Second Site Suppressor Mutations of a GTPase-deficient G-Protein α-Subunit

G proteins transmit signals from cell surface receptors to intracellular effectors. The intensity of the signal is governed by the rates of GTP binding (leading to subunit dissociation) and hydrolysis. Mutants that cannot hydrolyze GTP (e.g.GsαQ227L, Gi2αQ205L) are constitutively activated and can lead to cell transformation and cancer. Here we have used a genetic screen to identify intragenic suppressors of a GTPase-deficient form of the Gα in yeast, Gpa1Q323L. Sequencing revealed second-site mutations in three conserved residues, K54E, R327S, and L353Δ (codon deletion). Each mutation alone results in a complete loss of the βγ-mediated mating response in yeast, indicating a dominant-negative mode of inhibition. Likewise, the corresponding mutations in a mammalian Gi2α (K46E, R209S, L235Δ) lead to inhibition of Gβγ-mediated mitogen-activated protein (MAP) kinase phosphorylation in cultured cells. The most potent of these βγ inhibitors (R209S) has no effect on Gi2α-mediated regulation of adenylyl cyclase. Despite its impaired ability to release βγ, purified recombinant Gpa1R327S is fully competent to bind and hydrolyze GTP. These mutants will be useful for uncoupling Gβγ- and Gα-mediated signaling events in whole cells and animals. In addition, they serve as a model for drugs that could directly inhibit G protein activity and cell transformation.

The actions of many hormones and neurotransmitters are mediated through a cell surface receptor, heterotrimeric "G protein," and an intracellular effector that propagates the signal. Upon agonist binding to the receptor, G␣ undergoes guanine nucleotide exchange and a conformational change leading to dissociation from ␤␥. The subunits remain in the active state until GTP is hydrolyzed, at which time they reassociate and signaling stops.
Not surprisingly, disturbances in the cycle of G protein activation and inactivation can lead to disease. The severe and often fatal diarrhea associated with cholera is caused by the pathogenic exotoxin of Vibrio cholerae (cholera toxin) which promotes ADP-ribosylation of G s ␣ in the gut epithelium. Similarly, whooping cough is caused by a toxin from Bordetella pertusssis and ADP-ribosylation of G i ␣ (1). A number of germ line and somatic cell defects in G␣ proteins have also been described. These include both activating and inactivating mutations, which can sometimes lead to cell transformation and cancer (2)(3)(4). An early example of an activated G protein allele was described by Landis, et al. (5), who showed that certain types of human pituitary tumors are associated with GTPasedeficient mutants of G s ␣. Another inherited disorder, known as pseudohypoparathyroidism type Ia, is associated with lossof-function mutations in G s ␣ (R231H, A366S). Interestingly, G s ␣ A366S is unstable and inactive at 37°C but is stable and constitutively active at the slightly lower temperature of the male testis, resulting in a paradoxical combination of pseudohypoparathyroidism and testotoxicosis in affected individuals (6).
G proteins can transform cells in at least two ways. First, mutations in G␣ can lead to direct activation of certain effector enzymes, which promote cell proliferation. However, this mechanism appears to be operative in only a minority of cases (3). Alternatively, activating mutations in G␣ can lead to constitutive dissociation from G␤␥. The ability of ␤␥ subunits to promote cell proliferation occurs through activation of small GTPbinding proteins such as Rac and Ras, which in turn lead to the activation of Jun and mitogen-activated protein (MAP) 1 kinases, respectively (3).
A strikingly similar signaling cascade has been identified through genetic analysis of the mating response pathway in yeast Saccharomyces cerevisiae. Haploid yeast cells secrete small peptide pheromones that bind to G protein coupled receptors. The G protein ␣ subunit (GPA1 gene product, Gpa1) does not directly activate any known effector; rather it is ␤␥ (Ste4/Ste18) that activates a signaling cascade that includes a MAP kinase homologue, Fus3. This in turn triggers a coordinated series of events required for mating, including cell fusion, new gene transcription, and cell cycle arrest in G 1 (7).
A principal goal of pharmacology is to identify drugs that bypass or otherwise compensate for the molecular defects that lead to disease. One emerging strategy is to use genetics to identify protein binding partners of a dysfunctional gene product because any interacting proteins represent alternative (nonmutated) targets for drug therapy (8). Another approach has been to screen for second-site mutations that compensate for the primary defect in a gene. Genetic suppression of a disease-causing mutant provides strong presumptive evidence that pharmacological suppression can also be achieved (9). Indeed, a strategically designed genetic screen can be used to search for drugs having a similar compensatory effect. If structural information is available, it may even be possible to design drugs that mimic the effect of the second-site suppressor mutations.
Past efforts to find second-site suppressors of transforming Ras or G-protein mutants have been unsuccessful (9). Here, we have exploited the genetic tractability of yeast to identify intragenic suppressors of a GTPase-deficient mutant, Gpa1 Q323L . Our goal was to find mutants that are locked in the inactive conformation, in both the absence and presence of GTP. Our expectation was that mutations that block subunit dissociation would inhibit G␤␥ signaling in a dominant manner. Significantly, the mutations identified in Gpa1 have similar properties when introduced into mammalian G␣ subunits. These findings illustrate the utility of the yeast system to identify G protein mutants with highly selective, and highly potent dominant negative properties.

MATERIALS AND METHODS
Strains, Plasmids, Mutagenesis-Established methods were used for the growth of bacteria Escherichia coli and the manipulation of plasmids (10). All molecular biology reagents were purchased from New England Biolabs and used according to manufacturer instructions. Yeast expression plasmids were pRS316, pRS316-GPA1 (11), pG1501, and pG1501-GPA1 (12). The transcription reporter plasmid was pBJ207 (13), which contains the lacZ gene under the control of FUS1 promoter (14,15). The bacterial expression plasmid was pET15b (Novagen) (16). The mammalian G␣ expression plasmid was pcDNA1amp (containing EE-epitope tagged versions of G i ␣, G s ␣, and G q ␣) (17,18). Other mammalian receptor expression vectors are described in the references provided below.
cAMP Assay-cAMP accumulation in intact cells was assayed as described (6,25). Briefly, 24 h after transfection, cells were replated in 24-well plates at 1.5 ϫ 10 5 cells/well and labeled with [ 3 H]adenine (4 Ci/ml, Amersham Pharmacia Biotech) for an additional 24 h. After pretreatment with pertussis toxin (where indicated, PTX, 200 ng/ml for 4 h), cells were washed with medium and stimulated with the appropriate agonist in the presence of isobutylmethylxanthine for 25 min. cAMP and ATP fractions were resolved, and cAMP accumulation was estimated by determining the ratio of cAMP radioactivity to the sum of radioactivity of cAMP and ATP.
Inositol Phosphate Accumulation-Inositol phosphate (IP) accumulation in intact cells was assayed as described (26,27). Briefly, 24 h after transfection, cells were replated in 24-well plates at 1.5 ϫ 10 5 cells/well and labeled with myo-[ 3 H]inositol (6 Ci/ml, Amersham Pharmacia Biotech) for 24 h. After washing with a medium containing 5 mM LiCl for 10 min, cells were incubated with the appropriate agonist in the presence of 5 mM LiCl for 45 min. IP and total inositol fractions were resolved on a Dowex AG 1-X8 formate column (Bio-Rad), and IP accumulation was estimated by determining the ratio of IP radioactivity to the sum of radioactivity of IP and total inositol.
Measurement of p44 HA-MAPK Activity-HA-MAPK activity was assayed as described (26,28) with modifications. Cells were transfected in 6-well plates at 7 ϫ 10 5 cells/well, placed in serum-free medium after 28 h, and assayed after an additional 20 h. After pretreatment with PTX (where indicated, 200 ng/ml for 4 h), cells were stimulated with the appropriate agonist for 8 min. HA-MAPK was immunoprecipitated from cell lysates (300 l, representing 4 ϫ 10 5 cells) with 2 g of 12CA5 antibody and 35 l of protein A-agarose (50% slurry). After washing once with lysis buffer and once with kinase buffer, the agarose beads were incubated at 22°C for 20 min in 50 l of kinase buffer (28) containing 250 g/ml myelin basic protein and 50 M [␥-32 P]ATP (2 Ci/tube, 700 cpm/pmol, NEN Life Science Products). The reaction was stopped with 5 l of 88% formic acid, and radioactivity incorporated into myelin basic protein was determined by filtration on Whatman P81 membranes (28).

RESULTS
Screening for Intragenic Suppressors of gpa1 Q323L -Our objective here was to identify dominant-negative-type inhibitors of ␤␥-mediated signaling. Our approach was to screen for intragenic suppressors of a GTPase-deficient allele, gpa1 Q323L . To this end, the gpa1 Q323L mutant was expressed using the galactose-inducible GAL1/10 promoter (plasmid pG1501) in cells lacking GPA1 (gpa1⌬, strain YGS5). The gpa1⌬ mutation ordinarily leads to constitutive signaling and cell division arrest; however, YGS5 is viable at 34°C because of a temperature-sensitive mutation that blocks the signal downstream of ␤␥ (ste11 ts ). Cells were initially maintained in galactose medium at 34°C and then plated and shifted to 24°C. Rare colonies that grew under these restrictive conditions were picked, patched, and tested for galactose-dependent growth (Fig. 1, top). Plasmids were isolated from 68 independent colonies and tested for retention of the original Q323L substitu- tion (by restriction digestion). Ten plasmids that satisfied these criteria were sequenced across the entire GPA1 open reading frame. As expected, all retained the Q323L mutation, and all contained one additional mutation at a second position. Two of these were codon substitutions (K54E, R327S) and one was a codon deletion (L353⌬). We then examined the properties of each mutant without overexpression, in the absence or presence of the Q323L substitution. Single-site and double-site mutations were prepared and expressed using a low copy plasmid (pRS316) and the normal GPA1 promoter. In this case, single-site mutants were able to complement the gpa1⌬ mutation but the Q323L and double-site mutants could not (Fig. 1, and data not shown).
Gpa1 Mutants Inhibit the Mating Response Pathway in Yeast-Complementation of gpa1⌬ reflects an ability of each mutant to bind ␤␥ in vivo. We then tested the ability of the mutants to inhibit receptor-dependent signaling in a GPA1 ϩ strain. In this case, all three single-site mutants led to a complete inhibition of pheromone response (determined by the growth inhibition "halo assay," Fig. 2A). The double mutants can also inhibit pheromone signaling (and will even complement gpa1⌬) but only when overexpressed (data not shown). In comparison, a 2-fold overexpression of wild-type GPA1 led to a more modest inhibition of the pheromone response, and the double-site mutants were without effect ( Fig. 2A). The mutant and wild-type forms of Gpa1 were expressed at equal levels, as determined by immunoblotting (data not shown). Similar results were obtained using a pheromone-induced transcription reporter assay (FUS1 promoter, lacZ reporter; Fig. 2B).
Inhibition of G␤␥-mediated MAP Kinase Activity in COS-7 Cells-The results presented in Figs. 1 and 2 reveal that all three single-site mutants will complement a gpa1⌬ mutant and can inhibit signaling even in the presence of wild-type Gpa1, consistent with a dominant-negative mode of action. We then examined if the corresponding mutations in a mammalian G␣ would also inhibit ␤␥ signaling in cultured cells, using the MAP kinase phosphorylation assay as a read out. G i2 ␣ was chosen for these experiments because it does not regulate MAP kinase activity directly, in contrast to G q and G s (28).
We constructed versions of G i2 ␣ with the K46E, R209S, and L235⌬ mutations and transfected these together with the D2dopamine receptor and an epitope-tagged version of MAP kinase (HA-MAPK). MAP kinase activity was measured after immunoprecipitation by monitoring incorporation of 32 P to myelin basic protein.
As shown in Fig. 3, all three G i2 ␣ mutants reduced basal MAP kinase activity by ϳ20%, as compared with control (empty vector) transfected COS-7 cells. Upon stimulation with a D2 agonist quinpirole, phosphorylation was attenuated 20 -40% by the mutants, with R209S being the most potent inhibitor. In comparison, two known inhibitors of ␤␥ binding-the ␤␥-binding domain of ␤-adrenergic receptor kinase (␤ARK⌬) and the transducin ␣ subunit (␣ t )-inhibited basal phosphorylation by ϳ30% and quinpirole-stimulated activity by 50 -60%. Similar results were obtained using transfected CHO cells (Fig. 3B). Wild-type G i2 ␣ does not inhibit ␤␥ signaling (data not shown), presumably because the wild-type protein responds normally to activation by the receptor.
Having determined that G i2 ␣ R209S potently inhibits ␤␥-mediated signaling, we also examined if the mutant has any effect on G␣-mediated signaling events. CHO cells were transfected with G i2 ␣ R209S , the D2 dopamine receptor, and the luteinizing hormone/hCG receptor. Whereas, G i2 ␣ R209S inhibited agonistdependent MAP kinase activity by ϳ40% (Fig. 3B), the same mutant had no effect on hCG-mediated stimulation (via G s ) or quinpirole-mediated inhibition (via G i ) of adenylyl cyclase (Fig.  4A). In comparison, signaling via G i ␣ was potently and selectively inhibited by PTX treatment.
The results presented in Figs. 3 and 4A reveal that G i2 ␣ R209S can inhibit signaling through ␤␥ but not through G␣. We also examined whether the corresponding mutations in G s ␣ and G q ␣ would behave similarly, at least with respect to G␣-mediated signaling events. G s ␣ R231S did not alter hCG-stimulated (G s ␣-mediated) production of cAMP (Fig. 4B); and G q ␣ R207S had no effect on carbachol-stimulated (G q ␣-mediated) inositol phosphate production (Fig. 4C). Thus the Arg-to-Ser mutation does not interfere with coupling between receptors and the endogenous (wild-type) G␣, or between G␣ and its downstream effectors, in all three cases tested.
Biochemical Properties of the Gpa1 R327S Mutant-The Argto-Ser mutation leads to a dramatic, dominant-negative-type inhibition of G␤␥ activity, at least in the two cases where this could be measured (Gpa1 R327S , G i2 ␣ R209S ). There are at least two ways that such a mutant could block ␤␥ signaling. First, it could alter the conformation or subunit binding affinity of G␣ for ␤␥. Second, it could simply prevent binding to GTP. To rule out this more trivial explanation, two types of experiments were performed using purified recombinant Gpa1 R327S . First, we measured the rate of pseudo-irreversible binding of [ 35 S]GTP␥S, which is limited by the rate of GDP dissociation (16,29). As shown in Fig. 5A, Gpa1 R327S is able to exchange GDP for GTP␥S. As a second measure of GTP binding, we determined the steady state rate of GTP hydrolysis by Gpa1, which also reflects the rate of guanine nucleotide exchange  (40 -43). Residues are numbered according to the Gpa1 sequence. A, K54E (K42 in G t ␣) results in a charge reversal in the guanine nucleotide binding pocket, near the ␥ phosphate of GTP. B, L353⌬ (L230 in G t ␣) would shorten the loop that connects the ␤4 strand and the ␣3 helix (above), thereby altering the conformation of the ␣2 helix (below) that binds to ␤␥. R327S is likely to disrupt an ion-pair interaction between the guanidinium of R327 (␣2 helix) and the carboxylate of E364 (␣3 helix) (R204, E241 in G t ␣). C, same as panel B but in the GTP-bound state showing the conformational change in the ␣2 helix (switch II domain). (29). As shown in Fig. 5B, the overall rate of GTP hydrolysis by Gpa1 R327S is comparable with that of wild-type. Thus there is no evidence that the R327S substitution prevents GTP binding to Gpa1. DISCUSSION Using a genetic screening strategy in yeast, we have isolated dominant-negative mutations that potently inhibit G␤␥-effector coupling and yet do not interfere with coupling between the receptor and G␣, or between G␣ and its effector. This is in contrast to previously described dominant-negative alleles (e.g. G s ␣ G226A , G s ␣ G49V ) which do affect receptor-G␣ coupling (30 -36). Other mutants have more selective inhibitory effects on G␤␥ signaling (G o ␣ S47C , G o ␣ N270D ), but these are only modestly better than wild-type G␣ at inhibiting ␤␥ function (37). Analogous mutations in yeast (Gpa1 G322L , Gpa1 N388D , Gpa1 G50V ) have similar, weakly dominant-negative properties in vivo (20,36,38,39). None of these mutations were found in our screen and have never been shown to suppress a GTPase-deficient allele of G␣. The highly selective nature of our mutants must stem in part from the powerful selection process used for their identification. Specifically, all three mutants were screened for their ability to suppress an activating mutation in G␣, a mutation that acts downstream of (and independently of) the receptor.
These results can be interpreted in the context of the available crystal structures of G t (transducin) (40 -43) and G i1 (44). Those studies revealed three regions in G␣-designated switch I, II, and III-that exhibit significant conformational differences in the active (G␣-GTP␥S-bound), transition (G␣-GDP-AlF 4 Ϫ ), and inactive (G␣-GDP-␤␥) states. All three switch regions contribute to the binding of guanine nucleotides and G␤␥ and are extremely well conserved among all heterotrimeric G proteins including Gpa1. On the basis of these structures, the dominant-negative activity of all three mutants can be rationalized. Thus, K54E results in a charge reversal, introducing a negatively charged side chain in the guanine nucleotide binding pocket and possibly interfering with binding to the ␥ phosphate of GTP (Fig. 6A). This mutation could compensate for the loss of GTPase activity by simply binding GDP preferentially over GTP. The Leu deletion would shorten a loop that constrains the switch II helix in a conformation needed for dissociation of ␤␥ (Fig. 6, B versus C). Finally, R327S is likely to disrupt an ion-pair interaction (in the GTP-bound state) between the guanidinium of Arg-327 and the carboxylate of Glu-364. The loss of this high energy interaction could also perturb the conformational change needed for release of ␤␥ (Fig. 6, B versus C). Indeed, a substitution at the corresponding Arg of G s ␣ (R231H) is associated with pseudohypoparathyroidism type Ia and exhibits a loss-of-function phenotype in cultured cells (22,45). A substitution of the ion-pair partner, E268A, has a similar phenotype. 2 Several challenges remain. Currently we are co-expressing the Arg-to-Ser mutant (as G i ␣) with oncogenic receptors, G proteins, and effectors in cell culture to determine its ability to block MAP kinase signaling and cell transformation. A long term goal is to express these mutants in animals, to determine how inhibition of G␤␥ signaling affects their growth and behavior in vivo. These experiments will help to distinguish G␣versus G␤␥-mediated signaling processes in a variety of systems, from yeast to humans.