrine cells secrete caveolin-1 in a regulated manner. Cav-1(S80A) is not secreted by AR42J pancreatic adenocarcinoma cells even in the presence of dexamethasone, an agent that induces the secretory phenotype. Conversely, Cav-1(S80E) is secreted to a greater extent than wild-type caveolin-1 following dexamethasone treatment. We conclude that caveolin-1 phosphorylation on invariant into the regu- lated secretory pathway. In exocrine cells, phosphorylation of caveo-lin-1 invariant residue Ser 80 is required for targeting of the protein to the ER lumen. Blocking this phosphorylation event by mutation to alanine prevents ER targeting and subsequent secretion. Conversely, mimicking phosphorylation by mutation to glutamate increases regulated exocytosis.

A 22-kDa phosphoprotein first detected in Rous sarcoma virus-transformed fibroblasts (1), and normal fibroblasts overexpressing an activated form of the transforming kinase Src (2), was identified simultaneously as caveolin (now termed caveolin-1), the principal coat protein of caveolae (3), and as VIP21, an integral membrane protein component of detergent resistant, trans-Golgi derived vesicles (4). These initial descriptive reports all foreshadowed the functions of this molecule. First, caveolin-1 is a transformation-dependent substrate of v-Src, and is down-regulated in a number of cancer-derived cell lines (reviewed in Ref. 5). Second, it is a component of the cell's transport machinery, contributing to the influx and efflux of cellular cholesterol (6 -10), and to the polarized transport of proteins (4,11,12). Finally, it is the driving force for caveolae formation (13)(14)(15)(16)(17), actively contributing to the organization of lipids (18 -22) and signaling proteins (reviewed in Ref. 23) within these vesicular invaginations of the plasma membrane.
A wealth of biochemical and cell biological knowledge has now been gathered about the structure of caveolins. Several laboratories have validated the hypothesis that a central hydrophobic segment anchors the protein to the membrane, and splits the molecule into two cytoplasmic domains (24): first, antibodies directed against the extreme N and C termini fail to stain caveolin-1 in unpermeabilized cells (25,26). Likewise, artificial glycosylation sites introduced at the extreme N and C termini are not glycosylated (25). Third, caveolin-1 is not labeled when cells are subjected to surface biotinylation, suggesting that there is no extracellular domain (27). Finally, the N terminus of caveolin-1 undergoes tyrosine phosphorylation (1,2,28), and the C terminus of caveolin-1 undergoes palmitoylation (26), both cytoplasmic modifications.
Two other mammalian caveolins have been identified and characterized (29 -31). Caveolin-2 is expressed in many cell types that express caveolin-1 (12,32); while caveolin-3 is limited to striated muscle cells (cardiac and skeletal) and neuroglia (33,34). Caveolin-1 and caveolin-3 can each form homotypic, high molecular mass oligomers containing ϳ14 -16 individual molecules (25,27,31). In the ER, 1 homo-oligomerization is mediated by a 40-aminoacyl residue domain (residues 61-101 in caveolin-1) (27). In the Golgi, adjacent homo-oligomers undergo a second stage of oligomerization; they interact with one another through contacts between the C-terminal domains of caveolin, in a side by side packing scheme (35,36). These oligomer-oligomer interactions then produce an interlocking network of caveolin molecules that gives rise to the striated caveolar coat seen by scanning electron microscopy (3).
Although much has been learned about caveolin structure and function, the role of phosphorylation has not been defined. Constitutively active v-Src phosphorylates caveolin-1 on Tyr 14 in vitro (28). At the permissive temperature, expression of temperature-sensitive v-Src results in tyrosine phosphorylation on Tyr 14 in vivo (37). A reversible flattening and aggregation of caveolae at the cell membrane occurs as a result of temperature-sensitive v-Src activation (37,38).
In addition to these changes in caveolae morphology, we recently uncovered a novel interaction between caveolin-1 and the SH2 domain containing adapter molecule GRB7. Caveolin-1 binds GRB7 following growth factor-stimulated (and Srccatalyzed) phosphorylation of caveolin-1 tyrosine residue 14. Mutation of this residue to prevent phosphorylation undermines GRB7 association. When coexpressed, Src, Cav-1, and GRB7 dramatically increase anchorage-independent growth and cell migration (39). Downstream effector molecules remain to be identified, but this series of protein-protein interactions defines the first biological function for caveolin-1 tyrosine phosphorylation.
Here, we used a mutational approach to characterize the possible functional significance of phosphorylation at two serine residues conserved across all caveolin isotypes from all species examined to date. Mutation of Ser 80 to glutamate, a substitution that mimics chronic phosphorylation, leads to loss of caveolin-1 from caveolae membranes and tight association with endoplasmic reticulum (ER) membranes of fibroblasts. Mutation of Ser 80 to alanine neither hinders efficient targeting of caveolin-1 to caveolae, nor disrupts interaction with wildtype caveolin-1. However, in cells of the exocrine pancreas, where caveolin-1 undergoes regulated secretion, we find that mutation of Ser 80 to glutamate results in enhanced secretion, while mutation of Ser 80 to alanine blocks exocytosis completely. These findings suggest that phosphorylation of caveolin-1 on Ser 80 regulates caveolin-1 entry into the regulated secretory pathway by binding to ER membranes.
Plasmids-The canine caveolin-1 cDNA was as we described previously (41). Mutagenesis was performed using standard methods and verified by dideoxynucleotide sequencing.
Cells, Media, and Transfection Methods-Madin-Darby canine kidney epithelial cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% donor calf serum, 2 mM glutamine, 100 units/ml penicillin, 100 g/ml streptomycin. Human embryonic kidney 293T and COS-7 cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS), 2 mM glutamine, 100 units/ml penicillin, 100 g/ml streptomycin. AR42J cells were obtained from the American Type Cell Collection (CRL-1492) and were propagated in Kaighn's modified Ham's F-12 nutrient mixture (F-12K) supplemented with 20% FBS, 2 mM glutamine, 100 units/ml penicillin, 100 g/ml streptomycin. 293T cells were transfected by calcium phosphate precipitation. COS-7 cells (grown to 40% confluence in a 3.5-mm diameter dish) were transfected with Fugene 6 using 2 g of DNA and 3 l of transfection agent (diluted in 97 l of serum-free medium) as directed by the manufacturer. AR42J cells were transfected with Lipo-fectAMINE PLUS as described below.
Immunoblotting-Samples were subjected to SDS-PAGE under reducing conditions. Proteins were transferred to nitrocellulose mem-branes, and stained with Ponceau S (Sigma). Membranes were then washed in 10 mM Tris, pH 8.0, 150 mM NaCl, 0.05% Tween 20 (wash buffer), blocked in wash buffer supplemented with 2% nonfat milk and 1% bovine serum albumin (blocking buffer), incubated with primary antibody, washed again, and incubated with a secondary antibody conjugated with horseradish peroxidase. Bound IgGs were detected using a chemiluminescent substrate according to the manufacturer's instructions (Pierce). Monoclonal mouse IgG and donkey-anti mouse IgG were diluted in wash buffer supplemented with 1% bovine serum albumin. Polyclonal rabbit IgG and donkey anti-rabbit IgG were diluted in blocking buffer.
Protease Protection Assay-The topology of caveolin-1 was monitored as described (42,43). 293T cells were transfected with caveolin-1 cDNA. Thirty-six hours after transfection, cells were scraped into the culture medium, pelleted by centrifugation, resuspended in PBS, and pelleted again. All subsequent steps were performed at 4°C. Cells were then resuspended in 0.4 ml of iso-osmotic buffer (20 mM Tricine, pH 7.8, 250 mM sucrose, 0.1 mM EDTA), and homogenized by 30 passes through a 22-gauge needle. Nuclei were removed by centrifugation at 1000 ϫ g for 5 min. Membranes were then pelleted by centrifugation at 20,000 ϫ g for 10 min and resuspended in 0.4 ml of digestion buffer (100 mM sodium phosphate, pH 7.4, 150 mM NaCl, 4 mM KCl, 2 mM MgCl 2 , and FIG. 1. At steady-state phosphorylated caveolin-1 targets to caveolae. Madin-Darby canine kidney cells were labeled with [ 32 P]orthophosphoric acid. Cells were then subjected to lysis in cold 1% Triton X-100 buffer supplemented with phosphatase and protease inhibitors. The lysate was homogenized, mixed with an equal volume of 80% sucrose (4 ml total), prepared in buffer lacking detergent, and placed at the bottom of an ultra-clear centrifuge tube. The sample was overlaid with a 5-30% discontinuous sucrose gradient (4 ml each, both solutions prepared in buffer lacking detergent). The samples were then subjected to ultracentrifugation at 200,000 ϫ g for 18 h. A light scattering band was observed at the interface of the 5 and 30% sucrose solutions. Twelve 1-ml fractions were collected from the top of the gradient. Each fraction was mixed with 110 l of 100% trichloroacetic acid, and proteins were precipitated on ice for 30 min. Following centrifugation at 21,000 ϫ g for 10 min, the sucrose solutions were decanted, and precipitated proteins were resuspended by boiling in RIPA buffer (see "Experimental Procedures"). Caveolin-1 was immunoprecipitated with anti-caveolin-1 rabbit IgG N-20. A, equal volume aliquots of the resuspended proteins from the gradient were subjected to SDS-PAGE and autoradiography. Note that the majority of the incorporated [ 32 P] remained at the bottom of the gradient (fractions 9 -12). B, the caveolin-1 immunoprecipitate was subjected to SDS-PAGE, transfer to nitrocellulose, and autoradiography. C, the blot shown in B was immunoblotted with anti-caveolin-1 mouse IgG 2297. Note that caveolin-1 immunoreactivity and 32 P incorporation coincide.
FIG. 2. Ser 80 is a major site of caveolin-1 phosphorylation. A, sequence alignment of caveolin proteins. Twelve residues (boxed) are conserved in all caveolin isoforms across all species from C. elegans to H. sapiens (human). Of these invariant residues, Ser 80 and Ser 168 are the only ones that can undergo phosphorylation (marked with arrows). The homo-oligomerization domain (dashed line) spans residues 61-101, and is followed by the transmembrane domain (solid line, residues 102-134). B, 293T cells were transfected and labeled with [ 32 P]orthophosphate as detailed under "Experimental Procedures." After labeling, cells were washed and proteins were subjected to immunoprecipitation with anticaveolin-1 antibodies. Immunoprecipitated proteins were subjected to SDS-PAGE and were transferred to nitrocellulose. Autoradiography was performed to detect incorporation of [ 32 P]phosphate (upper panel). The membrane was then immunoblotted with anti-caveolin-1 antibodies to assess the level of protein present (lower panel). Note that Cav-1(S80E) showed significantly less [ 32 P]phosphate incorporation than either wild-type caveolin-1 or Cav-1(S168E). 0.02% (w/v) NaN 3 . One-hundred microliter aliquots were then incubated with or without trypsin for 30 min. Where indicated, samples were adjusted to 60 mM n-octylglucoside prior to addition of trypsin. Proteins were precipitated with trichloroacetic acid, pelleted by centrifugation, resuspended in 1% SDS, separated by SDS-PAGE, and transferred to nitrocellulose sheets.
Immunoprecipitation-Thirty-six hours post-transfection, 293T cells (grown to confluence in a 100-mm diameter dish) were washed twice with ice-cold PBS and scraped into 1 ml of IP buffer (10 mM Tris, pH 8.0, 150 mM NaCl, 1% Triton X-100, 60 mM n-octylglucoside, 0.1 mM Na 3 VO 4 , 50 mM NaF, 30 mM Na 4 P 2 O 7 , 0.1 g/ml okadaic acid supplemented with protease inhibitors). Debris was removed by centrifugation at 21,000 ϫ g for 10 min. Lysates were pre-cleared by incubation with 30 l of Protein-A Sepharose for 45 min at 4°C and then transferred to fresh tubes containing 30 l of a 1:1 slurry of Protein A-Sepharose and IP buffer. Two micrograms of IgG were added to the mixture. For coimmunoprecipitation studies, rabbit anti-GFP IgG or mouse anti-caveolin-3 IgG were used as an irrelevant control. Following a 4-h incubation at 4°C, immune complexes were collected by centrifugation, washed six times in 1 ml of IP buffer and four times with 1 ml of 10 mM Tris, pH 8.0, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA; and were disrupted by boiling in 1% SDS.
Triton X-100 Extraction, Purification of Caveolae-enriched Membranes, Hypotonic Lysis and Fractionation, and Alkaline Carbonate Extraction-Triton X-100 extraction, purification of caveolin-rich membranes, hypotonic lysis, and fractionation of proteins into soluble and particulate fractions, and extraction with alkaline sodium carbonate were all performed exactly as we described in detail previously (36).
Metabolic Labeling-[ 3 H]Palmitate incorporation was monitored as we described previously (44). Briefly, 16 h post-transfection, cells were washed twice with PBS and cultured for 1 h in labeling medium (DMEM, 5% dialyzed FBS, 5 mM sodium pyruvate). Cells were labeled for 4 h with 300 Ci of [ 3 H]palmitic acid in 2 ml of labeling medium, washed twice with PBS, and subjected to lysis in immunoprecipitation buffer. Proteins were immunoprecipitated, and separated by SDS-PAGE under nonreducing conditions. The gel was impregranted with sodium salicilate, dried, and processed for autoradiography, as we described previously (45). An aliquot of the immunoprecipitate was subjected to SDS-PAGE, transferred to nitrocellulose and detected by immunoblot analysis.
Steady state incorporation of [ 32 P]orthophosphate was monitored as follows: Madin-Darby canine kidney cells, grown to confluence in two 150-mm diameter dishes, were serum-starved for 12 h, and then cultured in phosphate-free DMEM, supplemented with 5% dialyzed FBS, 1 mCi/ml of [ 32 P]orthophosphate for 4 h. Cells were then scraped into 2 ml Cytoplasmically directed N and C termini are separated by a putative membrane spanning, hydrophobic domain. The protein undergoes palmitoylation (downward hatches) on three cysteinyl residues (Cys 133 , Cys 143 , and Cys 156 ) in the C-terminal third of the molecule. The positions and identities of the four caveolin-1 serine mutants characterized in this study are also indicated. All constructs were tagged with a c-Myc epitope at the C terminus. B, expression and detection of serine mutant caveolin-1. 293T cells were transfected with the indicated cDNAs. Lysates were prepared 36 h post-transfection. Proteins were subjected to SDS-PAGE under reducing conditions, and transferred to nitrocellulose. Immunoblotting with anti-c-Myc antibody revealed that all four mutants were expressed to comparable levels as wild-type (WT) caveolin-1.

FIG. 4. Mutation of conserved serine residues does not perturb caveolin-1 homo-oligomerization.
A, 293T cells were transfected with the indicated cDNAs. Thirty-six hours post-transfection, lysates were prepared and caveolin-1 homo-oligomerization was assessed using velocity gradient centrifugation, as described under "Experimental Procedures." Molecular weight standards were subjected to velocity gradient centrifugation in a parallel experiment, and their migration is indicated at the top of the figure. Like wild-type (WT) caveolin-1, all four mutants formed oligomers of ϳ14 -16 subunits. Blots were probed with anti-caveolin-1 rabbit IgGs (N-20). B, cells were transfected with cDNAs encoding untagged wild-type caveolin-1 and caveolin-1 cDNAs bearing a C-terminal c-Myc epitope tag, as indicated. The lysates were then subjected to immunoprecipitation with anti-c-Myc rabbit IgG A-14. Immunoprecipitates were then separated by SDS-PAGE (IP), along with one-tenth of the lysates (input), after transfer to nitrocellulose and blotted with anti-caveolin-1 mAb 2234 (67). This allowed unambiguous identification of the c-Myc epitope-tagged caveolin-1 proteins, and untagged wild-type caveolin-1. Note that untagged caveolin-1 was not precipitated by anti-c-Myc IgG unless tagged caveolin-1 was coexpressed. C, c-Myc-tagged Cav-1(S80A) or Cav-1(S80E) cDNAs were co-transfected with untagged, wild-type caveolin-1 cDNA. Lysates were prepared and subjected to immunoprecipitation as in B. As additional controls, equal fractions of the lysates were subjected to precipitation with Protein-A Sepharose alone (Protein A), and an irrelevant rabbit IgG (anti-GFP, FL). After recovery, proteins were subjected to SDS-PAGE. Note that both Cav-1(S80A) and Cav-1(S80E) correctly heterooligomerized with wild-type caveolin-1, and were not precipitated by Protein A alone, or the irrelevant IgG. of 25 mM Mes, pH 6.5, 150 mM NaCl, 1% Triton X-100, 0.1 mM Na 3 VO 4 , 50 mM NaF, 30 mM Na 4 P 2 O 7 , 0.1 g/ml okadaic acid, 100 nM calyculin A, supplemented with protease inhibitors. The lysate was homogenized and caveolae were purified by sucrose density centrifugation as described (17,36). Twelve 1-ml fractions were collected from the gradient. Proteins were precipitated with trichloroacetic acid (10% final concentration) and resuspended by boiling in RIPA buffer (50 mM Tris pH, 7.5, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS) for 1 min. Caveolin-1 was immunoprecipitated as described above using anticaveolin-1 rabbit IgG N-20. Following SDS-PAGE and transfer to nitrocellulose, proteins were detected by autoradiography and immunoblot analysis.
[ 32 P]Orthophosphate incorporation into mutated caveolin-1 was monitored as follows: 16 h post-transfection, 293T cells were washed twice with PBS and cultured for 3 h in phosphate-free DMEM supplemented with 1 mCi/ml [ 32 P]orthophosphate. Caveolin-1 was immunoprecipitated. [ 32 P]orthophosphate and protein expression were assessed as described above.
Velocity Gradient Centrifugation-Thirty-six hours post-transfection, 293T cells grown to confluence in a 10-cm dish were subjected to lysis in 1 ml of 25 mM Mes, pH 6.5, 150 mM NaCl, 60 mM octylglucoside (MBS-OG), supplemented with protease inhibitors. Two-hundred microliters of the lysate was loaded atop a 5-40% linear sucrose gradient (prepared in MBS-OG) cast in an 11 ϫ 45-mm ultraclear centrifuge tube. Samples were subjected to centrifugation at 200,000 ϫ g (44,000 rpm in Sorval rotor TH-660) for 12 h. Twelve 375-l fractions were collected and equal volume aliquots of each were subjected to SDS-PAGE and immunoblot analysis (27).
Immunofluoresence Localization-Thirty-six hours post-transfection, COS-7 cells were washed twice in phosphate-buffered saline supplemented with 1 mM MgCl 2 and 0.1 mM CaCl 2 (PBS-CM); fixed in ice-cold methanol/acetone, 1:1 (v/v), for 10 min at Ϫ20°C; and air-dried at room temperature. Cells were then rehydrated in PBS-CM. Epitopetagged caveolin-1 was detected with anti-c-Myc antibodies (mAb 9E10 or pAbs A14). Fluorescently conjugated secondary antibodies were used to visualize bound IgG on an Olympus IX70 inverted microscope. Images were obtained using a Photonics Cooled CCD camera or a Noran Oz laser scanning confocal microscope with a ϫ60 NA 1.4 objective lens. All microscopy was performed at the Analytical Imaging Facility of the Albert Einstein College of Medicine.
Cholesterol Depletion-COS-7 cells were washed twice with PBS, and then treated for 5 min at 37°C with 10 mM methyl-␤-cyclodextrin (46). After cholesterol depletion, cells were washed, fixed, and processed for immunofluorescence microscopy.
Caveolin-1 Secretion Assay-Caveolin-1 secretion by cells of the exocrine pancreas was monitored as described (47), with minor modifications. AR42J rat pancreatic adenocarcinoma cells grown to 60% confluence in a 100-mm diameter plate were transfected with LipofectAMINE PLUS as follows: caveolin-1 cDNA (6 g of DNA) was mixed with 20 l of PLUS reagent and 30 l LipofectAMINE in a final volume of 1.5 ml of Opti-MEM, exactly as suggested by the manufacturer. Cells were washed twice with PBS; the medium was replaced with 5 ml of Opti-MEM; and the DNA-lipid complex was added. After 3 h, cells were washed twice in PBS, and cultured for an additional 33 h in F-12K, 20% FBS. Cells were washed again, cultured in 2.5 ml of F-12K, 20% FBS, 4 M leupeptin, 10 g/ml soybean trypsin inhibitor, 500 M benzamidine, 5 g/ml pepstatin A, and were treated with 100 nM dexamethasone (or vehicle, ethanol) for 12 h. The medium was collected, adjusted to pH 7.9 with HEPES (50 mM final concentration). Triton X-100 was added (1% final concentration) and caveolin-1 was immunoprecipitated as described above using mAb 2234, and detected by immunoblotting with rabbit anti-c-Myc IgG A14 or mouse anti-c-Myc IgG 9E10.

Phosphorylated Caveolin-1 Targets to Caveolae at Steady
State-Caveolin-1 is phosphorylated on serine and tyrosine residues. Phosphoamino acid analysis reveals that the ratio of phosphoserine to phosphotyrosine is 10:1 (48). The role of caveolin-1 tyrosine phosphorylation in signal transduction and subcellular localization has been examined recently (39). However, the role of serine phosphorylation in caveolin-1 function has not been studied.
We performed metabolic labeling of Madin-Darby canine kidney epithelial cells to assess the steady state distribution of phosphorylated caveolin-1. Cells were labeled with [ 32 P]orthophosphoric acid, cultured in complete medium, and then subjected to lysis in cold Triton X-100 buffer. The lysate was then fractionated in an established sucrose density gradient that separates caveolae membranes from the bulk of cellular proteins (41,49). Fig. 1 shows that the bulk of the 32 P label remained at the bottom of the gradient (loading zone fractions 9-12). In contrast caveolin-1 immunoreactivity and [ 32 P]orthophosphate incorporation were detected in the low density fraction (fraction 5) enriched in purified caveolae. Since caveolin-1 is phosphorylated on serine and tyrosine residues in a ratio of 10:1 (48), caveolin-1 phosphate incorporation at steady state can be attributed largely to serine phosphorylation.
Mutation of Ser 80 Blocks Caveolin-1 Phosphorylation-Inspection of the caveolin family of primary sequences reveals that only 12 residues are conserved across all isoforms from every species examined to date (i.e. from Caenorhabditis elegans to Homo sapiens) (50). Of these 12 invariant residues, only Ser 80 and Ser 168 (in caveolin-1) could serve as potential phosphorylation sites (Fig. 2A). Thus, we mutated Ser 80 and Ser 168 individually to glutamate (a substitution that mimics chronic phosphorylation), and compared the amount of [ 32 P]orthophosphate incorporated by the mutants to the amount incorporated by wild-type caveolin-1. Fig. 2B shows   FIG. 5. Cav-1(S80E) is partially soluble in Triton X-100 and is excluded from purified caveolae. 293T cells were transfected with the indicated cDNAs. A, 36 h post-transfection, cells were extracted with an ice-cold solution of 1% Triton X-100. The extractable fraction was collected and the inextractable fraction was resuspended in the same volume of 1% SDS. Equal volume aliquots of both fractions were separated by SDS-PAGE. Proteins were transferred to nitrocellulose, and sheets were immunoblotted with anti-c-Myc antibodies. Note that Cav-1(S80E) was partially soluble in Triton X-100, with roughly half the protein in the soluble fraction (S), while wild-type (WT) caveolin-1, Cav-1(S80A), Cav-1(S168A), and Cav-1(S168E) were insoluble, with Ͼ95% of each protein in the insoluble fraction (I). B, 36 h post-transfection, caveolin-rich, detergent-resistant membranes were isolated by floatation in a sucrose gradient. Equal masses of protein from each fraction were separated by SDS-PAGE, transferred to nitrocellulsoe, and subjected to immunoblot analysis. Note that wild-type caveolin-1, Cav-1(S80A), Cav-1(S168A), and Cav-1(S168E) were enriched in low density fractions (5 and 6, arrows), while Cav-1(S80E) was preferentially excluded.
Expression and Characterization of Serine Mutant Caveolin-1 Proteins-To systematically assess the role of caveolin-1 phosphorylation on either of the invariant serine residues, we next prepared two more mutants that mimic chronic dephosphoryaltion (i.e. Cav-1(S80A) and Cav-1(S168A)). All four mutants were expressed in HEK 293T cells (Fig. 3) to comparable levels indicating that mutation of these conserved residues did not alter protein stability.
Caveolin-1 forms homotypic oligomers of 14 -16 monomers (25,27). Homo-oligomerization is mediated by a 40-aminoacyl residue domain in the N terminus, i.e. residues 61-101 in caveolin-1 (27). Also, the C-terminal domain mediates oligomer-oligomer interaction in an isoform-specific manner (35). Therefore, we assessed whether mutation of the conserved serine residues would perturb oligomerization. We were particularly concerned that because Ser 80 resides in the homooligomerization domain, mutation of this residue might disrupt oligomerization, as occurs when the oligomerization domain is deleted (17,35), or subjected to penta-alanine scanning mutagenesis (51). Conversely, formation of inappropriately large aggregates would indicate that either Ser 80 or Ser 168 is critical to the structural integrity of caveolin-1 and mutation to another residue would grossly compromise this integrity. This latter concern arises from our recent characterization of the two limb-girdle muscular dystrophy (type 1C) caveolin-3 proteins; both mutant molecules reside in the Golgi as high molecular weight aggregates that undergo rapid degradation (44).
Importantly, Fig. 4A shows that all four serine mutant caveolins form oligomers of the appropriate size. Thus, mutation of Ser 80 or Ser 168 does not cause global derangements in caveolin-1 structure, as manifested by changes in the size of the caveolin-1 oligomer. Thus, any differences observed in the localization or function of these mutant caveolin-1 molecules may be attributed reasonably to mimicry of chronic phosphorylation or dephosphorylation.
To assess the effect of mutation of Ser 80 to either Ala or Glu on hetero-oligomerization with wild-type caveolin-1, untagged wild-type caveolin-1 was coexpressed with Cav-1(S80A) or Cav-1(S80E). Cells were subjected to lysis, and proteins were immunoprecipitated with anti-c-Myc rabbit IgGs. After SDS-PAGE and transfer to nitrocellulose, blots were probed with mouse anti-caveolin-1 IgG 2234 to unambiguously identify FIG. 6. Immunolocalization of serine mutant caveolin-1. COS-7 cells were transfected with the indicated cDNAs, fixed, and stained with anti-c-Myc antibody; note that vector alone transfectants showed no staining. Wild-type caveolin-1 Cav-1(S80A), Cav-1(S168A), and Cav-1(S168E) expressing cells had the expected punctate pattern of staining. However, Cav-1(S80E) transfectants had a reticular pattern of staining throughout the cytoplasm and a strong perinuclear staining pattern. tagged and untagged caveolin-1 proteins (Fig. 4B). Fig. 4C shows that both Cav-1(S80A) and Cav-1(S80E) hetero-oligomerize with wild-type caveolin-1. This result further supports the contention that mutation of this invariant residue to either Ala or Glu does not result in gross disruption of the protein's structure.
To determine where Cav-1(S80E) resides in cells, we transfected COS-7 cells with wild-type and serine mutant caveolin-1 cDNAs and performed immunofluorescence microscopy. This cell line, like 293T and other cancer cell lines, shows little or no expression caveolin-1 endogenously, and was selected for these studies because it adheres tightly to coverslips, whereas the 293T line is less adherent (36). Fig. 6 illustrates that, like wild-type caveolin-1, Cav-1(S80A), Cav-1(S168A), and Cav-1(S168E) have punctate distribution patterns suggestive of concentration in vesicles at or near the plasma membrane. Interestingly, Cav-1(S80E) appears to be distributed in a retic-

FIG. 8. Cav-1(S80E) is tightly associated with membranes. A,
36 h post-transfection, 293T cells were subjected to lysis in hypotonic buffer. Nuclei were discarded by low speed centrifugation and postnuclear lysates were separated into soluble (S) and pellet (P) fractions by high-speed centrifugation. The soluble fraction was collected and the pellet was resuspended in the same volume of 1% SDS. Equal volume aliquots of both fractions were separated by SDS-PAGE. Proteins were transferred to nitrocellulose, and sheets were immunoblotted with antic-Myc antibodies. Like wild-type (WT) caveolin-1, all four caveolin-1 mutants were found in the pellet. As internal controls, we verified that a representative cytosolic protein (GDP dissociation inhibitor, GDI), and a glycosylated membrane spanning protein (lysosome associate membrane protein-1; LAMP-1) were found in the expected fractions. B, 36 h post-transfection, 293T cells were washed twice in PBS, and once in 150 mM NaCl. Cells were then scraped into 100 mM Na 2 CO 3 , pH 11.5. Following Dounce homogenization, proteins were fractionated into extractable (E) and inextractable (I) components. The extractable fraction was collected and the inextractable fraction was resuspended in the same volume of 1% SDS. Equal volume aliquots of both fractions were separated by SDS-PAGE. Proteins were transferred to nitrocellulose, and detected with anti-c-Myc antibodies. Like wild-type caveolin-1 (WT), all four mutant caveolin-1 proteins resisted solubilization by Na 2 CO 3 , whereas the peripherally associated membrane protein EEA1 was completely solubilized.
FIG . 9. Cav-1(S80E) has the same membrane topology as wildtype caveolin-1. Thirty-six hours after transfection, cells were collected in PBS, and subjected to iso-osmotic homogenization and fractionation as described under "Experimental Procedures." A, the membrane fraction was treated with trypsin and octylglucoside as indicated. Samples were split in two after digestion, and equal volume aliquots for each were separated by SDS-PAGE, transferred to nitrocellulose. Blots were probed with antibodies against the N-terminal 20 residues of caveolin (rabbit IgG N-20) and against the C-terminal c-Myc epitope tag (mouse IgG 9E10). Note that both the N-and C-terminal domains of caveolin-1 are sensitive to protease digestion. Like wild-type caveolin-1, Cav-1(S80E) was sensitive to tryptic digestion, indicating that both the N and C termini of Cav-1(S80E) are directed toward the cytoplasm. To confirm the integrity of the membrane preparation, blots were probed with antibodies against the ER luminal protein proteindisulfide isomerase. As expected, protein-disulfide isomerase was only sensitive to protease digestion when membranes were solubilized with octylglucoside.
The subcellular location of caveolin-1 is sensitive to cellular cholesterol levels (54). Indeed, depletion of cellular cholesterol results in relocation of caveolar caveolin-1 to the lumen of the ER (42,55). This pharmacological manipulation reflects the underlying physiologic role of caveolin-1 as an intracellular shuttle of cholesterol (7-9). Since Cav-1(S80E) colocalized with an ER marker, we used an established fractionation scheme (17) to determine whether the protein is soluble, or membranebound. Fig. 8A shows that nearly all of wild-type caveolin-1 is in the particulate fraction. Likewise, Cav-1(S80E) partitions in the pellet fraction. As internal controls to validate our fractionation method, blots were probed with antibodies against a cytosolic protein (GDP dissociation inhibitor), and a veritable integral membrane protein (lysosome associate membrane protein-1). Both control proteins were detected in the appropriate fractions, with GDP dissociation inhibitor partitioning in the supernatant (56), and lysosome-associated membrane protein-1 partitioning in the pellet (57).
To determine whether Cav-1(S80E) was tightly associated with membranes, cells were extracted with alkaline sodium carbonate, an agent that strips membranes of peripherally associated proteins, but not plasma membrane-bound caveolin-1 (11,17,56). Fig. 8B shows that like wild-type caveolin-1, Cav-1(S80E) resisted alkaline sodium carbonate extraction. As a control, we verified that the peripherally associated membrane protein early endosomal antigen 1 is solubilized completely by alkaline sodium carbonate (58).
ER Membrane-bound Cav-1(S80E) Has the Same Membrane Topology as Plasma Membrane-bound Wild-type Caveolin-1, but Undergoes Less Palmitoylation-Our fractionation results indicated that Cav-1(S80E) was not a soluble resident of the ER. Rather it was tightly associated with the ER membrane. Plasma-membrane bound caveolin-1 is believed to assume an unusual conformation with cytosolically directed N and C termini separated by a hydrophobic domain that is believed to anchor the protein to the membrane (11,24,26). Both the Nand C-terminal domains undergo post-translational modifications, with Tyr 14 (N terminus) undergoing phosphorylation (28) and three C-terminal cysteinyl residues undergoing thiopalmitoylation (26).
We used an established protease protection assay to determine the membrane topology of Cav-1(S80E) (42,43). Following isotonic lysis and homogenization, caveolin-1 is sensitive to tryptic digestion, as the N and C termini are directed toward the cytoplasm, and there is no ecto-domain (Fig. 9). However, the protein enters the lumen of the ER in response to cholesterol depletion, and is thus protected from tryptic digestion in this setting (42). Since Cav-1(S80E) was membrane-bound, we predicted it would remain sensitive to tryptic digestion, provided that it had the same topology as plasma membranebound caveolin-1. Fig. 9 shows that like wild-type caveolin-1, Cav-1(S80E) is sensitive to tryptic digestion: neither antibodies directed against the caveolin-1 N terminus (N-20), nor a Cterminal c-Myc epitope tag (9E10) detected caveolin-1 in protease-treated samples. The integrity of the membrane preparations was verified by blotting for a luminal ER protein (protein-disulfide isomerase) which was resistant to protease protection until membranes were dissolved with octylglucoside.
Finally, we determined if the C-terminal cysteinyl residues of Cav-1(S80E) are thio-acylated. To this end, cells were labeled with tritiated palmitic acid. After detergent lysis, caveolin-1 was recovered by immunoprecipitation and palmitate incorpo-ration was assessed by autoradiography. Fig. 10 shows that Cav-1(S80E) is palmitoylated to significantly lower levels than wild-type caveolin-1. Cav-1(S80A), Cav-1(S168A), and Cav-1(S168E) are palmitoylated to the same extent as wild-type caveolin-1. Although these results are uniformative regarding topology, they suggest that palmitoylation occurs once caveolin-1 has traversed the ER. Indeed, both Golgi-retained limbgirdle muscular dystrophy 1C caveolin-3 mutants are palmitoylated to the same extent as wild-type caveolin-3 (44). This suggests that palmitoylation of caveolins occurs after exit from the ER, at the level of the Golgi.
Although Cav-1(S80E) was bound to ER membranes, and was not in the lumen of this compartment, we suspected that phosphorylation of caveolin-1 on this residue might be a regulatory switch for targeting to the ER in response to cholesterol depletion. However, immunofluorescence microscopy revealed that the overwhelming majority of Cav-1(S80A) colocalized with calnexin when cells were treated with the cholesterol depleting agent methyl-␤-cyclodextrin (Fig. 11), suggesting that phosphorylation of Ser 80 does not govern this aspect of the caveolin-1 intracellular traffic (i.e. ER retrieval is not regulated by phosphorylation of Ser 80 ).
Caveolin-1 Phosphorylation on Ser 80 Regulates Entry into the Regulated Secretory Pathway of Pancreatic Acinar Cells-Recently, Anderson and colleagues (47) reported that caveolin-1 is secreted by cells of the exocrine pancreas after mice are treated with the secretagogues dexamethasone, cholecystekinin, or secretin (47). Furthermore, recombinant expression of caveolin-1 in the rat pancreatic adenocarcinoma cell line AR42J recapitulates the regulated secretion seen in whole animals (47). Although the function of secreted caveolin-1 remains elusive, understanding the regulation of this process may aid in unraveling its purpose.
We verified that AR42J cells secrete wild-type caveolin-1 when treated with dexamethasone, but not with vehicle alone FIG. 11. Phosphorylation of caveolin-1 on Ser 80 is not required for caveolin-1 retrieval to the endoplasmic reticulum during cholesterol depletion. COS-7 cells were transfected with the indicated caveolin-1 cDNAs. Thirty-six hours after transfection, cells were washed twice with PBS and treated with serum-free medium supplemented with 10 mM methyl-␤-cyclodextrin (M␤CD). After a 5-min incubation at 37°C, cells were processed for immunofluorescence microscopy. Note that both wild-type caveolin-1 and Cav-1(S80A) colocalized with the ER marker calnexin after treatment with M␤CD. (Fig. 12). To assess whether serine phosphorylation of caveolin-1 influences secretion in this cell system, we expressed Cav-1(S80A) and Cav-1(S80E) in AR42J cells. Twelve hours after administration of dexamethasone, the cell culture medium was collected and subjected to immunoprecipitation with anti-caveolin-1 antibodies. Total cellular proteins were collected as well. Fig. 12B shows that Cav-1(S80A) was not secreted, while wild-type caveolin-1 and Cav-1(S80E) were, suggesting that serine phosphorylation of caveolin-1 is required for secretion. Indeed, 2-2.5 times as much Cav-1(S80E) was recovered from the cell culture supernatant than wild-type caveolin-1. Conversely, more wild-type caveolin-1 was found in the intracellular pool than Cav-1(S80E) (Fig. 12B, lower panel). Cav-1(S168A) and Cav-1(S168E) were secreted to the same level as wild-type caveolin-1 (Fig. 12D).
To assess whether mimicking chronic phosphorylation of caveolin-1 on Ser 80 is sufficient to drive secretion, we compared the amount of Cav-1(S80E) secreted in the presence or absence of dexamethasone. Fig. 12C shows that like wild-type caveolin-1, Cav-1(S80E) was not detected in the cell culture medium in the absence of dexamethasone. Because dexamethasonedriven reorganization of internal membranes is required to induce the secretory phenotype of AR42J cells (59), this finding does not exclude the possibility that in normal exocrine pan-FIG. 12. Caveolin-1 phosphorylation on Ser 80 is necessary for caveolin-1 secretion in AR42J Cells. To assess the possible role of serine phosphorylation in caveolin-1 secretion, we transfected AR42J cells with the indicated cDNAs. Thirty-six hours post-transfection, cells were treated with 100 nM dexamethasone. Twelve hours later, the cell culture medium was collected, and immunoprecipitated with anti-caveolin-1 IgG 2234, which recognizes an epitope in the first 20 aminoacyl residues (67). Cellular proteins were recovered in 1% SDS. A, blots were probed with anti-c-Myc epitope rabbit IgGs (A14) to verify that full-length protein was recovered. Note that wild-type caveolin-1 was only detected in the extracellular compartment when cells were treated with dexamethasone. B, cells were transfected with the indicated caveolin-1 cDNAs and treated with dexamethasone as described above. Note that wild-type caveolin-1 and Cav-1(S80E) were detected in both the intra-and extracellular compartments, while Cav-1(S80A) was completely intracellular. Band densitometry was performed with Chemimager software and the results are shown below the secreted caveolin panel. Cellular proteins ("Cell"), and a trichloroacetic acid precipitate of the PBS used to wash the cells ("Wash"), after collection of the medium, were run in parallel to demonstrate that the intracellular caveolin-1 was not contaminated with extracellular protein. Extracellular caveolin-1 was detected with anti-c-Myc mouse IgG (9E10), and the light chain of the precipitating anti-caveolin-1 IgG (2234) was therefore detected. C, as with wild-type caveolin-1 (see panel A), Cav-1(S80E) was only secreted when cells were treated with dexamethasone. D, like wild-type caveolin-1, Cav-1(168A) and Cav-1(S168E) were secreted when AR42J cells were treated with dexamethasone.  creatic cells, caveolin-1 phosphorylation on Ser 80 is sufficient to target caveolin-1 to the exocytic pathway.

DISCUSSION
Because caveolin-1 is a multifunctional protein that moves among membrane compartments, and is even found in complex with cytosolic chaperone proteins (8), the molecular events associated with caveolin-1 localization have been studied intensely for nearly a decade (2). Cell density (24,60,61) and cholesterol levels (7-9, 42, 55) are the best characterized external determinants of caveolin-1 subcellular localization. However, an emerging body of work indicates that caveolin-1 has localization signals in its primary sequence that govern its subcellular localization (17,36,51,62). Superimposed on these determinants are reversible modifications like phosphorylation that might alter the protein's position and function in the cell (37,38).
What information does the sequence of caveolin-1 itself contain regarding the protein's subcellular traffic? We recently demonstrated that the caveolin scaffolding domain (residues 82-101 of caveolin-1) is necessary for caveolin-1 targeting to caveolae (17); and is sufficient to target a heterologous soluble protein to these membrane microdomains (36). We also found that caveolin-1 contains a trans-Golgi localization signal in its C-terminal domain (36). The steady state localization of caveolin-1 reflects, in part, these two cues. The role of phosphorylation is less clear, but Src-dependent tyrosine phosphorylation of caveolin-1 causes a reversible flattening and aggregation of caveolae at the cell membrane (37,38).
Here, we have taken a mutational approach to determine the role of serine phosphorylation in the subcellular traffic of caveolin-1. We find that mimicking chronic phosphorylation of invariant serine residue 80 by mutating it to glutamate results in exclusion from caveolae membranes, and tight association with ER membranes in fibroblasts (293T and COS-7). Importantly, this mutation does not disrupt homo-oligomerization, or hetero-oligomerization with wild-type caveolin-1. In contrast, mutation of this invariant serine residue to alanine, mimicking chronic dephosphorylation, does not alter the subcellular localization of caveolin-1 in these cell types. Table I summarizes the properties of the mutant caveolin-1 proteins examined in this study.
Finally, the phenotypes of Cav-1(S168A) and Cav-1(S168E) were indistinguishable from that of wild-type caveolin-1. Since Ser 168 is an invariant residue, it is unlikely that it is insensitive to mutation. The C-terminal domain of caveolin-1 mediates oligomer-oligomer interaction (35) and we recently demonstrated that the final 10 residues of caveolin-1 are required for oligomer-oligomer interaction. This stretch of amino acid residues includes Ser 168 , and mutation of this residue to either Ala or Glu hinders GFP-fused Cav-1(135-178) binding to wild-type caveolin-1. Thus, this invariant residue serves an important structural role that is only revealed in the context of the isolated C terminus (36).
Since Cav-1(S80E) did not show complete loss of phosphate incorporation (Fig. 2B), there are likely other sites of serine phosphorylation that contribute to the subcellular localization of caveolin-1. Our finding that phosphorylated caveolin-1 resides in caveolae of epithelial cells at steady state ( Fig. 1) supports this hypothesis. Combinations of serine mutants may be required to dissect the hierarchy of signals governing subcellular localization. We also note that mimicry of serine phosphorylation by mutation to glutamate may not be completely equivalent to sustained phosphorylation; protein-protein interactions dependent on phosphoserine may be lost in mutating the serine residue to glutamate. Alternatively, we cannot exclude the possibility that the S80E mutation does not ade-quately mimic the phosphorylated form, but rather is inducing a novel phenotype.
Serine Phosphorylation Regulates Caveolin-1 Entry into the Secretory Pathway by Targeting to the ER-In professional secretory cells of the exocrine pancreas, caveolin-1 undergoes regulated exocytosis, with secretion being stimulated by a panel of secretagogues (47). Here, we found that mutation of Ser 80 to Glu increases the amount of caveolin-1 protein secreted in response to dexamethasone. Conversely, mutation of Ser 80 to Ala blocks secretion completely.
The opposing effects of mutating Ser 80 to either Ala or Glu on caveolin-1 secretion are reminiscent of those exerted by similar mutations of the polymeric immunoglobulin receptor (pIgR) at Ser 664 . pIgR transcytosis in polarized epithelia is regulated by phosphorylation of Ser 664 ; and pIgR(S664A) undergoes transcytosis at a slower rate than wild-type pIgR, while pIgR(S664D) shows an increased transcytotic rate compared with wild-type pIgR (63). These results reflect an increase (S664A) or decrease (S664D) in the rate of pIgR recycling to the basolateral surface. Thus, serine phosphorylation of membrane proximal Ser residues may serve as a general signal for targeting proteins to distinct membrane compartments. Other signals like tyrosine phosphorylation by Src family kinases (37,64) and homo-oligomerization (44,51,65) regulate caveolin-1 and pIgR subcellular traffic as well.
Although studies with pIgR establish a precedent for serine phosphorylation mediated regulation of membrane protein trafficking, the point of caveolin-1 entry into the regulated secretory pathway contrasts with the route taken by most FIG. 13. A binary switch gaits caveolin-1 entry into the regulated secretory pathway. In exocrine cells, phosphorylation of caveolin-1 invariant residue Ser 80 is required for targeting of the protein to the ER lumen. Blocking this phosphorylation event by mutation to alanine prevents ER targeting and subsequent secretion. Conversely, mimicking phosphorylation by mutation to glutamate increases regulated exocytosis.
proteins that undergo regulated exocytosis in endocrine and exocrine cells. Typically, such proteins are packaged into nascent secretory granules derived from the trans-Golgi network where they undergo proteolytic processing and aggregation (66). Why is caveolin-1 entry into the regulated secretory pathway different? First, conventional sorting mechanisms like proteolytic processing in the trans-Golgi network to generate cargo for immature secretory granules cannot be used to target caveolin-1 to the regulated secretory pathway: full-length caveolin-1, not a proteolytic fragment thereof, undergoes regulated exocytosis. Second, caveolin-1 is usually an oligomeric protein bound to the cytoplasmic face of the plasma membrane, whereas secreted products are soluble proteins. Thus, the professional secretory cell can control caveolin-1 entry into the regulated secretory pathway by gating its traffic across the ER membrane (Fig. 13).