Caveolin-1 Interacts with 5-HT 2A Serotonin Receptors and Profoundly Modulates the Signaling of Selected G (cid:1) q -coupled Protein Receptors*

5-Hydroxytryptamine 2A (5-HT 2A ) serotonin receptors are important for a variety of functions including vascular smooth muscle contraction, platelet aggregation, and the modulation of perception, cognition, and emo-tion. In a search for 5-HT 2A receptor-interacting pro- teins, we discovered that caveolin-1 (Cav-1), a scaffold-ing protein enriched in caveolae, complexes with 5-HT 2A receptors in a number of cell types including C6 glioma cells, transfected HEK-293 cells, and rat brain synaptic membrane preparations. To address the functional significance of this interaction, we performed RNA inter-ference-mediated knockdown of Cav-1 in C6 glioma cells, a cell type that endogenously expresses both 5-HT 2A receptors and Cav-1. We discovered that the in vitro knockdown of Cav-1 in C6 glioma cells nearly abolished 5-HT 2A receptor-mediated signal transduction as measured by calcium flux assays. RNA interference-me-diated knockdown of Cav-1 also greatly attenuated endogenous G (cid:1) q -coupled P2Y purinergic receptor-medi- ated signaling without altering the signaling of PAR-1 thrombin receptors. Cav-1 appeared to modulate 5-HT 2A signaling by facilitating the interaction of 5-HT 2A receptors Transfection of Cells Cells— kidney encoding

enously expressed 5-HT 2A receptors associate with Cav-1 in C6 glioma cells and rat brain synaptic membrane preparations. We also report that the RNAi-mediated knockdown of Cav-1 has profound functional consequences for 5-HT 2A -mediated signal transduction in C6-glioma cells, a cell type that endogenously expresses both Cav-1 and 5-HT 2A receptors. Most importantly, screening of C6 glioma cells for other G␣ q -coupled receptors showed that knockdown of Cav-1 modulated the signaling of purinergic receptors but not of PAR-1 thrombin receptors in C6 glioma cells, suggesting that only a subset of G␣ q -coupled receptors is regulated by Cav-1.

EXPERIMENTAL PROCEDURES
cDNA Constructs and Reagents-The construct coding for FLAGtagged wild type (wt) rat 5-HT 2A receptor (FLAG-5-HT 2A ) (12) was modified by addition of an amino-terminal and cleavable signal peptide sequence (MKTIIALSYIFCLVFA; see Ref. 23) from influenza hemagglutinin and has been described elsewhere (24). The caveolin-1-myc cDNA was a gift from Gary Landreth (Case Western Reserve University), and the constitutively active G␣ q mutant (Q229L) was from David Siderovski (University of North Carolina, Chapel Hill). Thrombin receptor-activating peptide (TRAP) was a gift from Paul Di Corleto (Lerner Research Institute, Case Western Reserve University). All constructs containing inserts in the appropriate orientation were verified Transfection of HEK-293 Cells and C6 Glioma Cells-Human embryonic kidney 293 (HEK-293) cells and C6 glioma cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 1 mM sodium pyruvate, penicillin (100 units/ml), and streptomycin (100 mg/ml) (Invitrogen) at 37°C and 5% CO 2 . Transient transfection of HEK-293 cells with FuGENE 6™ (Roche Applied Science) using a FuGENE to DNA ratio of 5:1 was performed according to the manufacturer's recommendations. A total of 6 g of DNA was used in each co-transfection, 3 g of which were FLAG-5-HT 2A receptor or constitutively active G-protein (G␣ q *) and 3 g of which were DNA encoding cav-1-myc or GFP. In case of triple transfections with FLAG-5-HT 2A receptor, cav-1-myc, G␣ q , and/or GFP, 2 g of each plasmid DNA was used. C6 glioma cells were transfected using Lipo-fectAMINE 2000 (Invitrogen). A total of 24 g of DNA was used with 60 l of LipofectAMINE as recommended by the manufacturer.
Immunocytochemistry, Agonist-mediated Internalization, and Image Quantification-Dual-labeling immunocytochemistry was performed essentially as described before (12), and images of cells treated with vehicle and or agonist 5-HT (10 M) were acquired digitally using a Zeiss 410 confocal microscope (Oberkochen, Germany) without saturation of pixel intensities. Images were subsequently analyzed for percent receptor internalization using MetaView software (Universal Imaging) as detailed previously (24).
Co-immunoprecipitation and Immunoblot-Co-immunoprecipitation and immunoblotting were performed essentially as detailed previously (24,25). For visualization and immunoprecipitation of Cav-1, a polyclonal Cav-1 antibody (1:2000) (BD Transduction Laboratories) was used, and a monoclonal M2-anti-FLAG antibody preconjugated to agarose beads was used to immunoprecipitate FLAG-tagged 5-HT 2A receptors. Wheat germ agglutinin (WGA) preconjugated to agarose beads (Sigma) was used to concentrate native 5-HT 2A receptors from C6 glioma cells after solubilization in 1.5% CHAPS. A rabbit polyclonal FLAG antibody (1:1500) (Sigma) was used to detect FLAG-5-HT 2A receptors. A monoclonal mouse Cav-1 antibody and protein A/G-agarose were used to co-immunoprecipitate native receptor and Cav-1 from C6 glioma cells and rat brain synaptic membrane preparation. Rat brain synaptic membranes were prepared from rat frontal cortex as described previously (26). The native 5-HT 2A receptor was detected by using a polyclonal 5-HT 2A carboxyl-terminal antibody (27), a gift from Jon Backstrom (Vanderbilt University). The G␣ q in immunoprecipitates, the constitutively active G␣ q * mutant, and the endogenous wt G␣ q in cell lysates were detected by a rabbit polyclonal G␣ q antibody (1:2000) (Santa Cruz Biotechnology, Santa Cruz, CA). The immunoblots shown are representative of at least three independent experiments.
RNAi-mediated Knockdown of Cav-1-A DNA vector-based siRNAi was designed to stably knock down the expression of Cav-1 in C6 glioma cells by Genscript. Briefly, a commercial algorithm for designing a specific siRNA was used (www.genscript.com/ssl-bin/app/rnai). The length of the siRNA target, GC % range, and the rat cDNA sequence of Cav-1 were entered into the algorithm and potential siRNA targets For these experiments, HEK-293 cells were either untransfected (1st and 2nd lanes) or transiently co-transfected with Cav-1 (3rd and 4th lanes), FLAG-5-HT 2A ϩ GFP (5th and 6th lanes; treated with vehicle), FLAG-5-HT 2A ϩ GFP (7th and 8th lanes; treated with 10 M 5-HT for 5 min), FLAG-5-HT 2A ϩ Cav-1 (9th and 10th lanes; treated with vehicle), and FLAG-5-HT 2A ϩ Cav-1 (11th and 12th lanes; treated with 10 M 5-HT for 5 min). FLAG-tagged native 5-HT 2A receptors were immunoprecipitated (IP) by a monoclonal FLAG antibody conjugated to Sepharose beads. A polyclonal Cav-1 antibody detected Cav-1myc, and a polyclonal FLAG antibody on Western blots detected FLAG-5-HT 2A . A polyclonal G␣ q antibody was used to probe for G-protein as lysate control. Representative immunoblots (IB) from a single experiment that has been replicated three times with equivalent results are shown. A and B, immunoblots from immunoprecipitates. C and D, immunoblots from cell lysates.  (ϩ)). The 3rd and 4th panels from the top represents the immunoblots used to detect Cav-1 and 5-HT 2A receptor proteins, respectively, in the total C6 cell lysate. C, 5-HT 2A receptors were immunoprecipitated with Cav-1 from rat brain synaptic membrane preparations. Top panel shows 5-HT 2A receptors were co-immunoprecipitated with Cav-1 antibody and protein A/G-agarose (2nd lane) but not with protein A/G-agarose beads alone (Ϫ). The 2nd panel from the top shows that Cav-1 was found in the immunoprecipitates (lane 2). The 3rd and 4th panels from the top shows immunoblots (IB) used to detect Cav-1 and 5-HT 2A receptors, respectively, from total membrane lysates. Shown are representative immunoblots from a single experiment that has been replicated three times with equivalent results. generated. The ranking of siRNA candidates was based on the commercial algorithm. The target sequence CACACAGTTTCGACGGCATCT (52.38% GC content) was cloned into pRNA-U6.1/neo under the control of U6 promoter and neomycin selection marker, which was used to select stable lines. This insert-containing vector was then transfected into C6 glioma cells, as described above, and the subsequent isolation of Cav-1-depleted C6 glioma cells was done after selection of stable individual clones by using 0.6 mg of geneticin/liter of cell culture medium. Forty eight clones were individually picked and expanded in selection medium. After passaging the cells a minimum of three times, lysates were prepared as described above. To evaluate the stable knockdown of expression of Cav-1, 10 g of protein lysates were subjected to SDS-PAGE followed by Western blot analysis probing for Cav-1. The clones, which showed maximal knockdown, were then maintained in the selection media containing 0.6 mg of geneticin/liter for further study. Lysates prepared from the clone with maximal knockdown of Cav-1 was also probed with rabbit polyclonal RSK1 (1:1000) (Santa Cruz Biotechnology), mouse monoclonal Cav-2 (1:500) (BD Transduction Laboratories), and polyclonal rabbit G␣ q antibodies.
Microarray Analysis of C6 Glioma Cells-Nearly confluent plates of C6 cells were harvested using RNase-free conditions, and a sample preparation was done by the Gene Expression Array Core Facility at Case Western Reserve University (www.geacf.net), following methods recommended by Affymetrix (see Supplemental Material for full details). Briefly, total RNA was extracted using Trizol (Invitrogen) followed by RNA clean up using Qiagen columns that were used to clean up cDNA and/or RNA as required using the manufacturer's protocol. This was followed by cDNA synthesis using an oligo(dT) primer coupled to T7 RNA polymerase promoter. Reverse transcription of RNA was done using Superscript II reverse transcriptase in a 20-l reaction at 42°C for 1 h. Second strand synthesis was carried out immediately in the presence of Escherichia coli DNA polymerase I, RNase H, and DNA ligase. The reaction mixture was incubated for 2 h at 16°C and an additional 5 min in the presence of T4 DNA polymerase. The reaction was terminated by addition of EDTA followed by cDNA clean up using Qiagen columns and storage overnight at Ϫ20°C. In vitro transcription was used to generate cRNA by using a Bioarray High Yield ENZOkit (Affymetrix) followed by clean up of RNA samples. Purified in vitro transcribed cRNA was subjected to fragmentation at 94°C for 35 min using 1ϫ fragmentation buffer (40 mM Tris acetate, pH 8.1, 100 mM KOAc, 30 mM MgOAc) and then placed on ice. Preconditioning for hybridization was done in 1ϫ hybridization mixture (100 mM MES,1 M [Na ϩ ], 20 mM EDTA,0.01% Tween 20). Herring sperm and acetylated bovine serum albumin were added to a final concentration of 0.1 and 0.5 mg/ml, respectively. A 15-l aliquot of a 20ϫ mixture of in vitro transcripts of bacterial genes bioB, bioC, bioD, and cre were added to the mixture to give final concentrations of 1.5, 5, 25, and 100 pM, respectively. Control oligonucleotide was added to a final concentration of 50 pM. The amount of fragmentation reaction containing 15 g of cRNA was added to the mixture, and the remaining volume was made up with molecular biology grade water. Preconditioning of the array chip was done in 1ϫ hybridization buffer for 10 -15 min at 45°C with rotation (45 rpm). The preconditioning buffer was then removed from the chip chamber. Sample hybridization mixture was added to the chip and hybridized overnight (16 h) at 45°C with rotation (45 rpm). For posthybridization washing and staining, samples were recovered from the chips and stored in their original vials, and the hybridization chamber was filled with Buffer A (nonstringent, 6ϫ), and hybridization was overnight (16 h) at 45°C with rotation (45 rpm) in the sample hybridization mixture. For post-hybridization washing and staining samples were recovered from the chips and stored in their original vials. The hybridization chamber was filled with Buffer A (nonstringent, 6ϫ SSPE (3 M NaCl, 0.2 M NaH 2 PO 4 , 0.02 M EDTA), 0.01% Tween 20). Modules in the Fluidics Station 400 were primed according to the Affymetrix protocol.
The amplification (antibody) solution mixture was as follows: 50 mM MES, 0.5 M [Na ϩ ], 0.025% Tween 20, 2 mg/ml acetylated bovine serum albumin, 0.1 mg/ml normal goat IgG, 3 g/ml biotinylated antibody. All Chips were scanned twice. For data analysis, images obtained were converted into Microsoft Excel format using MAS5.0 software (Affymetrix). All chips were scaled to mean target intensity of 1500. Affymetrix present and absent calls were used. For comparisons between chips, genes with 2-fold differences in expression compared with controls were considered to be significantly differentially expressed.
Radioligand Binding and Second Messenger Studies-Phosphoinositide hydrolysis assays with the constitutively active G␣ q * were performed as described previously (24). Kinetic binding parameters (B max and K d ) were determined from binding assays with [ 3 H]ketanserin, and the results were replicated in at least three separate experiments as detailed previously (24). Phosphoinositide hydrolysis data were analyzed by nonlinear regression using Prism 3.0 software (GraphPad, San Diego, CA), and saturation-binding data were analyzed by using GraphPad Prism. Measurements of p42/44 ERK phosphorylation were performed as described previously by using total p42/44 ERK for normalization (28). Measurements of intracellular calcium mobilization on a Molecular Devices Flexstation were performed essentially as described recently (29), with response normalized to the maximum with each agonist (5-HT, ATP, and/or TRAP).
Statistical Analysis-Statistical significance for all studies was determined using a Student t test, and statistical significance was defined as p Ͻ 0.05. For these experiments C6 glioma cells were transfected with pRNA-U6.1/Neo-RNAi-Cav-1 vector, and clones were stably selected (see "Experimental Procedures" for details). A, representative immunoblot (IB) of five clonally isolated cell lines screened for the presence of Cav-1 (equal amount of protein was loaded in duplicates). B, representative immunoblot for RNAi clone showing knockdown of protein Cav-1 with respect to wt C6 glioma cell lysates. Samples were loaded in duplicate and also probed for total G␣ q and RSK1 to assess nonspecific knockdown of unrelated proteins. C, representative immunoblot for RNAi clone showing knockdown of protein Cav-2 with respect to wt C6 glioma cell lysates. Equal amounts of protein were loaded in duplicate and, as a control, probed for total G␣ q in the samples. seen, FLAG-tagged 5-HT 2A receptors were co-immunoprecipitated robustly with Cav-1 (Fig. 2, 9th to 12th lanes), indicating that 5-HT 2A receptors and Cav-1 associate with each other and are present in a complex in vitro. To establish the specificity of the interaction, we showed that cells co-transfected with Cav-1 and empty vector yielded minimal pull-down of Cav-1 (Fig. 2,  3rd and 4th lanes). Furthermore, no Cav-1 was pulled down from untransfected cells (Fig. 2, 1st and 2nd lanes) or from cells co-expressing 5-HT 2A receptors and GFP (Fig. 2, 5th and 6th  lanes). The interaction of FLAG-5-HT 2A and Cav-1 was unaf- For these experiments wt C6 glioma cells and RNAi-mediated knocked down Cav-1 cells were plated onto 96-well plates. Cells were place in serum-free media for a minimum of 18 h and then exposed to agonist as detailed previously (29). Left panel shows the sigmoid dose response to various agonists (A-C, 5-HT, ATP, and TRAP, respectively) in normalized relative fluorescence units (RFU). D shows the bar graph representing the maximal response of wt and Cav-1 knockdown C6 cell to calcium ionophore A23187. The panels on the right show the changes in fluorescence of wt and RNAi-mediated Cav-1 knockdown C6 cells with time when exposed to vehicle, maximal agonist, and calcium ionophore. (A-D, 5-HT, ATP, TRAP, and A23187, respectively). A shows the 5-HT-mediated dose response for calcium flux for wild-type and Cav-1 knockdown cells; B shows the ATP-mediated dose response for calcium flux for wild-type and Cav-1 knockdown cells; C shows the thrombin-mediated PAR-1 receptor response of wt C6 glioma and C6 Cav-1 RNAi cells; and D shows the maximal response of wt and Cav-1 knockdown cell response to calcium ionophore A23187.

Caveolin-1 Associates with and Regulates 5-HT 2A Receptors
fected by agonist exposure (Fig. 2, compare 9th to 12th lanes). These findings demonstrated that Cav-1 forms a complex with 5-HT 2A receptors in an agonist-independent fashion.
We also examined whether 5-HT 2A receptors and Cav-1 are associated in a milieu in which both 5-HT 2A receptors and Cav-1 are endogenously expressed. Several cell lines reported to express 5-HT 2A receptors were tested including an aortic smooth muscle cell line (A7R5 (30)), NIH 3T3 cells (31), and C6 glioma cells (13). As controls, we also examined HEK-293 cells, which express little Cav-1, and GF-62 cells, which we have demonstrated previously (21) express little Cav-1. We found that C6 glioma cells expressed the highest levels of Cav-1, and these were therefore chosen for further study.
Co-immunoprecipitation studies were then performed in C6 glioma cells and rat brain synaptic membrane preparations wherein endogenous Cav-1 was immunoprecipitated, and the immunoprecipitates were probed for endogenous 5-HT 2A -like immunoreactivity. Fig. 3B shows that 5-HT 2A receptors were co-immunoprecipitated with Cav-1 in C6 glioma cells, and Fig.  3C shows co-immunoprecipitation of native 5-HT 2A receptors by Cav-1 from solubilized rat brain synaptic membrane preparations. As controls to verify the specificity of the interaction, 5-HT 2A receptors were immunoprecipitated with protein-A/G beads conjugated to agarose in the absence (Ϫ) and/or presence of a monoclonal Cav-1 antibody (ϩ). In rat brain membrane preparations 5-HT 2A receptors were immunoprecipitated with protein-A/G beads conjugated to agarose without (Ϫ) or with a monoclonal Cav-1 antibody (ϩ). As shown in Fig. 3C, top panel, 1st lane (Ϫ) shows no immunoreactivity to 5-HT 2A receptor antibody, and the 2nd lane (ϩ) shows that 5-HT 2A receptors were co-immunoprecipitated with Cav-1. The 2nd panel from the top of Fig. 3C shows the 1st lane wherein minimal Cav-1 immunoreactivity to polyclonal Cav-1 antibody was detected in the negative control, and the 2nd lane shows that Cav-1 in the immunoprecipitate was detected (e.g. positive control). The 3rd panel from the top of Fig. 3C represents the immunoblots from the cell lysates to detect endogenous Cav-1. The bottom panel of Fig. 3C represents the immunoblot from rat membrane preparation to verify the expression of 5-HT 2A receptors. Taken together, these results indicate that 5-HT 2A receptors are complexed with Cav-1 in both transfected cells and in their native milieu.
RNAi-mediated Knockdown of Cav-1 in C6 Glioma Cells-We next examined whether Cav-1 functions as a physiological modulator of 5-HT 2A signaling by stably suppressing the expression of Cav-1 in C6 glioma cells using RNAi-mediated gene silencing. We clonally isolated 48 separate C6 lines that survived selection, and we quantified the level of relative Cav-1 expression by Western blot analysis. Shown in Fig. 4A are representative results from five lines screened for Cav-1 expression. Fig. 4B shows representative results from one of the surviving lines, clone 2 of the several screened for Cav-1 expression. All the signaling studies were subsequently carried out with clone 2. The Cav-1 knockdown C6 glioma cells appeared to be morphologically similar to wild-type C6 cells, although they had a slower doubling time than wt C6 cells (data not shown). In our initial studies we examined the relative expression of two unrelated proteins, G␣ q and ribosomal S6 kinase-1 (RSK1), to determine the specificity of the RNAimediated knockdown of Cav-1. We also examined the expression of Cav-2, another member of caveolin family of proteins. As can be seen in Fig. 4B, both the parental and the RNAi-Cav-1 lines expressed similar amounts of G␣ q and RSK1. However, as shown in Fig. 4C there is a significant down-regulation in the expression of protein Cav-2 with no change in G␣ q protein levels suggesting that the expression of Cav-2 protein is regulated by Cav-1, as reported previously (33).
RNAi-mediated Knockdown of Cav-1 Profoundly Suppresses 5-HT 2A -mediated Signaling-We next examined 5-HT 2A -mediated signal transduction in wt and Cav-1 knockdown C6 cells. As shown in Fig. 5A, 5-HT 2A -mediated intracellular Ca 2ϩ mobilization was nearly abolished in Cav-1 knockdown C6 cells. To determine whether this represented a generalized effect on G␣ q -mediated signaling, we performed a cDNA microarray experiment to identify other G␣ q -coupled GPCRs that are endogenously expressed in C6 cells. As shown in Table I and in the Supplemental Material, several other GPCRs were expressed, including the P2Y purinergic receptor and the coagulation factor II receptor (PAR-1), both of which are coupled to G␣ q . Fig.  5B shows that the ATP-induced Ca 2ϩ transients induced by P2Y purinergic receptor activation were also greatly attenuated in Cav-1 knockdown C6 cells. However, the TRAP-induced Ca 2ϩ flux mediated by PAR-1 was unaffected (Fig. 5C). Estimates of agonist potency (EC 50 ) and efficacy (E max ) values revealed that Cav-1 knockdown altered the efficacy without significantly altering potency (Table II). As a control, we also examined the response induced by the calcium ionophore A23187 response (Fig. 5D). These results indicate that Cav-1 modulates signaling of some but not all G␣ q -coupled receptors. The results also imply that Cav-1 knockdown does not induce a generalized dysfunction of G␣ q -mediated signaling.
RNAi-mediated Knockdown of Cav-1 Does Not Alter 5-HT 2A Receptor Binding or Expression-To address the possibility that the impairment in 5-HT 2A receptor signaling was a consequence of decreased 5-HT 2A receptor expression in Cav-1 knockdown cells, we concentrated the receptor protein by using WGA-agarose and subjected the eluate to immunoblot analysis in order to estimate receptor protein expression in Cav-1 knockdown cells versus wt C6 glioma cells. Total cell lysates were also probed for G␣ q as a loading control. Fig. 6A shows a representative immunoblot of 5-HT 2A expression in wt C6 cells  Fig. 7A is a representative immunoblot that shows an apparent increase in basal p42/44 phosphorylation in Cav-1 knockdown cells, whereas the total ERK levels were equivalent within each set (Fig. 7B). Exposure to 10 M 5-HT for 5 min increased p42/44 ERK phosphorylation in wild-type but not in Cav-1 knockdown cells. Fig. 7C depicts the quantification of the increase in phosphorylation upon 5-HT exposure over basal (normalized to total ERK). The average data from three independent experiments showed a significant decrease (p Ͻ 0.05) in p42/44 ERK phosphorylation levels upon agonist exposure in Cav-1 knockdown cells when normalized to total ERK. Our data show that upon 5-HT exposure there is an increase in phospho-ERK that is equivalent in both wt and Cav-1 knockdown C6 cells. As shown in Fig. 7D, the data are represented as a 5-HT-induced increase in phospho-ERK normalized to wt basal, where the maximal increase in phospho-ERK is significantly increased (p Ͻ 0.05) in both wt and Cav-1 knockdown C6 glioma cells.
Cav-1 Overexpression Minimally Affects G␣ q *-mediated Activation of Phospholipase C-To investigate the possibility that Cav-1 modulates the phospholipase C signaling pathway by associating with downstream effectors (i.e. G␣ q and phospholipase C), we examined whether Cav-1 modulated inositol phosphate accumulation in HEK-293 cells transiently expressing a constitutively active form of G␣ q (G␣ q *) with GFP and or Cav-1. G␣ q * activates phospholipase C and elevates immunoprecipitation levels independent of the activation of G␣ q -coupled receptors (32). In the absence of 5-HT 2A receptors, overexpression of Cav-1 modestly but significantly attenuated the G␣ q *-stimulated inositol phosphate accumulation when compared with GFP alone (p Ͻ 0.05; Fig. 8A). It has been reported previously that the association of Cav-1 with G-proteins invariably attenuates their activity. Most importantly, Cav-1 did not cause any change in the relative expression of G␣ q * (Fig. 8B). These findings support the hypothesis that Cav-1 modulates intracellular signaling at the level of receptor-effector coupling and not via a receptor-independent attenuation of G␣ q signaling.
Caveolin-1 Potentiates the Interaction between G␣ q and 5-HT 2A Receptors-To investigate further the role of Cav-1 in modulation of 5-HT 2A receptor signaling, we performed immunoprecipitation studies in HEK-293 cells co-transfected with 5-HT 2A receptors and G␣ q in the presence and absence of Cav-1. Forty-eight hours post-transfection cells were serumstarved for 18 h and exposed to agonist for various times. Cells were then lysed, and 5-HT 2A receptors were immunoprecipitated as described earlier (see "Experimental Procedures"), and G␣ q was detected in the immunoprecipitates using anti-G␣ q antibody (Fig. 9A). Immunoblot analysis showed that in presence of Cav-1 there is a significant increase of G␣ q detected in the immunoprecipitates under conditions of no agonist and at 2 min after agonist exposure (Fig. 9A compare the 4th and 5th  with the 7th and 8th lanes). Fig. 9B shows the quantitative analysis of the net pixel intensities of bands from three independent experiments normalized to the total G␣ q in the lysate. The data show a significant increase in G␣ q in the immunoprecipitates in presence of Cav-1. These findings imply that Cav-1 plays a major role in modulating the signaling of 5-HT 2A receptors by promoting a functional interaction between 5-HT 2A receptors and G␣ q .

TABLE II
RNAi-mediated knockdown of Cav-1 modulates 5-HT 2A -serotonin and P2Y purinergic but not PAR-1-thrombin receptor-mediated signaling Agonist potencies (EC 50 ) and efficacies (E max ) were determined for agonist-mediated activation of intracellular Ca 2ϩ mobilization as described under "Experimental Procedures." Data from radioligand binding assays include estimates of affinity (K d in nM) and the maximal receptor number (B max in fmol/mg) for [ 3 6. Knockdown of Cav-1 expression has no effect on 5-HT 2A protein expression. For these experiments wt C6 glioma cells and Cav-1 knockdown cells were prepared for immunoblot (IB) analysis and radioligand binding assays as described previously. Shown here is a representative immunoblot, which reveals a modest increase in expression of WGA affinity-purified 5-HT 2A receptors in RNAi-knockdown Cav-1 cells as compared with wt C6 cells.

DISCUSSION
The major findings of the present work are that Cav-1 associates with 5-HT 2A receptors in vivo and in vitro, promotes the association of 5-HT 2A receptors with G␣ q , and that the RNAimediated knockdown of Cav-1 profoundly impairs signaling of 5-HT 2A receptors and selected G␣ q -coupled GPCRs. Caveolin-1 has thus emerged as a novel modulator of 5-HT 2A receptor signaling. We also demonstrate that the RNAi-mediated knockdown of Cav-1 expression in C6 glioma cells profoundly impairs signaling for extracellular ATP acting on P2Y purinergic receptors without altering PAR-1 thrombin receptor-mediated signaling. Taken together, these results imply that Cav-1 plays a major role in modulating the signal transduction of selected G␣ q -coupled GPCRs. We also show that the impairment of signaling in Cav-1 knockdown cells was not a consequence of altered receptor or G␣ q protein content. Additionally, our studies with a constitutively active G␣ q imply that Cav-1 does not directly potentiate G␣ q signaling. Instead, our results suggest that Cav-1 facilitates functional interactions between G␣ q and selected GPCRs.
Prior studies by Razani et al. (33) in Cav-1 null mouse fibroblasts have shown that in the absence of Cav-1 there is a drastic reduction in Cav-2 levels because Cav-2 is not targeted to the membrane and is subsequently degraded intracellularly. In another study, an antisense strategy used to deplete Cav-1 in NIH 3T3 cells had no effect on expression of Cav-2 (34). In our present study we report that the stable knockdown of Cav-1 had no significant effect on the mRNA levels of Cav-2; Cav-3 was not measured because it is not expressed in C6 glioma cells (microarray data not shown). However, we report a significant reduction in Cav-2 protein expression suggesting that Cav-2 protein is degraded in Cav-1 knockdown C6 glioma cells, as For these experiments, a constitutively active form of G␣ q (Q229L or G␣ q *) was co-transfected with either GFP or Cav-1myc in HEK-293 cells. A, Cav-1 (Cav-1 ϩ G␣ q *) modestly attenuates G␣ q *-stimulated inositol phosphate accumulation over G␣ q * ϩ GFP in the absence of 5-HT 2A receptors. B shows a representative immunoblot of total G␣ q from all sample groups (in duplicate). Results shown represent the mean (ϮS.E.) of three independent experiments.
Cav-1 is an integral membrane protein, which associates with numerous lipid-modified signaling molecules including Ha-Ras, c-Src, and endothelial nitric-oxide synthase, and almost invariably attenuates the activity of the signaling molecule when overexpressed (35,36). Additionally, Cav-1 directly interacts with the EGF receptor and inhibits its activity (37), and such an interaction leads to internalization of EGF receptor via caveolar domains (38). More recent studies have shown that transcriptional up-regulation of Cav-1 in senescent cells attenuates EGF-mediated signaling thus implicating the role of Cav-1 in EGF receptor-mediated signaling and the general unresponsiveness of senescent cells to growth stimuli (39). The ability of Cav-1 to regulate oncogenes and the downstream signaling cascades has been observed in many overexpression studies wherein Cav-1 is a potent inhibitor of the Ras-p42/44 MAP kinase cascade (40). On the other hand antisense-mediated down-regulation of Cav-1 leads to hyperactivation of the p42/44 MAP kinase cascade (34). These studies suggest that Cav-1 is a modulator of multiple signaling cascades. In this regard, we found that Cav-1 knockdown in C6 glioma cells significantly elevates basal p42/44 ERK phosphorylation and impairs 5-HT 2A agonist-mediated p42/44 ERK phosphorylation. Thus, Cav-1 knockdown appears to induce a dysregulation of GPCR-mediated p42/44 ERK phosphorylation.
Our results are especially intriguing in light of a recent study that has indirectly implicated caveolae in 5-HT 2A -mediated signal transduction (22) in vascular smooth muscle cells. In those studies, the authors reported that 5-HT 2A receptors were enriched in "lipid rafts" isolated by sucrose density gradients. We have also found that a small fraction of 5-HT 2A receptors is located in the caveolar fraction when Cav-1 is overexpressed in HEK-293 cells and in C6 glioma cells where both proteins are endogenously expressed. 2 The localization of other GPCRs such as B2 bradykinin receptors (41) and ␤ 1 -and ␤ 2 -adrenergic receptors in the caveolar fraction has been reported previously (42), where the role of caveolae has been implicated as trafficking centers where GPCRs translocate into or out of the caveolae. Our findings that the effect of Cav-1 knockdown is seen with at least one other G␣ q -coupled GPCR family (P2Y purinergic receptors) and not with PAR-1 implies that Cav-1 selectively modulates G␣ q -coupled receptor signaling, perhaps by promoting association with G␣ q . Recent in vivo studies with Cav-1 knockout mice (19) suggest an indispensable role of Cav-1 in normal life span and cardiac, pulmonary, and vascular functioning (43,44). Our studies, wherein Cav-1 is knocked down, revealed that in the absence of Cav-1 the signalings of 5-HT 2A and P2Y receptors were nearly totally abolished. Because 5-HT 2A receptors represent the principal vascular smooth muscle 5-HT receptor and a main pulmonary and cardiac 5-HT receptor, and because purinergic receptors are ubiquitously expressed, our results suggest that the profound cardiovascular phenotype found in Cav-1 knockout mice may result, in part, from impaired serotonergic and purinergic signaling.
A preliminary microarray analysis of Cav-1 knockdown C6 glioma cells shows that the Cav-1 knockdown cells exhibited no global changes in gene expression compared with the parental cells, although Cav-1 but not Cav-2 mRNA was significantly diminished (not shown). Most important, mRNA levels for 5-HT 2A serotonin receptors and PAR-1 receptors remained unchanged along with the relative mRNA levels for all of the various proteins involved in G␣ q -mediated signaling (not shown). However, the mRNA encoding the P2Y 2 purinergic receptors was absent in Cav-1 knockdown cells and present in wt C6 cells. Most intriguingly, for another G␣ q -coupled P2Y receptor subtype (P2Y 5 ), the steady-state mRNA levels remain unchanged in parental and Cav-1 knockdown cells (not shown). The selective knockdown of P2Y 2 likely accounts for at least some of the decreased response to ATP seen in the Cav-1 knockdown cells. The precise role of Cav-1 in selectively regulating the expression of a single subtype of P2Y family receptors (e.g. P2Y 2 and not P2Y 5 ) will require further study.
In summary we have discovered a novel role for Cav-1 in regulating the activity of serotonergic and purinergic receptors but not thrombinergic G␣ q -coupled receptors. Our results indicate that Cav-1 associates with the 5-HT 2A receptors in vitro (HEK-293 and C6 cells) and in vivo (rat brain synaptic membranes) and that this interaction has profound functional significance. Given the widespread distribution of serotonergic receptors (e.g. platelets, gastrointestinal smooth muscle, uterine smooth muscle, kidneys, the cardiovascular system, and the brain) and the correspondingly ubiquitous distribution of Cav-1, it is likely that the 5-HT 2A -Cav-1 interactions represent a functionally significant protein-protein interaction of potentially profound biological significance. For these experiments, G␣ q was cotransfected with 5-HT 2A receptors and either GFP or Cav-1 in HEK-293 cells. Cells were put in serum-free media for a minimum of 18 h, and then exposed to vehicle (0) or 10 M 5-HT for 2 and 5 min. Lysates were prepared and subjected to immunoprecipitation (IP) using M2 FLAG preconjugated agarose ("Experimental Procedures"). A, shown is a representative immunoblot (IB) from a single experiment that has been replicated three times with equivalent results. B, quantification of the net pixel intensities of G␣ q in the immunoprecipitates normalized to total G␣ q . Each set was then normalized to the maximum in each set. Result shown here are the means Ϯ S.E. from all three experiments.