Mutations at the Domain Interface of Gsα Impair Receptor-mediated Activation by Altering Receptor and Guanine Nucleotide Binding*

G protein α subunits consist of two domains, a GTPase domain and a helical domain. Receptors activate G proteins by catalyzing replacement of GDP, which is buried between these two domains, with GTP. Substitution of the homologous αi2 residues for four αs residues in switch III, a region that changes conformation upon GTP binding, or of one nearby helical domain residue decreases the ability of αs to be activated by the β-adrenergic receptor and by aluminum fluoride. Both sets of mutations increase the affinity of αs for the β-adrenergic receptor, based on an increased amount of high affinity binding of the β-adrenergic agonist, isoproterenol. The mutations also decrease the rate of receptor-mediated activation and disrupt the ability of the β-adrenergic receptor to increase the apparent affinity of αs for the GTP analog, guanosine 5′-O-(3-thiotriphosphate). Simultaneous replacement of the helical domain residue and one of the four switch III residues with the homologous αi2 residues restores normal receptor-mediated activation, suggesting that the defects caused by mutations at the domain interface are due to altered interdomain interactions. These results suggest that interactions between residues across the domain interface are involved in two key steps of receptor-mediated activation, promotion of GTP binding and subsequent receptor-G protein dissociation.

G protein ␣ subunits consist of two domains, a GTPase domain and a helical domain. Receptors activate G proteins by catalyzing replacement of GDP, which is buried between these two domains, with GTP. Substitution of the homologous ␣ i2 residues for four ␣ s residues in switch III, a region that changes conformation upon GTP binding, or of one nearby helical domain residue decreases the ability of ␣ s to be activated by the ␤-adrenergic receptor and by aluminum fluoride. Both sets of mutations increase the affinity of ␣ s for the ␤-adrenergic receptor, based on an increased amount of high affinity binding of the ␤-adrenergic agonist, isoproterenol. The mutations also decrease the rate of receptor-mediated activation and disrupt the ability of the ␤-adrenergic receptor to increase the apparent affinity of ␣ s for the GTP analog, guanosine 5-O-(3-thiotriphosphate). Simultaneous replacement of the helical domain residue and one of the four switch III residues with the homologous ␣ i2 residues restores normal receptor-mediated activation, suggesting that the defects caused by mutations at the domain interface are due to altered interdomain interactions. These results suggest that interactions between residues across the domain interface are involved in two key steps of receptor-mediated activation, promotion of GTP binding and subsequent receptor-G protein dissociation.
Heterotrimeric G proteins transmit signals from cell surface receptors to effector proteins that modulate a wide variety of cellular processes (1,2). The ␣ and ␤␥ subunits of G proteins are associated in the inactive GDP-bound form. Receptors activate G proteins by catalyzing replacement of GDP by GTP on the ␣ subunit. Receptor-catalyzed nucleotide exchange is thought to involve an "opening" of the guanine nucleotide binding pocket that facilitates GDP release and increases the relative affinity for GTP compared with GDP (3,4). The transient empty state of the G protein has a high affinity for the hormone-receptor complex. However, this state is short-lived due to the high intracellular concentration of GTP. Binding of GTP leads to dissociation of the receptor from ␣⅐GTP and ␤␥, each of which can transmit signals to effectors. Hydrolysis of GTP by the ␣ subunit regulates the timing of deactivation and reassociation of ␣ with ␤␥.
␣ subunit structures consist of two domains, a GTPase domain that resembles the oncogene protein p21 ras and a helical domain consisting of ␣ helices and connecting loops. The bound GDP is buried between the two ␣ subunit domains, suggesting that the helical domain may present a barrier to GDP release. Three regions in the GTPase domain (switches I-III) assume different conformations in the structures of GTP␥S 1 -bound versus GDP-bound ␣ subunits (5)(6)(7)(8). Switches I and II correspond to conformational switch regions in the structures of both p21 ras and EF-Tu. Like the helical domain, switch III, which is located at the interface of the two domains, is unique to the structures of heterotrimeric G protein ␣ subunits. The conformational switch regions are important for the interaction of ␣ subunits with effectors (9), ␤␥ (10, 11), and RGS (regulators of G protein signaling) regulators of G proteins (12). Most likely, they play a role in receptor-mediated activation as well.
We previously identified a cluster of four switch III residues 2 in ␣ s at the interface between the GTPase and helical domains in which substitutions with ␣ i2 homologs in the mutant construct N254D/M255L/I257L/R258A␣ s decreased receptor-mediated activation of adenylyl cyclase in transiently transfected cells (13). The activation defect caused by substituting ␣ i2 residues for these ␣ s residues was corrected by replacing the helical domain of ␣ s with that of ␣ i2 in a chimera, ␣ sis , in which ␣ i2 homologs were substituted for ␣ s residues 62-235, extending from the end of the ␣1 helix to the end of the ␣2 helix (13). Thus, matching ␣ i2 residues on both sides of the domain interface of ␣ s restored receptor-initiated activation.
We now report a detailed analysis of the activation defects caused by these switch III substitutions and of a mutation that replaces a nearby helical domain residue, Asn 167 , with the homologous ␣ i2 residue, arginine. Measurements in stably transfected cells of isoproterenol binding to the ␤-adrenergic receptor, and the time course and dose-dependence of adenylyl cyclase stimulation by the hydrolysis-resistant GTP analog, GTP␥S, in the presence and absence of isoproterenol indicate that the mutations increase the affinity of ␣ s for the ␤-adrenergic receptor, decrease the rate of receptor-mediated activation, and block receptor-stimulated increases in GTP␥S affinity. Additional mutational analysis refines the nature of the interdomain interactions that play a role in receptor-mediated activation by demonstrating that of the switch III substitutions, R258A alone causes a defect in receptor-mediated activation, that this defect is corrected when the helical domain of ␣ s is replaced with that of ␣ i2 , and that the defect caused by the N167R substitution is corrected when combined with the N254D substitution but not the R258A substitution. These results suggest that interdomain interactions are involved in the transmission of signals between the receptor and the guanine nucleotide binding pocket.

EXPERIMENTAL PROCEDURES
Construction of ␣ Subunit Mutants-␣ s mutant constructs were generated from rat ␣ s cDNA (14). Chimeric ␣ subunits were constructed from rat ␣ s cDNA and mouse ␣ i2 cDNA (15). Subcloning and mutagenesis procedures were verified by restriction enzyme analysis and DNA sequencing. All ␣ subunit constructs produced in this study contain an epitope, referred to as the EE epitope (16), which was generated by mutating ␣ s residues DYVPSD (189 -194) to EYMPTE and ␣ i2 residues SDYIPTQ (166 -172) to EEYMPTE (single letter amino acid code; mutated residues are underlined). This epitope does not affect the ability of ␣ s to activate adenylyl cyclase in response to stimulation by the ␤-adrenergic receptor (17).
The ␣ s cDNA was subcloned into the expression vector, pcDNA I/Amp (Invitrogen) as a HindIII fragment. N167R␣ s , N167A␣ s , N254D␣ s , M255L␣ s , I257L␣ s , R258A␣ s , and N254D/M255L/I257L/R258A␣ s were produced by oligonucleotide-directed in vitro mutagenesis (18) using the Bio-Rad Muta-Gene kit. N167R/N254D␣ s and N167R/R258A␣ s were produced by ligating Alwn I fragments containing either the N254D or the R258A mutations into N167R␣ s in place of the analogous fragment to produce an ␣ s cDNA containing both mutations. Construction of the ␣ sis chimera, in which ␣ s residues 62-235 are replaced by the homologous ␣ i2 residues, has been described elsewhere (13). R258A␣ sis was produced by ligating a BamHI fragment containing the R258A mutation into ␣ sis in place of the analogous fragment.
Receptor-independent cAMP accumulation in transiently transfected cells was measured after introducing a second mutation (RC) that substitutes cysteine for the arginine at position 201 and causes constitutive activation by decreasing GTPase activity (19). ␣ s RC versions of N254D␣ s , M255L␣ s , I257L␣ s , and R258A␣ s were produced by ligating BamHI fragments containing the mutations into ␣ s RC in place of the analogous fragment. ␣ s RC versions of N167R␣ s and N167A␣ s were produced by ligating Alwn I fragments containing the RC mutation into N167R␣ s and N167A␣ s , respectively, in place of the analogous fragment. ␣ s RC versions of N167R/N254D␣ s and N167R/R258A␣ s were produced by ligating Alwn I fragments containing the RC mutation and either the N254D or the R258A mutation into N167R␣ s in place of the analogous fragment. The ␣ sis RC version of R258A␣ sis was produced by ligating a BamHI fragment containing the R258A mutation into ␣ sis RC in place of the analogous fragment.
Preparation of Stable Cell Lines-␣ s constructs were subcloned as HindIII fragments into the retroviral vector pMV7 (20) and then stably expressed as described (21), in a subclone of cyc Ϫ S49 lymphoma cells, cyc Ϫ kin Ϫ (22), in which cAMP-dependent protein kinase is inactivated. Single colonies containing the pMV7 vector were obtained using limiting dilution in microtiter wells and selection in G418 (1 mg/ml). Clones expressing ␣ s constructs were identified by immunoblotting with the anti-EE monoclonal antibody. Cell membranes were prepared after nitrogen cavitation as described (23).
Immunoblots-25 g of membrane proteins were resolved by SDSpolyacrylamide electrophoresis (10%), transferred to nitrocellulose, and probed with a monoclonal antibody to the EE epitope (16). The antigenantibody complexes were detected using an anti-mouse horseradish peroxidase-linked antibody according to the ECL Western blotting protocol (Amersham Pharmacia Biotech).
Adenylyl Cyclase Assay-Conversion of [␣ 32 P]ATP to [ 32 P]cAMP in the presence of various activators was measured as described (23). Membranes were incubated at 30°C. Reactions shown in Figs. 1 and 4 were preincubated for 5 min in the absence of [␣ 32 P]ATP and then incubated for 30 min. For the time courses shown in Fig. 3, membranes were preincubated in the absence of [␣ 32 P]ATP and activators for 6 min. At time ϭ 0, [␣ 32 P]ATP and either GTP␥S or GTP␥S and isoproterenol were added and aliquots were removed at the indicated times for cAMP determination. To determine EC 50 values for stimulation of adenylyl cyclase by GTP␥S shown in Fig. 4, the observed adenylyl cyclase activity was fitted to the equation, where X is the concentration of GTP␥S, Y is the observed adenylyl cyclase activity, a is the adenylyl cyclase activity observed in the absence of GTP␥S, b is the maximum observed adenylyl cyclase activity, c is the half-maximal effective concentration (EC 50 ) of GTP␥S, and d is the slope factor.
Receptor Binding Assay-Membranes were incubated with 75 pM [ 125 I]ICYP in competition with a range of concentrations of isoproterenol (10 Ϫ11 to 10 Ϫ3 M) in the presence or absence of 300 M GTP for 1 h at 30°C as described (24). At the end of this time, the membranes were diluted and washed on Whatman GF/C filters, and bound [ 125 I]ICYP was measured. The experimental data were analyzed for competition at two sites by nonlinear least-squares curve fitting as described (24). K L and K H , the low and high affinity dissociation constants, respectively, were assumed to be the same in the presence and absence of GTP. When K L and K H were allowed to vary in the two conditions, improved fits to the data were obtained. Therefore, the two-state model may be an oversimplification of receptor behavior, as has been suggested (25).
cAMP Accumulation Assay in Transiently Transfected cyc Ϫ S49 Lymphoma Cells-␣ subunit constructs were introduced by electroporation into a subclone of cyc Ϫ S49 lymphoma cells (26) that stably expresses Simian virus 40 large T antigen, and cAMP accumulation was measured after labeling with [ 3 H]adenine as described (13). Nucleotides were separated on ion-exchange columns (27), and cAMP accumulation was expressed as Receptor-independent cAMP accumulation was determined by measuring basal cAMP levels in cells transfected with the ␣ s RC versions of the mutant constructs.

Mutations at the Domain Interface of ␣ s Decrease Activation by the ␤-Adrenergic
Receptor and by Aluminum Fluoride-The ability of N254D/M255L/I257L/R258A␣ s to be activated by the ␤-adrenergic receptor and by AlF 4 Ϫ , which mimics the ␥-phosphate of GTP, was measured after expression in cyc Ϫ S49 lymphoma cells (26), which lack endogenous ␣ s (28). At equal expression levels (Fig. 1A), adenylyl cyclase activity stimulated by both isoproterenol and by AlF 4 Ϫ was reduced by 80% in membranes of cells expressing N254D/M255L/I257L/R258A␣s compared with membranes of ␣ s -expressing cells (Fig. 1B). Stimulation by the hydrolysis-resistant GTP analog, GTP␥S, not only was intact, but increased by almost 2-fold.
Because the activation defect of N254D/M255L/I257L/ R258A␣ s , as assessed in transiently transfected cells, was suppressed when the helical domain consisted of ␣ i2 residues (13), we analyzed the x-ray crystal structures of ␣ subunits to identify nearby helical domain residues that might be responsible for the conditional defect of these switch III substitutions. Comparison of the GTP␥S-and GDP-bound ␣ subunit structures (5-8) reveals slight changes in the positions of helical domain residues in the ␣D/␣E loop, which is in contact with switch III in the GTP␥S-bound form. In ␣ s , the only residue in this loop that is close to any of the four residues and differs in the sequences of ␣ s and ␣ i2 is Asn 167 . Mutation of Asn 167 to the homologous ␣ i2 residue to produce N167R␣ s decreased both isoproterenol-stimulated and AlF 4 Ϫ -stimulated adenylyl cyclase activity by 60% in membranes of cells expressing equal amounts of protein compared with membranes of ␣ s -expressing cells (Fig. 1). As with N254D/M255L/I257L/R258A␣ s , stimulation by GTP␥S was increased ϳ2-fold ( Fig. 1).
Mutations at the Domain Interface of ␣ s Increase the Apparent Affinity of ␣ s for the ␤-Adrenergic Receptor-Because N254D/M255L/I257L/R258A␣ s and N167R␣ s exhibited decreased receptor-mediated activation, we used a competitive binding assay to determine whether these mutant ␣ subunits exhibit alterations in binding to the ␤-adrenergic receptor. This assay measures an ␣ s -dependent increase in the affinity of the ␤-adrenergic receptor for the agonist, isoproterenol (24,29), which occurs in the absence of bound guanine nucleotide. The high affinity hormone binding state of the receptor is thought to reflect its interaction with G s in the nucleotide-free state. In the presence of GTP, receptors in membranes of ␣ s -expressing cells were predominantly in the low affinity state ( Fig. 2A). In the absence of GTP, ␣ s caused the appearance of high affinity binding sites for isoproterenol on the receptor ( Fig. 2A). Like ␣ s , both N254D/M255L/I257L/R258A␣ s and N167R␣ s increased the affinity of the ␤-adrenergic receptor for isoproterenol in the absence of GTP compared with in its presence (Fig. 2, B and C). However, in membranes of cells expressing these constructs, the affinity of the receptor for isoproterenol in both the presence and absence of GTP was greater than in membranes from ␣ s -expressing cells, due to decreases in K L and K H , the low and high affinity dissociation constants, respectively, as well as increases in the percentage of receptors in the high affinity form, % R H . This increase in hormone-receptor binding was particularly striking in cells expressing N167R␣ s , in which 50% of the receptors were in the high affinity form in the presence of GTP. The simplest explanation of the increased amount of hormone-receptor binding observed in the presence of N254D/ M255L/I257L/R258A␣ s and N167R␣ s is that these mutant ␣ subunits increase the affinity of G s for the receptor in both the nucleotide-bound and -free states.
Mutations at the Domain Interface of ␣ s Decrease the Rate of Activation by the ␤-Adrenergic Receptor-Because dissociation of G s from the activated receptor must precede adenylyl cyclase activation, we investigated whether N254D/M255L/I257L/ R258A␣ s and N167R␣ s exhibited altered rates of receptor-mediated activation. To estimate relative rates of receptor-mediated activation, we determined the effects of isoproterenol on the time courses of adenylyl cyclase activation by GTP␥S. In membranes of cells expressing ␣ s , GTP␥S activated adenylyl cyclase with a time lag that was greatly reduced by isoproterenol (Fig. 3A). This decreased time lag reflects receptor-stimulated increases in the rates of GDP dissociation and GTP␥S binding. In the absence of isoproterenol, GTP␥S activated ad-  (40). K L and K H are the low and high affinity dissociation constants, respectively, and % R H is the percentage of receptors in the high affinity form. K L and K H were assumed to be the same in the presence and absence of GTP. In B and C, the binding curves for membranes from ␣ s -expressing cells, from A, are redrawn as dotted lines. Similar results for each construct were obtained in two additional experiments using a second independent clone of stably transfected cyc Ϫ cells. The binding assay and data analysis were performed as described (24). enylyl cyclase in membranes containing N254D/M255L/I257L/ R258A␣ s or N167R␣ s with somewhat shorter time lags than in ␣ s membranes (Fig. 3, B and C). Isoproterenol increased the activation rates of these mutants, but not to the same extent as for ␣ s . Thus, in the presence of isoproterenol, the time lags of the mutants were longer than that of ␣ s (Fig. 3, B and C). N167R␣ s , which caused the appearance of the greatest amount of high affinity binding to the receptor (Fig. 2C), exhibited the longest time lag in the presence of isoproterenol. Thus, N254D/ M255L/I257L/R258A␣ s and N167R␣ s exhibit decreased rates of receptor-mediated activation, which could reflect decreased rates of GTP-dependent dissociation from receptors. Alternatively, or in addition, decreased rates of receptor-mediated activation could be due to defects in receptor-stimulated GTP binding, which we investigated as described below.
Mutations at the Domain Interface of ␣ s Disrupt the Ability of the ␤-Adrenergic Receptor to Promote Binding of GTP␥S-Receptors stimulate guanine nucleotide exchange on G proteins by increasing the rate of GDP release and by causing a preference for GTP compared with GDP (3,4). For ␣ s , this results in an approximately 8-fold decrease in the half-maximal effective concentration (EC 50 ) for GTP␥S stimulation of adenylyl cyclase in the presence of isoproterenol compared with in its absence (Fig. 4A). In the absence of isoproterenol, N254D/M255L/ I257L/R258A␣ s and N167R␣ s exhibited EC 50 values for GTP␥S stimulation of adenylyl cyclase that were slightly lower than that of ␣ s (Fig. 4, B and C). However, these EC 50 values were unchanged by isoproterenol (Fig. 4, B and C) so that in the presence of isoproterenol, their apparent affinities for GTP␥S were less than that of ␣ s . Thus, although isoproterenol increases the rate at which these mutant ␣ subunits exchange nucleotide (Fig. 3), it does not increase their apparent affinities for GTP.
Localization of a Single Switch III Residue on the GTPase Side of the Domain Interface That Is Important for Receptormediated Activation of ␣ s -We individually tested each of the four switch III residues that were mutated in N254D/M255L/ I257L/R258A␣ s to determine their roles in receptor-mediated activation. Receptor-dependent stimulation of cAMP synthesis was measured in transiently transfected cyc Ϫ S49 lymphoma cells. The only substitution that decreased receptor-mediated activation was R258A (Fig. 5A). Receptor-independent cAMP accumulation was also measured after introducing a second mutation (the RC mutation) that substitutes cysteine for the arginine at position 201 (19). ␣ s RC has decreased GTPase activity and is constitutively activated. R258A␣ s RC produced receptor-independent cAMP accumulation similar to that of ␣ s RC, indicating that, as is the case for N254D/M255L/I257L/ R258A␣ s , R258A␣ s can activate adenylyl cyclase when it is in the GTP-bound form.
Complementation of the Activation Defect of R258A␣ s -Because the activation defect of N254D/M255L/I257L/R258A␣ s was corrected by replacing the helical domain of ␣ s with that of ␣ i2 in a chimera, ␣ sis , in which ␣ i2 homologs were substituted for ␣ s residues 62-235, (Fig. 6) (13), we investigated whether introducing the single homolog substitution (R258A) responsible for the defect of N254D/M255L/I257L/R258A␣ s into ␣ sis would result in normal activation properties. R258A␣ sis exhibited activation properties similar to those of ␣ s rather than those of R258A␣ s (Fig. 5A). Thus, the defect produced by the R258A substitution appears to be due to an alteration in interactions with ␣ s residue(s) in the helical domain.
Combining the N167R and R258A Substitutions Results in an Additive Defect in Receptor-mediated Activation-We investigated the effect of combining the substitutions on each side of the domain interface, N167R and R258A, that caused significant decreases in receptor-mediated activation. Although Asn 167 and Arg 258 are both at the domain interface, they are not close enough to make contact in the x-ray crystal structures of ␣ subunits (see Fig. 6). N167R/R258A␣ s exhibits a more severe activation defect (Fig. 5B) than either N167R␣ s (Fig. 5C) or R258A␣ s (Fig. 5A) does, although receptor-independent activation by N167R/R258A␣ s RC is normal. Because the defects of the N167R and R258A substitutions are additive, the contributions of these substitutions to defects in receptor-mediated activation are independent. Furthermore, some other residue(s) in the helical domain other than the ␣ i2 homolog of Asn 167 must be responsible for the suppression of the R258A defect in R258A␣ sis .
Combining the N167R and N254D Substitutions Corrects the Defect of the N167R Substitution-According to the x-ray crystal structures of ␣ subunits, Asn 167 is close to Asn 254 (see Fig.  6). Therefore, we hypothesized that the N167R substitution might cause a conditional defect, depending on the identity of the residue at position 254. According to this hypothesis, replacing Asn 254 with aspartate should correct the defect caused by the N167R mutation. Indeed, we found that an ␣ subunit with both substitutions, N167R/N254D␣ s , exhibits activation properties similar to those of ␣ s (Fig. 5B). Thus, the N254D substitution, which on its own does not disrupt receptor-mediated activation, corrects the activation defect caused by the N167R substitution.

Substitution of Asn 167 by Alanine Does Not Cause a Defect in
Receptor-mediated Activation of ␣ s -To further investigate the mechanism by which the N167R substitution causes a defect in receptor-mediated activation, we determined the effect of mutating Asn 167 to alanine. Alanine substitutions eliminate the side chain beyond the ␤ carbon but generally do not alter the main chain conformation or impose significant electrostatic or steric effects (30). Therefore, if the activation defect resulting from the N167R substitution is due to a steric or electrostatic incompatibility with Asn 254 , then alanine substitution might not cause an activation defect. However, if the N167R substitution removes a favorable interaction between Asn 167 and Asn 254 , then alanine substitution should also cause a defect. Because N167A␣ s exhibited normal receptor-stimulated cAMP production (Fig. 5C), the activation defect of N167R␣ s appears to be due to a steric or electrostatic incompatibility that is reversed by the N254D substitution. cAMP accumulation was measured in cyc Ϫ cells transiently expressing the indicated ␣ s constructs. cAMP values from unstimulated cells and from cells stimulated with 0.1 mM isoproterenol are dark gray and light gray, respectively. Cells were electroporated with 30 g of each construct, with the following exceptions: N254D␣ s , 10 g; N167R/N254D␣ s , 10 g; N254D/M255L/I257L/R258A␣ s , 20 g; and N167R/R258A␣ s , 60 g. At these plasmid doses, equivalent amounts of receptor-independent cAMP accumulation (mean, 7.3; S.D., 1.5) were produced by versions of the constructs in which arginine 201 was replaced by cysteine, which causes constitutive activation by decreasing GTPase activity (19). cAMP levels in [ 3 H]adenine-labeled cells were determined as described under "Experimental Procedures." Within each panel, the values for ␣ s and the vector alone represent the activities obtained on the days that the indicated mutants were tested. Day-to-day variation in cAMP accumulation in transiently transfected cells accounts for the differences in these values in between A, B, and C. All values represent the means Ϯ S.E. of at least three independent experiments.

DISCUSSION
Our analysis of ␣ s mutants with substitutions at the interface of the GTPase and helical domains suggests that interdomain interactions play a role in the bi-directional transmission of signals between receptors and the nucleotide binding site. Interaction between activated receptors and G proteins promotes GTP binding by accelerating GDP release and increasing the relative affinity for GTP compared with GDP (3, 4). Conversely, nucleotide binding decreases the affinity of G proteins for receptors (24,29). Substitution of the homologous ␣ i2 residues for four ␣ s residues (Asn 254 , Met 255 , Ile 257 , and Arg 258 ) in switch III of the GTPase domain or of one nearby helical do-main residue (Asn 167 ) in the ␣D/␣E loop causes defects in both directions of this communication process. Signal transmission from the receptor to the nucleotide binding site is defective in that the affinities of these ␣ s mutants for GTP␥S are unchanged by isoproterenol. Conversely, altered communication between the guanine nucleotide binding pocket and the receptor binding site(s) is demonstrated by high affinity hormonereceptor binding in the presence of 300 M GTP.
Contacts between the ␣D/␣E loop in the helical domain and three regions of the GTPase domain have been implicated as playing a role in receptor-mediated activation. In the heterotrimer-based ␣ subunit model shown in Fig. 6, the helical FIG. 6. Mapping of ␣ s residues at the domain interface that are important for receptor-mediated activation onto the structure of an ␣␤␥ heterotrimer. View of the ␣ subunit based on the x-ray crystal structure of an ␣ t /␣ i1 chimera complexed with ␤ t ␥ t (10). The ␤␥ subunits have been omitted for clarity. The helical domain is to the left of the GDP, which is yellow. The GTPase domain is to the right. The spheres are centered on the ␣-carbons of the corresponding residues and are numbered according to the ␣ s sequence. Asn 167 and Arg 258 , in which substitutions with ␣ i2 homologs disrupt receptor-mediated activation, are red spheres. Met 255 and Ile 257 , which can be mutated without impairing receptormediated activation, are green spheres. Substitution of the homologous ␣ i2 residue for Asn 254 (magenta sphere) does not impair receptor-mediated activation, but corrects the defect caused by replacing Asn 167 with the homologous ␣ i2 residue. Asp 173 and Lys 293 (dark blue spheres) form a salt bridge that is required for receptor-mediated activation (31). The amino-terminal portion of the ␣ sis chimera, consisting of ␣ s residues (light blue), extends from the amino terminus to the end of ␣1. The middle portion of the chimera, consisting of ␣ i2 residues (pink), extends from the ␣ s /␣ i2 junction (labeled s/i) to the ␣ i2 /␣ s junction (labeled i/s) at the end of ␣2. The carboxyl-terminal portion of the chimera consists of ␣ s residues (light blue). Switches (Sw) I-III are gold. This figure was drawn using MidasPlus, developed by the Computer Graphics Laboratory at the University of California, San Francisco.  (9) is on the right and that of ␣ i1 from the x-ray crystal structure of ␣ i1 ⅐GTP␥S (6) is on the left. The pink and cyan tubes trace the overall fold of the helical and GTPase domains, respectively, at this interface, as in Fig. 6. The sequences of ␣ i1 and ␣ i2 are identical in the regions shown. The side-chains of residues that differ in the sequences of ␣ s and ␣ i2 are shown. Oxygens are red, nitrogens are blue, and sulfurs are yellow. The dotted yellow line indicates the hydrogen bond between Asn 167 and Asn 254 in ␣ s and between the corresponding residues in ␣ i2 , Arg 145 and Asp 232 . This figure was drawn using MidasPlus. domain side of the interface "above" the GDP consists of the ␣D/␣E loop. Moving up from the GDP toward the top of the ␣ subunit, the corresponding GTPase side of the interface consists of the ␤5/␣G, ␤4/␣3, and ␣G/␣4 loops. Closest to the GDP, a salt bridge interaction between Asp 173 in the carboxyl-terminal portion of the ␣D/␣E loop and Lys 293 in the ␤5/␣G loop (Fig.  6, dark blue) is required for activation by the ␤-adrenergic receptor and by AlF 4 Ϫ , but not by GTP␥S (31). These residues are highly conserved among ␣ subunits, and Lys 293 is located in the NKXD motif, which is important for GTP binding by monomeric GTPases. Mutation of Asp 173 increases GTP affinity, consistent with the idea that the mutation "frees" Lys 293 from Asp 173 to interact with GTP, whereas mutation of Lys 293 decreases GTP affinity. Because Asp 173 and Lys 293 are adjacent to the bound guanine nucleotide, they are more directly involved in regulating guanine nucleotide binding than the residues mutated in our study are. The effects of mutating Asp 173 and Lys 293 on receptor affinity and receptor-dependent changes in GTP affinity have not been determined.
Arg 258 (Fig. 6, red) is located further up, in the ␤4/␣3 loop, which includes switch III. Interestingly, a mutation that substitutes tryptophan for Arg 258 was found in a patient with Albright hereditary osteodystrophy, and R258W␣ s exhibited more severe defects than did R258A␣ s . 3 The defect caused by the R258A substitution is suppressed (Fig. 5A) by substituting the entire helical domain of ␣ i2 for that of ␣ s in the ␣ sis chimera (Fig. 6), but we have not identified the helical domain residue(s) responsible. In addition to Asn 167 , there are two other ␣ s residues in the ␣D/␣E loop, Ile 172 and Cys 174 , that differ in the sequences of ␣ s and ␣ i2 . Cys 174 is not close to Arg 258 when modeled onto the heterotrimeric G protein structures (10,11) or in the structure of ␣ s ⅐GTP␥S (32). However, in the latter structure, the side chains of Ile 172 and Arg 258 are within ϳ4 Å of each other (see Fig. 7).
Asn 167 (Fig. 6, red) is located further away from the GDP in the ␣D/␣E loop. The activation defect caused by replacing Asn 167 with its ␣ i2 homolog (arginine) is corrected (Fig. 5B) by simultaneously replacing Asn 254 (Fig. 6, magenta) in the ␤4/␣3 loop with its ␣ i2 homolog (aspartate). In the ␣ subunit structures, the corresponding residues are hydrogen bonded to each other via the side chain of the residue corresponding to Asn 167 and either the side chain (in ␣ t (5, 7) and an ␣ t /␣ i1 chimera complexed with ␤ t ␥ t (10)) or the backbone carbonyl (in ␣ s (9,32) and ␣ i1 (6)) of the residue corresponding to Asn 254 (see Fig. 7). The N254D substitution in ␣ s might correct the defect caused by the N167R substitution via a charge neutralization mechanism. This idea is supported by the fact that all ␣ subunits with an arginine at the position corresponding to Asn 167 (␣ q , ␣ 11 , ␣ 14 , ␣ 15 , ␣ 16 , and ␣ o ) have an aspartate at the position corresponding to Asn 254 and by the observation (Fig. 5C) that the N167A substitution in ␣ s leaves receptor-mediated activation intact. It is not surprising that the activation defect of N167R/ R258A␣ s is worse than those of N167R␣ s and R258A␣ s (Fig. 5), because these residues are not within contact distance. Because the two mutations cause additive defects, the two residues also do not appear to influence each other through electrostatic or steric effects (33).
We previously found that substitution of ␣ i2 residues for ␣ s residues 304, 305, and 307-311 in the ␣G/␣4 loop (furthest from the nucleotide on the GTPase side of the interface in Fig.  6) disrupts receptor-mediated activation in the context of ␣ s but not ␣ sis (13). Of the mutated ␣ s residues, only Lys 305 and Tyr 311 are close to the interface in the structure of ␣ s ⅐GTP␥S (32).
Although interactions between residues in switch III and the ␣D/␣E loop are important for receptor-mediated activation, the known receptor binding sites of ␣ s , the carboxyl terminus of ␣5 (13,21,34) and possibly the ␣4/␤6 loop (34), are not near this interface. ␣ subunits bind to ␤␥, which is required for receptormediated activation, via switches I and II and the amino terminus (10,11), which are also not near this interface. Thus, receptors initiate activating signals at a significant distance from the domain interface, possibly via an interaction between switches II and III. N254D/M255L/I257L/R258A␣ s and N167R␣ s exhibit two characteristics in the absence of receptor stimulation that are not normal and that resemble those of wild-type ␣ s upon activation by hormone-bound receptors. They exhibit slightly elevated basal rates of activation (Fig. 3) and somewhat increased basal affinities for GTP␥S (Fig. 4). The basal activation rates of these ␣ s mutants, which reflect basal nucleotide exchange rates, are not nearly as elevated as in an ␣ s mutant, A366S␣ s , which is both thermolabile and constitutive activated (35). However, increased rates of basal GDP dissociation in N254D/ M255L/I257L/R258A␣ s and N167R␣ s would account for their observed defects in activation by aluminum fluoride, which requires the presence of bound GDP.
The defects in guanine nucleotide and receptor binding of N254D/M255L/I257L/R258A␣ s and N167R␣ s may be interrelated. For instance, decreased receptor-stimulated GTP binding would stabilize the high affinity hormone-receptor-G protein complex, which forms when the G protein is in the nucleotide-free state. Conversely, higher affinity receptor-G protein interactions could decrease nucleotide binding, because activated receptors can cause dissociation of both GDP and GTP analogs (36,37). However, although there are other reported ␣ s mutants with defects in GTP binding and receptormediated activation, there is no precedent for an associated increase in receptor affinity. R231H␣ s , containing a mutation in the ␣2 helix (38), and S54N␣ s , containing a substitution in the ␣1 helix (39), exhibit impaired activation by receptors and AlF 4 Ϫ but can be activated by GTP␥S. The affinities of these ␣ s mutants for GTP are decreased upon receptor stimulation. The affinity of R231H␣ s for the receptor is normal, and the affinities of S54N␣ s and of A366S␣ s , which exhibits accelerated GDP release (35), were not determined. The defect of S54N␣ s appears to be due to altered interactions with the bound Mg 2ϩ , with which Ser 54 interacts. Thus, the activation defects of these ␣ s mutants are distinct from those of N254D/M255L/I257L/ R258A␣ s and N167R␣ s .
Further studies will be required to elucidate how the ␣ subunit domain interface mediates communication between the residues responsible for receptor binding and the guanine nucleotide binding pocket. This type of regulation appears to be unique to heterotrimeric G proteins, which interact with seventransmembrane-spanning receptors, compared with monomeric GTPases, which lack switch III and the helical domain and which utilize different exchange factors.