A Transient Interaction between the Phosphate Binding Loop and Switch I Contributes to the Allosteric Network between Receptor and Nucleotide in Gαi1*

Background: G proteins couple receptor binding to nucleotide release via an allosteric network. Results: Mutation of allosteric sites of Gαi1 stabilizes a transient signaling conformation and may highlight an allosteric connection between receptor and nucleotide. Conclusion: The P-loop interacts with Switch I in the K345L variant of Gαi1. Significance: G protein signaling is critical for numerous cellular functions, and GDP release is the rate-limiting step of the cycle. Receptor-mediated activation of the Gα subunit of heterotrimeric G proteins requires allosteric communication between the receptor binding site and the guanine nucleotide binding site, which are separated by >30 Å. Structural changes in the allosteric network connecting these sites are predicted to be transient in the wild-type Gα subunit, making studies of these connections challenging. In the current work, site-directed mutants that alter the energy barriers between the activation states are used as tools to better understand the transient features of allosteric signaling in the Gα subunit. The observed differences in relative receptor affinity for intact Gαi1 subunits versus C-terminal Gαi1 peptides harboring the K345L mutation are consistent with this mutation modulating the allosteric network in the protein subunit. Measurement of nucleotide exchange rates, affinity for metarhodopsin II, and thermostability suggest that the K345L Gαi1 variant has reduced stability in both the GDP-bound and nucleotide-free states as compared with wild type but similar stability in the GTPγS-bound state. High resolution x-ray crystal structures reveal conformational changes accompanying the destabilization of the GDP-bound state. Of these, the conformation for Switch I was stabilized by an ionic interaction with the phosphate binding loop. Further site-directed mutagenesis suggests that this interaction between Switch I and the phosphate binding loop is important for receptor-mediated nucleotide exchange in the wild-type Gαi1 subunit.

Heterotrimeric G proteins (G␣␤␥) switch between activation states to elicit cellular responses (1). In the G␣ subunit, signaling competence is encoded into surface conformations that reflect the identity of bound guanine nucleotide. The GTPbound form of the G␣ subunit is considered the activated state as its Switch regions adopt conformations that can interact with effector molecules. Hydrolysis of GTP to GDP changes the conformations of these regions and converts the G␣ subunit to a state that instead associates with G␤␥. GDP-bound G␣␤␥ then traffics to the membrane where it can interact with G proteincoupled receptor (GPCR). 3 The GPCR in turn facilitates release of GDP from the G␣ subunit. Subsequent binding of GTP both completes the G protein signaling cycle and is the rate-limiting step of G protein signaling (1).
The importance of the C terminus in receptor binding was first revealed using peptide mimetics of various regions of the G protein (3). Interestingly, combinatorial libraries of peptide mimetics of the G␣ t C terminus containing systematic sequence substitutions identified mutations with improvements in affinity for rhodopsin over the parent sequence (20). The greatest increase in affinity was ϳ200-fold and was associated with a lysine-to leucine substitution (K341L in G␣ t , K345L in G␣ i1 ) (20,21). Later, a G␣ t peptide harboring this substitution was used as a tool to stabilize opsin and metarhodopsin II for structure determination (22)(23)(24)(25).
Surprisingly, however, when the K341L mutation was introduced into chimeric G␣ t /G␣ i1 , it did not similarly improve the affinity for receptor (26). Because both G proteins and cognate GPCRs transmit signals via allosteric networks, one possibility for this apparent contradiction is that this C-terminal mutation influences the allosteric network within the G␣ subunit in a way that disfavors the receptor binding state and thus reduces receptor coupling efficiency and GDP release. Here, we used the K345L variant of G␣ i1 to probe the allosteric pathway linking receptor binding to guanine nucleotide exchange. Our results combine measurements of basal and receptor-catalyzed nucleotide exchange, receptor binding studies, high resolution crystallography, and stability measurements to demonstrate how the K345L mutation alters the different nucleotide-bound states of the G␣ subunit. An interaction between the P-loop and Switch I in the crystal structure of the K345L G␣ i1 subunit identifies one potential conformational intermediate of nucleotide exchange. This allows us to propose a new connection between structural elements of G␣ subunits, and suggests how the P-loop and Switch I help to coordinate nucleotide release.

EXPERIMENTAL PROCEDURES
G␣ i1 Expression and Purification-The cDNA encoding Rattus norvegicus G␣ i1 was amplified from the pPAL7 vector (generously donated by Dr. Ongun Onaran (Ankara University)) using appropriate primers and subcloned into the pSV277 expression vector (Vanderbilt University) to include an N-terminal hexahistidine tag and a thrombin cleavage site. The E43A and K345L mutations were introduced using the QuikChange Lightning site-directed mutagenesis ki (Stratagene) and confirmed by DNA sequencing (GenHunter).
Before purification, cell pellets were thawed and resuspended in ice-cold lysis buffer (50 mM NaHPO 4 , 300 mM NaCl, 2 mM MgCl 2 , 5 mM ␤-mercaptoethanol, 20 M GDP, pH 8.0) supplemented with 0.1 mM PMSF or 1 mM Pefabloc and 1 g/ml aprotinin, leupeptin, and pepstatin. The resuspended cells were disrupted by sonication, and the lysate was clarified by centrifugation for 1 h at 220,000 ϫ g (50,000 rpm in a Ti70 rotor). The supernatant was treated with 10 g/ml DNase and RNase, filtered, and added in batch to TALON Cobalt affinity resin equilibrated in lysis buffer. After 1 h at 4°C, the resin was transferred to a gravity flow column and washed with lysis buffer, then lysis buffer supplemented with 5 mM imidazole pH 8.0. Protein was eluted with lysis buffer supplemented with 100 mM imidazole, pH 8.0, concentrated using an Amicon 10-kDa molecular weight cutoff centrifugal concentrator at 2000 ϫ g, then diluted 20-fold. Thrombin was added (1 unit/mg of purified protein), and the sample was incubated overnight at 4°C to cleave the N-terminal affinity tag. The protein sample was further purified over TALON Cobalt affinity resin equilibrated in lysis buffer and by size exclusion chromatography on a Superdex S200 10/300GL column equilibrated in storage buffer (50 mM Tris-Cl, 200 mM NaCl, 2 mM MgCl 2 , 1 mM EDTA, 1 mM DTT, 20 M GDP, pH 8.0). Purified G␣ i1 was concentrated to 10 mg/ml as determined by the BCA assay (Pierce) in triplicate, and glycerol was added to 10% (v/v) before storage at Ϫ80°C. The purified protein contained two additional N-terminal residues (Gly-Ser) derived from the thrombin recognition sequence.
Basal and Receptor-mediated Nucleotide Exchange-The rate of GDP exchange for GTP␥S in G␣ i1 was determined by monitoring the increase in intrinsic tryptophan fluorescence ( ex ϭ 290 nm, em ϭ 340 nm) using a Varian Cary Eclipse fluorescence spectrometer. The fluorescence signal from basal nucleotide exchange was measured at 21°C. In this assay the basal fluorescence from 500 nM wild-type or variant G␣ i1 (in 50 mM Tris-Cl, 200 mM NaCl, 2 mM MgCl 2 , 1 mM DTT, pH 7.5) was collected for 5 min, after which time 10 M GTP␥S was added and mixed by pipetting. The fluorescence was then recorded for a minimum of 60 min. Receptor-mediated nucleotide exchange was monitored for heterotrimeric G protein, G␣ i1 ␤ 1 ␥ 1 , reconstituted by incubating wild-type or variant G␣ i1 with G␤ 1 ␥ 1 purified from endogenous transducin in a 1:1 molar ratio for at least 20 min. 2 M dark rhodopsin in ureawashed ROS membranes was added to 500 nM heterotrimeric G protein in assay buffer and incubated in the dark for a minimum of 20 min to allow complex association. GTP␥S was added to a final concentration of 10 M mixed by pipetting and fluorescence-monitored for an additional 5 min to measure the basal signal. Finally, the samples were light-activated and mixed by pipetting, and the fluorescence signal from receptor-mediated nucleotide exchange was monitored at 21°C for a minimum of 60 min. Time courses for both basal and receptor-mediated nucleotide exchange experiments were determined empirically by monitoring fluorescence signal decay after the addition of GTP␥S and were 60 and 90 min, respectively, in the experiments reported here. Nucleotide exchange rates were calculated from data for three independent experiments with four replicates per experiment and fit in Prism Version 6.0c using a one-site exponential association equation of the form where t is time (min), F o is the magnitude of the fluorescence signal at t ϭ 0, F max is the maximum fluorescence signal, F is the fluorescence signal observed at time ϭ t, and k is the rate constant (min Ϫ1 ). Values were normalized to wild-type G␣ i1 .
Rhodopsin Binding-Wild-type and variant G␣ i1 binding to rhodopsin was measured in urea-washed ROS membranes as previously described (29). G␣ i1 (5 M) was incubated with G␤␥ (10 M) and rhodopsin (50 M) in binding buffer (50 mM Tris-Cl, 100 mM NaCl, 2 mM MgCl 2 , pH 8.0) for 30 min at 4°C and assessed under three different conditions: dark-adapted, after light activation, and after light activation with the addition of GTP␥S (100 M). Supernatants were separated from membranes by centrifugation at 200,000 ϫ g for 1 h, and the darkadapted supernatants were removed under dim red light. Isolated fractions were boiled, visualized by Coomassie-stained SDS-PAGE, and quantified by densitometry using a Bio-Rad Multimager. Quantities of 37-kDa G␣ i1 in either the soluble or insoluble fraction are expressed as a percentage of the total protein in both. Data reported are the average of at least three independent experiments.
Extra Metarhodopsin II Formation-The formation of metarhodopsin II in ROS membranes was measured on an Aminco DW2000 spectrophotometer in the presence of increasing concentrations of G proteins previously described (4). Heterotrimeric G protein was reconstituted from purified wild-type or variant G␣ i1 in the same manner as used for receptor-mediated nucleotide exchange measurements. Rhodopsin (2 M) in darkadapted urea-washed ROS membranes were incubated with varying concentrations of wild-type or K345L G␣ i1 ␤ 1 ␥ 1 in Extra Meta II buffer (50 mM HEPES, 100 mM NaCl, 1 mM MgCl 2 , 1 mM DTT, pH 8.0) on ice for ϳ10 min. Absorption spectra for both dark and light-adapted samples were then collected at 4°C. After collection of a dark-adapted spectrum, samples were exposed to 2 quick flashes of light ϳ30 s apart. The light-adapted spectrum was then immediately collected. The extra meta II signal was calculated as the change in meta II formation (⌬A 380 nm Ϫ ⌬A 440 nm ) before and after light activation at 4°C. The EC 50 values for wild-type and mutant G␣ i1 were calculated by plotting data as a function of the G␣ i1 ␤ 1 ␥ 1 concentration and fit to a four-parameter, variable slope equa- where Y is the meta II signal, top is the Y value at the top plateau, bottom is the Y value at the bottom plateau, x is the log of the concentration of G␣ i1 ␤␥, and hill slope is the slope factor or the Hill slope (unit-less).
In Silico Modeling of G␣ C Terminus Binding to Rhodopsin-The structure of opsin bound to the high affinity peptide of the G␣ C terminus (PDB entry 3DQB (22)) and the structure of the ␤ 2 -adrenergic receptor bound to G␣ s (PDB entry 3SN6 (15)) were superimposed and prepared for in silico analysis in Maestro (Schrödinger LLC) (30). The position equivalent to G␣ i1 -345 in peptide and in G␣ s was mutated in silico to either the amino acid originally in the structure (Arg or Leu) or the converse and set to the most favorable rotamer. A truncated-Newton energy minimization of all four structures was performed in the Maestro Workspace GUI using the program Prime (31,32) and the OPLS_2005 all atom molecular mechanics force field (33). The Surface Generalized Born (VSGB) continuum solvation model (34) was also applied. To relax the structures around the mutations, the side chains were left unrestrained during the minimization. The energy minimizations were performed for two iterations and 65 steps per iteration. Resultant changes in receptor-G␣ interactions were evaluated visually in Coot (35). After this procedure, structures of the crystallized entities (positive control calculations) were in close agreement with the deposited coordinates and structure factors.
Differential Scanning Fluorimetry to Measure Thermostability-Differential scanning fluorimetry was performed as described (36). Protein samples were diluted to a final concentration of 5 M in assay buffer containing 5ϫ SYPRO Orange (Bio-Rad) and 50 M GDP or GTP␥S. Thermostability was screened in both the extra meta II (EMB) assay buffer and nucleotide exchange (NEB) assay buffer. Triplicate samples were prepared in 20-l volumes and transferred to a clear low profile 96-well PCR plate (Bio-Rad) and equilibrated at 25°C for 2 min in a Bio-Rad CFX96 Real-time system (C1000 Thermal Cycler) before a temperature ramp from 25 to 95°C in 0.2°C increments at a rate of 1°C/min. Changes in the fluorescence signal were monitored upon a temperature ramp from 25 to 95°C. Data were analyzed using differential scanning fluorimetry analysis tools Version 3.0.2 (36). Sigmoidal regions of the data were fit with the Boltzmann equation to determine the T m values.
Crystallization, Data Collection, Structure Determination, and Refinement-Crystals of GDP-and GTP␥S-bound forms of the K345L G␣ i1 variant protein were grown under the same conditions using the hanging-drop vapor diffusion method and previously reported reservoir conditions (37). Briefly, purified protein in 80 mM HEPES, 120 mM succinic acid, 8 mM DTT, pH 8.0, was incubated with either 1 mM GTP␥S or 20 M GDP, 40 M AlCl 3 , and 16 mM NaF. Crystals were grown from hanging drops consisting of 4.8 l of protein (10 mg/ml) and 1.2 l of reservoir solution (2.0-2.2 M ammonium sulfite, pH 8.0, 5-20 mM magnesium sulfate) equilibrated against 1 ml of reservoir solution at 20°C. Crystals were cryo-protected by briefly soaking in reservoir solution containing glycerol at a final concentration of 17.5% and flash-cooled by plunging in liquid nitrogen.
Diffraction data were collected at the Advanced Photon Source (APS) beamlines 21-ID-D and 21-ID-G. All data were collected at 100 K and recorded on a Mar 300 CCD detector. Crystals formed in two different space groups: P3 2 21 for GTP␥S-bound K345L G␣ i1 and I4 for GDP-bound K345L G␣ i1 . All data were indexed, integrated, and scaled using HKL2000 (38). Crystallographic data processing and refinement statistics are reported in Table 1.
The structure of GTP␥S-bound K345L G␣ i1 was determined by molecular replacement in Phaser (39) using the structure of K349P G␣ i1 as a search model (PDB entry 2ZJY (37)). The struc-ture of GDP-bound K345L G␣ i1 was determined by molecular replacement in Phaser (39) using the GTP␥S-bound K345L G␣ i1 structure as a search model. Initial model building used composite omit maps and was performed in Coot (35). Subsequent iterative rounds of refinement in Phenix (40) and model building in Coot (35) were performed to improve model quality, with sequential omit mapping used during the entire refinement process to minimize bias to the search model. Geometry was assessed in Procheck (41) and Molprobity (42). Figs. 1, 5, 6, and 9 were prepared using PyMOL (43). Supplemental Movie 1 was prepared in PyMOL (43) and Chimera (44).

RESULTS
Basal and Receptor-mediated Nucleotide Exchange of the K345L G␣ i1 Subunit-A hallmark of G␣ subunits is the ability to exchange GDP for GTP. Recombinant G␣ i1 activation was assessed by comparing normalized basal and receptor-mediated nucleotide exchange rates calculated from the increase in intrinsic tryptophan fluorescence upon the addition of the nonhydrolyzable GTP analog, GTP␥S. The normalized K345L G␣ i1 exchange rates were ϳ30 and 20% slower under basal and receptor-mediated conditions, respectively, than wild-type G␣ i1 ( Fig. 2; Table 2).
Binding and Activation of the K345L G␣ i1 ␤ 1 ␥ 1 by Rhodopsin-Receptor-catalyzed GDP release from G␣ subunits depends on efficient coupling between GDP-bound G␣␤␥ and activated receptor to promote and stabilize the nucleotide-free state of G␣. The effect of the K345L mutation in the G␣ i1 subunit on the ability of the G protein to couple to receptor was assessed by measuring the amount of reconstituted G␣ i1 ␤ 1 ␥ 1 bound to rhodopsin in ROS membranes (Fig. 3, A and B). Comparison of wild-type and K345L G␣ i1 subunits recovered in the membrane-bound fraction after light activation revealed reduced binding of K345L G␣ i1 . Conversely, the relative amount of G␣ i1 released into the soluble fraction after light activation was greater for K345L G␣ i1 . This indicates a lowered coupling efficiency of G protein containing the K345L G␣ i1 with light-activated receptor.
The extra meta II assay was used as a complementary method to clarify whether the lowered functional coupling efficiency  was a result of reduced receptor affinity. Using this method, the affinity between activated receptor and wild-type or K345L G␣ i1 ␤ 1 ␥ 1 (Fig. 3C; Table 2) was quantified. The affinity between rhodopsin and wild-type G␣ i1 ␤ 1 ␥ 1 (K d ϭ 1.03 Ϯ 0.10 M) was found to be comparable to that previously determined by kinetic light scattering (K d ϭ 0.72 Ϯ 0.05 M) (26). The K345L G␣ i1 ␤ 1 ␥ 1 (K d ϭ 1.86 Ϯ 0.13 M) had a modestly reduced affinity.
The difference in affinity between peptide containing the K345L mutation (20) and intact protein harboring this sequence might have several origins. For example, the orientations of the bound peptide and the G␣ s C terminus differ significantly in the opsin-G␣ t peptide costructure and the ␤ 2 -adrenergic-G␣ s costructure (15,22,23) and could have different contacts. Accordingly, in silico modeling of rhodopsin bound to wild-type or K345L G␣ i1 in the context of either a peptide or the intact subunit was based upon the opsin-G␣ t peptide costructure (22) and the G␣ s -␤ 2 -adrenergic receptor costructure (15), respectively. These calculations (not shown) did not offer an obvious explanation, such as steric clash or loss of hydrogenbonding interactions, for the observed difference in affinity of the K345L mutation in the context of G␣ subunits (26) versus the corollary C-terminal G␣ peptide (20,21). This prompted speculation that the allosteric network connecting receptor binding to nucleotide release had been modified.
Thermostability of the K345L G␣ i1 Variant-To identify how the K345L G␣ i1 substitution influences the stability of the nucleotide binding states of G␣ i1 , the thermostability of the wild-type and the K345L variant was measured in the presence of GDP (Fig. 4A; Table 2) or GTP␥S ( Fig. 4B; Table 2). Consistent with previous studies, both wild-type and K345L G␣ i1 are more stable when bound to GTP␥S than when bound to GDP (45)(46)(47). The T m values for GTP␥S-bound wild-type and K345L G␣ i1 were statistically identical. In comparison, the T m of the GDP-bound K345L G␣ i1 variant was decreased by 2.89 Ϯ 0.04°C as compared with wild-type G␣ i1 (Table 2) indicating that this substitution selectively destabilizes the GDP-bound state.

Structures of the GTP␥S-and GDP-bound K345L G␣ i1
Subunit-To examine the architectural changes in the K345L G␣ i1 variant, crystal structures in the GDP-and GTP␥S-bound states were determined to 2.1 and 1.5 Å resolution, respectively ( Table 1). The resolution of these structures is among the highest observed for any G␣ subunit with each respective nucleotide. Superposition of corresponding backbone C␣ atoms of GTP␥S-bound K345L G␣ i1 and GTP␥S-bound wild-type G␣ i1 (PDB entry 1GIA (47)) resulted in a root mean square deviation of 0.330 Å (308 C␣ atoms aligned out of 321 total) suggesting little overall conformational change accompanied the mutation. It is notable that residues at the N and C termini that have not previously been resolved in crystal structures of GTP␥Sbound G␣ i1 proteins were clearly observed in the electron density in the current structure (Fig. 5, A and B). These additional residues contribute to the formation of an extended ␤1 strand at the N terminus (Fig. 5A) and assignment of the ␣5 helix C terminus to residue 351 (Fig. 5B). At this time it is unclear whether the observation of the residues at the termini is a direct result of the mutation or is caused by crystalto-crystal variation.
Superposition of GDP-bound K345L G␣ i1 and GDP-bound wild-type G␣ i1 (PDB entry 1GDD (48)) resulted in a root mean square deviation of 0.473 Å between C␣ atoms (321 C␣ atoms aligned out of 330 total). Interestingly, the majority of the structural differences observed in the GDP-bound K345L G␣ i1 structure occurred within functionally important motifs. The Switch regions in the GDP-bound K345L G␣ i1 structure displayed the most significant conformational changes. The Switch I position was marked by an inward shift (Fig. 6A) as compared with wild type. This reorientation positioned Switch I in closer proximity to the ␣and ␤-phosphates of the bound GDP. Accompanying the new position of Switch I in the K345L G␣ i1 structure was the formation of a salt bridge (2.3 Å) between Arg-178 from Switch I and Glu-43 of the P-loop (Fig.  6B). Moreover, Switch II and III differed in both structure and orientation as compared with wild-type G␣ i1 . Although Switch II is still not completely visible, 5 of the 17 Switch II residues and two of the six Switch III residues that were not observed in GDP-bound wild-type G␣ i1 were clearly resolved (Fig. 6C). Interestingly, these adopted a conformation that differed from GDP-bound G␣ t (49) (Fig. 6D) and GTP␥S-bound G␣ i1 (47) (Fig. 6E). It is possible that the altered Switch I orientation and the P-loop/Switch I salt bridge observed in the GDP-bound K345L G␣ i1 structure help promote these unique Switch II and Switch III conformations.
Validation of the Conformational Changes in the K345L G␣ i1 as a Part of the Allosteric Network-The functional characterization presented here suggests that the K345L mutation stabilizes a transiently formed structural state sampled during G␣ i1 activation. Consistent with the role of these conformations in allosteric signaling, most of the residues that adopt conformations different from wild-type G␣ i1 have previously been mutagenized and exhibit altered signaling characteristics (18, 19, 50 -53).
To further validate that the observed conformational changes are important for allosteric signaling, we selected Glu-43 as a target for mutagenesis. In the GDP-bound K345L G␣ i1 structure, Glu-43 forms a salt bridge to Arg-178 that links   the P-loop and Switch I (Fig. 6B). If this salt bridge is transiently formed in the wild-type G␣ i1 subunit during physiological allosteric signaling, its alteration would be anticipated to reduce receptor-mediated nucleotide exchange rates. Biochemical analysis ( Fig. 7A; Table 2) indeed showed a 35% reduction in receptor-mediated nucleotide exchange while leaving basal nucleotide exchange unaffected. As assessed by the meta II assay, the binding between the E43A G␣ i1 variant and rhodopsin was reduced ϳ2-fold (K d ϭ 2.15 Ϯ 0.69 M) ( Fig. 7B; Table 2) but is still comparable to the affinity observed between rhodopsin and the K345L G␣ i1 variant given the standard error of the measurement. The effect of this mutation on receptor binding and receptor-mediated activation, but not basal nucleotide exchange, is consistent with Glu-43 being a part of the allosteric network in wild-type G␣ i1 . The T m of the E43A G␣ i1 variant was reduced by a similar amount (ϳ2°C) in the presence of either GDP or GTP␥S (Fig. 7, C and D, Table 2). This equivalent reduction in thermodynamic stability may reflect loss of the salt bridge.

DISCUSSION
The K345L Substitution in G␣ i1 Influences the Guanine Nucleotide Binding States Differently-G proteins and cognate GPCRs mediate signaling via conformational changes in response to either bound guanine nucleotides or agonist, respectively. In GPCRs, it has been proposed that the signaling states are not necessarily discrete, but that most receptors instead adopt a continuum of conformations along an energy landscape (54), with crystal structures representing snapshots of transient states formed along the activation pathway (55-59). Indeed, many receptors exhibit basal activity, and these can spontaneously sample the active state in the absence of agonist (58,60,61). The analogous observation of basal nucleotide exchange in G␣ subunits requires sampling of the nucleotide-free state in the absence of receptor. This suggests that G proteins similarly have intrinsic conformational heterogeneity that allows spontaneous sampling of multiple conformational states, albeit to a substantially lower Ϫ -bound G␣ i1 (PDB entry 1GFI (47); teal). B, stereoview of the salt bridge between Glu-43 and Arg-178 in the GDP-bound K345L G␣ i1 subunit. Coloring for the overlaid structures is the same as panel A. The closest distance between a Glu-43O⑀ and the Arg-178N atom in any of the wild-type structures is 3.1 Å in the GTP␥S-bound G␣ i1 (PDB entry 1GIA (47); purple). This could be considered a long hydrogen bond, but it is noted that the temperature factors of both side chains are elevated in that structure. C-E, SwII and SwIII from GDP-bound K345L G␣ i1 (green) superimposed with GDP-bound wild-type G␣ i1 (PDB entry 1GDD (48); pink) (C), GDP-bound wild-type G␣ t (PDB entry 1TAG (49); blue), (D) and GTP␥S-bound wild-type G␣ i1 (PDB entry 1GIA (47); purple) (E). FIGURE 7. Biochemical properties of wild-type and E43A G␣ i1 . Data for E43A G␣ i1 subunits are shown as dashed bars, data for wild-type G␣ i1 subunits are shown as solid bars. A, basal and receptor-mediated nucleotide exchange rates measured as a function of intrinsic fluorescence. B, rhodopsin affinity for G␣ i1 ␤ 1 ␥ 1 determined by the extra meta II assay for wild-type (F) and E43A (OE) G␣ i1 ␤ 1 ␥ 1 . C and D, fluorescence analysis of heat-induced melting of wild-type and E43A G␣ i1 . Protein sample in NEB or EMB buffer was incubated with GDP (C) or GTP␥S (D). Results shown are the mean Ϯ S.E. of at least two independent experiments (**, p Ͻ 0.01; ****, p Ͻ 0.0001). extent. Subsequent biophysical investigations suggest that the nucleotide-free form of the G protein is the most conformationally dynamic and is destabilized relative to the GDP-or GTP-bound forms; however, it retains an intact fold competent for signaling (62)(63)(64)(65).
Numerous transient conformational states may bridge the conversion between the GDP-and GTP-bound forms of the G␣ subunit. The majority of published G␣ mutations are associated with an increase in both intrinsic and receptor-mediated nucleotide exchange (for example, Refs. 18, 29, and 51-53)), which can result either from a shift of the conformational equilibrium to favor the nucleotide-free and GTP-bound states or from destabilization of the fold to result in spontaneous nucleotide release. In contrast, a reduction in both basal and receptor-mediated nucleotide exchange was observed in K345L G␣ i1 . The mechanistic basis for this reduction in nucleotide exchange could potentially be attributed to one of many physical changes, but cannot be ascribed to a destabilization of the fold.
One intriguing possibility is that this mutation has altered the energetic landscape between the nucleotide-bound states of G␣ by shifting the stability of the allosteric states. Several lines of evidence suggest that Lys-345 is directly involved in the allos-teric pathway linking receptor binding to GDP release in G␣ i1 subunits. First, the K345L mutation has a 200-fold increase in receptor affinity in the context of a peptide (20) but a modestly lower affinity in the context of the intact G␣ subunit (26). In silico modeling did not identify obvious physical phenomena that would explain this difference in affinity. Together, this suggests that the mutation instead reduces the ability of the G protein to bind receptor, possibly by decreasing the stability of the K345L G␣ i1 subunit in the nucleotide-free state. In agreement with this interpretation, the nucleotide-exchange (Fig. 2) and receptor coupling results (Fig. 3) suggest that the K345L G␣ i1 subunit is impaired in its ability to adopt the nucleotidefree state. The statistically significant decrease in T m of the GDP-bound state of the K345L G␣ i1 subunit suggests that this variant also disfavors the GDP-bound state as compared with wild type (Fig. 4A) and that only the GTP-bound state is unaffected by (Fig. 4B). Taken together, one interpretation of our measurements and modeling can be qualitatively conceptualized as a shifted energy diagram for the nucleotide binding states in K345L G␣ i1 (Fig. 8).
Structural Changes in GDP-bound K345L G␣ i1 Identify Connections between Known Allosteric States-The mechanism of allosteric coupling of receptor binding to guanine nucleotide release in G proteins has been investigated by multiple methods (7, 15, 18, 19, 29, 50 -53, 66 -72). A hypothetical model summarizing both literature results and our results here is shown in Fig. 9 and supplemental Movie 1. One major focus in the field is the role of the ␣5 helix (Fig. 9, red), which is an extension of the G␣ C terminus that connects the receptor binding surface to its nucleotide binding pocket. For example, site-directed spin labeling combined with EPR spectroscopy measurements (18,19,50) revealed changes in both mobility and solvent accessibility along the ␣5 helix upon receptor activation. These changes were attributed to a rigid-body rotation and translation motion adopted by the helix. Uncoupling of this motion from receptor interactions by the insertion of a glycine linker into the ␣5 helix abolished nucleotide exchange within the G␣ i1 subunit (18,68). In complementary studies cysteine mutagenesis was   Fig. 1, with the Glu-43 and Arg-178 side chains highlighted in blue. In the left panel, the GDP-bound G␣ subunit is shown in its G␤␥-bound conformation (G␤␥ omitted for clarity) and is based on the G␣ i1 ␤␥ heterotrimer structure (PDB entry 1GP2 (16)). In the center panel the conformation of the C terminus of the G␣ subunit is based on the K341L G␣ t peptide-bound meta II costructure (PDB entry 3PQR (23)) with the conformation of the G␣ subunit a hybrid between the GDP-bound K345L G␣ i1 structure, and the receptor-bound G␣ s structure (PDB entry 3SN6 (15)). In the right panel, the conformation of helical domain is from a computational model of the rhodopsin-G t costructure (67) and is based on the direction of rotation observed in the receptor-bound G␣ s structure (PDB entry 3SN6 (15)) and the magnitude of the rotation measured in site-directed spin labelingdouble electron-electron resonance and electron microscopy studies (69,79). used to stabilize the position of the ␣5 helix in an activated conformation, which resulted in a dramatic increase in basal nucleotide exchange and suggested that the ␣5 helix dipole could contribute to GDP release (29). Finally, the structure of the ␤ 2 -adrenergic receptor in complex with cognate G␣ s showed a significant roto-translation of the ␣5 helix as compared with the position observed in any other structure of a G␣ subunit (15).
Our results identify a conformation of the G␣ i1 subunit that we hypothesize may be transiently adopted during activation. In this conformation, the position of Switch I is stabilized by its interaction with the P-loop (Fig. 9, blue). A similar interaction is observed in structures of inhibited G␣ subunits (76 -78) and in the context of the G␤␥ subunits (16,17) prompting an analogy to a "seatbelt" that prevents GDP release (16). Consistent with the role of this interaction in controlling nucleotide release, the distance between the closest Glu-43O⑀ and Arg-178N atoms lengthens under conditions that promote nucleotide exchange (75).
However, if the sole functional role of this interaction were to prevent GDP release, we would anticipate that the mutation of Glu-43 to alanine would dramatically increase receptor-mediated nucleotide exchange. Our observed reduction of receptormediated nucleotide exchange in the E43A G␣ i1 subunit is thus initially counterintuitive. There are numerous possibilities for this unanticipated biochemical behavior. For example, significant denaturation or conversion of the E43A G␣ i1 subunit to a nonfunctional state could have occurred over the time course of the experiments; however, we did not detect behaviors reflecting folding problems. Instead, we propose that the interaction between the P-loop and Switch I is a transient feature of G␣ activation. Recent hydrogen-deuterium exchange measurements in G␣ s are consistent with the hypothesis that this interaction is indeed a conformation of the receptor-associated G␣ subunit before GDP release (76). Given the increased ordering of Switch I and Switch II in the structure of the GDP-bound K345L G␣ i1 variant, we hypothesize that the formation of the salt bridge between Glu-43 and Arg-178 (Fig. 9, blue) stabilizes these loops before GDP release by receptor-associated G␣ i1 .
The final step in nucleotide release is rotation of the helical domain of the G␣ i1 subunit away from the GTPase domain ( Fig.  9, violet). Cross-linking of the GTPase and helical domains established that this domain separation is required for receptor-catalyzed nucleotide exchange (69). The magnitude and direction of the rotation is suggested by site-directed spin labeling-double electron-electron resonance (69), electron microscopy (79), and the crystal structure of the ␤ 2 -adrenergic receptor in complex with cognate G␣ s (15), respectively.
Conclusions-This study allows the connection of structural elements involved in nucleotide exchange in G␣ subunits via a putative transient interaction between the P-loop and Switch I. We hypothesize that interaction between the P-loop and Switch I may stabilize the G␣ i1 subunit before GDP release. Taken within the context of previous literature, these findings add to the known structural elements that contribute to the allosteric mechanism of GDP release from G␣ subunits. The complexity of this process suggests that there must be precise coordination between numerous components of the G␣ subunit.