Caveolin-1 Regulates Transforming Growth Factor (TGF)-β/SMAD Signaling through an Interaction with the TGF-β Type I Receptor

Transforming growth factor-β (TGF-β) signaling proceeds from the cell membrane to the nucleus through the cooperation of the type I and II serine/threonine kinase receptors and their downstream SMAD effectors. Although various regulatory proteins affecting TGF-β-mediated events have been described, relatively little is known about receptor interactions at the level of the plasma membrane. Caveolae are cholesterol-rich membrane microdomains that, along with their marker protein caveolin-1 (Cav-1), have been implicated in the compartmentalization and regulation of certain signaling events. Here, we demonstrate that specific components of the TGF-β cascade are associated with caveolin-1 in caveolae and that Cav-1 interacts with the Type I TGF-β receptor. Additionally, Cav-1 is able to suppress TGF-β-mediated phosphorylation of Smad-2 and subsequent downstream events. We localize the Type I TGF-β receptor interaction to the scaffolding domain of Cav-1 and show that it occurs in a physiologically relevant time frame, acting to rapidly dampen signaling initiated by the TGF-β receptor complex.

(see Refs. 1 and 2 for reviews). The prototype member of the group, TGF-␤1, whose mechanism of action has been a focus of research for the past decade, classically initiates signaling at the plasma membrane by binding to a heterotetrameric complex consisting of two transmembrane serine/threonine kinases, known as Type I and Type II TGF-␤ receptors (T␤R-I and T␤R-II) (3,4). The activation of this membrane complex occurs via the ligand-dependent phosphorylation of T␤R-I by T␤R-II (5,6). In turn, T␤R-I can act to phosphorylate its immediate downstream effectors, Smad-2 and Smad-3, members of the SMAD family of intracellular signaling molecules (7,8). This phosphorylation induces a conformational change in Smad-2 and -3, thereby facilitating their heteromerization with another member of the family, Smad-4 (9). The SMAD complex then translocates to the nucleus, where it acts to regulate the transcription of various target genes (7,9).
As in numerous other cellular processes, TGF-␤-mediated signaling is subject to regulation and inhibition by a variety of mechanisms. The p42/44 MAP kinase pathway, interferon-␥/ STAT cascade, NF-B, SnoN/Ski oncoproteins, among others have been shown to act as negative regulators of TGF-␤ signaling primarily by interacting with, modifying, or regulating the SMAD proteins (10 -15). Various groups have also observed regulation at the receptor level by the showing the interaction of certain intracellular proteins with the T␤R-I. Of specific interest are the inhibitory proteins, Smad-6 and Smad-7, a functionally divergent subset of the SMAD protein family that can inhibit TGF-␤ signaling by directly interacting with the T␤R-I (16 -18). Other proteins such as FKBP12 have also been shown to interact with and negatively regulate T␤R-I (19).
However, surprisingly little is known about the interactions of the TGF-␤ receptor complex with other membrane proteins. Although the compartmentalization of signaling molecules via scaffolding proteins is an emerging theme in signal transduction biology (20), the behavior and regulation of the TGF-␤ receptors at the membrane has thus far been a matter of speculation. Previous reports addressing the heterotetramerization of the T␤R-I/II complex have shown a characteristically punctate membrane distribution (21,22), and recently, the cloning of a SMAD anchor protein, SARA, also revealed a punctate distribution for both SARA and the TGF-␤ receptor complex (23).
Caveolae are ϳ50 -100 nm vesicular invaginations of the plasma membrane and are thought to form as a result of a local accumulation of cholesterol, glycosphingolipids, and caveolin-1 (24 -26). Caveolin-1, a 21-24-kDa integral membrane protein, is a principal component of caveolae membranes in vivo (27)(28)(29). Structurally, a hallmark of caveolin-1 and by extension, caveolae, is a characteristically punctate staining pattern at the plasma membrane (30). Although caveolae function in vesicular and cholesterol trafficking (31,32), they have also been * This work was supported by grants from the the National Institutes of Health, the Muscular Dystrophy Association, American Heart Association, and the Komen Breast Cancer Foundation (to M. P. L.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Given the punctate plasmalemmal distribution of Cav-1 and its emerging role in various signaling cascades, we set out to investigate its relationship with the TGF-␤ receptor complex and substrate SMADs. Here, we demonstrate that Cav-1 and T␤R-I are highly colocalized at the membrane, that T␤R-I, T␤R-II, and Smad-2 (but not Smad-4) cofractionate with caveolin-1 in caveolae enriched microdomains, and that caveolin-1 directly interacts with T␤R-I in both heterologous and endogenous settings. We show that this interaction has functional consequences because Cav-1 is able to suppress TGF-␤-mediated transcriptional activation. In addition, we demonstrate that Cav-1 diminishes the phosphorylation of Smad-2, disrupts its interaction with Smad-4, and prevents Smad-2 translocation to the nucleus in the ligand-activated state. This inhibition is mediated by an interaction between T␤R-I and the scaffolding domain of Cav-1 and occurs in a physiologically relevant time frame. We show a rapid increase in the T␤R-I/Cav-1 interaction upon ligand binding, and, by using an antisense strategy, we demonstrate that targeted down-regulation of caveolin-1 is sufficient to hyperactivate TGF-␤ signaling.

EXPERIMENTAL PROCEDURES
Materials and Cell Culture-The caveolin-1 mAb 2297 (used for immunoblotting) and mAb 2234 (used for immunofluorescence and immunoprecipitation) (30) were the gifts of Roberto Campos-Gonzalez (Transduction Laboratories, Inc.). The phospho-specific anti-SMAD2 rabbit pAb was a gift of Peter ten Dijke (Ludwig Institute for Cancer Research). The anti-FLAG tag mAb (Sigma), the anti-SMAD2 mAb (Transduction Laboratories, Inc.), and the anti-HA as well as the anti-TGF-␤ type I receptor rabbit pAb (Santa Cruz Biotechnology) were obtained commercially. Cell culture reagents were from Life Technologies, Inc. Recombinant human TGF-␤1 was obtained from R&D systems. NIH-3T3 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% donor calf serum, 2 mM glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin at 37°C and 5% CO 2 . 293T cells were similarly cultured, with the exception of serum (10% fetal bovine serum in lieu of donor calf serum).
Immunoblotting-48 h post-transfection, cells were washed with PBS and incubated with lysis buffer (10 mM Tris, pH 7.5, 50 mM NaCl, 1% Triton X-100, 60 mM octyl glucoside) containing protease inhibitors (Roche Molecular Biochemicals). Where indicated, protein concentrations were quantified using the BCA reagent (Pierce). Samples were separated by SDS-PAGE (12.5% acrylamide) and transferred to nitrocellulose. The nitrocellulose membranes were stained with Ponceau S (to visualize protein bands) followed by immunoblot analysis. All subsequent wash buffers contained 10 mM Tris, pH 8.0, 150 mM NaCl, 0.05% Tween-20, which was supplemented with 1% bovine serum albumin and 2% nonfat dry milk (Carnation) for the blocking solution and 1% bovine serum albumin for the antibody diluent. Primary antibodies (either polyclonal or monoclonal) were used at a 1:500 dilution. Horseradish peroxidase-conjugated secondary antibodies (1:5000 dilution, Transduction Laboratory) were used to visualize bound primary antibodies with a chemiluminescent substrate (Pierce).
Immunofluorescence-The procedure was performed as we previ-ously described (35). For colocalization studies, NIH-3T3 cells (transfected with caveolin-1 and either T␤R-I w.t. or T␤R-I (T204D)) were fixed for 30 min in PBS containing 2% paraformaldehyde, rinsed with PBS, and quenched with 50 mM NH 4 Cl for 10 min. The cells were then incubated in permeabilization buffer (PBS, 0.2% bovine serum albumin, 0.1% Triton X-100) for 10 min, washed with PBS, and double-labeled with a 1:400 dilution of anti-caveolin-1 mAb 2234 and 1:200 dilution of anti-HA rabbit pAb for 60 min. After rinsing three times with PBS, secondary antibodies (7.5 g/ml) [(lissamine-rhodamine-conjugated goat anti-rabbit and fluorescein isothiocyanate-conjugated goat antimouse) antibodies] were added for a period of 60 min. Cells were washed three times with PBS, and slides were mounted with Slow-Fade anti-fade reagent (Molecular Probes). A Bio-Rad MR600 confocal fluorescence microscope was used for visualization of bound secondary antibodies. For assessment of SMAD2 translocation to the nucleus, a similar procedure was followed with the following differences: NIH-3T3 cells were transfected with caveolin-1, serum-starved for 20 h, and treated with TGF-␤1 (4 ng/ml) for 45 min. A 1:400 dilution of anticaveolin-1 mAb 2234 and a 1:200 dilution of anti-SMAD2 mAb were used for the caveolin-1 and endogenous SMAD2 staining, respectively.
Purification of Caveolae-enriched Membrane Fractions-Caveolaeenriched membrane fractions were purified essentially as previously described (44,50). 293T cells plated on a 150-mm diameter plate were transfected with the appropriate plasmid(s). 36 h post-transfection, the cells were washed twice in cold PBS, scraped into 2 ml of MBS (25 mM Mes, pH 6.5, 150 mM NaCl) containing 1% Triton X-100, passed 10 times through a loose fitting Dounce homogenizer, and mixed with an equal volume of 80% sucrose (prepared in MBS lacking Triton X-100). The sample was then transferred to a 12-ml ultracentrifuge tube and overlaid with a discontinuous sucrose gradient (4 ml of 30% sucrose, 4 ml of 5% sucrose, both prepared in MBS, lacking detergent). The samples were subjected to centrifugation at 200,000 ϫ g (39,000 rpm in a Sorval rotor TH-641) for 16 h. A light scattering band was observed at the 5/30% sucrose interface. Twelve 1-ml fractions were collected, and 50-l aliquots of each fraction were subjected to SDS-PAGE and immunoblotting.
Coimmunoprecipitation of Caveolin-1 with TGF-␤ RI-For coimmunoprecipitation of heterologously expressed proteins, 293T cells plated on 100-mm dishes were transfected with the appropriate plasmids. 36 h post-transfection, cells were lysed in lysis buffer (see "Immunoblotting"), clarified by centrifugation at 15,000 ϫ g for 15 min, and precleared by incubation with protein A-Sepharose (Amersham Pharmacia Biotech) for 1 h at 4°C. Supernatants were transferred to separate 1.5-ml microcentrifuge tubes containing anti-HA pAb or appropriate control antibodies (beads alone, preimmune serum) prebound to protein-A Sepharose. After incubation by rotating overnight at 4°C, immunoprecipitates were washed three times with lysis buffer and subjected to immunoblot analysis with the anti-caveolin-1 2297 mAb probe. For coimmunoprecipitation of endogenous proteins, NIH-3T3 cells were plated on 100-mm dishes and lysed at confluency. The procedure was as described above with the anti-caveolin-1 2234 mAb, anti-T␤R-I pAb, or preimmune serum pAbs serving as the precipitating antibodies.
In Vivo Phosphorylation Experiments-NIH-3T3 cells plated on 100-mm dishes were transfected with the appropriate plasmids. 36 h post-transfection, cells were washed twice in Dulbecco's modified Eagle's medium lacking phosphate and incubated for 3 h in Dulbecco's modified Eagle's medium lacking phosphate supplemented with 1 mCi [ 32 P]orthophosphate/100-mm dish. Indicated plates were then additionally treated with TGF-␤1 (4 ng/ml) for 45 min. Cells were washed in ice-cold PBS and subjected to lysis in RIPA/Nonidet P-40 buffer and immunoprecipitation with anti-FLAG M2 mAb as outlined above. SDS-PAGE and subsequent autoradiography visualized the phosphorylated SMAD2 proteins.
Intra-caveolar SMAD2 Phosphorylation-293T cells plated on a 150-mm diameter plates were transfected with T␤R-I (T204D), SMAD2, and either caveolin-1 or empty vector. Caveolae-enriched membrane fractions were purified as outlined above. Dilution with 1ϫ MBS and subsequent centrifugation was used to concentrate the caveolar membranes into a 50-l volume. Approximately 5 g of these membranes was mixed 1:1 with 10 l of 2ϫ kinase reaction buffer (40 mM Hepes, pH 7.4, 10 mM MgCl 2 , and 2 mM MnCl 2 ), and the reaction was initiated by the addition of 4 mM ATP (Sigma) for 15 min. Extent of phosphorylation was determined by immunoblot analysis with anti-phospho-SMAD2 pAb.
Cytoplasmic/Nuclear Fractionation-NIH-3T3 cells plated on 100-mm dishes were transfected with the appropriate plasmids. 36 h post-transfection, cells were washed once with PBS, scraped in 1 ml of Hypotonic lysis buffer (10 mM Hepes, pH 7.9, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 0.5% Nonidet P-40, 1 mM dithiothreitol, and protease inhibitors), passed 10 times through a loose fitting Dounce homogenizer, and centrifuged at 3000 rpm for 3 min. The supernatant (i.e. cytoplasmic fraction) was saved, whereas the pellet (nuclear fraction) was washed twice with hypotonic lysis buffer before resuspending in the same buffer and sonicating to disrupt nuclear membranes. Protein concentrations from both fractions were quantified using the BCA reagent, and equal amounts of protein were subjected to immunoblot analysis with the anti-FLAG mAb as the probe.
Luciferase Reporter Assays for TGF-␤ Activity-The A3-lux reporter construct and the TGF-␤ transcriptional coactivator Fast-1 were the gifts of Malcolm Whitman (Harvard Medical School) (51) and the 3TPlux reporter construct was the gift of Joan Massagué (Memorial Sloan Kettering Cancer Center) (52). Briefly, 150,000 NIH-3T3 cells were seeded per well in 6-well plates 12-24 h prior to transfection. For assays involving TGF-␤1 stimulation, each plate was transfected with the reporter of interest (A3-lux/Fast-1 or 3TP-lux), caveolin-1 (pCB7-Cav-1), or empty vector (pCB7) and pRSV-Gal (Promega, Madison, WI). 12 h post-transfection, the cells were rinsed with PBS, incubated in 0.2% donor calf serum starvation medium for 20 h and then, where indicated, treated with TGF-␤1 (4 ng/ml). 8 h post-stimulation, the cells were lysed in 200 l of extraction buffer, 100 l of which was used to measure luciferase activity as described (53). Luciferase activities were normalized for galactosidase activity assayed by a galactosidase assay system (Promega). For the analysis of the effects of caveolin-1 and the deletion mutant Cav-1 ⌬61-100 on the constitutively active TGF-␤ type I receptor, all points were additionally transfected with T␤R-I (T204D), and the cells were serum-starved but not stimulated with TGF-␤1 posttransfection. Results were expressed as a ratio of luciferase activity to galactosidase activity. Each experimental value represents the average of three separate transfections performed in parallel; error bars represent the observed S.D. All experiments were performed at least three times independently and yielded virtually identical results.
TGF-␤ RI in Vitro Kinase Assay-The kinase substrate used in this assay, the GST-SMAD2 fusion protein, was a gift of Mark de Caestecker (National Cancer Institute) (54). The purification of GST-SMAD2 was as we described previously (30,40,55). Briefly, after expression in Escherichia coli (BL21 strain; Novagen, Inc.), GST-SMAD2 was affinity purified on glutathione-agarose beads, using the detergent sarcosyl for initial solubilization (56) and washed three times with TNET buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 5 mM EDTA, 1% Triton X-100). SDS-PAGE followed by Coomassie staining was used to determine the approximate molar quantities of the fusion protein/100 l of packed bead volume. GST-SMAD2 was then eluted with an appropriate volume of TNET buffer containing 20 mM reduced glutathione (Sigma). An anti-HA rabbit pAb was used to immunoprecipitate HA-tagged T␤R-I (T204D) from transiently transfected 293T cells. In vitro kinase assays were performed as described previously (35), with minor modifications. Briefly, immunoprecipitates and an appropriate fraction of the GST-SMAD2 eluate were equilibrated with kinase reaction buffer (20 mM Hepes, pH 7.4, 5 mM MgCl 2 , and 1 mM MnCl 2 ), and the reaction was initiated by the addition of 4 mM ATP (Sigma). After 15 min of incubation at 25°C, the reaction was terminated by addition of 2ϫ SDS-PAGE sample buffer and boiling for 4 min. Phosphorylated GST-SMAD2 was detected by immunoblotting with the anti-phospho-SMAD2 rabbit pAb. In reactions involving the use of caveolin-derived peptides, prior to initiating the reaction, the immunoprecipitates were preincubated in kinase reaction buffer with the indicated peptide for 30 min at 4°C. The caveolin-based peptides were synthesized using standard methodology and subjected to amino acid analysis and mass spectroscopy (Massachusetts Institute of Technology Biopolymers Laboratory and Research Genetics) to confirm their purity and composition, as we described previously (35,48,49,57,58). Peptides were dissolved in Me 2 SO, and 100ϫ stock solutions were prepared for use in experiments. In vitro kinase assays assessing the autophosphorylation of T␤R-I were exactly as described above except for the addition of GST-SMAD2 and the use of 10 Ci of [␥-32 P]ATP in lieu of ATP. Phosphorylated T␤R-I was visualized by autoradiography.

TGF-␤ Type I Receptor Colocalizes, Cofractionates, and Interacts with Caveolin-1 in Caveolae-enriched Domains-Given
the punctate membrane immunostaining previously reported for the T␤R-I/II complex (21-23), a distribution resembling that of caveolin-1 and caveolae (30), we first investigated the possibility of colocalization between the two proteins in vivo. We utilized two different T␤R-I cDNA constructs (the HAtagged wild-type form or a constitutively active T204D mutant). It is important to note that the T␤R-I (T204D) construct is a constitutively active mutant that can initiate TGF-␤ signaling independent of ligand binding or heteromerization with the T␤R-II (47). For most of the following studies we used NIH-3T3, cells readily responsive to TGF-␤/SMAD signaling, or 293T, a Cav-1-negative cell line readily amenable to heterologous Cav-1 expression (50,59). NIH-3T3 cells cotransfected with Cav-1 and either T␤R-I w.t. or T␤R-I (T204D) were immunostained and visualized by confocal microscopy. Confocal slices of areas delineating the plasma membrane revealed significant colocalization between the proteins (Fig. 1). T␤R-I (T204D) was highly aligned with Cav-1, indicating that the two proteins can colocalize independent of ligand activation and T␤R-II. The coalignment for cells transfected with T␤R-I w.t. was not complete, however, possibly indicating that the baseline distribution of T␤R-I is not solely restricted to areas of Cav-1 expression (Fig. 1).
Biochemically, we have previously shown that caveolar microdomains can be separated from other cellular constituents using a sucrose gradient ultracentrifugation procedure. Via this FIG. 1. Caveolin-1 colocalizes with T␤R-I w.t. and T␤R-I (T204D). NIH-3T3 cells were cotransfected with the cDNAs encoding caveolin-1 and either T␤R-I w.t. or T␤R-I (T204D). Cells were then doubly immunostained with antibodies that specifically recognize caveolin-1 and HA (the hemaglutanin tag fused to each receptor). The bound primary antibodies were visualized with distinctly tagged secondary antibody probes (see "Experimental Procedures"). Both cell populations revealed significant membrane colocalization between caveolin-1 and T␤R-I. Positive and negative controls omitting a given primary antibody and with cells singly transfected with a given cDNA were also performed and yielded the expected results (data not shown).
Caveolin-1 and TGF-␤/SMAD Signaling method, it is possible to concentrate Cav-1, the caveolae marker protein, by over 2000-fold with respect to total cellular protein (45,60). By transfecting 293T cells with appropriate cDNAs, we applied this separation scheme to various members of the TGF-␤/SMAD pathway. The outputs of this centrifugation, which consists of 12 equal fractions (of which fractions 4 -6 and 9 -12 are considered of caveolar and noncaveolar origin, respectively), are shown in Fig. 2 (A and B). Distinct pools of T␤R-I (T204D), T␤R-I w.t., and T␤R-II w.t. cofractionate with Cav-1 in caveolae enriched domains. Note that the overwhelming majority of cellular proteins are noncaveolar, as observed by Ponceau S staining ( Fig.  2A, top panel). Therefore, the enrichment of a signaling protein cofractionating with Cav-1 is actually higher than indicated. A comparison of two of the above fractions (one of caveolar origin (fraction 5), and one of noncaveolar origin ( fraction 11)) via equal protein (rather than equal volume) serves to illustrate this point (Fig. 2C); note that T␤R-I is concentrated in the caveolar fractions. Similarly, our fractionation scheme applied to the receptorphosphorylated Smad-2 and the common-mediator Smad-4 showed that although Smad-2 is enriched in caveolae, Smad-4 is distinctly excluded from these fractions (Fig. 2, B and C). The presence of Smad-2 in caveolae microdomains is intriguing in light of the cloning of SARA, an anchoring protein shown to recruit Smad-2 to subcellular locations enriched in the TGF-␤ receptors (23). It should be noted that as a general rule, receptors and other transmembrane molecules are normally excluded during the purification of caveolae, or they are washed out by the harsh treatments employed (36,57,(61)(62)(63)(64)(65)(66)(67). In this regard, it is intriguing that several members of the TGF-␤/SMAD cascade remain associated with caveolae.
Because 293T cells do not endogenously express Cav-1, it should be noted that the fractionation of T␤R-I, T␤R-II, and Smad-2 to caveolar microdomains occurs independently of Cav-1 expression. This indicates that these signaling molecules are more generally localized to caveolae and caveolae-related domains (a term used to denote membrane regions rich in cholesterol and glycosphingolipids irrespective of the presence of Cav-1). This observation raises the issue as to the function of Cav-1 in TGF-␤ signaling, a point we address in the following experiments.
As Cav-1 is a marker protein for caveolae and has been shown to interact with certain signaling molecules in vivo (reviewed in Ref. 41), we next investigated its potential interaction with T␤R-I. 293T cells were transfected with HA-tagged T␤R-I (T204D) and Cav-1 either separately or together and subjected to immunoprecipitation using either anti-HA pAb or relevant negative controls (preimmune serum pAb and beads alone). Cav-1 is immunoprecipitated only in cells coexpressing T␤R-I (T204D) and Cav-1 (Fig. 3A). It should be noted that this interaction occurs to a similar extent with T␤R-I w.t. (see Fig.  9A), indicating that Cav-1 does not discriminate between the constitutively active and wild-type forms of the receptor.
Since these experiments were performed in a heterologous setting, we immunoprecipitated T␤R-I from untransfected NIH-3T3 cells in an attempt to assess the endogenous Cav-1/ T␤R-I interaction. Fig. 3B shows that a significant amount of Cav-1 is associated with T␤R-I at base-line. As compared with the immunoprecipitation of Cav-1 using a high affinity mAbclone 2234, an antibody capable of immunodepleting Cav-1 from cell lysates, 2 we estimate the association of endogenous Cav-1 with T␤R-I to be on the order of ϳ5-10% in vivo (Fig.  3B).
Caveolin-1 Inhibits TGF-␤/SMAD Signaling by Blocking SMAD Activation-Because T␤R-I plays a pivotal role in the propagation of TGF-␤ signaling from the membrane to the nucleus, we were interested in the functional consequences of its interaction with Cav-1. First, we investigated the response of two commonly used TGF-␤ transcriptional reporter assays to heterologous Cav-1 expression. The A3-lux/Fast-1 and 3TP-lux systems utilize TGF-␤-responsive promoter elements to drive the expression of a luciferase reporter gene (51,52). NIH-3T3 cells were cotransfected with the appropriate luciferase reporters and a combination of T␤R-I (T204D), Cav-1, or empty vector controls. Both reporters displayed robust activation in the presence of the constitutively active T␤R-I, an effect that was dramatically reverted in cells coexpressing Cav-1 (Fig. 4A). In lieu of T␤R-I (T204D), we used the same reporters to look at the activation of the endogenous TGF-␤ receptor complex by treating NIH-3T3 cells with TGF-␤1 (4 ng/ml) for 8 h. Cav-1 again displayed inhibitory capacity in this respect, diminishing the ligand-activated state 3-4-fold (Fig. 4B). A K232R (kinase dead) mutant of the T␤R-I had previously been shown to lack kinase activity and to act in dominant negative fashion by dimerizing with the wild-type T␤R-I (46). A comparison of the effects displayed by Cav-1 with those of T␤R-I (K232R) reveals a similar inhibitory capacity for both proteins (Fig. 4B). We have independently performed these assays using COS-7 cells yielding similar results, 2 indicating that the Cav-1-mediated inhibition is not necessarily cell type-specific. The receptor-activated SMADs (Smad-2 and -3) are the first step in the propagation of TGF-␤ signals from the plasma membrane to the nucleus. The ligand-activated T␤R-I directly phosphorylates Smad-2 and -3 at a C-terminal SSXS motif (7,68), a modification that facilitates Smad-2/-3 release and subsequent heteromerization with Smad-4 (9, 69). This interaction is followed by nuclear translocation of the complex with pleiotropic effects on the transcription of target genes (reviewed in Ref. 2). Because T␤R-I activation is the initiating event in this cascade, the interaction of Cav-1 with the receptor should by extension disrupt Smad-2/-3 phosphorylation, heteromerization with Smad-4, and translocation to the nucleus. We first investigated ligand-activated phosphorylation of SMADs by transfecting NIH-3T3 cells with FLAG-tagged SMAD2 and either Cav-1 or empty vector, labeling with 32 PO 4 , and selectively treating with TGF-␤1. As previously reported, the phosphorylation of Smad-2 was readily apparent in the TGF-␤1-stimulated cells (54) (Fig. 5A). However, in cells coexpressing Cav-1, a dramatic reduction of this phosphorylation was observed. Note that although base-line levels are reduced, Cav-1 more dramatically attenuates Smad-2 phosphorylation in the TGF-␤1-stimulated state.
The role of Cav-1 in SMAD phosphorylation was also investigated in a novel manner. Based on our sucrose density gradient experimentation above, we observed that T␤R-I and Smad-2 are localized to caveolae-enriched microdomains, independent of Cav-1 coexpression. To test the functional significance of Cav-1, we cotransfected 293T cells with T␤R-I (T204D), Smad-2, and either Cav-1 or empty vector and subjected the purified and concentrated caveolar fractions to an in vitro kinase reaction by adding 4 mM ATP. Fig. 5B shows that the constitutively active receptor can effectively phosphorylate cofractionated Smad-2 in these microdomains in the absence of Cav-1. In contrast, the fractions also containing Cav-1 significantly attenuate this process.
Finally, we tested the ability of Cav-1 to suppress TGF-␤mediated nuclear translocation of Smad-2 both biochemically and via microscopy. NIH-3T3 or 293T cells were cotransfected with T␤R-I (T204D), FLAG-tagged Smad-2, and either Cav-1 or empty vector. Via hypotonic lysis, cells were fractionated into with anti-caveolin 2297 mAb, anti-HA pAb, and anti-FLAG mAb was used to detect the Cav-1, T␤R-I w.t./T␤R-I (T204D)/T␤R-II w.t./SMAD4, and SMAD2 proteins, respectively. The Type I and Type II receptors and SMAD2 cofractionate with Cav-1 (fractions 4 -6), whereas SMAD4 is entirely excluded from these caveolae-enriched fractions. The distribution of total cellular protein (as analyzed by Ponceau S staining) is shown in the top panel, indicating that only a minute portion of total cellular protein actually exists in caveolae. C, analysis of two fractions from selected gradients in A (fraction 5, caveolar origin; fraction 11, noncaveolar origin) using equal protein quantities. Note that Cav-1 as well as T␤R-I w.t., T␤R-I (T204D), and Smad-2 are concentrated in caveolar fractions in contrast to Smad-4, which is entirely noncaveolar.

FIG. 2. TGF-␤ Type I receptor and SMAD2 cofractionate with caveolin-1 in caveolae-enriched microdomains.
A and B, 293T cells were individually transfected with the indicated cDNAs and subjected to sucrose gradient centrifugation after homogenization in buffer containing 1% Triton X-100 (see "Experimental Procedures"), a method which separates Triton-resistant caveolae-rich domains (fractions 4 -6) from other cellular components (fractions 9 -12). Immunoblot analysis with anti-caveolin 2297 mAb, anti-HA pAb, and anti-FLAG mAb was

FIG. 3. Caveolin-1 interacts with TGF-␤ Type I receptor in both heterologous and endogenous settings.
A, 293T cells were transfected with HA-tagged T␤R-I (T204D) and Cav-1 either separately or together, as indicated. Cell lysates were prepared and immunoprecipitated (IP) with either anti-HA pAb or relevant negative controls (preimmune serum pAb and beads alone). Immunoprecipitates were resolved by SDS-PAGE and subjected to immunoblot analysis with anticaveolin-1 2297 mAb. Note that Cav-1 is immunoprecipitated only in cells coexpressing T␤R-I (T204D) and Cav-1 (top panel). B, NIH-3T3 cells grown to confluence were subjected to immunoprecipitation with either anti-Cav-1 2234 mAb, anti-T␤R-I pAb, or appropriate controls (irrelevant mAb, preimmune serum pAb, and beads alone). Immunoblotting with anti-Cav-1 2297 mAb reveals an endogenous interaction between T␤R-I and Cav-1. As compared with immunoprecipitated Cav-1 (first lane), the amount of Cav-1 associated with T␤R-I can be estimated to be on the order of ϳ5-10%. distinct cytoplasmic and nuclear fractions and the translocation of Smad-2 was analyzed. In both cell lines, Cav-1 reduced Smad-2 levels in the nuclear fraction (Fig. 7A). We used immunofluorescence confocal microscopy to corroborate these observations by transfecting NIH-3T3 cells with Cav-1 and comparing the TGF-␤1-induced translocation of endogenous Smad-2 in transfected and nontransfected cells. Fig. 7B shows a mid-line confocal slice delineating the cytoplasms/nuclei of two closely juxtaposed cells. Although there is near complete nuclear translocation of Smad-2 in the untransfected cell, the neighboring Cav-1-expressing cell has a significant cytoplasmic pool of Smad-2.

The Caveolin-1 Scaffolding Domain Mediates the Functional Interaction with TGF-␤ Type I Receptor-
The demonstration of an association between caveolin-1 and T␤R-I and its inhibitory effects on downstream SMAD signaling led us to determine the Cav-1 domains possibly mediating this functional interaction via in vitro kinase assays. We affinity purified a GST-Smad-2 fusion protein and used it as a physiologically relevant substrate for immunoprecipitated T␤R-I (T204D) in vitro. We have previously described a series of caveolin-derived peptides spanning various regions of the caveolin molecule (Fig. 8A) (35,70). Using these peptides in combination with T␤R-I (T204D) and GST-Smad-2, we were able to localize an inhibitory region in the Cav-1 molecule. Of the several peptides derived from the Nor C-terminal regions, only two displayed a potent suppression of GST-Smad-2 phosphorylation, namely residues 61-101 (the FIG. 4. Caveolin-1 functionally regulates TGF-␤/SMAD signaling at the transcriptional level. A, NIH-3T3 cells were transfected with either A3-lux/Fast-1 (left panel) or 3TP-lux (right panel) (the TGF-␤-responsive luciferase reporters) and a combination of T␤R-I (T204D), Cav-1, or empty vector controls. Note that Cav-1 inhibits signaling mediated by the constitutively activate T␤R-I in both reporter systems. B, as in A, the A3-lux/Fast-1 and 3TP-lux reporter systems were utilized. However, instead of the constitutively active receptor, TGF-␤1 ligand (4 ng/ml) was used to stimulate endogenous T␤R-I in NIH-3T3 cells for 8 h, as indicated. In addition, the ability of Cav-1 to diminish signaling was compared with that of T␤R-I (K232R), a kinasedead receptor mutant. Note that Cav-1 significantly diminishes the transcriptional response, an effect on the same order of potency as the dominant negative K232R mutant receptor. In both panels, luciferase activities are expressed as ratios normalized to ␤-galactosidase activity, and each experimental value represented graphically is the average of three separate transfections performed in parallel. Error bars represent the observed S.D.  Cav-1 scaffolding domain). Importantly, when the Cav-1 scaffolding domain is divided into two halves (residues 84 -92 and 93-101), this inhibition is completely abrogated (Fig. 8B). The T␤R-I (T204D) has been shown to retain autophosphorylation activity in vitro (46,47). Therefore, we conducted kinase assays as above using only the immunoprecipitated T␤R-I (T204D). The same peptides that displayed inhibition of Smad-2 phosphorylation also abrogated the autophosphorylation of T␤R-I (Fig. 8C), indicating that the Cav-1 scaffolding domain acts to block T␤R-I kinase activity.
If the scaffolding domain is indeed the region of caveolin that binds to T␤R-I, it would be predicted that coincubation of NIH-3T3 cell lysates with Cav-1 peptides containing this domain (as also utilized in Fig. 8) would competitively disrupt the Cav-1/T␤R-I complex. Fig. 9C shows that indeed only peptides containing the scaffolding domain (i.e. 61-101 and 82-101) are capable of abrogating the interaction of Cav-1 with immunoprecipitated T␤R-I.
To demonstrate that the scaffolding domain is also functionally important in vivo, we utilized the previously tested A3-lux/ Fast-1 luciferase reporter system. NIH-3T3 cells were cotransfected with the A3-lux/Fast-1 reporter and a combination of T␤R-I (T204D), Cav-1 FL, Cav-1 ⌬61-100, or empty vector controls. Cav-1 FL inhibits the signaling mediated by the constitutively activate T␤R-I (also see Fig. 4B), whereas the Cav-1 ⌬61-100 mutant has no effect (Fig. 9D).

The Caveolin-1/TGF-␤ Type I Receptor Interaction Occurs in a Physiologically Relevant Time Frame and Is Important for
Dampening TGF-␤ Signaling-Various investigators have reported the phosphorylation kinetics of Smad-2 to occur gradually with a t1 ⁄2 of ϳ5-10 min, peaking at 20 -30 min (2, 54). The demonstration of an endogenous interaction between Cav-1 and T␤R-I (Fig. 3B) led us to investigate whether this association occurs with altered kinetics in the ligand-stimulated state and whether it occurs within the time frame of SMAD phosphorylation. Serum-starved NIH-3T3 cells (grown to confluence) were treated with TGF-␤1 (4 ng/ml) over an 80-min period and subjected to immunoprecipitation with anti-T␤R-I. The Cav-1/T␤R-I interaction gradually increases from base line, peaking at 40 min post-stimulation (Fig. 10, top panel). The total expression of Cav-1 is unaffected upon TGF-␤1 treatment (Fig. 10, middle panel), indicating that the observed interaction is independent of transcriptional regulation. Furthermore, note that the base-line Cav-1/T␤R-I interaction in this serum-starved and ligand-unstimulated setting is minimal in contrast to the ϳ5-10% serum-stimulated interaction observed in Fig. 3B. Cells were fractionated into cytoplasmic and nuclear components via hypotonic lysis (see "Experimental Procedures") and immunoblotted for the relative translocation of SMAD2 with anti-FLAG mAb. Note that in the presence of Cav-1 a significant portion of SMAD2 remains in the non-nuclear (cytoplasmic) fraction. B, NIH-3T3 cells were transfected with Cav-1 and 36 h post-transfection were treated with TGF-␤1 (4 ng/ml) for 45 min to allow for the translocation of endogenous SMAD2. Cells were then doubly immunostained with antibodies that specifically recognize caveolin-1 (anti-Cav-1 N20 pAb) and SMAD2 (anti-SMAD2 mAb). The bound primary antibodies were visualized with distinctly tagged secondary antibody probes (see "Experimental Procedures"). Note that only the cell overexpressing Cav-1 (left panel) shows significant cytoplasmic SMAD2 staining, whereas the untransfected neighboring cell shows nuclear SMAD2 staining that is characteristic of TGF-␤1-treated cells.
We also determined the phosphorylation state of Smad-2 and observed an increase in phosphorylation with expected kinetics (i.e. maximal at ϳ20 -30 min and plateauing thereafter; Fig.  10, lower panel). Note that the peak of Cav-1/T␤R-I interaction (ϳ30 -40 min) occurs slightly after the peak of phosphorylation level of Smad-2 (ϳ20 -30 min). Given this time frame, it is plausible that Cav-1 can act to dampen TGF-␤ signaling by gradually sequestering more of the available T␤R-I pool.
To test this hypothesis, we studied the kinetics of TGF-␤ signaling in cells with perturbed caveolin levels. We have previously described the use of an antisense construct in downmodulating caveolin-1 levels in NIH-3T3 cells (72). Cells harboring antisense caveolin-1 display significantly reduced Cav-1 protein levels and a concomitant loss of morphological caveolae (72). Confluent plates of serum-starved parental NIH-3T3 cells and their antisense-Cav-1 counterparts were treated with TGF-␤1 (4 ng/ml) over a 60-min period and subjected to immunoblot analysis. As expected, base-line phosphorylation of Smad-2 was negligible in both cell types and gradually increased in the TGF-␤1-stimulated state (Fig. 11A). However, starting at the 15-min time point, cells harboring antisense Cav-1 displayed significantly higher Smad2 phosphorylation than the parental cells. Note that the total Smad2 levels remain equal and unaltered in both cell types. In addition, the antisense-expressing cells still produce caveolin-1 but at dra-matically reduced levels (Fig. 11A). Quantitation of the Smad-2 phosphorylation levels in both cell types was also conducted by densitometry of the above results (Fig. 11B). Given the time frame in which Smad2 is activated in the antisense cells (i.e. unaltered at base-line but hyperactivated thereafter), caveolin-1 can act to physiologically dampen TGF-␤ signaling. DISCUSSION By using several independent and complementary approaches, we have examined the role of Cav-1 in TGF-␤/SMAD signaling. We demonstrated significant colocalization between the punctate distributions previously reported for both Cav-1 and T␤R-I and observed that T␤R-I, T␤R-II, and Smad-2 (but not Smad-4) cofractionate with Cav-1 in caveolae-enriched domains. Support for a direct interaction between Cav-1 and T␤R-I was provided by coimmunoprecipitation studies in both heterologous and endogenous settings. This interaction has functional consequences because Cav-1 was able to suppress TGF-␤-mediated transcriptional activation. In addition, we showed that Cav-1 diminishes the phosphorylation of Smad-2, disrupts its interaction with Smad-4 and prevents the nuclear translocation of Smad-2 in the ligand-activated state. This inhibition was mediated via an interaction between T␤R-I and the scaffolding domain of Cav-1, because only peptides derived from this region displayed potent inhibition of T␤R-I kinase activity in vitro and were able to disrupt the Cav-1/T␤R-I interaction in vivo. Furthermore, a Cav-1 mutant harboring a deletion of this domain was unable to either interact with the receptor or functionally suppress TGF-␤ signaling. We also demonstrated that the endogenous Cav-1/T␤R-I association occurs rapidly after ligand-activation and showed that antisensemediated down-regulation of caveolin-1 in NIH-3T3 cells was sufficient to hyperactivate TGF-␤-stimulated Smad-2 phosphorylation. Taken together, our results support a novel role for caveolin-1 as an important negative regulator of TGF-␤ signaling.
To date, a variety of physiological regulators of TGF-␤ have been identified. Many of these molecules (e.g. the p42/44 MAP kinases, SnoN/Ski oncoproteins, STAT proteins, etc) act to alter the function of the SMAD proteins by either modifying their phosphorylation state, disrupting their interaction with downstream partners, or preventing their capacity to affect transcription (11). The known repertoire of molecules affecting TGF-␤ signaling at the level of receptor is also expanding. Smad-6 and Smad-7, a functionally divergent subset of the SMAD protein family, can inhibit TGF-␤ signaling by directly interacting with T␤R-I (16 -18). Although Smad-6 is a more potent inhibitor of bone morphogentic protein signaling, Smad-7 seems to exclusively act on TGF-␤ pathways and in fact participates in an inhibitory feedback loop with T␤R-I, because its transcription is rapidly induced upon TGF-␤1 stimulation (17). Recently, a T␤R-I-related protein called BAMBI was cloned and shown to inhibit activin, bone morphogentic protein, and TGF-␤ signaling by acting as a pseudoreceptor (73). In addition, FKBP12 has been shown to negatively regulate T␤R-I by interacting with the receptor in the inactive state. This inhibition is releaved upon ligand binding, however, whereupon FKBP12 is released from the T␤R-I/-II complex (19,74).
The regulation of signaling mediated by Cav-1 and its kinetics of interaction with T␤R-I imply a distinct mechanism of TGF-␤ inhibition. In contrast to Smad-7, which is a TGF-␤inducible gene (17), Cav-1 expression remains unaffected in the first 80 min of TGF-␤1 treatment. Because the transcriptional response of Smad-7 occurs maximally at 60 min (14,17) and presumably longer for robust protein expression, the negative regulation mediated by Smad-7 is clearly different than that of Cav-1. The inhibitory effect of FKBP12 on T␤R-I is releaved FIG. 9. The region of caveolin-1 containing the scaffolding domain (residues 61-100) mediates the ability of Cav-1 to interact with T␤R-I w.t. and functionally inhibit TGF-␤ signaling. A, schematic diagram summarizing the domain organization of wild-type full-length caveolin-1 (␣-isoform residues 1-178) (Cav-1 FL) and the deletion mutant (lacking residues 61-100) (Cav-1 ⌬61-100). B, 293T cells were cotransfected with HA-tagged T␤R-I w.t. and either c-Myc-tagged Cav-1 FL or c-Myc-tagged Cav-1 ⌬61-100, as indicated. Cell lysates were prepared and immunoprecipitated (IP) with either anti-HA pAb or the negative control (preimmune serum pAb). Immunoprecipitates were resolved by SDS-PAGE and subjected to immunoblot analysis with anti-Myc mAb. Note that Cav-1 FL specifically interacts with immunoprecipitated T␤R-I w.t., whereas the Cav-1 ⌬61-100 mutant does not. C, NIH-3T3 cells grown to confluence were subjected to immunoprecipitation with anti-T␤R-I pAb. Where indicated, immunoprecipitates were coincubated with the caveolin-derived peptides (20 M), the sequences of which are displayed in Fig. 8A. Immunoblotting with anti-Cav-1 2297 mAb reveals the endogenous interaction between T␤R-I and Cav-1. The only peptides that disrupt this interaction are ones containing the Cav-1 scaffolding domain (i.e. peptides 61-101 and 82-101). DMSO, dimethyl sulfoxide. D, NIH-3T3 cells were transfected with the A3-lux/Fast-1 TGF-␤-responsive luciferase reporter system and a combination of T␤R-I (T204D), Cav-1 FL, Cav-1 ⌬61-100, or empty vector controls. Note that as in Fig. 4B, Cav-1 FL inhibits signaling mediated by the constitutively activated T␤R-I, whereas the Cav-1 ⌬61-100 mutant has no effect. Luciferase activities are expressed as ratios normalized to ␤-galactosidase activity, and each experimental value represented graphically is the average of three separate transfections performed in parallel. Error bars represent the observed S.D. upon ligand binding, indicating that its cellular function might be to control aberrant TGF-␤ signaling in the ligand-independent state (19). This is again in contrast to Cav-1, where its interaction with T␤R-I actually increases upon ligand activation and plateaus at 40 min. Various investigators have reported the phosphorylation kinetics of Smad-2 to occur gradually with a t1 ⁄2 of ϳ5-10 min, peaking at 20 -30 min (2,54). This response rate is inversely correlated with the observed gradual increase in Cav-1/T␤R-I interaction, leading credence to the possibility of a Cav-1-mediated dampening mechanism. In support of these observations, we showed that NIH-3T3 cells harboring an antisense Cav-1 construct behave similar to parental cells under serum-starved conditions, but display a 2-2.5-fold hyperactivation of Smad2 phosphorylation upon TGF-␤-stimulation.
At this time, we cannot rule out the presence of intervening proteins which mediate the Cav-1/T␤R-I interaction. Given the ability of peptides derived from the Cav-1 scaffolding domain to potently inhibit T␤R-I enzymatic activity in vitro and disrupt the Cav-1/T␤R-I complex in vivo, a direct interaction between the two proteins is likely. By using phage display libraries, we have previously identified ligands for the caveolin scaffolding domain. These peptide ligands or "caveolin-binding motifs" are as follows: ⌽X⌽XXXX⌽, ⌽XXXX⌽XX⌽, and ⌽X⌽XXXX⌽XX⌽, where ⌽ indicates an aromatic residue, Trp, Phe, or Tyr (75). More recent analysis indicates that motifs with a mixture of appropriately spaced aromatic and hydrophobic residues (i.e. Leu, Ile, and Val) could also serve to bind caveolin (76). Because functional caveolin-binding motifs have been deduced in tyrosine kinases, serine/threonine kinases, and endothelial nitric oxide synthase (reviewed in Ref. 41), the Cav-1/T␤R-I interac-tion could presumably occur in this manner. Indeed, there are several candidate caveolin-binding motifs in T␤R-I ( 424 YQLPYYDLV, 388 INMKHFESF, and 393 FESFKRADIY) (3). Most of these motifs are present in the intracellular kinase domain of the receptor (more specifically, subdomains IX and X as delineated in Ref. 77). Therefore, the inhibitory effects of Cav-1 on T␤R-I kinase activity and downstream signaling could be mediated by a direct interaction of the Cav-1 scaffolding domain with the T␤R-I kinase domain.
What physiological roles might the inhibitory interaction of Cav-1 with T␤R-I serve? Although we addressed only a subset of TGF-␤ growth factors (i.e. TGF-␤1), their pleiotropic effects on cellular physiology includes some intriguing highlights. In contrast to its role as an anti-mitogen in the early stages of tumor growth, TGF-␤1 appears to act as a promoter of metastasis and tumor cell migration in the later stages (78). By cooperating with matrix metalloproteinases on the deposition and remodeling of the extracellular matrix, TGF-␤ signals appear to promote tumor invasion and angiogenesis (79). These responses are in direct contrast to ones elicited by Cav-1. We have recently shown that Cav-1 inhibits lamellipod extension and cellular migration in a metastatic mammary adenocarcinoma cell line (80). In addition, VEGF, a potent angiogenesis FIG. 10. The interaction of endogenous caveolin-1 with T␤R-I is enhanced upon TGF-␤1 stimulation and occurs in a physiologically relevant time frame. Serum-starved NIH-3T3 cells grown to confluence were treated with TGF-␤1 (4 ng/ml) for the indicated times and subjected to immunoprecipitation (IP) with anti-T␤R-I. Immunoblotting with anti-Cav-1 2297 mAb reveals the kinetics of the interaction between endogenous T␤R-I and Cav-1. The Cav-1/T␤R-I interaction is minimal at base line but gradually increases, reaching a maximum at 40 min (top panel). Total Cav-1 levels and the phosphorylation state of SMAD2 (as determined by anti-phospho-SMAD2 pAb) are indicated in the middle and bottom panels, respectively. Given the time frame of SMAD2 phosphorylation (t1 ⁄2 ϭ ϳ5-10 min) (54), the interaction of Cav-1 receptor could serve as a dampening mechanism for TGF-␤ signaling.
FIG. 11. Antisense-mediated down-regulation of caveolin-1 protein levels is sufficient to hyperactivate TGF-␤1-induced SMAD phosphorylation. A, serum-starved NIH-3T3 cells (either parental or ones harboring the Cav-1 antisense construct) (72) were grown to confluence and treated with TGF-␤1 (4 ng/ml) for the indicated times. Cell lysates containing phosphatase inhibitors were prepared and subjected to immunoblot analysis with phospho-specific anti-Smad2 pAb. In the nonstimulated state (time 0), both cell types have an identically low Smad2 phosphorylation. Upon addition of TGF-␤1, the Cav-1 antisense cells show a distinct 2-2.5-fold hyperactivation of Smad2 activation. Total Smad2 and Cav-1 levels in both cell types are also shown. B, quantitation of the results shown in A were performed by densitometric analysis.
Caveolin-1 and TGF-␤/SMAD Signaling factor, is capable of down-regulating Cav-1 expression in an endothelial-derived cell line (81). Consequently, a loss of Cav-1 regulation on TGF-␤ signaling might be an important step in the progression to cellular migration and metastasis.
Furthermore, TGF-␤ signaling plays an extremely important role in cellular differentiation. The progression of Schwann cell, myocyte, adipocyte, endothelial cell, and other lineages are regulated by TGF-␤, and in many cases the attainment of a terminal phenotype depends on a cessation of TGF-␤ signaling (reviewed in Ref. 82). Cav-1 and other members of the caveolin family are up-regulated during such processes and expressed at high levels in terminally differentiated cells, including many of the TGF-␤-responsive lineages (83). For example, in the adipogenesis model system 3T3-L1, Cav-1 expression is upregulated 25-fold in the transition from 3T3-L1 fibroblasts to adipocytes (84). In contrast, TGF-␤ signaling has been shown to potently inhibit this adipocyte conversion (85). Therefore, the attenuation of TGF-␤ signals by Cav-1 could be an important mechanism for the controlled progression of developmental events.