Structural Basis for the Selectivity of the RGS Protein, GAIP, for Gαi Family Members

GAIP is a regulator of G protein signaling (RGS) that accelerates the rate of GTP hydrolysis by some G protein α subunits. In the present studies, we have examined the structural basis for the ability of GAIP to discriminate among members of the Gαi family. Gαi1, Gαi3, and Gαo interacted strongly with GAIP, whereas Gαi2 interacted weakly and Gαs did not interact at all. A chimeric G protein composed of a Gαi2N terminus and a Gαi1 C terminus interacted as strongly with GAIP as native Gαi1, whereas a chimeric N-terminal Gαi1 with a Gαi2 C terminus did not interact. These results suggest that the determinants responsible for GAIP selectivity between these two Gαis reside within the C-terminal GTPase domain of the G protein. To further localize residues contributing to G protein-GAIP selectivity, a panel of 15 site-directed Gαi1 and Gαi2 mutants were assayed. Of the Gαi1 mutants tested, only that containing a mutation at aspartate 229 located at the N terminus of Switch 3 did not interact with GAIP. Furthermore, the only Gαi2 variant that interacted strongly with GAIP contained a replacement of the corresponding Gαi2 Switch 3 residue (Ala230) with aspartate. To determine whether GAIP showed functional preferences for Gα subunits that correlate with the binding data, the ability of GAIP to enhance the GTPase activity of purified α subunits was tested. GAIP catalyzed a 3–5-fold increase in the rate of GTP hydrolysis by Gαi1 and Gαi2(A230D) but no increase in the rate of Gαi2 and less than a 2-fold increase in the rate of Gαi1(D229A) under the same conditions. Thus, GAIP was able to discriminate between Gαi1 and Gαi2 in both binding and functional assays, and in both cases residue 229/230 played a critical role in selective recognition.

Heterotrimeric G proteins associate with the cytoplasmic surfaces of 7-transmembrane spanning receptors and function to transduce signals from receptors activated by extracellular ligands to intracellular effectors (1). One of the most recent developments in the study of G protein regulation is the identification of a novel family of proteins known as regulators of G protein signaling or RGS proteins (2). RGS proteins are char-acterized by the presence of an RGS domain that is structurally conserved across evolution (3,4). These molecules function to desensitize G protein-coupled responses in organisms from yeast to man by directly interacting with the ␣ subunit of heterotrimeric G proteins and increasing their rate of GTP hydrolysis (5). Direct interaction between G protein ␣ subunits and RGS molecules was first demonstrated by DeVries et al. (6), who isolated the cDNA for the RGS GAIP (G alpha interacting protein) using a yeast two-hybrid screen for G␣ i3 -interacting proteins. A number of studies quickly followed revealing GAP (GTPase-activating protein) 1 activity to be the mechanism by which RGSs turned off G protein activation (7)(8)(9)(10). Both the structural interaction between RGS and G␣ subunits and the mechanism of RGS GAP activity were further elucidated by the co-crystallization of RGS4 with G␣ i1 (11). However, much remains to be revealed about the function of individual members of the RGS family, their specificities for interacting proteins, and the structural determinants that define these interactions.
Most of the initially described RGS proteins showed both binding and functional selectivity for the G␣ i family of G proteins (7)(8)(9)12). More recently, a number of RGS molecules have demonstrated binding or functional interactions with G␣ q and/or G␣ s signaling pathways (13)(14)(15)(16)(17), and p115RhoGEF was shown to be a functional RGS for the G␣ 12 /G␣ 13 family of G proteins (18 -20). However, there has been little information about the ability of any RGS to discriminate among the closely related members of the G␣ i family. Evidence for some specificity of RGS binding to distinct G␣ i family members was demonstrated by DeVries et al. (6), who showed strong interaction of GAIP with G␣ i1 , G␣ i3 , and G␣ o but weak interaction with G␣ i2 and no interaction with G␣ s . The differential binding characteristics of G␣ i1 and G␣ i2 are particularly intriguing because these two G proteins are highly homologous, having an amino acid sequence identity of 88%. Differences in RGS binding may reveal structural differences in these two G proteins that have implications for their ability to differentially activate divergent downstream signaling pathways.
To evaluate the structural basis for the selectivity of the RGS GAIP for individual members of the G␣ i family, we have expressed native, chimeric, and mutant G␣ proteins and compared their abilities to bind GAIP and act as substrates for GAIP GAP activity. The results show a preference of GAIP for G␣ i1 over G␣ i2 in both binding assays and GAP assays. This preference was reversed by mutating residue Asp 229 in G␣ i2 to alanine and making the reciprocal mutation (A230D) in G␣ i1 . Interestingly, the selectivity of GAIP for G␣ i1 over G␣ i2 was lost when GTPasedeficient mutants of these two G␣ i s were tested for GAIP binding. Thus, the structural preference of GAIP for G␣ i1 versus G␣ i2 in their ground (presumably GDP-bound) states has functional consequences in their respective GAP activities.

EXPERIMENTAL PROCEDURES
Generation of Yeast Two-hybrid Fusion Constructs-Rat G protein ␣ subunits were PCR amplified with oligonucleotides containing 5Ј EcoRI restriction sites and 3Ј SalI restriction sites. PCR products were then subcloned into the pCRII vector (Invitrogen, Carlsbad, CA) and sequenced to ensure fidelity to the template. Inserts were excised with EcoRI and SalI and subcloned into the pGBT9 Gal 4 DNA-binding domain fusion vector (CLONTECH).
Human GAIP was PCR amplified from a human heart cDNA library using oligonucleotides containing a 5Ј NarI restriction site and a 3Ј SalI restriction site. PCR products were subcloned and sequenced as above, then removed from pCRII with Nar I and SalI, and subcloned into the pGAD Gal 4 activation domain fusion vector (CLONTECH).
Generation of G Protein ␣ Subunit Chimeras-The G␣ s/i3 chimera was generated by removing the N-terminal BamHI site in the G␣ s cDNA via site-directed mutagenesis (see below) and then ligating the BamHI-digested N-terminal 700-bp fragment of G␣ s to the 430-bp Cterminal fragment of BamHI-digested G␣ i3 cDNA. The G␣ i3/s chimera was generated by ligating the N-terminal 630-bp G␣ i3 fragment to the C-terminal 516-bp G␣ s fragment of the same digestions. Both chimeras were subcloned into the pGBT9 vector and characterized with BamHI and EcoRI as well as with BamHI and SalI digestions to ensure correct constructions.
G␣ i1/i2 and G␣ i2/i1 chimeras were made by engineering a BamHI site into the G␣ i1 cDNA at the same site as a naturally occurring BamHI in G␣ i2 . G␣ i2 and mutant G␣ i1 cDNAs were digested with BamHI, and the N-terminal 635-bp fragment of G␣ i1 was ligated to the C-terminal 433-bp fragment of G␣ i2 to generate G␣ i1/i2 . Similarly, G␣ i2/i1 consists of the N-terminal BamHI fragment of G␣ i2 ligated to the C-terminal BamHI fragment of G␣ i1 .
Site-directed Mutagenesis of G Protein ␣ Subunits-Site-directed mutants of G␣ i1 and G␣ i2 were made using Stratagene QuickChange site-directed mutagenesis kit according to the manufacturer's protocols. Template pGBT9-G␣ i1 or pGBT9-G␣ i2 was amplified for 14 cycles of 12-min extensions, each using overlapping forward and reverse primers encoding the applicable mutation. All mutants were sequenced throughout the entire coding region to ensure desired mutagenesis as well as to screen against unwanted PCR-induced mutations.
Transformation of Competent Yeast-Saccharomyces cerevisiae of strain HF7c␣ were co-transformed with pGBT9 (containing Trp marker) and pGAD (containing Leu marker) vector constructions by standard lithium acetate procedures (CLONTECH Matchmaker twohybrid system). Briefly, single yeast colonies were grown overnight at 30°C with continuous shaking to an A 600 of 0.6. Cells were harvested by centrifugation for 10 min at 3000 rpm, washed once in sterile H 2 0, and resuspended in 2 ml of cold 100 mM lithium acetate. After shaking at 30°C for 1 h, 100 l of competent cells was added to 1-2 g of transforming DNA in the presence of 5 g of carrier salmon sperm DNA and 0.7 ml of 40% polyethylene glycol. Cells were heat shocked at 42°C for 15 min, then collected with a quick spin, and plated on -Leu-Trp selective dropout agar medium to grow for 3 days at 30°C. Four colonies of each construct were streaked on -Leu-Trp agar to propagate for assay.
Immunoblotting-Yeast transformants were grown overnight to high density in 4-ml cultures, harvested, and resuspended in binding buffer (0.2 M Tris, pH 8.0, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 10 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EDTA, 20 g/ml pepstatin A). Cells were lysed by vortexing three times for 1 min in the presence of glass beads at 4°C and spun for 10 min at 12,000 ϫ g to remove cell debris. 50 g of lysate was loaded per lane onto SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose, and probed with rabbit antibody common to G protein ␣ subunits (Calbiochem, La Jolla, CA) at 1:500 dilution in Tris-buffered saline/5% milk. Immunoreactivity was detected with horseradish peroxidase-conjugated donkey anti-rabbit secondary antibody (1:2000 dilution) and developed using ECL reagents according to the manufacturer's protocols (Amersham Pharmacia Biotech).
Liquid ␤-Galactosidase Assays-Single colonies of transformed cells were innoculated into 5 ml of SC-Leu-Trp agar and grown overnight to an A 600 of 0.8. Cells were collected by centrifugation, washed once in Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 P0 4 , 10 mM KCl, 1 mM MgSO 4 ), resuspended in 300 l of the same, and lysed by four freeze/thaw cycles. To start the assay, 100 l of this cell lysate was suspended in 0.7 ml of Z buffer containing 0.27% ␤-mercaptoethanol and then added to 0.16 ml of Z buffer containing 4 mg/ml o-nitrophenyl ␤-D-galactopyranoside substrate. Suspensions were vortexed and incubated for 2 h at 30°C. Color reactions were stopped with 0.4 ml of Na 2 CO 3 and read at A 420 after spinning out cell debris. ␤-Galactosidase units (21) were calculated according to the manufacturer's protocols (CLONTECH), as follows: ␤-galactosidase units ϭ 1000 ϫ A 420 /(t ϫ v ϫ A 600 ), where t is 120 min of incubation, v is 0.1 ml of reaction volume⅐concentration factor, and A 600 was 0.8 for the culture.
Histidine Growth Assays-5-ml cultures of yeast transformants were grown to an A 600 of 1.0 and then 3 l of 1:10 serial dilutions of confluent growths were spotted on either SC-Leu-Trp or SC-Leu-Trp-His agar plates and allowed to grow at 30°C for 3 days.
Protein Expression and Purification-Full-length G protein ␣ subunits G␣ i1 , G␣ i2 , G␣ i1 (D229A), and G␣ i2 (A230D) and full-length GAIP were expressed as GST fusion proteins by subcloning cDNAs downstream of the GST tag using EcoRI/SalI sites of the vector pGEX-6P-1 (Amersham Pharmacia Biotech). Each plasmid construct was transformed into bacterial strain BL21, grown overnight, and induced to express protein with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside. Cells were harvested by centrifugation, sonicated in TE containing 0.1

FIG. 1.
A, liquid ␤-galactosidase assays of native G protein ␣ subunit interactions with GAIP. Yeast clones co-expressing indicated Gal 4-binding domain-G␣ fusions with activation domain-GAIP fusions were assayed for the interaction-dependent activation of a lacZ reporter. The amount of ␤-galactosidase released was measured colorimetrically using the substrate ONPG. Two clones of each transformant were assayed and normalized to the interaction of GAIP with G␣ i1 defined as 100%. The results shown are the mean Ϯ S.E. for n ϭ 4 -18 in triplicate. B, histidine-minus growth assays of native versus mutant G protein ␣ subunit interactions with GAIP. Yeast clones co-expressing indicated binding domain-G␣ fusions with activation domain-GAIP were assayed for their interaction-dependent activation of a histidine reporter. Each clone was grown and plated as detailed under "Experimental Procedures." Plates on the left show limiting dilutions of clones grown on tryptophan-and leucine-lacking agar medium to control for noninteraction dependent growth. Plates on the right show identical dilutions of clones grown on tryptophan-minus, leucine-minus, and histidine-lacking medium to assay for interaction-dependent histidine reporter expression. This assay has been performed twice with identical results. mM phenylmethylsulfonyl fluoride and 1 mM ␤-mercaptoethanol, and solubilized with 1% Triton X-100. Lysates were cleared by centrifugation at 12,000 ϫ g for 10 min, and supernatants were applied to pre-washed glutathione-Sepharose columns (Amersham Pharmacia Biotech). Columns were washed with TE containing phenylmethylsulfonyl fluoride and ␤-mercaptoethanol and GST fusion proteins eluted with 10 mM glutathione. Purified proteins were buffer exchanged into TED buffer (20 mM Tris-HCl, pH 8, 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol), concentrated to 1 mg/ml in Nanosep spin columns (Pall Filtron Corp.), and stored at Ϫ80°C. Size and homogeneity of purified proteins were verified via Coomassie-stained SDS-polyacrylamide gel electrophoresis, and in-frame translation of G proteins was verified via immunoblot using a G␣ i1 /G␣ i2 -selective antibody (kind gift of Dr. David Manning, University of Pennsylvania, Philadelphia, PA).
GTP␥S Competition Curves-100 nM purified GST-tagged G protein ␣ subunits were shaken for 4 h at 30°C in the presence of 100 nM GTPase Assays-100 nM purified GST-tagged G protein ␣ subunits were loaded with 1 M [␥-32 P]GTP (8000 cpm/pmol) for 20 min at 30°C in 600 l of GTPase buffer (0.1% lubrol PX, 50 mM HEPES, pH 7.5, 1 mM dithiothreitol, 5 mM EDTA). Reactions were chilled at 4°C for 10 min, and assays were conducted at 6°C. A 50-l aliquot was removed immediately before initiating the reaction and quenched with 750 l of 5% Norit activated charcoal in 50 mM NaPO 4 , pH 3. To initiate the reaction, 100 M cold GTP and 15 mM MgSO 4 (final concentrations) in the presence versus absence of 500 nM GST-tagged GAIP were added to reaction mixtures, and 50-l aliquots were removed after 10 s, 20 s, 40 s, 1 min, 2 min, 3 min, 4 min, and 5 min and stopped as just described. Charcoal was precipitated by centrifugation for 15 min at 12,000 ϫ g, and 400-l free phosphate-containing supernatants were counted to determine the amount of P i released per reaction.

Interactions of G Protein
Fusions with GAIP-To explore the structural basis for the differences in GAIP binding by the different members of the G␣ i family, we engineered a panel of G␣ protein chimeras and mutants. As a first step, native and engineered G proteins were assayed for the ability to bind GAIP using the yeast two-hybrid system. To make use of this system, G␣ protein cDNA constructs were subcloned downstream of a Gal 4-binding domain cDNA and coexpressed with a GAIP-Gal 4 activation domain fusion in the S. cerevisiae strain HF7C␣. All fusions were immunoblotted to control for relative expression levels. An anti-G protein ␣ subunit antibody raised against the internal GTP-binding sequence common to all heterotrimeric G protein ␣ subunits recognized a protein of the appropriate molecular mass (about 65 kDa) for a G protein ␣ subunit fused to the Gal 4-binding domain in each of the clones transformed with a G protein fusion (data not shown). All of the clones expressed comparable levels of G protein fusion, and no protein of the same size was seen in clones transformed with pGBT9-binding domain alone.
Given such a similar background of G protein fusion expression, a measure of the strength of interaction between various G proteins and GAIP can be estimated from the relative activation of Gal 4-dependent reporters. The yeast strain HF7C␣ was stably transformed with cDNAs encoding both ␤-galactosidase and histidine reporters downstream of a Gal 4 promotor. In this system, the promotor is activated in proportion to the degree of interaction between the Gal 4-binding domain and activation domain fusions (23). Thus, two different reporters were used to measure the relative strength of the interaction between the G protein-binding domain fusion and the GAIP activation domain fusion.
According to both histidine and ␤-galactosidase reporter systems, robust interaction of GAIP was seen with G␣ i1 , G␣ i3 , and G␣ o , whereas the interaction with G␣ i2 was weak, and the interaction with G␣ s was undetectable (Fig. 1). These results are consistent with those obtained by DeVries et al. (6). Due to the quantitative nature of the assays, liquid ␤-galactosidase assays were used for interaction comparisons henceforth. Because G␣ i1 gave a strong interaction with GAIP in its native conformation, which was statistically indistinguishable from that of G␣ i3 and G␣ o , and because this G protein was tested in every assay conducted, this level of interaction was designated as 100% for comparison with all other G protein constructs. 100% interaction in these assays corresponds to 1.4 ␤-galactosidase units (21). The ␤-galactosidase activity generated by GAIP co-transfected with pGBT9 vector alone (0.12 ␤-galactosidase units) was considered background and was subtracted from all values for G protein-GAIP interactions before normalization.
Chimeras-Because G␣ i3 interacted strongly with GAIP, whereas G␣ s did not interact at all, chimeras of G␣ i3 with G␣ s were generated in an attempt to localize the regions of G␣ i required for GAIP binding. A BamHI site that cuts both cDNAs roughly two-thirds into the length of the coding region was used to generate both chimeras (Fig. 2). This BamHI site conveniently separates all of the N-terminal ␣-helical domain from most of the GTPase domain (a small part of which is encoded at the very N terminus of the cDNA). The binding characteristics of these chimeras could thus substantiate the relative importance of these two domains in GAIP binding. However, neither chimera bound to GAIP (Fig. 2A). These results potentially indicate that both domains of G␣ i3 contribute important determinants for RGS binding, but the divergence in ␣ s sequence from that of the ␣ i family presents a number of other possible interpretations.
To discriminate among these possibilities, chimeras composed of the initial two-thirds of ␣ i1 fused to the distal one-third of ␣ i2 and the reciprocal G␣ i2/i1 chimera were prepared using an engineered BamHI site (Fig. 2B). G␣ i2 is highly homologous to G␣ i1 , yet its interaction with GAIP is negligible compared with G␣ i1 . The G␣ i2/i1 chimera interacted with GAIP just as strongly as native G␣ i1 , whereas the reverse G␣ i1/i2 chimera, like wild type G␣ i2 , showed little binding to GAIP (Fig. 2B). These results suggest that the G␣ i1 C terminus is required for GAIP interaction. The results may also imply that the determinants contributing to GAIP binding are entirely contained within the GTPase domain of the G protein, but there may be additional determinants that are conserved between the N termini of G␣ i1 and G␣ i2 that remain to be identified.
Site-directed Mutants-As a next step, site-directed mu-tagenesis of G␣ i1 and G␣ i2 was used to further localize determinants contributing to the selectivity of GAIP interaction. Because G␣ i1 and G␣ i2 are 88% identical at the amino acid level but show vastly different GAIP binding capacities in the yeast two-hybrid system, the primary sequences of the two proteins were compared with identify candidate residues that might contribute to differential GAIP binding. Of the amino acids that differed between G␣ i1 and G␣ i2 , reciprocal mutants were generated at eight different positions in the primary sequence based on the likelihood that a given position would affect RGS binding given its location in the three-dimensional crystal structure of G␣ i1 bound to RGS4 (11). The effects of C-terminal mutants were of particular interest due to the results of the chimeras, but a number of N-terminal mutants were also studied because they appeared to be close to potential RGS contact sites in the crystal structure (11). Of the five G␣ i1 mutants C-terminal to the BamHI site that were tested, several impaired binding to GAIP, but only D229A abolished it (Fig. 3A). Even more significantly, the reciprocal mutation in the corresponding residue in G␣ i2 (G␣ i2 (A230D)) 2 produced a variant G␣ i2 that bound to GAIP as strongly as G␣ i1 (Fig. 3A). Thus, G␣ i1 (D229) appears to be particularly important for GAIP interaction.
In addition to the C-terminal mutants shown in Fig. 3A, three N-terminal G␣ i1 mutants and the corresponding reciprocal G␣ i2 mutants were also assayed for ␤-galactosidase activity. Consistent with the results of the G␣ i1/i2 chimeras, all of the N-terminal G␣ i1 mutants bound to GAIP, and none of the corresponding G␣ i2 mutants bound GAIP as strongly as G␣ i1 (Fig. 3B). Thus, none of these residues appears to be a necessary determinant for GAIP binding.
GTPase-deficient Mutants-To determine whether different nucleotide-dependent conformations of these G proteins affected their relative GAIP affinities, GTPase-deficient mutants of G␣ i1 and G␣ i2 were generated to "trap" the ␣ subunits in their GTP-bound forms and assayed for binding to GAIP. In contrast to the wild type proteins, the "activated" forms of both G␣ i1 and G␣ i2 interacted at least as strongly with GAIP as wild type G␣ i1 (Fig. 4). Both G␣ i1 (Q204L) and G␣ i2 (Q205L) generated about a 4-fold increase in GAIP binding activity relative to that seen with wild type (nonactivated) G␣ i1 , so that the selectivity of GAIP for G␣ i1 over G␣ i2 appears to be restricted to the interaction with their ground state (presumably GDP-bound) conformations. Two additional GTPase-deficient mutants, G␣ i1 (R178C) and G␣ i2 (R179C), were also tested and interacted very strongly with GAIP although less strongly than the Q204L/Q205L mutants.
Nucleotide Binding Affinity-To explore the mechanism of the selectivity of GAIP for G␣ i1 over G␣ i2 in their GDP-bound states, the position of G␣ i1 aspartate 229 in relation to the bound RGS4 molecule in the published crystal structure was examined (Fig. 5). In the AlF 4 -activated state in which this G protein was crystallized, Asp 229 appears closer to the nucleotide-binding site than to the RGS-binding site of this G protein. Therefore, we examined the relative GTP␥S affinities of both G␣ i1 and G␣ i2 to determine whether there were differences in nucleotide binding affinity that in turn might affect their affinities for GAIP. Recombinant full-length G␣ i1 , G␣ i2 , G␣ i1 (D229A), and G␣ i2 (A230D) were GST-tagged, expressed in bacteria, and purified to homogeneity over glutathione affinity columns. The ability of unlabeled GTP␥S to displace placement was the same for all four proteins (Fig. 6), so differences in GAIP binding are not reflective of differences in nucleotide binding affinities.
GAP Activity-Finally, to determine whether any functional differences might correlate with selective binding capacity, we tested the ability of GAIP to catalyze the GTPase activities of G␣ i1 , G␣ i2 , G␣ i1 (D229A), and G␣ i2 (A230D). GAIP catalyzed a 5-fold increase in the rate of GTP hydrolysis by G␣ i1 (Fig. 7A) but caused no increase in the GTPase rate of G␣ i2 (Fig. 7B) under the same conditions. In addition, GAIP only slightly increased the GTPase activity of G␣ i1 (D229A) (from K obs of 2.1 in the absence of GAIP to K obs of 3.7 in the presence of GAIP) (Fig. 7C). Of particular interest, the rate of GTP hydrolysis seen for this mutant form of G␣ i1 in the presence of GAIP is similar to the GTPase rate of G␣ i2 in the presence of GAIP (K obs ϭ 4.2). Similarly, G␣ i2 (A230D) now behaves more like G␣ i1 in that there is a significant increase in GAIP activation, and the GTPase rate seen in the presence of GAIP is similar to that seen for G␣ i1 in the presence of GAIP (K obs ϭ 5.2 for the former and 5.6 for the latter) (Fig. 7D). Therefore, the ability of GAIP to act as a GAP for these two G␣ i proteins and their reciprocal mutants correlates with its affinities for these proteins in their "ground states" as measured in the yeast two-hybrid assay. DISCUSSION RGS proteins are a family of G protein regulators that downregulate G protein-coupled responses by stimulating the GTPase activity of the G␣ subunits to which they bind (3,5). Both the G protein binding and GAP activity of RGS molecules have been localized to a 130-amino acid domain (RGS domain) that is conserved among all RGS proteins (6,10,24). Within this domain, a number of residues have been shown to serve as contact points for G␣ protein binding (11,25,26).
Elucidation of the sites on G proteins with which RGS proteins interact and the selectivity of RGS proteins for different forms of G␣ have important implications for the mechanism by which RGSs stimulate ␣ subunit GTPase activity. The observation that RGS4 binds more strongly to the AlF 4 -GDP-Mg 2ϩbound state of G␣ i than to the GDP or GTP-bound states suggests that RGSs exhibit GAP activity by stabilizing the transition state for GTP hydrolysis by G␣ (7-9, 27, 28). The crystal structure of AlF 4 -GDP-Mg 2ϩ -G␣ i bound to RGS4 further reveals that the RGS interacts directly with the Switch regions of G␣ i , reducing their flexibility in this transition state mimic and thus further supporting this proposed GAP mechanism (11). It has also been observed that the sites on G␣ to which RGS proteins bind may interfere with the binding of the effector PLC␤1, suggesting another possible mechanism for G␣ i down-regulation by RGSs (13).
The sites on G protein ␣ subunits responsible for the selectivity with which RGS proteins bind have been less well studied. DeVries et al. (29) showed a significantly reduced GAIP interaction with a 10-amino acid truncation of G␣ i3 , but a chimeric G␣ q containing the last 10 residues of G␣ i3 did not bind to GAIP, indicating that other determinants remain to be identified. More recently, Lan et al. (30) showed that a G184S mutation in G␣ o and the equivalent mutation in G␣ i1 prevents both binding to and activation by RGS4, extending the observation by DiBello et al. (31) that a mutant Gpa1 prevented a functional interaction with the yeast RGS sst2. However, because this glycine is a highly conserved Switch 1 residue, it appears to be required for all G␣ interactions with RGS molecules rather than a determinant for specificity. Finally, Natochin and Artemyev (32) showed that the interaction of G␣ t with human retinal RGS could be abolished by mutating serine 202 to the corresponding G␣ s aspartate, providing one candidate G␣ s site that might interfere with RGS binding. They recently extended this finding by showing that mutation of this G␣ s aspartate (G␣ s Asp 229 ) to the serine which occurs in G␣ i family members at the corresponding Switch 1 position promotes binding to an RGS (33).
To extend the characterization of RGS/G protein specificities and their structure/function relationships, we sought to identify regions in the G␣ subunit that contributed to GAIP binding selectivity by testing the relative interaction strengths of GAIP with a number of native G protein ␣ subunits, mutants, and chimeras using the yeast two-hybrid system. In this system, GAIP interacts equally strongly with native forms of G␣ i1 , G␣ i3 , and G␣ o but very weakly with G␣ i2 and not at all with G␣ s . Both G␣ s/i3 and G␣ i3/s chimeras disrupted GAIP binding, indicating either that both the N and C termini of the G␣ i subunit contain determinants required for binding or that divergent sequences in the G␣ s protein relative to G␣ i may interfere with GAIP contact points. G␣ i1/i2 and G␣ i2/i1 mutants gave more interpretable results, indicating that the C-terminal domain of G␣ i1 is required for GAIP binding. This region constitutes most of the GTPase domain of the G protein, which is consistent with reports showing that GAIP binds in a groove within this domain (11). By comparison, the failure of either G␣ s chimera to bind may indicate that N-terminal inserts in the G␣ s sequence (such as amino acids 72-86) relative to G␣ i interfere with the RGS-G␣ binding surface or that other divergent residues in the G␣ s N-terminal portion interfere with RGS contact. The interfering aspartate (G␣ s residue 229) proposed by Natochin and Artemyev (32,33) is in fact in the N-terminal portion of our chimeras, consistent with this possibility.
To further localize the region in the G protein C terminus responsible for GAIP selectivity, site-directed mutants were generated in which residues in G␣ i1 and G␣ i2 were swapped. Candidate residues were chosen on the basis of their conservation in G␣ i1 and G␣ i3 and divergence in G␣ i2 . The mutation of aspartate 229 of G␣ i1 to the alanine present in G␣ i2 nearly abolished GAIP binding. Conversely, when aspartate was substituted for the alanine normally present at the same site in G␣ i2 , the mutant G␣ i2 bound GAIP to the same extent as native G␣ i1 . These results reveal the importance of aspartate 229 for the binding of G␣ i subunits in their native state to GAIP and potentially suggest a site of physical contact with GAIP. Yet, upon inspection of the G␣ i1 -RGS4 crystal structure, this aspartate appears quite far from the sites of RGS4 interaction. Due to the location of G␣ i1 aspartate 229 at the far N terminus of Switch 3, it is possible that the position of this amino acid in the AlF 4 transition state analogue in which it was co-crystallized with RGS4 differs from its position in the nonactivated state in which the G␣ i s show selectivity for binding to GAIP. That is, it may be that in its GDP-bound (ground state) conformation, G␣ i1 Asp 229 is in closer proximity to GAIP than in its AlF 4 -Mg 2ϩ -GDP-bound conformation.
Closer inspection of the RGS4-G␣ i1 crystal structure presents an alternative explanation. In this structure, aspartate 229 appears to be involved in a relay system that connects its carbonyl through a water molecule to lysine 270, which in turn maintains a hydrophobic interaction with GDP in the RGS4-G␣ i1 crystal structure. We hypothesized that removal of the carbonyl group at this position by mutation to an alanine might disrupt this relay system, destabilizing the binding of nucleotide and hence the binding of RGS, because its binding is dependent on the nucleotide-bound state of the G protein. To test this possibility, IC 50 values for the ability of GTP␥S to compete [ 35 S]GTP␥S binding by G␣ i1 , G␣ i2 , G␣ i1 (D229A), and G␣ i2 (A230D) were compared. The displacement curves were identical in all cases, implying that differences in nucleotide binding capacities do not account for RGS binding differences.
Finally, to determine whether there is also selectivity by GAIP for G␣ i1 versus G␣ i2 in their GTP-bound forms, GTPasedeficient mutants of both G␣ i1 and G␣ i2 were engineered and tested for GAIP binding in the yeast two-hybrid system. Interestingly, both G␣ i1 (Q204L) and G␣ i2 (Q205L) exhibited similarly high binding affinities to GAIP (about four times the native G␣ i1 interaction), consistent with an inability by GAIP to discriminate between the two proteins in their GTP-bound states. The G␣ i1 (R178C) and G␣ i2 (R179C) GTPase-deficient mutants interacted less strongly than the Q204L/Q205L mutants, although still more strongly than their native counter- FIG. 7. Effect of GAIP on single turnover GTPase activity of purified GST-tagged native and mutant G␣ subunits. Squares, G protein alone; triangles, G protein in the presence of GAIP. GST-GAIP (500 nM) was added to 100 nM [ 32 P]GTP-loaded G␣ subunits in the presence of Mg 2ϩ and excess unlabeled GTP to initiate reactions. A, G␣ i1 ; B, G␣ i2 ; C, G␣ i1 (D229A); D, G␣ i2 (A230D). Aliquots were removed at the indicated times, and free 32 P i released was measured. An average of 470 fmol 32 P i was released per assay, which was normalized to 100%. Values given are the means of four experiments for A, B, and C and the means of six experiments for D. The observed rate constants (K obs ) for each reaction were calculated based on an exponential association curve fit using GraphPad Prism.
parts. This may reflect the ability of RGS proteins to partially restore the GTPase activity of R178C mutants, but not Q204L mutants (7), such that Q204L mutants remain in their GTPbound states, but R178C mutants may reflect a mixture of conformations. These data also bring up an alternative explanation for the preferential binding of GAIP to nonmutated G␣ i1 over G␣ i2 , namely that there is a greater population of GTPbound G␣ i1 than GTP-bound G␣ i2 in the yeast cell. This could result from different rates of GTP/GDP exchange or GTP turnover by the two ␣ subunits. Formally, that remains a possibility. However, because mammalian G␣ proteins do not couple to yeast G protein-coupled receptors (34) and because G proteins remain GDP-bound in the absence of receptor stimulation (35), we find it more likely that there is a structural difference between the two G␣ i s that is recognized by GAIP only in their nonactivated states.
To determine whether the ability of GAIP to discriminate between G␣ i1 and G␣ i2 only in their GDP-bound states has any functional significance, we measured the GAP activity of GAIP with each of these proteins and their mutants. GAIP enhanced the rate of GTP hydrolysis of G␣ i1 but not G␣ i2 under similar conditions. Furthermore, as predicted by the binding studies, G␣ i1 (D229A) was a poor substrate for GAIP GAP activity compared with native G␣ i1 , and G␣ i2 (A230D) was comparable with G␣ i1 as a substrate for GAIP GAP activity. Although Berman et al. (7) showed GAIP-catalyzed increases in GTPase activity of both G␣ i1 and G␣ i2 , Heximer et al. (17) also showed a greater enhancement by GAIP of G␣ i1 over G␣ i2 GTPase activity. Our results indicate that GAIP preferentially enhances G␣ i1 over G␣ i2 GTPase activity and that this activity correlates with the binding selectivity shown for G␣ i s in their ground state conformations. In addition, because GTPase-deficient mutants of both ␣ i1 and ␣ i2 subunits bind tightly to GAIP, these results may imply that GAP binding is not sufficient for GAP catalytic activity. Indeed, differential effects on G␣ binding versus GAP activity were discerned by Chen et al. (25) using various RGS mutants, consistent with this idea. It may be that the difference in the binding affinities for GTP-bound versus GDP-bound G␣ conformations drives GTP hydrolysis, so that binding to the activated G protein conformation is not the only indicator of RGS functional selectivity.
The functional selectivity displayed by GAIP and other RGS proteins for G protein partners in vivo remains to be explored. The contributions of additional interacting partners, including C-terminal tails of GPCRs (36) and additional effector proteins (18 -20), and post-translational modifications (37) will have to be considered to determine how individual RGS proteins modulate specific G protein signaling pathways.