Structural Determinants of Affinity Enhancement between GoLoco Motifs and G-Protein α Subunit Mutants

Abstract

GoLoco motif proteins bind to the inhibitory G_i subclass of G-protein α subunits and slow the release of bound GDP; this interaction is considered critical to asymmetric cell division and neuro-epithelium and epithelial progenitor differentiation. To provide protein tools for interrogating the precise cellular role(s) of GoLoco motif/Gα_i complexes, we have employed structure-based protein design strategies to predict gain-of-function mutations that increase GoLoco motif binding affinity. Here, we describe fluorescence polarization and isothermal titration calorimetry measurements showing three predicted Gα_i1 point mutations, E116L, Q147L, and E245L; each increases affinity for multiple GoLoco motifs. A component of this affinity enhancement results from a decreased rate of dissociation between the Gα mutants and GoLoco motifs. For Gα_i1^Q147L, affinity enhancement was seen to be driven by favorable changes in binding enthalpy, despite reduced contributions from binding entropy. The crystal structure of Gα_i1^Q147L bound to the RGS14 GoLoco motif revealed disorder among three peptide residues surrounding a well defined Leu-147 side chain. Monte Carlo simulations of the peptide in this region showed a sampling of multiple backbone conformations in contrast to the wild-type complex. We conclude that mutation of Glu-147 to leucine creates a hydrophobic surface favorably buried upon GoLoco peptide binding, yet the hydrophobic Leu-147 also promotes flexibility among residues 511–513 of the RGS14 GoLoco peptide.

Introduction

Protein-based therapeutics and tools are increasingly utilized for manipulating cell biology (1–5). A vital aspect of their effective function is affinity for a target protein or ligand. An effective approach for enhancing the affinity of protein/protein interactions is structure-based computational prediction of affinity-enhancing point mutations (6–9). The development of reliable in silico procedures promises to provide a more efficient means of protein design than labor-intensive combinatorial screening techniques such as phage display or ribosome display (10–12). Previously, a structure-based protocol for predicting affinity-enhancing point mutations was developed (7), in part using as a template the high-resolution crystal structure of the heterotrimeric G-protein subunit Gα_i1 in complex with the RGS14 (regulator of G-protein signaling 14) GoLoco motif (13). Residues at the protein/peptide binding surface were sequentially mutated to nonpolar amino acids in silico, producing 33 predicted affinity-enhancing mutations. Four of six experimentally tested mutations to either Gα_i1 or the GoLoco motif enhanced binding affinity, as predicted by Rosetta (7).

As a critical component of seven-transmembrane domain receptor signaling, Gα subunits normally cycle between an inactive GDP-bound state and an active GTP-bound state (reviewed in Refs. 14 and 15). GoLoco motif proteins such as LGN, AGS3, RGS12, and RGS14 act as guanine nucleotide dissociation inhibitors, specifically for the adenylyl cyclase inhibitory subclass of Gα subunits (reviewed in Refs. 16 and 17). GoLoco motifs prevent GDP dissociation by binding across the Ras-like and all-helical domains of Gα, as well as by contributing an arginine side chain from the highly conserved (E/D)QR GoLoco triad (16) to contacts made to the bound guanosine diphosphate (13). Not only are GoLoco motif-containing proteins thought to modulate seven-transmembrane domain receptor signaling in various physiological responses (18–20), but these proteins are also critical to cell fate determinant sorting and cell division processes in multiple contexts across multiple species (21–24). Highlighting the value of mutation-derived affinity modulation to cell biological applications, a Gα_i1^N149I mutant with selective loss of GoLoco motif affinity has recently provided direct evidence for the Gα_i/GoLoco protein interaction in these cell division processes (25, 26).

We previously predicted that mutation to leucine of one of three polar residues, Glu-116, Gln-147, or Glu-245, on the surface of Gα_i1 enhances binding of Gα_i1-GDP to the RGS14 GoLoco motif (7, 27). Although the algorithms implemented in the Rosetta program predict a favorable change in the Gibbs free energy of binding, our computational tools do not provide a clear mechanism of affinity enhancement. Knowledge of changes in the thermodynamic and/or kinetic parameters of binding, such as the rate of dissociation, will be important in the future application of these computational design strategies to biological and therapeutic questions. Here, we assessed the effects of these predicted Gα_i1/GoLoco affinity-enhancing mutations on the kinetics and thermodynamics of binding using fluorescence polarization, isothermal titration calorimetry, and surface plasmon resonance. Although mutations that promote binding to a target protein are desirable, protein design applications often require the preservation of other protein functions. Accordingly, we also determined the effects of these three affinity-enhancing mutations on the nucleotide cycle of Gα_i1, including GTP binding and hydrolysis. To further elucidate the molecular mechanism of affinity enhancement for one mutant, we determined the x-ray crystal structure of Gα_i1^Q147L-GDP bound to the RGS14 GoLoco motif. In addition, we investigated the conformational flexibility of the GoLoco peptide backbone and side chains surrounding the Leu-147 residue of the Gα_i1 mutant using Monte Carlo simulations.

EXPERIMENTAL PROCEDURES

Chemicals and Other Assay Materials

Unless otherwise noted, all chemicals were the highest grade available from Sigma or Fisher Scientific. Peptides were synthesized by Fmoc (N-(9-fluorenyl)methoxycarbonyl) group protection, purified via HPLC, and confirmed using mass spectrometry by the Tufts University Core Facility (Medford, MA). Peptide sequences are as follows: FITC-GPSM2(GL2), FITC-β-alanine-NTDEFLDLLASSQSRRLDDQRASFSNLPGLRLTQNSQS-amide; FITC-RGS12 GoLoco, FITC-β-alanine-DEAEEFFELISKAQSNRADDQRGLLRKEDLVLPEFLR-amide; FITC-RGS14 GoLoco, FITC-β-alanine-SDIEGLVELLNRVQSSGAHDQRGLLRKEDLVLPEFLQ-amide; and RGS14 GoLoco, DIEGLVELLNRVQSSGAHDQRGLLRKEDLVLPEFLQ-amide.

Protein Expression and Purification

For biochemical experiments, N-terminally truncated, hexahistidine-tagged WT and mutant Gα_i1 proteins were expressed and purified as described previously (28). Point mutations were made using QuikChange site-directed mutagenesis (Stratagene, La Jolla, CA). For crystallization, the Gα_i1^Q147L mutant was expressed and purified as described previously (29). Purified Gα_i1^Q147L was concentrated to 15 mg/ml and stored in crystallization buffer (10 mm Tris (pH 7.5), 1 mm magnesium chloride, 5% (v/v) glycerol, and 5 mm DTT).

Fluorescence Polarization Measurements

All polarization experiments were conducted using a PHERAstar microplate reader (BMG Labtech, Offenburg, Germany). Excitation wavelength was 485 ± 6 nm, and emission was recorded at 520 ± 15 nm. Gains of the parallel and perpendicular channels were calibrated so that 350 pm FITC probe alone had a polarization of ∼35 milli-Polarization units (mP) for each experiment. All experiments were conducted at 25 °C in triplicate using 96-well black-bottom plates (Costar, Corning, NY). The final volume of each well was brought to 180 μl with PheraBuffer (10 mm Tris (pH 7.5), 150 mm NaCl, 100 μm GDP, and 0.05% (v/v) Nonidet P-40). For experiments conducted in the presence of the transition-state mimetic, aluminum magnesium tetrafluoride (AMF),⁴ an additional 10 mm MgCl₂, 10 mm NaF, and 30 μm AlCl₃ were added to the standard PheraBuffer. Polarization was determined from raw intensity values of the parallel and perpendicular channels using PHERAstar Version 1.60 software (BMG Labtech). Binding isotherms were fit and equilibrium dissociation constants were determined using GraphPad Prism Version 5.0 as described previously (28).

Surface Plasmon Resonance Assays

Optical detection of protein/protein interactions by surface plasmon resonance was performed using a Biacore 3000 (GE Healthcare). Carboxymethylated dextran (CM5) sensor chips (GE Healthcare) with covalently bound anti-GST antibody surfaces were created as described previously (30). The GST-GoLoco fusion proteins GST-RGS12 (amino acids 1184–1228) (29), GST-RGS14 (amino acids 497–532) (29), and GST-GPSM2(GL2) (31), as well as GST alone (serving as a negative control), were loaded onto separate surfaces to levels of 1372, 963, 1290, and 1061 resonance units (RU), respectively. All experiments were performed with the sensor surface, fluidics, and pump equilibrated with Biacore running buffer (10 mm HEPES (pH 7.4), 150 mm NaCl, 0.05% (v/v) Nonidet P-40, 100 μm GDP, and 50 μm EDTA). Injections of 200 μl of WT or mutant Gα_i1 at 50 μm for kinetic analyses were performed using the KINJECT command with a dissociation phase of 2000 s at a flow rate of 40 μl/min. The surface was regenerated between runs by removal of bound Gα_i1 using 325-μl injections of AMF buffer (Biacore running buffer with an additional 10 mm MgCl₂, 10 mm NaF, and 30 μm AlCl₃). Nonspecific binding to the GST only-loaded surface was subtracted from each curve (BIAevaluation Version 3.0 software, GE Healthcare). To establish off-rates, dissociation-phase sensorgrams were normalized to percent bound, and the averages of triplicate experiments were fit to a single exponential function for dissociation without constraints using Prism Version 5.0.

Nucleotide Binding and Hydrolysis Assays

The [³⁵S]GTPγS filter-binding assay used to measure rates of spontaneous GDP release was conducted as described previously (32). Briefly, WT or mutant Gα_i1 (100 nm) was incubated in assay buffer (20 mm Tris (pH 7.5), 1 mm C₁₂E₁₀, 1 mm DTT, 1 mm EDTA, 100 mm NaCl, and 25 mm MgCl₂) or nonspecific buffer (assay buffer plus 100 μm GTPγS) at 20 °C. Radioligand binding was initiated by the addition of 6.25 nm [³⁵S]GTPγS. At timed intervals, 100-μl aliquots of the reaction were filtered through nitrocellulose membranes and washed four times with 7.5 ml of ice-cold wash buffer (20 mm Tris (pH 7.5), 120 mm NaCl, and 25 mm MgCl₂). Filters were dried under a vacuum before measuring β-radiation via liquid scintillation counting.

Intrinsic GTP hydrolysis rates of Gα_i1 subunits were assessed by monitoring the production of ³²P-labeled inorganic phosphate during a single round of GTP hydrolysis as described previously (33). In brief, WT or mutant Gα_i1 (100 nm) was incubated for 10 min at 20 °C with 1 × 10⁶ cpm of [γ-³²P]GTP (specific activity of 6500 dpm/Ci; PerkinElmer Life Sciences) in reaction buffer (50 mm Tris (pH 7.5), 0.05% (v/v) C₁₂E₁₀, 1 mm DTT, 10 mm EDTA, 100 mm NaCl, and 5 μg/ml BSA). The nucleotide-loaded protein was then chilled on ice for 5 min prior to initiation of hydrolysis by the addition of 10 mm MgCl₂ and 100 μm GTPγS (final concentrations). At timed intervals, 100-μl aliquots were quenched in 900 μl of charcoal slurry (5% (w/v) activated charcoal in 50 mm H₃PO₄ (pH 3.0)) and centrifuged at 4 °C for 10 min at 4000 × g. Subsequently, the β-radiation of 600-μl aliquots of the supernatant was counted via liquid scintillation.

Intrinsic Tryptophan Fluorescence Measurements of Gα Activation

Changes in tryptophan fluorescence of Gα_i1 subunits were measured to assess activation by the transition-state mimetic GDP·AlF₄⁻. Activation of Gα_i1 results in translocation of Trp-211, a residue in the conformationally flexible switch II region, to a hydrophobic pocket; the change in environment results in an increased quantum yield of Trp-211 fluorescence (34, 35). Tryptophan fluorescence of WT or mutant Gα_i1 (1 μm) was measured using a POLARstar Omega plate reader (BMG Labtech) in 187 μl of assay buffer (100 mm NaCl, 100 μm EDTA, 2 mm MgCl₂, 2 μm GDP, and 20 mm Tris (pH 8.0)). Intrinsic fluorescence was measured using a 280-nm excitation filter and 350-nm emission filter (each with 10-nm cutoffs). After a 20-s equilibration at ambient temperature, activation was induced by sequential injections of 8 μl of 0.5 m NaF and 5 μl of 1.2 mm AlCl₃ to each well. Tryptophan fluorescence was recorded for 2 min following activation. Fluorescence intensity traces shown represent the average of experiments performed in triplicate.

Isothermal Titration Calorimetry

Isothermal titration calorimetry (ITC) was performed using a MicroCal^TM Auto-ITC200 system (GE Healthcare). WT and mutant Gα_i1 subunits were subjected to size-exclusion chromatography in ITC buffer (50 mm HEPES (pH 7.5), 150 mm NaCl, 10 μm GDP, 5% (v/v) glycerol, and 0.005% (v/v) β-mercaptoethanol) and subsequently concentrated to 30 μm using a VivaSpin 20 ultrafiltration spin column (Sartorius Stedim Biotech, Göttingen, Germany) at 4 °C and 3000 × g. Lyophilized RGS14 GoLoco peptide was dissolved in ITC buffer to a concentration of 350 μm. Titrations were performed at 25 °C and consisted of 20 2-μl injections of peptide into WT Gα_i1 or Gα_i1^Q147L solutions at 250-s intervals. Binding isotherms and curve fittings were obtained using Origin for ITC analysis software (MicroCal). Enthalpy of binding was determined for three titrations each of RGS14 GoLoco peptide into solutions of wild-type Gα_i1 or Gα_i1^Q147L, and average values were compared with a Student's t test using Prism Version 5.0. Because ITC sensitivity did not allow us to approach a Gα_i1 or peptide concentration near the K_D, Gibbs free energy of binding was calculated from the more accurate fluorescence polarization binding affinity measurements, using ΔG = −R(T)ln(1/K_D), where R is the universal gas constant and T is temperature in Kelvin. Entropy of binding was then estimated with ΔS = (ΔH − ΔG)/T, where ΔH is the average enthalpy of binding derived from isothermal titrations.

Crystallization and Structure Determination

The complex of Gα_i1^Q147L and RGS14 GoLoco peptide was obtained by mixing Gα_i1^Q147L and GoLoco peptide to concentrations of 400 and 600 μm, respectively, in 10 mm Tris (pH 7.5), 1 mm MgCl₂, 10 μm GDP, 5% (v/v) glycerol, and 5 mm DTT. Crystals of the Gα_i1^Q147L·GDP/GoLoco peptide heterodimer were obtained by vapor diffusion from hanging drops containing a 1:1 (v/v) ratio of protein/peptide solution to well solution (1.5–1.6 m ammonium sulfate, 100 mm MgCl₂, 100 mm sodium acetate (pH 4.8), and 10% (v/v) glycerol). Crystals (∼200 × 200 × 50 μm) formed in 1–4 days at 18 °C exhibited the symmetry of space group P222₁ (a = 70.3, b = 83.7, and c = 190.1 Å, α = β = γ = 90°) and contained two Gα_i1^Q147L/GoLoco heterodimers in the asymmetric unit. For data collection at 100 K, crystals were transferred to well solution supplemented with 25% saturated sucrose for ∼30 s, followed by plunging into liquid nitrogen. A native data set was collected at the SER-CAT 22-BM beamline at the Advanced Photon Source (Argonne National Laboratory). Data were processed using the HKL-2000 program (36). The crystal structure of the WT Gα_i1/RGS14 GoLoco heterodimer (Protein Data Bank code 2OM2 (7)), excluding the GoLoco motif peptide, nucleotide, waters, and magnesium, was used as a search model for molecular replacement using the Phaser program (37). Refinement was carried out using the Refmac5 program (38), consisting of conjugate gradient minimization and refinement of individual atomic displacement and translation-libration-screw parameters, interspersed with manual revisions of the model using the Coot program (39). The current model contains two Gα_i1^Q147L·GDP/GoLoco peptide heterodimers in the asymmetric unit. For data collection and refinement statistics and a list of residues that could not be located in the electron density, see supplemental Table S1. All structural images were made with PyMOL (Schrödinger LLC, Portland, OR) unless indicated otherwise.

RESULTS AND DISCUSSION

Comparing Different Point Mutations in Gα_i1 That Enhance Affinity for GoLoco Motifs

Three polar-to-nonpolar point mutations in Gα_i1 (E116L, Q147L, and E245L) have been described individually (7, 27) as enhancing GoLoco motif binding affinity; the location of these three positions within Gα_i1 is highlighted in supplemental Figs. S1 and S2. To determine and compare directly the affinities of these three separate Gα_i1 point mutants (Gα_i1^E116L, Gα_i1^Q147L, and Gα_i1^E245L) for the GoLoco motif of RGS14, we used a synthetic FITC-RGS14 GoLoco motif peptide probe in fluorescence polarization (FP) binding assays similar to those performed with RGS12 and WT Gα_i1 (28). Holding the peptide concentration at 350 pm, we added increasing concentrations of WT or mutant Gα_i1 protein (30 pm to 200 nm) and generated equilibrium binding isotherms (Fig. 1). The dissociation constants (K_D) were determined to be 14, 1.6, 2.3, and 0.69 nm for GDP-bound Gα_i1^WT, Gα_i1^E116L, Gα_i1^Q147L, and Gα_i1^E245L, respectively. These results confirm our earlier reports of affinity enhancement obtained with a tenascin-GoLoco fusion FP probe (7) and, for Gα_i1^E245L, mutation in the context of other affinity-modifying changes (27). Consistent with the selective binding of GoLoco motifs to ground-state Gα_i subunits (i.e. Gα_i·GDP) (reviewed in Ref. 16), the transition-state mimetic form (i.e. GDP·AlF₄⁻-bound) of Gα_i1^WT, Gα_i1^E116L, and Gα_i1^Q147L did not demonstrate any appreciable binding to the RGS14 GoLoco motif (Fig. 1); however, the aluminum tetrafluoride-activated Gα_i1^E245L was observed to bind the FITC-RGS14 GoLoco FP probe with a dissociation constant of 47 nm.

FIGURE 1.

WT Gα_i1 and three affinity-enhanced mutants bind to the RGS14 GoLoco motif in a nucleotide-dependent manner. Binding isotherms were generated using FP measurements by increasing the concentration of Gα_i1 protein while maintaining the concentration of FITC-labeled RGS14 GoLoco motif peptide constant at 350 pm. All experiments were performed in triplicate. The error bars represent S.E. A, Gα_i1^WT·GDP binds the RGS14 GoLoco motif with a dissociation constant (K_D) of 14 nm (95% CI, 11–17 nm). No appreciable binding was observed when the Gα_i1 protein was in its transition-state mimetic form, created with AMF. B, Gα_i1^E245L binds the RGS14 GoLoco motif with K_D values of 0.69 nm (95% CI, 0.61–0.77 nm) and 47 nm (95% CI, 37–56 nm) in its GDP and GDP·AMF forms, respectively. C, Gα_i1^E116L·GDP binds the RGS14 GoLoco motif with a K_D of 1.6 nm (95% CI, 1.4–1.7 nm), whereas no appreciable binding is observed with the AMF-activated Gα_i1^E116L. D, Gα_i1^Q147L·GDP binds the RGS14 GoLoco motif with a K_D of 2.3 nm (95% CI, 1.9–2.6 nm), whereas no appreciable binding is observed in the AMF-activated form. mP, milli-Polarization units.

These affinity-enhancing mutations of Gα_i1 were computationally predicted based on the known structure of the WT Gα_i1/RGS14 GoLoco motif heterodimer (7, 27). To determine whether these mutations differentially affected binding to other known GoLoco motifs, we performed similar FP assays with GoLoco motifs from RGS12 and GPSM2 (the latter protein is also known as LGN or mammalian Pins). Both Gα_i1^E116L·GDP and Gα_i1^E245L·GDP had enhanced affinities toward the RGS12 and GPSM2 GoLoco motif FP probes, whereas the K_D values obtained for Gα_i1^Q147L·GDP were statistically indistinguishable from those for Gα_i1^WT·GDP for both probes (Fig. 2).

FIGURE 2.

Gα_i1 point mutants exhibit enhanced affinity for multiple GoLoco motifs as measured by FP. A–C, log linear binding isotherms were generated from FP measurements performed in triplicate and obtained by increasing the concentration of Gα_i1 to 200 nm while the FITC-GoLoco probe concentration was held constant at 350 pm. D, apparent dissociation constants (K_D) were calculated using nonlinear regression in GraphPad Prism. Values in parentheses represent 95% confidence intervals.

To determine the dissociation rates (k_off) of the WT and mutant Gα_i1 subunits from their respective complexes with the RGS12, RGS14, and GPSM2(GL2) GoLoco motifs, we used surface plasmon resonance biosensors. We generated surfaces of GST-RGS12 GoLoco, GST-RGS14 GoLoco, and GST-GPSM2(GL2) fusion proteins, each bound to separate anti-GST antibody-coated biosensor flow cells (30), and injected 200 μl of 50 nm Gα_i1^WT, Gα_i1^E116L, Gα_i1^Q147L, or Gα_i1^E245L (in their GDP-liganded forms). Dissociation rates were determined by fitting the resultant surface plasmon resonance sensorgrams (Fig. 3) with a single exponential dissociation function. RGS12 GoLoco motif-bound Gα_i1^WT, Gα_i1^E116L, Gα_i1^Q147L, and Gα_i1^E245L were observed to have k_off rates of 1.09, 0.85, 0.79, and 0.78 (1000 s)⁻¹, respectively (Fig. 3A). Although the rates of dissociation from the RGS12 GoLoco motif were significantly decreased among all mutants versus WT Gα_i1, alterations in dissociation rates from the RGS14 GoLoco motif were more pronounced (Fig. 3B). Gα_i1^E116L, Gα_i1^Q147L, and Gα_i1^E245L had, respectively, 2-, 3-, and 4-fold decreases in their rates of dissociation compared with Gα_i1^WT. Gα_i1^E116L and Gαi1^Q147L also had significant, albeit modest, decreases in their rates of dissociation from GPSM2(GL2), 13.8 and 11.3 (1000 s)⁻¹, respectively, compared with 14.8 (1000 s)⁻¹ for Gα_i1^WT. By contrast, the rate of dissociation for Gα_i1^E245L was actually increased slightly to 16.6 (1000 s)⁻¹.

FIGURE 3.

Gα_i1 point mutants exhibit decreased rates of dissociation from GoLoco motif complexes. Surface plasmon resonance sensorgrams were generated by injecting 200 μl of 50 nm WT or mutant Gα_i1 over immobilized GST-RGS12 GoLoco motif (A), GST-RGS14 GoLoco motif (B), GST-GPSM2(GL2) (second GoLoco motif of LGN) (C), or GST alone at a flow rate of 40 μl/min with a dissociation phase of 2000 s. Dissociation curves are the plotted average of three independent experiments that have been normalized to percent of maximal binding. D, dissociation rates (k_off, expressed as per 1000 s) were derived from fitting to a single exponential function using GraphPad Prism. 95% confidence intervals are shown in parentheses.

Assessment of Nucleotide Binding and Hydrolysis Properties of the Affinity-enhanced Gα_i1 Mutants

Using our 2.2-Å crystal structure of the WT Gα_i1/RGS14 GoLoco motif interface (Protein Data Bank code 2OM2 (7)), we employed the Rosetta program to predict mutations with the goal of increasing GoLoco motif binding affinity while maintaining other Gα functionality. To determine whether the Gα_i1 mutants retain their ability to bind GTP, we performed radionucleotide-binding assays with [³⁵S]GTPγS. Whereas all three Gα_i1 mutants were capable of binding GTPγS, Gα_i1^E245L had a noticeably slower rate of binding (k_obs), 0.04 min⁻¹ compared with 0.07, 0.06, and 0.07 min⁻¹ for Gα_i1^WT, Gα_i1^E116L, and Gα_i1^Q147L, respectively (Fig. 4A). To determine any effects on the intrinsic GTP hydrolysis rate, we also performed single-turnover [γ-³²P]GTP hydrolysis assays (Fig. 4B). Gα_i1^WT, Gα_i1^E116L, and Gα_i1^Q147L were observed to have similar k_cat values of 0.48, 0.58, and 0.65 min⁻¹, respectively, whereas Gα_i1^E245L had a reduced rate of GTP hydrolysis of 0.15 min⁻¹ (95% confidence interval (CI), 0.12–0.17). Despite the rate changes observed with Gα_i1^E245L, all four Gα_i1 subunits demonstrated identical rates of activation by aluminum tetrafluoride, as measured by increased Trp-211 intrinsic fluorescence (supplemental Fig. S3). However, the reduced rates of GTP binding and hydrolysis exhibited by Gα_i1^E245L led us to exclude this particular mutant from further pursuit.

FIGURE 4.

Rates of guanine nucleotide binding and hydrolysis by affinity-enhanced Gα_i1 mutants. A, nucleotide binding assays were initiated by the addition of [³⁵S]GTPγS to 100 nm WT or mutant Gα_i1 protein in either assay buffer or nonspecific buffer (the latter with excess unlabeled nucleotide). At the indicated time points, 100 μl of the protein mixture was vacuum-filtered through nitrocellulose, washed four times, and counted by scintillation. Assays were conducted in duplicate. The error bars represent S.E. Rates of binding (k_obs) were determined using nonlinear regression to be 0.07 min⁻¹ (95% CI, 0.04–0.10), 0.04 min⁻¹ (95% CI, 0.02–0.05), 0.06 min⁻¹ (95% CI, 0.04–0.09), and 0.07 min⁻¹ (95% CI, 0.06–0.08) for Gα_i1^WT, Gα_i1^E245L, Gα_i1^E116L, and Gα_i1^Q147L, respectively. B, intrinsic GTPase activities of WT and mutant Gα_i1 subunits were determined using [γ-³²P]GTP single-turnover assays. 100 nm Gα_i1 protein was loaded with [γ-³²P]GTP at 20 °C for 10 min before assays were initiated by the addition of Mg²⁺ (10 mm) and GTPγS (400 μm). At the indicated time points, 100 μl of the protein mixture was quenched with activated charcoal, and released [³²P]P_i was counted by scintillation. Assays were conducted in duplicate. The error bars represent S.E. GTPase rates (k_cat) were determined using nonlinear regression to be 0.48 min⁻¹ (95% CI, 0.40–0.56), 0.15 min⁻¹ (95% CI, 0.12–0.17), 0.58 min⁻¹ (95% CI, 0.46–0.70), and 0.65 min⁻¹ (95% CI, 0.50–0.81) for Gα_i1^WT, Gα_i1^E245L, Gα_i1^E116L, and Gα_i1^Q147L, respectively.

Assessment of Thermodynamic Binding Parameters of the Affinity-enhanced Gα_i1^Q147L Mutant

Rosetta makes predictions of changes in Gibbs free energy (ΔG) upon mutation of single residues, but because there is no explicit entropy term provided as output from the program, one cannot accurately predict whether affinity enhancement arises from favorable enthalpic (ΔH) or entropic (ΔS) thermodynamic changes (40). We therefore interrogated by isothermal titration calorimetry which thermodynamic parameter(s) were responsible for the higher affinity of GoLoco motif binding by Gα_i1^Q147L compared with Gα_i1^WT (Fig. 5). Enthalpy of binding was favorably exothermic for Gα_i1^Q147L at −7.13 kcal/mol compared with Gα_i1^WT at −4.14 kcal/mol, whereas entropy decreased unfavorably: 15.6 cal/mol/°C for Gα_i1^Q147L versus 22.0 cal/mol/°C for Gα_i1^WT (Fig. 5C).

FIGURE 5.

Thermodynamic contributions to RGS14 GoLoco motif binding by WT and Gα_i1^Q147L subunits as determined by ITC. 30 μm solutions of each indicated Gα_i1 subunit were separately titrated with repeated additions of 350 μm RGS14 GoLoco motif peptide. Calorimetric traces of heat released and enthalpic binding curves (insets) fitted with a one-site binding model are shown for Gα_i1^WT (A) and Gα_i1^Q147L (B). The stoichiometry of binding was ∼1:1 in all experiments (three independent titrations were performed for each protein). C, average values and 95% confidence intervals for enthalpy of binding (ΔH) are given for the triplicate titration experiments (p = 0.006). Gibbs free energy (ΔG) is calculated from FP binding affinity measurements, and entropy of binding is derived by ΔS = (ΔH − ΔG)/T.

Structural Determinants of the Gα_i1^Q147L Mutant Bound to the RGS14 GoLoco Motif

To elucidate the structural determinants of GoLoco affinity enhancement for Gα_i1^Q147L, we obtained x-ray diffraction data to 2.38-Å resolution from a single crystal of the Gα_i1^Q147L·GDP/RGS14 GoLoco motif peptide complex. A structural model was obtained by molecular replacement using the model of the WT Gα_i1·GDP/RGS14 complex (Protein Data Bank code 2OM2) as a search model and subsequently refined (supplemental Table S1). The overall conformations of the Gα_i1^Q147L subunit and bound GoLoco motif peptide were essentially identical (root mean square deviation of 0.38 Å, 2545 atoms used in comparison) (supplemental Fig. S1) to those observed in the WT crystal structure (7). The N-terminal helix of the RGS14 GoLoco motif contacts the Ras-like domain of Gα_i1^Q147L in a manner similar to the WT structure (Protein Data Bank code 2OM2 (7)), whereas the N-terminal helix of a WT structure reported earlier (code 1KJY (13)) is shifted by one half-turn. The conserved (E/D)QR GoLoco triad (Asp-514, Gln-515, Arg-516) loops across the nucleotide-binding pocket, positioning the Arg-516 side chain to contact the α- and β-phosphate groups of GDP, and thus inhibits nucleotide dissociation (13). However, very weak electron density was observed for the preceding three GoLoco motif residues Gly-511, Ala-512, and His-513 that surround a well defined leucine side chain on Gα_i1 at position 147 (i.e. site of the Q147L substitution) (Fig. 6). Thus, the placement of a hydrophobic leucine on the surface of Gα_i1 appears to lead to disorder in the surrounding GoLoco motif peptide region. These findings suggest an increased conformational flexibility of the RGS14 GoLoco motif peptide in the region of amino acids 511–513 upon binding to Gα_i1^Q147L compared with WT Gα_i1.

FIGURE 6.

Differences in electron density observed within the GoLoco motif between WT and Q147L mutant Gα_i1·GDP/RGS14 GoLoco motif complexes. A crystal structure (using diffraction data to 2.38-Å resolution) of the Gα_i1^Q147L·GDP/RGS14 GoLoco motif complex (Protein Data Bank code 3ONW) was obtained. Overall, the Gα_i1^Q147L protein binds the RGS14 GoLoco motif protein similarly (root mean square deviation of 0.38 Å) to Gα_i1^WT (code 2OM2 (7)). However, electron density for the GoLoco motif peptide residues Gly-511, Ala-512, and His-513 is much weaker in the Gα_i1^Q147L structure (B) than in the WT structure (code 2OM2) (A), indicating disorder in the peptide surrounding the mutated Gα_i1 residue Leu-147. Both 2F_o − F_c electron density maps shown are contoured at σ = 1.5.

Modeling of the Q147L Substitution by Monte Carlo Simulations

The Gα_i1^Q147L/GoLoco crystal structure suggest that residues 511–513 may be more dynamic in the mutant than in WT Gα_i1, and therefore, an increase in conformational entropy may contribute to the enhanced affinity of Gα_i1^Q147L for GoLoco motifs. However, the ITC results show that the net change in the entropy component upon binding (Fig. 5) is more favorable for WT Gα_i1 than for Gα_i1^Q147L. In general, it is very difficult to predict relative changes in binding entropies and enthalpies because enthalpy/entropy compensation is a dominant feature of solvated systems (41). Changes in water networks in either the bound or unbound state could change the number of hydrogen bonds within the system and also change the rotational freedom of each water. Given our observation of disorder within the RGS14 GoLoco motif, we used Rosetta loop modeling simulations to examine the structural preferences of GoLoco motif residues 510–513 when bound to either Gα_i1^WT or Gα_i1^Q147L. For each case, 100 independent structure prediction simulations were performed. With WT Gα_i1, all 100 simulations predicted a loop conformation similar to that observed in the crystal structure (Fig. 7A). However, with Gα_i1^Q147L, two alternative conformations were observed for residues 510–513 (Fig. 7B). In 40% of the models, residues 510–513 adopt a conformation similar to the WT structure; in the other 60%, residues 511 and 512 are displaced by >4 Å. The alternate conformation places the backbone groups adjacent to both methyl groups on the Gα_i1 residue Leu-147. The solvation energy term in Rosetta disfavors the alternate conformation for Gα_i1^WT because both the nitrogen and oxygen atoms of Gln-147 are desolvated without the formation of a compensating hydrogen bond. These results are consistent with the poor electron density observed for residues 511–513 in the GoLoco/Gα_i1^Q147L crystal structure.

FIGURE 7.

Rosetta loop modeling predictions for the RGS14 GoLoco motif bound to Gα_i1^WT and Gα_i1^Q147L. Six representative models are shown from 100 independent structure prediction trajectories. GoLoco motif residues 510–513 are shown in stick mode. Gln-147 (A) and Leu-147 (B) are shown in space filling representation. Nitrogens are colored blue, and oxygens are colored red.

Conclusions

Our findings confirm the ability of Rosetta to predict point mutations that enhance protein/protein binding affinities, a technique that should be readily applicable to other systems with high-resolution structural information available (7). Furthermore, our findings have provided insight into the kinetic, thermodynamic, and structural mechanisms underlying the favorable accommodation of affinity-enhancing mutations. Two of the three Gα_i1 mutants studied (Gα_i1^Q147L and Gα_i1^E116L) exhibit enhanced binding to GoLoco motifs without disrupting guanine nucleotide binding or hydrolysis functions (an important consideration for their utility in examining selective Gα activities in an integrated cellular context). The selective loss of GoLoco motif binding of the Gα_i1^N149I point mutant was instrumental in establishing direct involvement of a Gα_i/GoLoco motif complex in microtubule dynamics underlying spindle orientation and chromosomal segregation during cell division (25). Other groups have since employed the Gα_i1^N149I mutant to demonstrate that cell polarization-mediated spatial restriction of spindle orientation determinants is critical for epithelial morphogenesis (26) and that the multiple-GoLoco motif protein AGS4 may couple Gα subunits to receptor-proximal events affected by agonist application (42). The verified gain-of-function mutants Gα_i1^Q147L and Gα_i1^E116L described in this work should complement the loss-of-function Gα_i1^N149I mutant and find equal utility in investigating the role of Gα/GoLoco motif interactions in asymmetric cell division, epithelial progenitor differentiation, and G-protein-coupled receptor signaling regulation.