Gβγ Isoforms Selectively Rescue Plasma Membrane Localization and Palmitoylation of Mutant Gαs and Gαq *

Mutation of Gαq or Gαs N-terminal contact sites for Gβγ resulted in α subunits that failed to localize at the plasma membrane or undergo palmitoylation when expressed in HEK293 cells. We now show that overexpression of specific βγ subunits can recover plasma membrane localization and palmitoylation of the βγ-binding-deficient mutants of αs or αq. Thus, the βγ-binding-defective α is completely dependent on co-expression of exogenous βγ for proper membrane localization. In this report, we examined the ability of β1–5 in combination with γ2 or γ3 to promote proper localization and palmitoylation of mutant αs or αq. Immunofluorescence localization, cellular fractionation, and palmitate labeling revealed distinct subtype-specific differences in βγ interactions with α subunits. These studies demonstrate that 1) α and βγ reciprocally promote the plasma membrane targeting of the other subunit; 2) β5, when co-expressed with γ2 or γ3, fails to localize to the plasma membrane or promote plasma membrane localization of mutant αs or αq; 3) β3 is deficient in promoting plasma membrane localization of mutant αsand αq, whereas β4 is deficient in promoting plasma membrane localization of mutant αq; 4) both palmitoylation and interactions with βγ are required for plasma membrane localization of α.

G proteins 1 are found on the cytoplasmic face of the plasma membrane (PM) where they transduce signals from heptahelical receptors to effector proteins (1,2). Activation of the receptor via agonist binding induces a conformational change in the G protein ␣␤␥ heterotrimer, which triggers its dissociation from the receptor in the form of a GTP-bound ␣ subunit and a ␤␥ complex. Each subunit is then capable of regulating the function of various effector proteins. Because proper membrane localization is a prerequisite for the correct functioning of this system, numerous studies have examined the requirements for targeting of the G protein ␣ subunits to the plasma membrane. The two major requirements seem to be covalent lipid modifications (3,4) and binding to the ␤␥ complex (5,6).
Two types of lipid modification of G protein ␣ subunits have been described, both of which occur at the extreme N terminus of the protein. These subunits are either palmitoylated, in the case of ␣ s , ␣ q , ␣ 12 , and ␣ 13 , or myristoylated and palmitoylated in the case of ␣ i , ␣ o , and ␣ z . Myristate, a 14-carbon fatty acid, is attached co-translationally and irreversibly, whereas palmitate, a 16-carbon fatty acid, is attached post-translationally and reversibly. These lipid modifications help anchor the G protein ␣ subunit to the PM, but in the case of ␣ s and ␣ q , and presumably ␣ 12 and ␣ 13 as well, palmitoylation and proper localization of the ␣ subunit at the PM requires stable binding between the ␣ and ␤␥ subunits (6). The ␤␥ complex is anchored at the PM with the help of its own lipid modification, the post-translational attachment of either a farnesyl or geranylgeranyl moiety to a cysteine at a C-terminal CAAX box on the ␥ subunit.
Although formation of the ␣␤␥ heterotrimer is clearly essential for cell signaling, the selectivity of interactions among different subunit subtypes has not been well defined. Given that there are at least 23 different ␣ subunits (including splice variants), 5 ␤ subunits and 12 ␥ subunits, this allows for more than 1300 different heterotrimers. It has been previously documented that some G protein subunits are expressed in subtype-specific patterns in tissues and cells (7)(8)(9)(10)(11). This obviously determines the limits of subtype-specific heterotrimer formation in a particular cell but it does not address what ␤␥ dimers and ␣␤␥ heterotrimers are actually formed. However, a number of reports have described subtype-specificity in ␤␥ dimer and ␣␤␥ heterotrimer formation and function (12). Purification and identification of ␤␥ dimers from tissues (13)(14)(15) and in vitro binding data (16,17) have provided evidence for the existence of specific ␤ and ␥ combinations. A variety of approaches, including overexpression of subunits, reconstitution studies, antisense, and ribozyme studies, have revealed subtype-specific differences in ␣␤␥ heterotrimer coupling to receptors (18 -23) and subtype-specific differences in ␤␥ heterodimer regulation of effectors (24 -29). Yet, little information is available on differences in the abilities of specific ␣ subtypes to interact with specific ␤␥ dimers. It is not known if observed combinations are indicative of restricted functional associations between the ␣ and ␤␥ subunits, or whether all combinations are possible and selectivity is derived solely through variations in G protein expression and the kinetics of G protein coupling to receptors or effectors.
To begin to address questions of selectivity in ␣␤␥ heterotrimer formation and the role of such selectivity in proper subcel-lular localization, we have expressed different combinations of ␤ and ␥ subunits along with wild-type or ␤␥-binding-deficient mutants of ␣ s or ␣ q . We recently demonstrated that mutation of amino acids, which are predicted to contact ␤␥, in the Nterminal region of ␣ s or ␣ q resulted in ␣ subunits that failed to localize at the plasma membrane or undergo palmitoylation when transiently expressed in HEK293 cells (6). Thus, it appeared that endogenous ␤␥ was unable to effectively interact with the ␤␥-binding mutants of ␣ s or ␣ q . However, the mutants still contain a C-terminal region, which is known to contact ␤␥, and although this region seems to be insufficient for endogenous levels of ␤␥ to functionally interact with the mutant ␣ subunits, it may be possible to recover a functional interaction by overexpression of the ␤␥ complexes. In this report, we test the hypothesis that overexpressed ␤␥ can overcome the defects observed with the ␤␥-binding-defective mutants of ␣ s and ␣ q .
Furthermore, expression of unique ␤␥ combinations would allow us to test different ␤␥ complexes for their ability to recover the localization of the ␤␥-binding-defective mutants of ␣ s and ␣ q . Thus, rescue of PM localization could provide a model system for assaying subtype-specific differences in ␤␥ interactions with ␣ subunits that may be masked by the high levels of overexpressed protein displaying normal binding affinities. Herein, we present the first test of this system by examining the ability of ␤ 1-5 in combination with ␥ 2 or ␥ 3 to promote plasma membrane targeting of the ␤␥-binding mutants of ␣ s or ␣ q . Importantly, this analysis revealed distinct subtype-specific differences in ␤␥ interactions with ␣ subunits.
We also tested the importance of ␣ subunit palmitoylation in this system. We demonstrate that overexpression of ␤␥ recovers palmitoylation of ␤␥-binding mutants of ␣ s and ␣ q . Significantly, overexpression of ␤␥ with cysteine to serine mutants of ␣ s and ␣ q , which are incapable of being palmitoylated, fails to recover membrane localization of the ␣ subunits.

EXPERIMENTAL PROCEDURES
Materials-HEK293 cells were obtained from the American Type Culture Collection (CRL-1573). [9,10-3 H]Palmitic acid was from PerkinElmer Life Sciences. 12CA5 mouse monoclonal antibody was a gift from Henry Bourne. c-Myc mouse monoclonal antibody was a gift from Matthew Hart. All other primary antibodies were from Santa Cruz Biotechnology. Tissue culture reagents were from Life Technologies, Inc. Other reagents were from Fisher Scientific and Sigma Chemical Co.
Cell Culture and Transient Transfection-HEK293 cells were maintained in culture as previously described (6). Transfections were carried out in either 6-well cell culture plates with 1 g of total plasmid DNA or 6-cm plates with 3 g of total plasmid DNA using FuGENE 6 Transfection Reagent (Roche Molecular Biochemicals) according to the manufacturer's protocol. Cells were transfected overnight, transferred to new plates the next day, and grown for 24 -30 h prior to subsequent manipulations.
Immunofluorescence Localization-Immunofluorescence localization was performed as previously described (6). Briefly, HEK293 cells were transfected overnight in 6-well plates with the indicated amounts of each expression plasmid. The cells were replated on glass coverslips and grown for 24 to 48 h before fixing with 3.7% formaldehyde in phosphate-buffered saline for 15 min. Cells were permeabilized and then incubated with the indicated primary antibodies for 1 h at the following dilutions: 12CA5 monoclonal (3 g/ml), c-Myc monoclonal (1 g/ml), HA-probe (Y-11) rabbit polyclonal (2 g/ml), c-Myc (A-14) rabbit polyclonal (2 g/ml), G ␥2 (A-16) rabbit polyclonal (2 g/ml) and G ␥3 (K-20) rabbit polyclonal (2 g/ml). The cells were washed and incubated in a 1:100 dilution of the indicated secondary antibody: Alexa Fluor 488 goat anti-rabbit, Alexa Fluor 488 goat anti-mouse, Alexa Fluor 594 goat anti-rabbit, and Alexa Fluor 594 goat anti-mouse (Molecular Probes, Eugene, OR), for 30 min. The coverslips were washed and mounted on glass slides with Prolong Antifade reagent (Molecular Probes, Eugene, OR) and microscopy was performed with an Olympus BX60 microscope equipped with a Sony DKC-5000 digital camera. A minimum of 100 cells was examined for each transfection, and in most cases this number was at least 500. Many transfections were repeated once, and two or three representative pictures were taken of cells displaying a typical expression pattern for each transfection. Only cells displaying low to intermediate levels of expression were utilized. Similar or identical f-stops were used for imaging the fluorescence from the ␣ or ␤ subunits in each experiment to verify that imaged cells were expressing approximately equal amounts of each subunit. Images were processed with Adobe Photoshop.
Subcellular Fractionation-Soluble and particulate fractions were isolated as previously described (6). Briefly, HEK293 cells were transfected with the indicated amounts of each expression plasmid in 6-cm plates overnight, transferred to 10-cm plates, and grown for 48 h. The cells were washed with phosphate-buffered saline and lysed in 0.5 ml of lysis buffer by passage through a syringe. Nuclei and intact cells were removed by centrifugation, and the supernatant was centrifuged at 150,000 ϫ g to obtain the soluble and particulate fractions. Fractions were resolved by 12% SDS-PAGE, transferred to PVDF-Plus (Micron Separations Inc.) and probed with the 12CA5 monoclonal antibody. The bands were visualized by chemiluminescence and quantitated using a Kodak DC40 imaging system.
Palmitoylation and Immunoprecipitation-Radiolabeling and immunoprecipitation of ␣ s and ␣ q were performed as previously described (6). HEK293 cells were transfected with the indicated amounts of expression plasmids, grown for 48 h, and incubated with 1 ml of Dulbecco's modified Eagle's medium containing 10% dialyzed fetal bovine serum, 5 mM sodium pyruvate, and [9,10-3 H]palmitic acid (1 mCi/ml for ␣ s or 0.5 mCi/ml for ␣ q ) for 2 h in a 6-cm dish. Cells were washed once with phosphate-buffered saline and lysed in 1 ml of radioimmune precipitation buffer. Cell extracts were tumbled for 1 h at 4°C, and nuclei and insoluble material were removed by microcentrifugation at 16,000 ϫ g for 3 min. Samples containing HA-tagged ␣ q or its mutants were adjusted to 0.1% SDS. 1 g of 12CA5 antibody was added, and the samples were tumbled for 2 h at 4°C. Next, 20 l of Protein A/G PLUS-agarose (Santa Cruz Biotechnology, Santa Cruz, CA) was added, and the sample was tumbled overnight at 4°C. The sample was centrifuged for 30 s at 200 ϫ g to pellet the beads. The supernatant was discarded, and the beads were washed three times with 1 ml of radioimmune precipitation buffer. SDS-PAGE sample buffer containing 10 mM dithiothreitol was added to the washed beads, and the samples were heated at 65°C for 1 min. An aliquot was analyzed by 10% SDS-PAGE. Gels were incubated for 20 min in an aqueous solution of 50% methanol/10% acetic acid, followed by 10% methanol/10% acetic acid for 20 min, and, finally, Amplify (Amersham Pharmacia Biotech) for 20 min. Gels were dried and subjected to fluorography at Ϫ80°C using Hyperfilm MP (Amersham Pharmacia Biotech).

RESULTS
Interdependence of ␣, ␤, and ␥ Subunits for Plasma Membrane Localization-To examine the intracellular targeting and localization of the G protein ␣ and ␤␥ subunits, these subunits were transiently expressed in HEK293 cells independently or with complementary subunits. As previously shown by us, ␣ s expressed in the presence of only endogenous ␤␥ is found predominately at the plasma membrane but some remains in the cytoplasm (6). Because ␤␥ binding was determined to be essential for membrane localization of ␣ s , it is reasonable to expect that overexpression of ␤␥ complexes with the ␣ s subunit would be able to target more ␣ s to the PM than endogenous ␤␥ alone. Co-expression of ␤ 1 ␥ 2 with ␣ s in HEK293 cells resulted in prominent PM localization of ␣ s (Fig. 1A1). Comparison of these results with those of ␣ s when expressed alone showed more pronounced PM localization in a larger number of cells. The ␤ 2-4 ␥ 2 subunits were also effective at promoting more pronounced PM localization of ␣ s (Table I). Co-expression with ␤ 5 ␥ 2 , a ␤␥ complex that does not interact with ␣ s (31), resulted in a mixed PM/cytoplasmic localization of ␣ s that was similar to ␣ s expressed alone (Fig. 1B1). The presence of ␥ 3 rather than ␥ 2 in the ␤␥ complex did not change the effectiveness of PM targeting for either ␤ 1 or ␤ 5 (Fig. 1, C1 and D1). Examination of the localization of the ␤␥ complex in each case showed that it co-localized with the ␣ subunit when the ␣ subunit was exclusively at the PM (Fig. 1, A2 and C2) but failed to do so otherwise (Fig. 1, B2 and D2). In contrast to ␣ s , ␣ q displayed strong PM localization when expressed alone (6). As expected, ␣ q retained a predominant PM localization when co-expressed with each of the ␤␥ complexes, and the localization of the ␤␥ complexes did not vary from the results with ␣ s (Table I).
We have demonstrated that the ␤␥ complex is critical for PM localization of ␣ s and ␣ q , but it is not clear how important the ␣ subunit is for ␤␥ complex localization. To help determine this, myc-tagged ␤ subunits were expressed alone, co-expressed with ␥ subunits, or co-expressed with both ␣ and ␥, and visualized by immunofluorescence. The ␤ 1 subunit displayed some plasma membrane staining but seemed to be most prominent in the cytoplasm or intracellular membranes (Fig. 2). Similar results were obtained with the ␤ 4 subunit (data not shown). ␤ 2 , ␤ 3 , and ␤ 5 displayed only intracellular staining, although ␤ 2 and ␤ 3 displayed some exclusion from the nucleus and ␤ 5 did not (data not shown). Co-expression with ␥ 2 subunits resulted in little to no increase in PM localization for all ␤ subunits despite the fact that the ␥ 2 subunit should be prenylated and able to anchor to the PM ( Fig. 2 and data not shown). Upon co-expression with ␣ s and ␥ 2 , essentially all of the ␤ 1 subunits were targeted to the PM where they co-localized with ␣ s (Figs. 1A2 and 2), as were ␤ 2-4 ( Table I). Overlaying of the two pictures ( Fig. 1, A1 and A2) showed conclusively that ␤ 1 co-localizes with ␣ s at the plasma membrane, as demonstrated by the bright yellow plasma membrane coloring (Fig. 1A3). Similar to ␣ s , expression of ␣ q induced increased PM localization of ␤ 1-4 ␥ 2 (Table I).
Although co-expression of the ␥ subunit with the ␤ subunit did not seem to affect the localization of the ␤ subunit, ␤ subunit expression did affect the expression and localization of ␥ in our system. Because the ␥ 2 antibody did not work reliably in our hands, we used the ␥ 3 subunit for these experiments. Previous localization of endogenous ␥ 3 in cardiac fibroblasts showed diffuse staining throughout the cell (32). Transient transfection of the ␥ 3 subunit alone resulted in almost no detectable expression. The few cells that did express the protein displayed a small level of intracellular expression only (Fig. 2). Co-expression with the ␤ 1 subunit resulted in expression in more cells with protein present both intracellularly and at the PM (Fig. 2). However, it was not until the ␣ s subunit was also transfected that predominately PM localization was seen for the ␥ 3 subunit and there was expression in a large number of cells (Fig. 2). Thus, it appears that expression and localization of each subunit of the heterotrimeric G protein is dependent in some way on the others and no single subunit determines the expression and localization of the others.
Unlike all the other ␤ subunits, the ␤ 5 subunit never displayed any PM localization. When ␤ 5 was expressed alone or with the ␥ 3 subunit, its expression was evenly distributed throughout the cell (Fig. 1D2). However, when it was co-expressed with the ␥ 2 subunit, ␤ 5 expression was restricted to perinuclear and cytoplasmic membranes. Additional expression of ␣ s (Fig. 1B2) or ␣ q (Table I) did not change the localization of ␤ 5 . This unique localization of the ␤ 5 subunit was found to be caused by co-expression of the ␥ 2 subunit, which, when expressed alone, shows this same localization in perinuclear and cytoplasmic membranes (Ref. 33 and data not shown). These results suggest that ␤ 5 is able to co-localize and associate with ␥ 2 but not ␥ 3 . This is supported by studies that have FIG. 1. Immunofluorescence microscopy displaying the localization of G protein ␣ and ␤ subunits overexpressed in HEK293 cells. HEK293 cells were transiently transfected with HA-tagged ␣ s and myc-tagged ␤ 1 ␥ 2 (A), ␤ 5 ␥ 2 (B), ␤ 1 ␥ 3 (C), or ␤ 5 ␥ 3 (D) and fixed as detailed under "Experimental Procedures." Visualization of the ␣ and ␤ subunits was done by indirect immunofluorescence with Y-11 HApolyclonal and Alexa 488 anti-rabbit for ␣ s and anti-myc and Alexa 594 anti-mouse antibodies for ␤ 1 and ␤ 5 . ␣/␤ (row 3) shows an overlay of ␣ and ␤ immunofluorescence images.

TABLE I Summary of immunofluorescent subcellular localization of ␣ and ␤␥
The indicated combinations of subunits were expressed in HEK293 cells, and the subcellular localization of the expressed ␣ or ␤ was determined as described under "Experimental Procedures," using antibodies directed against an HA epitope (␣) or myc epitope (␤). ϩϩ, prominent PM localization of the ␣ or ␤ subunit with little or no observable staining of cytoplasm or intracellular structures; ϩ, both PM and cytoplasmic localization in individual cells with the relative amounts of PM and cytoplasm staining varying among cells; ϩ/Ϫ, barely detectable PM localization; Ϫ, homogeneously distributed throughout the cytoplasm, except where indicated by footnotes, with no PM localization.
demonstrated signal transduction pathways utilizing overexpressed ␤ 5 ␥ 2 subunits but not ␤ 5 ␥ 3 subunits (24). ␤ Subtype-specific Targeting of Mutant ␣ s -The results above demonstrated the ability of the ␤␥ complex to help promote PM binding of wild-type ␣ s . Because our previous results demonstrated that mutation of the N-terminal ␤␥ contact region of ␣ s resulted in its inability to localize at the plasma membrane, we wanted to test whether co-expression of ␤␥ with a ␤␥-binding-deficient mutant of ␣ s could overcome the effects of decreased ␤␥ binding ability. The previously characterized ␤␥-binding mutant, ␣ s IEKϩ, was co-transfected with ␤ 1-5 and ␥ 2 . The ␤ subunits and ␣ s IEKϩ were then visualized by indirect immunofluorescence. As shown previously, ␣ s IEKϩ was totally cytoplasmic with no visible membrane staining (6). Coexpression with ␤ 1 ␥ 2 , ␤ 2 ␥ 2 , and ␤ 4 ␥ 2 resulted in ␣ s IEKϩ localizing at the plasma membrane as was the case with the wildtype ␣ s (Fig. 3, A1, B1, D1). Visualization of ␤ 1 , ␤ 2 , and ␤ 4 demonstrated that all three ␤ subunits co-localized with ␣ s IEKϩ at the plasma membrane (Fig. 3, A2, B2, D2). However, co-expression of ␤ 3 ␥ 2 was unable to target ␣ s IEKϩ to the plasma membrane (Fig. 3C1). Visualization of the ␤ 3 subunit showed that at least some of it was in fact localized at the plasma membrane, and the merged photo clearly shows the red membrane staining of the ␤ 3 subunit and green cytoplasmic staining of the ␣ s IEKϩ (Fig. 3, C2 and C3). As would be expected from the results with the wild-type ␣ s subunit (Fig.   1B), ␤ 5 ␥ 2 was unable to target the ␣ s IEKϩ mutant to the plasma membrane (Fig. 3E1). Not only was ␤ 5 ␥ 2 unable to target ␣ s IEKϩ to the plasma membrane, but the two subunits failed to co-localize within the cytoplasm. ␣ s IEKϩ displayed an even cytoplasmic distribution with no concentration in the perinuclear or Golgi regions where the ␤ 5 subunit was localized (Fig. 3E3).
As a complementary approach, we utilized cellular fractionation to determine the localization of the ␣ s subunits in the entire population of expressing cells. HEK293 cells were transiently transfected with the various combinations of subunits and grown for 3 days. The cells were lysed, the soluble and particulate fractions representing the cytoplasmic and membrane cellular fractions were isolated, and the level of ␣ s subunit in each fraction was determined. The results clearly show that co-expression of nearly every ␤␥ complex combination tested was able to recover the membrane binding ability of ␣ s IEKϩ (Fig. 6A). The order of effectiveness was which is exactly what was seen with immunofluorescence. Unexpectedly, fractionation results seem to indicate that ␤ 3 ␥ 2 can recover minimal membrane binding of ␣ s IEKϩ even though immunofluorescence did not detect any ␤ 3 ␥ 2 -dependent PM binding of ␣ s IEKϩ. Unlike immunofluorescence, however, these fractionation experiments do not differentiate between PM and intracellular membrane binding. The only subunits that were completely ineffectual at targeting ␣ s IEKϩ to the membrane fraction were the ␤ 5 subunits, which agrees with the immunofluorescence results.
␤ Subtype-specific Targeting of Mutant a q -The experiments above were repeated using ␣ q IE, a ␤␥-binding-defective mutant of ␣ q . Different ␤␥-binding-defective mutants of ␣ s and ␣ q were used in these experiments because we wanted to use ␣ subunits with the smallest number of mutations capable of blocking PM localization. Our previous results demonstrated that two mutations, I19A and E20A, are sufficient for ␣ q to lose all PM localization, whereas ␣ s required additional N-terminal mutations to lose PM binding (6). Co-expression of the ␤ x ␥ 2 subunits with ␣ q IE demonstrated a subtype-specific effect on targeting of ␣ q IE to the plasma membrane (Fig. 4). Unlike ␣ s we could not visualize the ␣ and ␤ subunits in the same cells. This was because the polyclonal antibody against the HA tag could not detect the HA tag in ␣ q , thus we were forced to use monoclonal antibodies to visualize the ␣ and ␤ subunits in different cells. ␤ 1 ␥ 2 and ␤ 2 ␥ 2 were each able to recruit ␣ q IE to the PM as they were with ␣ s IEKϩ (Fig. 4). However, ␤ 3 ␥ 2 and ␤ 4 ␥ 2 differed in their ability to promote PM binding of ␣ q IE, compared with their effect on ␣ s IEKϩ. Although ␤ 3 ␥ 2 was unable to target ␣ s IEKϩ to the plasma membrane and co-localize with it ( Fig.   FIG. 2. Effect of the expression of complementary G protein subunits on ␤ and ␥ expression and localization in HEK293 cells. HEK293 cells were transiently transfected with myc-tagged ␤ 1 or ␥ 3 alone or with other G protein subunits, which are able to complex with them and form functional G proteins. The transfected proteins were fixed and then visualized by indirect immunofluorescence with antimyc and Alexa 594 anti-mouse antibodies for ␤ 1 and anti-␥ 3 and Alexa 488 anti-rabbit for ␥ 3 . 3), it was able to promote weak PM staining of ␣ q IE (Fig. 4). The most profound difference involved ␤ 4 ␥ 2 . This subunit was poor at targeting ␣ q IE to the plasma membrane even though it was one of the most efficient subunits at targeting of ␣ s IEKϩ to the PM. Examination of Fig. 4 indicates that ␤ 4 ␥ 2 was, in fact, localized at the PM with seemingly the same efficiency as ␤ 1 ␥ 2 and ␤ 2 ␥ 2 but was unable to recruit ␣ q IE to the PM with the same efficiency (Fig. 4). ␤ 5 ␥ 2 failed to promote PM targeting of ␣ q IE and did not co-localize with ␣ q IE; when co-expressed, ␤ 5 ␥ 2 was localized to the perinuclear region and intracellular membranes while ␣ q IE remained cytoplasmic (Fig. 4).
Cellular fractionation results (Fig. 6B) from ␣ q IE paralleled the results of immunofluorescence. ␤ 5 ␥ 2 was totally ineffectual at recovering membrane binding, and the other ␤␥ dimers were more effective in order of ␤ 1 ␥ 2 ϭ ␤ 2 ␥ 2 Ͼ ␤ 4 ␥ 2 Ͼ ␤ 3 ␥ 2 . Comparison of the ␣ q IE fractionation to that of ␣ s IEKϩ indicates that ␤␥ complexes are less efficient at recovering the plasma membrane binding of ␣ q IE than ␣ s IEKϩ, despite the fewer mutations in ␣ q IE.
Effect of ␥ Subunits on the Targeting and Co-localization of the ␣ Subunits-So far our results have indicated a ␤ subtypespecific targeting of ␣ s and ␣ q mutants, which have had their N-terminal ␤␥-binding region disrupted. Although co-expression of ␤␥ complexes is able to recover proper PM targeting and co-localization in most cases, ␤ 3 ␥ 2 was unable to do so for ␣ s IEKϩ and ␣ q IE, ␤ 4 ␥ 2 was unable to do so for ␣ q IE, and no co-localization was ever seen between ␤ 5 and any ␣ subunits (summarized in Table I). It is possible that such differences could be due to improper or inefficient binding between these particular ␤ and ␥ subunits. One report demonstrated that the ␥ 2 subunit associates weakly with the ␤ 3 , ␤ 4 , and ␤ 5 subunits but the ␥ 3 subunit binds to all five ␤ subunits with similar high affinities (17). To determine if this could be the cause of our ␤ subtype-specific results, we repeated the experiments above using ␥ 3 . ␤ 1 ␥ 3 recruited/co-localized with wild-type HA-tagged ␣ s (Fig. 1C) and ␣ q (Table I) indistinguishably from ␤ 1 ␥ 2 . The ability of ␤ 2-4 ␥ 3 to recruit/co-localize with wild-type HA-tagged ␣ s and ␣ q was indistinguishable from ␤ 2-4 ␥ 2 with the exception of ␤ 3 ␥ 3 ( Table I), suggesting that both ␥ 2 and ␥ 3 form functional complexes in our system. However, co-expression with the ␤␥-binding mutants, ␣ s IEKϩ and ␣ q IE, displayed interesting similarities and differences (Figs. 4 and 5). Immunofluorescence localization showed that ␤ 3 ␥ 3 was unable to target ␣ s IEKϩ to the PM just as ␤ 3 ␥ 2 was unable to do (Fig. 5C). Subcellular fractionation results indicate that ␤ 3 ␥ 3 may be a little worse than ␤ 3 ␥ 2 at PM targeting, but the deficiency seems to be largely a property of ␤ 3 in general and is not due to the complexed ␥ subunit. ␤ 1 ␥ 3 functioned just as well as ␤ 1 ␥ 2 at targeting and co-localization of ␣ s IEKϩ (Figs. 5A and 6A). However, ␥ 3 was less effective FIG. 4. Immunofluorescence localization of a ␤␥-binding mutant of ␣ q when co-transfected with different ␤␥ complexes. HEK293 cells were transiently transfected with 600 ng of ␣ q IE, 250 ng of myc-tagged ␤ 1 , ␤ 2 , ␤ 3 , ␤ 4 , or ␤ 5 and 150 ng of ␥ 2 or ␥ 3 . The transfected proteins were fixed and then visualized by indirect immunofluorescence with 12CA5 and Alexa 488 anti-mouse for ␣ q or anti-myc and Alexa 594 anti-mouse antibodies for ␤ 1-5 . than ␥ 2 at targeting the ␤ 2 and ␤ 4 subunits to the PM, and ␣ s IEKϩ was not effectively recruited to the PM, as determined by immunofluorescence (Fig. 5, B and D). The same results were observed when ␤ 2 ␥ 3 and ␤ 4 ␥ 3 were co-expressed with ␣ q IE (Fig. 4). However, fractionation results for ␣ s IEKϩ and ␣ q IE show no significant difference between ␤␥ complexes containing ␥ 2 and ␥ 3 except for ␣ s IEKϩ/␤ 3 ␥ 3 and ␣ q IE/␤ 1 ␥ 3 (Fig. 6). ␤ 5 , when co-expressed with ␥ 3 and either ␣ s IEKϩ or ␣ q IE, no longer localized at the perinuclear/Golgi region but instead displayed diffuse intracellular localization (Figs. 5E2 and 4). Thus, the ␤ subunits demonstrate ␥ subunit-independent (␤ 3 ) and -dependent (␤ 2 and ␤ 4 ) ␤ subtype-specific ability to promote PM localization of the mutant ␣ subunits.
Palmitoylation-We previously demonstrated that ␤␥-binding-deficient ␣ subunits of ␣ s and ␣ q , which are not targeted to the plasma membrane are less efficiently palmitoylated (6). In this report we have shown that co-expression of ␤␥ complexes is able to recover the plasma membrane binding of ␣ subunit mutants. We next wanted to determine whether ␤␥ complex overexpression also restored normal levels of palmitoylation to ␤␥-binding-deficient ␣ subunits of ␣ s and ␣ q , as determined by incorporation of radiolabeled palmitate. Because wild-type ␣ s incorporates extremely low levels of palmitate when expressed in HEK293 cells, COS cells were used in these studies. Cells were transiently transfected with the HA epitope-tagged ␣ subunits with and without various ␤␥ complexes, metabolically labeled with [ 3 H]palmitic acid and immunoprecipitated with the 12CA5 monoclonal antibody. Levels of incorporated palmitate were visualized by fluorography of immunoprecipitated proteins after SDS-PAGE fractionation. Co-expression of the ␤ 2-4 ␥ 2/3 complexes resulted in increases of incorporation of radiolabeled palmitate into ␣ s IEKϩ, which corresponded quite well with their ability to promote membrane localization of the ␣ subunit (Fig. 7A, lanes 5-10). Unexpectedly, ␤ 1 ␥ 2 promoted the incorporation of significantly more radiolabeled palmitate than ␤ 1 ␥ 3 , ␤ 2 ␥ 2/3 , or ␤ 4 ␥ 2 , despite the fact that all of these ␤␥ complexes were just as efficient at targeting ␣ s IEKϩ to the PM (Fig. 7A, lane 3 and Fig. 6). Similar results were seen with ␣ q IE except that in this case it was the two ␤ 4 complexes that promoted the incorporation of more radiolabeled palmitate than expected based on the fractionation data (Fig. 6B). The only ␤␥ complexes that were completely ineffective at promoting palmitoylation of the ␣ subunit were those containing ␤ 5 (data not shown). With this only exception, co-expression of ␤␥ complexes with the ␤␥-binding mutants of both ␣ s and ␣ q , was able to stimulate incorporation of more radiolabeled palmitate than seen with expression of the wild-type ␣ subunits alone. This was particularly true for ␣ s IEKϩ, which labeled better than wt-␣ s when co-expressed with each ␤␥ complex.
Because co-expression of ␤␥ complexes is able to recover palmitoylation and membrane binding of ␣ subunit mutants, this raises the question of whether palmitoylation is always required for membrane binding or if overexpression of ␤␥ complexes can overcome this requirement. To examine this, we tested whether co-expression of ␤␥ complexes can recover membrane binding of palmitoylation-deficient ␣ subunits. ␣ s C3S and ␣ q C9S, C10S were previously shown to be non-palmitoylated, unable to target to the PM, and incapable of coupling receptor stimulation to effector activation (34). Co-expression of each of these non-palmitoylated mutants with ␤ 1 ␥ 2 failed to promote any PM binding of the ␣ subunits (Fig. 8, C and G). Examination of the localization of the ␤ 1 ␥ 2 subunit (Fig. 8D) showed PM localization, which indicates that the transfected G protein ␤␥ complexes were being expressed and properly localized in these cells. All five pairs of ␤ x ␥ 2 subunits were unable to promote PM binding by the ␣ subunit when co-expressed with ␣ s C3S or ␣ q C9S, C10S. However, although ␣ q C9S, C10S did not localize at the PM upon co-expression of ␤ 1 ␥ 2 , it did get ex-FIG. 6. Subcellular fractionation of ␤␥-binding mutants of ␣ s and ␣ q when co-transfected with different ␤␥ complexes. A, HEK293 cells were transiently transfected with 2 g of ␣ s or ␣ s IEKϩ and 1 g of pcDNA3 or 700 ng of ␤ 1-5 and 300 ng of ␥ 2 or ␥ 3 . B, HEK293 cells were transiently transfected with 2 g of ␣ q or ␣ q IE and 1 g of pcDNA3 or 700 ng of ␤ 1-5 and 300 ng of ␥ 2 or ␥ 3 . Soluble and particulate fractions were isolated as described under "Experimental Procedures." The results shown are the means Ϯ S.D. for two experiments assayed two times. FIG. 7. Palmitoylation of ␤␥-binding mutants of ␣ s and ␣ q when co-transfected with different ␤␥ complexes. A, COS7 cells were transiently transfected with 2 g of ␣ s and 1 g of pcDNA3, or 2 g of ␣ s IEKϩ and 1 g of pcDNA3, or 2 g of ␣ s IEKϩ and 700 ng of ␤ 1-4 and 300 ng of ␥ 2 or ␥ 3 . B, COS7 cells were transiently transfected with 2 g of ␣ q and 1 g of pcDNA3, or 2 g of ␣ q IE and 1 g of pcDNA3, or 2 g of ␣ q IE and 700 ng of ␤ 1-4 and 300 ng of ␥ 2 or ␥ 3 . The ␣ subunits were metabolically labeled with [ 3 H]palmitic acid (1 mCi for ␣ s and 0.5 mCi for ␣ q ) and immunoprecipitated as described under "Experimental Procedures." The upper panel for each subunit shows the radiolabeled palmitate incorporated by each subunit and visualized by fluorography (exposures: A, 10 days; B, 66 days). Western blot analysis (lower panels) of aliquots of each immunoprecipitation show the level of each ␣ subunit.
cluded from the nucleus (Fig. 8G). This suggests that the ␤␥ complexes are, in fact, interacting with the non-palmitoylated ␣ subunits. The non-palmitoylated G proteins are simply unable to localize at the PM. ␤ 4 ␥ 2 and ␤ 5 ␥ 2 , which were shown earlier to be unable to target ␣ q IE to the PM, were similarly ineffective at excluding ␣ q C9S, C10S from the nucleus, suggesting that their deficiency is indeed in binding to the ␣ subunit.

DISCUSSION
Stable binding between the ␣ and ␤␥ subunits of G proteins is critical for their proper functioning. This binding can be disrupted by point mutations in the N-terminal region of the ␣ subunit, a region shown to contact ␤␥ (35,36). This binding deficiency gives rise to improper cellular localization, lack of palmitoylation, and impaired signaling when the ␣ s IEKϩ or ␣ q IE ␤␥-binding-deficient mutants are overexpressed in cell lines (6). In this paper we demonstrated that overexpression of ␤␥ complexes is capable of overcoming this binding deficiency of the ␤␥-binding "weakened" mutant ␣ subunits. Moreover, as described in this report, these mutants have allowed us to develop a novel system for the detection of subunit-specific differences in ␣␤␥ complex formation and intracellular targeting. In this report we demonstrated that 1) ␣ and ␤␥ reciprocally promote the PM targeting of the other subunit; 2) ␤ 5 , when co-expressed with ␥ 2 or ␥ 3 , fails to localize to the PM or promote PM localization of mutant ␣ s or ␣ q ; 3) ␤ 3 is deficient in promoting PM localization of mutant ␣ s and ␣ q , while ␤ 4 is deficient in promoting PM localization of mutant ␣ q ; and 4) palmitoylation, in addition to ␤␥ interaction, is required for PM localization of the ␣ s and ␣ q subunits.
The function of the ␤␥ complex in helping to localize the ␣ subunit to the PM has been demonstrated recently (6,33,37), but little is known about the localization of ␤␥ itself except that it is considered to be anchored at the PM by prenylation (38).
Here we have demonstrated that ␤␥ localizes to the PM very poorly when overexpressed in HEK293 cells. However, co-expression with an ␣ subunit leads to an increase in expression of ␤␥ and greater PM localization. This was seen using antibodies directed at both the ␤ and ␥ subunits. These results indicate that each of the subunits comprising the G protein heterotrimer are required for proper expression and targeting of the others. The ␤␥ complex seems to be more sensitive to this than the ␣ subunit, because overexpression of the ␣ subunits results in significant levels of expression at the PM, while overexpression of ␤␥ results in little PM localization. This difference between overexpressed ␣ and ␤␥ may reflect differences in degradation rates of the subunits or may reflect differences in the stoichiometry of ␣:␤␥ inside a cell; little is known regarding either possibility. Defining such parameters as degradation rates and stoichiometry will be critical not only for understanding further the regulation of G protein heterotrimer formation and signaling, but will also be an important component of understanding specificity among all signaling pathways inside a cell.
Every pair of ␤ and ␥ subunits examined, with the exception of the ␤ 5 ␥ 2 and ␤ 5 ␥ 3 , displayed the ability to localize efficiently at the PM when co-expressed with wild-type ␣ s and ␣ q (Table I). Studies utilizing the ␤ 5 ␥ 2 complex have demonstrated its ability to form a functional dimer that interacts selectively with ␣ q (31) and stimulates downstream effectors in a subtype-specific manner (24, 28, 39 -41). Although the ␤ 5 subunit failed to localize at the PM under any of the conditions tested, it did co-localize with the ␥ 2 subunit in perinuclear membranes, suggesting that it had dimerized with it ( Fig. 1B2 and data not shown). ␤ 5 did not co-localize with ␥ 3 , indicating that it does not dimerize with this subunit (data not shown). In addition to the inability of the ␤ 5 ␥ 2 complex to co-localize with either ␣ s or ␣ q , results from other assays were also consistent with a lack of interaction between ␣ q and ␤ 5 . First, the ␤ 5 ␥ 2 complex was unable to promote the exclusion from the nucleus of the nonpalmitoylated mutant of ␣ q (Fig. 8I). When the non-palmitoylated mutant of ␣ q was co-expressed with ␤ 1 ␥ 2 , non-palmitoylated ␣ q was excluded from the nucleus even though it could not localize at the PM (Fig. 8G). Presumably this was due to an interaction between ␣ q and ␤ 1 ␥ 2 . Second, co-expression of ␤ 5 ␥ 2 or ␤ 5 ␥ 3 with ␣ q IE had no effect on the palmitoylation of the ␣ subunit (data not shown), in contrast to the palmitoylationinducing effect of other ␤␥ combinations (Fig. 7). Taken together, our results fail to find any evidence of an ␣ q ␤ 5 ␥ 2 complex in transfected cells.
In fact, our studies are more consistent with the hypothesis that the main role of ␤ 5 is interaction with RGS proteins rather than ␥ subunit, and ␤ 5 may not participate in the formation of a classical G protein heterotrimer in vivo (42). Despite our results and those of others indicating that transfected ␤ 5 and ␥ 2 FIG. 8. Immunofluorescence localization of non-palmitoylated mutants of ␣ s and ␣ q when co-transfected with different ␤␥ complexes. HEK293 cells were transiently transfected with 600 ng of ␣ s (A); 600 ng of ␣ s C3S (B); 600 ng of ␣ s C3S, 250 ng of myc-tagged ␤ 1 , and 150 ng of ␥ 2 (C and D); 600 ng of ␣ q (E); 600 ng of ␣ q C9S, C10S (F); 600 ng of ␣ q C9S, C10S, 250 ng of myc-tagged ␤ 1 and 150 ng of ␥ 2 (G); 600 ng of ␣ q C9S, C10S, 250 ng of myc-tagged ␤ 4 and 150 ng of ␥ 2 (H); and 600 ng of ␣ q C9S, C10S, 250 ng of myc-tagged ␤ 5 and 150 ng of ␥ 2 (I). The transfected proteins were fixed and then visualized by indirect immunofluorescence with Y-11 HA-polyclonal and Alexa 488 anti-rabbit for ␣ s (A-C), 12CA5 and Alexa 488 anti-mouse for ␣ q (E-I), or anti-myc and Alexa 594 anti-mouse antibodies for ␤ 1 (D) or ␤ 4 or ␤ 5 (not shown).
can dimerize, recent studies of the composition of ␤ 5 complexes in vivo and in vitro were unable to find ␤ 5 complexed with ␥ 2 (43,44). All of the ␤ 5 complexes examined contained RGS6, RGS7, or RGS9 rather than ␥ 2 . However, our results demonstrating a co-localization of ␤ 5 with ␥ 2 raises the interesting possibility that there are, in fact, ␤ 5 ␥ 2 complexes in the Golgi that form heterotrimers with ␣ subunits. A previous report suggested that there are more ␣ subunits in the Golgi than ␤ 1-4 subunits, which would result in excess ␣ subunits that do not form ␣␤␥ heterotrimers (45). It is possible, given our results, that these extra ␣ subunits in the Golgi may be complexed with ␤ 5 ␥ 2 . This possibility will require further study to determine what the status of ␤ 5 ␥ 2 is in the cell and whether it may function in the Golgi.
Excluding the ␤ 5 subunit, which shows no interaction with either ␣ s or ␣ q when expressed with either ␥ 2 or ␥ 3 , the ␤ 3 subunit is the least efficient at translocating ␣ s or ␣ q to the PM. Importantly, this deficiency in ␤ 3 was only revealed through the use of the ␤␥-binding-deficient ␣ s IEKϩ and ␣ q IE, validating the utility of this expression system for detecting subtypespecific differences in ␣␤␥ complex formation. Although the inefficiency of ␤ 3 in promoting PM localization and palmitoylation of mutant ␣ s and ␣ q could be explained by the inability of ␤ 3 to bind ␥, several lines of evidence suggest the ␤ 3 subunit does complex with ␥ 2 : 1) ␤ 3 ␥ 2 is able to promote PM localization of wt-␣ s and co-localize with ␣ s and ␣ q at the PM (Table I); 2) ␤ 3 ␥ 2 complexes have been purified after co-expression (26); and 3) ␤ 3 ␥ 2 can activate effectors after co-expression in mammalian cells (24,26).
However, other reports have indicated that ␤ 3 and ␥ 2 do not interact (16,17,46). Indeed, one report suggested that ␥ 3 interacted with ␤ 3 better than ␥ 2 did. However, in our system, ␤ 3 ␥ 3 also inefficiently promoted PM localization and palmitoylation of ␣ s IEKϩ and ␣ q IE. The lack of consensus as to the ability of ␤ 3 to interact with ␥ 2 may relate to the different methods used to detect functional ␤ 3 ␥ 2 complexes. A key component may be the presence of the ␣ subunit to aid in ␤ 3 ␥ 2 complex formation. Thus, methods in which ␣ is absent would fail to detect ␤ 3 ␥ 2 complexes (16,17,46), while methods, such as analyses of localization or function in cells, in which ␣ is present are consistent with ␤ 3 ␥ 2 complex formation (24,26). Little is known about the temporal relationship between ␤␥ dimer formation and formation of the heterotrimer, but one report does suggest that there may be interactions between ␣ and ␤␥ within a few minutes of translation, before the proteins reach the PM (47). It may be that the simple affinities of specific ␤ and ␥ subunits to couple in vitro are not a good indication of what happens in vivo. The ␣ subunit as well as other factors may be critical in determining specificity of ␤␥ dimer formation, as was previously suggested based on in vivo localization studies (10). Regardless of the exact mechanism, our results utilizing ␤␥-binding mutants of ␣ s and ␣ q demonstrate that ␤ 3 ␥ 2 and ␤ 3 ␥ 3 fail to promote PM localization of the mutant ␣ subunits. Importantly, the system described in this report will allow us to test other ␥ subunits. It is possible that ␤ 3 , when expressed with a unique ␥ subtype, will be able to more efficiently recover PM localization of ␣ s IEKϩ and ␣ q IE.
The use of ␤␥-binding-deficient ␣ s IEKϩ and ␣ q IE also revealed a striking difference in the ability of ␤ 4 to promote PM localization of mutant ␣ s versus mutant ␣ q . The PM binding of ␣ s IEKϩ was recovered with ␤ 4 dimers just as well as with ␤ 1 dimers (Fig. 6A). However, ␤ 4 -containing dimers were unable to recover PM localization of ␣ q IE (Fig. 6B). Although this was the only ␤ subunit that displayed a profound difference in coupling between ␣ s IEKϩ and ␣ q IE, it was not the only ␤␥ dimer to preferentially target one ␣ subunit to the PM over another. ␤ 1 ␥ 3 and ␤ 3 ␥ 3 were much worse at targeting ␣ q IE and ␣ s IEKϩ, respectively, to the PM than their ␥ 2 containing counterparts. To our knowledge, this is the first demonstration of a ␥-dependent modulation of the interaction between the ␣ and ␤␥ subunits.
In a previous paper we showed that lack of membrane binding by ␣ s or ␣ q caused by lack of ␤␥ binding resulted in a corresponding decrease in palmitoylation (6). In this report we examined this further by overexpression of ␤␥ complexes to recover membrane binding of the ␤␥-binding-deficient ␣ subunits and assayed them for their ability to incorporate radiolabeled palmitate. We have found that recovery of PM localization does not occur without a corresponding increase in palmitate incorporation. However, it is possible to get a significant increase in incorporation of palmitate without a corresponding increase in PM localization. This is particularly true of ␣ s IEKϩ, which was able to incorporate more than 20 times the amount of radiolabeled palmitate when co-expressed with ␤ 1 ␥ 2 compared with wt-␣ s expressed alone. ␣ s IEKϩ, co-expressed with ␤ 1 ␥ 2 , and wt-␣ s , expressed alone, were expressed at similar levels and localized to the PM to similar degrees. Although co-expression of ␤ 1 ␥ 2 with wt-␣ s was able to increase palmitate incorporation by the wild-type subunit, it was only increased 2-fold (data not shown). It seems unlikely that the steady-state level of palmitoylation is increased for these mutant ␣ subunits without a visible increase in PM localization. A more likely explanation is that the palmitate turns over more rapidly in these mutants due to their poor ␤␥ binding. In this scenario, ␤␥ promotes PM binding and palmitoylation of both wt-␣ s and the mutant ␣ s IEKϩ, but ␣ s IEKϩ can more rapidly exchange unlabeled palmitate for the radiolabeled palmitate. This suggestion agrees with studies showing that ␤␥ protects ␣ s from depalmitoylation by thioesterases (48 -50). This could also explain the curious result that ␤ 4 was able to stimulate as much palmitate incorporation in ␣ q IE as ␤ 1 did, even though ␤ 4 was unable to promote efficient PM localization of ␣ q IE.
The importance of palmitoylation of ␣ s and ␣ q was highlighted further by our demonstration that ␤␥ fails to induce any PM localization of ␣ s C3S or ␣ q C9S, C10S, palmitoylation site mutants of ␣ s and ␣ q (Fig. 8). Previous studies of cysteine to serine palmitoylation site mutants of ␣ s and ␣ q have relied upon overexpression of the ␣ subunits alone in the absence of co-expressed ␤␥, and differential effects on the ␣ subunits' ability to bind membranes have been observed (4,51). However, the results presented here clearly demonstrate that Nterminal cysteine residues, and by inference palmitoylation, are essential for PM localization of ␣ subunits. Under conditions in which overexpression of wt forms of ␣ s or ␣ q together with ␤␥ results in efficient PM localization of all subunits, ␣ s C3S and ␣ q C9S, C10S fail to display any PM localization. Thus, palmitoylation, in addition to interactions with ␤␥, is required for stable binding of the ␣␤␥ G protein heterotrimer to the PM.
In summary, the results presented in this report demonstrate not only the reciprocal importance of ␣ and ␤␥ for PM targeting of the G protein heterotrimer, but also describe a novel system for detecting subunit subtype selectivity by employing ␤␥-binding mutants of ␣ s and ␣ q . We predict that this system will be useful for defining further ␣␤␥ subtype specificity. For example, ␤␥-binding mutants of other members of the ␣ subunit family can be created and assayed in a similar manner. In addition, this system allows us to test rapidly the ability of other ␥ subunits to complex with ␤ 1-5 and form functional complexes capable of promoting PM localization of mutant ␣ subunits. Finally, this analysis can be extended beyond subcellular localization, and it provides a means to ascer-tain subunit subtype differences in a heterotrimer's ability to couple to unique receptors.