Pivotal Role of Extended Linker 2 in the Activation of Gα by G Protein-coupled Receptor*

Background: G protein-coupled receptors mainly signal through heterotrimeric G-proteins. Results: We have demonstrated that mutations in the extended Linker 2 impaired the activation of Gαs by β-adrenergic receptors. Conclusion: We have proposed a potential novel conduit from β-adrenergic receptors to the helical domain of Gαs subunit via the extended Linker 2. Significance: Our systematic mutagenesis studies provide insights into the activation mechanism of G-proteins by receptors. G protein-coupled receptors (GPCRs) relay extracellular signals mainly to heterotrimeric G-proteins (Gαβγ) and they are the most successful drug targets. The mechanisms of G-protein activation by GPCRs are not well understood. Previous studies have revealed a signal relay route from a GPCR via the C-terminal α5-helix of Gα to the guanine nucleotide-binding pocket. Recent structural and biophysical studies uncover a role for the opening or rotating of the α-helical domain of Gα during the activation of Gα by a GPCR. Here we show that β-adrenergic receptors activate eight Gαs mutant proteins (from a screen of 66 Gαs mutants) that are unable to bind Gβγ subunits in cells. Five of these eight mutants are in the αF/Linker 2/β2 hinge region (extended Linker 2) that connects the Ras-like GTPase domain and the α-helical domain of Gαs. This extended Linker 2 is the target site of a natural product inhibitor of Gq. Our data show that the extended Linker 2 is critical for Gα activation by GPCRs. We propose that a GPCR via its intracellular loop 2 directly interacts with the β2/β3 loop of Gα to communicate to Linker 2, resulting in the opening and closing of the α-helical domain and the release of GDP during G-protein activation.

A structurally diverse repertoire of ligands, from photons to many hormones and neurotransmitters, activate G proteincoupled receptors (GPCRs) 4 to elicit their physiological functions (1,2). GPCRs comprise a large and diverse superfamily, and family members have been identified in organisms as evolutionarily distant as yeast and human. Heterotrimeric G-proteins (G␣␤␥) directly relay the signals from GPCRs (3)(4)(5). These G-proteins are composed of ␣, ␤, and ␥ subunits. The ␤ and ␥ subunits are tightly associated and can be regarded as one functional unit. G-proteins function as molecular binary switches with their biological activity determined by the bound nucleotide (3)(4)(5). Upon ligand-binding, GPCRs increase the exchange of GDP bound on the G␣ subunit with GTP, in the presence of G␤␥ subunits. This leads to the dissociation of the G␣ subunit from the G␤␥ dimer resulting in two functional subunits (G␣ and G␤␥). Both G␣ and G␤␥ subunits signal to various cellular pathways.
Over the past 30 years, great progress has been made in understanding the mechanisms by which heterotrimeric G-proteins regulate their downstream targets (6,7). Recently a series of crystal structures of GPCRs in the inactive and active states, bound with antagonists, inverse agonists, or agonists, have elucidated the structural basis for the modulation and activation of GPCRs by ligands (1,8,9). A crystal structure of the complex of ␤ 2 -adrenergic receptor and G s has revealed the structural changes in ␤ 2 -adrenergic receptor and G s , the interacting regions, and residues between a GPCR and a G-protein (10). However, the molecular mechanisms by which GPCRs activate G-proteins are still not completely understood (11,12).
The structure of G␣ subunits consists of two domains: a Raslike GTPase domain and an ␣-helical domain (6) (Fig. 1, A and  B). These two domains are linked by Linker 1 and Linker 2 (Fig.  1B). Between these two domains lies a deep cleft within which GDP or GTP is tightly bound (Fig. 1, A and B). The nucleotide is essentially occluded from the bulk solvent, leading to a proposal that the helical domain is the inhibitory barrier and provides the regulatory entry point by GPCRs or G␤␥ subunits (13)(14)(15)(16).
One of the major remaining problems in the biology of GPCR/G-protein signaling is to experimentally demonstrate whether GPCRs or G␤␥ subunits play the catalytic exchange role in G␣ protein activation (17). In theory, the question could be straightforwardly answered with purified proteins of GPCRs, G␣ and G␤␥ subunits. However, purified GPCRs had no guanine-nucleotide exchange effect on G␣ in the absence of G␤␥ subunits (18,19). A GPCR could only activate G␣ in the presence of G␤␥ (18,19), leading some to believe that GPCRs were not the real catalysts, although they were required to initiate the activation event from agonist binding. Yet, purified G␤␥ subunits alone also could not accelerate the guanine nucleotide exchange on G␣ subunits (3,20). In fact, G␤␥ subunits inhibit the basal nucleotide exchange activity of G␣ sub-units and behave as a guanine-nucleotide dissociation inhibitor (20). When the structure of G␤ was revealed to be similar to RCC1, a guanine-nucleotide exchange factor for the small GTPase Ran, G␤␥ was proposed to be the real guanine-nucleotide exchange factor for G␣, whereas GPCRs served as a trigger (15,21). Therefore, it remains a fundamental unsolved question: which one, a GPCR and/or G␤␥ subunit, has the ability to catalyze the nucleotide exchange on G␣.
We reasoned that if GPCRs or G␤␥ subunits are the nucleotide exchangers, we should be able to find mutants of GPCRs or G␤␥ subunits that would accelerate the nucleotide exchange on G␣ subunits in the absence of G␤␥ or GPCRs, respectively. Alternatively, we might be able to find mutant G␣ subunits that could be activated by GPCRs alone or by G␤␥ alone. Here we describe our finding of a direct activation of some G␣ s mutant proteins by ␤-adrenergic receptors without G␤␥ subunits. Although our data do not exclude a possible additional catalytic role for G␤␥, it clearly demonstrates that GPCRs by themselves have the catalytic ability to activate G␣ subunits. Furthermore, from this systematic study, a pivotal role for the ␣F/Linker 2/␤2 region (extended Linker 2) in G␣ activation has been uncovered. Therefore, we propose that the opening and closing of the two domains (the Ras-like GTPase domain and the helical domain) with Linker 2 as a pivot provide one of the mechanisms for G␣ activation by GPCRs.

EXPERIMENTAL PROCEDURES
G-protein Purification-G␣ s wild-type and mutant proteins were purified from Escherichia coli. The pGEX-6P-G␣ s plasmid was transformed into bacteria strain BL21(DE3). One liter of bacterial culture was grown at room temperature until the absorbance at 600 nm was ϳ1. G-protein expression was induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside for 18 h at room temperature. The bacterial pellet was resuspended in lysis buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1% Triton X-100, 0.1 mg/ml of lysozyme, and 0.2 mM PMSF) and incubated on ice for 30 min. After sonication, the lysate was spun down at 10,000 ϫ g for 120 min at 4°C. Glutathioneagarose resin (0.5 ml, from Sigma) was added to the supernatant after pre-equilibration of the resin with lysis buffer. The mixture was gently agitated at 4°C for 3 h. After washing 3 times with 10 ml of washing buffer (50 mM Tris, pH 8.0, 100 mM NaCl, and 0.2 mM PMSF), GST-tagged G␣ s proteins were eluted with 0.5 ml of elution buffer (50 mM Tris, pH 8.0, 100 mM NaCl, 1 mM EDTA, 1 mM DTT, and 10% glycerol). preScission protease (Amersham Biosciences) was used to cleave GST off at 4°C overnight.
G␤␥ Proteins Were Purified from Insect Hi5 Cells-One liter of Hi5 cell pellet was resuspended into 50 ml of lysis buffer (50 mM Tris, pH 8.0, 1 mM EDTA, and protease inhibitors: 10 g/ml of leupeptin, 1 g/ml of pepstatin A, 1 mM benzamidine, and 0.2 mM PMSF). After sonication, the lysate was spun down at 150,000 ϫ g for 90 min at 4°C. The membrane pellet was resuspended in 50 ml of lysis buffer. After homogenization, the lysate was centrifuged again. The final pellet was resuspended in 50 ml of extraction buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 2% dodecyl-␤-D-maltoside (DM) and protease inhibitors). After centrifugation at 10,000 ϫ g for 120 min at 4°C, 1 ml of nickel-nitrilotriacetic acid-agarose beads pre-equilibrated with extraction buffer was added to the supernatant. The mixture was gently agitated overnight at 4°C. After washing 3 times with 10 ml of washing buffer (50 mM Tris, pH 8.0, 100 mM NaCl, 5 mM imidazole, and 0.2 mM PMSF), G␤␥ was eluted with 10 ml of elution buffer (50 mM Tris, pH 8.0, 100 mM NaCl, 10 mM EDTA, and 200 mM imidazole).
G␤␥ Binding Assay (with Purified G␤ 1 ␥ 2 Proteins)-Ten g of G␣ s protein (with an N-terminal GST tag, attached to glutathione beads) were incubated in 100 l of binding buffer (20 mM Hepes, pH 8.0, 2 mM GDP, 1 mM DTT, 150 mM NaCl, and 0.02% DM) with 100 nM G␤ 1 ␥ 2 proteins overnight at 4°C. After centrifugation, beads were washed three times with binding buffer and then eluted with binding buffer containing 10 mM reduced glutathione. After SDS-PAGE, Western blots were performed with anti-G␣ s and anti-G ␤ antibodies (Santa Cruz Biotechnology, Inc.).
G␤␥ Binding Assay (with Membrane Preparations)-Membrane preparations were made from 293T cells. After solubilization, the supernatant (in 20 mM Hepes, pH 8.0, 2 mM GDP, 1 mM DTT, 150 mM NaCl, 1% DM, and protease inhibitors) was pre-cleared with glutathione beads. 500 l of supernatant was mixed with 10 g of G␣ s protein (with an N-terminal GST tag, attached to glutathione beads). The mixtures were rolled at 4°C overnight. After centrifugation, beads were washed three times with binding buffer and then eluted with binding buffer containing 10 mM reduced glutathione. After SDS-PAGE, Western blots were performed with anti-G␣ s and anti-G ␤ antibodies.
Size Exclusion Chromatography-Size exclusion chromatography was used to examine the binding of G␣ s and its mutants to G␤ 1 ␥ 2 in solution. G␣ s (1 M) and G␤ 1 (C68S)␥ 2 (2 M) were mixed and incubated on ice for 30 min. Samples were loaded on a Superdex 200 column (GE Healthcare Life Sciences) equilibrated with 20 mM Hepes, pH 8.0, 50 mM NaCl, and 1 mM EDTA at a flow rate of 0.5 ml/min. The elution was monitored at 280 nm, and 0.8-ml fractions were collected for subsequent SDS-PAGE analysis.
In Vivo cAMP Assay-Cells were plated on 6-well plates and treated with 1 mM isobutylmethylxanthine for 30 min at 37°C. After washing twice with HEM buffer (20 mM Hepes, pH 7.4, 135 mM NaCl, 4.7 mM KCl, 1.2 mM MgSO 4 , 2.5 mM NaHCO 3 , 0.1 mM Ro-20 -1724, 0.5 unit/ml of adenosine deaminase, and 1 mM isobutylmethylxanthine), cells were treated with 0, 1 nM, 10 nM, 100 nM, 1 M, 10 M, and 100 M isoproterenol in HEM buffer for 5 min at 37°C. After two additional washes with HEM buffer, cells were harvested and lysed with 0.5% Triton X-100 containing 1 mM isobutylmethylxanthine. The amount of cAMP was measured with a Direct Cyclic AMP Enzyme Immunoassay kit (Assay Designs, Inc.).
In Vitro Adenylyl Cyclase Activation Assay-Adenylyl cyclase activation assay of Sf9 membrane preparations was performed as previously described (22). Briefly, adenylyl cyclase isoform V (dog) was recombinantly expressed in Sf9 cells, and the adenylyl cyclase-containing membranes were prepared. G␣ s protein was activated by adding a buffer consisting of 10 mM NaF, 10 mM MgCl 2 , 30 M AlCl 3 and incubated at 30°C for 1 h. Activated G␣ s proteins with membrane preparations of adenylyl cyclase V were incubated in buffer of 50 mM Tris, pH 7.6, 2 mM isobutylmethylxanthine, 1 mM ATP, 10 mM MgCl 2 , 20 mM creative phosphate, 100 units/ml of creative phosphokinase at 30°C for 10 min. The samples were boiled for 3 min to stop the reaction. After spinning for 3 min at 16,000 ϫ g, the supernatant was used for cAMP measurement with Direct Cyclic AMP Enzyme Immunoassay kit.
Purification of Turkey ␤ 1 -Adrenergic Receptor Proteins-Turkey ␤ 1 -adrenergic receptor protein was purified as described previously (23). Turkey ␤ 1 -adrenergic receptor (residues 34 -424 with a mutation C116L) cDNA was subcloned into the pVL1393 vector. Hi5 cells were infected with the recombinant baculovirus carrying the turkey ␤ 1 -adrenergic receptor at a density of 2 ϫ 10 6 /ml. Sixty-hours later, cells were harvested and resuspended in lysis buffer (50 mM Tris-HCl, pH 8, 1 mM EDTA, 10 g/ml of leupeptin, 9 mM benzamidine, 5 g/ml of pepstatin A, and 2 mM PMSF). After homogenization, the cell lysate was centrifuged at 2,000 ϫ g for 10 min and the supernatant was centrifuged again at 150,000 ϫ g at 4°C for 1.5 h. The membrane pellet was resuspended in lysis buffer and homogenized again. After spinning down at 150,000 ϫ g at 4°C for 1.5 h, the final pellet was then resuspended in membrane extraction buffer (50 mM Tris-HCl, pH 8, 350 mM NaCl, 3 mM imidazole, 2% DM, 10 g/ml of leupeptin, 9 mM benzamidine, 5 g/ml of pepstatin A, and 1 mM PMSF) and rolled at 4°C for 3 h. After centrifugation at 150,000 ϫ g at 4°C for 1 h, solubilized membrane proteins were incubated with nickel-nitrilotriacetic acidagarose (Qiagen) overnight. After washing with buffer containing 50 mM Tris-HCl, pH 8, 350 mM NaCl, 3 mM imidazole, 0.1% DM, 10 g/ml of leupeptin, 9 mM benzamidine, 5 g/ml of pepstatin A, and 1 mM PMSF, the receptor protein was eluted down by an imidazole gradient.
GTP␥S Loading Assay-Agonist-stimulated GTP␥S loading to G-proteins was performed as previously described (19). G␣ s (with or without 1 M G␤␥) and turkey ␤ 1 -adrenergic receptor (20 nM) together with 10 M alprenolol or isoproterenol in 200 l of loading buffer (50 mM Tris, pH 8.0, 100 mM NaCl, 10 mM MgCl 2 , 1 mM EDTA, 0.02% DM, and 5 M GDP) were incubated on ice for 20 min. After incubation at 30°C (for wild-type and G␣ s 219) or at room temperature (for G␣ s 263 and G␣ s 195) for 5 min, 100 nM GTP␥S was added. At various times, 40-l aliquots were removed and added to 1 ml of termination buffer (20 mM Tris, pH 8.0, 100 mM NaCl, and 25 mM MgCl 2 , ice cold) and loaded onto nitrocellulose membrane (Schleicher & Schuell BioScience). After 3 washes with 1 ml of termination buffer, 4 ml of scintillation liquid were added to the membrane and 35 S was counted to measure GTP␥S loading. 5 mM GDP was used to determine nonspecific binding. G␣ s and Receptor Binding Assay-Different concentrations of G␣ s proteins (with an N-terminal GST tag, attached to glutathione beads) were incubated in 100 l of binding buffer with 80 nM ␤-AR proteins (with a FLAG tag) at 4°C for 1 h. GST (600 nM) was used as control. After centrifugation, beads were washed three times with washing buffer. After SDS-PAGE, Western blots were performed with anti-G␣ s and anti-FLAG M2 antibodies.

Identification of G␣ s Mutants Defective in G␤␥ Binding-To
identify G␣ subunits that could be activated by GPCRs in the absence of G␤␥ subunits, we first characterized G␣ s mutants that were defective in G␤␥ binding. The crystal structures of G␣ and G␣␤␥ have been solved (24,25). The conformational differences between the free and G␤␥-bound forms of G␣-GDP mainly involve residues that directly interact with G␤ (24,25). Based on the crystal structures of G␣ t G␤ 1 ␥ 1 and G␣ i1 G␤ 1 ␥ 2 complexes, residues in the N-terminal, Switch I (Arg-185 to Ile-193, encompassing Linker 2 region), and Switch II region of G␣ subunits are involved in contacting or binding of G␤␥ (24,25) (Fig. 1C, red residues). Furthermore, from the crystal structures of the Ras superfamily GTPases and their guanine nucleotide-exchange factors, such as Ras and Sos1, Arf1 and Sec7, Rac1 and Tiam1, EF-Tu and EF-Ts, as well as Ran and RCC1 (26 -30), guanine nucleotide-exchange factors interact extensively with and remodel the Switch I and II regions of GTPases. Therefore, we have performed alanine scanning mutagenesis of every residue in these regions as well as some residues in Switch III and its adjacent regions (Table 1 and Fig. 1D).
First we identified G␣ s mutants that could not bind G␤␥. We purified recombinant proteins of wild-type G␣ s and 66 mutant G␣ s proteins from E. coli (Table 1 and some examples shown in Fig. 2A). (Here we used the short spliced variant of bovine G␣ s . The difference between the short and long splice forms of G␣ s is an insertion of 14 amino acids after residue Gly-70.) We also purified recombinant G␤ 1 ␥ 2 proteins from insect Hi5 cells ( Fig.  2A). To test the functionality of these purified G␣ s mutant proteins, we performed in vitro GTP␥S binding assays and adenylyl cyclase activation assays for all G␣ s mutants (Table 1). Other than eight mutants (Table 1), all G␣ s mutant proteins were able to bind to GTP␥S, implying these purified proteins were stable and functional. Furthermore, using membrane preparations from Sf9 insect cells infected with recombinant baculoviruses carrying adenylyl cyclase type V, we performed in vitro adenylyl cyclase activation assays (Table 1, and some examples shown in Fig. 2, B-E). In addition to the 8 mutants that could not bind GTP␥S in vitro, 4 more mutants could not activate adenylyl cyclase in vitro (Table 1). Three (Asn-265, Arg-266, and Trp-267) of these 4 residues are in the ␣3/␤5 loop, which directly contacts the catalytic domain of adenylyl cyclase as revealed by the crystal structure of the complex of G␣ s and the catalytic domains of adenylyl cyclase (31). Thus, most of the purified mutant G␣ s proteins were stable and functional.
We next tested the interaction of these G␣ s mutants with G␤␥. For the in vitro binding experiments, we incubated wildtype G␣ s and mutant G␣ s proteins with purified G␤ 1 ␥ 2 proteins. Glutathione beads were used to pulldown G␣ s proteins. Co-precipitation of G␤ 1 ␥ 2 was detected with anti-G␤ antibody. As shown in Table 1 (and some examples in Fig. 2F), wild-type G␣ s and 48 mutant G␣ s proteins pulled down G␤ 1 ␥ 2 , 18 other mutant G␣ s proteins (including G␣ s E195A, G␣ s K219A, or G␣ s W263A) did not bind G␤ 1 ␥ 2 . To confirm these binding data and to show that the 18 mutant G␣ s proteins could not interact with other G␤ subunits in addition to G␤ 1 , we used membrane preparations of 293T cells as the source of G␤␥ subunits (293T cells at least express G␤ 1 , G␤ 2 , and G␤ 4 ). The results were the same as the binding experiments with purified G␤ 1 ␥ 2 (Table 1 and some examples shown in Fig. 2G). As a third approach, we used size exclusion chromatography to verify the inability of some G␣ s mutants binding to G␤␥. As shown in Fig. 2H, wildtype G␣ s , G␣ s K219A, and G␤ 1 ␥ 2 eluted as single major peaks from the size exclusion column, demonstrating the homogeneity of the subunit preparations. When combined with excess G␤ 1 ␥ 2 , wild-type G␣ s showed a two-peaks elution profile: one was the free G␤ 1 ␥ 2 and the other was the complex of G␣ s -G␤ 1 ␥ 2 with a shorter retention time as compared with the isolated wild-type G␣ s (Fig. 2H). On the other hand, G␣ s K219A showed a two-peak elution prolife with the same retention times as free G␤ 1 ␥ 2 and free G␣ s K219A, indicating that G␤ 1 ␥ 2 had little effect on the retention time of G␣ s K219A and that G␣ s K219A was unable to form a complex with G␤ 1 ␥ 2 (Fig. 2H). Similar results were observed with G␣ s E195A and G␣ s W263A, which could not form trimers with G␤ 1 ␥ 2 . Together, these data demonstrate that 18 G␣ s mutants (including G␣ s E195A, G␣ s K219A, and G␣ s W263A) are unable to interact with G␤␥ subunits, and the locations of these 18 G␣ s mutants are displaced in Fig. 3 (red and green residues).
Activation of G␣ s Mutants by ␤-Adrenergic Receptors in Cells-Because our goal was to find G␣ s mutants that could be potentially activated by GPCRs in the absence of G␤␥, we next investigated the activation of all 66 G␣ s mutants by GPCRs in G␣ s -deficient cells. G␣ s -deficient mouse embryonic fibroblast (MEF) cells were derived from G␣ s knock-out mouse embryos (32,33). Because exon 2 of the G␣ s gene was deleted, none of the two alternative spliced variants of G␣ s were present in these G␣ s Ϫ/Ϫ MEF cells (32,33). We have made stable cell lines with these G␣ s Ϫ/Ϫ MEF cells expressing G␣ s and all 66 G␣ s mutants (in pcDNA3.1/hygromycin vector). Stimulation of these cells with isoproterenol, which activates the endogenous G s -coupled ␤-adrenergic receptors, increased cellular cAMP accumulation in cells expressing wild-type and 37 mutant G␣ s proteins ( Table  1, and some examples shown in Fig. 4). Among the 18 G␣ s mutants that could not interact with G␤␥, 8 mutants (G␣ s D182A, G␣ s L184A, G␣ s R185A, G␣ s E195A, G␣ s F198A, G␣ s K219A, G␣ s W263A, and G␣ s N23A/I26A/E27A/L30A/ D33A) could still mediate receptor activation of adenylyl cyclases leading to cAMP increase (Fig. 4, B-E, and green in Fig. 1D). As control, G␣ s Ϫ/Ϫ MEF cells did not show any cAMP increase upon isoproterenol stimulation (Fig. 4A). The functionality of these G␣ s mutant proteins was also verified by cholera toxin, a direct activator of G␣ s (Fig. 4). These cellular studies demonstrate that 8 G␣ s mutants are able to functionally couple to ␤-adrenergic receptors, are activated by the receptors, and stimulate the downstream effector adenylyl cyclases, leading to the cellular accumulation of cAMP, even though they could not bind to G␤␥.

Activation of Purified G␣ s Mutants by Purified ␤-Adrenergic Receptors in Vitro in the Absence of G␤␥-
To directly demonstrate that a GPCR has the capability to accelerate the guaninenucleotide exchange on G␣ without G␤␥, we performed biochemical studies with purified recombinant proteins of a GPCR and G␣ s mutants in vitro. We purified recombinant turkey ␤ 1 -adrenergic receptors from insect Hi5 cells (23) (Fig. 5A). G␣ s proteins were initially purified as GST fusion proteins and the GST tag was removed by protease cleavage (Fig. 5B). Among the 8 G␣ s mutants identified above (able to be activated by ␤-AR in cells but unable to bind G␤␥), 5 (D182A, L184A, R185A, F198A, and N23A/I26A/E27A/L30A/D33A) were unstable after removal of the GST tag. Therefore, we examined the activation of the remaining 3 G␣ s mutants (E195A, K219A, and W263A) by ␤-adrenergic receptor. The activation was monitored by the initial rate of GTP␥S loading onto G␣ s subunits. As reported previously, purified ␤-adrenergic receptors had no effect on the rate of GTP␥S loading (ϳ90 fmol/min) onto wild-type G␣ s proteins in the absence of G␤␥ subunits (19,34) (Fig. 5C). The rate of GTP␥S loading on G␣ s was the same in the presence of isoproterenol (an agonist for ␤-adrenergic receptor) or alprenolol (an antagonist). The initial rate of GTP␥S loading to G␣ s alone was similar to that of G␣ s with ␤-adrenergic receptor in the presence of the antagonist alprenolol. In contrast, in the presence of purified G␤␥ proteins, isoproterenol increased the initial rate of GTP␥S loading onto G␣ s (ϳ450 fmol/min) by 4-fold compared with that in the presence of alprenolol (ϳ115 fmol/min) (Fig. 5D). The fold-increase is similar to that reported in previous reconstituted systems, reflecting a relatively high basal nucleotide exchange rate of G␣ s (19,34).

TABLE 1 Summary of the characterization of 66 G␣ S mutants
GTP␥S binding and in vitro activation of adenylyl cyclase (AC) were shown as percentage of the values obtained with wild-type G␣ S . The increase of cAMP in cells is shown as the fold increase from each individual experiment. ⌬N31 and ⌬N38 are deletion mutants of the N-terminal 31 or 38 amino acids, respectively.

FIGURE 2. Functional characterization of G␣ s and its mutants.
A, Coomassie Blue staining shows the purified G␣ s proteins (some representatives), GST and G␤ 1 ␥ 2 . B-E, in vitro activation assays of adenylyyl cyclase by G␣ s and its mutants (some representatives). F and G, in vitro binding of G␣ s and representative mutants to G␤ 1 ␥ 2 . H, size exclusion chromatography of heterotrimer formation. Upper panels: elution profiles are shown for wild-type G␣ s alone, G␣ s K219A alone, G␤ 1 ␥ 2 alone, the mixture of wild-type G␣ s /G␤ 1 ␥ 2 , and the mixture of G␣ s K219A/G␤ 1 ␥ 2 . Lower panels, SDS-PAGE analysis of the elution fractions. One representative experiment from three independent experiments is shown for each case. WB, Western blot.
With purified G␣ s E195A, we found that, without G␤␥ subunits, isoproterenol increased the rate of GTP␥S loading by about 2.4-fold compared to that with alprenolol (from ϳ39 fmol/min with alprenolol to ϳ95 fmol/min with isoproterenol) (Fig. 5E). Addition of G␤␥ subunits did not significantly change the rates with either isoproterenol (ϳ105 fmol/min) or alprenolol (ϳ40 fmol/min) (Fig. 5F). Similar results were obtained with G␣ s K219A and G␣ s W263A mutants: without G␤␥ subunits, isoproterenol increased the rate of GTP␥S loading by 2ϳ3-fold on G␣ s K219A and G␣ s W263A mutants, compared to that with alprenolol. Furthermore, the mutations did not seem to alter the interaction between G␣ s and ␤-AR (Fig. 5G). Moreover, increasing the concentrations of G␤␥ enhanced the initial rate of GTP␥S loading onto wildtype G␣ s , but not G␣ s K219A (Fig. 5H). Additionally, increasing the concentrations of ␤-AR elevated the initial rate of GTP␥S loading onto both wild-type G␣ s (in the presence of G␤ 1 ␥ 2 ) and G␣ s K219A (in the absence of G␤ 1 ␥ 2 ) (Fig. 5H). These biochemical experiments clearly demonstrate that purified ␤ 1 -adrenergic receptors can accelerate the guanine nucleotide exchange on G␣ s in the absence of G␤␥ subunits. We should note that previous studies with rhodopsin and transducin had indicated that high concentrations of rhodopsin might activate transducin in the absence of G␤␥, and that G␤ 1 or G␥ 1 knock-out mice still had some light response (35)(36)(37)(38)(39)(40). Thus, GPCRs possess the ability to catalyze the nucleotide exchange on G␣ subunits.
Role of Extended Linker 2 in G␣ Activation-How do the 8 G␣ s mutations (G␣ s D182A, G␣ s L184A, G␣ s R185A, G␣ s E195A, G␣ s F198A, G␣ s K219A, G␣ s W263A, and G␣ s N23A/I26A/E27A/ L30A/D33A) alleviate the requirement of G␤␥ for the activation of G␣ s by ␤-adrenergic receptors? When mapped onto the crystal structure of the complex of ␤ 2 -AR and G s , 3 of these 8 residues (Asp-182, Leu-184, and Arg-185) are in ␣F helix (Fig. 6 A  and B). Two of the 8 residues (E195 and F198) are in ␤2 sheet (Fig. 6B). ␣F and ␤2 flank Linker 2, which connects the Ras-like GTPases domain and the ␣-helical domain of G␣ s (Figs. 6, A  and B, and 1D). Also, within Linker 2, mutations of Arg-187, Leu-189/Thr-190, Ile-193, and Thr-196 all blocked ␤-AR induced cAMP increases in cells even though these mutants could bind to G␤␥, GTP␥S, and activate adenylyl cyclase in vitro (Table 1). Hence, almost all residues in the extended Linker 2 are critical for G␣ activation. Although some mutations (such as R185A and E195A) enable G␣ activation by FIGURE 3. Spatial locations of 18 G␣ s mutants that are defective in G␤␥ binding. The structure displayed is G␣ i1 ␤ 1 ␥ 2 (PDB code 1GP2). G␣, yellow; G␤, gray; G␥, gold. Corresponding residues for the 18 G␣ s mutants are highlighted (red and green). Green labeled residues are the 8 G␣ s mutants that are defective in G␤␥ binding but can still be activated by ␤-ARs. Red labeled residues are the 10 G␣ s mutants that are defective in G␤␥ binding and could not be activated by ␤-ARs.  GPCR in the absence of G␤␥, other mutations (such as S191A and I193A) block G␣ activation by GPCR in cells. When comparing the structures of inactive G␣ i -GDP and active G␣ i -GTP␥S, or the structures of inactive G␣ t -GDP and active G␣ t -GTP␥S, Linker 2 is moved toward the ␥-phosphate, bringing the side chain oxygen of Thr-190 into the coordination sphere of the Mg 2ϩ ion where it replaces one of the water molecules observed in the structure of the GDP form (24,25). Furthermore, it has been shown that Linker 2 changes the conformation upon GPCR activation (41). A critical role for the extended Linker 2 in G␣ activation is consistent with the observation from the crystal structure of the complex of ␤ 2 -AR and G s , which shows a large rotation (ϳ127 o ) of the ␣-helical domain relative to the Ras-like GTPase domain upon G-protein activation (Fig. 6, A and E) (10). ␣F/Linker 2, through the ␤ 2 -strand, is connected to the ␤ 2 /␤ 3 loop, which interacts with intracellular loop 2 of ␤ 2 -AR (Fig. 6F). Furthermore, experiments using various biophysical measurements suggest a "clam-shell" like opening model for G␣ activation by GPCRs, in which the helical domain opens away from the Raslike GTPase domain (42)(43)(44)(45) (Fig. 6D). Here the clustering of several G␣ s mutants critical for GPCR activation of G␣ s around the extended Linker 2 suggests a possibility that Linker 2 may serve as the "hinge" in this clam-shell model (Fig. 6D). Similarly, in the "rolling-top" model (in which the ␣-helical domain slides and rotates away along the sideways from the Ras-like GTPase domain) observed in the ␤ 2 -AR⅐G s complex structure, this hinge is stretched (some residues in the linkers were disordered and thus were unmodeled in the structure) (10) (Fig. 6E). Mutations of residues in this hinge could make it easy to open the interface between the two domains of G␣ or to enhance the interdomain motions. Preventing this interdomain opening by cross-linking inhibits GPCR-catalyzed G-protein activation (42). We should note that, in addition to Linker 2, Linker 1 also connects the Ras-like GTPase domain and the ␣-helical domain. Indeed a point mutation (G56P) in Linker 1 of transducin increased the basal exchange rate and exhibited some degrees of activation at high levels of rhodopsin in the absence of ␤ 1 ␥ 1 (35). The role of Linker 1 in G␣-protein activation by GPCRs requires future systematic investigation. Here we propose that a GPCR via its intracellular loop 2 directly interacts with the ␤ 2 /␤ 3 loop of G␣ to communicate to Linker 2 and ␣F, resulting in the rotation of the helical domain and the release of GDP (Fig. 6, E and F).
Linker 2 as the Target Site of a Natural Product Inhibitor for G q -The crystal structure of G q and a small molecule natural product inhibitor shows the inhibitor binds to a hydrophobic cleft and directly contacts Linker 2 (Fig. 6G) (46). This inhibitor, YM-254890, inhibits the guanine-nucleotide exchange reaction by preventing the GDP release (46). Also, this inhibitor blocks the AlF 4 Ϫ -induced conformational change in G␣ q (46). It is proposed that this inhibitor stabilizes an inactive GDP-bound form (46). These structural data suggest that Linker 2 might be critical for G-protein activation, and that the Linker 2 region is a potential therapeutic targeting site.
Role of ␣N, ␣3, and Switch Regions in G␣ Activation in Cells-In addition to the extended Linker 2, our data have uncovered some other residues that are essential for G␣ s activation in cells by ␤-ARs. The N-terminal residues (Asn-23, Ile-26, Glu-27, Leu-30, and Asp-33) and Trp-263 are located at or close to the receptor-interacting surface (10,15,47,48) (Fig. 6, A and C). Trp-263 is close to the ␣ 3 /␤ 5 loop. This region is analogous to GEF contact sites in other GTPases such as Ras and EF-Tu (26,27). The crystal structure of ␤ 2 -AR/G s has revealed that ␤ 2 -AR interacts with the ␣N/␤ 1 hinge; this may explain the role of the 5 N-terminal residues (Asn-23, Ile-26, Glu-27, Leu-30, and Asp-33) in the activation of G␣ by GPCR.
Lys-219 is located within Switch II (Figs. 1D and 6C). Lys-206 of G␣ t (corresponding to Lys-219 in G␣ s ) had been identified to be important for G␣ t activation by rhodopsin (49). Several residues near Lys-219 of G␣ s such as Gly-212 and Gln-213 are critical for triggering conformational changes in Switch II through an interaction with the ␥-phosphate in GTP and for stabilizing the transition state for GTP hydrolysis, respectively (13,14,50,51). Indeed, most G␣ s mutations in the Switch I and II regions (such as Ile-193, Thr-196, Asp-209, Gly-212, Arg-214, Asp-215, Arg-217, Arg-218, Trp-220, Ile-221, and Gln-222) could not be activated by ␤-adrenergic receptors in cells, despite their ability to bind to G␤␥ subunits, bind to GTP␥S, and activate adenylyl cyclases in vitro (Table 1). Arg-187 and Thr-190 (in Switch I) as well as Gly-212 (in Switch II) contact the ␥-phosphate of GTP (6). Together, these data demonstrate that ␣N, ␣3, Switch I and Switch II are critical for G␣ activation by GPCRs.
The critical role of Switch I and II regions in G␣ activation is similar to the activation of Ras-superfamily GTPases by their GEFs. GEFs for small GTPases have different structures (26 -30). They contact GTPases at the same as well as different amino acid residues and induce different conformational changes on GTPases to drive out the GDP. However, they all utilize a two-sided attack to release positive charges (the Mg 2ϩ ion and the invariant lysine residue in the P loop (phosphatebinding loop)) from their interaction with the phosphates of the nucleotide (52). These GEFs interact extensively with and remodel Switch I and II regions of GTPases, which form part of the binding pocket for Mg 2ϩ and the ␥-phosphate of GTP. Thus, although GPCRs are unusual GEFs because they do not contact these switch regions directly, they still require/remodel these switch regions for the activation of G␣.
Possible Role of G␤␥ in G␣ Activation-Our mutagenesis studies unexpectedly reveal a role for residues Trp-263 (in ␣3) and Asn-265 (in the ␣3-␤5 loop) in G␤␥ binding (Table 1). From the structure of the G␣ i ⅐GDP/G␤␥ complex, this region is not directly involved in contacting with G␤␥. However, this region is right next to Switch III and forms an elaborate interdependent network of polar interactions with the Switch II region, which is critical for interacting with G␤␥ (24). Although we did not address the role of G␤␥ in the guanine-nucleotide exchange on G␣, we should mention that there were two models that proposed a catalytic role for G␤␥ (15,16,21). G␤␥ subunits contact the switch regions of G␣ subunits (24,25), and have a structure similar to RCC1 (30). Hence, G␤␥ subunits would be more like GEFs if only G␤␥ subunits could be proven to possess GEF activity without GPCRs. The role of G␤␥ in G␣ activation could be complex. G␤␥ stabilizes GDP binding on G␣, thus serving as a guanine-nucleotide dissociation inhibitor. GPCR activation could release the inhibition of G␤␥. However, the nucleotide exchange rate of G␣, in the absence of G␤␥, is still slower than that in the presence of GPCRs. This would indicate that, in addition to releasing G␤␥, GPCRs act on the nucleotide exchange on G␣. G␤␥ subunits could serve an additional catalytic role to augment the complex formation between GPCRs and G␣ subunits, similar to the role of ELMO in facilitating nucleotide exchange on Rac by Dock180 (53). Indeed, our data ( Table 1, Fig. 5, E and F) show that, whereas G␣ s E195A could be activated by ␤-ARs in the absence of G␤␥, the extent of activation was suboptimal relative to wild-type G␣ s in the presence of G␤␥. This implies that G␤␥ somehow contributes to the activation process.
Previously two models were proposed for a catalytic role for G␤␥ subunits in activating G␣ subunits. In the "lever" model (15,21), GPCRs cause a tilt (or a rotation) of G␤␥ relative to G␣. The membrane proximal part (interacting with the G␣ N-terminal helix) of G␤␥ moves closer to G␣. The opposite end of G␤␥ moves away from G␣ and, at the same time, pulls along the interacting parts of G␣. These interacting residues from Switch I and II of G␣ are part of the lip of the nucleotide binding pocket. Therefore, GDP can exit permitting exchange by GTP. In the second "gear shift" model (16), GPCRs also cause a tilt of G␤␥. However, the rotating direction is opposite of that in the lever model. G␤␥ moves closer to G␣. The N terminus of G␥ (the membrane distal end of G␤␥) moves to contact and displace the helical domain of G␣, creating an exit route for GDP. The exact role of G␤␥ in the exchange reaction needs further investigation.
Conclusions-Upon ligand-binding, GPCRs initiate the exchange of GDP bound on G␣ subunits of G-proteins with GTP. However, the detailed molecular mechanisms by which GPCRs activate G-proteins are not well understood. Here we reveal some new insights on this process. We demonstrate that a GPCR could activate some G␣ mutants in the absence of G␤␥, and that the extended Linker 2 in G␣ is critical for G␣ activation by GPCRs. These data will advance our understanding of this critical cellular signaling process.
The closed and open states of the cleft (the GDP/GTP binding site) are determined by relative pivoting of the Ras-GTPase domain and the ␣-helical domain at Linker 2. The closed state of the cleft is represented in all crystal structures of G␣ subunits. One of the open states of the cleft is captured in the ␤ 2 -AR/G s crystal structure. This model of activation of G␣ subunits suggests that factors binding at the interface of the Ras-GTPase domain and the helical domain might facilitate cleft closure, thus serving as an inhibitor of G␣ activation (Fig. 6G). Given this pivot model, the two domains of G␣ could be either clam-shell opening or rolling-top expansion including a relative rotation of the GTPase domain and the helical domain around an axis through the linkers (Fig. 6, D and E). Results from several biophysical studies are consistent with the clam-shell opening, and that the crystal structure of the ␤ 2 -AR⅐G s complex is in line with the rolling-top movement. Therefore, in addition to the established role for ␣5-helix of G␣ connecting GPCRs to the guanine-nucleotide binding pocket, we show here that the extended Linker 2 connects GPCRs to the nucleotide binding pocket as well as the ␣-helical domain of G␣. In conclusion, our data demonstrate that GPCRs possess the capability to catalyze the nucleotide exchange on G␣ subunits, and that the extended Linker 2 is critical for G␣ activation by GPCRs.