Thermodynamic Characterization of the Binding of Activator of G Protein Signaling 3 (AGS3) and Peptides Derived from AGS3 with Gαi1*

Activator of G protein signaling 3 (AGS3) is a guanine nucleotide dissociation inhibitor (GDI) that contains four G protein regulatory (GPR) or GoLoco motifs in its C-terminal domain. The entire C-terminal domain (AGS3-C) as well as certain peptides corresponding to individual GPR motifs of AGS3 bound to Gαi1 and inhibited the binding of GTP by stabilizing the GDP-bound conformation of Gαi1. The stoichiometry, free energy, enthalpy, and dissociation constant for binding of AGS3-C to Gαi1 were determined using isothermal titration calorimetry. AGS3-C possesses two apparent high affinity (Kd ∼ 20 nm) and two apparent low affinity (Kd ∼ 300 nm) binding sites for Gαi1. Upon deletion of the C-terminal GPR motif from AGS3-C, the remaining sites were approximately equivalent with respect to their affinity (Kd ∼ 400 nm) for Gαi1. Peptides corresponding to each of the four GPR motifs of AGS3 (referred to as GPR1, GPR2, GPR3, and GPR4, respectively, going from N to C terminus) bound to Gαi1 with Kd values in the range of 1-8 μm. Although GPR1, GPR2, and GPR4 inhibited the binding of the fluorescent GTP analog BODIPY-FL-guanosine 5′-3-O-(thio)triphosphate to Gαi1, GPR3 did not. However, addition of N- and C-terminal flanking residues to the GPR3 GoLoco core increased its affinity for Gαi1 and conferred GDI activity similar to that of AGS3-C itself. Similar increases were observed for extended GPR2 and extended GPR1 peptides. Thus, while the tertiary structure of AGS3 may affect the affinity and activity of the GPR motifs contained within its sequence, residues outside of the GPR motifs strongly potentiate their binding and GDI activity toward Gαi1 even though the amino acid sequences of these residues are not conserved among the GPR repeats.

The ␣ subunits of heterotrimeric G proteins (G␣) act as molecular switches to modulate intracellular signaling pathways. In the classical G protein signaling cycle, agonist-bound G protein-coupled receptors catalyze the exchange of GDP for GTP at the catalytic site of G␣. Thus activated, G␣ subunits subsequently dissociate from the heterodimeric complex of G␤ and G␥ subunits (G␤␥) (1). GTP-bound G␣ subunits modulate the activity of effector enzymes, including adenylyl cyclase, phospholipase C␤, and nucleotide exchange factors for the small GTPase, Rho (2)(3)(4). Signaling is terminated by the intrinsic GTPase activity of G␣; upon hydrolysis of GTP, G␣⅐GDP reassociates with G␤␥ to form the inactive heterotrimer. Several molecules can regulate the G protein reaction cycle by modulating the rate of GTP hydrolysis or nucleotide exchange. The most familiar of these are regulators of G protein signaling (RGSs), 1 which increase the GTPase activity of G␣ subunits (5). In recent years, a new class of regulators has been discovered, typified by activator of G protein signaling 3 (AGS3). Like G␤␥, AGS3 selectively binds to the GDP-bound form of G␣ and acts as a guanine nucleotide dissociation inhibitor (GDI) (6,7).
AGS3 is a 650-residue protein from Rattus norvegicus first identified in a yeast expression screen as a receptor-independent activator of G␤␥-dependent signaling (7,8). The N-terminal half of AGS3 contains seven tetratricopeptide repeats, which have been shown to act as protein interaction domains in multiprotein complexes (9). The C-terminal half of AGS3 (AGS3-C) contains a series of four G protein regulatory (GPR) motifs (6,7,10). AGS3-related proteins, having similar arrangements of tetratricopeptide repeat-rich and GPR-rich domains, include the human protein LGN, Drosophila protein partner of inscuteable (PINS), and a related protein from Caenorhabditis elegans (7,11,12).
The GoLoco motif was first identified in the Drosophila protein "loco," the homolog of mammalian RGS12 (13). GoLoco motifs occur either singly or as tandem repeats in proteins that interact with G i and G o class ␣ subunits (6,14). Both RGS12 and RGS14 contain one GoLoco motif as do Purkinje cell protein-2 (Pcp2) and Rap1GAP (10,15,16). GPR and GoLoco motifs are one and the same.
AGS3 binds specifically to the GDP-bound forms of all three G␣ i isoforms, to G␣ t , and weakly to G␣ o and G␣ q (6,7,10,12) but exhibits GDI activity only toward G␣ i and G␣ t (6,12,14). GPR-containing proteins Pcp2 and Rap1GAP also display GDI activity toward G␣ o (10). None of the GPR/GoLoco-containing proteins that have been characterized have GDI activity toward G␣ s (10,14). By binding to G␣⅐GDP, AGS3 blocks association of G␣ with G␤␥ (12,14) and thereby prevents G protein * This work was supported by National Institutes of Health Grant DK46371, Robert A. Welch Foundation Grant I-1229, and the John W. and Rhonda K. Pate Professorship in Biochemistry (to S. R. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The GPR motifs of AGS3-C are necessary and sufficient for the GDI activity of AGS3 (6,12). Indeed a fully active splice variant of AGS3 expressed in cardiac muscle consists only of the GPR-containing C-terminal domain of AGS3 (24). Peptides that correspond to the GPR motif of RGS12 and a synthetic peptide that bears the consensus amino acid sequence of the four GPR motifs in AGS3 have potent GDI activity (10,11,15). Mutagenic scanning experiments have revealed a conserved functional core within GPR repeats corresponding to the sequence FFXLLXXXXXXXMX(D/E)QR that is required for GDI activity (16). Experiments with chimeric constructs of G␣ i1 and G␣ s identified the switch regions and helical domain of G␣ as GoLoco binding sites (25). The recently determined crystal structure of the complex between a peptide that contains the GoLoco motif of RGS14 and the GDP-bound form of G␣ i1 reveals the structural basis of GDI activity and specificity (26). The GoLoco peptide reorganizes and stabilizes the switch regions of G␣ i1 , and the arginine residue located in the conserved Asp-Gln-Arg (DQR) triad of the GoLoco motif is inserted into the active site of G␣ i1 and interacts with the ␤ phosphate of GDP. Residues C-terminal to the GoLoco motif interact with non-conserved residues in the helical domain of G␣ i1 and are therefore proposed to be determinants of specificity (26).
Inasmuch as a single GoLoco/GPR motif can possess GDI activity, the biochemical advantage conferred by proteins such as AGS3, which possess multiple copies of GPR motifs in tandem repeats, is worthy of investigation. The presumption that each of the GPR motifs in AGS3 possesses GDI activity has received experimental support, suggesting that AGS3 and its homologs might serve as scaffolds for GDP-bound G␣ subunits (12). Although it is possible that each of the four GPR motifs in AGS3 can bind one molecule of G␣ i1 , the actual stoichiometry of the complex has not been determined. It has been observed that an AGS3 construct containing only two GPR repeats is substantially less potent than one containing all four (14). Therefore, GPR motifs might function cooperatively, for example, by adopting a more stable tertiary structure in the context of AGS3 than would individual GPR motifs in isolation. It is also possible that the GPR repeats within AGS3 differ in their G␣ specificity, affinity, or GDI activity.
To address these questions, we used isothermal titration calorimetry (ITC) to determine the stoichiometry and thermodynamic parameters associated with binding of G␣ i1 ⅐GDP to AGS3-C and constructs of AGS3-C from which one or more GPR repeats had been deleted. In a complementary series of experiments, we measured the affinity and GDI activity of peptides derived from each of the GPR repeats of AGS3 to establish a correlation between affinity and biological activity. These experiments provided direct insight into the thermodynamic basis for the GDI activity of AGS3 and the relationship between activity and binding affinity.

MATERIALS AND METHODS
Reagents-Peptides that correspond in amino acid sequence to the GPR motifs 1-4 ( Fig. 1A) of rat AGS3 were synthesized in the Protein Chemistry Technology Core at The University of Texas Southwestern Medical Center at Dallas: GPR1, 469 EECFFDLLSKFQSSRMDDQRCPL 492 ; GPR2, 524 EEFFDLIASSQSRRLDDQRASV 545 ; GPR3, KK-571 GDEFFNM-LIKYQSSRIDDQRCPP 593 ; GPR4, 606 EDFFSLIQRVQAKRMDEQRVD-L 627 . Two lysine residues were added to the N terminus of the GPR3 peptide to enhance its solubility. In addition, the GPR consensus peptide (11), TMGEEDFFDLLAKSQSKRMDDQRVDLAG, was synthesized. The C terminus of each peptide was blocked by amidation. The purity of the synthetic peptides was verified using electrospray mass spectrometry and analytical high performance liquid chromatography. Peptides were desalted by elution through Sep-pack C18 cartridges (Waters), lyophilized, and stored at Ϫ20°C. BODIPY-FL-GTP␥S (BODIPY® FL thioether, sodium salt) was purchased from Molecular Probes Inc. (Eugene, OR). pGEX-4T-1 expression vectors (Amersham Biosciences) encoding peptides corresponding to AGS3 GPR motifs extended with native N-and Cterminal flanking residues (Fig. 1B) were a gift from Dr. Stephen M. Lanier (Department of Pharmacology, Louisiana State University Health Sciences Center). These peptides were expressed as N-terminal fusion proteins with glutathione S-transferase (GST) (12). The fusion peptides corresponding to the extended forms of the first three GPR motifs are referred to as GPR1ex (Pro 463 -Glu 501 ), GPR2ex (Ser 516 -Leu 555 ), and GPR3ex (Gly 563 -Thr 602 ), respectively.
AGS3 Expression Plasmids-The C-terminal domain of rat AGS3, residues 465-650, subcloned as an N-terminal His 6 fusion protein in the expression vector pQE30 was a gift from Dr. Stephen M. Lanier. This domain contains four GPR motifs: GPR1, residues 470 -489; GPR2, residues 524 -542; GPR3, residues 572-590; and GPR4, residues 606 -624. AGS3-C was cloned into the pDEST-15 destination vector as a GST fusion protein using the Gateway cloning system (Invitrogen). A tobacco etch virus protease cleavage site was inserted between coding regions for GST and AGS3. Cleavage by tobacco etch virus protease introduces an extra glycine residue at the N terminus. Two deletion mutants of AGS3 were also created: AGS3-⌬34 encompasses the first two (N-terminal) GPR domains of AGS3, residues 465-548; and AGS3-⌬4 includes the first three GPR domains, residues 465-597. Both mutants were created by inserting stop codons at the desired C terminus of the open reading frame of AGS3 using the QuikChange TM sitedirected mutagenesis kit (Stratagene).
Expression and Purification of Recombinant Proteins-Recombinant G␣ i1 was expressed and purified as described previously (27). AGS3 and the deletion mutants of AGS3 were expressed in transformed BL21 (DE3) strains of Escherichia coli cells as GST fusion proteins. Cells were grown in 1-3 liters of LB medium at 37°C to A 600 of ϳ0.9 and induced with 200 M isopropyl-␤-D-thiogalactopyranoside at 30°C for 5 h for expression of recombinant proteins. Induced cells were harvested by centrifugation, flash frozen in liquid nitrogen, and stored at Ϫ80°C. Frozen cells were thawed and resuspended in lysis buffer (50 mM Na ϩ ⅐HEPES, pH 7.5, 100 mM NaCl, 5 mM dithiothreitol (DTT), and 1ϫ protease mixture PTT (1000ϫ PTT contains 23 mg/ml phenylmethylsulfonyl fluoride, 21 mg/ml N ␣ -p-tosyl-L-lysine chloromethyl ketone, and 21 mg/ml N-tosyl-L-phenylalanine chloromethyl ketone dissolved in 1:2 (v/v) solution of dimethyl sulfoxide and isopropanol)) containing 3 mg/ml hen egg white lysozyme for 30 min at 4°C with continuous stirring followed by sonication for 5 min on ice (5-s pulse and 5-s idle cycle). The cell lysate was centrifuged at 35,000 rpm for 40 min at 4°C in a Beckman Ti45 rotor. Clear supernatant was filtered using a 0.45-m syringe filter and loaded on glutathione-Sepharose 4B resin (Amersham Biosciences). The resin was washed with lysis buffer, and the GST-tagged proteins were eluted with elution buffer (20 mM Tris, FIG. 1. Multiple sequence alignment of GPR peptides. A, sequences of the four synthetic GPR peptides of AGS3-C are shown along with the GoLoco peptide derived from RGS14 (26) and the GPR consensus peptide (11). The highly conserved residues that are critical for GDI activity (16) are colored red, absolutely conserved but functionally noncritical residues are colored green, other highly conserved but functionally non-critical positions are colored blue, and non-native residues added to enhance solubility are colored orange. The length of each peptide is listed at the right of its sequence. B, multiple sequence alignment of the extended GPR peptides GPR1ex, GPR2ex, and GPR3ex. Flanking residues are colored gray. pH 8.0, 100 mM NaCl, 5 mM DTT, 25 mM reduced glutathione, and 1ϫ PTT). The purified GST fusion proteins were cleaved using 10 g of recombinant tobacco etch virus protease/mg of fusion protein at 4°C overnight and dialyzed against low salt buffer (20 mM Tris, pH 8.0, 2 mM DTT, 1 mM EDTA, 1ϫ PTT, 5% glycerol). The protein was loaded on a Hi-Trap Q TM column (Amersham Biosciences) and eluted with a 75-ml linear gradient of 100 -750 mM NaCl. The proteins were eluted at 150 -250 mM NaCl concentration. Fractions containing AGS3 or its deletion mutants were pooled, concentrated, and loaded on tandem Superdex TM 200 and 75 gel filtration columns (Amersham Biosciences) equilibrated with low salt buffer supplemented with 100 mM NaCl. The peak fractions were pooled, concentrated, and stored at 4°C. Protein concentration was estimated by the Bradford assay using the Bio-Rad protein assay kit. GST fusion proteins with extended GPR peptides were expressed in BL21 (DE3) strains. 2 Cells (1 liter) were grown to an A 600 of 0.5-0.6 and induced with 100 M isopropyl-␤-D-thiogalactopyranoside at 30°C for 3 h. Cells were harvested by centrifugation, flash frozen in liquid nitrogen, and stored at Ϫ80°C. Frozen cells were thawed and resuspended in lysis buffer containing 1ϫ PBS (1 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 2 mM KCl, 140 mM NaCl, pH 7.4), 5 mM EDTA, 5 mM DTT, and complete protease inhibitor mixture (Roche Applied Science). Cell lysis was performed by brief 1-min sonication on ice followed by incubation with 1% Triton X-100 with continuous stirring at 4°C. The cell lysate was centrifuged at 35,000 rpm for 40 min at 4°C in a Beckman Ti45 rotor. Cleared lysate was loaded on glutathione-Sepharose 4B resin pre-equilibrated with lysis buffer. The resin was washed with lysis buffer, and the fusion proteins were eluted using buffer containing 50 mM Tris, pH 8.0, 100 mM NaCl, 5 mM DTT, 5 mM EDTA, and 25 mM reduced glutathione. The fusion proteins were dialyzed against a buffer containing 20 mM Tris, pH 8.0, and 1 mM EDTA, concentrated, and stored at 4°C. Protein concentration was estimated by measuring A 280 and using theoretically predicted extinction coefficients.
Preparation and Purification of AGS3-C⅐G␣ i1 ⅐GDP Complexes-The complex of G␣ i1 ⅐GDP with AGS3-C was formed by incubating the proteins together in molar ratios (G␣ i1 ⅐AGS3-C) exceeding 4:1 to ensure saturation of AGS3-C. Molar ratios exceeding 3:1 and 2:1 were used for complex formation of G␣ i1 with AGS3-⌬4 and AGS3-⌬34, respectively. In a typical experiment, 100 l of 500 M AGS3-C was mixed with 500 l of 350 M G␣ i1 in buffer containing 50 mM Tris, pH 8.0, 2 mM DTT, 1 mM EDTA, and 1 mM GDP and incubated on ice for 2 h. The resulting complex was separated at 4°C on tandemly connected Superdex 200 and 75 gel filtration columns (Amersham Biosciences) at a flow rate of 0.4 ml/min with 20 mM Tris, pH 8.0, 100 mM NaCl, 1 mM EDTA, 2 mM DTT, and 1ϫ PTT as the running buffer. The apparent molecular weights of the complexes were estimated using gel filtration standards (Bio-Rad).
Isothermal Titration Calorimetry-Isothermal titration calorimetry was performed at 20°C (293 K) using a MicroCal VP-ITC (MicroCal, Northhampton, MA) calorimeter. Protein samples were dialyzed against titration buffer (20 mM Tris, pH 8.0, 2 mM DTT, and 10 M GDP). Because peptides could not be dialyzed due to their lower molecular weight, lyophilized peptides were dissolved directly in titration buffer. Samples were centrifuged at 13,000 rpm in a bench-top microcentrifuge (Forma Scientific) for 5 min before loading in the sample cell or syringe. Contents of the sample cell were stirred continuously at 300 rpm during the experiment. A typical titration of G␣ i1 with GPR peptides involved 25-30 injections at 3-min intervals of 8 -10 l of peptide solution (1 mM) into a sample cell containing 1.4 ml of G␣ i1 (45-50 M). For the titration of G␣ i1 with AGS3-C or its deletion mutants, G␣ i1 (125-200 M) was injected through the syringe into a sample cell containing AGS3-C (5-10 M); the injection volume was decreased to 5-6 l, and the number of injections was increased to 45-50. For the GST fusion extended GPR peptides, G␣ i1 (175-225 M) was injected (35 injections of 8 l) into the sample cell containing GPRex fusion peptide (20 -25 M), and NaCl (100 mM) was added to the titration buffer to avoid nonspecific interactions with GST. The heats of dilution of the titrants were subtracted from the titration data for base-line correction. The base line-corrected data were analyzed with MicroCal Origin TM 5.0 software to determine the enthalpy (⌬H), association constant (K a ), and stoichiometry of binding (N). Thermal titration data were fit to one or more of the three association models available in the software: "single set of identical sites," "two sets of independent sites," and "sequential binding sites" (28,29). The models were compared by visual inspection of the fitted curves and by comparing the 2 values obtained after the computation. The model resulting in the lowest value of 2 was consid-ered the best model to describe the molecular mechanism of binding. Free energy change (⌬G) and entropy change (⌬S) were calculated from ⌬H and K a using standard free energy relationships. Several titrations were performed (two to four) for each sample set to evaluate reproducibility.
Nucleotide Exchange Assay-The rate of guanine nucleotide exchange on G␣ i1 was assayed by monitoring the rate at which the fluorescent GTP analog BODIPY-FL-GTP␥S replaces bound GDP in the catalytic site of G␣ i1 (30). Binding of the fluorescent nucleotide analog is accompanied by an increase in fluorescence of the BODIPY moiety. G␣ i1 (final concentration, 200 nM) in assay buffer (50 mM Tris, pH 8.0, 1 mM EDTA, and 10 mM MgCl 2 ) was incubated in the presence or absence of AGS3-C, AGS3-⌬34, AGS3-⌬4, GPR peptides, or GST fusion GPRex (at concentrations indicated in Figs. 4, 6, and 7) for 1-2 h at 25°C. Samples were transferred into a 3-ml quartz cuvette with a path length of 1 cm containing 1 M BOPDIPY-FL-GTP␥S in assay buffer, and nucleotide exchange was initiated by rapid mixing. Binding of fluorescent nucleotide to G␣ i1 was monitored by the intensity of fluorescence emission at 510 nm using a PerkinElmer Life Sciences LS50B spectrophotometer (data shown in Figs. 4, A and B, and 6A) or a custom-built fluorescence spectrophotometer (Photon Technology International, Fig. 7) The excitation wavelength was set to 485 nm with the slit widths for excitation and emission maintained at 2.5 nm. All assays were performed at 30°C. Typically fluorescence intensity was recorded at 30-s intervals with 10-s averaging time over a period of 1 h after mixing of the samples. Two to three data sets were measured for each experiment-and the base-line fluorescence (intensity at time t ϭ 0) was subtracted from the data sets. The data were averaged and smoothed using five-point adjacent averaging. The data were fit to a first order exponential association model, Y ϭ Y s (1 Ϫ e Ϫkt ), where k is the rate constant (s Ϫ1 ) and Y and Y s represent concentrations of BODIPY-FL-GTP␥S-bound G␣ i1 at time t and at maximum saturation, respectively. Initial rates were estimated by linear approximation to the change in fluorescence intensity during the first 10 min after initiation of the exchange reaction.
Circular Dichroism Spectroscopy-CD spectra were measured with a Jasco Model J715 spectropolarimeter (Jasco Inc., Easton, MD) using a 0.05-cm path length cylindrical cell. AGS3-C was dialyzed overnight in 10 mM Tris, pH 8.0, and lyophilized GPR consensus peptide was dissolved directly in the buffer. Spectra were recorded of 8 and 16 M AGS3-C and of 100 M GPR consensus peptide. CD spectra were measured at ϳ25°C with a 1-nm spectral bandwidth, scan speed of 50 nm/min, and a response time of 1 s. Data were collected at 0.1-nm intervals, and 15 accumulations were averaged to obtain each spectrum. The spectra of AGS3-C at the two different concentrations were averaged to obtain the final spectrum. CD data were smoothed by the Savitzky-Golay method using the program provided by Jasco and ⑀ L Ϫ ⑀ R was calculated (in units of M Ϫ1 cm Ϫ1 /residue) at 1-nm intervals. The CD spectra over the range of 250 -190 nm were analyzed for fractional content of secondary structures using CDPRO software CONTINLL (31), SELCON3 (32), and CDSSTR (33) with a reference set containing 43 proteins.

Stoichiometry and Affinity of the Binding of AGS3-C and Its
Deletion Mutants to G␣ i1 -AGS3-C (residues 465-650) contains four tandemly repeated GPR motifs, each of which has been shown in immunoprecipitation assays to be capable of binding to G␣ i1 (10), suggesting that a single AGS3-C domain can bind up to four molecules of the ␣ subunit. We used gel filtration chromatography to estimate the stoichiometry of the interaction between AGS3-C and G␣ i1 . Upon incubation together in the presence of GDP, G␣ i1 and AGS3-C formed a stable complex. The complex could be purified by gel filtration chromatography (Fig. 2) and eluted with an apparent molecular mass (ϳ207 kDa) consistent with four G␣ i1 subunits bound to one molecule of AGS3-C (molecular mass, ϳ180 kDa). Hence it appears that the each of the GPR motifs binds to G␣ i1 . To further test this hypothesis, C-terminal deletion mutants of AGS3-C were created that contain the first two (AGS3-⌬34) or the first three (AGS3-⌬4) GPR motifs, respectively (see "Materials and Methods"). Both deletion mutants formed complexes with G␣ i1 in the presence of GDP and could be separated from their constituents by gel filtration chromatography. An esti-mate of the molecular masses of the complexes by gel filtration using standards of known molecular mass was consistent with a stoichiometry of 1:3 for AGS3-⌬4⅐G␣ i1 (ϳ151-kDa complex) and 1:2 for AGS3-⌬34⅐G␣ i1 (ϳ110 kDa). The estimated molecular masses for both of these complexes were ϳ20 kDa higher than the masses expected for the predicted complexes.
We then used ITC to precisely determine the stoichiometry and affinity of the interaction between AGS3-C and G␣ i1 . The dependence of heat evolved upon titration of G␣ i1 ⅐GDP into a solution containing AGS3-C (Fig. 3A) yielded an overall stoichiometry of 4.3 (G␣ i1 ⅐AGS3-C), consistent with a 4:1 stoichiometry of binding. Because the amino acid sequences of the four GPR repeats differ from each other, the most general model for binding of G␣ i1 to AGS3-C would require four association constants. If some of the affinity constants are of similar magnitude, their individual values cannot be accurately determined from the titration data. On the other hand, binding of G␣ i1 to AGS3-C was not well approximated by a model in which four identical independent binding sites are assumed. A sequential four-site binding model, in which the dissociation constant is a function of the number of binding sites occupied, provided a better fit to the data. However, a yet simpler model, in which two sets of independent binding sites are assumed, provided the best fit to the titration data (Fig. 3A). Accordingly the data were consistent with two strong (K d ϳ 20 nM) and two weak (K d ϳ 300 nM) G␣ i1 binding sites in AGS3-C (Table I). The two types of sites differed only by ϳ1.5 kcal/mol in binding free energy. Binding to the higher affinity sites was accompanied by a larger positive change in entropy but a smaller negative enthalpy than is binding to the low affinity sites ( Table I).
Binding of the two AGS3-C deletion mutants to G␣ i1 was also investigated using ITC. Analysis of the thermal titration data indicated that AGS3-⌬4 and AGS3-⌬34 bind to three and two molecules of G␣ i1 ⅐GDP (Table I), respectively, which again is consistent with the gel filtration data. However, unlike the binding of AGS3-C to G␣ i1 , the titration data for both of the deletion mutants was most consistent with a model in which a single set of identical binding sites is present (Fig. 3B). Although the deletion mutants share a similar binding mechanism, AGS3-⌬34 had almost 3-fold greater affinity for G␣ i1 than did AGS3-⌬4. The apparent change in the G␣ i1 binding , and AGS3-⌬34 (closed squares) with G␣ i1 are superimposed. Nonlinear least squares fit to the "two sets of independent sites" model for AGS3-C is shown as a solid line and the "single set of identical sites" models used to fit the ITC profile of AGS3-⌬4 and AGS3-⌬34 are shown as dotted and dashed lines, respectively. mechanism of AGS3-C upon deletion of one or two of its Cterminal GPR domains could reflect an alteration of the tertiary structure or structural environment of the remaining GPR repeats. Alternatively the four repeats may differ substantially in their affinity for G␣ i1 ⅐GDP.
The AGS3-C constructs used in these experiments inhibit the GDP exchange activity of G␣ i1 as expected from previous studies (Fig. 4A). The apparent association rate for the binding of a fluorescent non-hydrolyzable GTP analog (BODIPY-FL-GTP␥S) to G␣ i1 was measured in the presence of AGS3-C and the two C-terminal truncation mutants. In keeping with earlier reports (14), the GDP exchange rate of G␣ i1 was reduced 50% at 100 nM AGS3-C (Fig. 4A). Deletion of one or two GPR repeats from AGS3-C did not have a severe effect on nucleotide exchange inhibitory activity. At the same concentration, AGS3-⌬4 and AGS3-⌬34 were almost equally potent inhibitors of nucleotide exchange on G␣ i1 (Fig. 4B), indicating that the GPR repeats can function independently and that deletion of individual repeats from AGS3-C does not result in loss of its GDI activity.
Thermodynamic Analysis of the Binding of GPR Peptides to G␣ i1 -The apparent differences in the binding models for AGS3-C and that of its truncation mutants could reflect differences in the affinities of individual GPR motifs for G␣ i1 . These might arise from sequence variation at positions other than those that are conserved and critical for function (Fig. 1A) (16).
To investigate this possibility, we sought to determine the contribution of each GPR motif to the binding of AGS3-C to G␣ i1 . Peptides corresponding to each of the four GPR motifs in  I Thermodynamic parameters of the binding of AGS3-C and its deletion mutants to G␣ i1 ⅐GDP Thermodynamic parameters for the binding of AGS3-C and its deletion mutants with G␣ i1 were determined using ITC at 20 o C in 20 mM Tris, pH 8.0, 2 mM DTT, and 10 M GDP. A binding model that assumes two sets of independent sites best described the titration data for AGS3-C. The two different types of sites are referred to as strong sites and weak sites in the table. Titration data for AGS3-⌬4 and AGS3-⌬34 were fit to a single set of identical sites model. K d , ⌬H, ⌬G, ⌬S, and N represent the dissociation constant, enthalpy, free energy, entropy, and stoichiometry, respectively, and the units corresponding to these parameters are shown in the AGS3 were synthesized, and the binding of these peptides to G␣ i1 was analyzed by ITC (Fig. 5). The interaction of the previously described GPR consensus peptide (11) with G␣ i1 was also analyzed. The titration data for the GPR peptides are best described by a single binding site model. Each of the peptides bound to G␣ i1 with a stoichiometry of 1:1. The dissociation constant (K d ) for the peptides ranged from 1-8 M, and the enthalpy of binding ranged from Ϫ3.5 to Ϫ9.8 kcal/mol (Table  II). The differences in the enthalpy changes were balanced by entropic terms such that binding free energies of individual peptides differed by no more than 1.2 kcal/mol (Table II). The affinity of the consensus peptide for G␣ i1 was somewhat higher than that of the four GPR peptides. The average free energy of binding for the four GPR peptides was Ϫ7.3 kcal/mol, which is significantly less in absolute value than the average free energy of binding per site for AGS3-C (Ϫ10.3 kcal/mol) and for AGS3-⌬4 and AGS3-⌬34 (Ϫ8.6 and Ϫ9.2 kcal/mol, respectively). Since sequence variation among the AGS3-C GPR motifs appears to have little effect on their affinity for G␣ i1 , it is possible that residues outside of the GPR motif are involved in binding or that the three-dimensional structure of AGS3-C contributes to the stability of the complex. GDI Activity of GPR Peptides-To investigate the possibility that GPR peptides differ in GDI activity, the nucleotide exchange rate of G␣ i1 was determined in the presence of each peptide. Peptides were present in the assay at a concentration of 10 M, which exceeds the K d of GPR1, which has the lowest affinity of the four for G␣ i1 . Under these conditions, all the GPR peptides with the exception of GPR3 demonstrated nucleotide exchange inhibitory activity (Fig. 6A). Relative GDI activity of the peptides followed the rank order: GPR consensus Ͼ GPR1 Ͼ GPR2 Ͼ GPR4 Ͼ Ͼ GPR3 (Fig. 6B). In contrast, the rank order of binding affinity as determined by ITC was: GPR consensus Ͼ GPR4 Ͼ GPR3 ϳ GPR2 Ͼ GPR1. GPR3 also failed to inhibit the increase in intrinsic tryptophan fluorescence of G␣ i1 in the presence of excess GTP␥S, again indicating that it cannot inhibit nucleotide exchange (data not shown) (34). In general, the GDI activity of the GPR peptides was not correlated with their affinity for G␣ i1 .
Affinity and GDI Activity of Extended GPR Peptides-In light of the observation that residues extending C-terminal to the GoLoco motif of RGS14 contribute to G␣ specificity and GDI activity (26), we investigated the effect of residues outside of the GPR motifs of AGS3. The extended peptides (Fig. 1B) were expressed as N-terminally GST-tagged fusion proteins. The proteins were purified to near homogeneity by glutathione affinity chromatography and used without further purification. Because GST-GPR4ex was rapidly degraded despite the presence of protease inhibitors, it was not tested for G␣ i1 binding or GDI activity. GST-GPR3ex was a potent GDI toward G␣ i1 (Fig.   FIG. 6. Inhibition of BODIPY-FL-GTP␥S binding to G␣ i1 by GPR peptides derived from AGS3. A, G␣ i1 (200 nM) was incubated with 10 M GPR1 (red), GPR2 (green), GPR3 (blue), GPR4 (cyan), and GPR consensus (magenta). The preincubated solutions were mixed with 1 M BODIPY-FL-GTP␥S, and fluorescence intensity was monitored at 510 nm. The black curve represents the binding of BODIPY-FL-GTP␥S with G␣ i1 in the absence of any peptide. B, histogram showing the GDI activity of different peptides at 10 M concentration. Fractional exchange at any time point was calculated as a ratio of fluorescence intensity of the sample with peptide to that without peptide. Percentage of inhibition ϭ (1 Ϫ fractional exchange) ϫ 100. The peptides can be arranged in descending order of activity as GPR consensus Ͼ GPR1 Ͼ GPR2 Ͼ GPR4 Ͼ Ͼ GPR3. GPR-con, GPR consensus.

TABLE II
Thermodynamic parameters for binding of GPR peptides to G␣ i1 ⅐GDP Thermodynamic parameters describing the binding of GPR peptides (GPR1, GPR2, GPR3, GPR4, and GPR consensus) and extended GPR peptides (GPR1ex, GPR2ex, and GPR3ex) with G␣ i1 were determined at 20°C using ITC in 20 mM Tris, pH 8.0, 2 mM DTT, and 10 M GDP. The titration data were fit to a single site binding model.  7) with an apparent IC 50 in the micromolar range. The addition of GPR-flanking residues also increased the potency of GPR2 ϳ10-fold but had less effect upon the activity of GPR1. The rank order of the extended peptides with respect to GDI activity (GPR3ex ϳ GPR2ex Ͼ GPR1ex) was almost the reverse of that for the core GPR peptides. The extended GPR peptides also bound with 10 -50-fold higher affinity to G␣ i1 ⅐GDP than did the smaller peptides that contain only few (approximately three) flanking residues in addition to the core GoLoco consensus motif (Table II). Binding constants for these extended peptides were within the range measured for AGS3 itself. Interestingly, much like GPR2, which had the lowest enthalpy of binding among the GPR peptides, GPR2ex also had a low enthalpy of binding, and binding was to a great extent driven by entropy. The greater affinity of GPR1ex relative to GPR1 was also marked by a large increase in the entropy of binding. In contrast, the binding enthalpy of GPR3ex was substantially greater than that of GPR3. For the extended peptides, the rank order with respect to G␣ i1 affinity followed that of GDI activity. GST itself had no affinity or GDI activity toward G␣ i1 (data not shown). Solution Structure of AGS3 and a GPR Peptide-To compare the content of secondary structure in GPR peptides and AGS3, CD spectra were obtained for the GPR consensus peptide and AGS3-C in the wavelength range of 190 -250 nm. Analysis of the CD spectra (Fig. 8) indicated higher ␣-helical content in AGS3-C (20.4% or ϳ38 residues of 186 residues) than in the GPR consensus peptide (8% or ϳ2 residues of 28 residues), suggesting that AGS3-C could have native helical structures that are in the proper conformation to interact with G␣ i1 . The analysis also suggested that AGS3-C and the GPR peptide have approximately equal fractions of turn (ϳ25%) and disordered structure (ϳ33%), but the latter has a higher content of ␤-strand structure (33.5%) than AGS3-C (22.2%).

DISCUSSION
The residues that confer GDI activity in GPR/GoLoco motifs have been defined by mutagenesis (16), and the structure of a GoLoco repeat bound to G␣ i1 ⅐GDP provides insight into its mechanism of action (26). Several proteins such as AGS3 possess multiple GPR repeats, and therefore the question arises how these repeats function biochemically as an ensemble.
Using isothermal titration calorimetry and gel filtration analysis, we demonstrated that AGS3-C could bind up to four molecules of G␣ i1 , equal to the number of GPR motifs present in its amino acid sequence. Successive deletion of GPR motifs from the C terminus of AGS3-C did not abrogate the G␣ i1 binding activity of the remaining motifs. Hence GPR repeats can function independently within the context of the AGS3-C scaffold. Indeed the naturally occurring splice variant of AGS3 known as AGS3-SHORT contains three complete GPR motifs and is known to be a functionally active molecule (24). Nevertheless the ITC data for AGS3-C and its truncation mutants suggest that the structural context of the GPR repeats within AGS3-C may affect their affinity for G␣ i1 . Surprisingly AGS3-C and its truncation mutants had nearly equal potency as GDIs when tested at 100 nM concentration, which is close to their average K d as determined by ITC.
Although all conformed to the GPR consensus sequence, the four GPR repeats otherwise differed in sequence from each other, and therefore AGS3-C must have four non-equivalent binding sites. These sites could function independently or with some degree of cooperativity. The experimental errors inherent in the ITC measurements do not allow thermodynamic parameters to be extracted for each binding site in the most general four-site model. However, the binding data clearly did not fit the simplest model in which four identical, independent sites are assumed. Although the titration data could be fit to a more general sequential four-site binding model (eight free parameters), the uncertainties associated with derived constants did not justify a model of this complexity, and a simple two-site model (six free parameters) gave an adequate fit to the binding data. From this latter analysis, two high affinity (K d ϳ 20 nM) and two low affinity (K d ϳ 300 nM) binding sites for G␣ i1 could be derived. However, this binding model cannot be considered a complete or quantitative thermodynamic description of the interaction between AGS3-C and G␣ i1 . The key observation is that the four binding sites in the context of AGS3-C were not equivalent even though the GPR peptides bound to G␣ i1 with approximately equal affinity (see below). Successive deletion of the two C-terminal GPR motifs from AGS3 destroyed the biphasic character of G␣ i1 binding. The truncated constructs bound G␣ i1 with lower affinity, and binding could be defined by a single association constant. The apparent change in binding mechanism upon truncation may reflect the loss of tertiary structure that is present in AGS3-C as well as the loss of specific G␣ i1 -binding elements.
The GPR motif of RGS14 adopts a helical conformation while bound to G␣ i1 (26). However, CD and 1 H NMR analysis (data not shown) of the consensus GPR peptide showed little evidence of ␣-helical structure. The CD spectrum of AGS3-C indicated that a modest fraction of the molecule is helical, so that it is possible that, within the context of AGS3-C, the GPR motifs adopt the secondary structure consistent with the requirements of the G␣ i1 binding site. This may account for some small fraction of the 2.5 kcal/mol (per binding site) increase in binding energy for GPR repeats in AGS3-C relative to that for individual GPR peptides.
It was evident that the sequences that flank the GPR consensus motif include residues that contribute substantially to G␣ i1 binding energy. These additional residues augmented GDI activity as well, and for GPR3, they were required. The affinity of any one of the extended GPR peptides for G␣ i1 was comparable to that of the high or low affinity sites in AGS3-C. Even the GPR consensus peptide, which is longer by a few residues at both termini than that of the other four GPR peptides, had a greater free energy of binding to G␣ i1 . The dissociation constants derived for the binding of AGS3-C to G␣ i1 may therefore simply represent the distribution of affinities of the individual extended GPR motifs. Nevertheless the possibility that the tertiary organization of the G␣ i1 -binding motifs in AGS3-C influences their affinity for G␣ i1 cannot be dismissed. That AGS3-⌬34 bound to G␣ i1 more strongly than did AGS3-⌬3 suggests steric interference between sites in the latter, which may be relieved by truncation of the third motif. More rigorous structural and thermodynamic analysis of the AGS3-C⅐G␣ i1 binding interaction will be required to address these questions in depth.
While it is clear that residues outside of the GoLoco motif contribute binding energy toward G␣ i1 , it is not apparent whether the N-or C-terminal residues are equally important. The structure of the complex between G␣ i1 and the GoLoco motif of RGS14 reveals a substantial interface between the helical domain of G␣ i1 and the residues that extend from the C terminus of the GoLoco motif (26). In contrast, the N-terminal boundary of the GoLoco motif almost extends beyond the surface of the Ras-like domain of G␣ i1 . Hence it is probable that it is the C-terminal flanking residues of GPRex peptides that contribute the additional binding energy toward G␣ i1 . It is remarkable that these residues are not conserved among the four GPR repeats in AGS3-C (Fig. 1B). Again the structures of the complexes between extended peptides and G␣ i1 should reveal the structural basis for their contribution to binding.
Residues that flank the GPR/GoLoco motif not only confer specificity, as demonstrated for RGS14 (26), but may be critical for GDI activity as our data indicated for GPR3 of AGS3-C. Apparently the binding mode of GPR3 is nonproductive for GDI activity even though the conserved core of residues common to all GPR motifs is retained in GPR3 (Fig. 1A). The few significant differences between the sequence of GPR3 and those of the other peptides are as follows: 1) presence of a hydrophobic isoleucine residue (Ile 579 ) at a position that is normally occupied by either a solvent-exposed polar residue or alanine (26), 2) substitution of an isoleucine for a methionine at position 586, which has been shown to be important for GDI activity (16), and 3) occurrence of a pair of proline residues immediately following the catalytic arginine that binds the ␤ phosphate of GDP in the G␣ i1 ⅐GDP complex (26). Whether these or other factors account for the inactivity of GPR3, residues that flank this motif are compensatory to the extent that GPR3ex exhibits potent GDI activity.
Several protein families have evolved multiple GoLoco motifs, although only a single GoLoco motif is necessary and sufficient for biochemical activity. Molecules such as AGS3 could act as scaffolding molecules (24,35) to bring several G␣ subunits together and thereby enhance the efficiency of signaling. The strong sites in AGS3-C differ from the weak sites by 1.5 kcal/mol in binding free energy for G␣ i1 (ϳ15-fold difference in K d ) suggesting that a nearly linear response in GDI activity is possible over a ϳ10 -500 nM range in effective G␣ i concentration. The broad dynamic range of AGS3-C might play a critical role in the mechanism by which it inhibits G␣ i1 signaling in vivo.