Gβγ signaling to the chemotactic effector P-REX1 and mammalian cell migration is directly regulated by Gαq and Gα13 proteins

G protein–coupled receptors stimulate Rho guanine nucleotide exchange factors that promote mammalian cell migration. Rac and Rho GTPases exert opposing effects on cell morphology and are stimulated downstream of Gβγ and Gα12/13 or Gαq, respectively. These Gα subunits might in turn favor Rho pathways by preventing Gβγ signaling to Rac. Here, we investigated whether Gβγ signaling to phosphatidylinositol 3,4,5-trisphosphate–dependent Rac exchange factor 1 (P-REX1), a key Gβγ chemotactic effector, is directly controlled by Rho-activating Gα subunits. We show that pharmacological inhibition of Gαq makes P-REX1 activation by Gq/Gi-coupled lysophosphatidic acid receptors more effective. Moreover, chemogenetic control of Gi and Gq by designer receptors exclusively activated by designer drugs (DREADDs) confirmed that Gi differentially activates P-REX1. GTPase-deficient GαqQL and Gα13QL variants formed stable complexes with Gβγ, impairing its interaction with P-REX1. The N-terminal regions of these variants were essential for stable interaction with Gβγ. Pulldown assays revealed that chimeric Gα13-i2QL interacts with Gβγ unlike to Gαi2–13QL, the reciprocal chimera, which similarly to Gαi2QL could not interact with Gβγ. Moreover, Gβγ was part of tetrameric Gβγ–GαqQL–RGS2 and Gβγ–Gα13-i2QL–RGS4 complexes, whereas Gα13QL dissociated from Gβγ to interact with the PDZ–RhoGEF–RGS domain. Consistent with an integrated response, Gβγ and AKT kinase were associated with active SDF-1/CXCL12–stimulated P-REX1. This pathway was inhibited by GαqQL and Gα13QL, which also prevented CXCR4-dependent cell migration. We conclude that a coordinated mechanism prioritizes Gαq- and Gα13-mediated signaling to Rho over a Gβγ-dependent Rac pathway, attributed to heterotrimeric Gi proteins.

GPCR-dependent 2 chemotactic signaling is a phylogenetically conserved process that plays a central role during development and homeostatic control of multicellular organisms (1,2). Life-threatening pathologies such as metastatic cancer also involve directional cell migration guided by chemotactic GPCRs (3,4). Shallow gradients of chemokines elicit a spatiotemporal signaling response that defines cell polarity by generating an asymmetric distribution of actin nucleation sites. The force generated by actin polymerization pushes the cell to move forward. This process is particularly relevant for individual migrating cells (1,2).
Chemotactic GPCRs, such as CXCR4 and LPA1, are frequently coupled to different families of G proteins, including G i , G q/11 , and G 12/13 , but their ability to sustain a migratory response is commonly sensitive to inhibition by pertussis toxin (5)(6)(7)(8). This is consistent with the widely accepted idea that signaling-ready G␤␥ complexes, suitable for chemotactic events, preferentially come from G i (9 -11), which has recently been confirmed by synthetic biology strategies (12). Mechanistically, this process involves G i -dependent activation of Rho GTPases, particularly Rac and Cdc42, downstream of a G␤␥-regulated group of effectors (4). These include phosphoinositide 3-kinases PI3K␤ and PI3K␥ (13-17) and 3-phosphoinositide-dependent Rac guanine nucleotide exchange factors P-REX1 and P-REX2 (18 -20). Interestingly, aberrantly overexpressed chemotactic GPCRs, such as CXCR4, highly abundant in various types of metastatic cancer, abnormally couple to G13 leading to a predominant Rho-dependent migratory pathway (21).
G␣ subunits (particularly G␣ 12/13 and G␣ q ) activate Rho via direct interaction with RGS-RhoGEFs and p63RhoGEF/TRIO, respectively (22)(23)(24)(25)(26). Because Rho and Rac have opposite effects on cell morphology, they are considered contradictory signals that have to be spatiotemporally regulated in migrating cells where Rho promotes contraction, whereas Rac stimulates extension (27)(28)(29). Therefore, diverse molecular mechanisms are believed to exert a spatiotemporal control of their activity. In the case of G 12/13 -and G q -coupled receptors, their signals to Rho are prevalent over the potential G␤␥-dependent mechanisms activating Rac that would be expected to result by the action of G␤␥ subunits from these heterotrimeric G proteins. These observations opened the possibility that G␤␥ signaling by G 12/13 and G q heterotrimers is directly regulated by the respective G␣ subunits. This hypothesis would be supported by the possibility that inhibition of certain G␣ subunits, such as G␣ q , might improve G␤␥ signaling. In addition, GTPase-deficient G␣ subunits might stably interact with G␤␥ dimers regulating their availability to bind their downstream effectors. Here, we tested these possibilities by analyzing the pathway linking G␤␥ signaling to P-REX1, its chemotactic effector, and the G␤␥-dependent activation of AKT and cell migration.

G q /G i -coupled LPA receptors exhibit increased signaling to P-REX1 when G␣ q is inhibited
G␤␥-dependent pathways promoting chemotactic processes are usually derived from G i , considered one of the most abundant heterotrimeric G proteins and characterized by their sensitivity to pertussis toxin (4,30,31). We hypothesized that G␤␥ signaling to Rac, via P-REX1, is mainly elicited by G i not only because of its high expression but also because other G␣ subunits, particularly those that activate Rho, such as G␣ q and G␣ 13 , may prevent G␤␥ availability to interact with P-REX1. Therefore, inhibition of G␣ q , for instance, would lead to higher effect of G␤␥ on P-REX1. As an endogenous system to test this possibility, we used MCF-7 cells to study signaling by G q /G icoupled LPA receptors (Fig. 1A). We chose MCF-7 cells because they express P-REX1 that is known to be regulated by G i -coupled CXCR4 receptors (32). They also express LPA receptors, which are characterized by their ability to couple to G i , G q , and G 13 families of heterotrimeric G proteins (33). P-REX1 activation was measured by RacG15A pulldown as previously described (34). In these cells, LPA activated ERK via G␣ q (Fig. 1B) but not G i (Fig. 1C) and had a limited G i -dependent effect on P-REX1, which decreased below its basal activation levels when G i was inhibited with pertussis toxin (Fig. 1C). The activation of ERK by LPA was sensitive to FR900359 (FR, iGq), a recently described specific inhibitor of G␣ q (35), and sotrastaurin (iPKC), a PKC inhibitor (Fig. 1B, top and bottom panels, respectively), but insensitive to pertussis toxin (Fig. 1C). In contrast, the small effect on P-REX1 elicited by LPA was inhibited by PTX (Fig. 1C). Because P-REX1 is known to be activated by G␤␥, the lack of activation of P-REX1 by G q , expected to release G␤␥, opens the possibility that G␣ q might inhibit G␤␥ signaling. To test this possibility, we assessed the activation of P-REX1 in cells in which G␣ q was inhibited. Consistent with this idea, P-REX1 reached higher levels of activation by LPA when G␣ q was inhibited with FR (iGq) (Fig. 1D). In these cells, with transfection with RGS2, which blocks G␣ q (36), the basal levels of active P-REX1 were increased, making the effect of LPA hardly evident but still significant compared with the effect in cells not overexpressing RGS2 (Fig. 1E). A similar effect on AKT activation was observed (Fig. 1F), RGS2 overexpression increased basal pAKT, which in this case was still sensitive to LPA stimulation. The increased activation of P-REX1 by LPA in RGS2-transfected cells was sensitive to pertussis toxin (Fig. 1F), which partially reduced the effect on AKT (Fig. 1F).

Chemogenetic evidence showing that G i -coupled but not G q -coupled receptors activate P-REX1
To confirm that endogenous G i preferentially activates P-REX1, we followed a chemogenetic approach using genetically modified receptors exclusively coupled to G i ( Fig. 2A, G i -DREADD) and compared with others exclusively coupled to G q (Fig. 2B, G q -DREADD), which according to our previous experiments would not provide enough signaling-ready G␤␥ to activate P-REX1. These receptors are susceptible to be activated solely by clozapine-N-oxide (CNO), a synthetic ligand without effects on cells not expressing DREADD (37,38). First, we assessed the time course, dose response, and specificity of CNO to verify the suitability of the DREADD system to evaluate AKT/ERK and P-REX1 signaling pathways in COS7 cells. G i -DREADD transfected COS7 cells were stimulated with 1 M CNO in time-course experiments; in these cells, G i -DREADD strongly activated AKT (Fig. 2C, pAKT, top panel and graph) and ERK (Fig. 2D, pERK, top panel and graph) with maximum effects between 5 and 15 min. As expected, phosphorylation of AKT and ERK in G i -DREADD transfected cells stimulated with CNO was dose-dependent (Fig. 2, E and F, respectively), with a maximum response reached at 1 M, which was selected for next experiments. Even at maximum concentration, CNO was inactive in control cells not expressing DREADD (Fig. 2, E and F, control, graphs and lower panels), confirming the suitability of the system to characterize specific G protein-dependent signaling pathways. We further confirmed the distinct G protein coupling of G i -and G q -DREADD using G i and G q specific inhibitors and evaluated AKT and ERK activation in response to CNO. Consistent with the reported specificity, G i -DREADD effects on AKT and ERK were inhibited with pertussis toxin (PTX, a specific G i inhibitor) but not by G q family specific inhibitor FR900359 (35) (Fig. 2G, upper panels, iGq), whereas the activation of AKT and ERK by G q -DREADD was prevented by the G q inhibitor FR, but not by PTX (Fig. 2H, upper panels).
Subsequently, we tested the time course of P-REX1 activation by CNO in G i -DREADD transfected COS7 cells. As shown in Fig. 2I, maximal activation of P-REX1 was detected between 15 and 30 min. Consistently, G␤␥ was found associated with active P-REX1 in the RacG15A pulldowns (Fig. 2I, G␤ in PD). Finally, we compared the activation of P-REX1, AKT, and ERK by G i -DREADD versus G q -DREADD. Consistent with the differential ability of G i versus G q to provide signaling-ready G␤␥, G i -DREADD but not G q -DREADD activated P-REX1 (Fig. 2J). In addition, G i -DREADD activated AKT more effectively than G q -DREADD (Fig. 2K, pAKT, top left panel and graph), whereas both receptors were equally effective activators of ERK (Fig. 2K, pERK, top right panel and graph). In this case, the activation of ERK by G q -DREADD occurred via G␣ q (Fig. 2H, pERK, second panel) and was susceptible to PKC inhibition (not G␣ q -QL and G␣ 13 -QL prevent G␤␥-dependent P-REX1 activation Figure 1. G q /G i -coupled LPA receptors exhibit increased signaling to P-REX1 when G␣ q is inhibited. A, proposed model of G␤␥ signaling by G q /G icoupled LPA receptors in MCF7 cells. Accordingly, G␤␥ signaling to P-REX1 is elicited by G i but not G q because of an inhibitory role of G␣ q on G␤␥. Therefore, inhibition of G␣q would increase G␤␥-mediated activation of P-REX1. B, LPA activates ERK via G q and not G i . MCF7 cells were stimulated with 5 M LPA in the absence or presence of either FR900359 (250 nM, iGq), 1 M sotrastaurin (PKC inhibitor), or PTX (100 ng/ml overnight, C), and ERK activation was assessed by Western blotting against phosphorylated ERK (pERK). C, LPA activates P-REX1 via G i . MCF7 cells were preincubated or not with PTX (100 ng/ml overnight) and used to detect the effect of 5 M LPA on P-REX1 activation assayed by RacG15A pulldown. D, LPA-dependent activation of P-REX1 is increased as a consequence of G␣ q inhibition. The cells were pretreated with FR900359 (iGq) and stimulated with 5 M LPA, and activation of P-REX1 and ERK was tested. The graph represents the means Ϯ S.E. of three independent experiments. **, p Ͻ 0.01; ***, p Ͻ 0.001; n.s., no significance (two-way ANOVA followed Tukey). E, RGS2 overexpression increases the effect of LPA on P-REX1 activation. MCF7 cells were transfected with EGFP or HA-RGS2 and selected for 4 days with G418 (500 g/ml); then the cells were starved overnight and stimulated with 5 M LPA for 5 and 15 min, and the fraction of active P-REX1 was isolated by pulldown and detected by Western blotting. The graph represents the means Ϯ S.E. of three independent experiments. **, p Ͻ 0.01, two-way ANOVA followed Tukey. F, the combined effect of RGS2 and LPA on P-REX1 activation was mediated by G i . MCF7 cells expressing HA-RGS2 were starved, incubated with PTX (100 ng/ml overnight), and then were stimulated with LPA (5 M) for 5 min. Activation of P-REX1 was detected by pulldown. Western blots of the indicated phospho-and total proteins in total cell lysates, shown in the bottom panels, were assayed in the same experiments. The graph represents the means Ϯ S.E. of three independent experiments. **, p Ͻ 0.01; ***, p Ͻ 0.001, two-way ANOVA followed Tukey.

G␣ q -QL and G␣ 13 -QL prevent G␤␥-dependent P-REX1 activation
shown). To further confirm that endogenous P-REX1 is preferentially activated by G i , we used MCF7 cells transfected with G i -DREADD or G q -DREADD. As shown in Fig. 2L, endogenous P-REX1 was indeed preferentially activated by G i -DREADD but not by G q -DREADD, both effective activators of AKT; the latter was the more effective activator of ERK in this cellular system. Consistent with a previous report (12), chemogenetic activation of G i , but not G q , promoted cell migration (Fig. 2M).

GTPase-deficient G␣ q and G␣ 13 Q 3 L mutants form stable complexes with G␤␥: Critical role of G␣ N-terminal helix
The mechanism by which not all G proteins are good providers of signaling-ready G␤␥ might in part be linked to different dynamics of heterotrimer dissociation. Thus, we hypothesized that GTPase-deficient G␣ subunits, particularly Rho-activating G␣ q and G␣ 13 , might directly prevent G␤␥ signaling to Rac by interfering its interaction with P-REX1, a RacGEF known to be directly activated by G␤␥ (18,39), and G i -dependent cell migration (Fig. 3A).
To test the possible interaction between G␤␥ and different GTPase-deficient G␣ subunits, HEK293T cells were transfected with GST-G␤ 1 ␥ 2 and HA-tagged versions of representative WT or Q 3 L G␣ subunits (G␣ i , G␣ s , G␣ q , and G␣ 13 ) and subjected to pulldown assays (Fig. 3B, TCL and PD:GST, respectively). We found that GTPase-deficient G␣ q and G␣ 13 Q 3 L mutants, but not G␣ i or G␣ s (G␣ i QL and G␣ s QL), interacted with G␤␥ (Fig. 3B, upper panel). As expected, all WT G␣ subunits did maintain stable interactions with G␤␥ (Fig. 3B, upper panel). To confirm that GTPase-deficient G␣ q and G␣ 13 did interact with G␤␥ in live cells and not after cell lysis, we assessed whether co-expression of G␤␥ and either WT or GTPase-deficient G␣ q or G␣ 13 was required for their interaction to occur. Our results showed that G␣ q , as well as G␣ 13 , either WT or QL, had to be co-expressed with G␤␥ to interact ( Fig. 3, C and D, respectively). No complexes were detected when lysates from cells independently expressing these proteins were mixed (Fig. 3, C and D, upper panels, Western blotting anti-HA in the GST pulldowns). The expression of transfected proteins was confirmed by Western blotting in total cell lysates.
Structural analysis has revealed that G␣ subunits have two independent interfaces of interaction with G␤␥: one at the N-terminal ␣-helix (␣N) and other at switch I and switch II regions (40,41). The interaction interface at the switch regions suffers important conformational changes linked to G␣ activa-tion as a consequence of GTP binding that have been considered part of the dissociation process (40,42). To map the region of G␣ 13 QL that maintains interactions with G␤␥, we then used chimeric G␣ 13-i2 QL and G␣ i2-13 QL proteins. The signaling properties of these G protein chimeras have been previously described (43). These chimeras were tested by GST-pulldown analysis for their potential interactions with co-expressed GST-G␤␥. Using antibodies that detected either the N-or the C-regions of G␣ 13 to identify which chimera formed heterotrimers with GST-G␤␥, we revealed that G␣ 13 QL, as well as the chimera containing the N-terminal side of G␣ 13 (G␣ 13-i2 QL), including the interaction interface at the ␣N helix, maintained stable interactions with G␤␥ ( Fig. 3E, PD, upper panel). Diagrammatic representation of a heterotrimeric G protein, based on the structure of G i (41), shown in Fig. 3F highlights the two interaction interfaces between G␣ and G␤␥. The G protein constructs have swapped their helical domains (except HF, the last helix in the helical domain) and have chimeric GTPase domains resulting from joining in the tridimensional structure the N regions (HN, S1, H1, and their joining loops) of one G␣ subunit with the C regions (S2, S3, H2, S4, H3, S5, HG, H4, S6, H5, and their joining loops) of other G␣, which together compose the GTPase domain (44). In the case of the G␣ 13-i2QL chimera, the peptide regions contributing to the GTPase domain are G␣ 13 M1-D77 and G␣ i2 T178 -F355, whereas in the case of G␣ i2-13 QL chimera, they are G␣ i2 M1-S62 and G␣ 13 A199 -Q377. Therefore, although these chimeras are QL mutants, their nucleotide binding and release properties likely differ from those of G␣ i2 or G␣ 13 . The primary structure of these constructs is represented in Fig. 3F (bottom panel) (G␣ 13(M1-S192)-i2(Q172-P366) QL and G␣ i2(M1-T171)-13(Q193-Q377) QL) (43). According to the previous results (Fig. 3E), the N-terminal ␣-helix (␣N) contains the interacting interface essential to keep G␣ 13 QL and G␣ 13-i2 QL bound to G␤␥. These interactions are schematically represented in Fig. 3F. The G␣ i2-13 QL chimera, with a GTPase domain containing the second G␤␥-interacting interface, was unable to maintain stable interactions with G␤␥. Both G␣ 13 QL and G␣ i2-13 QL promoted strong cytoskeletal reorganization in endothelial cells (Fig. 3G, second and fourth panels, respectively; EGFP cotransfected cells pointed with arrowheads), which is indicative of their ability to activate Rho GTPase-dependent pathways likely through RGS-RhoGEFs (45)(46)(47). In contrast, the G␣ 13-i2 QL chimera Figure 2. Chemogenetic evidence that G i but not G q efficiently activates P-REX1. G i -DREADD (A), but not G q -DREADD (B), is expected to release G␤␥. Activation of AKT (pAKT, C and E) and ERK (pERK, D and F), by time course (C and D) and concentration response to CNO (E and F, 5 min) was assessed in G i -DREADD-expressing COS7 cells. The graphs represent the means Ϯ S.E. of four (C and D) or three (E and F) experiments. C and D, one-way ANOVA followed Dunnett's. E and F, two-way ANOVA followed Tukey (control versus G i -DREADD). *, p Ͻ 0.001. G and H, specificity of G i -and G q -DREADDs was confirmed with PTX (100 ng/ml, G) or FR900359 (500 nM, iGq, H), specific inhibitors of G i and G q . The cells were stimulated with CNO (5 min). I, time course of P-REX1 activation in G i -DREADD-expressing COS7 cells stimulated with CNO (1 M). Active FLAG-P-REX1 was isolated with RacG15A (shown stained with Ponceau). The graph represents the means Ϯ S.E. of three independent experiments. *, p Ͻ 0.015 (one-way ANOVA followed Dunnett's; all versus basal). J and L, G i but not G q activates P-REX1 in COS7 cells (J, FLAG-P-REX1) and MCF7 cells (L, endogenous P-REX1) expressing G i -or G q -DREADD stimulated with CNO (1 M, 15 min). K, comparative effect of G i and G q signaling on AKT and ERK activation by G i -DREADD or G q -DREADD in COS7 cells stimulated with CNO (1 M, 15 min). The graph (J and K) represents the means Ϯ S.E. of five independent experiments. J, t test and Mann-Whitney. n.s., no significance; *, p Ͻ 0.05; **, p ϭ 0.008. K, one-way ANOVA followed Tukey. n.s., no significance; *, p Ͻ 0.01; **, p Ͻ 0.001. C-H, K, and L, phospho-and total AKT and ERK were detected by Western blotting in TCLs. M, G i -DREADD induces cell migration but not G q -DREADD. Wound-closure experiments were done with HeLa cells that expressed empty vector (control) or G ior G q -DREADD and stimulated with CNO (1 M). The graph represents the means Ϯ S.E. of three independent experiments. *, p Ͻ 0.001; n.s., no significance (two-way ANOVA followed Tukey).

G␤␥ interacts with GTPase-deficient G␣ q QL or chimeric G␣ 13-i2 QL bound to RGS proteins
Because RGS proteins are known to interact with active G␣ subunits accelerating their GTPase activity, thus contributing to turn them into their inactive conformation, we wanted to discern whether GTPase-deficient G␣ q QL, G␣ 13 QL and the G␣ 13-i2 QL chimera, shown in the previous experiments to maintain stable interactions with G␤␥, expose their GTPase domain to interact with RGS proteins, forming tetrameric complexes including G␤␥. As shown in Fig. 4, GST-G␤␥ pulldown experiments similar to those described in the previous section, but here also including RGS2, RGS4, or RGL domains of RH-RhoGEFs, revealed that tetrameric complexes including G␤␥-G␣ q QL-RGS2 (Fig. 4A) and G␤␥-G␣ 13-i2 QL-RGS4 (Fig. 4B) were formed. WT G␣ subunits not expected to interact with RGS proteins or conditions where the chimeric constructs were excluded served as controls (Fig. 4, A-C). Consistent with previous experiments, GST-G␤␥ pulldowns contained G␣ q WT and G␣ q QL subunits; remarkably, RGS2 interacted with G␣ q QL bound to G␤␥, but not with G␣ q WT (Fig. 4A). In this situation, G␤␥ remains bound and putatively inactive (Fig. 4A, bottom panel, model illustrating nondissociable, G q ). Regarding the tetrameric G␤␥-G␣ 13-i2 QL-RGS4 complex (Fig. 4B), the model at the bottom highlights the role of ␣N, the N-terminal ␣-helix of G␣ 13 , to maintain stable interactions between G␤␥ and this chimera. In the case of G 13 , we detected WT and QL versions of G␣ 13 bound to G␤␥, but none of the RGL domains of its RhoGEF effectors (PDZ-RhoGEF, p115-RhoGEF, or LARG) were detected as part of the complex (Fig. 4C). In fact, the interaction of G␣ 13 QL with G␤␥ was reduced in cells transfected with a low concentration of PDZ-RhoGEF's RGL domain. This seems to be dynamically linked to a tight control of the G-protein cycle, as indicated by the effect of higher concentrations of this domain, which increased the G␤␥-G␣ 13 QL complex (Fig. 4D, upper panel, PD). This result raises the possibility of an effector-dependent mechanism that would contribute to G 13 dissociation (Fig. 4E, effector-dependent dissociation of G 13 ). After GST pulldown, the interaction of the RGL domain and G␣ 13 QL was confirmed by co-immunoprecipitation (Fig. 4D, middle panel, IP). Expression of transfected proteins was confirmed in lysates (total cell lysates (TCLs)). Together, these results are consistent with the idea that by maintaining stable interactions with G␤␥, G␣ q QL and G␣ 13 QL decrease the availability of this signaling heterodimer to interact with P-REX1.

GTPase-deficient G␣ q and G␣ 13 Q 3 L mutants inhibit G␤␥-dependent activation of P-REX1 and AKT and CXCR4dependent cell migration
Based on the demonstrated stable interactions between G␤␥ and G␣ q QL and G␣ 13 QL mutants, we postulated that these GTPase-deficient G␣ q and G␣ 13 mutants could negatively modulate G␤␥-signaling to Rac by preventing G␤␥ interaction with P-REX1. We first confirmed in COS7 cells that G␤␥ overexpression activates P-REX1 and Rac (Fig. 5A, first and second panels, respectively; expression of transfected proteins was demonstrated in TCL). Active P-REX1 was detected by pulldown assays using nucleotide-free GST-RacG15A bound to GSH-Sepharose (34,48). Then we evaluated the effect of G␣ q QL or G␣ 13 QL on the activation of P-REX1 and AKT by G␤ 1 ␥ 2 . As shown in Fig. 5B (upper panel and graph), P-REX1 activation elicited by G␤ 1 ␥ 2 overexpression in COS7 cells was lost when G␣ q QL or G␣ 13 QL were co-expressed. Interestingly, G␤␥ remained associated with the active fraction of P-REX1 only in the case that G␣ q QL or G␣ 13 QL were absent (Fig. 5B, Western blotting anti-G␤ in PD). Consistent with previous reports (49 -51), G␤␥ overexpression activated AKT (Fig. 5B, lower panel and graph). This effect was also attenuated by both G␣ q QL and G␣ 13 QL (Fig. 5B). Overexpression of G␤␥ was confirmed in total cell lysates, as well as expression of transfected HA-G␣ subunits and endogenous total and phosphorylated AKT (Fig. 5B, TCL, lower panel). We then tested whether this effect was due to a direct interference on the interaction between G␤␥ and P-REX1. We assessed this possibility in HEK293T cells transfected with GST-G␤ 1 ␥ 2 , FLAG-P-REX1, and the Q 3 L versions of G␣ q and G␣ 13 subunits. Then GST-G␤ 1 ␥ 2 pulldown assays were used to detect, using anti-FLAG antibodies, P-REX1 bound to G␤ 1 ␥ 2 . As shown in Fig. 5C, GST-G␤ 1 ␥ 2 interacted with FLAG-P-REX1, and this interaction was significantly lower in cells co-expressing G␣ q QL or G␣ 13 QL.
To test the effects of G␣ q QL and G␣ 13 QL on SDF-1-dependent migratory response, we first confirmed the participation of P-REX1 in CXCR4 signaling (32,52). Initially, we directly assessed the activation of endogenous P-REX1 in MCF7 cells stimulated at different times with SDF-1 (Fig. 5D). Active P-REX1 was captured by pulldown using RacG15A, and endogenous G␤␥ and AKT, but not ERK, were detected in the pulldowns from SDF-1-stimulated cells. Expression of endogenous proteins and phosphorylation of ERK in response to SDF-1 was confirmed in total cell lysates (Fig. 5D, TCL). Then we con- Figure 3. G␣ q -QL and G␣ 13 -QL but not G␣ i -QL or G␣ s -QL form stable complexes with G␤␥. A, chemotactic G i -coupled receptors such as CXCR4 activate P-REX1 and the PI3K/AKT signaling pathway via G␤␥. The drawing shows a hypothetical inhibitory effect of GTPase-deficient G␣ subunits on G␤␥ signaling to P-REX1 and AKT. B, characterization of G␤␥ interaction with WT and GTPase-deficient G␣ subunits in transfected HEK293T analyzed by pulldown of GST-G␤␥ (second panel) and detection of HA-tagged G␣ subunits associated to it (upper panel). C, pulldown assays of GST-G␤␥ show that GTPase-deficient G␣ q (G␣ q QL, upper panel). D, G␣ 13 (G␣ 13 QL, upper panel), as well as WT versions of these GTPases (G␣ q WT and G␣ 13 WT), do interact with G␤␥ only when they are co-expressed (Co-exp) but not when lysates of cells independently transfected with these proteins were mixed (Mixed). E, the ␣N-helix is critical to maintain GTPase-deficient G␣ 13 QL and G␣ 13-i2 QL bound to G␤␥. Pulldown assays of GST-G␤␥ were followed by Western blotting to detect G␣ 13 QL and chimeric G␣ 13-i2 QL or G␣ i2-13 QL, using antibodies recognizing epitopes at the N-or C-regions of G␣ 13 QL. Expression of transfected proteins is confirmed in TCLs (bottom panels). F, schematic representation of G protein complexes highlighting the importance of the ␣N-helix of G␣ 13 to maintain the interaction of this GTPase and chimeric G␣ 13-i2 QL with G␤␥. G, effects of G␣ 13 QL, G␣ 13-i2 QL, or G␣ i2-13 QL on the actin cytoskeleton of endothelial porcine aortic endothelial cells. More than 90% of G␣ 13 QL and G␣ i2-13 QL transfected cells exhibited a contracted phenotype, whereas more than 95% EGFP and G␣ 13-i2 QL-transfected cells showed an extended phenotype similar to untransfected cells. Transfected cells were fixed and stained with phalloidin. Arrowheads point to transfected cells in which EGFP (inset) was used as a marker. Scale bars, 25 m.

G␣ q -QL and G␣ 13 -QL prevent G␤␥-dependent P-REX1 activation
firmed the critical role of P-REX1 in Rac activation in the SDF-1-CXCR4 pathway in P-REX1 knockdown MCF7 cells. As shown in Fig. 5E, P-REX1 expression was required for SDF-1dependent activation of Rac (upper panel and graph) and AKT, but not ERK (lower panel, pAKT and pERK, respectively). In addition, we proved that HGF, a growth factor receptor agonist, activates Rac1 via P-REX1 (Fig. 5E).
We next tested the effect of SDF-1 in wound-healing assays. We first compared the effect of SDF-1 in wound-healing assays in MCF7 and HeLa cells and selected the later because we did not detect a significant migratory response in MCF7 cells. Therefore, in HeLa cells, we initially confirmed that SDF-1 promotes cell migration via CXCR4 receptors in a G i -dependent manner, as evidenced by the inhibitory effect of AMD3100 (10 g/ml), a CXCR4 antagonist, and PTX (200 ng/ml), known to inhibit G i (Fig. 5F). The G i -dependent migratory response of HeLa cells to SDF-1 was indeed inhibited by G␣ q and G␣ 13 Q 3 L mutants (Fig. 5G). Cell confluence before scratch was comparable within migration assays. Furthermore, the inhibitory effect of G␣ q and G␣ 13 Q 3 L mutants or PTX was not observed when cells were stimulated with 10% serum (Fig. 5H); therefore, the anti-migratory effect of G␣ q and G␣ 13 Q 3 L mutants was specific on GPCR/G i -dependent migration.

Discussion
G␤␥ signaling to Rac and chemotactic cell migration is mainly associated to G i -linked GPCRs (9,11,12), which raises the question of why other heterotrimeric G proteins, also containing G␤␥, are less efficient to promote cytoskeletal reorganization, cell protrusion, and migration via P-REX1, a RacGEF directly activated by G␤␥ (18). Various mechanisms have been proposed to explain why G␤␥ signaling is more effective when it comes from G i . These include the higher availability of G␤␥ caused by abundant expression of G i (30,31), existence of G i heterotrimers as preassembled complexes with receptors and effectors (53), more effective dissociation of G i compared with other heterotrimers (54), the influence of the subcellular localization (55,56), and the nature of the G␥ subunits (57), among other possibilities. Here, we tested the possibility that G␤␥ signaling to P-REX1, AKT, and cell migration might be directly regulated by Rho-activating G␣ q and G␣ 13 subunits, which would establish a signal coordination mechanism to prevent G␤␥-dependent activation of Rac, known to oppose Rho-dependent cytoskeletal effects (27)(28)(29). We speculated that GTPase-deficient G␣ q and G␣ 13 subunits might maintain stable interactions with G␤␥, directly inhibiting it. Consistent with this possibility, we found that GTPase-deficient G␣ q and G␣ 13 , but not G␣ i or G␣ s mutants, form stable complexes with G␤␥ preventing the activation of P-REX1, AKT, and cell migration by this signaling heterodimer. We propose the existence of direct regulatory interactions by which G␣ q and G␣ 13 subunits interfere with the G␤␥-P-REX1-Rac1 signaling axis (Fig. 6). This model adds family-specific differences among G protein heterotrimers in terms of direct regulatory mechanisms preventing the activation G␤␥-dependent pathways linked to actin cytoskeleton remodeling during cell extension, hence contributing to explaining the general observation that G i -linked GPCRs are more effective in promoting chemotactic migration (12,29,58), which has been mainly attributed to the higher abundance of G i heterotrimers (30,31). Although we speculate that the effects of overexpressed G␣ q and G␣ 13 mutants on G␤␥ signaling reflect maintained association of active (GTPbound) heterotrimers (as inferred by the direction of nucleotide exchange indicated in Fig. 6), we cannot rule out the possibility that some of these effects reflect sequestration of G␤␥ subunits by inactive (GDP-bound) QL heterotrimers.
To address the effect of different G proteins on G␤␥ signaling to P-REX1, we studied the regulation of this RacGEF by G i /G qcoupled LPA receptors in MCF7 cells, in which G i -dependent activation of P-REX1 has been previously demonstrated (32). We found that LPA receptors fine-tune P-REX1 activity via contrasting G␣ q -and G i -dependent effects. Through G i , LPA slightly activated P-REX1, whereas pharmacological inhibition of G␣ q made P-REX1 activation by LPA more efficient. Pertussis toxin not only prevented P-REX1 activation by G i but also revealed the inhibitory action of LPA reducing active P-REX1 below its basal levels. Putatively this effect was mediated by G␣ q , whose chronic inhibition by RGS2 overexpression had the opposite effect, increasing the basal levels of active P-REX1.
The N-terminal ␣-helix of different G␣ subunits interacts with G␤␥ (40,41,59,60). Very interestingly, driver oncogenic mutations in G␤ have been found at the region known to interact with this G␣ N-terminal ␣-helix, rendering G␤␥ constitutively active because of a lack of interaction with G␣ subunits from all families (61). Considering the recently revealed dynamics of G␣ helical domain, it seems likely that the ␣N helix, relevant to keep GTPase-deficient G␣ subunits bound to G␤␥, serves as a lever that sets the axis for the movement of the G␣ helical domain. Molecular dynamics simulations revealed that G␣ domains (helical and Ras homology) are constantly adjusting their distance from each other, reaching an almost

. G␤␥ forms tetrameric complexes composed by G␤␥-G␣ q QL-RGS2 and G␤␥-G␣ 13-i2 QL-RGS4, whereas G␣ 13 QL modulates its binding to G␤␥ by interacting with the RGS domain of its RhoGEF effectors.
A, G␣ q QL simultaneously interacts with RGS2 and G␤␥. HEK293T cells transfected with GST-G␤␥, HA-RGS2, and either HA-G␣ q QL or WT HA-G␣ q (WT) or empty vector (Ϫ) were subjected to GST-G␤␥ pulldowns followed by Western blotting that revealed a tetrameric complex composed by G␤␥-G␣ q QL-RGS2. Accordingly, the model at the bottom postulates that G q adjusts its conformation to interact with RGS2 without releasing G␤␥. B, HEK293T cells were transfected with GST-G␤␥ and HA-RGS4 and either G␣ i2 QL, chimeric G␣ 13-i2 QL, G␣ i2-13 QL, or empty vector (Ϫ) and subjected to GST pulldown analysis that revealed the existence of a tetrameric G␤␥-G␣ 13-i2 QL-RGS4 complex illustrated at the bottom panel. C, G␣ 13 QL does not simultaneously interact with G␤␥ and the RGS-like (RGL) domains of PDZ-RhoGEF (RGL), LARG or p115-RhoGEF. GST-G␤␥ revealed its interaction with G␣ 13 WT and G␣ 13 QL but not the RGS domain of RhoGEFs known as G␣ 13 effectors. D, G␣ 13 QL interacts with its RGL effector domain (from PDZ-RhoGEF) controlling its binding to G␤␥. GST-G␤␥ pulldown from HEK293T cells transfected with increasing amounts of RGL were used to detect the effect of this G␣ 13 QL effector on the association/dissociation of G␣ 13 QL to G␤␥. The remaining lysates were used to confirm by immunoprecipitation the interaction between G␣ 13 QL and RGL (IP: AU1). G␣ 13 QL was revealed as part of independent complexes with G␤␥ and RGL in PD: GST and IP: AU1, respectively. In contrast to the finding of a stable complex including G␤␥-G␣ q QL and RGS2, no equivalent complex was detectable with G␤␥-G␣ 13 QL and the RGL (RGS-like) domains of RGS-RhoGEFs. E, model proposing an effector-dependent dissociation of G 13 . It also considers a reciprocal regulation by the effector, consistent with the increased interaction between G␤␥ and G␣ 13 QL detected in the presence of increasing amounts of RGL.

G␣ q -QL and G␣ 13 -QL prevent G␤␥-dependent P-REX1 activation
open conformation nearly ready to release GDP. GPCRs stabilize this open conformation (62). How does the communication between the helical and Ras-homology domains result in G␤␥ release after nucleotide exchange is currently unknown. An interesting possibility is that in cases where G␤␥ signaling is restricted, effector proteins recognizing GTP-bound G␣ sub-G␣ q -QL and G␣ 13 -QL prevent G␤␥-dependent P-REX1 activation units might facilitate their full dissociation from G␤␥. Although G␤␥ would be expected to be excluded from complexes formed by active G␣ subunits and RGS proteins (42), our results demonstrating the existence of tetrameric G␤␥-G␣ q QL-RGS2 and G␤␥-G␣ 13-i2 QL-RGS4 complexes supports an important role for the ␣N helix of G␣ q and G␣ 13 regulating its rate of dissociation from G␤␥, because a conformational change is implied, based on available structures (63,64), to explain the existence of these tetramers. Then to release G␤␥, this interacting interface has to be disrupted. In this sense, it has been observed that RGS3 and RGS4 increase available G␤␥ to regulate inwardly rectifying K ϩ channels (65). Furthermore, our findings are consistent with the dynamics of G 13 recently reported by Bünemann and co-workers (66), who demonstrated by FRET that upon receptor activation, G␣ 13 is not fully dissociated from G␤␥. Interestingly, interaction of G␣ 13 with LARG resulted in a decreased FRET signal between G␣ 13 and G␤␥ (66). These results and our biochemical data and profile of P-REX1 activa-tion by G␤␥ in the presence of GTPase-deficient G␣ q and G␣ 13 might be interpreted in terms of the differential ability of these G␣ subunits to directly control G␤␥ heterodimers. Thus, in the case of G␣ q and G␣ 13 , we speculate that receptor-promoted nucleotide exchange leads to interaction of effectors with GTPbound G␣, while keeping G␤␥ dimers inactive or less available to interact with their downstream effectors.
The fact that CXCR4 receptors normally transduce chemotactic signals via G␤␥ released from G i , even though they might also be coupled to G q and G 12/13 , is consistent with our observations that GTPase-deficient G␣ q and G␣ 13 directly interact with G␤␥ and inhibit its interaction with P-REX1, which plays a central role in SDF-1␣-CXCR4 chemotactic effect (32,52). Observations that G q -coupled receptors are not chemotactic, whereas G 13 -coupled receptors have a chemotactic role linked to their expression levels, as in the case of CXCR4 overexpressed in various metastatic cancers (21), further indicate that G␤␥ signaling might be regulated by G␣ q and G␣ 13 depending Figure 5. GTPase-deficient G␣ q and G␣ 13 (Q 3 L mutants) inhibit G␤␥-dependent activation of P-REX1, AKT, and G i -dependent cell migration. A, G␤␥ activates P-REX1 and Rac1. Active P-REX1 and Rac were detected by pulldown in COS7 cells expressing FLAG-P-REX1 and His 6 -G␤␥. B, GTPase-deficient G␣ q and G␣ 13 inhibit G␤␥-dependent P-REX1 activation (upper panel) and AKT (lower panel). The graphs represent the means Ϯ S.E. densitometric value of three to four independent experiments. *, p Ͻ 0.05, t test. Lower graph, *, p ϭ 0.029; **, p ϭ 0.005, Mann-Whitney tests. C, GTPase-deficient G␣ q and G␣ 13 inhibit the interaction between G␤␥ and P-REX1. The graph represents the means Ϯ S.E. densitometric value of P-REX1 interacting with G␤␥ obtained in four to five independent experiments. *, p ϭ 0.006; **, p Ͻ 0.001 (t tests). D, representative time course of P-REX1 activation by SDF-1/CXCL12 in MCF7 cells. Active endogenous P-REX1 was detected by pulldown using RacG15A. Endogenous G␤␥ and AKT interacting with active P-REX1 were revealed in the respective pulldown in which ERK was not detectable. Four independent experiments showed interacting G␤␥ and AKT with the active fraction of P-REX1. E, SDF-1-CXCL12 and HGF activate Rac via P-REX1. MCF7 cells were transfected with esiRNA-P-REX1 (consisting on a mixture of siRNAs) or esiRNA-GFP (as control) and activation of endogenous Rac, AKT, and ERK was detected by PAK-CRIB pulldown (for active Rac) and Western blotting against phosphorylated AKT and ERK in cells stimulated with SDF-1-CXCL12 (50 ng/ml) or HGF (10 ng/ml) for 5 min. The graph represents the mean densitometric values Ϯ S.E. of three independent experiments showing the fraction of active Rac. *, p ϭ 0.003; **, p Ͻ 0.001; n.s., no significance (two-way ANOVA followed Tukey). A-E, expression of transfected and endogenous proteins was detected in TCLs. F, SDF-1/CXCL12 promotes G i -dependent cell migration via CXCR4 receptors. Wound-closure experiments were done with HeLa cells stimulated with SDF-1/CXCL12 (100 ng/ml) alone or with AMD3100 (10 g/ml, a CXCR4 antagonist), or PTX (200 ng/ml, G i inhibitor), added 1 h before stimulation, or FBS (10%). G, GTPase-deficient G␣ q and G␣ 13 Q 3 L mutants inhibit cell migration induced by SDF-1/CXCL12 (100 ng/ml) (G), but not by fetal bovine serum (10%) (H). The graphs at the left represent the means Ϯ S.E. of four independent experiments. F, *, p ϭ 0.002; **, p Ͻ 0.001. G, n.s., no significance; *, p ϭ 0.013; **, p Ͻ 0.01. H, *, p Ͻ 0.001 one-way ANOVA followed Tukey. Figure 6. Cell migration elicited by chemotactic GPCRs via the G␤␥-P-REX1-Rac pathway is mainly mediated by G i over G q or G 13 because of the tight control that G␣ q and G␣ 13 subunits exert on G␤␥. Based on our results, the model postulates that at the front of a migrating cell (left panel) G i -coupled receptors release G␤␥ to activate P-REX1/Rac and actin polymerization to move the cell forward, whereas Rho activating G␣ q and G␣ 13 directly prevent G␤␥-dependent activation of Rac via P-REX1, contributing to explaining the general observation that chemotactic signaling transduced by GPCRs occurs via G i . Because all heterotrimeric G proteins contain G␤␥, the lack of Rac activation by G q and G13 is, according to our model, at least in part because of direct inhibition of G␤␥ signaling by its G␣ q and G␣ 13 counterparts. This model contributes to explaining why G␤␥ signaling is preferentially linked to G i -coupled receptors, which so far has been mainly attributed to higher abundance of G i heterotrimers.

G␣ q -QL and G␣ 13 -QL prevent G␤␥-dependent P-REX1 activation
on their levels of activity. This is also supported by our experiments showing that pharmacological inhibition of G␣ q enhances P-REX1 signaling upon activation of G q /G i -coupled LPA receptors and by the assays using the DREADD systems. In the latter, we demonstrated that under comparable settings, G i -coupled but not G q -coupled receptors activate P-REX1. Furthermore, we found that AKT was activated by G i as part of an interplay with the active fraction of P-REX1. Indeed, AKT stimulation was reduced in P-REX1 knockdown cells, suggesting that P-REX1 serves as a signaling platform for AKT activation, likely coordinated by distinct domains within this RacGEF. These findings are consistent with previous observations in which the activation of AKT1 by insulin-like growth factor requires P-REX1 (67), which responds to mTORC2 signaling (67,68).
GTPase-deficient G␣ subunits with inactivating mutations that impair GTP hydrolysis have been widely used as constitutively active mutants to reveal fundamental signaling mechanisms by different families of G proteins (69 -75). Very recently, this approach successfully revealed a mechanism elicited by G␣ i to inhibit Rap-dependent cell adhesion promoting cell migration (75). In addition, recent deep sequencing analysis of cancer genomes revealed that some GTPase-deficient G␣ subunits are driver oncogenes (76 -78). Remarkably, oncogenic G␤ mutants constitutively activate the PI3K/mTOR signaling pathway (61), which is consistent with the recent demonstrated ability of G␤␥ to stimulate AKT and interact with mTOR directly promoting its activation (79).
In conclusion, our results reveal that G␣ q QL and G␣ 13 QL mutants inhibit G␤␥-dependent P-REX1 and AKT activation and cell migration induced by the SDF-1-CXCR4 -G i axis. In the case of G␣ 13 , the N-terminal ␣-helix of this G␣ subunit is critical to keep a stable complex with G␤␥, thus contributing to our understanding of the prominent role of the G␤␥-interacting site located at this region of G␣ to control G␤␥ signaling. Furthermore, the finding that G␣ q QL simultaneously interacts with RGS2 and G␤␥ indicates that this G␣ subunit is able to recognize its signaling partners while preventing G␤␥ from doing the same.
Lysates in 1.5-ml tubes were incubated with 30 l of GSH-Sepharose beads for 30 min in an ice bath with constant shaking. Then beads were washed three times with lysis buffer, boiled for 5 min in 1ϫ Laemmli buffer with ␤-mercaptoethanol, and centrifuged for 5 min at 13,000 rpm before being subjected to Western blotting analysis.
For immunoprecipitations, cell lysates obtained from HEK293T cells transfected with HA-G␣ 13 and AU1-RGL (RGS-like domain from PDZ-RhoGEF) were incubated with anti-AU1 (1:1000) in ice/rocking platform overnight at 4°C. Next day, 30 l of protein G-Sepharose (Millipore, catalog no.  were added, and incubation continued for 3 h. The beads were washed three times with lysis buffer and subjected to Western blotting analysis.

P-REX1 knockdown in MCF7 cells
Reverse transfection of MCF7 cells was performed with Lipofectamine RNAiMAX and 30 pmol of esiRNA-P-REX1 (Sigma-Aldrich, catalog no. EHU136571) or esiRNA-EGFP (Sigma-Aldrich, catalog no. EHUEGFP), as control; siRNA against P-REX1 is in fact a pool including a heterogeneous mixture of siRNAs (esiRNAs are endoribonuclease-prepared siRNAs). Lipofectamine and esiRNA mixes were prepared in OptiMEM, according to the manufacturer's instructions. The cells were trypsinized and suspended in DMEM containing 5% FBS without antibiotics, then Lipofectamine-esiRNA complexes were added, and cells were seeded on p60 dishes and incubated overnight at 37°C in a 5% CO 2 atmosphere. Next day, the cells were washed with PBS and cultured with DMEM containing 10% FBS and antibiotics. Experiments with knockdown cells were done 72 h after transfection. The cells were starved in serumfree medium for 18 h before being stimulated for 5 min with CXCL12-SDF-1␣ (50 ng/ml, PeproTech, catalog no. 300-28A) or HGF (10 ng/ml hepatocyte growth factor recombinant human; RD Systems, catalog no. 294-HGN), before lysis.

Chemogenetic strategy with G i -DREADD and G q -DREADD
COS7 and MCF7 cells were transfected with empty vector, G i -DREADD or G q -DREADD specifically coupled to G i or G q . Then cells were stimulated 48 h after with CNO, a synthetic ligand for DREADD (Tocris, catalog no. 4936) as indicated in Fig. 4. To confirm the signaling specificity of G i -and G q -DRE-ADD, serum-starved transfected cells were incubated overnight with pertussis toxin (100 ng/ml, G i inhibitor), or 2 h with the G q inhibitor (FR900359, 500 nM), before stimulation with CNO. In experiments in which PKC was inhibited, 1 M sotrastaurin was preincubated for 2 h before stimulation with 5 M LPA (Biomol, catalog no. LP-100).

Active Rac-GTP and RhoGEF pulldowns
The cells transfected with P-REX1, G␤ 1 ␥ 2 and G␣ q QL or G␣ 13 QL were starved for 14 h with serum-free DMEM before lysis. Then to capture Rac-GTP, cell lysates were incubated with PAK-PBD-Sepharose beads for 45 min on ice/shaker (68). Active P-REX1 was isolated from cell lysates incubated with RacG15A-Sepharose beads for 1.5 h. The beads were washed three times with lysis buffer and processed for Western blotting analysis (34,48). For DREADD experiments, when evaluated G i -DREADD versus G q -DREADD, the cells were stimulated for 15 min with 1 M CNO, before lysis to detect active P-REX1 by pulldown. To address the effect of RGS2 in MCF7 cells, HA-RGS2-transfected cells were selected with 500 g/ml G418 (Sigma, catalog no. A1720) for 4 days before analysis

Wound-closure assay
HeLa cells were seeded on 0.02% gelatin in p35 6-well plates and transfected using PolyFect (Qiagen, catalog no. 301105) with empty vector, G␣ subunits, G i -DREADD, or G q -DREADD as indicated in the respective figures. The cells were starved with serum-free DMEM for 4 h. After 2 h of starvation, mitomycin C was added (12 M, Sigma-Aldrich, catalog no. M0440). Pertussis toxin (200 ng/ml, Calbiochem, catalog no. 516560) or AMD3100 (10 g/ml, Sigma-Aldrich, catalog no. A5602) were used to assess signaling by G i and CXCR4, respectively. To initiate the migration assays, cell monolayers were wounded with a pipette tip, washed three times with PBS, and subjected to stimulation with CXCL12-SDF-1␣ (100 ng/ml, PeproTech, catalog no. 300-28A) or 10% FBS in 2 ml of DMEM containing the indicated inhibitors. Migration in G i -and G q -DREADD-transfected cells was assessed with 1 M CNO. After 24 h, the cells were fixed with 4% paraformaldehyde, stained with crystal violet, washed with PBS, and photographed.

Fluorescence microscopy
Porcine aortic endothelial cells seeded on gelatin-coated glass-bottomed dishes were transfected with G␣ 13 QL or the indicated chimeras using Polyfect. pCEFL-EGFP was included as a marker to identify transfected cells. 48 h post-transfection, the cells were starved for 8 h and fixed in 4% paraformaldehyde, washed with PBS, and stained with phalloidin and 4Ј,6Ј-diamino-2-phenylindole. The cells were photographed in a Nikon Eclipse Ti inverted fluorescence microscope.

Statistical analysis
The data are presented as means Ϯ S.E. of at least three to five independent experiments. Densitometric quantitation of Western blots and wound-closure assays was done with ImageJ software. Active proteins, phosphorylated proteins, interactions in pulldowns were normalized with total proteins and pulldown efficiency. Test and control samples in the functional assays were compared using one-way analysis of variance (ANOVA) followed Tukey tests, and the Western blots were compared with t test or Mann-Whitney, one-way and two-way ANOVA followed by Dunnett or Tukey, as indicated in figure legends. Statistical significance was considered for values of p Ͻ 0.05. Statistical analysis was performed using Sigma Plot 11.0 and graphs using GraphPad Prism software V6.0.

Data availability statement
Any reasonable request for materials, data, and associated protocols originally described in this work will be available to readers without undue qualifications in material transfer agreements.