Conformational dynamics of a G-protein α subunit is tightly regulated by nucleotide binding

Significance

G proteins are molecular switches for cellular signaling induced by G-protein–coupled receptor activation. The Gα subunit is the central timer of signal transduction regulated by GTP hydrolysis, which returns the system to its inactive state. Although previous work has characterized the structural states of Gα during the GTPase cycle, we show here that Gα is highly dynamic in the apo and GDP-bound states but in complex with GTP is completely rigid and is locked in a defined domain orientation. These insights help demonstrate that the conformational plasticity of G proteins is a central feature of their switching functionality.

Abstract

Heterotrimeric G proteins play a pivotal role in the signal-transduction pathways initiated by G-protein–coupled receptor (GPCR) activation. Agonist–receptor binding causes GDP-to-GTP exchange and dissociation of the Gα subunit from the heterotrimeric G protein, leading to downstream signaling. Here, we studied the internal mobility of a G-protein α subunit in its apo and nucleotide-bound forms and characterized their dynamical features at multiple time scales using solution NMR, small-angle X-ray scattering, and molecular dynamics simulations. We find that binding of GTP analogs leads to a rigid and closed arrangement of the Gα subdomain, whereas the apo and GDP-bound forms are considerably more open and dynamic. Furthermore, we were able to detect two conformational states of the Gα Ras domain in slow exchange whose populations are regulated by binding to nucleotides and a GPCR. One of these conformational states, the open state, binds to the GPCR; the second conformation, the closed state, shows no interaction with the receptor. Binding to the GPCR stabilizes the open state. This study provides an in-depth analysis of the conformational landscape and the switching function of a G-protein α subunit and the influence of a GPCR in that landscape.

Heterotrimeric G proteins are localized at the inner leaflet of the plasma membrane where they convey signals from cell-surface receptors to intracellular effectors (1). Heterotrimeric G proteins consist of two functional units, an α subunit (Gα) and a tightly associated βγ complex. The Gα subunit harbors the guanine nucleotide-binding site. In the inactive GDP-bound state, the Gα subunit is associated with the βγ complex. Exchange of GDP for GTP on the Gα subunit, triggered by interaction with the agonist-bound G-protein–coupled receptor (GPCR), results in a conformational change leading to GDP release and ultimately to GTP binding and subunit dissociation. The complexity of the mechanism by which a GPCR activates the Gα subunit based on available crystal structures has been discussed recently (2, 3). Both the Gα subunit and the βγ subunit interact with downstream effectors and regulate their activity. The intrinsic GTP hydrolysis of the Gα subunit returns the protein to the GDP-bound state, thereby increasing its affinity for the Gβγ subunit, and the subunits reassociate (Fig. 1A), ready for interaction with the agonist-bound GPCR. Throughout this cycle, the Gα subunit is engaged in specific interactions with the GPCR and/or the βγ subunit that stabilize the flexible parts of the protein, e.g., its switch regions. Only the GTP-bound form is stable enough to mediate downstream signaling.

Fig. 1.

Ligand binding and thermal stability assays of an inhibitory Gαi1 subunit. (A) Mechanism of G-protein activation by a GPCR. A GPCR ligand binds to the receptor and induces its active conformation, i.e., binding to a heterotrimeric G protein located at the intracellular side of the cell membrane. This activation stimulates a conformational change within the Gαi1 subunit leading to the exchange of bound GDP for GTP and to the subsequent dissociation of the heterotrimer and downstream signaling. (B) FRET-based nucleotide-binding assay using fluorescently (MANT)-labeled GDP and GMP-PNP. Intrinsic tryptophan residues in the protein were excited at 280 nm, and the FRET signal at 445 nm was monitored upon the stepwise addition of Gαi1. (C and D) CD spectra (C) and CD-detected thermal melting curves (D) of Gαi1Δ31 bound to various ligands. (E) Binding mode of GMP-PNP to Gαi1. Residues engaged in additional hydrogen bonds (red lines) occurring only in the GMP-PNP–bound form are highlighted.

Crystallographic (4–9), biochemical (10), and biophysical (11–13) studies have elucidated details of the conformational states of the Gα subunit during the GTPase cycle. The Gα subunit has two structural domains, a nucleotide-binding domain (the Ras-like domain) and a helical domain (the α-H domain) that partially occludes the bound nucleotide (Fig. 1A). Because of this steric consideration, nucleotides can be released only upon at least partial opening of the two domains. The Ras-like domain is associated with the membrane surface through N-terminal myristoylation, and its position is further restricted by interaction with the βγ complex and the GPCR. The helical domain is connected to the nucleotide-binding domain through two flexible linkers, and linker 1 (switch I) undergoes conformational changes upon receptor binding (12). The relative orientation of these two subdomains has been investigated previously at lower resolution by double electron–electron resonance (DEER) spectroscopy (13–15). One of these studies (15) also used extensive molecular dynamics (MD) simulations to monitor conformational changes within the Gα subunit. In these calculations the authors identified large conformational changes taking place in the microsecond time scale.

To determine experimentally the changes in the relative orientation and dynamics of the two subdomains of the Gα subunit in its apo form and upon binding to GDP and GTP, we performed a thorough NMR spectroscopic characterization of an inhibitory Gα subunit, Gαi1. We combined our NMR experiments with small-angle X-ray scattering (SAXS), circular dichroism (CD), and fluorescence spectroscopy to gain detailed insights into the conformational states of Gαi1 along the nucleotide-binding trajectory. In addition, we performed Carr–Purcell–Meiboom–Gill (CPMG) relaxation dispersion experiments, MD simulations, and 2D NMR transverse relaxation optimized spectroscopy (TROSY) experiments to monitor the dynamics of Gαi1 in the minute-to-microsecond time scale in the apo form and in complex with different nucleotides. The results show that Gαi1 adopts a more open conformation in the apo and GDP-bound forms, but its conformation is fairly compact and rigid in complex with GTP analogs. The degree of Gα domain opening as determined here is less pronounced than shown in a recent GPCR-stimulatory heterotrimeric G protein complex structure (3) but is in line with previous electron paramagnetic resonance (EPR) studies (13) at lower resolution and with MD simulations (15). In contrast to those studies, here we present structural models of the Gα conformational states at a per-residue resolution based on various structural methods. Furthermore, we provide detailed insights into the functional dynamics of Gαi1 extracted from CPMG relaxation dispersion and 2D TROSY NMR experiments. These data show that the Gα subunit is dynamic in the apo and GDP-bound form but is rigid in complex with GTP. We also show that the GPCR-binding ability of Gα correlates positively with the degree of conformational plasticity. Thus, this study describes previously unknown dynamical properties of the Gα subunit in solution and in the conformational states associated with GPCR binding.

Results

Active Gαi1 Can Be Produced in Escherichia coli for NMR Structural Studies.

To produce human Gαi1 (UniProt accession no. P63096) for biophysical studies and NMR spectroscopy, we constructed a fusion protein containing Gαi1 and GB1 (protein G B1 domain) serving as an expression and solubility tag (16) during expression in E. coli. Low-temperature cell growth and application of low concentrations of isopropyl β-d-1-thiogalactopyranoside (IPTG) for induction (see Methods and previous protocols in ref. 17) allowed us to obtain 5–10 mg of active Gαi1 per liter of cell culture. Purified protein was free of bound GDP as probed by ¹H NMR spectroscopy (Fig. S1). Initial NMR experiments performed with full-length Gαi1 resulted in poor spectral quality, so we designed a truncated variant lacking the N-terminal 31 residues (Gαi1Δ31). The deleted part of the protein is not resolved in the crystal structure [Protein Data Bank (PDB) ID code: 1cip (18)] and is unfolded in solution as shown here by NMR and elsewhere (19). This truncated protein variant resulted in high-quality NMR spectra as probed with 2D [¹⁵N,¹H]-TROSY and TROSY for rotational correlation times (TRACT) experiments (Fig. S2). To validate the nucleotide-binding activity of Gαi1Δ31, we used FRET assays with methylanthraniloyl (MANT)-labeled nucleotides (MANT-GDP and MANT–GMP-PNP) (Fig. 1B). Upon excitation of protein tryptophan side chains at 280 nm, the emitted light leads to excitation of the MANT moiety and finally emission at 445 nm. Subsequent titration of protein to a 1-µM MANT–nucleotide solution led to a proper binding curve that permitted the extraction of binding constants in the order of 10 µM with slightly tighter binding of the GTP analog. Next, we characterized the secondary structure, content, and thermal stability of Gαi1 in the presence of different nucleotides and inhibitors (Fig. 1C). The minima in the CD signal of Gαi1Δ31 at 208 and 222 nm, indicative of the α-H secondary structure, are markedly increased in the presence of GTP analogs and are less pronounced in complex with GDP and the G-protein inhibitor NF-023. In line with these findings, the thermal stability of these complexes as monitored with CD is highest with guanosine 5′-O-(3-thiotriphosphate (GTPγS) and lowest in the apo form, indicating a strong stabilizing effect of GTP analogs. These differences can be partly explained by the formation of additional hydrogen bonds between the γ-phosphate of the GTP analog and Gαi1Δ31 (Fig. 1D and Fig. S3). We investigated the structural and internal mobility details of these different structural states using solution NMR spectroscopy.

Fig. S1.

NMR spectroscopy of Gαi1Δ31 preparations. Gα extracted from E. coli culture was purified with Ni-NTA and size-exclusion chromatography. This procedure yielded pure Gαi1 that was free of any bound nucleotides, as can be seen by the spectral differences in 1D [¹H]-NMR and 2D [¹⁵N,¹H]-TROSY in the apo (red spectra) and the GDP-bound (blue spectra) protein. (A) Signals arising from GDP are completely absent in the 1D spectrum of the apo sample. Signals labeled with an asterisk originate from the reducing agent DTT in the sample buffer [20 mM sodium phosphate (pH 7.0), 50 mM NaCl, 5 mM MgCl₂, 5 mM DTT]. (B) A 2D [¹⁵N,¹H]-TROSY spectrum shows that GDP induced chemical shift perturbations in Gα and indicates that the apo form is able to bind to nucleotides.

Fig. S2.

Initial NMR analysis of Gα constructs. (A) 2D TROSY spectra of full-length Gαi1 and the optimized truncated variant Gαi1Δ31. The signal intensity of resonances around 8 ppm is reduced significantly for the Δ31 variant, indicating a lower content of amino acids in an unfolded conformation. The spectral quality in the GMP-PNP–bound form is improved compared with the apo state. (B) ¹⁵N-TRACT experiments to determine the apparent correlation time and molecular weight of Gαi1 and Gαi1Δ31 in various nucleotide-bound states. Gαi1Δ31 showed improved relaxation properties and thus was used for subsequent NMR experiments.

Fig. S3.

(Upper) Comparison of crystal structures of Gαi1 bound to GMP-PNP (Left) and to GDP (Right). Regions in the protein not resolved in the GDP-bound structure are marked by lines, and corresponding residues are labeled. (Lower) LigPlot+ (50) diagram of the interaction between Gαi1 and GMP-PNP and between Gαi1 and GDP.

Gαi1 Displays Ligand-Dependent Changes as Probed with NMR Chemical Shift Perturbations.

To characterize the structure and dynamics within Gαi1 upon binding to various nucleotides, we produced U-[²H,¹³C,¹⁵N]–labeled Gαi1Δ31 and prepared a sample with a protein concentration of ∼250 µM, either in the apo form or in the presence of 3 mM GDP, guanosine 5′-[β,γ-imido]triphosphate (GMP-PNP), or GTPγS. We performed the following TROSY-based 3D triple-resonance NMR experiments (20) on these samples: HNCO, HN(CA)CO, HNCA, HN(CO)CA, HNCACB, and 3D-¹⁵N–edited TROSY-NOESY. All 3D experiments were recorded with Poisson-gap nonuniform sampling (NUS) data acquisition (21) and iterative soft threshold (IST) reconstruction (22). We could obtain backbone assignments of 89% (288/323) of the residues for the GMP-PNP–bound (Fig. 2A), 88% (284/323) for the GTPγS-bound, 78% (252/323) for the GDP-bound, and 77% (259/323) for the apo state. Chemical shift-based estimation of the secondary structure based on differences in random coil chemical shifts in Cα and Cβ nuclei (23) correlates well with the secondary structure elements in the crystal structure of Gαi1 in complex with GMP-PNP (PDB ID code: 1cip) (Fig. 2B). We further analyzed the NMR spectral changes in Gαi1Δ31 upon the addition of various nucleotides and the Go/i-selective Suramin-based small-molecule inhibitor NF-023 (Fig. 2C) (24). The data show that GDP induces relatively small (up to 0.15 ppm) chemical shift perturbations within Gαi1Δ31. These chemical shift changes cluster specifically to the Ras-like domain with little effect on the helical domain (gray and orange coloring in Fig. 2 D and E). The inhibitor induces a similar pattern of chemical shift changes but in addition leads to the disappearance of 18 resonances, most likely because of exchange broadening (negative black bars in Fig. 2C) located around the nucleotide-binding site and along the GPCR-binding site in the Ras domain. In contrast, the addition of GTP analogs (GMP-PNP and GTPγS) leads to the appearance of 38 additional resonances (red negative bars in Fig. 2C and red color in Fig. 3 D and E), which again are clustered around the nucleotide-binding site and all over the Ras-domain of Gαi1. The appearance of additional resonances points toward the quenching of exchange processes and a more rigid structure of Gαi1Δ31 in the GTP-bound form, in line with crystallographic studies (5) in which the electron density of parts of the Ras domain could not be observed in the GDP-bound form, presumably because of increased flexibility (Fig. S3). Furthermore, GTP analogs induce larger chemical shift perturbations than GDP clustered around the docking site of the two subdomains (Fig. 2 C and D). Regions around residues 204–217 and residues 233–239 are not resolved in the crystal structures of Gαi1 in the GDP-bound form but are visible in the complex structure with GMP-PNP (Fig. S3).

Fig. 2.

NMR assignment and mapping of binding sites of Gαi1. (A) 2D [¹⁵N,¹H]-TROSY spectra of Gαi1Δ31 in complex with the nonhydrolyzable GTP analog GMP-PNP. Around 85% of all backbone resonances could be assigned, as labeled. (B) Secondary chemical shift of Gαi1Δ31 plotted against residue number. Positive and negative values indicate α-helical and β-sheet conformation, respectively. As a comparison, the secondary structure elements in the crystal structure (PDB ID code: 1cip) are shown as blue bars. (C) [¹⁵N,¹H]-averaged chemical shift perturbations within Gαi1Δ31 upon the addition of various ligands. Negative black and red bars indicate resonances disappearing or reappearing, respectively, upon ligand binding. (D) Chemical shift perturbations mapped onto the structure of Gαi1 color-coded in orange and red according to the positive and negative bars in C. (E) Cartoon representation of Gαi1 bound to GDP and to GTPγS color-coded according to NMR chemical shift perturbations. Structural elements within Gαi1 are labeled. Nucleotides are shown as green spheres.

Fig. 3.

Interaction between Gαi1Δ31 and a GPCR in phospholipid nanodiscs. (A) 2D [¹⁵N,¹H]-TROSY experiments were used to extract chemical shift changes of 50 µM ²H,¹⁵N-labeled Gαi1Δ31 (black spectra) in the apo, GDP-bound, and GMP-PNP–bound forms upon the addition of an equimolar amount of the neurotensin-bound activated neurotensin-1 receptor variant HTGH4 L167R (28) in phospholipid nanodiscs (red spectra) or 50-µM empty nanodiscs as a reference (blue spectra). Chemical shift perturbations are mapped onto the structure and are color-coded according to the chemical shift perturbations (C.S.P.) bar. (B) Empty nanodiscs induce moderate chemical shift perturbations on Gαi1Δ31 that are more pronounced in the open apo and GDP-bound forms than in the closed GMP-PNP–bound form. (C) In contrast, the chemical shift perturbations induced by GPCR-containing nanodiscs clearly cluster within the Ras domain around helix 5, the main site of interaction with the GPCR. These effects are most pronounced in the apo state and are absent in the closed GMP-PNP–bound state. MSP, membrane scaffold protein.

Gαi1 Interacts with an Activated GPCR in Phospholipid Nanodiscs.

We next studied the interaction between uniformly ²H,¹⁵N-labeled Gαi1Δ31 with a thermostabilized (25, 26), signaling-competent (27) variant of rat neurotensin receptor subtype 1 [HTGH4 L167R (28)], which was purified from E. coli as described previously (27, 29) and then incorporated in phospholipid nanodiscs (30, 31). The binding process was monitored by 2D-[¹⁵N,¹H]-TROSY experiments with ²H,¹⁵N-labeled Gαi1Δ31 in the apo, GDP-, or GMP-PNP–bound forms. We recorded NMR experiments with Gαi1Δ31 samples alone and after the addition of empty nanodiscs assembled with 1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) lipids and membrane scaffold protein 1D1 (MSP1D1) and finally in complex with nanodiscs containing neurotensin-activated HTGH4 (Fig. 3 and Fig. S4).

Fig. S4.

Assembly of phospholipid nanodiscs containing NTR1. For nanodisc assembly, we used a 20-fold molar excess of MSP1D1 over rat NTR1 (HTGH4 L167R variant) (28) and a POPC-to-MSP ratio of 25:1. After assembly, empty nanodiscs were removed by cation exchange on an SP Sepharose HiTrap column (GE Healthcare). GPCR-containing nanodiscs eluted from the column at a sodium chloride concentration of ∼150 mM; empty discs did not bind to the column. Eluted protein was applied to a Superdex S200 10/300 GL size-exclusion column (24 mL bed volume) (GE Healthcare). a.u., arbitrary units.

Differences in NMR resonance positions in these experiments were used to calculate chemical shift perturbations of Gαi1Δ31 upon GPCR binding (color coding in Fig. 3). We observe a weak effect between empty nanodiscs and Gαi1Δ31 that was more pronounced in the apo and GDP-bound states and was weaker in the closed GMP-PNP–bound state. Sites in Gαi1Δ31 affected by binding to the phospholipid bilayer are located in sheets β2 and β3, in switch III, and around helix αG and helix α5 (Fig. 3A). However, the addition of GPCR-containing nanodiscs induced more pronounced chemical shift perturbations in the fast chemical-exchange regime within Gαi1Δ31 in the apo and GDP-bound states but not in the GMP-PNP–bound state (Fig. 3B). These effects are most prominent for the apo state and are located in sheets β1, β2, β3, and β5, helix αG, helix α4, sheet β6, and the C-terminal helix α5 in the Ras domain. The C-terminal helix previously has been shown to undergo translational and rotational motion upon GPCR binding and represents one of the main sites of interaction with the activated GPCR (3, 11). This interaction might be more pronounced in the presence of the Gβγ subunit and with myristoylated Gα.

Relative Gαi1 Domain Orientation Is Dependent on Nucleotide Binding.

To obtain further structural insights about the orientation of the two subdomains in Gαi1, the Ras-like domain and the helical domain, we performed SAXS experiments on the apo, GDP-, and GTPγS-bound forms of Gαi1Δ31 (Fig. 4A and Table 1). SAXS data yield the overall shape and radius of gyration of the biomolecule of interest.

Fig. 4.

Structural changes within Gαi1 in the apo and GDP- or GTPγS-bound forms using SAXS and NMR RDCs. (A) SAXS data and back-calculated scattering profiles of Gαi1Δ31 in the apo, GDP-, or GTPγS-bound forms. (B) Structural models derived from SAXS experiments were refined further with amide proton-nitrogen (HN)-RDCs obtained with 8 mg/mL Pf1 phage-aligned Gαi1Δ31 in the apo and GDP- or GTPγS-bound form. Correlation and quality factors of the agreement between back-calculated and experimental RDCs are shown. (C) Structural models obtained after the two-step refinement in each case. The apo and GDP-bound forms adopt a more open conformation than the GTP-bound form. (D) MD simulation of apo-Gαi1Δ31 at 37 °C. The distance between the COM (center of mass) of the Ras and α-H domains (yellow spheres) is plotted against simulation time, and structural snapshots at the indicated times are shown.

Table 1.

SAXS data collection and data analysis statistics

	Gαi1-GTPγS	Gαi1-GDP	Gαi1-apo
Data collection
Instrument	BIOSAXS1000	BIOSAXS1000	BIOSAXS1000
Beam geometry	Point	Point	Point
Wavelength, Å	1.54187	1.54187	1.54187
Q range, Å−1	0.09–0.7	0.09–0.7	0.09–0.7
Exposure time, h	0.5	0.5	0.5
Concentration range, mg/mL	2–10	2–10	2–10
Temperature, °C	20	20	20
Structural parameters*
Rg, Å [from P(r)]	22.08 ± 0.05	21.93 ± 0.05	21.98 ± 0.04
Rg, Å [from Guinier]	22.1 ± 0.2	22.6 ± 0.2	22.6 ± 0.2
Dmax, Å	72.5	69.4	67.8
Porod volume estimate, Å3	54,600	53,089	55,469
Molecular mass determination
Molecular mass Mr, Porod volume	44	43	45
Calculated monomeric Mr from sequence	37	37	37
Software used
Primary data reduction	SAXSLab v3.0.1r1	SAXSLab v3.0.1r1	SAXSLab v3.0.1r1
Data processing	Primus, Coral (atsas 2.5.0–2)	Primus, Coral (atsas 2.5.0–2)	Primus, Coral (atsas 2.5.0–2)

^*Reported for the 10-mg/mL measurement.

For an independent determination of the domain orientation in more structural detail, we recorded NMR TROSY and semi-TROSY experiments using Pf1 phage–induced aligned Gαi1Δ31 samples to extract backbone amide residual dipolar couplings (RDCs). In the semi-TROSY experiment, the signal position is shifted in the ¹H dimension by the ¹H,¹⁵N coupling constant plus the residual dipolar coupling (¹J + ¹D) compared with the TROSY experiment. RDCs are well suited for determining the relative orientation of protein domains. The RDC data were used to refine the structural models obtained with SAXS. As can be seen in Fig. 4B, the overall correlation between the experimental RDCs and the RDCs back-calculated from the refined structures is very good, yielding correlation coefficients of at least 0.95. The fitted components of the alignment tensor are shown in each subpanel. The resulting overall structural model fitted to the SAXS-based envelope in each case is shown in Fig. 4C, and the corresponding structural statistics are reported in Table 2. The main differences between these structures are slight changes in distance and rotational angle between the two subdomains. These changes in subdomain orientation lead to an overall opening of the nucleotide-binding site in the apo state and the GDP-bound state compared with the closed GTP-bound conformation. The GTP-bound model is basically identical to an existing crystal structure of GMP-PNP–bound Gα [PDB ID code 1cip (18)] in which the two domains are docked tightly against each other and the secondary structure elements are formed at the interface. This structural picture is in line with the thermal stabilities measured with CD spectroscopy in which the GTP-bound form shows an ∼25 °C higher melting point than the apo and GDP-bound forms, indicating better domain docking in the presence of GTP. To probe the degree of freedom in relative domain orientation, we further used MD simulations of apo-Gαi1Δ31 at 37 °C. In Fig. 4D, the distance between the centers of mass (COM) of the Ras and α-H domains is plotted against simulation time, and representative snapshots are shown above the plot. These simulations show that Gαi1 is very dynamic, with distances between the two COM ranging from 25–32 Å, leading to varying positions of the two domains with respect to each other. Thus, imperfections in correlation between the SAXS envelopes and the refined structures are most likely caused by the inherent dynamics of the two domains.

Table 2.

Structural statistics of Apo, GDP-, and GTP-bound Gαi1

	Apo	GDP	GTPγS
Deviations from restraints and idealized geometry
NH RDCs*	Rms = 2.27 Hz	Rms = 2.53 Hz	Rms = 2.09 Hz
	R = 0.952, Q = 0.186	R = 0.991, Q = 0.083	R = 0.974, Q = 0.149
	A = 9.4 Hz, Rh = 0.23	A = 22.6 Hz, Rh = 0.20	A = 11.05 Hz, Rh = 0.09
SAXS χ2†	1.15	1.17	1.12
Bonds, Å	2.60e−3 ± 3.11e−5	3.49e−3 ± 1.71e−4	2.51e−3 ± 4.87e−5
Angles, °	0.67 ± 0.01	0.96 ± 0.05	0.66 ± 0.01
Impropers, °	1.87 ± 0.05	3.61 ± 0.34	1.57 ± 0.04
Ramachandran map analysis, %‡
Most favored	89.1	82.8	89.7
Additionally allowed	9.4	14.9	8.5
Generously allowed	1.5	1.7	0.6
Disallowed	0.0	0.7	1.2
Structural statistics§
Backbone rmsd, Å	0.66 ± 0.11¶	0.43 ± 0.29¶	0.53 ± 0.06¶
	0.35 ± 0.05#	0.19 ± 0.05#	0.32 ± 0.04#
	0.31 ± 0.07‖	0.17 ± 0.05‖	0.26 ± 0.05‖
Heavy atom rmsd, Å	0.82 ± 0.10¶	0.46 ± 0.28¶	0.71 ± 0.08¶
	0.49 ± 0.04#	0.24 ± 0.05#	0.45 ± 0.05#
	0.52 ± 0.07‖	0.23 ± 0.05‖	0.47 ± 0.07‖

^*Determined with the program Pales (36).

^†Calculated with the SAXS module in Chimera using the FoXS server.

^‡Calculated with ProcheckNMR (37).

^§Rmsd analysis of the best 20 total energy structures:

^¶All structured parts (residues 33–348).

^#Ras domain only (residues 33–56,184–348).

^‖α-H domain only (residues 62–177).

This aspect was investigated further by fluorescence spectroscopy with MANT-labeled nucleotides (MANT-GDP and –GMP-PNP). We first probed the environment of MANT-GDP and MANT–GMP-PNP provided by Gαi1Δ31 using this assay. The fluorescence spectra of MANT are sensitive to the hydrophobicity of the surrounding environment. As a reference, we measured a spectrum of MANT–GMP-PNP in buffer solution. Without bound protein, the two MANT nucleotides give identical spectra. As can be seen in Fig. 5A, there are pronounced changes in the fluorescence emission spectra of MANT-GDP and MANT–GMP-PNP after binding to Gαi1Δ31. The emission intensity of MANT–GMP-PNP is higher, and its maximum is blue-shifted compared with MANT-GDP, a strong indication that Gαi1Δ31 entraps MANT–GMP-PNP more tightly than the GDP adduct. To probe direct contact between the MANT-nucleotides and the protein, we used FRET; we irradiated the protein at 280 nm and recorded emission spectra from 350–600 nm (Fig. 5B). The FRET peak of the bound MANT dye appears around 425 nm. Because of the lack of protein in the reference MANT–GMP-PNP sample in buffer, no FRET signal was observed. The Gαi1-MANT-GDP sample showed only slightly increased FRET efficacy. In contrast, the FRET emission of the Gαi1–MANT–GMP-PNP sample was around three times higher than that of the MANT–GDP sample, again indicating tight binding between MANT–GMP-PNP and the protein.

Fig. 5.

Nucleotide-binding properties and exchange kinetics monitored with fluorescence spectroscopy. (A) Fluorescence spectra of the free and Gαi1Δ31-bound MANT-labeled nucleotides obtained by direct excitation of the dye at 356 nm. The fluorescence intensity and emission maximum increase subsequently from the free to the GMP-PNP–bound form. (B) FRET spectra of the MANT-labeled nucleotides obtained by excitation of tryptophan side chains at 280 nm. (C) Association kinetics of 5 µM MANT-GDP to 20 µM Gαi1Δ31. (D) Dissociation kinetics of the same sample upon the addition of 50 µM GDP or GMP-PNP, respectively. (E) Association kinetics of MANT–GMP-PNP to Gαi1Δ31. (F) Dissociation of the complex by the addition of an excess of GDP or GMP-PNP is much slower than in the MANT-GDP case. Calculated K_d values for the interaction between Gαi1Δ31 and MANT-GDP and MANT–GMP-PNP are 16 and 3 µM, respectively.

We next investigated the kinetic properties of nucleotide association and dissociation. We used MANT-labeled nucleotides and observed the buildup of FRET intensity over time after the addition of Gαi1Δ31. As shown in Fig. 5C, the association of 5 µM MANT-GDP with 20 µM Gαi1Δ31 is a single-exponential process with an observed association rate constant (k_on) of 1,340 1/(s·M). This complex can be dissociated rapidly, and rebinding is prevented by a 10-fold excess of unlabeled GDP or GMP-PNP, following the same observed rate constant (k_off) of 0.022/s (Fig. 5D), yielding a calculated K_d (K_d = k_off/k_on) of 16 µM. The same association–dissociation experiment conducted with MANT-labeled GMP-PNP yields an observed association rate constant (k_on) of 260 1/(s·M) (Fig. 5E), which is considerably slower than the GDP association rate and most likely is caused by structural rearrangements within Gαi1 to accommodate the GTP fully. Dissociation of this tight complex is slow and is dependent on the nucleotide type used as a competitor. Dissociation in the presence of excess GDP is characterized by a k_off of 0.00065/s, whereas with GMP-PNP as a competitor slightly faster dissociation is observed with a k_off of 0.00075/s (Fig. 5F). The calculated K_d for GMP-PNP is 3 µM. These experiments confirm not only that the Gαi1-GTP complex is tighter than the GDP complex but also that it is characterized by markedly reduced association and very slow dissociation rates.

Nucleotide Binding Strongly Alters Gαi1 Dynamics.

To investigate the nucleotide-dependent dynamics of the Gαi1 subunit in the nanosecond-to-picosecond, millisecond-to-microsecond, and slower time scales, we performed various NMR experiments in the apo, GDP-, and GTPγS-bound forms. First, {¹H}¹⁵N-heteronuclear NOE experiments, reporting on dynamics in the nanosecond-to-picosecond time scale, were recorded on the three samples (Fig. S5). These experiments showed that GMP-PNP–bound Gαi1Δ31 is very rigid throughout the sequence. In contrast, the apo or GDP-bound forms show greater flexibility at the N-terminal β1-strand and more pronounced fluctuations in dynamics throughout the protein. To probe dynamics in the millisecond-to-microsecond time scale, we performed ¹⁵N and ¹³C CPMG relaxation dispersion experiments, which are particularly suitable for probing conformational changes such as protein-folding events (32) and motions associated with enzyme catalysis (33). Initially, we recorded ¹⁵N CPMG relaxation dispersion experiments on the apo, GDP-, and GTPγS-bound forms on backbone amide resonances in the protein. The resulting data indicate chemical exchange in the apo and GDP-bound but not in the GTPγS-bound form, where only flat dispersion profiles could be observed, as shown for His188 in Fig. 6A. The dispersion profiles of the apo and GDP-bound forms differ slightly, indicating alterations in the chemical-exchange properties of these two states. To evaluate the chemical-exchange processes of the apo and GDP sample further, we made use of the enhanced NMR signal intensity of methyl groups suitable for studying dynamics of large protein systems (34). 2D [¹³C,¹H]-heteronuclear multiple quantum coherence (HMQC) experiments of selectively Ileδ1, Leuδ2, Valγ2, Alaβ ¹H,¹³C-methyl-, and otherwise ²H,¹²C,¹⁵N-labeled Gαi1Δ31 were of excellent quality and thus are suitable for extracting chemical shift perturbations upon ligand binding and for conducting ¹³C-CPMG relaxation dispersion experiments (Fig. S6). We therefore measured single-quantum ¹³C-CPMG relaxation dispersion experiments at static magnetic field strengths of 600 and 900 MHz and obtained dispersion profiles of sufficient quality for data analysis (Fig. 6 B and C), except for parts that are severely exchange-broadened in the apo and GDP-bound states (sheet 1, helix 2), as labeled in Fig. S6A. Methyl groups in Gαi1Δ31 showing chemical exchange in the millisecond-to-microsecond time scale are color-coded in the apo and the GDP-bound form, respectively (Fig. 6 B and C). Residues involved in fast exchange, residues involved in slow exchange, and residues that are severely exchange-broadened are shown as blue, red, and yellow spheres, respectively. Regions exhibiting chemical exchange cluster around the nucleotide- and GPCR-binding sites but also are present in the α-H domain. Switch regions (switches I, II, and III) are exchange-broadened or show slow chemical exchange in the apo and GDP-bound forms. Overall, differences in exchange rates between the apo and the GDP-bound form can be detected only in the Ras domain, where the average exchange rate drops from ∼900 to 740 Hz. Dynamics of the α-H domain is not affected by GDP binding, indicating that binding of GDP to the Ras domain does not induce its docking to the α-H domain. This observation is in line with thermal denaturation analyses of Gαi1Δ31 in both forms that showed only a slight increase in the transition temperature in the GDP-bound form versus the apo form (Fig. 1C). We further analyzed populations of the ground and excited states in the millisecond-to-microsecond time scale for helix 5, the main interaction site with GPCRs, in the apo and GDP-bound forms in more detail. A comparison of the two forms shows that the population of the ground state (p_A) of residues in and surrounding helix 5 is around 5% lower in the GDP-bound state than in the apo state. Therefore, we speculate that GDP stabilizes the excited state, which possibly is similar the GTP-bound closed state in which helix 5 is tightly bound to the Ras domain. Possible effects of the G-βγ subunit on the overall dynamics of the Gα subunit remain to be shown in future studies.

Fig. S5.

{¹H}¹⁵N-heteronuclear NOE experiments of Gαi1Δ31 at a field strength of 600 MHz in the apo form and in complex with GDP or GMP-PNP, as indicated, measured at 30 °C. ¹H irradiation time was 2 s. The number of transients was 96 and 128 complex points in the indirect ¹⁵N dimension. The error bars underestimate the true uncertainty of the presented data.

Fig. 6.

Millisecond-to-microsecond dynamics of Gαi1Δ31. (A) ¹⁵N-CPMG relaxation dispersion profiles recorded at magnetic field strengths of 600 and 700 MHz. His188 of Gαi1 is shown as an example to illustrate the effect of nucleotides on the inherent dynamics of the protein. (B and C) Millisecond-to-microsecond dynamics of Gαi1 in the apo (B) and GDP-bound (C) forms extracted from ¹³C-CPMG relaxation dispersion experiments measured at 600 (black) and 900 MHz (orange) proton frequency, respectively. Residues in fast exchange are shown in blue, residues in slow exchange are shown in red, and severely exchange broadened residues are shown in yellow. (D) Populations of the ground state in the apo and GDP-bound state of helix 5 and nearby residues. The addition of GDP induces a slightly higher population in the excited state.

Fig. S6.

(A) 2D [¹H,¹³C] HMQC experiments of selective Ileδ1, Leuδ2, Valγ2, and Alaβ ¹H,¹³C-methyl– and otherwise [²H,¹²C,¹⁵N]-labeled Gαi1Δ31 in the apo form and in complex with GDP or GTPγS, as indicated, measured at 30 °C. Signals that are severely broadened in the apo and GDP samples are labeled in the GTPγS spectrum. (B) Averaged ¹H,¹³C chemical shift perturbation of GNAI1 in the apo (red) or GDP-bound (blue) form versus the GTPγS-bound form. Negative bars indicate missing signals in the apo and GDP-bound samples. (C) Gαi1 structure color-coded to show ¹H,¹³C chemical shift perturbations between the apo and the GTPγS-bound forms. The most-affected residues are located within the Ras domain and in the inner core of the α-H domain. Residues in helix 5 are also affected by GTP binding.

A closer look at the 2D TROSY spectra of the different states provides additional evidence for the existence of two conformational states of the Gα subunit in very slow exchange (Fig. 7A). We observed a second signal set for a subset of residues located in the Ras domain in both the apo and GDP-bound forms but not in the closed GTP-bound state. The α-H domain was not affected, as shown exemplarily by the single signal of the amide resonance of Gly112. All other resonances in Fig. 7A are located in the Ras domain and give rise to a second signal set in the apo and GDP-bound form, except for F191, which shows one signal in the apo form and two in the GDP-bound form. The resonance position that most likely corresponds to the GTP-bound closed conformation is indicated by the intersection of the horizontal and vertical dashed lines, implying that the closed species is also populated in the apo and GDP-bound forms. Analysis of the relative peak intensities suggests that GDP binding leads to a markedly higher population in the excited state than present in the apo form. Overall, these effects cluster to the Ras domain in general and to helix 5 in particular, and the chemical shift differences between the two states are more pronounced in the GDP-bound form than in the apo form (Fig. 7B). Helix 5 is the main interaction site with a GPCR, and inherent changes in its conformation therefore are crucial for the binding event. We analyzed the relative heights of the corresponding NMR signals and calculated the populations of the ground state, which are ∼65% for the apo form and ∼35% for the GDP-bound form (Fig. 7C, Left). Next we investigated the influence of an activated GPCR on the relative population of the ground state. We used nanodiscs incorporating rat neurotensin receptor-1 variant HTGH4 L167R (28), performed 2D [¹⁵N,¹H]-TROSY experiments, and analyzed the change in NMR signal intensity of the ground and excited states. The population of the ground state in the apo form is increased by 5% upon GPCR binding; this increase is less pronounced than the 20% increase determined for the GDP-bound state (Fig. 7C, Right). These experiments show that interaction with an activated GPCR is mediated by the Gα ground state. The addition of nucleotides leads to the preferred or complete formation of a second conformation. Binding of a GPCR enhances the population of the ground state, and this effect is more pronounced with the GDP-bound Gα subunit, thus favoring the nucleotide-exchange process. As shown in Fig. 3, the closed Gα-GTP complex does not interact with a GPCR, and, in turn, the apo state shows the most pronounced binding effects upon the addition of a GPCR. We assume that the presence of the Gβγ subunit will lead to a more defined conformational state of the Gα subunit in complex with GDP, and the GPCR might shift the conformational equilibrium toward a low-affinity ground state that also is present in apo Gα. However, because of current limitations in investigating such a large complex with solution-state NMR, this question remains to be addressed in future studies.

Fig. 7.

Gα conformational states are modulated by nucleotide and GPCR binding. (A) Regions of 2D [¹⁵N,¹H]-TROSY spectra of Gα in the apo (black), GDP-bound (orange), and GMPPNP-bound (blue) states. The apo and GDP-bound forms show two conformational states in slow exchange (longer than seconds) that give rise to the appearance of a second signal set in the NMR experiment. The GMP-PNP–bound state exists in one defined conformation and consequently gives rise to only one set of NMR signals. (B) C.S.P. (color coding: gray to red) between the two Gα NMR signal sets are mapped onto the structure. These data show that only the Ras domain is affected and that chemical shift perturbations are more pronounced in the GDP-bound form. (C) Population of the ground state in the apo and GDP-bound forms. The addition of GDP leads to a significantly decreased population of the ground state. (D) The influence of GPCR-binding on the population of the ground state. This influence is less pronounced in the apo state than in the GDP-bound state. The minus and plus signs indicate the absence or presence, respectively, of nanodisc-incorporated and -activated neurotensin receptor variant HTGH4 L167R.

Discussion

In this study our goal was to conduct a detailed survey of the structure and internal mobility of Gαi1 in complex with different nucleotides and an activated GPCR in phospholipid nanodiscs by solution NMR, CD, and fluorescence spectroscopy, SAXS, and MD simulations. We were able to assign the NMR backbone resonances of apo Gαi1Δ31, as well as the GDP- and GTP-bound forms. These NMR assignments could be used in combination with SAXS experiments to extract differences in chemical shift and to obtain NMR RDCs for the refinement of the relative orientations of the Ras and α-H domains of Gαi1 in the various ligand-bound forms. The results revealed a nucleotide-dependent conformational change in Gαi1, in which GTP induces a very tight and stable domain-docked state (Fig. 4). In contrast, in the apo form both domains adopt a more open topology that permits binding of GDP or GTP to the nucleotide-binding site.

The relative changes in domain orientation of Gαi1 domains shown here are less pronounced than described in a previous crystal structure of a stimulatory G protein in complex with β2-adrenergic receptor (3), which also contained a nanobody bound to the Gα–Gβ interface, but are in better agreement with more recent studies based on molecular modeling and DEER spectroscopy (13–15). In addition to these earlier studies, we were able to probe the dynamics and the populations of the associated conformational states. We could clearly show that the populations of these states change upon the addition of GDP and that binding of GTP completely abolished internal dynamics as measured by our assays. We could detect these effects in the millisecond-to-microsecond as well as in the minute-to-second time scale. Of particular interest is the finding that the major conformational state of the apo state is less populated in the GDP-bound state and is completely absent in the GTP-bound state. In addition, the appearance of this ground state correlates well with the binding behavior of Gα to an activated GPCR. The relative intensity of NMR signals corresponding to this state increases when Gα is in complex with a GPCR. This result suggests that only one conformational species of Gα, highly populated in the apo form but also present to some extent in the GDP-bound form, is capable of high-affinity binding to an activated GPCR. The addition of GDP pushed the equilibrium toward the excited (lower-affinity) state. This tendency can be reverted by binding to a GPCR, providing evidence that a GPCR promotes nucleotide exchange by stabilizing the nucleotide-free conformation of a Gα subunit. In contrast, the GTP-bound form is present in a single closed and rigid conformation whose population cannot be altered by a GPCR because of a lack of interaction. These findings demonstrate a tight interplay between nucleotide and GPCR binding mediated by allosteric structural changes. Residues that show two conformations in slow exchange cluster to the upper face of the Ras domain (Fig. 7B), and basically the same interface is affected by GPCR binding (Fig. 3). Conformational changes in the Gα subunit upon binding to a GPCR have been reported in a previous EPR study (12) and crystal structure (3). These reports suggest that helix 5 must undergo a translational and rotational motion to interact with the receptor. A more recent MD simulation study provides further insights, suggesting that this motion is restricted by GDP binding (15). Interaction with a GPCR leads to a disorder-to-order transition within the C-terminal part of helix 5, as recently summarized for existing G-protein structures (2).

Our findings can be included in the well-explored activation cycle of a heterotrimeric G protein mediated by GPCR in which, as a first step, GDP-bound heterotrimeric G protein interacts with an activated receptor. The receptor then pushes the Gα conformation slightly toward a low-affinity state for GDP. The GDP-bound Gα subunit is quite flexible, and we speculate that even in the complex with the βγ subunit there would be enough conformational space to mediate such slight structural changes. The resulting apo heterotrimeric G protein eventually binds to GTP, leading to subunit dissociation and loss of affinity with the receptor. The change in affinity in the Gα subunit can be directly correlated with its conformational states. The apo form showing the highest affinity for the receptor is present mostly in the ground state, as determined by NMR (Fig. 7). In the GDP-bound form, both, the ground and the excited state occur, resulting in a reduced affinity for the receptor. Finally, in complex with GTP, the Gα subunit lacks any dynamics and exists exclusively in the excited state, thus losing affinity with the GPCR.

Methods

Protein production was done in E. coli as described previously (17). Proteins were purified with NiNTA and size-exclusion chromatography. We used a FRET assay to extract binding affinities between Gα and MANT-labeled GDP or GMP-PNP nucleotides. Secondary structure estimation and thermal stability screens of Gα preparations were done with CD spectroscopy. NMR assignment experiments with U-²H,¹³C,¹⁵N–labeled Gα were recorded in a NUS manner (21, 22). Dynamical parameters were extracted from ¹⁵N- and ¹³C-CPMG relaxation dispersion experiments (35). Structural changes upon nucleotide binding were detected with SAXS, and the conformational flexibility of Gα was further determined with MD simulations. More experimental details can be found in SI Methods.

SI Methods

Protein Production.

The cDNA of the human Gαi1 gene (GenBank accession no. BC026326) was inserted into an in-house–modified pQE30 vector (Qiagen) using PCR methods, resulting in a DNA construct encoding (N terminal-to-C terminal) a His₆-tag, GB1, a tobacco etch virus (TEV) protease cleavage site, and Gαi1. For NMR optimization, another construct, Gαi1Δ31, was made in which the first 31 amino acids were deleted. Protein expression was achieved in the E. coli Tuner (DE3) pLacI strain (Novagen). Cells were shaken at 37 °C until an OD₆₀₀ of 0.6 was reached, at which point the temperature was lowered to 20 °C before the addition of 150 µM IPTG for the induction of protein production. Cells were shaken for additional 20 h and harvested by centrifugation at 5,000 × g. Cells were resuspended in 50 mM NaPi (pH 8.0), 200 mM NaCl, 10 mM β-mercaptoethanol, 1% Triton X-100, and 10% glycerol (buffer A) plus protease inhibitor (Roche). Cells were lysed by lysozyme treatment and sonication and were spun down at 50,000 × g for 45 min. Cell extract then was applied to an Ni-NTA column, was washed first with buffer A with 0.5% Triton X-100 and then with buffer A containing 10 mM imidazole and no Triton X-100, and was eluted with buffer B [50 mM NaPi (pH 8.0), 200 mM NaCl, 10 mM β-mercaptoethanol, 10% glycerol, 300 mM imidazole, 20 µM GDP]. Protein was buffer exchanged into 20 mM NaPi (pH 7.5), 50 mM NaCl, 3 mM DTT, 0.5 mM EDTA, 10% glycerol and was digested with TEV protease (1 A₂₈₀ for 100 A₂₈₀ protein). After TEV digestion, the buffer was replaced with 20 mM NaPi (pH 7.8), 200 mM NaCl, 5 mM β-mercaptoethanol, 5% glycerol and was reapplied to an Ni-NTA column. The collected flow-through and wash fractions were pooled, concentrated using 30 kDa MWCO centrifugal devices (Millipore), and applied to a S75 Hi Load 16/600 size-exclusion column (GE Healthcare) equilibrated in 20 mM NaPi (pH 7.0), 50 mM NaCl, 5 mM MgCl₂, 5 mM DTT. The protein typically elutes from the column at ∼70 mL (124 mL column volume).

The neurotensin-bound, activated neurotensin receptor 1 variant HTGH4 L167R (28) was produced and purified as described (27, 29). Nanodisc incorporation was achieved by adding decyl maltoside solubilized receptor (concentration = 20 µM) to the nanodisc assembly mixture consisting of 200 µM MSP1D1 (38) and 5 mM POPC lipids solubilized with 20 mM sodium cholate in 20 mM Tris⋅HCl (pH 7.5), 100 mM NaCl, 0.5 mM EDTA, and 2 mM DTT. Nanodisc assembly then was initiated by dialysis using a 20-kDa MWCO cassette. GPCR-containing nanodiscs were isolated with SP Sepharose and S200 size-exclusion chromatography. Neurotensin peptide (1 µM) was added to the size-exclusion chromatography buffer [20 mM sodium phosphate (pH 7.0), 50 mM NaCl, 5 mM DTT] to keep the receptor in an activated state.

CD Spectroscopy.

CD spectra and thermal melting curves were recorded on a Jasco J-715 spectrometer using a cuvette with a 1-mm path length. The protein concentration was 10 µM in 10 mM NaPi (pH 7.0), 20 mM NaCl, 1 mM MgCl₂, and 1 mM DTT. The ligand concentration was 50 µM. CD spectra were recorded from 190 to 260 nm using five scans for each sample. The response time was 1 s, and the bandwidth was 1 nm. Thermal scans were run with a heating rate of 60 °C/h. Curve fitting was done with an equation described in ref. 39.

Fluorescence Experiments.

Fluorescence experiments were conducted with a Horiba-Jobin Fluoromax-4 instrument. To determine the apparent K_d between GDP and GMP-PNP to Gαi1, we performed a FRET-based binding assay in which 1 µM of MANT-labeled nucleotides (Invitrogen) in buffer [20 mM Hepes (pH 7.0), 100 mM NaCl, 2 mM DTT] were titrated with increasing amounts of Gαi1. During the course of the titration, protein aromatic residues were excited by irradiation at 280 nm (2-nm bandwidth), and the FRET emission signal intensity of the MANT label was detected at 425 nm (5-nm bandwidth). Fluorescence emission and FRET spectra were recorded between 350 and 600 nm and between 320 and 550 nm, respectively.

SAXS Experiments.

Measurements in SAXS experiments were performed with Gαi1Δ31 samples on a Rigaku BIOSAXS-1000 instrument using a Pilatus 100K detector (at a distance from the sample of 480 mm) with a HF007 microfocus generator equipped with a Cu target at 40 kV and 30 mA. Transmissions were measured with a photodiode beam stop, and q-calibration was made by a silver-behenate measurement. Samples of 10, 5, and 2 mg/mL protein containing 1 mM of GTPγS, GDP, or no nucleotide were measured. All samples were filtered using a 0.1-μM filter (Whatman) immediately before data collection. Thirty microliters of each sample were loaded in a capillary (1-mm diameter) and exposed to an X-ray beam for 30 min at 20 °C. Circular averaging was done with the Rigaku SAXSLab software v 3.0.1r1. Background subtraction was done within the Origin program (Microcal). Measurements of different concentrations were normalized to exclude concentration-dependent effects. The 10-mg/mL samples then were used for the rigid body modeling with CORAL, ATSAS package v 2.5.0-2 (36). We used the crystal structure of GTP-bound Gα (PDB ID: 1cip) for our modeling calculations. To allow the unrestricted domain movement required to optimize the correlation between experimental and back-calculated SAXS scattering curves, two flexible linkers (five and six amino acid residues in length, respectively) were defined between the Ras and α-H domains. Five independent structural models were calculated, and the model with the lowest χ² value was used for further RDC refinement.

NMR Assignment.

Sequence-specific assignment of backbone ¹HN, ¹⁵N, ¹³Cα, ¹³Cβ, and ¹³C′ resonances was first carried out with a 200 µM Gαi1Δ31 + 3 mM GMP-PNP sample labeled with ²H, ¹³C, and ¹⁵N. Triple-resonance experiments used for this purpose include TROSY-based HNCA, HN(CO)CA, HNCACB, HN(CA)CO, and HNCO (20). Additionally, the assignments were validated with a 3D [¹H,¹H]-NOESY-¹⁵N–edited TROSY spectrum recorded with an NOE mixing time of 200 ms. These experiments were recorded in an NUS manner using Poisson-gap sampling (21) at 30 °C on a 600-MHz Bruker spectrometer equipped with a cryogenic TXI probe. Processing of NUS spectra was done with NMRPipe (40) and IST reconstruction (22). Resonance assignments were obtained for ∼85% of all backbone and ¹³Cβ resonances of the GMP-PNP complex. The same set of experiments was applied to assign resonances in the apo and GDP-bound forms and in the GTPγS complex. Data analysis was done with Sparky 3 software (T. D. Goddard and D. G. Kneller, University of California, San Francisco). Chemical shift perturbations for both ¹⁵N and ¹³C correlation experiments were calculated using the empirical formula Δ_av= [(ΔdH² + (ΔN/5)²)/2]^0.5.

RDCs and Calculation of Structural Models.

Weak alignment of ²H,¹⁵N-labeled Gαi1Δ31 was induced by the addition of 8 mg/mL Pf1 phage medium (Asla Biotech). RDCs were measured with TROSY and semi-TROSY experiments in the isotropic and aligned state. RDC values were extracted with the software Sparky.

For RDC refinement, we used the best (lowest χ value) structural model obtained with SAXS in the apo and the GDP-bound form. First, we used the obtained SAXS structural models to back-calculate NOE restraints between heteroatoms in the protein to fix the secondary and tertiary structures of the individual domains during subsequent structure-calculation steps. No restraints between the two domains (Ras and α-H) were included at this point. We then performed structure calculation runs using XPlor-NIH (41) and standard protocols for refinement. RDC back-calculation and fitting of the alignment tensor was done with the program Pales (42). The structures showing the best RDC correlations, i.e., R-factor, were further ranked by their agreement with the initial SAXS scattering curve using Chimera (University of California, San Francisco) (43). The overlay between the final structures and the envelope determined with SAXS using the program DAMMIN [ATSAS package, European Molecular Biology Laboratory Hamburg (EMBL Hamburg) (36)] was done with the program SUBCOMB [ATSAS package, EMBL Hamburg (36)].

NMR Relaxation Dispersion Measurements.

Backbone ¹⁵N single-quantum CPMG relaxation dispersion experiments were performed on 600-MHz and 700-MHz Agilent DD2 spectrometers equipped with a third-generation cryogenic probe. The CPMG dispersion experiment measures the modulation of the transverse relaxation rate [R₂ (1/τ_cp)] as a function of the delay (τ_cp) between 180° rf pulses. As shown in the data analysis, the R₂ versus τ_cp data can be fitted to determine equilibrium populations (p_a, p_b), kinetics (k_ex = k₁ + k₋₁), and chemical-shift differences (Δω) between the two interconverting conformations. The CPMG experiments use a 20-ms constant-time ¹⁵N relaxation delay with τ_cp spacing corresponding to CPMG-based rf field strengths, ν(CPMG), ranging from 100–900 Hz (44). All spectra were recorded at 30 °C with 0.25 mM [²H,¹⁵N]-labeled Gαi1Δ31 in the apo form or in complex with 3 mM GDP or GTPγS, respectively. One hundred twenty-eight complex data points were recorded in the indirect dimension with 96 scans per increment.

For probing side-chain dynamics, a ¹³C single-quantum CPMG constant-time (20 ms) relaxation dispersion experiment featuring HMQC-based indirect chemical shift evolution optimized for large proteins (35) was recorded on selectively Ileδ1, Leuδ2,Valγ2, Alaβ [¹H,¹³C]-methyl- and otherwise [²H,¹²C,¹⁵N]-labeled Gαi1Δ31 in the apo and GDP-bound forms on Bruker Avance III or HD spectrometers with cryogenic probes operating at proton frequencies of 600 and 900 MHz. CPMG-based RF field strengths ranged from 75 to 975 Hz. For methyl spectra, 96 complex data points were recorded in the indirect dimension with 64 scans. The spectra with different τ_cp values were recorded in an interleaved fashion, and duplicate datasets were recorded at two τ_cp values for error analysis.

CPMG data were analyzed using the program Relax (45). Initial starting values were determined by grid search, and data subsequently were fitted to motional models [no exchange, fast exchange (46), and a general two-site exchange model (47) for all time scales, and to a two-site numerical solution (44)]. Each spin system was assigned to the best-fitting model based on χ² calculations. Finally, the statistical significance of each model was estimated with Monte-Carlo simulations.

MD Simulations.

The MD simulation setup with Gαi1Δ31 in its apo form was done with VMD (48). An MD simulation of 1-µs duration was carried out at 310 K with the program NAMD (49) using standard parameters settings. Analysis of the trajectory was performed with VMD or Gromacs (37). Figures were prepared with VMD or Chimera (University of California, San Francisco) (43).