A Structural Determinant That Renders Gαi Sensitive to Activation by GIV/Girdin Is Required to Promote Cell Migration*

Although several non-receptor activators of heterotrimeric G proteins have been identified, the structural features of G proteins that determine their interaction with such activators and the subsequent biological effects are poorly understood. Here we investigated the structural determinants in Gαi3 necessary for its regulation by GIV/girdin, a guanine-nucleotide exchange factor (GEF) that activates Gαi subunits. Using G protein activity and in vitro pulldown assays we demonstrate that Gαi3 is a better substrate for GIV than the highly homologous Gαo. We identified Trp-258 in the Gαi subunit as a novel structural determinant for GIV binding by comparing GIV binding to Gαi3/Gαo chimeras. Mutation of Trp-258 to the corresponding Phe in Gαo decreased GIV binding in vitro and in cultured cells but did not perturb interaction with other Gα-binding partners, i.e. Gβγ, AGS3 (a guanine nucleotide dissociation inhibitor), GAIP/RGS19 (a GTPase-activating protein), and LPAR1 (a G protein-coupled receptor). Activation of Gαi3 by GIV was also dramatically reduced when Trp-258 was replaced with Tyr, Leu, Ser, His, Asp, or Ala, highlighting that Trp is required for maximal activation. Moreover, when mutant Gαi3 W258F was expressed in HeLa cells they failed to undergo cell migration and to enhance Akt signaling after growth factor or G protein-coupled receptor stimulation. Thus activation of Gαi3 by GIV is essential for biological functions associated with Gαi3 activation. In conclusion, we have discovered a novel structural determinant on Gαi that plays a key role in defining the selectivity and efficiency of the GEF activity of GIV on Gαi and that represents an attractive target site for designing small molecules to disrupt the Gαi-GIV interface for therapeutic purposes.

Heterotrimeric G proteins are molecular switches that control signal transduction. G protein cycling between active and inactive states is controlled via interaction with regulatory proteins. Activation is triggered by guanine nucleotide exchange factors (GEFs), 5 and deactivation is greatly enhanced by GTPase-activating proteins (GAPs) (1)(2)(3). Because the duration and extent of G protein-mediated signaling is determined by the lifetime of G␣ in the GTP-bound state, it is crucial to define the molecular machinery that triggers G protein activation to understand how this signal transduction pathway functions. Ligand-occupied G protein-coupled receptors (GPCRs) are the canonical GEFs of which Ͼ800 genes have been identified in the human genome (4). They regulate a myriad of physiological functions and are the most common target for marketed drugs (ϳ30%) (5). Recently, a few non-receptor GEFs have been described, i.e. AGS1 (6), Ric-8 (7,8), CSP␣ (9), and Arr4 (10). In contrast to GPCRs, these non-receptor GEFs are structurally unrelated, and their physiological roles are just beginning to be elucidated (8,(11)(12)(13). The lack of information on non-receptor GEFs has limited their exploitation as pharmacological targets.
We recently demonstrated that GIV is a non-receptor GEF for G␣ i subunits (11). Originally GIV was identified by its ability to interact with G␣ i3 in a yeast two-hybrid screen (14). Work from other groups indicated that GIV (also known as girdin) enhances Akt signaling (15) and plays a critical role in cell migration via its interaction with Akt and the actin cytoskeleton (16). GIV was shown to be required for cancer metastasis in murine models by virtue of its ability to control cell migration and actin remodeling (17). We subsequently found that active G␣ i3 , like GIV, promotes Akt signaling, remodeling of the actin cytoskeleton, and tumor cell migration (18).
Moreover, we recently reported that GIV activates G␣ i3 subunits via an evolutionarily conserved GEF motif and that this novel regulatory motif provides the structural and biochemical basis for the pro-metastatic features of GIV (11). We identified the GEF motif of GIV based on its sequence homology with the synthetic GEF peptide KB-752 (19) and showed that mutational disruption of the ability of GIV to activate G␣ i subunits via this motif abolished the enhanced Akt activation (15), actin cytoskeleton remodeling (16,17,20), and cell migration (16,17) seen in metastatic tumor cells (11).
GIV is the first non-receptor GEF whose function has been shown to be governed by a defined motif. Because the GEF * This work was supported, in whole or in part, by National Institutes of Health function of GIV appears critical for cancer metastasis, disruption of the interface formed between the GEF motif of GIV and G␣ i is potentially of therapeutic significance, and defining the molecular basis and properties of this interface is crucial for the future development of pharmacological agents that target this interface. Here we investigated in depth the structural determinants in the G␣ i3 subunit required for it to interact with GIV and be activated. Using the G␣ selectivity of GIV to identify such determinants, we found that residues outside of the previously described G␣ i -GIV interface (11) define the selectivity and efficiency of the GEF activity of GIV on G␣ i in living cells and in vitro. These data provide valuable insights that can be used in the design of pharmacological agents that selectively disrupt the G␣ i -GIV interface for therapeutic purposes.

EXPERIMENTAL PROCEDURES
Reagents and Antibodies-Unless otherwise indicated all reagents were of analytical grade and obtained from Sigma-Aldrich. Cell culture media were purchased from Invitrogen. All restriction endonucleases and Escherichia coli strain DH5␣ were purchased from New England Biolabs (Cambridge, MA). E. coli strain BL21(DE3) was purchased from Invitrogen. Pfu ultra DNA polymerase was purchased from Stratagene (La Jolla, CA). [␥-32 P]GTP and [ 35 S]GTP␥S were from PerkinElmer Life Sciences. Rabbit antisera against AGS3 (21) and the coiledcoil region of GIV (14) were raised as described. Goat antirabbit and goat anti-mouse Alexa Fluor 680 or IRDye 800 F(abЈ) 2 were from Li-Cor Biosciences (Lincoln, NE). Mouse monoclonal antibodies against hexahistidine (His), FLAG (M2), and ␣-tubulin were obtained from Sigma-Aldrich. Rabbit anti-pan-G␤ (M-14) IgG was from Santa Cruz Biotechnology (Santa Cruz, CA) and anti-Akt and phospho-Akt (S473) IgGs were from Cell Signaling (Beverly, MA).
Plasmid Constructs and Mutagenesis-Cloning of rat G␣ i3 into pGEX-4T-1 or pET28b and GIV-CT-(1623-1870) into pET28b were described previously (11,18). Rat G␣ o (isoform 1, G␣ o1 , hereafter referred to as G␣ o ) was cloned from pGBT9-G␣ o (22) and inserted between the EcoRI and NotI restriction sites of the pGEX-4T-1 vector to generate GST-G␣ o or between the NdeI and EcoRI restriction sites of the pET28b vector to generate His-G␣ o . GIV-CT-(1623-1870) was cloned from pcDNA 3.1-GIV (16) and inserted between the EcoRI and NotI restriction sites of the pGEX-4T-1 vector to generate GST-GIV-CT-(1623-1870). GIV-CTs-(1660 -1870, "s" stands for "short") was cloned from pcDNA 3.1-GIV and inserted between the NdeI and EcoRI restriction sites of the pET28b vector to generate His-GIV-CTs-(1660 -1870) used in the GTPase and GTP␥S binding assays. To generate G␣ i3 with three FLAG sequences fused to the C terminus of the protein (G␣ i3 -FLAG), rat G␣ i3 was amplified by PCR from pET28b-G␣ i3 with primers designed to add HindIII and BamHI restrictions sites at the 5Ј-and 3Ј-ends, respectively, and to replace the stop codon of the original cDNA with a glycine codon. The PCR product was digested and inserted between the HindIII and BamHI restriction sites of p3XFLAG-CMV-14. Untagged rat G␣ i3 cloned into pcDNA3.1 was described previously (18). Mouse LPA receptor 1 (LPAR1) cloned into pFLAG-CMV-1 was a gift from Dr. Jerold Chun (Scripps Research Institute) and was described previously (23). G␣ i3 and G␣ o mutants were generated using specific primers (sequences available upon request) following the manufacturer's instructions (QuikChange II, Stratagene). All constructs were checked by DNA sequencing (University of California at San Diego Moores Cancer Center Sequencing Facility).
Cell Culture, Transfection, and Lysis-COS7 and HeLa cells were grown at 37°C in DMEM supplemented with 10% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, 1% L-glutamine, and 5% CO 2 . siRNA transfection of HeLa cells was carried out using Oligofectamine (Invitrogen) following the manufacturer's protocol. Oligonucleotides against human G␣ i3 were from Santa Cruz Biotechnology. When reversal of phenotype was attempted, pcDNA3.1-G␣ i3 (untagged) transfection was carried out 8 -10 h post-siRNA transfection using GeneJuice (Novagen) following the manufacturer's protocol, and cells were analyzed after ϳ38 -40 h. Transfection of COS7 cells with G␣ i3 -FLAG was also carried out using GeneJuice. Lysates used as a source for GIV for in vitro protein binding assays or for immunoprecipitation were prepared by resuspending the cells in lysis buffer (20 mM HEPES, pH 7.2, 5 mM Mg(CH 3 COO) 2 , 125 mM K(CH 3 COO), 0.4% Triton X-100, 1 mM DTT) supplemented with phosphatase (Sigma) and protease (Roche Applied Science) inhibitor mixtures, passed through a 28-gauge needle at 4°C, and cleared (10,000 ϫ g for 10 min) before use in subsequent experiments.
Steady-state GTPase Assay-This assay was performed as described previously (11). Briefly, His-G␣ i3 or His-G␣ o (100 nM) was preincubated with different concentrations of His-GIV-CTs-(1660 -1870), for 15 min at 30°C in assay buffer (20 mM sodium HEPES, pH 8, 100 mM NaCl, 1 mM EDTA, 2 mM MgCl 2 , 1 mM DTT, 0.05% (w/v) C12E10). GTPase reactions were initiated at 30°C by adding an equal volume of assay buffer containing 1 M [␥-32 P]GTP (ϳ50 cpm/fmol). Duplicate aliquots (50 l) were removed at 10 min, and reactions were stopped with 950 l of ice-cold 5% (w/v) activated charcoal in 20 mM H 3 PO 4 , pH 3. Samples were then centrifuged for 10 min at 10,000 ϫ g, and 500 l of the resultant supernatant was scintillation-counted to quantify released [ 32 P]P i . To determine the specific P i produced, the background [ 32 P]P i detected at 10 min in the absence of G protein was subtracted from each reaction. The % reduction in G protein activation of G␣ o and G␣ i3 mutants compared with wt G␣ i3 was determined using the formula, 100 Ϫ [100*(x Ϫ 1)/(y Ϫ 1)], where x is the -fold increase in G protein activity for G␣ o and G␣ i3 mutants and y is the -fold increase for wt G␣ i3 .
Single-turnover GTPase Assay-This assay was performed as previously described (25). Briefly, His-G␣ i3 (500 nM) was loaded for 30 min at 30°C with GTP in the absence of magnesium to prevent the hydrolysis of the nucleotide (20 mM sodium HEPES, pH 8, 5 mM EDTA, 1 mM DTT, 4 M [␥-32 P]GTP (ϳ100 cpm/ fmol)). After transferring the tubes to ice, the reaction was initiated by diluting ten times the GTP-loaded G protein with assay buffer (20 mM sodium HEPES, pH 8, 80 mM NaCl, 5 mM EDTA, 7.5 mM MgCl 2 , 1 mM DTT, 200 M GTP, 0.05% (w/v) C12E10) containing the desired amount of protein to be tested (His-GIV-CTs, 2 M, GST-GAIP, 1 M, or equal volume of buffer in the control reaction). The reactions were carried out on ice. Aliquots (50 l) were removed at different time points (0, 20, 40, 60, 90, 120, 180, 240, and 300 s), and reactions were stopped with 950 l of ice-cold 5% (w/v) activated charcoal in 20 mM H 3 PO 4 , pH 3. Samples were then centrifuged for 10 min at 10,000 ϫ g, and 500 l of the resultant supernatant was scintillation-counted to quantify released [ 32 P]P i . To determine the specific P i produced, the background [ 32 P]P i detected at 0 s was subtracted from each reaction.
GTP␥S Binding Assay-GTP␥S binding was measured using a filter binding method (8,26). His-G␣ i3 (100 nM) was preincubated with different concentrations of wild-type His-GIV-CTs-(1660 -1870) or His-GIV-CTs-(1660 -1870) F1685A mutant, for 15 min at 30°C in assay buffer (20 mM sodium HEPES, pH 8, 100 mM NaCl, 1 mM EDTA, 25 mM MgCl 2 , 1 mM DTT, 0.05% (w/v) C12E10). Reactions were initiated at 30°C by adding an equal volume of assay buffer containing 1 M [ 35 S]GTP␥S (ϳ50 cpm/fmol). Duplicate aliquots (50 l) were removed at 15 min, and binding of radioactive nucleotide was stopped by addition of 3 ml of ice-cold wash buffer (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 25 mM MgCl 2 ). The quenched reactions were rapidly passed through BA-85 nitrocellulose filters (Amersham Biosciences) and washed with 4 ml wash buffer. Filters were dried and subjected to liquid scintillation counting. To determine the specific nucleotide binding, the background [ 35 S]GTP␥S detected in the absence of G protein was subtracted from each reaction.
Trypsinization of G␣ subunits-His-G␣ i3 , His-G␣ o , or the indicated His-G␣ i3 mutants (0.5 mg/ml) were incubated for 120 min at 30°C in the presence of GDP (30 M) or GDP⅐AlF 4 Ϫ (30 M GDP, 30 M AlCl 3 , 10 mM NaF). After incubation, trypsin was added to the tubes (final concentration, 12.5 g/ml), and samples were incubated for an additional 10 min at 30°C. Reactions were stopped by adding SDS-PAGE sample buffer and boiling. Proteins were resolved by SDS-PAGE and stained with Coomassie Blue.
Cell Migration Assays-Scratch wound assays were done as described previously (18). Briefly, 1-mm wounds were created in monolayer cell cultures (ϳ100% confluent) with a 1-ml sterile pipette tip, and the cells were subsequently monitored by phase-contrast microscopy over the succeeding 24 h. To quantify cell migration (expressed as % wound area covered), images were analyzed using ImageJ (National Institutes of Health) software to calculate the difference between the wound area at 0 h and at the end of the migration assay divided by the area at 0 h ϫ 100.
Statistical Analysis-Experiments were repeated at least three times, and results were presented either as one representative experiment or as mean Ϯ S.E. when data from multiple independent experiments were pooled. Statistical significance between various conditions was assessed with the Student's t test. p Ͻ 0.05 was considered significant.

Validation of GIV as a Bona
Fide GEF for G␣ i subunits-In our previous work (11) we demonstrated that GIV was capable of increasing the steady-state GTPase activity of G␣ i3 . The steady-state GTP hydrolysis by G␣ subunits is a reaction with two major steps, nucleotide exchange (i.e. release of GDP and loading of GTP) and GTP hydrolysis. The GTP hydrolysis step is 10-to 100-fold faster than the nucleotide exchange step. For this reason nucleotide exchange is rate-limiting for the steadystate GTPase reaction (27). Because GIV increases the steadystate GTPase activity of G␣ i3 and this activity directly depends on the rate of nucleotide exchange, we proposed that GIV is a GEF for G␣ i subunits.
To rule out an effect of GIV on GTP hydrolysis we performed single-turnover GTPase assays. Under these experimental conditions the nucleotide exchange step is bypassed and the GTPase activity depends solely on the rate of GTP hydrolysis (28). We measured the single-turnover GTPase activity of purified His-G␣ i3 in the absence or presence of GIV-CTs-(1660 -1870) ("s" stands for "short"), which contains the GEF motif of GIV. His-GIV-CTs was used, because it behaves the same as His-GIV-CT-(1623-1870) (13) in terms of its binding to G␣ i3 (data not shown) and modulation of steady-state GTPase activity ( Fig. 1A) but gives greater protein yields in E. coli. As a positive control we used GST-GAIP, a well characterized GAP (1) that accelerates the rate of GTP hydrolysis by G␣ i subunits. GST-GAIP dramatically increased the single-turnover GTPase activity of G␣ i3 , whereas GIV-CTs had no   (Fig. 1A). This result demonstrates that GIV does not affect GTP hydrolysis in the steady-state GTPase reaction.
To further validate the role of GIV as a GEF, we performed GTP␥S binding experiments, which directly measure nucleotide exchange activity (8,26). Purified His-G␣ i3 was incubated in the presence of increasing amounts of wild-type His-GIV-CTs or the G␣ i3 binding-deficient His-GIV-CTs F1685A mutant (11). GTP␥S binding to G␣ i3 was increased by His-GIV-CTs in a dose-dependent manner but was not significantly affected by His-GIV-CTs F1685A (Fig. 1C). At the maximal concentration of His-GIV-CTs tested (2 M) GTP␥S binding was increased up to ϳ2.2-fold over basal binding. This result indicates that GIV increases nucleotide exchange by G␣ i3 via its previously described GEF motif (11). Taken together these results demonstrate that GIV is a bona fide GEF for G␣ i3 and validates the steady-state GTPase activity as a direct measure of its GEF activity.
Comparative Binding of GIV to G␣ i3 and G␣ o -We previously reported that GIV interacts with members of the G i (includes G␣ i1 , G␣ i2 , G␣ i3 , and G␣ o ) and G s subfamilies of G proteins in two-hybrid assays (14); however, in in vitro pulldown assays GIV preferentially binds to G␣ i3 versus G␣ s and binds as efficiently to G␣ i1 and G␣ i2 (11,18) as to G␣ i3 . To gain insights into the structural features responsible for the preferential binding of GIV to G␣ i subunits, we compared GIV binding to G␣ i3 and G␣ o1 (hereafter referred as G␣ o ), which also belongs to the G i family and is the closest to G␣ i subunits in sequence and structure (29). We found that both GST-G␣ i3 and GST-G␣ o interact with endogenous GIV in pulldown assays on COS7 lysates when the G protein is preloaded with GDP (inactive state ( Fig. 2A)); however, GST-G␣ i3 bound ϳ15to 20-fold more GIV than GST-G␣ o . Neither G␣ i3 nor G␣ o bound GIV when preloaded with GDP and AlF 4 Ϫ (mimicking the active state) ( Fig. 2A). Because the state-dependent interaction of GIV with inactive G␣ i is mediated by the GEF motif located in the C terminus, we next investigated the ability of both GST-G␣ i3 and GST-G␣ o immobilized on beads to bind purified His-GIV-CT-(1623-1870) in vitro. The findings were similar to those obtained for endogenous GIV from cell lysates: His-GIV-CT bound preferentially to inactive (GDP-bound) G␣ i3 and G␣ o , and binding to G␣ i3 was ϳ10-fold greater than to G␣ o (Fig. 2B). These results indicate that, although GIV-CT can bind to both G␣ i3 ⅐GDP and G␣ o ⅐GDP, binding to G␣ o ⅐GDP is much less efficient.
Comparative Activation of G␣ i3 and G␣ o by GIV-Based on our recent finding (11) that binding of the GEF motif of GIV to the G␣ subunit is required for GIV to exert its GEF function, we hypothesized that its decreased binding to G␣ o ⅐GDP might affect its GEF activity. To test if this is the case, we measured the steady-state GTPase activity of purified His-G␣ i3 and His-G␣ o in the presence of increasing amounts of purified GIV-CTs. The relative increase in the steady-state GTP hydrolysis exerted by GIV-CTs on G␣ o was significantly reduced compared to its effect on G␣ i3 at all concentrations tested (Fig. 2C). At the maximal concentration of GIV-CTs tested (2 M), activation of G␣ o was ϳ75% less than G␣ i3 (Table 1). Taken together, these results demonstrate that GIV is a more efficient GEF for G␣ i3 than for G␣ o , and thus is capable of discriminating among G␣ subunits of the G i subfamily.
Mapping the Region of the G␣ Subunit That Determines Preferential Binding of GIV to G␣ i -G␣ i1 , G␣ i2 , G␣ i3 , and G␣ o , share ϳ75% sequence identity and have a similar tertiary structure (29). In our previous work we found that the GEF motif of GIV binding to the switch II on G␣ i subunits (11), but this region is 100% identical in G␣ o . We reasoned that another region of G␣ i3 must be responsible for GIV binding. To localize the region that specifies the strength of the interaction of G␣ i3 with GIV, we generated a number of GST-G␣ i3 /G␣ o chimeras (Fig. 3A). wt G␣ i3 and G␣ i3/o chimera 1 (contains the all-helical Ϫ -loaded G proteins is absent in all cases (lanes 3, 5, 7, 9, 11, and 13). COS7 cell lysates were incubated with purified GST (lanes 2 and 3), GST-G␣ i3 (lanes 4 and 5) or the indicated GST-G␣ i3/o chimeras (lanes 6 -13) pre-loaded with GDP (lanes 2, 4, 6, 8, 10, and 12) or GDP⅐AlF 4 Ϫ (lanes 3, 5,7,9,11, and 13) immobilized on glutathione beads and analyzed as in Fig. 2A. A higher exposure of the same immunoblot (middle panel) shows the weak binding observed for GST-G␣ i3/o chimera 2 (lane 8) and 4 (lane 12). Equal loading of GST proteins was confirmed by Ponceau S staining (lower panel).

TABLE 1 Comparative activation of G␣ i3 and G␣ o by GIV-CTs
Experiments were performed as described for Fig 2C. Reduction in G protein activation by GIV-CTs was calculated as percent decrease in the -fold activation of G␣ o compared to -fold activation of G␣ i3 as described under "Experimental Procedures." All parameters are expressed as mean Ϯ S.E. of n ϭ 9 independent experiments. domain of G␣ o ), bound similar amounts of endogenous GIV from COS7 lysates, whereas G␣ i3/o chimera 2 (contains the Raslike domain of G␣ o ), showed dramatically reduced GIV binding (Fig. 3B). In addition, chimera 3 (contains the C-terminal half, aa 272-354) but not chimera 4 (contains the N-terminal half, aa 178 -271 of the Ras-like domain of G␣ o ) bound as much GIV as wt G␣ i3 (Fig. 3B). None of the chimeras bound GIV when they were preloaded with GDP⅐AlF 4 Ϫ (Fig. 3B). These results indicate . Trp-258 of G␣ i3 is responsible for GIV binding and activation of G␣ i3 . A, sequence alignment of G␣ o , G␣ i1 , G␣ i2 , and G␣ i3 indicating the G␣ i3 mutants studied. Rat G␣ o , G␣ i1 , G␣ i2 , and G␣ i3 sequences corresponding to G␣ i3 aa 178 -270 were obtained form the NCBI data base and aligned using ClustalW. Conserved identical residues are in black; similar residues are shaded in gray. The secondary structure elements (␣ ϭ ␣-helix, ␤ ϭ ␤-sheet) indicated below the alignment are named according to their crystal structures (29,33). Residues conserved among G␣ i1 , G␣ i2 , and G␣ i3 but different in G␣ o within the 178 -270 region were mutated in GST-G␣ i3 to the corresponding residue in G␣ o (indicated above with arrows). B, mutations W258F/T260I and W258F, in the ␣3/␤5 loop, impair endogenous GIV binding to GDP-loaded GST-G␣ i3 , whereas mutations G217D, L232Q, A235H/E239T/M240T, and K248M do not affect binding (upper panels). GIV binding to GDP⅐AlF 4 Ϫ -loaded G proteins is virtually absent in all cases (lower panels). COS7 cell lysates were incubated with ϳ5 g of purified GST-G␣ i3 or the indicated GST-G␣ i3 mutants pre-loaded with GDP (upper panels) or GDP⅐AlF 4

Reduction in G protein activation by GIV-CTs compared to wt G␣i3
Ϫ (lower panels) immobilized on glutathione beads and analyzed as in Fig. 2A. Equal loading of GST proteins was confirmed by Ponceau S staining. C, binding of GST-GIV-CT to His-G␣ i3 W258F (lane 6) is reduced (ϳ80%) compared with wt His-G␣ i3 (lane 3). ϳ3 g of purified wt His-G␣ i3 (lanes 1-3) or His-G␣ i3 W258F (lanes 4 -6) pre-loaded with GDP were incubated with where switches I, II, and III are located (Fig. 3A) contains the determinants that specify the preferential binding of GIV to G␣ i3 versus G␣ o .
Identification of a Single Residue That Determines GIV Preferential Binding and Activation of G␣ i3 -We reasoned that one or several of the residues within aa 178 -270 of G␣ i3 that are conserved among G␣ i1 , G␣ i2 , and G␣ i3 but different in G␣ o must be responsible for the preferential binding of GIV to G␣ i subunits. To identify such residues we aligned the sequences of G␣ o , G␣ i1 , G␣ i2 , and G␣ i3 (Fig. 4A). We mutated residues conserved among G␣ i subunits but differing from G␣ o to the corresponding amino acids of G␣ o (specified in Fig. 4A) and tested their ability to bind endogenous GIV in GST pulldown assays on COS7 cell lysates. The G␣ i3 W258F/T260I double mutant and the G␣ i3 W258F single mutant showed a dramatic reduction in GIV binding, whereas the remainder of the mutants were the same as wt G␣ i3 (Fig. 4B). Importantly, the GST-G␣ i3 W258F mutant bound as much AGS3 and G␤␥ as wt G␣ i3 (data not shown), suggesting that the decreased GIV binding is specific and not due to an overall effect on the structure of G␣ subunits. In addition to the mutants specified in Fig. 4A, two other mutants (G␣ i3 E193N/Y195H/K197R/M198L and G␣ i3 D229G/L232Q) were used in similar pulldown assays, but no differences in GIV binding were observed from wt G␣ i3 (data not shown).
To further confirm that the W258F mutation directly affects interaction between GIV and the G protein we carried out protein interaction assays with purified recombinant proteins. GST-GIV-CT was immobilized on glutathione-agarose beads and incubated with His-tagged wt G␣ i3 or G␣ i3 W258F. Binding of His-G␣ i3 W258F was dramatically reduced compared to wt G␣ i3 (Fig.  4C). Thus, Trp-258, which is located in the ␣3/␤5 loop of G␣ i , is a critical determinant of its interaction with GIV.
We next investigated the GEF activity of GIV on G␣ i3 W258F and found that the relative increase in steady-state GTP hydrolysis exerted by GIV-CTs on G␣ i3 W258F was significantly reduced compared with wt G␣ i3 (Fig. 4D). At the maximal concentration of the GEF tested (2 M), activation of G␣ i3 W258F was reduced Ͼ65% (Table 2), which is comparable to that observed for G␣ o (Fig. 2C and Table 1). In addition, we tested the GEF activity of GIV on a G␣ o mutant in which Phe-259 was replaced by the corresponding aa (Trp-258) in G␣ i3 . We found that mutation of Phe-259 to Trp enhanced the relative increase in steady-state GTP hydrolysis exerted by GIV-CTs on G␣ o ϳ2-fold (Fig. 4E), indicating that the F259W mutation is sufficient to make G␣ o a better substrate for the GEF activity of GIV. From these results we conclude that mutation of Trp-258 in the ␣3/␤5 loop of G␣ i3 to the corresponding amino acid (Phe) in G␣ o dramatically reduces GIV binding and accounts for the reduced GEF activity of GIV on G␣ o .
Trp-258 Is Critical for Activation of G␣ i3 by GIV-Mutation of Trp-258 to Phe (aromatic to an aromatic side chain) is a conservative mutation; yet it significantly reduces G␣ i3 activation by GIV. Based on this finding we reasoned that functional G␣ i3 -GIV coupling should be very sensitive to alterations in aa 258 of G␣ i3 . To test if this is the case, we performed further mutational analysis by replacing Trp-258 with amino acids of different nature, i.e. tyrosine (aromatic), leucine (aliphatic), serine (polar), histidine (basic), aspartate (acidic), and alanine (small). The mutants were then tested for their response to GIV in steady-state GTPase assays. All the G␣ i3 Trp-258 mutants showed reduced activation by GIV-CTs compared to wt G␣ i3 , but the relative decrease in GTPase hydrolysis varied (Fig. 5A). At the maximal concentration of GIV-CTs tested (2 M), G protein activation was as follows: Table 2). The W258A and W258D mutants showed a ϳ90 -95% reduction and W258Y a ϳ60% reduction in activation, a value very similar to that observed for W258F (Table 2). These results highlight the specific requirement for Trp in position 258 to achieve maximal activation by GIV.
Results from pulldown assays with purified recombinant proteins were consistent with the results of the GTPase assay: binding of GST-GIV-CT to His-G␣ i3 W258A or W258D was virtually abolished (Fig. 5B) and less than to His-G␣ i3 W258F (Fig. 5B). Thus the extent of GIV binding parallels the extent of G␣ i3 activation (Figs. 4D and 5A and Table 2). Some mutations that cause reduced activation of G␣ subunits by GPCRs also decrease their activation by AlF 4 Ϫ (30 -32), which activates G proteins by mimicking the ␥-phosphate of GTP in the transition state (33). To test the ability of the G␣ mutants to be activated by AlF 4 Ϫ we took advantage of a well established assay based on differential resistance to proteolysis (31,34). When G␣ subunits are in the inactive GDP-bound conformation, they are readily digested by trypsin, whereas upon AlF 4 Ϫ binding and adoption of the active conformation only a short sequence can be cleaved, and the remainder of the protein remains trypsin-resistant (34). All the His-G␣ i3 Trp-258 mutants (see Table 2) as well as His-G␣ o behaved like wt His-G␣ i3 in that they were hydrolyzed by trypsin when preloaded with GDP but generated a trypsin-resistant form when preloaded with GDP⅐AlF 4 Ϫ (Fig. 6). Thus all the G␣ subunits tested can efficiently adopt the active conformation upon AlF 4 Ϫ binding. Because activation by AlF 4 Ϫ requires the nucleotide binding site to contain GDP and the G protein to be in an appropriate conformation, these results also indicate that the G␣ proteins are properly folded. In addition, all the G␣ i3 Trp-258 mutants had similar basal rates of steady-state GTPase hydrolysis (see Table 2), suggesting that the spontaneous

Comparative activation of wt G␣i3 and G␣i3 W258 mutants by GIV-CTs
Experiments were performed as described for Fig 5A. Reduction in G protein activation by GIV-CTs was calculated as percent decrease in the -fold activation of each G␣i3 mutant compared to -fold activation of wt G␣i3 as described under "Experimental Procedures." All parameters are expressed as mean Ϯ S.E. of n ϭ 3-9 independent experiments. exchange of nucleotide is unaffected and that they fold properly and maintain their native properties. Collectively, these results support the conclusion that mutations in position 258 of G␣ i3  Ϫ . His-G␣ i3 and the indicated His-G␣ i3 mutants or His-G␣ o (0.5 mg/ml) were incubated in the presence of GDP or GDP⅐AlF 4

Basal
Ϫ and treated or not with trypsin as described under "Experimental Procedures." A Coomassie Blue-stained gel of a representative experiment is shown. The arrowhead denotes the position of the non-trypsinized fulllength proteins loaded, and the asterisk arrowhead denotes the trypsin-resistant form of the active, GDP⅐AlF 4 Ϫ -loaded His-G␣ subunit. specifically alter GIV-catalyzed activation without causing global structural changes in the G␣ subunit.
Mutation of Trp-258 Impairs G␣ i3 Binding to GIV in Cultured Cells-Next, we investigated the effect of mutating Trp-258 on the interaction between GIV and the G protein in cultured cells. COS7 cells were transfected with FLAG-tagged wt G␣ i3 and G␣ i3 mutants, and immunoprecipitation was carried out using anti-FLAG IgG followed by immunoblotting for GIV. We found that the amount of endogenous GIV that co-immunoprecipitated with the G␣ i3 mutants W258F, W258A, and W258D was dramatically reduced compared with wt G␣ i3 (Fig.  7). These results demonstrate that mutation of Trp-258 impairs the interaction of G␣ i3 with endogenous GIV in cultured cells, corroborating our observations in vitro.
Mutation of Trp-258 to Phe Does Not Affect Binding of G␤␥, AGS3, GAIP/RGS19, or LPAR1 to G␣ i3 -We next investigated if mutation of Trp-258 of G␣ i3 to Phe interferes with its interaction with other binding partners such as G␤␥ and AGS3 (a G␣ i -guanine nucleotide dissociation inhibitor (21,35)). We also investigated the behavior of G␣ i3 W258A and W258D, because these two mutants have reduced GIV binding both in vitro and in cultured cells (Figs. 5 and 7) and show the most dramatic reduction in activation by GIV (Fig. 5 and Table 2). We found that G␣ i3 W258F binds G␤␥ and AGS3 as efficiently as wt G␣ i3 in co-immunoprecipitation assays (Fig. 7), whereas binding of G␣ i3 W258A or W258D to G␤␥ and AGS3 was dra-matically reduced (Fig. 7). From these results we conclude that mutation of Trp-258 to Phe specifically impairs G␣ i3 interaction with GIV without affecting its interaction with G␤␥ and AGS3, whereas this is not the case for mutation of Trp-258 to Asp or Ala.
Thus mutation of Trp-258 specifically to Phe can be tolerated for the interaction of the G protein with binding partners other than GIV (e.g. G␤␥ and AGS3), whereas mutation of Trp-258 to Ala or Asp most likely affects the structural properties of the G protein such that they impair its interaction with G␤␥ and AGS3. We further investigated the ability of G␣ i3 W258F to interact with GAIP (RGS19) (22), a GAP for G␣ i subunits (36). Identical results were obtained with wt G␣ i3 and G␣ i3 W258F: GST-GAIP bound robustly to G␣ i3 preloaded with GDP-AlF 4 Ϫ but showed virtually no binding to inactive, GDP-loaded G␣ i3 (22) (Fig. 8A), demonstrating that this mutation does not compromise the interaction with GAIP.
We also investigated the ability of G␣ i3 W258F to interact with LPAR1, a GPCR that couples to G␣ i/o subunits (23,37,38). COS7 cells were co-transfected with FLAG-tagged LPAR1 and untagged wt G␣ i3 or G␣ i3 W258F, and immunoprecipitation was carried out using anti-FLAG IgG followed by immunoblotting for G␣ i3 . We found that the amount of G␣ i3 that co-immunoprecipitated with the receptor from lysates of COS7 cells transfected with either wt G␣ i3 or G␣ i3 W258F was virtually the same (Fig. 8B), indicating that mutation of Trp-258 to Phe does Endogenous G␣ i3 (lane 4) is not detected in FLAG-LPAR1 immunoprecipitates, and no G␣ i3 is present in FLAG-immunoprecipitates from controls (lanes 1-3). 48 h after transfection cells were harvested and lysates used for immunoprecipitation (IP) with anti-FLAG IgG (ϳ2 g) as described under "Experimental Procedures." IP was followed by immunoblotting (IB) for FLAG (LPAR1) and G␣ i3 . Equal IgG loading was confirmed by Ponceau S staining. Lower panel, aliquots of cell lysates (ϳ5%) were analyzed for expression of FLAG (LPAR1), G␣ i3 , and ␣-tubulin by immunoblotting (IB) to confirm the expression of the analyzed proteins.
not affect the interaction of G␣ i3 with LPAR1, a GPCR. Taken together, these results indicate that the W258F, but not the W258A or W258D mutation, specifically impairs activation of G␣ i3 by GIV without perturbing other known interactions of the G protein.
G␣ i3 W258F Fails to Enhance LPA-and Insulin-stimulated Akt Activation and to Promote Cell Migration-We have previously shown that activation of G␣ i3 enhances Akt signaling after stimulation of both GPCRs and receptor tyrosine kinases (18) and that the GEF motif of GIV is required for these functions (11). These effects might be triggered directly by activa-tion of the G protein by GIV, or alternatively, they could be enhanced by GIV-independent activation of G␣ i3 . To distinguish between these two possibilities we took advantage of the GIV-insensitive G␣ i3 W258F mutant. G␣ i3 was depleted (Ͼ95%) in HeLa cells using siRNA oligonucleotides that specifically target the human sequence of G␣ i3 (18), and Akt activation was measured in response to stimulation of either the LPA receptor, a GPCR that enhances Akt signaling by activating G i proteins (37,38), or the insulin receptor, a receptor tyrosine kinase. When serum-starved HeLa cells were stimulated with either LPA (Fig. 9A) or insulin (Fig.  9B), activation of Akt was dramatically reduced (ϳ70%) in G␣ i3 -depleted cells compared with controls, and this effect could be reversed by expression of wt rat G␣ i3 (which is insensitive to human G␣ i3 -specific siRNA oligonucleotides (18)). G␣ i3 depletion also impaired the ability of HeLa cells to migrate efficiently in scratch-wound assays (Fig. 9, C-E), and this effect was restored by expression of rat G␣ i3 wt. By contrast, expression of rat G␣ i3 W258F failed to restore Akt activation in response to either LPA or insulin (Fig. 9, A and B), and to reverse the defect on cell migration (Fig. 9, C-E). From these results we conclude that direct activation of G␣ i3 by GIV is required to enhance Akt signaling and to promote cell migration after growth factor or GPCR stimulation.

DISCUSSION
GIV is a recently characterized non-receptor GEF that can activate G␣ i3 (11). The major finding in this work is the identification of a novel structural determinant on G␣ i3 that renders the G protein sensitive to activation by GIV. This structural determinant is required to promote efficient cell migration and Akt signaling, two cell functions that we have previously shown to be triggered by active G␣ i3 (18). Using site-directed mutagenesis, we demonstrate here that Trp-258 located in the ␣3/␤5 loop of the Ras-like domain of G␣ i subunits is required to establish an efficient interaction with GIV and to activate the G protein. When Trp-258 is mutated to Phe, G␣ i3 is less efficiently activated by GIV, but it retains its ability to interact with G␤␥ subunits, AGS3, GAIP (RGS19), and FIGURE 9. G␣ i3 W258F fails to rescue Akt activation and cell migration defects observed upon depletion of endogenous G␣ i3 . A and B, HeLa cells treated with scrambled (Scr) or hG␣ i3 siRNA oligonucleotides and the indicated DNA plasmids (empty vector, rG␣ i3 WT, or rG␣ i3 W258F) were serum-starved for 6 h and stimulated with 10 M LPA (A) or 100 nM insulin (B) for 5 min. Cell lysates were analyzed by immunoblotting (IB) for total (tAkt) and S473 phospho-Akt (pAkt), G␣ i3 , and ␣-tubulin. Depletion of endogenous G␣ i3 reduces LPA-stimulated (A) and insulin-stimulated (B) Akt activation by ϳ70%, which is restored upon transfection of wt rG␣ i3 but not rG␣ i3 W258F. C, in controls (Scr siRNA), HeLa cells cover the majority of the experimental wound area after 24 h, whereas in G␣ i3 -depleted cells wound closure is greatly impaired. The ability to migrate and close the wound area at 24 h is restored by transfection of rG␣ i3 wt but not rG␣ i3 W258F. HeLa cell monolayers treated with the indicated siRNA oligonucleotides, and DNA plasmids were scratch-wounded and examined by light microscopy immediately (0 h) or 24 h after wounding. Scale bar ϭ 500 m. D, bar graph showing quantification of the wound area covered by cells in C. The area covered by cells was determined by calculating the difference between the wound area at 0 and 24 h expressed as percent of the wound area at 0 h. Results are shown as mean Ϯ S.D. of 8 -12 randomly chosen fields from three independent experiments. E, cell lysates from cells treated as in C were immunoblotted to assess the efficiency of siRNA depletion of G␣ i3 (ϳ95%) and the expression of rG␣ i3 WT or rG␣ i3 W258F.
Mutants that selectively abolish the ability of G␣ i subunits to be regulated by GAPs (44) or guanine nucleotide dissociation inhibitors (45) have been described and used to evaluate the role of these regulators in G i functions (44 -50). G␣ i3 W258F represents a new addition to the growing battery of mutants that can be used to finely dissect how different regulators of G protein activity control cell fate. Although G i -coupled GPCRs also couple efficiently to G␣ o subunits (51)(52)(53)(54), here we show that replacement of a single residue (Trp-258) in G␣ i3 for the corresponding residue in G␣ o (Phe) dramatically and specifically reduces G protein coupling to GIV. Using this mutant we provide evidence that cell migration and Akt signaling, cell functions previously described to be promoted by constitutively active G␣ i3 mutants (18), require G␣ i3 to be specifically activated by GIV. The G␣ i3 W258F mutant will also be useful in the future to distinguish other functions of G␣ i subunits controlled by the GEF activity of GIV.
The data presented here suggest that the footprint of the GIV GEF domain on G␣ i3 probably extends from the previously described binding site within switch II (11) to make contact with an additional binding site located in the ␣3/␤5 loop. Based on our homology modeling (11) depicted in Fig. 10B, the GEF motif of GIV docks within the groove formed between the switch II and the ␣3 helix of the G protein. The location of the novel structural determinant in G␣ i3 required for binding of GIV raises the interesting possibility that GIV residues C-terminal to the previously described GEF motif may be involved in making direct contact with the ␣3/␤5 loop region surrounding Trp-258. However, at this point allosteric effects cannot be ruled out to explain the decreased interaction between GIV and G␣ i3 upon mutation of Trp-258. Nevertheless, our unpublished work 6 favors the possibility of a direct contact site, because mutation of GIV residues C-terminal to the previously described GEF motif impairs the G␣ i3 -GIV interaction.
Our identification of a novel structural determinant in G␣ i3 required for GIV binding provides insights that may help in the design of selective pharmacological agents that disrupt the G␣ i -GIV interface for therapeutic purposes. We previously found that expression of full-length GIV is induced severalfold in cancer cell lines that are highly metastatic (18) and that mutational disruption of the GIV binding to switch II of G␣ i abolishes Akt signaling and tumor cell migration (11), which are hallmarks of cancer metastasis (55)(56)(57)(58). Here we describe that GIV binding and these functions are similarly abolished by mutational disruption of the ␣3/␤5 loop of G␣ i3 . Because alterations in the switch II may also impair the binding of other molecules to the G protein (59 -61), the ␣3/␤5 loop represents a more attractive target than the previously described switch II binding site. Moreover, we show here that alterations in aa 258 of G␣ i3 can impair GIV binding and abolish cell functions controlled by GIV without affecting the interaction of G␣ i3 with its other binding partners G␤␥, AGS3, GAIP/RGS19, and LPAR1. Thus we envision that small molecules that target this site might work as anti-metastatic agents by specifically disrupting GIV-G␣ i interaction.