Hsp90 Interactions and Acylation Target the G Protein Gα12 but Not Gα13 to Lipid Rafts*

The heterotrimeric G proteins, G12and G13, are closely related in their sequences, signaling partners, and cellular effects such as oncogenic transformation and cytoskeletal reorganization. Yet G12 and G13can act through different pathways, bind different proteins, and show opposing actions on some effectors. We investigated the compartmentalization of G12 and G13 at the membrane because other G proteins reside in lipid rafts, membrane microdomains enriched in cholesterol and sphingolipids. Lipid rafts were isolated after cold, nonionic detergent extraction of cells and gradient centrifugation. Gα12 was in the lipid raft fractions, whereas Gα13 was not associated with lipid rafts. Mutation of Cys-11 on Gα12, which prevents its palmitoylation, partially shifted Gα12 from the lipid rafts. Geldanamycin treatment, which specifically inhibits Hsp90, caused a partial loss of wild-type Gα12 and a complete loss of the Cys-11 mutant from the lipid rafts and the appearance of a higher molecular weight form of Gα12 in the soluble fractions. These results indicate that acylation and Hsp90 interactions localized Gα12 to lipid rafts. Hsp90 may act as both a scaffold and chaperone to maintain a functional Gα12 only in discrete membrane domains and thereby explain some of the nonoverlapping functions of G12 and G13 and control of these potent cell regulators.

A recent report shows that G␣ 12 interacts with Hsp90 (heat shock protein of 90 kDa), and this interaction is required for G␣ 12 signaling (18). The Hsp90 interaction was specific for G␣ 12 because G␣ 13 did not bind or show functional interactions with Hsp90. Hsp90 is an abundant and ubiquitously expressed molecular chaperone with the unique property of binding and maintaining the activity of numerous signal transduction proteins (19,20) including steroid hormone receptors, signaling kinases, nitric-oxide synthase, and thrombin receptors (21). Recently Hsp90 was discovered to serve as a scaffold to promote the interactions of the Akt kinase and its substrate, endothelial nitric-oxide synthase, through their binding to two of the multiple protein binding sites on Hsp90 (22).
Compartmentalization of eukaryotic cell membranes into dynamic microdomains, referred to as lipid rafts, has been detected in numerous studies using different methodologies (23)(24)(25). These microdomains are enriched in cholesterol and sphingolipids and can be isolated by their resistance to cold, nonionic detergent extraction. Proteins modified with multiple acyl chains or glycosylphosphatidylinositol groups are targeted to lipid rafts (23,26). The G␣ subunits of the G s , G i , and G q classes undergo palmitoylation and/or myristoylation on their amino-terminal ends (27) and localize to lipid rafts (28 -31). G␣ 12 and G␣ 13 also undergo palmitoylation, the reversible addition of palmitate to cysteine residues through a thioester bond (32)(33)(34)(35). A single cysteine residue (Cys-11) for G␣ 12 (32,33) and two or three cysteine residues (Cys-14, Cys-18, and Cys-37) for G␣ 13 are critical for this posttranslational modification (34,35).
We investigated the lipid raft localization of G␣ 12 and G␣ 13 to better understand the different signaling functions of these proteins. We found a distinct segregation of G␣ 12 and G␣ 13 at the membrane. G␣ 13 was not in lipid rafts, whereas G␣ 12 was in lipid rafts with both acylation and Hsp90 interactions critical for this targeting.

EXPERIMENTAL PROCEDURES
Materials-The QE antibody specific for G␣ 12 , HD antibody specific for G␣ 13 , AS antibody specific for G␣ i , and SW antibody specific for G␤ were prepared by immunizing rabbits with the carboxyl-terminal peptide sequence of the respective subunits and affinity-purified. Antibodies against the following proteins were purchased: G␣ 12 and G␣ 13 against their amino-terminal sequences and Fyn (Santa Cruz Biotechnology), caveolin and Hsp90 (BD PharMingen), and Na ϩ /K ϩ -ATPase (Biomol Research Laboratories). OptiPrep was purchased from Invitrogen. Geldanamycin was obtained from Tocris Cookson Inc. MG132, a proteasome inhibitor, was from Calbiochem. Methyl-␤-cyclodextrin, hydrocortisone, and epidermal growth factor were from Sigma.
Cell Culture and Treatment with Methyl-␤-cyclodextrin and Geldanamycin-Simian kidney (COS-7) and human embryonic kidney (HEK293) cells were maintained in Dulbecco's modified Eagles medium * 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.
with 10% fetal bovine serum. Human dermal microvascular endothelial cells that were transformed with SV-40 (HMEC1) (Biological Products Branch, Centers for Disease Control and Prevention, Atlanta, GA) were maintained in MCBD 131 medium supplemented with 5% fetal bovine serum, hydrocortisone (1 g/ml), and epidermal growth factor (0.01 mg/ml). NIH 3T3 mouse fibroblasts that were stably transfected with wild-type and mutant plasmids of G␣ 12 were prepared and maintained as described previously (32). In geldanamycin experiments, cells were incubated in serum-free medium for 1 h and preincubated with 10 M MG132 for 10 min, and then 1 g/ml geldanamycin was added and incubated for 1 h. In cholesterol depletion experiments, COS cells were preincubated with serum-free medium for 2 h and then treated with 20 mM methyl-␤-cyclodextrin in serum-free medium at 37°C for 30 min.
Preparation of Detergent-resistant Membranes (DRMs)-DRMs were prepared as described previously with some modifications (36). Cells were washed twice and harvested in ice-cold phosphate-buffered saline (pH 7.4). Cell pellets were obtained by centrifugation at 1,000 ϫ g for 10 min at 4°C. The cell pellet was suspended in phosphate-buffered saline, and the protein concentration was determined (Bio-Rad). An aliquot of the cell suspension corresponding to 1 mg of protein was centrifuged at 5,000 ϫ g for 5 min at 4°C. The cell pellet was resuspended in 0.5 ml of TNET buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA, 0.5% (v/v) Triton X-100) containing protease inhibitors, incubated on ice for 20 min, and then homogenized with 10 strokes in a Dounce homogenizer. The cell lysate was adjusted to 35% OptiPrep and overlaid with 3.0 ml of 30% OptiPrep in TNET and then with 200 l of 5% OptiPrep in TNET in Beckman SW55 tubes. Samples were centrifuged for 4 h at 170,000 ϫ g at 4°C and fractionated from the top (0.52 ml each for a total of nine fractions). The pellet was suspended in 0.52 ml of TNET buffer. SDS sample buffer containing 2-mercaptoethanol was added to aliquots of each fraction of the gradient.
Subcellular Fractionation and Triton X-100 Solubilization Experiments-NIH 3T3 cells expressing the wild-type G␣ 12 or COS cells were homogenized in TNE buffer (TNET without Triton X-100) containing protease inhibitors by passing through a 25-gauge needle 12 times. The particulate and soluble fractions were separated by centrifugation at 125,000 ϫ g for 1 h. Aliquots of soluble and particulate fractions were treated with 200 M MG132 for 10 min and then treated with 20 g/ml of geldanamycin for 1 h. Triton X-100 was added to give a final concentration of 0.5% and incubated at 4°C for 20 min, then diluted 3-fold in TNE buffer, and further incubated for 4 h. Detergent-soluble and -insoluble fractions were separated by centrifugation at 15,000 ϫ g and analyzed by immunoblotting.
Immunoblotting and Miscellaneous Procedures-Equal volumes of gradient fractions were loaded on 12% polyacrylamide gels (Novex precast gels, Invitrogen) and subjected to SDS-PAGE and immunoblotting using Enhanced Chemiluminescence (Amersham Biosciences) for detection of the antibodies per the manufacturer's directions. Densitometry was performed using a UMAX scanner (model UTA-II) and Scion Image software.

RESULTS AND DISCUSSION
DRM Partitioning of G␣ 12 but Not G␣ 13 -To determine whether G␣ 12 and G␣ 13 reside in lipid rafts, we prepared DRMs by treatment of COS cells with 0.5% Triton X-100 and centrifugation on an OptiPrep gradient. Caveolin, Fyn, and G␣ i were found in fractions 1 and 2 consistent with their known partitioning to DRMs and lipid rafts (Fig. 1A) (25). Na ϩ /K ϩ -ATPase, an integral plasma membrane protein not found in lipid rafts (36), was in the higher density fractions 6 -8. Endogenous G␣ 12 was found in the DRM fraction 1; G␣ 13 was in the higher density fractions 7-9. We found the same pattern of partitioning after detergent solubilization and gradient centrifugation for the endogenous G␣ 12 and G␣ 13 proteins in two other cell lines, HMEC (Fig. 1B) and HEK293 (data not shown). In these cells, G␣ 12 and the lipid raft markers were in the DRM fractions, and G␣ 13 and Na ϩ /K ϩ -ATPase were in the higher density fractions. The partitioning of G␣ 12 and the lipid raft-associated proteins to the DRM fraction was abolished and decreased, respectively, by treatment with 20 mM methyl-␤-cyclodextrin, which depletes cellular cholesterol and disrupts lipid rafts (Fig. 1C). These results show that G␣ 12 segregates with proteins that prefer lipid rafts and G␣ 13 segregates with a plasma membrane protein that avoids lipid rafts.
Acylation and the DRM Localization of G␣ 12 but Not G␣ 13 -We investigated the role of the cysteine site of palmitoylation for targeting G␣ 12 to lipid rafts. Mutation of Cys-11 on G␣ 12 prevents [ 3 H]palmitate incorporation, but the protein is still found in the particulate fraction (32). Both the wild-type G␣ 12 and a GTPase-deficient mutant of G␣ 12 (Q229L) were in the DRM fractions in transfected NIH 3T3 cells ( Fig. 2A). Most of the nonpalmitoylated C11S mutant of G␣ 12 was shifted to the higher density fractions ( Fig. 2A). This result is consistent with other studies showing that acylation can target proteins to lipid rafts (23,26). Further mutation of the amino terminus of G␣ 12 at Ser-2 and Arg-6 changes the acylation to myristoylation (32) and restored a small amount of the DRM localization of G␣ 12 ( Fig. 2A). Therefore, the less hydrophobic myristate on the amino terminus could partially substitute for palmitate in targeting G␣ 12 to the DRM fractions.
Given the importance of the cysteine site of acylation for the DRM partitioning of G␣ 12 and the lipid raft localization of other biacylated G␣ subunits including G␣ i in this study (29 -31), we were surprised that G␣ 12 with only one putative acylation site (32,33) was in the DRM fraction and that G␣ 13 with two or three putative acylation sites (34,35) was in the higher density fractions. Not all palmitoylated proteins reside in lipid rafts (37), and the stoichiometry of palmitoylation on G␣ 13 is not known, but our present result suggests that G␣ 13 evades lipid rafts. Clearly further studies are required to understand the interaction that may prevent G␣ 13 targeting to lipid rafts.
Hsp90 Interactions-Hsp90 binds to G␣ 12 but not G␣ 13 (18), and Hsp90 can be part of a complex with caveolin (38). Therefore, we investigated whether Hsp90 may be involved in the targeting of G␣ 12 , but not G 13 , to the DRM fractions. To probe the interaction between G␣ 12 and Hsp90, we used geldanamycin, a fungal benzoquinone ansamycin, that specifically and tightly binds to the ATP binding site in Hsp90 and inhibits the FIG. 1. DRM localization and methyl-␤-cyclodextrin treatment. COS-7 (A) and HMEC (B) cells were treated with 0.5% Triton X-100 on ice, homogenized, and subjected to floatation on an OptiPrep gradient and fractionated as described under "Experimental Procedures." Equal volumes from each fraction were subjected to SDS-PAGE followed by immunoblotting with the carboxyl-terminal antibody to G␣ 12 or the amino-terminal antibody to G␣ 13 . Immunoblotting was also carried out with antibodies to the raft marker proteins, caveolin, G␣ i , and Fyn, and a non-raft marker protein, Na ϩ /K ϩ -ATPase. C, COS cells were incubated with 20 mM methyl-␤-cyclodextrin at 37°C for 30 min and subjected to Triton X-100 solubilization, OptiPrep density gradient centrifugation, and immunoblotting as described above.
ability of Hsp90 to form complexes with its substrate proteins (20). We included the proteasome inhibitor MG132 (39) because geldanamycin can increase protein degradation through the ubiquitin/proteosomal pathway (19,40). Treatment with MG132 alone did not change the DRM localization of G␣ 12 (data not shown). Geldanamycin and MG132 treatment of COS cells displaced G␣ 12 from the DRM fractions and led to the appearance of a 220-kDa band in fractions 6 -9 (Fig. 3A). Densitometric analysis showed a loss of G␣ 12 from the DRM fraction of about 60 Ϯ 10% (mean Ϯ S.E. for three experiments) for cells treated with geldanamycin plus MG132 compared with cells treated with vehicle alone. The lipid raft localization of caveolin, G␣ i , G␤, and Fyn was not changed by geldanamycin treatment (Fig. 3B). Likewise, geldanamycin treatment did not alter the localization of G␣ 13 (Fig. 3B) and Na ϩ /K ϩ -ATPase (data not shown) to the higher density fractions. Only a trivial amount of Hsp90 was found in the DRM fraction, consistent with a previous report (41) and with the cytosolic location of this abundant protein (Fig. 3B) (19,20). Hsp90 was not found in the DRMs after geldanamycin treatment. Treatment with geldanamycin alone or with MG132 did not change the total amount of G␣ 12 (data not shown) in agreement with a previous report (18). These results indicate that the loss of G␣ 12 from the DRM fractions after geldanamycin treatment was not due to a general disruption of lipid rafts or degradation of G␣ 12 but rather suggest it was due to a loss of Hsp90 binding.
We also tested the effect of geldanamycin treatment on the DRM localization of the nonpalmitoylated C11S mutant of G␣ 12 to determine the relationship between acylation and Hsp90 binding in the lipid raft localization of G␣ 12 . Geldanamycin treatment of the transfected NIH 3T3 cells led to a further loss of the nonpalmitoylated C11S mutant from the DRM fraction so that the mutant G␣ 12 was only in the higher density fractions (Fig.  4A). An increase in the intensity of the 220-kDa band was also seen after geldanamycin treatment in these cells. This result indicates that acylation and Hsp90 binding act independently to promote the DRM localization of G␣ 12 .
We investigated whether Hsp90 interactions maintain G␣ 12 at lipid rafts. The crude membrane fraction of NIH 3T3 cells stably transfected with wild-type G␣ 12 was treated with geldanamycin and MG132 followed by solubilization with Triton X-100 under conditions similar to DRM preparation.

FIG. 4. Geldanamycin treatment of intact NIH 3T3 cells and its subcellular fractions.
A, NIH 3T3 cells expressing C11S mutant were preincubated with 10 M MG132 for 10 min and then treated with 1 g/ml geldanamycin for 1 h. Cells underwent cold detergent extraction and were subjected to OptiPrep gradient centrifugation as in Fig. 1. A mixture of carboxyl-and amino-terminal antibodies was used to detect G␣ 12 . B, NIH 3T3 cells expressing wild-type G␣12 were homogenized and separated into cytosol and crude membrane fractions by centrifugation. Aliquots of these fractions were treated with 200 M MG132 for 10 min and then treated with 20 g/ml geldanamycin (GA) for 1 h. Triton X-100 at a final concentration of 0.5% was added and incubated at 4°C for 20 min, then diluted 3-fold with TNE buffer, and further incubated for 4 h. Triton X-100-soluble (S) and -insoluble (P) fractions were separated by centrifugation and analyzed by immunoblotting with a mixture of carboxyl-and amino-terminal antibodies to G␣ 12 .

FIG. 2. Lipid raft localization of wild-type and mutant G␣ 12 in stably transfected NIH 3T3 cells.
A, NIH 3T3 cells expressing wildtype G␣ 12 and its mutants were homogenized in cold 0.5% Triton X-100 and separated into gradient fractions as described in the Fig. 1 legend. The antibody to the carboxyl terminus of G␣ 12 was used for all the G␣ 12 -expressing cells except the C11S mutant for which the aminoterminal antibody was used. B, NIH 3T3 cells transfected with wildtype G␣ 12 were prepared as described above. Immunoblotting was performed with antibodies to G␣ 13 (amino-terminal antibody), caveolin, G␣ i , and Na ϩ /K ϩ -ATPase. C, NIH 3T3 cells stably transfected with vector alone (V) or containing the cDNA of the wild-type (WT) or mutant G␣ 12 proteins (C11S or S2G,C11S) were homogenized, and 40 g of protein from the total cell lysate were separated by SDS-PAGE and analyzed by immunoblotting with either carboxyl-(C-term Ab) or amino-terminal (N-term Ab) antibodies to G␣ 12 .

FIG. 3. Geldanamycin treatment of COS cells.
A, COS cells were preincubated with 10 M MG132 for 10 min, and then treated with 1 g/ml geldanamycin for 1 h. Cells underwent cold detergent extraction and OptiPrep gradient centrifugation as described in the Fig. 1 legend. Fractions were analyzed by SDS-PAGE and immunoblotting with a mixture of carboxyl-and amino-terminal antibodies against G␣ 12 . B, immunoblotting was also carried out with antibodies to caveolin, G␣ i , G␤, Fyn, G␣ 13 (amino-terminal antibody), and Hsp90.
Geldanamycin treatment increased the detergent solubility of G␣ 12 (Fig. 4B, membrane) suggesting that Hsp90 interactions maintain G␣ 12 in the lipid rafts in addition to any possible role in directing G␣ 12 to these membrane microdomains.
To our knowledge, this is the first report of a change in lipid raft localization after disruption of Hsp90 binding. Hsp90 is an abundant cytosolic protein that is not directly targeted to lipid rafts so its role in the targeting of G␣ 12 would be either 1) maintenance of G␣ 12 in a conformation that permits direct interaction of G␣ 12 with lipids and proteins in lipid rafts or 2) formation of a complex with proteins that have a high affinity for lipid rafts. In regard to the latter, Hsp90 has multiple protein binding sites (22) and readily forms complexes with other proteins including proteins that directly associate with lipid rafts (38) or translocate to lipid rafts (41).
The Higher Molecular Weight Form of G␣ 12 -The presence of a 220-kDa band detected with an antibody to G␣ 12 in the high density fractions after geldanamycin and MG132 treatment is a novel finding. Higher molecular weight bands of G␣ 13 were not detected after geldanamycin treatment with antibodies to the amino and carboxyl terminus of G␣ 13 (data not shown). The 220-kDa band was not due to immunoreactivity of another protein that is expressed after geldanamycin treatment because we saw an increase in the 220-kDa band after geldanamycin treatment of the cytosolic fractions of NIH 3T3 cells transfected with the wild-type G␣ 12 (Fig. 4B, cytosol) or COS cells (data not shown). The relative amounts of the 43-and 220-kDa bands could not be determined because the antibody to the amino terminus poorly detects the 43-kDa form probably because palmitoylation obscured the epitope (Fig. 2B), and the antibody to the carboxyl terminus did not detect the 220-kDa form (data not shown).
The correlation of the loss of G␣ 12 immunodetection at 43 kDa in the DRMs (Figs. 3A and 4A) with an increase at 220 kDa suggests that this band is likely to contain the G␣ 12 protein, possibly as part of another protein or an SDS-resistant aggregate or oligomer. Hsp90 as a molecular chaperone maintains the proper folding of signal transduction proteins, which have an inherent instability because they undergo conformational changes as part of relaying signals (19). The presence of the 220-kDa band after geldanamycin treatment suggests that Hsp90 may be preventing aggregation of G␣ 12 .
Functional Implications-Signal transduction requires not only affinity between molecules but also proximity. Some of the defects seen in G 12 signaling after mutation of Cys-11, which prevents palmitoylation (32,33), or geldanamycin treatment, which blocks Hsp90 interactions (18,21), may be due to mistargeting of G␣ 12 away from its signaling partners. Concentrating G 12 in lipid rafts may significantly improve the kinetics of its signaling.
Heterogeneity and compartmentalization within cellular membranes can also keep signaling partners apart to prevent interactions and increase the specificity of signaling. The differential membrane targeting of G␣ 12 and G␣ 13 may be responsible for increased specificity of receptor coupling seen for G␣ 12 and G␣ 13 in intact cells under physiologic conditions compared with results using cell membranes (11). Specific control of the membrane location of G 12 and G 13 is especially important because they are potent inducers of cell growth, differentiation, shape change, and apoptosis. G 12 interactions with Hsp90 may explain the tight regulation of G 12 signaling and the nonoverlapping functions of G 12 and G 13 because Hsp90 is both a scaffold and chaperone. If G␣ 12 remains bound directly or indirectly to Hsp90, it maintains a functional conformation to interact with its signaling partners. If it strays from Hsp90 and its membrane microdomain, G␣ 12 aggregates and loses activity. G␣ 12 would then only work when it is in the correct location and thus effectively prevent random collisions with signaling partners during diffusion in the membrane. The inability of G␣ 12 to compensate for the loss of G␣ 13 in the G␣ 13 -deficient mice (16) may be the result of its poor mobility. Besides Hsp90, G␣ 12 has also been found to interact with two other scaffold proteins in specialized cells (6,13). G 12 and G 13 are like fraternal twins, similar but different. Our finding that they play their roles at different sites on the membrane gives a framework for understanding their regulation and the unfolding story of Hsp90 and G 12 signaling. More studies are needed to discover the ramifications of their interactions and membrane targeting for G protein signaling.