Structure of Giα1·GppNHp, Autoinhibition in a Gα Protein-Substrate Complex*

The structure of the G protein Giα1 complexed with the nonhydrolyzable GTP analog guanosine-5′-(βγ-imino)triphosphate (GppNHp) has been determined at a resolution of 1.5 Å. In the active site of Giα1·GppNHp, a water molecule is hydrogen bonded to the side chain of Glu43 and to an oxygen atom of the γ-phosphate group. The side chain of the essential catalytic residue Gln204 assumes a conformation which is distinctly different from that observed in complexes with either guanosine 5′-O-3-thiotriphosphate or the transition state analog GDP·AlF4 −. Hydrogen bonding and steric interactions position Gln204 such that it interacts with a presumptive nucleophilic water molecule, but cannot interact with the pentacoordinate transition state. Gln204 must be released from this auto-inhibited state to participate in catalysis. RGS proteins may accelerate the rate of GTP hydrolysis by G protein α subunits, in part, by inserting an amino acid side chain into the site occupied by Gln204, thereby destabilizing the auto-inhibited state of Gα.

The G ␣ subunits of heterotrimeric G proteins are GTP hydrolyases which, upon receptor activation, bind to GTP and regulate effector molecules (1,2). G ␣ -catalyzed hydrolysis of GTP to GDP releases the G ␣ ⅐GDP product complex from effector and allows its sequestration by inhibitory G ␤␥ subunits. The rate of the hydrolytic reaction thus determines, in part, the length of time during which the effector is regulated by activated G proteins.
Structures of the G ␣ subunits G t␣ (transducin), G i␣1 , and G s␣ complexed with Mg 2ϩ and the nonhydrolyzable GTP analog GTP␥S 1 have provided views of the active site in the ground state (7)(8)(9). In particular, the positions of the presumptive nucleophilic water (W nuc ), Mg 2ϩ , and the side chains of catalytically significant residues have been determined. Structures of G i␣1 and G t␣ complexed with GDP⅐AlF 4 Ϫ ⅐W nuc , a mimic of the pentavalent transition state for GTP hydrolysis, have also been determined (8,10).
Mutagenesis experiments demonstrate that two residues of G i␣1 , Arg 178 and Gln 204 , are required for catalysis (11,12). In crystals of G i␣1 ⅐GTP␥S, which mimics the ground state E⅐S complex, the side chains of these residues are partly disordered and do not interact with the substrate analog. In contrast, in complexes of G i␣1 and its homolog G t␣ with GDP⅐AlF 4 Ϫ , both residues are well ordered and interact with AlF 4 Ϫ and its axial water ligand (8,10). It was proposed that Arg 178 and Gln 204 stabilize the pentavalent transition state through an analogous set of interactions (8,10). It was also suggested that the low catalytic rate (k cat Ϸ 2-4 min Ϫ1 ) exhibited by G i␣1 (14) and its homologs might be attributed to a high activation energy for the conformational rearrangement of Arg 178 and Gln 204 from the ground state to the transition state (2). However, the nature of this rearrangement is not established, in part because the G i␣1 ⅐GTP␥S complex may not accurately mimic the true E⅐S complex. Specifically, the thiol substituent of the ␥-phosphate in GTP␥S may sterically perturb the catalytic site. The sulfur atom is both more bulky than the corresponding oxygen atom of GTP (van der Waals radii of 1.8 and 1.4 Å, respectively) and has a longer bond length with the ␥-phosphate atom (P-S, 1.86 Å; P-O, 1.52 Å). Hence, to obtain a more accurate view of the E⅐S ground state, we have determined a high resolution x-ray crystal structure of G i␣1 complexed with an alternative nonhydrolyzable GTP analog: GppNHp. In GppNHp all three terminal substituents of the ␥-phosphate group are oxygen atoms.
The active site of the G i␣1 ⅐GppNHp complex differs from that with GTP␥S in several respects, but most importantly in the conformation of the catalytic residue Gln 204 . In this conformation, Gln 204 interacts with W nuc , but neither contacts the nucleotide nor is positioned to interact with the pentavalent transition state. Thus, Gln 204 may participate directly in the G i␣1 ⅐GTP ground state but must reorient to stabilize the transition state. RGS4, a member of the RGS family of G protein stimulatory factors (15,16), may accelerate hydrolysis of GTP by G i␣1 , in part, by destabilizing the ground-state conformation of Gln 204 .

MATERIALS AND METHODS
Non-myristoylated, recombinant rat G i␣1 was synthesized in E. coli and purified as described previously (17). Crystals of G i␣1 complexed with GppNHp and Mg 2ϩ were grown and prepared for x-ray data collection as described except that GppNHp was substituted for GTP␥S (18). Crystals were transferred to a cryoprotection solution containing 15% (v/v) glycerol and frozen in liquid propane as described (19). Two x-ray diffraction data sets were collected at 100 K, each from a single crystal. The first was measured at the Cornell synchrotron source (CHESS) A1 line ( ϭ 0.91 Å) equipped with an ADSC Quantum 1 CCD detector. The crystal diffracted beyond 1.5 Å, but data were measured only to 1.7 Å. A second data set collected to 1.5 Å was collected at the CHESS F1 line ( ϭ 0.92 Å) equipped with an ADSC Quantum 4 CCD detector. Data were processed using the DENZO/SCALEPACK programming package (20). After simultaneous determination of the scale factors for both data sets using all observations, data between 15-2.24 Å from the first data set and between 2.24 -1.50 Å from the second were combined to obtain the final 15.00 -1.50-Å data set (Table I).
The structure of the G i␣1 ⅐GppNHp complex was solved by molecular replacement, using the G i␣1 ⅐GTP␥S complex (PDB accession code 1gia) with the nucleotide, Mg 2ϩ , and with waters removed as the starting model. SigmaA-weighted 2 F o Ϫ F c and F o Ϫ F c difference maps (21) indicated strong difference electron density for GppNHp, a previously * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The atomic coordinates and structure factors (code1cip) has been deposited in the Protein Data Bank, Brookhaven National Laboratory, Upton, NY.
¶ To whom correspondence should be addressed: Howard Hughes Medical Institute, The University of Texas Southwestern Medical Center at Dallas, 5323 Harry Hines Blvd., Dallas, TX 75235-9050. Tel.: 214-648-5008; Fax: 214-648-6336; E-mail: Sprang@howie.swmed.edu. 1 The abbreviations used are: GTP␥S, guanosine 5Ј-O-3-thiotriphosphate; RGS, regulator of G protein signaling; GppNHp, guanosine-5Ј-(␤␥-imino)triphosphate; GppCHp, guanosine-5Ј-(␤␥-methylene)triphosphate; W nuc , nucleophilic water; Gln cat and Arg cat , conserved catalytic residues in G␣ subunits. unobserved active site water molecule, and new side chain positions for Glu 43 and Gln 204 . Small changes in the side chain conformations of other residues were also observed. Simulated annealing omit maps (22) were used to confirm that the observed differences were not caused by model bias. The protein model was manually rebuilt and ligands were added using the interactive graphics model building program O (23). Positional and atomic temperature factor refinement was carried out on the rebuilt model using XPLOR (24) followed by additional rounds of model building and refinement. In the final rounds, a bulk solvent correction (25) was applied using XPLOR. Electron density about the active site is shown in Fig. 2. The free-R factor, computed using 5% of the data, was used to monitor the refinement (26). The structure exhibits good stereochemistry, with 94% of the residues in the most favored regions of the Ramachandran plot as analyzed by PROCHECK (27). Comparisons and superposition of the model with other proteins was carried out using O. Data collection and refinement statistics are given in Table I.
The model of the RGS4⅐G i␣1 ⅐GppNHp complex was made by superimposing the C␣ atoms of residues 34-343 of G i␣1 from the G i␣1 ⅐GppNHp complex onto the corresponding atoms of the RGS4⅐G i␣1 ⅐GDP⅐AlF 4 Ϫ complex (PDB accession code 1agr). G i␣1 ⅐GDP⅐AlF 4 Ϫ was then replaced by the superimposed G i␣1 ⅐GppNHp complex. Analysis and in silico mutations were then performed on the resulting RGS4⅐G i␣1 ⅐GppNHp model using O.

RESULTS
The structure of the G i␣1 ⅐GppNHp complex ( Fig. 2A) differs from that of G i␣1 ⅐GTP␥S (Fig. 2B) in two ways. First, there are small changes in the side-chain positions of several surface residues that are poorly ordered in crystals of G i␣1 ⅐GTP␥S. These differences are most likely because of damping of thermal motions in the frozen G i␣1 ⅐GppNHp crystals. In contrast, 2.0 Å data from G i␣1 ⅐GTP␥S crystals were collected at 14°C. Additionally, the G i␣1 ⅐GppNHp data set was measured to 1.5 Å, permitting unambiguous assignment of side chain conformations.
The second, and more interesting, group of changes relative to G i␣1 ⅐GTP␥S are observed in the active site and are most likely attributable to GppNHp ( Fig. 2A). For the most part, the active site is similar to that observed in the G i␣1 ⅐GTP␥S complex (Fig. 2B). All of the interactions between the guanine nucleotide, Mg 2ϩ , and the protein observed in the latter complex are also present in G i␣1 ⅐GppNHp. The presumptive water nucleophile (W nuc ) is also clearly observed (Fig. 1). A new feature of the active site is a highly ordered (B ϭ 21.9Å 2 ) water molecule (W 600 ) bound to the O1G oxygen of the ␥-phosphate ( Fig. 1 and Fig. 2A). Modeling indicates that this water molecule cannot bind to the G i␣1 ⅐GTP␥S complex because it would be sterically excluded by the ␥-thiophosphate sulfur atom of GTP␥S. The conformation of Glu 43 is altered, and its carboxyl-ate moiety forms a hydrogen bond to W 600 and consequently cannot form the hydrogen bond to Arg 242 that is present in G i␣1 ⅐GTP␥S (Fig. 2, A and B). In G i␣1 ⅐GppNHp, Glu 43 and Arg 178 form a doubly hydrogen-bonded ion pair ( Fig. 2A), in contrast to the less intimate ion pair interaction observed in the G i␣1 ⅐GTP␥S complex (Fig. 2B). This enhanced salt-bridge may strengthen binding of GTP in the ground state. Remarkably, the same interaction occurs in crystals of G i␣1 ⅐GDP complexed with G protein ␤␥ subunits (28). The conformations of Arg 178 in G i␣1 ⅐GppNHp and G i␣1 ⅐GTP␥S differ slightly, but in neither case does Arg 178 interact with the GTP substrate analog. In contrast, in the G t␣ ⅐GTP␥S complex, the corresponding residue, Arg 174 , interacts directly with the oxygen bridging the ␤and ␥-phosphate atoms and the ␥-thiophosphate sulfur atom. The bridging NH group of GppNHp would not be capable of this interaction however.
The orientation of Gln 204 in the GppNHp complex is particularly interesting. In the G i␣1 ⅐GTP␥S complex the side chain of this residue, which assumes the favored gauche ϩ 1 conformation, is poorly ordered and directed out of the active site, and does not appear to interact with any other residues (Fig. 2B).  Similar conformations for the corresponding catalytic glutamine (Gln cat ) are observed in the GTP␥S-G t␣ , -G s␣ , and -G s␣ ⅐adenylyl cyclase complexes (7,9,29). In contrast, Gln 204 exhibits strong density for the gauche Ϫ 1 conformation in the G i␣1 ⅐GppNHp complex and makes several contacts with other moieties (Figs. 1 and 2A). Notably, while Gln 204 does not interact directly with the substrate analog, its side chain amino group forms a hydrogen bond to W nuc . The orientation of the amide group is defined by the functionality of the three other moieties to which W nuc is hydrogen bonded: the O1G oxygen of the ␥-phosphate group, which is proposed to act as a catalytic base (30,31), the main chain carbonyl oxygen of Thr 181 , and the main chain NH group of Gln 204 ( Fig. 2A). Gln 204 is also anchored by a hydrogen bond between its carboxamide oxygen atom and the hydroxyl moiety of Ser 206 . The thiophosphate moiety of GTP␥S does not prevent Gln 204 from adopting the conformation observed in the GppNHp complex; however, it does block entry of W 600 to the active site. W 600 , in turn prevents Gln 204 from assuming the conformation that is partly populated in the GTP␥S complex. In the GppNHp complex, the hydrogen bond between the phosphate O1G oxygen and W nuc is shorter than the corresponding interaction in the GTP␥S complex (2.82 versus 3.27 Å). The position of W nuc closer to the ␥-phosphate permits a more favorable interaction with the side chain of Gln 204 . The well ordered conformation of Gln 204 in the GppNHp complex may be attributed to hydrogen bonds formed with W nuc and Ser 206 but also to the conformational restriction imposed by W 600 as well as the cryogenic data measurement conditions. In other G protein-GppNHp or -GppCp complexes, Gln cat also interacts with W nuc (32,33). However, in these cases Gln cat assumes the gauche ϩ 1 conformation.

DISCUSSION
The most intriguing feature of the G i␣1 ⅐GppNHp complex is the conformation of the side chain of Gln 204 and its implications for the mechanism of both intrinsic and RGS-stimulated GTP hydrolysis. As established by mutagenesis studies (11), Gln 204 , and the corresponding residues in other G proteins (12,34,35), is absolutely required for enzymatic activity. Studies of substrate-assisted catalysis using GTPase-deficient G s␣ have demonstrated that a hydrogen donor group near the ␥-phosphate can substitute for Gln cat (36). However, the function of Gln cat during GTP hydrolysis by G proteins has been debated. Hydrolysis of GTP in G proteins is believed to occur by a direct, in-line attack on the ␥-phosphate atom by a nucleophilic water (37)(38)(39). Crystal structures of G i␣1 and other G proteins have identified, near the ␥-phosphate group, a well ordered water molecule (W nuc ) that is positioned to carry out a nucleophilic attack (7-9, 32, 40, 41). Although Gln 204 is near W nuc in G i␣1 ⅐GTP␥S, it does not directly contact it. Further, W nuc is observed in Gln 204 3 Leu G i␣1 ⅐GTP␥S, and Gln 61 3 Leu Ras⅐GppNHp (Gln 61 is Gln cat in Ras) (8,42), indicating that Gln cat is not required for binding of W nuc in the ground state. It has been pointed out that the basicity of glutamine is low, and it is therefore unlikely that Gln cat acts as the general base that deprotonates W nuc (43). Rather, an oxygen of the presumably dianionic ␥-phosphate is proposed to serve this function in Ras (30,31). Hence, Gln cat may only polarize W nuc in the ground state.
X-ray crystallographic studies have indicated that Gln cat stabilizes the transition state (2), as originally proposed by Prive et al. (43). Structures of G i␣1 and G t␣ complexed with the transition state analog GDP⅐AlF 4 Ϫ , reveal Gln cat positioned within the active site and directly interacting with a fluorine substituent and W nuc of the GDP⅐AlF 4 Ϫ ⅐W nuc complex. Fig. 2C shows these interactions in the RGS4⅐G i␣1 ⅐GDP⅐AlF 4 Ϫ complex (44). It was proposed that the amino group of Gln cat stabilizes negative character on the equatorial oxygen of the transition state and its carbamoyl oxygen stabilizes the attacking nucleophilic water. A similar configuration is observed in the Ras-GAP⅐Ras⅐GDP⅐AlF 4 Ϫ and the Rho-GAP⅐Cdc42Hs⅐GDP⅐AlF4 Ϫ complexes (45)(46)(47). In addition, mutations that perturb the transition state conformation of Gln cat abolish GTPase activity (42,48). These observations indicate that Gln cat stabilizes the transition state for GTP hydrolysis.
The GppNHp complex provides novel insights into both the mechanism of GTP hydrolysis as well as to the role of both Arg cat and Gln cat in the ground state E⅐S complex. GppNHp, but not GTP␥S, permits a water molecule, W 600 , to occupy a position in which it could act as the ultimate proton acceptor from W nuc . Water molecules in similar, but not identical, positions are present in the GppNHp-or GppCp-bound complexes of Ras and Rac1 (32,40). A proton could be relayed from W nuc to W 600 via O1G of the GTP ␥-phosphate. This substituent does not otherwise participate in hydrogen bonds with the protein and corresponds to the thiol of GTP␥S. The basicity of W 600 may be enhanced by hydrogen bond formation with Glu 43 , which is well conserved in G␣ proteins with the exception of G z␣ where it is replaced with an asparagine residue. Glu 43 also forms a hydrogen-bonded ion pair with Arg 178 . In this conformation, Arg 178 is restrained from interacting with the ␥-phosphate of GTP. Transfer of a proton from W nuc to W 600 would tend to weaken this ion pair, releasing Arg 178 to stabilize the incipient pentacoordinate phosphoryl transition state. W 600 also blocks the side chain of Gln 204 from interacting with the pentacoordinate phosphate. Thus, until it diffuses from the active site, W 600 impedes the reorganization of the catalytic site that is required for transition state stabilization.  38 -48), blue (Switch I residues 178 -184), and yellow (Switch II residues 200 -208). The main chain segment of RGS4 in panels C and D is shown in red (residues 126 -131). The atoms and water molecules are colored as described in Fig. 1, except that the phosphorous atoms are in yellow, magnesium is blue, and the sulfur atom of GTP␥S is green. In panel D, the region of the hypothetical model where Asn 128 of RGS4 and Gln 204 of G i␣1 collide is highlighted in cyan.
Gln 204 is anchored in a noncatalytic conformation by hydrogen bonds to both W nuc and Ser 206 (Ser 206 is substituted by an Asp in G ␣s ,). In the ground state, Gln 204 could orient and perhaps activate W nuc ; however, to stabilize the transition state as represented by G protein GDP⅐AlF 4 Ϫ complexes, Gln 204 must sever its hydrogen bond with Ser 206 and W nuc and rotate Ϸ120°about 1 and Ϸ90°about 2 and 3 (to gaucheϩ and gaucheϪ, respectively) such that its carbamoyl group donates a hydrogen bond to the equatorial oxygen of the pentacoordinate ␥-phosphoryl group and accepts a hydrogen bond from W nuc . Such would incur a substantial penalty in catalytic efficiency and perhaps account, at least in part, for the low catalytic rate of GTP hydrolysis in G ␣ and perhaps in other G proteins.
We propose that the ground state G i␣1 ⅐GTP complex is "autoinhibited" with Gln cat locked into an unproductive conformation. Active site residues in the EF-Tu⅐GppNHp complex also assumes anti-catalytic positions; in this case His cat , the residue corresponding to Gln cat , cannot interact with the substrates because of steric interference by other active site residues (41). In G i␣1 , catalysis could occur only if the bonds that hold Gln cat in this position are broken, and the side chain freed to interact with the pentacoordinate transition state. This model predicts that changes that disrupt the ground state conformation of Gln 204 , while not otherwise compromising the active site, would increase k cat .
RGS proteins, which accelerate the rate of GTP hydrolysis by G i␣1 by 50 -100-fold, may act in part by destabilizing the ground state conformation of Gln cat , as well as stabilizing its productive conformation in the transition state (44). The crystal structure of the RGS4⅐G i␣1 ⅐GDP⅐AlF 4 Ϫ complex demonstrates that RGS4 stabilizes the active site of G i␣1 in the conformation corresponding to that of the transition state complex (Fig. 2C) (44). No residues from RGS4 are inserted into the active site except Asn 128 , which could enhance catalysis by aiding in binding, orienting, and polarizing W nuc in the pretransition state complex (44).
However, superposition of G i␣1 ⅐GppNHp and G i␣1 ⅐GDP⅐AlF 4 Ϫ from the RGS4⅐G i␣1 ⅐GDP⅐AlF 4 Ϫ complex reveals that the carbamoyl groups of Gln 204 and Asn 128 occupy nearly the same positions, although the side chains approach from opposite directions (Fig.  2D). Further, both residues are positioned such that they can bind the nucleophilic water and Ser 206 . We suggest that Asn 128 of RGS4 displaces the side chain of Gln 204 from its "anti-catalytic" position in the ground state, freeing it to participate in stabilization of the transition state.
Mutational analysis of RGS proteins supports this hypothesis. Mutation of Asn 131 in hRGSr (analogous to Asn 128 of RGS4) to either serine or glutamine resulted in a relatively small decrease in the k cat of G t␣ (49). In addition, hRGSr in which Asn 131 was mutated to leucine or alanine also retains substantial stimulatory activity, and the loss of activity that was observed could be attributed to weakened binding of these mutants to G t␣ . Similar mutagenic studies have been performed with RGS4 (50). Mutants of Asn 128 analogous to those of hRGSr Asn 131 were modeled in the structure of the "RGS4⅐G i␣1 ⅐GppNHp" complex. In all cases these residues were in steric conflict with Gln 204 . These findings indicate that the bulk and binding of the residue at position 128 is important to the stimulatory activity of RGS4 although it is unlikely that it has a direct catalytic role in stimulation of GTPase activity (49).
The evidence presented is consistent with a self-inhibited or anti-catalytic model of the ground state of G ␣ proteins, and a role for RGS proteins in stimulating GTPase activity by releas-ing G ␣ subunits from this ground state while stabilizing the transition state.