Role of Molecular Chaperones in G Protein β5/Regulator of G Protein Signaling Dimer Assembly and G Protein βγ Dimer Specificity*

The G protein βγ subunit dimer (Gβγ) and the Gβ5/regulator of G protein signaling (RGS) dimer play fundamental roles in propagating and regulating G protein pathways, respectively. How these complexes form dimers when the individual subunits are unstable is a question that has remained unaddressed for many years. In the case of Gβγ, recent studies have shown that phosducin-like protein 1 (PhLP1) works as a co-chaperone with the cytosolic chaperonin complex (CCT) to fold Gβ and mediate its interaction with Gγ. However, it is not known what fraction of the many Gβγ combinations is assembled this way or whether chaperones influence the specificity of Gβγ dimer formation. Moreover, the mechanism of Gβ5-RGS assembly has yet to be assessed experimentally. The current study was undertaken to directly address these issues. The data show that PhLP1 plays a vital role in the assembly of Gγ2 with all four Gβ1–4 subunits and in the assembly of Gβ2 with all twelve Gγ subunits, without affecting the specificity of the Gβγ interactions. The results also show that Gβ5-RGS7 assembly is dependent on CCT and PhLP1, but the apparent mechanism is different from that of Gβγ. PhLP1 seems to stabilize the interaction of Gβ5 with CCT until Gβ5 is folded, after which it is released to allow Gβ5 to interact with RGS7. These findings point to a general role for PhLP1 in the assembly of all Gβγ combinations and suggest a CCT-dependent mechanism for Gβ5-RGS7 assembly that utilizes the co-chaperone activity of PhLP1 in a unique way.

tein heterotrimer (composed of G␣, G␤, and G␥ subunits), which in turn causes the release of G␣-GTP from the G␤␥ dimer (2)(3)(4). Both G␣-GTP and G␤␥ propagate and amplify the signal by interacting with effector enzymes and ion channels (1,5). The duration and amplitude of the signal is dictated by receptor phosphorylation coupled with arrestin binding and internalization (6) and by regulators of G protein signaling (RGS) proteins, which serve as GTPase-activating proteins for the GTP-bound G␣ subunit (7,8). The G protein signaling cycle is reset as the inactive G␣-GDP reassembles with the G␤␥ dimer and G␣␤␥ re-associates with the GPCR (5).
To fulfill its essential role in signaling, the G protein heterotrimer must be assembled post-translationally from its nascent polypeptides. Significant progress has been made recently regarding the mechanism by which this process occurs. It has been clear for some time that the G␤␥ dimer must assemble first, followed by subsequent association of G␣ with G␤␥ (9). What has not been clear was how G␤␥ assembly would occur given the fact that neither G␤ nor G␥ is structurally stable without the other. An important breakthrough was the finding that phosducin-like protein 1 (PhLP1) functions as a co-chaperone with the chaperonin containing tailless complex polypeptide 1 (CCT) in the folding of nascent G␤ and its association with G␥ (10 -15). CCT is an important chaperone that assists in the folding of actin and tubulin and many other cytosolic proteins, including many ␤ propeller proteins like G␤ (16). PhLP1 has been known for some time to interact with G␤␥ and was initially believed to inhibit G␤␥ function (17). However, several recent studies have demonstrated that PhLP1 and CCT work together in a highly orchestrated manner to form the G␤␥ dimer (10 -15).
Studies on the mechanism of PhLP1-mediated G␤␥ assembly have focused on the most common dimer G␤1␥2 (10,13,14), leaving open questions about the role of PhLP1 in the assembly of the other G␤␥ combinations. These are important considerations given that humans possess 5 G␤ genes and 12 G␥ genes with some important splice variants (18,19), resulting in more than 60 possible combinations of G␤␥ dimers. G␤1-4 share between 80 and 90% sequence identity and are broadly expressed (18,19). G␤5, the more atypical isoform, shares only ϳ53% identity with G␤1, carries a longer N-terminal domain, and is only expressed in the central nervous system and retina (20). The G␥ protein family is more heterogeneous than the G␤ family. The sequence identity of the 12 G␥ isoforms extends from 10 to 70% (21), and the G␥ family can be separated into 5 subfamilies (21-23). All G␥ proteins carry C-terminal isoprenyl modifications, which contribute to their association with the cell membrane, GPCRs, G␣s, and effectors (9). Subfamily I G␥ isoforms are post-translationally farnesylated, whereas all others are geranylgeranylated (22,24).
There is some inherent selectivity in the assembly of different G␤␥ combinations, but in general G␤1-4 can form dimers with most G␥ subunits (25). The physiological purpose of this large number of G␤␥ combinations has intrigued researchers in the field for many years, and a large body of research indicates that GPCRs and effectors couple to a preferred subset of G␤␥ combinations based somewhat on specific sequence complementarity, but even more so on cellular expression patterns, subcellular localization, and post-translational modifications (18).
In contrast to G␤1-4, G␤5 does not interact with G␥ subunits in vivo, but it instead forms irreversible dimers with RGS proteins of the R7 family, which includes RGS proteins 6, 7, 9, and 11 (26). All R7 family proteins contain an N-terminal DEP (disheveled, Egl-10, pleckstrin) domain, a central G␥-like (GGL) domain, and a C-terminal RGS domain (8,26). The DEP domain interacts with the membrane anchoring/nuclear shuttling R7-binding protein, and the GGL domain binds to G␤5 in a manner similar to other G␤␥ associations (27,28). Like G␤␥s, G␤5 and R7 RGS proteins form obligate dimers required for their mutual stability (26). Without their partner, G␤5 and R7 RGS proteins are rapidly degraded in cells (26,29). G␤5-R7 RGS complexes act as important GTPase-accelerating proteins for G i/o ␣ and G q ␣ subunits in neuronal cells and some immune cells (26).
It has been recently shown that all G␤ isoforms are able to interact with the CCT complex, but to varying degrees (15). G␤4 and G␤1 bind CCT better than G␤2 and G␤3, whereas G␤5 binds CCT poorly (15). These results suggest that G␤1 and G␤4 might be more dependent on PhLP1 than the other G␤s, given the co-chaperone role of PhLP1 with CCT in G␤1␥2 assembly. However, another report has indicated that G␥2 assembly with G␤1 and G␤2 is more PhLP1-dependent than with G␤3 and G␤4 (30). Thus, it is not clear from current information whether PhLP1 and CCT participate in assembly of all G␤␥ combinations or whether they contribute to the specificity of G␤␥ dimer formation, nor is it clear whether they or other chaperones are involved in G␤5-R7 RGS dimer formation. This report was designed to address these issues.

EXPERIMENTAL PROCEDURES
Cell Culture-HEK 293T cells were cultured in Dulbecco's modified Eagle's medium/F-12 (50/50 mix) growth media containing L-glutamine and 15 mM HEPES supplemented with 10% fetal bovine serum (Sigma-Aldrich). The cells were subcultured regularly to maintain growth, but were not used beyond 25 passages.
RNA Interference Experiments-Short interfering RNAs (siRNAs) were chemically synthesized (Dharmacon) to target nucleotides 608 -628 of human lamin A/C (14), nucleotides 345-365 of human PhLP1 (14), and nucleotides 172-192 of human CCT-1 (32). HEK 293T cells were grown in 12-well plates to 50 -70% confluency at which point they were transfected with siRNA at 100 nM final concentration using Oligofectamine reagent (Invitrogen) as described previously (14). 24 h later, the cells were transfected with 0.5 g each of FLAG-G␤ and HA-G␥ or HA-RGS7 in pcDNA3.1(ϩ) using Lipofectamine 2000 according to the manufacturer's protocol (Invitrogen). The cells were harvested for subsequent immunoprecipitation experiments 72 h later. 10 g of cell lysate was immunoblotted with an anti-PhLP1 antibody (33) to assess the percent PhLP1 knockdown, and 20 g was immunoblotted with anti-CCT and anti-CCT⑀ antibodies (Santa Cruz Biotechnology) to determine the percent CCT knockdown.
Dominant Interfering Mutant Experiments-HEK 293T cells were plated in 6-well plates and grown to 70 -80% confluency. The cells were then transfected with Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Each well was transfected with 1.0 g of the empty vector control, wild-type PhLP1-Myc, or PhLP1 ⌬1-75-Myc along with 1.0 g each of the indicated FLAG-G␤ and HA-G␥ or HA-RGS7 cDNAs. The cells were harvested for immunoprecipitation 48 h after transfection.
Immunoprecipitation Experiments-Transfected HEK 293T cells were washed with phosphate-buffered saline (Fisher) and solubilized in immunoprecipitation buffer (phosphate-buffered saline, pH 7.4, 2% Nonidet P-40 (Sigma), 0.6 mM phenylmethylsulfonyl fluoride, 6 l/ml protease inhibitor mixture per ml of buffer (Sigma P8340)). The lysates were passed through a 25-gauge needle 10 times and centrifuged at maximum speed for 10 -12 min at 4°C in an Eppendorf microcentrifuge. The protein concentration for each sample was determined using the DC Protein Assay Kit II (Bio-Rad), and equal amounts of protein were used in the subsequent immunoprecipitations. Approximately 150 g of total protein was used in immunoprecipitations from cells in 12-well plates and 450 g from cells in 6-well plates. The clarified lysates were incubated for 30 min at 4°C with 2.5 g of anti-FLAG antibody (clone M2, Sigma), 1 g of anti-CCT⑀ (clone PK/29/23/8d Serotec), 1.75 g of anti-Myc or 1.5 g of anti-HA (clone 3F10, Roche Applied Science) for lysates from 12-well plates or with 6.25 g of anti-FLAG for lysates from 6-well plates. Next, 30 l of a 50% slurry of Protein A/G Plus-agarose (Santa Cruz Biotechnology) was added, and the mixture was incubated for 30 min at 4°C as described (14). The immunoprecipitated proteins were solubilized in SDS sample buffer and resolved on a 10% Tris-glycine-SDS gel or a 16.5% Tris-Tricine-SDS gel for G␥. The proteins were transferred to nitrocellulose and immunoblotted using an anti-FLAG (clone M2, Sigma), anti-c-Myc (BioMol), anti-HA (Roche Applied Science), or an anti-PhLP1 antibody (14). Immunoblots were incubated with the appropriate anti-rabbit, anti-mouse, anti-goat (Li-Cor Biosciences), or anti-rat (Rockland) secondary antibody conjugated with an infrared dye. Blots were scanned using an Odyssey Infrared Imaging System (Li-Cor Biosciences), and protein band intensities were quantified using the Odyssey software. The data are presented as the mean value Ϯ S.E. from at least three experiments.
Radiolabel Pulse-Chase Assays-Radiolabel pulse-chase assays were preformed as previously described (14). Briefly, siRNA-treated or transfected HEK 293T cells in 12-well plates were washed once with 1.5 ml of methionine-free Dulbecco's modified Eagle's medium media (Mediatech) and then incubated in 1 ml of this same media at 37°C for 1 h. The media was discarded, and 400 l of media supplemented with 200 Ci/ml radiolabeled L-[ 35 S]methionine (PerkinElmer Life Sciences) was added. The cells were incubated in the radiolabeled media for 10 min at 23°C. Subsequently, the cells were washed once with 1.6 ml of Dulbecco's modified Eagle's medium media supplemented with 4 mM non-radiolabeled L-methionine (Sigma) and 4 M cycloheximide to remove the radiolabeled media and then incubated in 0.8 ml of this same media at 23°C for the chase times indicated. After the chase periods, the cells were harvested for immunoprecipitation. Radiolabeled gels were visualized with a Storm 860 PhosphorImager, and the band intensities were quantified using ImageQuant software (Amersham Biosciences). The molar ratios were determined by normalizing the band intensities to the number of methionine residues found in FLAG-G␤1, HA-G␥, FLAG-G␤5, or HA-RGS7 and then calculating the ratios. The molar ratios were consistently substoichiometric, with the HA-G␥2/FLAG-G␤1 ratio reaching ϳ0.4 by 60 min of chase (Fig. 9B) and the HA-RGS7/ G␤5 ratio reaching ϳ0.1 by 60-min chase (Figs. 8 and 9A). The lower HA-RGS7/G␤5 ratio probably results from less efficient synthesis and folding of the nascent RGS7 compared with nascent G␤5. The rate data for G␤␥ and G␤5-RGS7 assembly were fit to a first-order rate equation with background correction to determine the rate constant for assembly.
Protein Purifications-G␤1␥2, G␤5␥2, and G␤5-RGS9-1 were expressed and purified from insect cells. Recombinant baculovirus constructs were generously provided by Narasimhan Gautam of Washington University (G␤1(34)), James Garrison from the University of Virginia (G␤5 and G␥2 (35)), and Ching-Kang Chen from Virginia Commonwealth University (RGS9-1 (36)). The G␥2 subunit contained both a His 6 tag and a FLAG epitope tag on the N terminus. The RGS9-1 subunit contained only a FLAG epitope tag on the C terminus. Sf9 cells (Amersham Biosciences) were grown to a density of 2 ϫ 10 6 cell/ml and then co-infected with a G␤ and G␥2 or RGS9-1 baculovirus at an multiplicity of infection of 5 for each virus type. Cells were grown in shaker culture for ϳ60 h and then pelleted by centrifugation at 250 ϫ g for 10 min at 4°C. The supernatant was discarded, and the cell pellet was snap frozen in liquid nitrogen and stored at Ϫ80°C.
G␤1␥2 and G␤5␥2 were purified by a modification of a previously described protocols (35,37). The cell pellet from 1 liter of cells was thawed and resuspended in 100 ml of homogenization buffer (20 mM HEPES, pH 7.5, 3 mM MgCl 2 , 150 mM NaCl, 2 g/ml aprotinin, 2 g/ml leupeptin, 2 g/ml pepstatin, 20 g/ml benzamidine, and 0.1 mM phenylmethylsulfonyl fluoride). The suspension was sonicated with a tip sonicator on ice and centrifuged at 100,000 ϫ g for 1 h. The pellet was homog-enized in 100 ml of extraction buffer (homogenization buffer ϩ 0.1% polyoxyethylene 10 laurel ether) using a Dounce homogenizer and stirred on ice for 1 h. The suspension was centrifuged again at 100,000 ϫ g for 1 h. The supernatant was collected and applied to a 5-ml M2 FLAG-agarose column (Sigma-Aldrich) equilibrated in extraction buffer. The column was washed with 30 ml of extraction buffer, and the G␤␥ dimers were eluted with 15 ml of FLAG elution buffer (extraction buffer plus 250 g/ml FLAG peptide). Fractions containing the purified dimers were combined and applied to a 2 ml nickelnitrilotriacetic acid column (Novagen) equilibrated in extraction buffer plus 30 mM imidazole. The column was washed with 20 ml of this buffer, and then eluted with 10 ml of extraction buffer plus 500 mM imidazole. Fractions containing G␤␥ were combined and dialyzed in extraction buffer plus 50% glycerol, which caused a 4-fold increase in protein concentration. G␤5-RGS9-1 was purified the same way except the nickel-nitrilotriacetic acid column was skipped because the RGS9-1 protein did not contain an His 6 tag. This procedure generally resulted in ϳ1 ml of ϳ1 mg/ml protein that was 90% pure.
Metabolically labeled 35 S-PhLP1 was prepared by transforming DE3 Escherichia coli cells with a PhLP1 pET15b vector (38) and inoculating 100 ml of M9 minimal media with a single colony of cells. The culture was incubated ϳ20 h at 37°C until the absorbance at 600 nm reached 0.6 -0.7. The cells were collected by centrifugation and resuspended in 100 ml of reduced Na 2 SO 4 M9 minimal media. At this point, 12 mg of isopropyl-␤-D-thiogalactopyranoside was added along with 500 l of 2 mCi/ml [ 35 S]H 2 SO 4 . The culture was grown for 3.5 h at 37°C to an absorbance at 600 nm of ϳ1.0. The labeled PhLP1 was then purified as described previously (38).
In Vitro Binding Assays-The binding of 35 S-PhLP1 to G␤␥ or G␤5-RGS9-1 dimers was determined by mixing 35 l of a 50% slurry of M2 FLAG-agarose beads equilibrated in assay buffer (extraction buffer without protease inhibitors) with purified G␤␥ or G␤5-RGS9-1 (final concentration, 0.5 M). The 35 S-PhLP1 was then added to the reaction mixture at final concentrations ranging from 0.01 M to 2 M in a total reaction volume of 150 l. The reaction mixture was incubated on a rotator at 4°C for 1 h. Each reaction was briefly vortexed, and 50 l of the mixture was counted in a scintillation counter to obtain the total amount of PhLP1 added. Each reaction was then centrifuged for 1 min at 1000 ϫ g to separate the bound from the free 35 S-PhLP1. A 50-l aliquot of the supernatant was then counted as described above to obtain the free counts. The free counts were subtracted from the total counts to determine the counts of bound 35 S-PhLP1. Nonspecific binding was determined by running the assay in parallel with FLAG-glutathione S-transferase in place of G␤␥ or G␤5-RGS9-1. The specific binding was determined by subtracting the nonspecific binding from the total binding. Counts were converted into concentration units using the specific activity of the 35 S-PhLP1. The concentration of specifically bound 35 S-PhLP1 was then plotted versus the free PhLP1 concentration, and the K d for the interaction was determined by fitting the data to a one-to-one binding equation where B is the amount of PhLP1 bound to the beads, B max is the maximal binding of PhLP1, and K d is the dissociation constant for the interaction.

RESULTS
It has been shown previously that, to mediate G␤␥ assembly, PhLP1 must bind G␤␥ with high affinity (13,14). As a first step toward determining the ability of PhLP1 to catalyze G␤␥ dimer formation with the five G␤ subunits, we measured the interaction of PhLP1 with each G␤ subunit in complex with G␥2 by co-immunoprecipitation. Equal amounts of Myc-tagged PhLP1, G␥2, and FLAG-tagged G␤1-5 were overexpressed in HEK-293T cells. After incubation, cells were harvested and immunoprecipitated with an anti-FLAG antibody and immunoblotted with anti-Myc and anti-FLAG antibodies. Protein band intensities were quantified, and the ratio of the PhLP1-Myc band to each FLAG-G␤ band was determined (Fig. 1A). The data show that G␤s 1-4 all co-immunoprecipitated similar amounts of PhLP1 while G␤5 co-immunoprecipitated significantly less, indicating that PhLP1 binds G␤5 complexes with a lower affinity than it does G␤1-4 complexes. All five G␤s expressed equally well under these conditions, so the differences in binding cannot be attributed to different G␤ expression levels (Fig. 1A). These results suggest that PhLP1 may be involved in G␤␥ assembly of G␤1-4, but perhaps not G␤5.
To directly measure the contribution of PhLP1 to the assembly of the five G␤ isoforms with G␥, the effect of siRNA-mediated PhLP1 knockdown on G␤␥ dimer formation was measured by co-immunoprecipitation of G␥2 with the G␤s. We chose G␥2 because it is a common isoform that associates to some extent with all G␤ subunits in vitro (25). HEK 293T cells were treated with PhLP1 siRNA, a control siRNA to lamin A/C or a mock treatment with no siRNA and then co-expressed with HA-G␥2 and one of the five FLAG-tagged G␤ subunits. Cell lysates were immunoprecipitated with an anti-FLAG antibody, and the precipitate was immunoblotted with anti-HA and anti-FLAG antibodies to detect the amount of G␥2 bound to each G␤ subunit. Fig. 1B shows the levels of PhLP1 in the cell extract and the amounts of FLAG-G␤1 and HA-G␥2 in the immunoprecipitate relative to the lamin A/C siRNA control. A 75% knockdown of PhLP1 resulted in a 50% decrease in G␤1 and a striking 85% decrease in G␥2 compared with the lamin A/C control. This pattern was consistent among all the G␤ subunits except for G␤5, which had no detectable G␥2 bound under After 48 h, cells were lysed, immunoprecipitated with an anti-FLAG antibody, and immunoblotted with anti-Myc or anti-FLAG antibodies. The graph represents the ratio of the PhLP1-Myc/FLAG-G␤ band intensities for all 5 G␤s. Bars represent the average Ϯ S.E. from three separate experiments. A representative blot is shown below the graph. B and C, cells were treated with siRNA against PhLP1, lamin A/C, or no siRNA as indicated. Twenty-four hours later, cells were transfected with the indicated FLAG-G␤ subunit and HA-G␥2 cDNAs. After 72 additional hours, cells were lysed, immunoprecipitated with an anti-FLAG antibody, and immunoblotted with anti-FLAG or anti-HA antibodies. Bands were quantified and expressed as a percentage of the lamin A/C control for G␤1␥2 in B or as the ratio of HA-G␥2/FLAG-G␤ for all five G␤s in C. PhLP1 knockdown was measured by quantifying the PhLP1 band intensity in immunoblots of 10 g of whole cell lysate. The average PhLP1 knockdown was between 60 and 76% compared with the lamin A/C control. Bars represent the average Ϯ S.E. from 3-5 separate experiments. A representative blot for G␤1␥2 is shown below the graph in B. The ncd indicates no complex detected under these conditions. This same abbreviation is also used in Figs. 2-4. these conditions (Fig. 1C). To more directly compare the effects of PhLP1 knockdown, the G␥2/G␤1-4 band intensity ratios in the immunoprecipitates were determined for the three siRNA conditions (Fig. 1C). In each case, much less G␥2 was associated with G␤ when PhLP1 was knocked down. The G␥2/G␤ ratio decreased between 65 and 84% compared with the lamin A/C control. These results indicate that PhLP1 does assist in the formation of G␤␥ complexes containing G␤s 1-4 with G␥2.
To further examine the role of PhLP1 in G␤␥ assembly with the different G␤ subunits, an alternative method to block PhLP1 function was employed. It has been shown previously that an N-terminally truncated PhLP1 variant in which the first 75 amino acids have been removed (PhLP1 ⌬1-75) acts in a dominant interfering manner to block G␤␥ assembly by forming a stable PhLP1 ⌬1-75-G␤-CCT ternary complex that does not release G␤ from CCT for association with G␥ (13,14). Coexpression of PhLP1 ⌬1-75 with FLAG-G␤1 and HA-G␥2 resulted in a dramatic reduction in the amount of G␥2 in the G␤1 immunoprecipitate compared with wild-type PhLP1 (expressed at comparable levels) or to an empty vector control ( Fig. 2A). This pattern was similar among G␤s 1-4. PhLP1 ⌬1-75 decreased the G␥2/G␤ ratios by 75-92% in the G␤1-4 immunoprecipitates (Fig. 2B). For G␤5, again very little G␥2 was associated with it under these conditions. Interestingly, coexpression of wild-type PhLP1 increased the amount of both G␤ and G␥2 in the FLAG-G␤ immunoprecipitate by 30 -50% for all five G␤ isoforms (see Figs. 2A, 4A, and 6B). This observation is consistent with a PhLP1-mediated enhancement of G␤␥ formation, resulting in a stabilization of G␤ and G␥ expression. Together, these findings confirm the siRNA knock-down results by showing that PhLP1 is important in the assembly of each of the G␤s 1-4 with G␥2.
A second question regarding the scope of PhLP1-mediated G␤␥ assembly is whether all 12 G␥ subunits or just a subset require PhLP1 to associate with G␤. To address this question, the effects of siRNA-mediated PhLP1 knockdown and PhLP1 ⌬1-75 overexpression on the association of the twelve G␥ subunits with G␤2 were measured. G␤2 was chosen because it forms dimers with most G␥ isoforms, yet it shows selectivity between the different G␥s (25). The siRNA knockdown experiments followed the same format as those in Fig. 1. HEK 293T cells were treated with PhLP1 siRNA, a control siRNA to lamin A/C or no siRNA and then co-expressed with FLAG-G␤2 and each of the 12 HA-tagged G␥ subunits. Cell lysates were immunoprecipitated with an anti-FLAG antibody, and the precipitate was immunoblotted with anti-HA and anti-FLAG antibodies to detect the amount of each G␥ subunit bound to G␤2. Fig. 3A shows the levels of PhLP1 in the cell extract and FLAG-G␤2 and HA-G␥2 in the immunoprecipitate relative to the lamin A/C siRNA control. The results were similar to the G␤1␥2 experiment. The PhLP1 knockdown was 80%, which resulted in a 30% decrease in G␤2 and a 90% decrease in G␥2 compared with the lamin A/C control. This pattern was consistent among all the G␤2G␥ combinations that formed dimers. G␤2 decreased by 20 -50% while the co-immunoprecipitating G␥s decreased by 80 -95% (data not shown). Fig. 3B compares the G␥/G␤2 band intensity ratios for the three siRNA conditions. In each case, much less G␥ was associated with G␤ when PhLP1 was knocked down. The G␥/G␤2 ratios decreased between 74 and 91% compared with the lamin A/C control, except for G␥s 1, 11, and 13, which did not form dimers with G␤2. These results indicate that all G␤2G␥ dimers depend upon PhLP1 for their assembly.
The dominant interference experiments with PhLP1 ⌬1-75 followed a similar pattern. Co-expression of PhLP1 ⌬1-75 with FLAG-G␤2 and HA-G␥2 resulted in a 50% reduction in the amount of G␤2 and a 95% reduction in the amount of G␥2 in the FLAG immunoprecipitate when compared with the wildtype PhLP1 control (Fig. 4A). Moreover, co-expression of wildtype PhLP1 increased both G␤2 and G␥2 levels by 50%, similarly to G␤1␥2 (Fig. 2A). The effect of PhLP1 ⌬1-75 on the G␥/G␤2 ratio was the same for all the G␥s that formed dimers with G␤2. The ratios were drastically reduced by amounts rang-  ing from 81-100%. Together with the PhLP1 knockdown data, these results clearly demonstrate that all G␥ subunits that interact with G␤2 require PhLP1 for dimer formation.
Another interesting observation that can be made from the data in Figs. 3B and 4B concerns the effect of PhLP1 on the specificity of G␤2␥ dimer formation. The G␥ subunits can be divided genetically into five subfamilies as shown in the phylogenetic tree of Fig. 5A. Members of subfamily II form dimers with G␤2 readily, whereas members of the other subfamilies interact weakly with G␤2 or not at all (Fig. 5, B and C). The order of dimer formation of the G␥ subfamilies with G␤2 is II Ͼ III Ͼ I/IV with no dimer formation found with subgroup V. This pattern of G␤2␥ dimer specificity is similar to in vitro data reported previously (25). Importantly, PhLP1 does not appear to influence the specificity of G␤2␥ dimer formation. The specificity pattern is the same no matter the level of PhLP1 activity. For example, when PhLP1 is siRNA-depleted, G␤2␥ formation with G␥ subfamily II is greater than with subfamilies III, I, and IV by a similar factor as when PhLP1 is at endogenous levels. Similarly, when PhLP1 function is blocked by the PhLP1 ⌬1-75 variant, G␤2␥ formation with subgroup II is greater than the other subgroups by a similar factor as when PhLP1 is overexpressed (Fig. 5, B and C). Thus, it appears that PhLP has no effect on which G␥ subunit will interact with G␤2.
A third important question regarding the scope of PhLP1mediated dimer assembly is whether PhLP1 assists in the formation of G␤5-R7 RGS protein complexes. G␤5 binds both CCT (15) and PhLP1 (Fig. 1) weakly compared with the other G␤ subunits, suggesting that CCT and PhLP1 may not be required for G␤5-R7 RGS dimer assembly. To begin to address this issue, the effects of PhLP1 knockdown and PhLP1 ⌬1-75 overexpression on the interaction of G␤5 with RGS7 were assessed by co-immunoprecipitation as in Fig. 1. PhLP1 knockdown decreased both G␤5 expression and RGS7 co-immunoprecipitation with G␤5 by 40% (Fig. 6A). This result is in contrast to the G␤␥ co-immunoprecipitation data, which showed a similar 40% decrease in G␤1 and G␤2 expression but exhibited a much greater decrease (80 -90%) in the amount of G␥ coimmunoprecipitating with G␤ upon PhLP1 knockdown (Figs. 1B and 3A). The results were similar in the dominant interference experiments (Fig. 6B). Overexpression of wild-type PhLP1 increased G␤5 expression by ϳ2-fold over the empty vector control, as was observed with G␤1 and G␤2 ( Figs. 2A and 4A). However, the proportional increase in co-immunoprecipitation seen with G␥2 ( Figs. 2A and 4A) was not observed with RGS7, which showed a much smaller increase. Moreover, overexpression of PhLP1 ⌬1-75 did not cause the dramatic decrease in RGS7 co-immunoprecipitation that was observed with G␥2 (Fig. 6B). These findings suggest that the effect of PhLP1 on the expression of G␤5 is similar to the other G␤s but that PhLP1 may not be as important in G␤5-RGS7 assembly as it is in G␤␥ assembly.
The findings of Fig. 6 point to potentially significant differences between the mechanisms of G␤5-RGS7 assembly and G␤␥ assembly. As a result, G␤5-RGS7 assembly was further investigated to better understand the role of PhLP1 and CCT in this process. If CCT were involved in G␤5 folding, the two would have to interact, yet G␤5 has been reported to bind CCT poorly in vitro (15). To further test the ability of G␤5 to interact with CCT, the co-immunoprecipitation of overexpressed G␤5 with endogenous CCT in HEK-293T cells was measured. G␤5 was readily detected in the CCT immunoprecipitate, but the amount was 20-fold less than that of G␤1 (Fig. 7A), confirming the finding that G␤5 binds CCT much less than other G␤s. Importantly, co-expression of PhLP1 increased G␤5 binding to CCT by nearly 10-fold (Fig. 7B), indicating that PhLP1 stabilizes the interaction of G␤5 with CCT considerably. In contrast, co-expression of RGS7 had no effect on G␤5 binding to CCT. These results are very different from the effect of PhLP1 and G␥2 on G␤1 binding to CCT in which both PhLP1 and G␥2 contributed significantly to the release of G␤1 from the CCT complex (13). Thus, it appears that the role of PhLP1 in the binding of G␤1 and G␤5 to CCT are opposite, with PhLP assist-ing in the release of a tightly binding G␤1 while stabilizing the weak interaction of G␤5.
To complete the investigation of CCT interacting partners in the G␤5-RGS7 dimer, the binding of RGS7 was also measured by co-immunoprecipitation with CCT. No RGS7 bound CCT when RGS7 was overexpressed alone, but co-expression of G␤5 caused a detectible amount of RGS7 to co-immunoprecipitate with CCT (Fig. 7C). In contrast, co-expression of PhLP1 with RGS7 did not cause RGS7 to bind CCT and co-expression of PhLP1 together with G␤5 and RGS7 did not increase RGS7 co-immunoprecipitation with CCT. The total amount of RGS7 in the cell lysate also increased significantly upon G␤5 co-expression, consistent with the fact that R7 RGS proteins require G␤5 for stable expression in the cell (29). These results suggest that in the process of G␤5-RGS7 assembly G␤5 recruits RGS7 to CCT. The lack of effect of PhLP1 on RGS7 binding to CCT suggests that PhLP1 does not play a role in this recruitment. To further test this notion, RGS7 and PhLP1 were co-expressed with and without G␤5, and their ability to co-immunoprecipitate each other was measured. Neither protein was found in the immunoprecipitate of the other in the presence or absence of G␤5 (Fig. 7D), indicating that RGS7 and PhLP1 do not exist in any complexes together. From these binding experiments, it appears that PhLP1 stabilizes the interaction of G␤5 with CCT  Effects of PhLP1 and RGS7 on the binding of G␤5 to CCT. A, binding of G␤5 to CCT was compared with that of G␤1 by co-immunoprecipitation. HEK 293T cells were transfected with cDNAs for FLAG-G␤1, FLAG-G␤5 or an empty vector control as indicated. After 48 h, cells were lysed, immunoprecipitated with an anti-CCT⑀ antibody, and immunoblotted with anti-FLAG antibodies. Bands were quantified and the binding of G␤5 to CCT was expressed relative to that of G␤1. Bars represent the average Ϯ S.E. from three separate experiments, and representative blots are shown below the graphs. (The G␤5 error bar is very small.) For all experiments A-D, the expression of each transfected cDNA was confirmed by immunoblotting 5 g of whole cell lysate with the antibodies indicated. B, the effect of PhLP1 and RGS7 on the binding of G␤5 to CCT was measured by co-immunoprecipitation as in panel A. Cells were transfected with the indicated cDNAs, and CCT was immunoprecipitated and immunoblotted for FLAG-G␤5. Bands were quantified and expressed relative to the FLAG-G␤5/HA-RGS7 sample. Data are from eight separate experiments. C, the effects of PhLP1 and G␤5 on RGS7 binding to CCT was measured by co-immunoprecipitation as in panel A. Cells were transfected with the indicated cDNAs, and CCT was immunoprecipitated and immunoblotted for HA-RGS7. Bands were quantified and expressed relative to the FLAG-G␤5/HA-RGS7 sample. Data are from three separate experiments. D, the ability of PhLP1 and RGS7 to co-exist in CCT or other complexes was tested by co-immunoprecipitation. Cells were transfected with cDNAs to PhLP1-Myc and HA-RGS7 with or without FLAG-G␤5, immunoprecipitated with anti-Myc or anti-HA antibodies and immunoblotted with these same antibodies as indicated. The resulting blots are shown. and that G␤5 recruits RGS7 to CCT, but only after PhLP1 has been released from the complex.
The data from Figs. 6 and 7 suggest that PhLP1 may be involved in the folding of G␤5 by stabilizing its interaction with CCT but that PhLP1 may not participate in G␤5-RGS7 assembly. This concept was further tested by measuring the effect of PhLP1 knockdown or overexpression on the rate of G␤5-RGS7 dimerization. In these experiments, PhLP1 was either siRNA-depleted or overexpressed in HEK-293T cells, and the rate of G␤5-RGS7 dimer formation was measured in a pulse-chase experimental format (14). PhLP1 knockdown resulted in a 2-fold decrease in the rate of G␤5-RGS7 dimerization compared with a control siRNA (Fig. 8A), which is somewhat less than the 5-fold decrease in the rate of G␤1␥2 dimerization observed with a similar PhLP1 knockdown (14). In contrast, the effects of PhLP1 overexpression on the rate of G␤5-RGS7 assembly were strikingly different than what was observed for G␤1␥2 assembly. PhLP1 overexpression actually caused a small decrease in the rate of G␤5-RGS7 assembly (Fig. 8B), whereas it resulted in a 4-fold increase in the rate of G␤1␥2 assembly (14). Interestingly, PhLP overexpression increased the amount of G␤5 produced during the 10-min pulse by 40%, which in turn caused a small increase in RGS7 co-immunoprecipitation. However, the net effect was a decrease in the RGS7/G␤5 ratio, indicating an inhibition of RGS7/G␤5 dimer formation despite the fact that more G␤5 was available for assembly. It is clear from these results that the role of PhLP1 in G␤5-RGS7 assembly is much different than its role in G␤␥ assembly. It appears that endogenous levels of PhLP1 may contribute to G␤5-RGS7 assembly by stabilizing the interaction of G␤5 with CCT, but that excess PhLP1 inhibits G␤5-RGS7 assembly, possibly by interfering with the G␤5-RGS7 interaction.
The recently published structure of the G␤5-RGS9 complex (39) suggests a possible reason for the observed inhibition of G␤5-RGS7 assembly by excess PhLP1. In the structure, the G␥-like domain interacts along the expected G␥ binding surface of G␤5, opposite the predicted PhLP1 binding site (39). However, the N-terminal lobe of RGS9 interacts with G␤5 on the same surface as PhLP1 (39,40). This overlap may preclude the formation of a PhLP1-G␤5-RGS7 complex analogous to the PhLP1-G␤␥ complex that is believed to be an intermediate in G␤␥ assembly (13,14). To test this possibility, the binding of PhLP1 to the G␤5-RGS9-1 complex was measured. An in vitro assay was preformed in which G␤5-RGS9-1 was immobilized on FLAG antibody-linked agarose beads via a FLAG tag on the RGS9-1. Increasing concentrations of metabolically labeled 35 S-PhLP1 were added to the beads and allowed to reach equilibrium. The beads were pelleted, and the amount of bound and free 35 S-PhLP1 was determined. The results show that indeed there was no measurable binding of PhLP1 to G␤5-RGS9-1 (Fig.  8C). In contrast, PhLP1 readily bound G␤1␥2 and to a lesser extent G␤5␥2 in this assay. The dissociation constants for the interactions were 83 Ϯ 13 nM for G␤1␥2 and 440 Ϯ 72 nM for G␤5␥2. The K d for G␤1␥2 binding is similar to the 107 nM K d reported previously for the PhLP1-G␤1␥1 interaction using surface plasmon resonance methods (38), so the assay appears to be measuring the binding accurately. The inability of G␤5-RGS9-1 to bind PhLP1 suggests that excess PhLP1 interferes with G␤5-RGS7 dimer formation, because it binds G␤5 in a manner that does not allow RGS7 to simultaneously interact.
The binding of G␤5 to CCT and the G␤5-dependent recruitment of RGS7 to CCT suggest an important role for CCT in the G␤5-RGS7 assembly process. This possibility was tested further by measuring the effect of CCT knockdown on the rate of G␤5-RGS7 dimerization using the pulsechase assay. An siRNA to CCT that results in a substantial knockdown of CCT complexes has been reported (32,41). Using this siRNA, CCT expression was decreased by 50% in HEK-293T cells (Fig. 9A). In addition, expression of the CCT⑀ subunit was also decreased by a similar amount (Fig.  9A), indicating that expression of the entire CCT complex was reduced by 50%. This reduction in CCT resulted in a proportional decrease in the rate of G␤5-RGS7 assembly of 50% (Fig. 9A), suggesting that G␤5-RGS7 assembly is very dependent on CCT. For comparison, the effect of this CCT knockdown on G␤␥ assembly, which is expected to be CCTdependent (13,15), was also measured. The 50% reduction in CCT caused a similar 50% decrease in the rate of G␤␥ assembly (Fig. 9B), confirming the importance of CCT in G␤␥ formation. The striking similarity of these effects of CCT knockdown on the rates of both G␤5-RGS7 and G␤␥ dimerization show that G␤5-RGS7 assembly is just as dependent on CCT as G␤␥ assembly. Together, the data in Figs. 8 and 9 indicate a similar role for CCT in both G␤5-RGS7 and G␤␥ assembly, but a much less critical role for PhLP1 in G␤5-RGS7 assembly compared with its essential role in G␤␥ assembly.

DISCUSSION
Post-translational assembly of stable G protein heterotrimers is a fundamental prerequisite for G protein signaling, yet the mechanism by which the assembly process occurs had been an enigma for more than two decades since the G protein heterotrimer was initially discovered. The most puzzling issue has been how the G␤ and G␥ subunits could come together to form a stable dimer when the individual polypeptides were structurally unstable. Recent studies have shed considerable light on the assembly process and have outlined a mechanism by which CCT and PhLP1 work as co-chaperones to fold G␤ and present it to G␥ for dimerization to occur (10 -15, 42, 43). G␥ itself appears to be held by another chaperone DRiP78 (30) until it can interact with PhLP1-G␤. Mechanistic studies have thus far focused on the most common G␤1␥2 dimer combination and have not addressed whether this assembly mechanism was general to the many other G␤␥ or G␤5-RGS protein dimers, or specific to only a subset. All the G␤ subunits have been recently shown to interact with CCT, but the interaction of G␤5 with CCT was much weaker that G␤1-4 (15). The current study has addressed the scope of PhLP1-mediated dimer assembly for many G␤␥ combinations. The results clearly show that PhLP1 is a general cochaperone for G␤␥ assembly. All G␤ subunits required PhLP1 for association with G␥2 ( Figs. 1 and 2), and all G␥ subunits that Chaperone-mediated G␤5-RGS7 and G␤␥ Assembly form dimers with G␤2 required PhLP1 for association with G␤2 (Figs. 3 and 4). It seems very likely that the other possible G␤␥ dimer combinations would also require PhLP1 for their assembly as well. Thus, it appears that all G␤␥s follow a similar mechanism of dimer formation.
Understanding the reasons why some G␤␥ combinations form dimers and other do not has been of interest in the field for some time (18). Apparent differences in G␤␥ specificity between in vitro assays and cell-based assays have suggested that cellular factors that are involved in the assembly process FIGURE 8. Effects of PhLP1 on the rate of G␤5-RGS7 dimer formation. A, the rate of G␤5-RGS7 dimer assembly was measured in HEK-293T cells with or without PhLP1 knockdown. Cells were treated with PhLP1 or lamin A/C siRNAs as indicated. Twenty-four hours later, the cells were transfected with FLAG-G␤5 and HA-RGS7 cDNAs. After 72 additional h, nascent polypeptides were labeled for 10 min with [ 35 S]methionine and then chased with unlabeled methionine and cycloheximide. At the chase times indicated, the FLAG-G␤5 was immunoprecipitated and the proteins were separated by SDS-PAGE. The radioactive bands were visualized and quantified using a PhosphorImager, and the molar ratio of G␤5 to RGS7 was calculated. The data points represent the average Ϯ S.E. from three separate experiments, and lines represent fits of the data to a first order rate equation. A representative gel is shown below the graph as is a PhLP1 immunoblot of 10 g of whole cell lysate showing the degree of siRNA knockdown. B, HEK-293 cells were transfected with FLAG-G␤5 and HA-RGS7 with and without PhLP1-Myc cDNAs for 48 h, and the rate of G␤5-RGS7 assembly was measured using the pulse-chase assay as in panel A. The data are from three separate experiments. C, the binding of the indicated concentrations of 35 S-PhLP1 to 0.5 M purified G␤1␥2 (E), G␤5␥2 (E), or G␤5-RGS9-1 (OE) was measured by in vitro co-immunoprecipitation (see "Experimental Procedures"). Symbols represent the average Ϯ S.E. from three separate experiments. Lines represent non-linear least squares fits of the data to a one-to-one binding equation. The fits yielded K d values of 83 Ϯ 13 nM for G␤1␥2, 440 Ϯ 70 nM for G␤5␥2, and no measurable value for G␤5-RGS9-1.
such as PhLP1 might influence G␤␥ specificity (25). However, this does not appear to be the case. As noted above, the specificity of G␤␥ dimer formation was not changed by increases or decreases in PhLP1 activity. Thus, it appears that PhLP1 is acting as a true catalyst in G␤␥ assembly by not influencing which G␤s can bind which G␥s but by simply facilitating the association of G␤␥ combinations that are intrinsically stable. In the case of the G␤2␥ combinations investigated here, dimer stability appears to be determined by sequence specificity, because G␥ binding segregated along subfamily lines according to sequence homology (Fig.  5). Hence, the major factors that determine G␤2␥ specificity appear to be limited to complementarity of the binding surfaces as determined by specific amino acid interactions, the expression of the complementary G␤␥ combinations in the same cell types, and the subcellular localization within the cell (18).
It is interesting to note that inhibition of PhLP1 activity through siRNA-mediated knockdown or overexpression of the PhLP1 ⌬1-75 dominant negative variant resulted in a surprisingly small decrease in G␤ expression (ϳ50%), despite the fact that very little of this residual G␤ was associated with G␥ (Figs. [1][2][3][4]. This finding indicates that G␤ can exist in the cell unassociated with G␥. It is likely that this pool of undimerized G␤ is associated with CCT, because it has been previously shown that G␤-CCT complexes are relatively stable in the absence of PhLP1 and G␥ (13). Thus, it appears that the role of CCT is to fold G␤ and protect it from aggregation or proteolytic degradation until it can be released by PhLP1 to interact with G␥.
In the case of G␤5-RGS7 dimers, the data suggest a very different assembly mechanism than that of G␤␥. An outline of a possible mechanism for G␤5-RGS7 assembly that is consistent with the data presented is depicted in Fig. 10. The decrease in the rate of G␤5-RGS7 assembly upon siRNA-mediated CCT knockdown (Fig. 9A) indicates that CCT is involved in the assembly process, most likely by folding the nascent G␤5 despite the weak interaction of G␤5 with CCT. Likewise, the decrease in the rate of G␤5-RGS7 assembly upon PhLP1 knockdown (Fig. 8A) shows that PhLP1 also contributes to the assembly process, possibly by increasing the efficiency of G␤5 folding by increasing the binding of G␤5 to CCT through the formation of a stable PhLP1-G␤5-CCT ternary complex (Fig. 7B). However, the decrease in the rate of G␤5-RGS7 assembly upon overexpression of PhLP1 (Fig. 8B) indicates that excess PhLP1 interferes with the assembly process. A logical explanation of this effect is that PhLP1 must be released from G␤5 prior to its interaction with RGS7 and that excess PhLP1 blocks the association of RGS7 with G␤5. Once PhLP1 is released, it appears FIGURE 9. Effects of CCT on the rate of G␤5-RGS7 dimer formation. A, HEK-293T cells were treated with CCT or lamin A/C siRNA for 96 h and the expression of CCT and CCT⑀ was measured by immunoblotting 20 g of whole cell lysate. Representative blots are shown. B, the rate of G␤5-RGS7 dimer assembly was measured in HEK-293T cells with or without CCT knockdown as in Fig. 8A. The data are from three separate experiments. C, the rate of G␤1␥2 dimer assembly was measured in HEK-293T cells with or without CCT knockdown as in Fig. 8A. The data are from three separate experiments.
that RGS7 can associate with G␤5 while still bound to CCT, given the fact that G␤5 initiates the co-immunoprecipitation of RGS7 with CCT (Fig. 7C). Once formed, the G␤5-RGS7 complex would be expected to readily release from CCT because of the relatively weak interaction of the complex with CCT (Fig. 7, B and C). The folded and assembled G␤5-RGS7 complex would then be able to interact with its R7 anchoring protein and with its G␣ targets.
The unique roles for PhLP1 in G␤␥ versus G␤5-RGS7 dimer formation can be understood by examining the structures of the complexes. In the case of G␤␥, the G␥ binding surface is on the opposite side of G␤ from the principal PhLP1 binding surface (40), allowing PhLP1 and G␥ to interact with G␤ simultaneously. It has been proposed that this configuration allows nascent G␥ to associate with G␤ while the G␤ ␤-propeller is being stabilized by PhLP1 (13). In the case of G␤5-RGS9, the N-terminal lobe of RGS9 covers a 2600-Å 2 area on the same face of G␤5 (39) predicted to bind PhLP1, based on the phosducin-G␤1␥1 structure (40). In fact, several residues of G␤5 that contact the N-terminal lobe of RGS9 are also expected to contact PhLP1 (39,40). Because of this overlap, assembly of the G␤5-RGS complex apparently cannot proceed through a PhLP1-G␤5-RGS intermediate.
A question that is not clear from the structures is how PhLP1 assists in the release of G␤1 from CCT, whereas it stabilizes the binding of G␤5 to CCT. More structural information on the PhLP1-G␤-CCT complexes for both the G␤1 and G␤5 complexes would be required to understand the underling molecu-lar basis for these disparate binding properties. Perhaps the differences lie more in the interactions of the G␤ subunits with CCT, with G␤1 making high affinity contacts and G␤5 making only low affinity contacts in the absence of PhLP1. Upon PhLP1 binding, it is possible that both G␤1 and G␤5 form a similar complex with CCT in which the high affinity contacts of G␤1 have been lost but indirect contacts with CCT through PhLP1 have been gained, thereby increasing the binding of G␤5 to CCT.
In conclusion, this work expands the role of PhLP1 as an essential cochaperone in the assembly of all G␤␥ combinations and outlines a mechanism for G␤5-RGS7 dimer formation. This mechanism is similar to G␤␥ assembly in its CCT dependence but differs significantly in its PhLP1 dependence. The data provide additional insight into the intricate means by which the cell utilizes its molecular chaperones to bring the unstable ␤-propeller fold of G␤ subunits together with their complementary G␥-like domains to create stable G␤␥ and G␤5-RGS7 dimers to perform their vital functions in G protein signaling.