Caveolin-3 Undergoes SUMOylation by the SUMO E3 Ligase PIASy

Caveolin (Cav) proteins in the plasma membrane have numerous binding partners, but the determinants of these interactions are poorly understood. We show here that Cav-3 has a small ubiquitin-like modifier (SUMO) consensus motif (ΨKX(D/E, where Ψ is a hydrophobic residue)) near the scaffolding domain and that Cav-3 is SUMOylated in a manner that is enhanced by the SUMO E3 ligase PIASy (protein inhibitor of activated STAT-y). Site-directed mutagenesis revealed that the consensus site lysine is the preferred SUMOylation site but that mutation of all lysines is required to abolish SUMOylation. Co-expression of a SUMOylation-deficient mutant of Cav-3 with β-adrenergic receptors (βARs) alters the expression level of β2ARs but not β1ARs following agonist stimulation, thus implicating Cav-3 SUMOylation in the mechanisms for β2AR but not β1AR desensitization. Expression of endothelial nitric-oxide synthase (NOS3) was not altered by the SUMOylation-deficient mutant. Thus, SUMOylation is a covalent modification of caveolins that influence the regulation of certain signaling partners.

lin-binding motif," but how this interaction is dynamically regulated is poorly understood. Because of their rapid kinetics and reversibility, control by post-translational modifications is a likely mechanism; however, no regulatory post-translational modifications have been identified on Cav-3. We searched for motifs important for post-translational modifications in the amino acid sequence of Cav-3 and found a conserved, SUMO consensus motif in Cav-3 and -1 but not Cav-2 (Fig. 1A).
SUMOylation is the reversible modification of proteins by small ubiquitin-like modifier (SUMO) proteins and can regulate protein-protein interactions by masking or creating binding surfaces on target proteins (6,7). SUMOylation, initially thought to occur only in nuclear or perinuclear compartments (8), also regulates cytoplasmic and plasma membrane proteins (9 -15). TGF␤ induces SUMOylation of the Type I TGF␤ receptors (16), which localize to caveolae and interact with Cav (17)(18). Other SUMOylated signaling proteins that interact with Cavs or localize to caveolae include Kv1.5, PTP1B, RGS-Rz, and PDE4D5 (19 -24). Thus, modification by SUMO may contribute to the binding and regulation of proteins that interact with Cavs. Here we define properties of the SUMOylation of caveolin-3 and the contribution of this SUMOylation to the expression of ␤-adrenergic receptors (␤ARs).
Site-directed Mutagenesis-The single and multiple Lys to Arg (KR) mutations were introduced into WT Cav-3-V5 using Multi-Site and Lightning QuikChange site-directed mutagenesis kits (Stratagene). The sequences for primers designed for creation of the KR mutants are available upon request. Cav-3-V5-K7R was used to create a series of Cav-3-V5 RK mutants. The sequences for primers designed for creation of the RK mutants are available upon request. All plasmids were sequenced to validate the presence of the desired mutations.
Cell Culture and Transfection-Human embryonic kidney cells (HEK AD-293) were obtained from Stratagene (catalog no. 240085) and cultured in a 37°C, 5.0% CO 2 incubator. Cells were grown in "HEK medium" (DMEM with 4.5 g/liter glucose, L-glutamine, and sodium pyruvate, supplemented with 10% FBS) without antibiotics. HEK AD-293 cells are derived from HEK 293 cells. Confluent HEK AD-293 cells were transfected 2-3 days after plating in 6-well plates. 4.0 g of plasmid DNA and 9 l of Lipofectamine 2000 (Invitrogen) were each diluted separately in 250 l of OptiMEM (Invitrogen) before being combined, incubated for 20 min at room temperature, and diluted with 500 l of HEK medium. Transfection complexes (1-ml total volume) were pipetted onto cells in 6-well plates to which 2 ml of fresh HEK medium had been added. Cells were incubated with transfection complexes overnight, and medium was replaced with 3 ml of fresh HEK medium the following day. Except where stated, cells were analyzed 48 h post-transfection. For co-expression of the SUMOylation machinery with Cav-3-V5, the following ratio was used: Cav-3-V5/SUMO-3/Ubc9/ PIASy ϭ 1:1:0.5:0.15-0.25. For the experiments that assessed the dose dependence of PIASy (Fig. 3B, lanes 3-7), the range was 0 -400 ng of plasmid. FLAG-␤ 2 AR, FLAG-␤ 1 AR, or eNOS was co-expressed in a 1:1 ratio with Cav-3-V5 using the same ratios of SUMOylation machinery components. The total DNA for all transfections was normalized to 4.0 g with empty pcDNA3.1 vector.
Immunoprecipitation and Immunoblotting-Transfected HEK AD-293 cells were washed with ice-cold PBS and scraped into 500 -800 l of ice-cold "in vivo SUMOylation" lysis buffer (20 mM Tris-HCl, pH 7.8, 150 mM NaCl, 0.1% SDS, 0.5% deoxycholate, 0.5% Triton X-100, 1 mM EDTA) supplemented with 20 mM NEM (prepared fresh) and Halt protease inhibitor mixture (Pierce) and pipetted into prechilled 1.5-ml Eppendorf tubes. Cells were incubated on ice for 30 min and sonicated in an ice water bath (3 ϫ 10 s). Lysates were clarified by centrifugation (5000 ϫ g for 15 min at 4°C), and protein concentrations were normalized by dilution with lysis buffer after performing BCA protein assays (Pierce). Clarified lysates were either directly resolved by SDS-PAGE and immunoblotted with appropriate antibodies or subjected to immunoprecipitation. Lysates were diluted with 4ϫ LDS sample loading buffer (Invitrogen) supplemented with 100 mM DTT and heated (except where stated) to 80°C for 5 min before being resolved by NuPAGE BisTris 4 -12% 12-well gels. Proteins were then transferred for 1.5-2.5 h at 55 V onto PVDF membranes that were immediately blocked for 1 h at room temperature in 4% milk, 0.1% TBST. In general, primary antibodies were diluted at 1:1000 in 4% milk, 0.1% TBST except for V5-HRP (1:5000), Cav-3 (1:2000), and PIASy (1:2000). For anti-SUMO-1 and anti-SUMO-2/3 antibodies, 5% BSA, 0.1% TBST was used for blocking and dilution instead of milk to reduce background signal. PVDF membranes were incubated overnight at 4°C with primary antibodies and washed at least three times for 10 min each with 0.1% TBST before incubation with secondary antibodies for 45-50 min at room temperature. Membranes were then washed (4 ϫ 10 min) with 0.1% TBST and detected by Amersham Biosciences ECL (GE Healthcare; catalog no. RPN2135V2) or SuperSignal West Dura (Pierce; catalog no. 34075). Images were collected with a Sensicam QE High Performance CCD camera (Cooke Corp.) mounted on an Epi Chemi II Darkroom (UVP BioImaging Systems) and analyzed using LabWorks 4.0 image acquisition and analysis software (UVP BioImaging Systems).
For immunoprecipitations, 300 -500 l of clarified lysates were precleared with 40 l of Protein G-agarose (Roche Applied Science), incubated overnight at 4°C with primary antibodies (1-2 g), followed by incubation with 40 l of Protein G-agarose for 1 h. Complexes were centrifuged at 10,000 ϫ g at 4°C for 1 min; supernatants were transferred to prechilled 1.5-ml Eppendorf tubes. The beads were washed four times with lysis buffer and eluted by heating to 80 -90°C for 5 min in 50 -100 l of 2ϫ LDS buffer with 100 mM DTT. Immunoprecipitates were resolved by SDS-PAGE and immunoblotted with appropriate antibodies. Immunoprecipitations with anti-FLAG M2 and anti-FLAG M2-agarose (Sigma) were performed per the manufacturer's instructions with FLAG immunoprecipita-tion lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100) supplemented with 20 mM NEM, 10% glycerol, and Halt protease inhibitor mixture. Isolation of adult rat cardiac myocytes for immunoprecipitation with anti-Cav-3, anti-SUMO-1, and anti-SUMO-2/3 antibodies was performed as described (25). Myocytes were washed with ice-cold PBS and incubated on ice with lysis buffer (0.2% SDS, 0.5% Nonidet P-40, 15 mM MgCl 2 , 1 mM DTT) supplemented with Halt protease inhibitor mixture and 20 mM NEM. Lysates were sonicated, clarified by centrifugation, and incubated overnight with primary antibodies at 4°C and processed by the Protein G-agarose method as described above.
In Vitro SUMOylation Assay-An in vitro protein synthesis kit (Qiagen EasyXpress Insect Kit II) was used to express recombinant Cav-3-V5 per the manufacturer's instructions. 2 l of protein synthesis reactions containing Cav-3-V5 were used directly in 20-l in vitro SUMOylation reactions (BioMol International) consisting of 2 l of 10ϫ SUMOylation buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 2 mM ATP), 1 l each of 20ϫ SUMO E1 and SUMO E2 enzyme solutions, and recombinant SUMO-1 or SUMO-3 proteins. Reactions were incubated at 37°C for 1 h and quenched with 2ϫ LDS sample buffer with 100 mM DTT. 10 l of the reaction product was analyzed by immunoblotting with anti-V5-HRP antibodies.
Expression and Purification of SUMOylated Caveolin-3 in E. coli-For production of SUMOylated Cav-3, BL21(DE3) competent cells were sequentially transformed with three E. coli expression plasmids: pKRSUMO, pBADE12, and pST39-Cav-3-V5. For negative controls, cells were transformed with pST39-Cav-3-V5 alone or with pKRSUMO and pBADE12. Transformed cells were stored as glycerol stocks at Ϫ80°C. For protein expression, transformed cells were grown at 37°C in LB supplemented with antibiotics (e.g. Amp, Cm, and Kan). The A 600 was monitored, and protein expression was induced at an A 600 of 0.5-0.6 for 3 h. Cultures were pelleted and resuspended in TNE buffer (25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA). Purifications were performed with Ni 2ϩ -NTA-agarose (Qiagen) and anti-V5-agarose (Bethyl Laboratories) resins per the manufacturer's instructions.
Transfection of HEK cells with increasing amounts of PIASy yielded increased modification of Cav-3-V5 by SUMO-3 (Fig.  3B, lanes 4 -7). A ladder of bands corresponding to Cav-3 modified at multiple lysine residues or by poly-SUMO-3 chains was seen (*, Fig. 3B) at 65 kDa and higher in addition to the 40 and 50 kDa bands observed in vitro (Fig. 2, B-D). Transfecting Ͼ400 ng of PIASy had no further effect on Cav-3-V5 SUMOylation and produced cell toxicity (data not shown).
Mutation of the Consensus Site Lysine Reduced but Did Not Abolish SUMOylation of Caveolin-3-We next sought to identify the lysine(s) modified and to create a SUMOylation-deficient mutant that would facilitate identification of biological functions influenced by SUMOylation of Cav-3. We initially mutated the consensus site lysine in Cav-3-V5 (Fig. 4A) to arginine (K38R) and assessed SUMOylation (Fig. 4B). K38R reduced but did not abolish SUMOylation of Cav-3-V5, indicating that either 1) multiple lysine residues are modified or 2) if the consensus site is mutated, a different lysine can be modified.
Identification of a SUMOylation-deficient Mutant-We mutated the seven lysines of Cav-3-V5 to arginine one at a time (KR mutants) and assessed SUMOylation in transfected HEK cells (Fig. 4C). K38R reduced SUMOylation relative to WT Cav-3-V5, but SUMOylation of other KR mutants was equal to or greater than that of WT. It is unclear whether this is due to increased SUMOylation or differences in stability or expression of the mutants (e.g. K30R and K69R). Lys 38 appeared to be the major site of SUMOylation; however, other sites can be modi- fied, suggesting that two or more lysines might need to be mutated. Two rounds of multisite mutagenesis were performed to generate multiple KR mutants. None of these mutants completely prevented SUMOylation of Cav-3 (Fig. 4D, lanes 6 -10). SUMOylation was abolished when all seven lysines were mutated (Cav-3-V5-K7R; Fig. 4, C (lane 7) and D (lanes 1-4)), implying that unavailability of the preferred SUMOylation site(s) allows other lysines to be SUMOylated.
Lys 38 Is the Preferred SUMOylation Site-To define precisely which lysines are potential versus preferred SUMOylation sites and determine their relative ability to be SUMOylated, we used Cav-3-V5-K7R (Fig. 5A) to create a series of RK mutants. HEK cells were transfected with WT and the Cav-3-V5-K7R mutant in parallel with the RK mutants (Fig. 5B). Immunoblotting with anti-V5-HRP antibodies detected SUMOylation at each lysine, but reintroduction of the consensus site lysine in the R38K mutant most effectively restored the SUMOylation of Cav-3-V5 (Fig. 5B, lane 6). Moreover, R38K was the only mutant that, akin to WT Cav-3-V5, was poly-SUMOylated (*, Fig.  5B, lane 1 versus lane 6), suggesting that PIASy stimulates SUMOylation by poly-SUMO-3 chains at this site.
To confirm that PIASy was required for SUMOylation at Lys 38 , we transfected the four N-terminal RK mutants with and without co-expression of the SUMOylation machinery (SUMO-3, Ubc9, and PIASy). Immunoblotting with anti-V5-HRP, anti-PIASy, and anti-GAPDH revealed that co-expression of the SUMOylation machinery had no effect on the low level of modification of R15K, R20K, or R30K but strongly enhanced modification of R38K (Fig. 5C, lane 6), which closely resembled WT (*, Fig. 5C, lane 2 versus lane 6). Thus, PIASy is necessary for SUMOylation, including poly-SUMOylation, of Lys 38 .

JOURNAL OF BIOLOGICAL CHEMISTRY 14835
The mechanisms of agonist-induced desensitization of the ␤ 2 AR have been extensively studied in HEK 293 cells (37,38). ␤-Adrenergic agonists (norepinephrine and epinephrine) are present in fetal bovine serum (FBS) (39). Therefore, during these experiments, the FLAG-␤ 2 AR-transfected HEK cells were exposed to ␤-agonists by use of FBS in the culture media.
The data shown thus far suggest that modification of Cav-3 by SUMO-3 may alter FLAG-␤ 2 AR expression levels by inhibiting, whereas the SUMOylation-deficient Cav-3 mutant may promote agonist-induced desensitization of ␤ 2 ARs. To determine whether the effect of the K7R mutant of Cav-3 is specific for ␤ 2 AR, we assessed the impact of K7R on expression of eNOS (NOS3), a well studied, caveolin-regulated protein (40,41). HEK 293 cells were transfected with eNOS alone or with WT or Cav-3-V5-K7R and with or without the SUMOylation machinery for 24 h (Fig. 6C). The expression level of eNOS was unaffected by expression of either WT Cav-3 or Cav-3-V5-K7R, suggesting that the effect on FLAG-␤ 2 AR is not a nonspecific inhibition of protein expression caused by the Cav-3-V5-K7R mutant.
Propranolol Attenuates the Effect of Cav-3-V5-K7R on FLAG-␤ 2 AR Expression Levels-We used the ␤AR antagonist propranolol to determine whether the decrease in FLAG-␤ 2 AR expression is a result of ␤-agonist stimulation of FLAG-␤ 2 AR by catecholamines present in the culture media. HEK 293 cells transfected with the SUMOylation machinery (SUMO-3, Ubc9, and PIASy) were co-transfected with either FLAG-␤ 2 AR alone, WT Cav-3-V5, or the K7R mutant in the absence or presence of 2 M propranolol. Treatment with propranolol blunted the effect of the Cav-3-V5-K7R mutant on FLAG-␤ 2 AR expression (Fig. 7B, lanes 1-3). Incubation of cells in media with reduced serum (1%) had a similar effect as did propranolol (Fig. 7B, lanes  4 -6). In each case, WT Cav-3-V5 increased FLAG-␤ 2 AR expression levels (Fig. 7B, lanes 2 and 5), presumably by inhibiting the agonist-induced desensitization of the receptor.
To quantify this effect, HEK 293 cells were co-transfected with FLAG-␤ 2 AR alone, WT Cav-3-V5, or the K7R mutant and treated with or without 2 M propranolol; FLAG-HRP immunoblots of anti-FLAG immunoprecipitates were then quantified by densitometry (Fig. 7, C and D). Immunoprecipitates and protein samples are typically heated prior to analysis by SDS-PAGE; however, this can cause self-aggregation of membrane proteins, such as G-protein-coupled receptors (43). To achieve accurate quantification, we did not heat the anti-FLAG immunoprecipitates prior to SDS-PAGE to reduce the formation of Ͼ100-kDa FLAG-␤ 2 AR aggregates (Fig. 7, A and B). We found that K7R reduced FLAG-␤ 2 AR expression levels by about 50% (p ϭ 0.005) in vehicle-treated cells and that treatment with propranolol returned the expression to control levels (Fig. 7, C and D). WT Cav-3-V5 had the opposite effect, increasing FLAG-␤ 2 AR expression by ϳ2-fold; expression was further increased by propranolol. Cav-3 and its ability to undergo SUMOylation thus appear to modulate the agonist-induced desensitization of ␤ 2 ARs.

DISCUSSION
These studies demonstrate for the first time that a caveolin, Cav-3, is post-translationally modified by SUMO proteins in vitro and in vivo. SUMOylation of Cav-3 appears to be a novel regulatory mechanism for agonist-induced desensitization of ␤ 2 ARs. In vitro SUMOylation assays and co-expression of recombinant Cav-3 in E. coli with the SUMOylation machinery showed that conjugation with SUMO-1 or SUMO-3 is possible. However, in mammalian cells, conjugation is specific for SUMO-3. In E. coli transformed with the SUMOylation machinery, Cav-3 was modified at multiple sites or by poly-SUMO-1 (Fig. 2, B-D). There is debate regarding the ability of SUMO-1 to form chains in vivo (44); this possibility cannot be ruled out because SENPs (which reverse SUMO conjugation and edit poly-SUMO chains) are absent in E. coli. The tandem SUMO-interacting motifs (SIMs) of Cav-3 (supplemental Fig. S1, C and D) may assist in the modification by poly-SUMO chains by recruiting Ubc9 thioesters containing polymerized SUMO. The specificity for modification by SUMO-3 in HEK 293 cells may result from its greater cytosolic localization or preferential interaction of Cav-3 SIMs with SUMO-3 rather than SUMO-1 (45)(46)(47).
We find that PIASy preferentially enhances SUMOylation of Cav-3 at a consensus site, but mutation of this site does not abolish SUMOylation. This may result from a lack of secondary structure of the N terminus of Cav. Amino acids 79 -96 in Cav-1, which comprise the CSD (aa 55-72 in Cav-3), form an ␣-helix, whereas the remainder of the N terminus lacks secondary structure (48). Perhaps non-consensus site lysines become accessible when Lys 38 is mutated. Fernandez et al. (48) proposed that the N-terminal tail region (aa 1-54 in Cav-3) wraps back around the ␣-helical region (Fig. 8). Mutation of Cav-1 aa 66 -70 (IDFED) to alanine dramatically affects oligomerization (48); Fernandez et al. (48) speculated that this acidic patch binds basic residues in the ␣-helix, thereby causing the tail to wrap around the CSD. The corresponding region in Cav-3 (aa 39 -43) is next to the SUMOylated residue in Cav-3, Lys 38 (Fig.  8A, EDIVK(SUMO)VDFED). A helical wheel projection illustrates clustering of basic residues along one side of the ␣-helical CSD (Fig. 8B). Thus, modification by SUMO at this site could block the interaction of the N-tail region with the CSD, acting as a switch to regulate binding of proteins to the CSD or oligomerization of caveolins.
The tandem SIMs in the N terminus of Cav-3 and the ability of caveolin to oligomerize may also contribute to the SUMOylation of other lysines, albeit at a much lower level than WT or R38K (Figs. 4 and 5). Proteins can be SUMOylated on non-consensus sites that are in close proximity to a SIM (e.g. USP25 (47)). This suggests that stabilization of the binding between Ubc9 ϳ SUMO thioesters and the target by a SUMO-SIM interaction or by substrate-E3 ligase interactions may facilitate the interaction of Ubc9 with non-consensus site lysines.
PIASy and other PIAS proteins have been considered nuclear proteins (49,50), whereas Cav-3 is primarily found in plasma membrane microdomains, with some cytosolic and perinuclear localization (51). A study in PIASy ϩ/ϩ MEFs showed that although PIASy was predominantly found in the nucleus, it could also be detected in the cytoplasm (52). PIASy accumulates in the cytoplasm and co-localizes with the E3 ubiquitin ligase TRIM32 when MG-132 is used to inhibit proteasomal degradation (50), which may explain why it is normally detected at low levels in the cytoplasm. Consistent with this idea, treatment of HEK 293 cells with MG-132 for 3 h enhanced SUMOylation of Cav-3 relative to untreated cells (Fig. 4D, lane   FIGURE 6. Wild-type Cav-3-V5 stabilizes, whereas the SUMOylation-deficient K7R mutant destabilizes, FLAG-␤ 2 AR. A and B, co-expression of Cav-3-V5-K7R reduces the expression of FLAG-␤ 2 AR protein, and WT has an opposite effect when co-transfected with the SUMOylation machinery. HEK 293 cells were co-transfected with FLAG-␤ 2 AR (lanes 1-11) and either WT Cav-3-V5 (lanes 3 and 4 and lanes 8 and 9) or the K7R mutant (lanes 5 and 6 and lanes 10 and 11), both with and without the SUMOylation machinery, SUMO-3, Ubc9, and PIASy (even versus odd lanes, respectively). The proteasomal inhibitor MG-132 was preincubated with cells 3 h prior to lysis (lanes 7-11). 48 h post-transfection, lysates were immunoprecipitated (IP) with anti-FLAG antibodies and analyzed by immunoblotting with anti-FLAG-HRP, anti-ubiquitin, anti-V5-HRP, and anti-PIASy antibodies. B, HEK 293 cells were transfected identically to those in lanes 1-6 of A; however, cells were harvested 24 h, rather than 48 h, post-transfection. Lysates were immunoprecipitated with anti-FLAG antibodies and analyzed by immunoblotting with anti-FLAG-HRP and anti-V5-HRP antibodies. C, eNOS expression is not affected by co-transfection WT or Cav-3-V5. HEK 293 cells were co-transfected with eNOS (lanes 1-6) and either WT Cav-3-V5 (lanes 3 and 4) or the K7R mutant (lanes 5 and 6) with and without the SUMOylation machinery, SUMO-3, Ubc9, and PIASy. Cells were harvested 24 h post-transfection, and anti-eNOS immunoprecipitates were analyzed by immunoblotting with anti-eNOS and anti-V5-HRP antibodies.

5), presumably by increasing the cytosolic expression of PIASy.
This finding is also consistent with data obtained with HEK 293 cells in which we sought to manipulate the expression of PIASy. TGF␤ stimulates a 6 -8-fold increase in PIASy mRNA expression in Hep2B cells that peaks after 12 h (53). We incubated HEK 293 cells with TGF␤ in the presence or absence of MG-132 (treatment for 12 h) and found that cells treated with MG-132 have a 3-fold increase in PIASy protein expression and an increase in 40-kDa Cav-3-V5 (data not shown).
PIASy appears to be highly expressed in cardiac myocytes (Fig. 3E), and transfection of HEK 293 cells (which express low levels of PIASy (Fig. 3A)), with as little as 40 ng of PIASy greatly enhances SUMOylation of Cav-3 (Fig. 3B). PIASy co-immunoprecipitates with Cav-3 in transfected HEK 293 cells (Fig. 3, C and D), and immunofluorescence studies of HEK 293 cells transfected with Cav-3-V5 alone or with the SUMOylation machinery show that the affinity-tagged Cav-3 construct localizes in the cytosol and plasma membrane puncta, similar to endogenous Cav-3, consistent with the idea that PIASy co-localizes with Cav-3 in these compartments (data not shown).
PIASy interacts with the type I TGF-␤ receptor (T␤RI) in MCF7 human breast cancer cells (54), and T␤RIs localize to caveolae and interact with Cav-1 and eNOS in endothelial cells (18). Whether Cav-3 and PIASy co-localize in vivo has not been investigated. The E3 ubiquitin ligase TRIM32 can control cytosolic levels of PIASy and, akin to what has been observed for Cav-3, TRIM32 is mutated in Limb-Girdle muscular dystrophy. The latter mutation prevents TRIM32 from binding PIASy (50), FIGURE 7. The SUMOylation-deficient K7R mutant does not affect the stability of ␤ 1 AR and destabilization of ␤ 2 AR by K7R is blocked by ␤-adrenergic antagonist. A, FLAG-␤ 1 AR is not destabilized by the K7R mutant. HEK 293 cells were co-transfected with FLAG-␤ 1 AR (lanes 1-3) or FLAG-␤ 2 AR (lanes 4 -6), the SUMOylation machinery (lanes 1-6), and either WT Cav-3-V5 (lanes 2 and 5) or the K7R mutant (lanes 3 and 6). Cell lysates were immunoprecipitated (IP) with anti-FLAG antibodies and analyzed by immunoblotting with anti-FLAG-HRP and anti-V5-HRP antibodies. B-D, the destabilization of FLAG-␤ 2 AR by Cav-3-V5-K7R is blocked by (Ϫ)-propranolol and reduced serum. B, HEK 293 cells were co-transfected with FLAG-␤ 2 AR and the SUMOylation machinery (lanes 1-6) with either WT Cav-3-V5 (lanes 2 and 5) or the K7R mutant (lanes 3 and 6). Cells were grown in media plus 2 M (Ϫ)-propranolol for 48 h (lanes 1-3). Cells were also grown with reduced serum for 24 h prior to harvest (lanes 4 -6). Anti-FLAG immunoprecipitates were analyzed by immunoblotting with anti-FLAG-HRP and anti-V5-HRP antibodies. C, HEK 293 cells were transfected as in A and B and were treated or not for 48 h with 2 M (Ϫ)-propranolol. Lysates were immunoprecipitated with anti-FLAG antibodies, but proteins were not heated prior to SDS-PAGE to reduce formation of FLAG-␤ 2 AR aggregates and aid in quantification. Immunoprecipitates and cell lysates were analyzed by immunoblotting with anti-FLAG-HRP and anti-GAPDH antibodies, respectively. D, FLAG-HRP immunoblots (C) were quantified by densitometry, normalized to GAPDH, and expressed as -fold change relative to vehicle-treated controls (lane 1).
suggesting that the regulation of cytosolic SUMOylation of Cav-3 and other targets by PIASy has functional consequences in terms of the pathophysiology of muscular dystrophy.
The findings presented here are consistent with recent data that show PIASy-dependent SUMOylation of a phosphodiesterase subtype, PDE4D5, which is recruited to activated ␤ 2 ARs (24,55,56). Caveolin and this PDE subtype share a high degree of similarity in their SUMOylation motifs (supplemental Fig.  S1B). Perhaps Cav-3, ␤ 2 AR, PDE4D5, and PIASy are part of a signaling complex localized in caveolae that regulates cellular responses to agonist stimulation.
The mechanism by which SUMOylation of caveolin affects ␤ 2 AR expression levels is not clear. Cav-3 might affect ␤ 2 ARs through its ability to couple to the machinery that regulates the agonist-induced internalization and desensitization of the receptors. For example, ␤-arrestin-2 and Mdm2 are recruited to activated receptors. Cav-1 interacts with Mdm2 following H 2 O 2 treatment (57), and oxidative stress caused by H 2 O 2 can stimulate altered patterns and increases in SUMO-2/3 conjugation (58). Moreover, Mdm2 and ␤-arrestin-2 are targets of SUMOylation (37, 59 -60), and PIASy cooperates with Mdm2 to regulate SUMOylation of p53 (61). ␤-Arrestin-2 preferentially interacts with the SUMOylated form of Mdm2 (37) (p90Mdm2). A search for SIMs similar to those in Cav-1 and -3 reveals the presence of tandem SIMs in ␤-arrestin-1 and ␤-arrestin-2 (supplemental Fig. S1C, ARRB1 and ARRB2), suggest-ing that these motifs may be responsible for the preferential binding of p90Mdm2. It will thus be of interest to test if the decreased expression of ␤ 2 AR caused by the Cav-3 K7R mutant results from SUMOylation-dependent alterations of Mdm2-␤arrestin-2 interactions.
In summary, our findings are the first to show that Cavs are SUMOylated and that SUMOylation influences the expression level of a binding partner of Cav. These results thus define a previously unappreciated mechanism that contributes to the interaction of Cavs, presumably by the CSD, with such binding partners. It will be of interest to define the full range of binding partners whose interaction is influenced by SUMOylation and whether SUMOylation of Cavs is altered during physiological and pathophysiological settings.