Caveolin-1 Associates with TRAF2 to Form a Complex That Is Recruited to Tumor Necrosis Factor Receptors*

Tumor necrosis factor (TNF) receptor-associated factor (TRAF) 2 is an intracellular adapter protein, which, upon TNF stimulation, is directly recruited to the intracellular region of TNF receptor 2 (TNFR2) or indirectly, via TRADD, to the intracellular region of TNF receptor 1 (TNFR1). In cultured human umbilical vein endothelial cells, endogenous TRAF2 colocalizes with the membrane-organizing protein caveolin-1 at regions of enrichment subjacent to the plasma membrane as detected by confocal fluorescence microscopy. Both endogenous and transfected TRAF2 protein coimmunoprecipitate with caveolin-1 in the absence of ligand. Upon TNF treatment, the TRAF2-caveolin-1 complex transiently associates with TRADD, and upon overexpression of TNFR2, the TRAF2-caveolin-1 complex stably associates with and causes redistribution of this receptor as detected by confocal fluorescence microscopy. In human embryonic kidney 293 cells, which have minimal endogenous expression of caveolin-1, cotransfection of TRAF2 and caveolin-1 results in spontaneous association of these proteins which can further associate with and redistribute transfected TNFR2 molecules. The association of caveolin-1 with TNFR2 depends upon TRAF2. Cotransfection of caveolin-1 protein increases TRAF2 protein expression levels in HEK 293 cells, which correlates with enhancement of TNF and TRAF2 signaling, measured as transcription of a NF-κB promoter-reporter gene, although the caveolin-enhanced response to TNF is attenuated at higher caveolin levels. These findings suggest that intracellular distribution of activated TNF receptors may be regulated by caveolin-1 via its interaction with TRAF2.

TNF 1 is a critical mediator of innate and adaptive immunity (1,2). It exerts its biological effects through binding to either of two receptor molecules, designated TNFR1 (CD120a) and TNFR2 (CD120b) (3). TNF exists as a stable homotrimer, and ligand binding induces homotypic clustering of the transmembrane receptor proteins (4 -6). The cytoplasmic tails of the clustered receptors initiate signaling by recruitment of adapter proteins (7). Clustered TNFR1 binds TNF receptor-associated death domain (TRADD) protein. TRADD, in turn, binds receptor-interacting protein (RIP) and TNF receptor-associated factor (TRAF) 2 (8 -12). The TRADD⅐RIP⅐TRAF2 complex initiates new gene expression through protein kinase cascades that result in activation of the transcription factors NF-B and AP-1 (7,9,13,14). Clustered TNFR2 directly binds TRAF2, initiating the same signaling pathways; TRAF2 recruits TRAF1 to the TNFR2 complex, and it is not known if RIP is involved in signaling through this receptor (10,15,16).
Many receptor signaling pathways have been linked to cholesterol-and sphingomyelin-enriched invaginations of the plasma membrane called caveoli (17)(18)(19). The inner surface of these membrane regions is coated with a protein scaffolding formed by members of the caveolin family (caveolin-1, -2, and -3). Caveolins have also been associated with other membrane organelles, such as the Golgi, and they may play a role in localizing signaling molecules to these intracellular compartments (19). Such an association has been observed for endothelial nitric-oxide synthase (called eNOS; also called NOS-3), which can exist association with either plasma membrane or Golgi (20). Since TNFR1 can also reside either in the plasma membrane or in the Golgi of human umbilical vein endothelial cells (HUVEC) (21), we wondered if it too might associate with caveolin. Although we did not find a direct association of TNFR1 with caveolin, we did find an interaction of caveolin-1 with TRAF2. This association is independent of ligand and results in the recruitment of TRAF2-caveolin-1 complexes to either TRADD or TNFR2.

EXPERIMENTAL PROCEDURES
Materials--Recombinant human TNF-␣ was purchased from R&D Systems Inc. (Minneapolis, MN). Vent ® DNA polymerase, T4 DNA ligase, and all restriction endonucleases were purchased from New England Biolabs (Beverly, MA). Monoclonal antibody against green fluorescence protein (GFP) was purchased from CLONTECH (San Francisco, CA); monoclonal antibody against human TRAF2 and polyclonal antibody against caveolin-1 were purchased from BD Biosciences-PharMingen (San Diego, CA) and BD Biosciences-Transduction Laboratories (Lexington, KY). Protein G Plus/Protein A-agarose and Nonidet P-40 were purchased from Calbiochem (La Jolla, CA). * This work was supported in part by grants from the National Institutes of Health (to J. S. P.) and from the National Kidney Research Fund (to J. R. B.). 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  , and rabbit control serum IgG was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Peroxidase-conjugated donkey anti-mouse lgG and anti-rabbit lgG, Texas Red ® dye-conjugated donkey anti-rabbit lgG, and fluorescein isothiocyanate (FITC)-conjugated donkey anti-mouse lgG were purchased from Jackson ImmunoResearch Laboratories (West Grove, PA). Texas Red ® dye-conjugated swine anti-rabbit IgG and FITC-conjugated goat anti-mouse IgG were purchased from DAKO (Carpinteria, CA). All other reagents are from Bio-Rad or Sigma.
Plasmid Constructs-cDNAs encoding mouse TRAF1, full-length human TRAF2, and a dominant-negative of human TRAF2 (TRAF2DN,  which has 86 amino acids deleted from the N-terminal end) were each amplified by polymerase chain reaction (PCR) from mammalian expression constructs pRK-TRAF1-Flag and pRK-TRAF2-Flag (both gifts from Dr. V. Dixit, Genentech Inc., South San Francisco, CA). The amplified cDNA fragments were directly inserted between BamHI and KpnI sites of the expression vector pBK-CMV (Stratagene, La Jolla, CA) to generate pBK-CMV-TRAF1, pBK-CMV-TRAF2, and pBK-CMV-TRAF2DN. To generate enhanced green fluorescent protein (EGFP) fusion construct, the EGFP coding sequence minus its stop codon was first amplified from the plasmid pEGFP-N1 (CLONTECH) using 5Јoligonucleotide primer 5Ј-GTGAACCGTCAGATCCGCTAG-3Ј (based on the sequence of pEGFP-N1 from 575 to 595) and 3Ј primer 5Ј-CCATCT-TGCGCGCCTTGTACAGCTCGTTCCATGC-3Ј (with the native sequence of pEGFP-N1 from 1376 to 1396 underlined). The PCR-amplified cDNA fragment, containing the EGFP cassette, was gel-purified, digested with BamHI and BssHII, and directly inserted between the BamHI and BssHII sites of plasmids pBK-CMV-TRAF1, pBK-CMV-TRAF2, or pBK-CMV-TRAF2DN (in which the BssHII site is located three base pairs upstream of the ATG start codon of TRAF1 or TRAF2). To make TNFR2(Ϫ16), TNFR2(Ϫ37), and TNFR2(Ϫ59), which are Cterminal deletion constructs of wild type TNFR2, nucleotides encoding 16, 37, and 59 amino acids, respectively, of the C-terminal end of TNFR2 were removed from the cDNA encoding the full-length receptor (in plasmid pRCX-3 (Ref. 22)) and a new stop codon was introduced by PCR. Each of the TNFR2s with C-terminal truncations was then directly inserted back into the same expression vector as the full-length cDNA. A mammalian expression construct of dog caveolin-1 (pCB7caveolin-1) was a gift from Dr. W. C. Sessa (Yale University, New Haven, CT). The pBIIXLuc plasmid which contains two NF-B sites from the Ig-enhancer fused with a firefly luciferase encoding cDNA, and the plasmid p␤-actin-Rluc, which contains the ␤-actin promoter fused to a Renilla luciferase encoding cDNA, were both gifts from Dr. S. Ghosh (Yale University). The sequences of all constructs were confirmed by DNA sequencing.
Cell Culture-HUVEC were isolated and serially cultured as described previously on gelatin (J. T. Baker Inc., Phillipsburg, NJ)-coated tissue culture plastic (Falcon, Lincoln Park, NJ) in Medium 199 supplemented with 20% (v/v) fetal bovine serum, 200 M L-glutamine (all from Life Technologies, Inc.), 50 g/ml endothelial cell growth factor (Collaborative Biomedical Products, Bedford, MA), 100 g/ml porcine heparin (Sigma), and 100 units/ml penicillin and 100 g/ml streptomycin (from Life Technologies, Inc.) (23). Transient transfection of HUVEC was performed using a DEAE-dextran method as described previously (24). HEK 293 cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and were maintained in Eagle's minimum essential medium (Life Technologies, Inc.) supplemented with 10% (v/v) fetal bovine serum in a 5% CO 2 incubator at 37°C. Cells were seeded at a density of 60 -80% and transfected using a modified calcium phosphate method with 1-12 g of plasmid (25).
Immunoprecipitation-Immunoprecipitation of control or transfected cells was conducted as follows. Cell monolayers were washed with cold PBS and solubilized in 0.5 ml of lysis buffer containing protease inhibitors (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.5% Nonidet P-40, 10 mM NaF, 1 mM sodium orthovanadate, 10 g/ml leupeptin, 10 g/ml aprotinin, 10 g/ml trypsin/chymotrypsin inhibitor, 5 g/ml pepstatin A, 1 mM phenylmethylsulfonyl fluoride) for 1 h. Lysates were cleared by centrifugation in Eppendorf tubes, then precleared with Protein G Plus/Protein A-agarose beads with rocking for 1 h at 4°C. 5 g of the specific antibody was added to the precleared lysates for 2-16 h at 4°C, Protein G Plus/Protein A-agarose beads were used to absorb immunoprecipitates with rocking for 1-8 h at 4°C, and the beads were then washed four times with lysis buffer.
Immunoblotting--Cell lysates or immunoprecipitates were separated by SDS-polyacrylamide gel electrophoresis and electrophoretically transferred onto nitrocellulose membranes (Bio-Rad). The membranes were blocked in PBS with 0.05% Tween 20 and 5% dried milk (Nestle Food Co., Glendale, CA), probed with specific antibodies and peroxidase-conjugated goat anti mouse or rabbit antibody (1:10000 dilution) and exposed using the Supersignal ® West Femto Chemiluminescence Western blotting detection system (Pierce). Indirect Immunofluorescence and Confocal Microscopy-Transfected cells were seeded on human plasma fibronectin-or collagen type I (Sigma)coated glass coverslips placed in six-well culture dishes at a density of 5 ϫ 10 5 cells/well for HEK 293 and 2.5 ϫ 10 5 cells/well for HUVEC. The cells were then rinsed briefly in PBS and fixed in 3.7% paraformaldehyde for 10 min. The fixed cells were permeabilized in PBS containing 0.2% Triton X-100 for 10 min and blocked in PBS containing 0.2% bovine serum albumin for 10 min. After 1 h of incubation with specific antibody, the cells were washed and incubated with Texas Red ® dye-conjugated donkey or swine anti-rabbit lgG, and/or FITC-conjugated donkey or goat anti-mouse lgG (1:100 dilution) for 1 h. The coverslips were mounted onto glass slides, and immunofluorescence was observed with a Zeiss LSM-410 laser scanning microscope at 568 and 488 nm excitation, respectively.
NF-B Promoter Reporter Assay-Cell lysates of reporter gene-transfected cells were prepared and assayed using Promega Dual-Luciferase ® reporter assay system (Promega, Madison, WI) according to manufacturer's instruction. Luciferase activity was measured in triplicate using a Berthold (Schwarzwald, Germany) model LB9501 luminometer according to the manufacturer's instructions. Activity of the NF-B promoter-firefly luciferase reporter was normalized to the activity of the cytokine-unresponsive ␤-actin promoter Renilla luciferase reporter.

TRAF2 Associates with Caveolin-1 in HUVEC and HEK 293
Cells-Our first approach to study the role of caveolin-1 in the TNF signaling pathway was to examine HUVEC cell lysates for an association of caveolin-1 with various signaling molecules by immunoprecipitation followed by immunoblotting. We did not observe an association of caveolin-1 with TNFR1, TRADD, or TNFR2 (data not shown). However, as shown in Fig. 1A, caveolin-1 protein was found in immunoprecipitates of TRAF2 by immunoblotting, but not in mock immunoprecipitates using control serum. (A much smaller amount of caveolin-1 was seen at long exposures in the absence of specific antibody, consistent with a weak nonspecific association.) Furthermore, immunoprecipitates of caveolin-1 contained TRAF2 as detected by immunoblotting (Fig. 1B). Treatment of the cells with TNF prior to lysis did not affect the TRAF2-caveolin association, but did cause the TRAF2-caveolin-1 complex to transiently associate with TRADD (maximal 5 min). Interestingly, this time course is consistent with previous studies of HUVEC which demonstrated formation of a transient TNF-induced complex at the plasma membrane containing TNFR1, TRADD, and TRAF2 (26,27).
We next turned to confocal immunofluorescence microscopy to confirm these biochemical findings. Both TRAF2 and caveolin-1 are diffusely present throughout the cytoplasm of HUVEC ( Fig. 2), limiting any conclusions that may be drawn regarding association. However, in many (but not all) cells, caveolin-1 is focally enriched in regions located subjacent to the plasma membrane (Fig. 2B). Such regions are more common when the cultures reach confluence. Interestingly, TRAF2 is also enriched within these same regions (Fig. 2, A and C), supporting our immunoprecipitation results of an associate between TRAF2 and caveolin-1. We did not observe any redistribution of TRAF2 or caveolin-1 induced by TNF treatment (data not shown).
To further explore the interaction of TRAF2 and caveolin-1, we turned to HEK 293 cells, which contain a little endogenous caveolin-1 compared with the level found in HUVEC (Fig. 1C).
In these experiments, HEK 293 cells were cotransfected with Flag-tagged TRAF2 and caveolin-1, and the cell lysates were immunoprecipitated with antibody to Flag and then immunoblotted. As shown in Fig. 1D, exogenous caveolin-1 was readily detected in the immunocomplex precipitated from the cell lysates expressing both Flag-tagged TRAF2 and caveolin-1. Only a small amount of caveolin-1 was found in the anti-Flag immu-nocomplex prepared from cells transfected with caveolin-1 only, again attributable to the intrinsic stickiness of this protein. These results strongly support the findings first made in HUVEC that TRAF2 associates with caveolin-1 prior to ligand signaling.
We again employed fluorescence confocal microscopy to confirm the results observed by immunoprecipitation. HEK 293 cells were transfected with GFP-TRAF2 alone or, were cotransfected with caveolin-1 and GFP-TRAF2. As shown in Fig. 3, GFP-TRAF2 alone exhibited diffuse fluorescence throughout the cytoplasm (Fig. 3A) (28). Caveolin-1 is not detectable in HEK 293 cells by immunofluorescence in the absence of transfection with this protein. Transfection of caveolin-1 alone showed diffuse cytosolic staining (data not shown). Cotransfection of caveolin-1 and GFP-TRAF2 induced a redistribution of TRAF2 into a punctuate pattern (Fig. 3B). To explore the specificity of this interaction, we transfected HEK 293 cells with TRAF1 instead of TRAF2. Transfection with GFP-TRAF1 alone, like GFP-TRAF2, resulted in diffused cytosolic fluorescence (Fig. 3C). However, in this case, caveolin-1 did not cause . GFP fluorescent signal (green) was visualized directly and TNFR2 and endogenous caveolin-1 were visualized by indirect immunofluorescence (red) staining using antibody to Flag tag for TNFR2 or antibody to caveolin-1. Experiments were performed on 5-10 cells in three independent experiments. The micrographs are representative of more than 75% of the cells observed. a redistribution of GFP-TRAF1 (Fig. 3D).
To further explore the effects of TNF receptor signaling, we examined HUVEC transfected with GFP-TRAF2 and TNF receptors. In single transfected cells, GFP-TRAF2 was evenly distributed throughout cytoplasm. Once again, endogenous caveolin-1 was also observed to be diffusely expressed, except where it sometimes coalesced along regions of the plasma membrane to form a characteristic "patch" pattern (Fig. 3A). In contrast to our findings with endogenous TRAF2, the GFP-TRAF2 signal was too bright to observe enrichment of the patch regions. We also treated the single transfected cells with TNF, but this treatment did not cause any obvious redistribution of GFP-TRAF2 or of caveolin-1 (data not shown), similar to our findings with endogenous TRAF2. We then proceeded to overexpress TNF receptors in the same cells. We have reported previously that transfected TNFR1 does not escape from the Golgi apparatus and desensitizes cell signaling (22). In contrast, TNFR2 is readily expressed at the plasma membrane and does not desensitize signaling. Therefore, we overexpressed TNFR2 receptor in the GFP-TRAF2-expressing HU-VEC. This maneuver causes ligand-independent receptor clustering and stable recruitment of TRAF2 to the clustered receptors (Fig. 4). Specifically, in cells transfected only with TNFR2, the receptor localized to the perinuclear region (probably Golgi and endoplasmic reticulum) as well as the plasma membrane (Fig. 4B), and isolated receptor did not show significant overlap with caveolin-1 expression. In HUVEC cells transfected with GFP-TRAF2 and Flag-TNFR2, GFP-TRAF2 fluorescence was both redistributed to plasma membrane and clustered into a punctate pattern in the cytoplasm that was largely coincident with that of transfected TNFR2 stained with antibody to Flag (Fig. 4C). Interestingly, when the cells cotransfected both with GFP-TRAF2 and TNFR2 were stained with antibody to caveolin-1, the subcellular localization of GFP-TRAF2 was now clearly colocalized with that of caveolin-1 (Fig.  4D), indicating that both TRAF2 and caveolin-1 were corecruited to the TNFR2 receptor. In other words, the redistribu-tion of TRAF2 induced by overexpressed TNFR2 allowed a clear demonstration of corecruitment of caveolin-1 to the activated TNF receptor in HUVEC.
The association of TNFR2, TRAF2, and caveolin-1 observed in HUVEC cells was also detected in HEK 293 cells by immunofluorescence (data not shown) and further analyzed by immunoprecipitation (Fig. 5). In these experiments, Flag-tagged TNFR2 and caveolin-1 were coexpressed with or without GFP-TRAF2. In the presence of TRAF2, TNFR2 was found to significantly associate with caveolin-1 (Fig. 5, lane 4), whereas, in the absence of TRAF2, much less interaction of TNFR2 with caveolin-1 was detected (Fig. 5, lane 3). These findings strongly suggest that the interaction of TNFR2 with caveolin-1 is mediated via TRAF2. The role of TRAF2 in linking caveolin-1 to TNFR2 was further supported by a mutational analysis of the TRAF2 binding region on the C-terminal region of TNFR2. To do this, we made a series of Flag-tagged TNFR2 proteins in which deletions were constructed by removing different numbers of amino acid residues from the C-terminal end of the receptor (Fig. 6A). These C-terminal deletions had no significant effect on intracellular distribution of the receptors, all of which localized both to plasma membrane and to cytoplasm concentrated in the perinuclear area (data not shown). The capacity of TNFR2 mutants to interact with TRAF2 was then examined by cotransfection of GFP-TRAF2 with individual wild type or mutant TNFR2 molecules (Fig. 6B). Like wild type TNFR2, coexpression of TNFR2(Ϫ16) with TRAF2 caused a redistribution of TRAF2, consistent with the fact that TNFR2(Ϫ16) retains the TRAF2 binding region (Fig. 6, B and  C). In contrast, TRAF2 remained evenly distributed in the cells in which TNFR2(Ϫ37) or TNFR2(Ϫ59) was expressed, indicating that loss of the ability of TNFR2(Ϫ37) and TNFR2(Ϫ59) to associate with TRAF2 (Fig. 6, D and E). Analysis of the same transfectants by indirect immunofluorescence showed that caveolin-1 was only redistributed by TNFR2 molecules (wild type and TNFR2(Ϫ16)), which are capable of interacting with TRAF2 (Fig. 7, A and B). In contrast, the distribution of caveolin-1 in cells transfected with TNFR2(Ϫ37) or TNFR2(Ϫ59) did not show any receptor association (Fig. 7, C and D). Similar results were also found in HEK 293 cells transiently transfected with GFP-TRAF2, caveolin-1, and different Flag-tagged TNFR2 and its deletion mutants (data not shown). Collectively, these data demonstrate that TRAF2 binding to activated TNFRs results in corecruitment of caveolin-1 into the receptor complex.
Effect of Caveolin-1 on TRAF2 Expression Levels and Signaling-Since caveolin-1 interacts with TRAF2 and TRAF2 plays critical roles in TNF signaling via NF-B pathway (7, 15), a possible functional linkage between interaction of TRAF2 and caveolin-1 and TRAF2-mediated NF-B activation was investigated. First, protein levels of GFP-TRAF2 in HEK 293 cells cotransfected with a constant amount of GFP-TRAF2 cDNA and varying amounts of caveolin-1 cDNA were examined by immunoblotting (Fig. 8). Under the condition of equal protein loading, the amount of TRAF2 protein was significantly increased by coexpression of caveolin-1 in a dose-dependent manner, even though the amount of cDNA encoding TRAF2 was held constant. In a final series of experiments, we examined the influence of caveolin-1 expression on TNF induced NF-B activation in HEK 293 using a NF-B promoter luciferase-based reporter system. At a suboptimal concentration of TNF (1 ng/ ml), the effect of caveolin-1 on activation of NF-B was biphasic, being initially augmented by increasing expression of caveolin-1 and then decreased at the highest level of transfected caveolin cDNA. The peak of this enhancement was about 2-3-fold over the cells not expressing exogenous caveolin-1 (Fig. 9A). At supra-optimal TNF concentrations (100 ng/ml), the effect of caveolin-1 was no longer observed (data not shown). However, the enhancement of TNF-induced NF-B activity was mimicked by ligand-independent increases of NF-B activity triggered by overexpression of TRAF2, but not with an inactive form of TRAF2 (TRAF2DN) (Fig. 9B). In this case, increasing caveolin-1 simply increased TRAF2 signaling without showing a diminution at higher levels. Thus the interaction of caveolin-1 with TRAF2 can potentiate TNF signaling mediated by TRAF2. It seems likely that this effect is explained by the increased level of TRAF2 protein observed in the presence of caveolin-1.
FIG. 6. Subcellular localization of GFP-tagged TRAF2 in HUVEC transfected with wild type TNFR2 and its deletion mutants. A, schematic of a set of deletions of TNFR2 generated by removing various lengths of C-terminal amino acid residues as indicated by the negative number. Also indicated is the filled rectangular area responsible for TRAF binding. B-E, representative thin section confocal fluorescence micrographs of HUVEC cells transiently transfected with 6 g of pBK-CMV-GFP-TRAF2 together with 6 g of pRCX-Flag-TNFR2 (A), pRCX-Flag-TNFR2(Ϫ16) (B), pRCX-Flag-TNFR2(Ϫ37) (C), and pRCX-Flag-TNFR2(Ϫ59) (D) respectively. GFP fluorescent signal was visualized directly, and TNFR2 was detected by indirect immunofluorescence (red) staining using anti-Flag primary antibody under a confocal microscope. The experiments were performed on 5-10 cells in three independent experiments. The micrographs are representative of more than 75% of the cells observed.

DISCUSSION
In this report, we describe an association of the membraneorganizing coat protein, caveolin-1, with the signal transducing adapter protein TRAF2. We demonstrate the association between endogenous TRAF2 and caveolin-1 in HUVEC and extend these findings using transfected proteins in HUVEC and HEK 293 cells. These proteins are associated in unstimulated HUVEC or in cotrasfected HEK 293. Upon receptor activation, either by TNF induction or by receptor overexpression, caveolin-1 is recruited to the receptor signaling complex through its interaction with TRAF2.
TRAF2 belongs to a family of adapter proteins including TRAF1, -2, -3, -4, -5, and -6 (29,30). These proteins link clustering of receptors belonging to the TNF receptor family and to the Toll/IL-1 receptor family to downstream signaling events such as activation of the transcription factors NF-B and AP-1. The sequences and in some cases, the partial threedimensional structure of TRAF proteins have been solved (10,(31)(32)(33)(34)(35)(36). Caveolin-1 has been observed to bind to various target proteins through consensus binding sequences ØXØXXXXØ and ØXXXXØXXØ, where Ø is aromatic amino acid Trp, Phe, or Tyr (37). The primary sequence of TRAF2 obtains a possible caveolin binding motif within its conserved TRAF domain (Phe 354 -Ile-Trp-Lys-Ile-Ser-Asp-Phe 361 ). A similar sequence is present in TRAF6, but not in TRAF1. This may explain why TRAF1, which we used as a specificity control for caveolin-1 association, did not FIG. 9. Effect of expression of caveolin-1 in HEK 293 on NF-B reporter activity. A, TNF-induced NF-B reporter activity was measured in HEK 293 cells transfected transiently with 0.5 g of pBIIXLuc and 0.1 g of p␤-actin-Rluc together with increased amount of pBC-caveolin-1 (0, 1, 2, and 3 g, respectively). Then the transfected cells were divided into two replicate cultures. One culture was stimulated with 1 ng/ml TNF, whereas the other culture remained unstimulated. The firefly and Renilla luciferase activity from the transfected cells were measured as described under "Experimental Procedures." Shown are ratios of normalized NF-B reporter activity with treatment by TNF versus without treatment by TNF. Similar results were obtained in two independent experiments. B, TRAF2-mediated receptor-independent NF-B activity was measured in HEK 293 cells were transiently transfected with 0.5 g of pBIIXLuc, 0.1 g of p␤-actin-Rluc, and 1 g of pBK-CMV, or pBK-CMV-GFP-TRAF2 or pBK-CMV-GFP-TRAFDN together with increased amount of pBC-caveolin-1 (0, 1.5, 3.0, and 4.5 g, respectively). The cells were harvested after 24-h transfection, and firefly and Renilla luciferase activity were measured as described under "Experimental Procedures." Shown are the representative results obtained in three independent experiments. interact with caveolin-1 in cotransfected HEK 293 cells. Substitution of this putative caveolin binding site of TRAF2 with the nonfunctional sequence from TRAF1 appears to reduce but not eliminate the association with caveolin-1. 2 Therefore, although this sequence may contribute to caveolin-1 association, it does not appear to fully account for it. Furthermore, the binding site is located within a region that is near the contact region of TRAF2 with the receptor (35), and it is not clear whether caveolin-1 binding to this site and receptor binding are simultaneously possible.
Our findings, which establish a biochemical protein-protein association, raise a number of questions about the functional implication of caveolin-1 binding to TRAF2. TRAF2 signaling may be terminated, in some systems, by TRAF2 degradation (38,39). In transfected HEK 293 cells, the presence of caveolin-1 increases TRAF2 expression levels. This effect could be due to TRAF2 protein stabilization by prevention of degradation. If true, then caveolin-1 binding to TRAF2 could prolong TNF signaling. We did observe a significant enhancement of TNF signaling in the presence of caveolin-1, but the effect was decreased at the higher levels of caveolin-1 expression and was not seen at supra-optimal TNF concentrations. In any event, caveolin-1 is not required for TRAF2 signaling since TRAF2 overexpression can activate a NF-B reporter gene in the absence of significant caveolin expression (e.g. in transfected HEK 293 cells). We do not know if caveolin-1 plays a role in TNF-mediated apoptosis since HEK 293 cells cannot be killed by TNF whether or not caveolin-1 is present. 3 An effect on apoptosis seems unlikely in any event since TRAF2 is not involved in the TNF-activated death pathway (9). Overall, our data suggest that caveolin-1 may modulate TRAF2-mediated signaling, but it is not essential for TNF responses.
A more likely role for the association of caveolin-1 and TRAF2 is the targeting of activated TNF receptor complexes to caveolin-enriched membrane compartments such as caveoli. Many other receptor systems appear to localize to caveoli, which are defined by their cholesterol-and sphingomyelinenriched lipid environments as well as by their morphological features (18). This microenvironment may alter the efficiency or receptor signaling and may be uniquely suited to allow cross-regulation by different types of hormone and cytokine receptors. If caveolin functions to recruit activated TNF receptors to a specific signaling environment, then excess free caveolin may inhibit TNF receptor recruitment to caveoli or related structures. This may explain why high levels of transfected caveolin-1 decreases ligand-induced responses but not signals caused by overexpression of TRAF2. It may also explain why TNF addition fails to cause a redistribution of TRAF2 molecules within the cell, i.e. it is the receptor rather than the TRAF molecule that is redistributed by ligand binding. To date, there are no reports that TNF-induced transcription requires the microenvironment of the caveoli, but TNF induced apoptosis in U937 cells have been reported to be initiated in caveoli-like regions (40). As we have noted above, TRAF2 has not been implicated in TNF-induced apoptosis (9) and may in fact inhibit through the induced expression of antiapoptotic proteins and through the recruitment of inhibitor of apoptosis proteins to the receptor complex (13,15,41). Activated TNF receptors are rapidly removed from the membrane by endocytosis, but this is thought to involve clathrin-coated pits and vesicles rather than caveoli (21,42). Further studies will be needed to determine whether caveolin-1 alters this pattern of receptor transport.
In summary, we have made the unexpected finding that the TNF signaling adapter protein TRAF2 exists in a complex with caveolin-1, allowing caveolin-1 to be recruited into activated receptor signaling complexes. The functional significance of this finding is unclear, and may relate to TRAF2 stabilization and/or to targeting of activated receptors within the cells. Indeed, our results may suggest that from a cell biology perspective, it is more accurate to describe the recruitment of ligandreceptor complexes to TRAF2-caveollin-1 binding sites than it is to describe the recruitment of TRAF2 to the occupied receptor.