Heterotrimer Formation, Together with Isoprenylation, Is Required for Plasma Membrane Targeting of G (cid:1)(cid:2) *

Nascent (cid:1) and (cid:2) subunits of heterotrimeric G proteins need to be targeted to the cytoplasmic face of the plasma membrane (PM) in order to transmit signals. We show that (cid:1) 1 (cid:2) 2 is poorly targeted to the PM and predomi- nantly localized to endoplasmic reticulum (ER) membranes when expressed in HEK293 cells, but co-expres-sion of a G protein (cid:3) subunit allows strong PM localization of the (cid:1) 1 (cid:2) 2 . Furthermore, C-terminal isopre- nylation of the (cid:2) subunit is necessary but not sufficient for PM localization of (cid:1) 1 (cid:2) 2 . Isoprenylation of (cid:2) 2 and localization of (cid:1) 1 (cid:2) 2 to the ER occurs independently of (cid:3) expression. Efficient PM localization of (cid:1) 1 (cid:2) 2 in the absence of co-expressed (cid:3) is observed when a site for palmitoylation, a putative second membrane targeting signal, is introduced into (cid:2) 2 . When a mutant of (cid:3) s is targeted to mitochondria, (cid:1) 1 (cid:2) 2 follows, consistent with an important role for (cid:3) in promoting subcellular localization of (cid:1)(cid:2) . Furthermore, we directly demonstrate the requirement for (cid:3) by showing that disruption of heterotrimer formation by the introduction of (cid:3) PM the for efficient PM targeting of (cid:2)(cid:3) not dimer, interaction of the (cid:2)(cid:3) dimer with the (cid:1) is critical for PM targeting of (cid:2)(cid:3) . In this report, we examined the importance of the (cid:1) subunit in PM targeting of (cid:2)(cid:3) . We show that isoprenylation of the (cid:3) subunit is necessary but not sufficient for PM localization of (cid:2)(cid:3) , and expression of (cid:1) is not required for (cid:3) isoprenylation. We demonstrate that introduction of an additional membrane targeting signal into the (cid:3) subunit can overcome the reliance of (cid:2)(cid:3) on (cid:1) for PM targeting. In addition, (cid:2)(cid:3) accompanies an (cid:1) subunit mis-targeted to mitochondria. Finally, we present the first direct test of the necessity for heterotrimer assembly for PM localization of (cid:2)(cid:3) by demonstrating that (cid:1) binding-deficient mutants of (cid:2)(cid:3) fail to

Heterotrimeric G proteins 1 are composed of ␣ and ␤␥ subunits. The ␤␥ complex only dissociates when denatured and hence is a functional monomer under physiological conditions. Upon receptor activation the ␤␥ dimer is freed from GTP-bound ␣ and relays signals to downstream molecules until it reassociates with GDP-bound ␣, re-forming the heterotrimer. To perpetuate this G protein cycle, the trimer must be tethered to the cytoplasmic face of the PM. This crucial subcellular localization is promoted by the covalent attachment of lipids to the subunits. Three lipid modifications have been found in G proteins, namely myristoylation and/or palmitoylation for the ␣ subunit and isoprenylation for the ␥ subunit. Myristoylation is the covalent attachment of a 14-carbon saturated myristate to an N-terminal glycine through an amide bond, whereas palmitoylation is a 16-carbon saturated palmitate linked to a cysteine via a thioester bond. Isoprenylation is a lipid modification in which an unsaturated, 15-carbon farnesyl isoprenoid or 20carbon geranylgeranyl isoprenoid is linked to a cysteine, via a thioether bond.
Mechanisms underlying the PM targeting of the ␣ subunit have been studied in some detail (1)(2)(3). The available data suggest a model in which myristoylation and/or binding to ␤␥ subunits constitutes an initial membrane targeting signal for the ␣ subunit. Subsequently, palmitoylation functions as a second signal that specifies localization to the PM. In contrast to the ␣ subunit, relatively less is known about how ␤␥ is targeted to the PM. Either a 15-carbon farnesyl or 20-carbon geranylgeranyl isoprenoid is linked to a cysteine residue in the so-called CAAX motif in the C terminus of the ␥ subunit (4,5). The CAAX box (where C is a cysteine, A is commonly an aliphatic amino acid, and X can be one of several amino acids) is a consensus sequence for isoprenylation. The X residue is thought to specify which isoprenoid group will be linked to the cysteine. Among 12 human ␥ subunits thus far identified, ␥ 1 , ␥ 9 , and ␥ 11 have serine in the X position and are farnesylated, and the rest of them have leucine and are modified with a geranylgeranyl group. It has been generally agreed that isoprenylation of the ␥ subunit is essential in PM targeting of the dimer but whether the modification is sufficient has not been defined. Other prenylated proteins, such as members of the Ras family of small GTPases, require an additional second signal for their PM targeting (6,7). H-Ras and N-Ras are palmitoylated at cysteines upstream of the prenylcysteine, whereas K-Ras contains polybasic lysines adjacent to the CAAX box. Ste18p, the ␥ subunit in Saccharomyces cerevisiae, is modified with palmitate at the cysteine immediately next to the prenylcysteine (8). However, none of the human ␥ have such cysteines or a stretch of basic residues flanking its CAAX motif.
Previously we found that the ␤␥ complex was localized poorly to the PM when transiently expressed in HEK293 cells, whereas co-expression of the ␣ subunit led to strong PM localization of ␤␥ (9). This suggests that the complete information required for efficient PM targeting of ␤␥ is not contained within the ␤␥ dimer, and interaction of the ␤␥ dimer with the ␣ subunit is critical for PM targeting of ␤␥. In this report, we examined the importance of the ␣ subunit in PM targeting of ␤␥. We show that isoprenylation of the ␥ subunit is necessary but not sufficient for PM localization of ␤␥, and expression of ␣ is not required for ␥ isoprenylation. We demonstrate that introduction of an additional membrane targeting signal into the ␥ subunit can overcome the reliance of ␤␥ on ␣ for PM targeting. In addition, ␤␥ accompanies an ␣ subunit mis-targeted to mitochondria. Finally, we present the first direct test of the necessity for heterotrimer assembly for PM localization of ␤␥ by demonstrating that ␣ binding-deficient mutants of ␤␥ fail to localize to the PM, even when co-expressed with ␣. The results presented herein are consistent with a model in which both heterotrimer assembly and lipid modifications, working in concert, target ␤␥ and ␣ to the PM.

EXPERIMENTAL PROCEDURES
Cell Culture-HEK293 and COS7 cells were obtained from the American Type Culture Collection (Manassas, VA) and grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and maintained at 37°C in a 95% air, 5% CO 2 -humidified atmosphere.
Transfection-Unless otherwise noted, cells were seeded 1 day before transfection, and 1 g of total plasmid DNA at a 6:3:1 ratio of ␣:␤:␥ was transfected into the cells using FuGENE 6 (Roche Applied Science). Cells were incubated overnight, transferred to new plates, and grown for 24 h prior to subsequent manipulation.
Immunofluorescence Microscopy-Cells were fixed with 3.7% formaldehyde in phosphate-buffered saline (PBS) for 15 min and permeabilized by incubation in blocking buffer (2.5% nonfat milk and 1% Triton X-100 in Tris-buffered saline) for 20 min. Cells were then incubated with the indicated primary antibodies in blocking buffer for 1 h. The cells were washed with blocking buffer and incubated in a 1:250 dilution of a goat anti-mouse or a goat anti-rabbit antibody conjugated with either Alexa 488 or Alexa 594 for 30 min. The coverslips were washed with 1% Triton X-100 in Tris-buffered saline, rinsed in distilled water, and mounted on glass slides with Prolong Antifade reagent (Molecular Probes, Eugene, OR). Microscopy was performed with an Olympus BX60 microscope. Images were recorded with a Sony DKC-5000 digital camera and transferred to Adobe Photoshop for digital processing.
Confocal Microscopy-Coverslips were prepared for confocal microscopy as described above under "Immunofluorescence Microscopy." Representative images were recorded by confocal microscopy at the Kimmel Cancer Center Bioimaging Facility using a Bio-Rad MRC-600 laser scanning confocal microscope running CoMos 7.0a software and interfaced to a Zeiss Axiovert 100 microscope with Zeiss Plan-Apo 63 ϫ 1.40 NA oil immersion objective. Dual-labeled samples were analyzed using simultaneous excitation at 488 and 568 nm. Images of "x-y" sections through the middle of a cell were recorded.
Ni-NTA Pull Down of ␤ 1 -Transfected cells were washed once with ice-cold PBS and lysed in lysis buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 1 mM EDTA, 2.5 mM MgCl 2 ) supplemented with protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin, 2 g/ml aprotinin). After 1 h of incubation on ice, nuclei and insoluble material were removed by centrifugation. Ni-NTA magnetic agarose beads (Qiagen, Valencia, CA) were added to the clarified lysate, and the samples were tumbled for 2 h at 4°C. The samples were washed three times and eluted with elution buffer containing 250 mM imidazole. Eluates were separated by SDS-PAGE followed by immunoblotting. Bands were visualized by chemiluminescence (Pierce). These experiments utilized the hexahistidine tag of N-terminal Myc-His-tagged ␤ 1 .
Prenylation Assay-COS7 cells were seeded in 60-mm culture plates. 24 h later the cells were transfected with the indicated plasmids (Myc-His tagged ␤ 1 and non-tagged ␥ 2 without or with HA-tagged ␣ s ). Cells were incubated overnight and labeled with 50 Ci/ml [ 3 H]mevalolactone (American Radiolabeled Chemical, St. Louis, MO) for another 18 h in the presence of 10 M mevastatin (Biomol, Plymouth Meeting, PA). Cells were washed with ice-cold PBS and lysed. The ␤␥ complex was pulled down and purified using Ni-NTA agarose beads as described above. Eluates were separated by SDS-PAGE and transferred onto polyvinylidene difluoride membrane. The membrane was sprayed with EnHance (PerkinElmer Life Sciences) and then exposed to Hyperfilm MP (Amersham Biosciences) at Ϫ80°C for 8 -15 days. After fluorography, the ␥ 2 subunit was detected by immunoblotting. Note that because an anti-␥ 2 polyclonal antibody did not recognize ␥ 2 C68S, His-tagged ␤ 1 (12) and Myc-tagged ␥ 2 C68S were used for the ␣ s ␤ 1 ␥ 2 C68S control sample, and ␥ 2 C68S was detected on an immunoblot with an anti-Myc monoclonal antibody.
Cell Fractionation Assay-Soluble and particulate fractions were isolated as described previously (10). Briefly, 48 h after transfection HEK293 cells were washed in ice-cold PBS and lysed in hypotonic lysis buffer (50 mM Tris-HCl, pH 8, 2.5 mM MgCl 2 , 1 mM EDTA, 1 mM dithiothreitol) with protease inhibitors (0.5 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, 5 g/ml aprotinin). Cells were passed through a 27-gauge needle 10 times. Lysed cells were centrifuged at 400 ϫ g for 5 min to remove nuclei and debris. The supernatant was centrifuged at 150,000 ϫ g for 20 min at 4°C. Fractions were analyzed by SDS-PAGE and immunoblotting using the indicated antibody.
Materials-pEYFP-Mito vector (Clontech, Palo Alto, CA) was provided by Emad Alnemri (Thomas Jefferson University). pEYFP-IBV-M1 encoding an ER marker protein was a generous gift from Mark R. Philips (New York University). 9E10 monoclonal antibody was from Covance (Berkeley, CA). 12CA5 monoclonal antibody was from Roche Applied Science. Anti-HA and anti-␥ 2 rabbit polyclonal antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA).

RESULTS
In the present study, we focused on the ␣ s ␤ 1 ␥ 2 G protein heterotrimer. Immunofluorescence microscopy was utilized to examine the subcellular localization of the ␣ and ␤␥ subunits. It has been shown that ␣ s is predominantly localized to the PM when expressed alone (9, 10). As described by us previously (10), when ␤ 1 and ␥ 2 were expressed together in HEK293 cells very little PM localization was observed ( Fig. 1, a and b). However, when ␣ s was co-expressed with ␤ 1 and ␥ 2 , ␣ s and ␤ 1 strongly co-localized at the PM (Fig. 1, c and d). Expression of ␣ q also strongly promoted PM localization of ␤ 1 ␥ 2 (not shown). ␥ 2 displayed PM localization when ␣ s , ␤ 1 , and ␥ 2 were all expressed together, but ␥ 2 was also found intracellularly and not co-localizing with ␣ s or ␤ 1 (Fig. 1, e-h). Apparently some of the ␥ 2 did not form a dimer with ␤ 1 . Because localization of ␤ 1 rather than ␥ 2 appeared to be a better representative of the ␤ 1 ␥ 2 complex, most of the experiments described herein followed the localization of ␤ 1 .
Replacement of Cysteine with Serine in the CAAX Motif of the ␥ 2 Subunit Resulted in Loss of PM Targeting but Not ␣ Binding-The C terminus of the ␥ 2 subunit contains the CAAX motif, specifically the sequence of cysteine, alanine, isoleucine, and leucine. The cysteine is modified with a 20-carbon geranylgeranyl isoprenoid. We substituted a serine for the cysteine and transiently expressed ␥ 2 C68S in HEK293 cells in conjunction with wild type ␤ 1 in the presence and absence of wild type ␣ s . In immunofluorescent staining, little ␤ 1 ␥ 2 C68S was found at the PM regardless of ␣ s expression ( Fig. 2A). To test whether poor PM localization of ␤ 1 ␥ 2 C68S, when expressed with ␣ s , resulted from an inability to form a heterotrimer, the ␤ 1 ␥ 2 dimer was pulled down with Ni-NTA beads, taking advantage of an N-terminal hexahistidine tag on ␤ 1 , and immunoblotted for the ␣ s subunit. ␤ 1 ␥ 2 was able to efficiently pull down ␣ s (Fig.  2B, lane 3). Similarly, the ␤ 1 ␥ 2 C68S dimer pulled down the ␣ s subunit (Fig. 2B, lane 6), implying that ␣ s and ␤ 1 ␥ 2 C68S are capable of assembling a heterotrimer. In this assay, efficient heterotrimer formation required co-expression of all three components. When lysates from cells expressing ␤ 1 ␥ 2 or ␤ 1 ␥ 2 C68S were mixed with a lysate from cells expressing only ␣ s , heterotrimer formation was not detected (Fig. 2B, lanes 4 and 7). The results with ␥ 2 C68S indicate that non-prenylated ␥ 2 can form a complex with wild type ␣ s and ␤ 1 , but the ␤ 1 ␥ 2 C68S dimer was not localized at the PM.
The ␤ 1 ␥ 2 Complex Was Prenylated in the Absence of the ␣ s Subunit-To determine whether ␤ 1 ␥ 2 displayed very poor PM targeting without the co-expressed ␣ s because of inefficient lipid modification of the ␥ 2 subunit, a prenylation assay was carried out. Because HEK293 cells, in our hands, detached from the bottom of culture plates upon treatment with mevastatin, a hydroxymethylglutaryl-CoA reductase inhibitor, the experiments were carried out using COS7 cells. Similar to its subcellular localization in HEK293 cells, ␤ 1 ␥ 2 is poorly targeted to the PM in COS7 cells unless an ␣ subunit is also co-expressed (13). The ␤ 1 and ␥ 2 subunits were transiently transfected into COS7 cells in the absence or presence of the ␣ s subunit. The transfected cells were labeled with [ 3 H]mevalolactone in the presence of mevastatin for 18 h. The ␤ 1 ␥ 2 complexes were isolated using Ni-NTA beads. Because the ␤ 1 subunit contains the N-terminal hexahistidine tag, only ␥ 2 subunits that are bound to ␤ 1 are pulled down in these experiments. The level of isoprenoid incorporated into ␥ 2 was visualized by fluorography and that of expression of ␥ 2 was assessed by Western blotting. Mock/pcDNA3 transfection showed no nonspecific uptake of the radioactivity (Fig. 2C, lane 1), and as expected, ␥ 2 C68S failed to incorporate radioactivity (Fig. 2C,   lane 4). The ␤ 1 ␥ 2 efficiently incorporated radioactive isoprenoid in the presence of the ␣ s subunit, and virtually no difference was seen in incorporation of radioactivity into the ␤ 1 ␥ 2 without the ␣ s subunit (Fig. 2C, lanes 2 and 3). Therefore, the defect in the PM localization of ␤ 1 ␥ 2 when expressed in the absence of ␣ was not due to failure of the ␥ 2 subunit to be prenylated. In other words, the ␤ 1 ␥ 2 dimer is capable of being modified with the isoprenoid group in the absence of the ␣ subunit. It is therefore conceivable that lipid modification takes place prior to trimer formation.
Prenylated ␤ 1 ␥ 2 Dimer Was Membrane-bound-Without coexpressed ␣ s , the ␤ 1 ␥ 2 complex is prenylated but localized to the PM very poorly. Next we examined the subcellular localization of the prenylated ␤ 1 ␥ 2 by a cell fractionation assay. After transient transfection, cells were lysed in hypotonic buffer, and the soluble and particulate fractions, representing cytoplasmic and membrane fractions, were separated by ultracentrifugation. Proteins in each fraction were analyzed by Western blotting using anti-Myc monoclonal antibody to detect the ␤ 1 subunit of the ␤ 1 ␥ 2 dimer. With co-expressed ␣ s , virtually all ␤ 1 ␥ 2 complexes were found in the particulate fraction (Fig. 3A, lane 2), presumably tethered to the PM (Fig. 1d). H]mevalolactone, cells were lysed, and ␤ 1 ␥ 2 was pulled down with Ni-NTA beads. Proteins were separated by SDS-PAGE and transferred onto a polyvinylidene difluoride membrane. The membrane was exposed to a film at Ϫ80°C (upper panel). Subsequently, the membrane was subjected to Western blotting for the ␥ 2 subunit (lower panel).
When expressed alone, the ␤ 1 ␥ 2 dimers, which localized very poorly at the PM (Fig. 1, a and b), were also found mostly in the particulate fraction (Fig. 3A, lane 4). This suggests that the prenylated ␤ 1 ␥ 2 was targeted to membranes other than the PM. To examine the intracellular localization of ␤ 1 ␥ 2 more closely, we compared the subcellular localization of ␤ 1 ␥ 2 with an ER marker protein using confocal microscopy. ␤ 1 ␥ 2 , when expressed alone, exhibited a subcellular distribution virtually identical to the ER marker (Fig. 3B, a and b), whereas ␤ 1 ␥ 2 , when expressed with ␣ s , displayed PM localization that was clearly distinct from the ER (Fig. 3B, c and d). These results are thus consistent with a model in which the ␤ 1 ␥ 2 dimer is geranylgeranylated in the cytosol by a cytoplasmic geranylgeranyltransferase (14) and then targeted to ER, prior to ␣-dependent transit to the PM.
Introduction of a Palmitoylation Site into ␥ 2 Allows ␣ s -independent PM Targeting of ␤ 1 ␥ 2 -It has been shown that isoprenylation is necessary but not sufficient for PM targeting of the Ras family of small GTPase, and a so-called second membrane targeting signal is required for PM targeting (6,7). H-Ras and N-Ras have been shown to be palmitoylated at cysteines upstream of their CAAX boxes. K-Ras possesses polybasic lysines flanking the prenylcysteine. Of interest, Ste18p, the yeast ␥ subunit, also is palmitoylated at a cysteine next to the prenylcysteine, and palmitoylation is necessary for avid PM mem-brane binding (8,15). We tested the possibility that the ␤ 1 ␥ 2 dimer becomes able to localize to the PM if ␥ 2 is bestowed with a "second" membrane targeting signal by constructing ␥ mutants with a potential palmitoylation site. A phenylalanine residue at the 66 or 67 position of the ␥ 2 subunit was replaced with a cysteine, based on the site of palmitoylation in H-Ras or Ste18p, respectively (Fig. 4A). The mutant ␥ 2 and wild type ␤ 1 were transiently expressed in HEK293 cells. Unlike the ␤ 1 ␥ 2 complex containing wild type ␥ 2 which displays predominant intracellular staining and weak or no PM staining (Fig. 4B, a), the dimer with ␥ 2 F66C or ␥ 2 F67C showed strong PM localization without overexpressed ␣ s (Fig. 4B, b and c). This result indicates that with a second signal the ␤ 1 ␥ 2 dimer is capable of trafficking to the PM in the absence of the ␣ subunit. These results suggest that ␤␥ requires a second membrane targeting signal for efficient PM localization. Palmitate linked to ␣ s may serve as the second signal for the ␤ 1 ␥ 2 dimer. Consistent with this model, ␤ 1 ␥ 2 fails to localize efficiently at the PM when co-expressed with a palmitoylation-deficient mutant of ␣ s , ␣ s C3S (Fig. 3B, f), or ␣ q , ␣ q C9S,C10S (not shown). ␤ 1 ␥ 2 Followed an ␣ s Artificially Directed to Mitochondria-To examine further a role for the ␣ subunit in ␤␥ localization, we utilized a strategy described previously (11) to study protein-protein interactions: target one protein to a location where it does not normally exist and assess the ability of the artificially directed protein to be accompanied by its partner protein.
We targeted the ␣ s subunit to mitochondria and tested whether the ␤␥ dimer follows. A mitochondria targeting signal sequence derived from Mas70p/Tom70 was fused to the N terminus of HA-tagged ␣ s to generate mito-␣ s . This peptide sequence has been shown to bring a fused protein to the mitochondrial outer membrane without subsequent import of the protein into the mitochondrial matrix (16). The integrated protein is therefore anchored at the cytosolic face of the mitochondrial outer membrane and is available to interact with other proteins. A mitochondria localization vector, pEYFP-Mito, was used as a mitochondrial marker. Wild type ␣ s exhibited pronounced PM staining, displaying little or no overlap with the mitochondrial marker (Fig. 5, a-c). On the other hand, mito-␣ s was co-localized with the mitochondrial marker (Fig. 5, d-f). Overlaying of the two pictures showed conclusively that mito-␣ s localized at  1 and 2) or without (lanes 3 and 4) pcDNA3 encoding ␣ s . The subunits were separated into soluble (S) (lanes 1 and 3) and particulate (P) (lanes 2 and 4) fractions as described under "Experimental Procedures." ␤ 1 was detected in an immunoblot using an anti-Myc monoclonal antibody. B, ␤ 1 and ␥ 2 were transiently expressed in HEK293 cells in the absence (a and b) or presence of ␣ s (c and d) or ␣ s C3S (e and f). pEYFP vector encoding a partial protein from infectious bronchitis virus was coexpressed as an ER marker (a, c, and e). Transfected ␤ 1 ␥ 2 was visualized by immunofluorescent staining using anti-Myc monoclonal antibody directed against Myc-His-tagged ␤ 1 and Alexa 594 anti-mouse antibody (b, d, and f). Images were recorded using confocal laser scanning microscopy, as described under "Experimental Procedures." Bar, 10 m. mitochondria, as demonstrated by the yellow color in Fig. 5f. When ␤ 1 and ␥ 2 were expressed with mito-␣ s , the ␤ 1 ␥ 2 dimer accompanied mito-␣ s to mitochondria (Fig. 5, j and k), as a superimposed image clearly demonstrates co-localization (Fig.  5, l). In contrast, the ␤ 1 ␥ 2 complex was found at the PM when expressed with wild type ␣ s (Fig. 5, g-i).
Impaired Interaction of ␤␥ with ␣ Prevented PM Localization of ␤␥-To address more directly the requirement for assembly of the ␣␤␥ heterotrimer in ␤␥ localization, we tested the subcellular localization of ␣-binding-deficient ␤ mutants. Two surfaces of ␤, the ␣ switch interface and the ␣ N-terminal interface, contain important residues for interaction with the ␣ subunit (17,18). Mutation of the putative ␣-contacting residues in the ␤ subunit resulted in a decreased affinity for the ␣ subunit (12,19). Based on previous findings, we introduced the mutations, I80A, N88A/K89A, L117A, D228R, or D246S (note that Ile-80, Asn-88, and Lys-89 are in the ␣ N-terminal interface, and Leu-117, Asp-228, and Asp-246 are in the ␣ switch interface), into Myc-His-tagged ␤ 1 and transiently expressed them in conjunction with wild type ␥ 2 in HEK293 cells. All mutant ␤ 1 ␥ 2 showed meager PM localization, similar to wild type ␤ 1 ␥ 2 (Fig. 6A, a). Just as wild type ␤ 1 ␥ 2 displayed much greater PM localization when co-expressed with wild type ␣ s , the mutant ␤ 1 ␥ 2 complex exhibited stronger PM membrane targeting when expressed with ␣ s (Fig. 6A, b). The ability of ␣ s to promote PM localization of the ␣-binding-deficient mutants of ␤ 1 suggests that the ␤ 1 mutants are not completely unable to interact with ␣. Consistent with this interpretation, others (12,19) using these ␤ 1 mutants observed varying degrees of loss of ␣ binding, depending upon the assay used.
Next, we expressed the ␤ 1 mutants and wild type ␥ 2 in conjunction with an ␣ s mutant, ␣ s IEKϩ, that contains mutations to five N-terminal amino acids at the ␤␥ binding interface. Our previous work (10) demonstrated that this mutant lost its ability to localize to the PM when expressed alone, but coexpression of wild type ␤␥ restored the PM localization of the ␣ s IEKϩ mutant. It was expected that a combination of the ␤␥-binding-deficient mutant of ␣ s and an ␣-binding-deficient mutant of ␤ 1 would result in more impaired heterotrimer formation. Consistent with this prediction, a ␤ 1 ␥ 2 complex containing ␤ 1 D246S (Fig. 6A, c) and, to lesser extent, ones with ␤ 1 D228R and ␤ 1 N88A/K89A (not shown) were poorly localized at the PM when co-expressed with ␣ s IEKϩ.
With these results, we sought to construct ␤ mutants with more severe mutations, to generate ones that are less capable of binding to wild type ␣. Three mutants were created by combining mutations in both interfaces. ␤ 1 NKD contains the were lysed 48 h after transfection, and ␤ 1 ␥ 2 dimers were pulled down using Ni-NTA beads by virtue of the hexahistidine tag on Myc-Histagged ␤ 1 or ␤ 1 mutants. Samples were separated by SDS-PAGE followed by immunoblotting using anti-␥ 2 antibody (lanes 1-4) or anti-HA antibody (lanes 5-7) (upper panel). The level of ␥ 2 or ␣ s expression in the lysates was assessed by Western blotting (lower panel). C, expression vectors encoding the following proteins were transfected into HEK293 cells: ␤ 1 NKD and ␥ 2 (a); ␣ s , ␤ 1 NKD, and ␥ 2 (b); ␤ 1 NKDD and ␥ 2 (c); ␣ s , ␤ 1 NKDD, and ␥ 2 (d). As above, localization of the ␤ 1 ␥ 2 dimer was visualized by immunofluorescent staining using anti-Myc monoclonal and Alexa 594 anti-mouse antibodies. mutations N88A, K89A, and D246S; ␤ 1 NKDD is like ␤ 1 NKD with an additional D228R mutation, and ␤ 1 INKDD is further mutated at I80A. To allay concerns that multiple mutations impede proper folding of the ␤ 1 protein, we checked their ability to bind the ␥ 2 subunit. The ␤ 1 NKD and ␤ 1 NKDD mutants showed ␥ 2 binding similar to wild type ␤ 1 as assessed by Ni-NTA pull-down assay (Fig. 6B, lanes 1-3). Interestingly, ␤ 1 INKDD, containing one additional mutation, failed to associate with the ␥ 2 subunit (Fig. 6B, lane 4), and thus ␤ 1 INKDD was not analyzed further. The abated capability of the mutants to interact with ␣ s was also confirmed by a Ni-NTA pull-down assay (Fig. 6B, lanes 6 and 7). Collectively, the ␤ 1 NKD and ␤ 1 NKDD mutants are correctly folded yet substantially defective in association with ␣ s . When these two mutants were expressed with wild type ␥ 2 , the ␤ 1 ␥ 2 complex was localized to the PM poorly, similar to wild type ␤ 1 (Fig. 6C, a and c). Importantly, co-expression of wild type ␣ s did not promote PM localization of the dimer containing either mutant (Fig. 6C, b  and d). Expression of the ␣ s subunit was confirmed by double staining of the subunit in the same cells. Collectively, impaired interaction of the ␤␥ dimer with the ␣ subunit resulted in poor PM targeting of the dimer. These results underscore the significance of proper heterotrimer formation and indicate that the ␣ subunit plays an important role in PM targeting of the ␤␥ dimer.

DISCUSSION
Data presented here refine the requirements for PM targeting of the G protein ␤␥ complex. In addition to demonstrating that isoprenylation of ␥ is required but not sufficient, our results reveal a crucial role for the G protein ␣ subunit. Thus, both heterotrimer assembly and lipid modifications function together to promote proper PM localization of ␤␥.
Substitution of cysteine 68 with serine in the C terminus of ␥ 2 prevented attachment of isoprenoid to it, and ␤ 1 ␥ 2 C68S exhibited virtually no PM localization, consistent with earlier immunofluorescence observations in COS cells (20). The additional co-expression of ␣ s failed to promote PM localization of ␤ 1 ␥ 2 C68S ( Fig. 2A), although ␣ s strongly promoted PM localization of ␤ 1 ␥ 2 (Fig. 1d) (9). In addition, we demonstrated that the ␥ 2 C68S mutant can form a heterotrimer with co-expressed ␣ s and ␤ 1 subunits as assessed by Ni-NTA pull-down assay (Fig. 2B). Although prenylation of the ␤␥ complex has been reported to increase its affinity for the ␣ subunit (14,21), prenylation is not a strict requirement for heterotrimer formation. Consistent with this, none of the subunits in a crystallized ␣␤␥ complex contained lipid modifications; the C68S ␥ mutant produced in Sf9 cells was able to assemble with ␣ and ␤ subunits (17). The inability of the ␤ 1 ␥ 2 C68S complex to localize to the PM when co-expressed with ␣ s , even though it is capable of binding ␣ s , implies that heterotrimer formation alone is not sufficient for PM targeting of the ␤␥ dimer.
Heterotrimer formation, however, appears to be necessary for localization of ␤␥, and several lines of evidence are consistent with a role for ␣ in PM targeting of ␤␥. First, we demonstrated previously that transiently expressed ␤␥ was poorly targeted to the PM (9) and was found predominantly at intracellular membranes (Fig. 1, a and b, Fig. 3B, b, and Fig. 4B, a). However, co-expression of an ␣ subunit promoted strong PM localization of the ␤␥ dimer (9) (Fig. 1, d and g, and Fig. 3B, d).
A recent report (13) confirmed these results using a green fluorescent protein-tagged ␥ in COS cells. Second, when ␣ s was targeted to mitochondria the ␤ 1 ␥ 2 subunits followed (Fig. 5). The ability of ␣ s , in this case misdirected to mitochondria, to mistarget ␤␥ is consistent with a prominent role for ␣ subunits in guiding ␤␥ to its appropriate cellular destination.
Third, we directly tested the effects of impaired ␣␤␥ assem-bly in the subcellular localization of the ␤␥ dimer by mutating putative ␣-binding residues in the ␤ subunit. Currently available crystal structure models of the ␣␤␥ complex indicate that the ␤ subunit contacts the ␣ subunit at two surfaces, termed the ␣ N-terminal interface and the ␣ switch interface (17,18). Others reported (12,19) that introduction of a mutation into the ␤ subunit in either interface resulted in reduced ability to form a heterotrimer properly. However, when we examined subcellular localization of mutant ␤ 1 ␥ 2 complexes in which the ␤ 1 contained single mutations, or the double N88A/K89A mutation, ␣ s was able to promote efficient PM localization of the mutant ␤␥. This is consistent with demonstrations that such ␤ 1 mutants still retain some ability to interact with ␣ (12, 19). Nonetheless, a defect in PM localization of mutant ␤ 1 ␥ 2 was revealed when an ␣-binding defective ␤ 1 mutant was expressed with ␥ 2 and a previously described ␤␥-binding defective ␣ s mutant ␣ s IEKϩ (9) (Fig. 6A, c). Combination of the ␤␥-binding defective ␣ and the ␣-binding defective ␤ impeded proper heterotrimer assembly, resulting in poor PM targeting. We further constructed the mutants ␤ 1 NKD and ␤ 1 NKDD with combined mutations in both ␣-binding interfaces to achieve more impaired interaction with wild type ␣ subunit. Importantly, coexpression of ␣ s failed to promote PM localization of ␤ 1 NKD or ␤ 1 NKDD (Fig. 6C, b and d). The ␤ 1 mutants were capable of binding ␥ 2 (Fig. 6B), and thus the failure of PM targeting was not due to misfolding of the mutant ␤ 1 subunit. Collectively, the results with ␣-binding defective ␤ 1 mutants clearly demonstrate that interaction of the ␤␥ subunit with the ␣ subunit is critical in PM targeting of the ␤␥ dimer. To our knowledge, these results are the first to show explicit evidence of the significance of heterotrimer assembly in ␤␥ localization at the PM.
Fourth, studies in model organisms indicate a role for ␣ in targeting ␤␥. In the yeast S. cerevisiae, ␤␥ is defective in localizing at the PM in an ␣ subunit (Gpa1) null mutant (22). Moreover, expression of the yeast ␣ Gpa1p rescues PM localization of a cytoplasmic ␤␥ mutant in which the palmitoylation site in ␥ is mutated (15). In addition, a recent study of G protein localization in Caenorhabditis elegans showed that depletion of an ␣ subunit resulted in failure of the ␤␥ subunits to localize properly (23). In C. elegans embryos, GPB-1, the ␤ subunit, and GOA-1, the ␣ i/o subunit, were found at the cell PM and on microtubule asters. When expression of GOA-1 and GPA-16, the widely expressed ␣ subunit with redundant functions to GOA-1 in C. elegans, was abrogated by RNA interference, GPB-1 lost its aster and PM localization (23). The orientation role of the ␣ subunit in ␤␥ localization may be widespread.
Recent findings (24,25) revealed that Ras undergoes prenylation and then transits via intracellular membranes to the PM rather than moving directly from the cytosol to the PM as was once thought. G protein ␤␥ subunits may take a similar pathway to the PM. The enzymes that catalyze proteolytic cleavage of the last three amino acids and methylation of the carboxyl group of the prenylcysteine have been cloned recently and identified as membrane-bound proteins at the ER (26,27). Thus, Ras is prenylated in the cytosol and then targeted to the cytoplasmic face of the ER where additional C-terminal processing takes place. The ␤␥ dimer is similarly prenylated in the cytoplasm and then presumably targeted to the ER to undergo subsequent CAAX processing. Interaction with an ␣ subunit does not appear to be required for the initial step in ␤␥ trafficking. ␤␥ is localized to intracellular membranes in the absence of ␣ co-expression (Fig. 3), and ␤␥ undergoes prenylation equally well in the absence or presence of ␣ expression (Fig.  2C). It has been known that, in addition to the CAAX processing, a second signal in the hypervariable region is required for