Membrane Recruitment of the Non-receptor Protein GIV/Girdin (Gα-interacting, Vesicle-associated Protein/Girdin) Is Sufficient for Activating Heterotrimeric G Protein Signaling*

GIV (aka Girdin) is a guanine nucleotide exchange factor that activates heterotrimeric G protein signaling downstream of RTKs and integrins, thereby serving as a platform for signaling cascade cross-talk. GIV is recruited to the cytoplasmic tail of receptors upon stimulation, but the mechanism of activation of its G protein regulatory function is not well understood. Here we used assays in humanized yeast models and G protein activity biosensors in mammalian cells to investigate the role of GIV subcellular compartmentalization in regulating its ability to promote G protein signaling. We found that in unstimulated cells GIV does not co-fractionate with its substrate G protein Gαi3 on cell membranes and that constitutive membrane anchoring of GIV in yeast cells or rapid membrane translocation in mammalian cells via chemically induced dimerization leads to robust G protein activation. We show that membrane recruitment of the GIV “Gα binding and activating” motif alone is sufficient for G protein activation and that it does not require phosphomodification. Furthermore, we engineered a synthetic protein to show that recruitment of the GIV “Gα binding and activating” motif to membranes via association with active RTKs, instead of via chemically induced dimerization, is also sufficient for G protein activation. These results reveal that recruitment of GIV to membranes in close proximity to its substrate G protein is a major mechanism responsible for the activation of its G protein regulatory function.

Heterotrimeric G proteins regulate virtually all physiological functions in humans, and their dysregulation is the cause of many diseases (1,2). They cycle between inactive (GDP-bound) and active (GTP-bound) states to control the flow of information from extracellular cues to intracellular effectors (1,3).
Under resting conditions the nucleotide-binding ␣ subunit (G␣) is loaded with GDP and forms a complex with G␤␥. According to the classic paradigm, heterotrimeric G protein activation in response to extracellular stimuli is carried out via G protein-coupled receptors (GPCRs) 2 located at the plasma membrane. GPCRs are guanine nucleotide exchange factors (GEFs) that promote the exchange of GDP for GTP on G␣␤␥ (1,3). The consequence of this is the dissociation of the G␣␤␥ heterotrimer into G␣-GTP and free G␤␥, which are the two active molecular species that signal to downstream effectors.
Despite the well established mechanism of heterotrimeric G protein activation by GPCRs, there is mounting evidence indicating that heterotrimeric G proteins also mediate signaling via other types of receptors. This is documented best for receptor tyrosine kinases (RTKs), for which the role of heterotrimeric G protein signaling is supported by experiments in isolated cells, as well as transgenic and knock-out mice (4 -10). However, the molecular mechanism of G protein activation by RTKs to support this functional evidence has remained rather obscure and controversial because neither RTKs nor their best characterized adaptors have GEF activity toward heterotrimeric G proteins. A plausible explanation for this quandary is provided by the existence of non-receptor GEFs (11)(12)(13)(14)(15), which could be activated by RTKs to relay signals to downstream effectors via heterotrimeric G proteins. Recent work indicates that GIV (G␣interacting, vesicle-associated protein, also known as Girdin) fulfills such role (reviewed in Ref. 16). GIV contains a ϳ30 -35amino acid sequence, namely the G␣ binding and activating (GBA) motif, conserved across functionally and evolutionarily divergent non-receptor GEF proteins (17)(18)(19)(20)(21). Among G␣ subunits, GIV binds predominantly to G␣ i (equally to all three members of mammals: G␣ i1 , G␣ i2 , and G␣ i3 ), although it also interacts weakly with G␣ o and G␣ s (17,(22)(23)(24). The GIV GBA motif activates G␣ i proteins in vitro via acceleration of nucleo-* This work was supported by National Institutes of Health Grants tide exchange (17,23), and it also promotes G protein activation in cells, as determined by readouts for either G␣ i -GTP (e.g. conformation-specific antibodies, cAMP dampening) (25)(26)(27) or free G␤␥ (e.g. PI3K-Akt signaling, resonance energy transferbased biosensors) (17,26,28). Importantly, GIV-mediated G protein activation operates downstream of receptor types different from GPCRs, like RTKs and integrins (16, 25, 26, 29 -31) and is crucial for the role of GIV in cancer progression (32)(33)(34)(35)(36)(37)(38) by promoting tumor cell migration, actin remodeling, and activation of the oncogenic PI3K-Akt pathway (10,39,40). Additional evidence indicates that the GIV GBA motif is also important in a wide range of cellular processes (e.g. autophagy, protein trafficking, or epithelial junction integrity) (22,(41)(42)(43) and diseases (e.g. liver fibrosis, diabetes, or kidney failure) (27,30,44,45).
A major gap in the understanding of GIV is the lack of clear mechanistic insights into how its G protein regulatory function becomes activated. Because GIV is not a transmembrane protein and its role is restricted to ligands that do not cross the plasma membrane, its activation must be regulated via receptor-mediated activation. The molecular events linked to GIVmediated G protein activation are characterized best in the context of RTKs. GIV directly binds to the phosphorylated tail of RTKs (25,26), and this is required for different signaling functions of GIV, e.g. G protein-dependent (25,26) and tyrosine phosphorylation-dependent signaling (46). In addition, stimulation of EGFR, a prototypical RTK, has been recently shown to lead to the phosphorylation of GIV at serine 1674, a residue adjacent to its GBA motif, which enhances the GEF activity (47). However, this enhancement is unlikely to fully account for all GIV-mediated G protein activation because it is modest (ϳ1.5-fold enhancement of GEF activity in vitro) and dependent on a particular kinase (cyclin-dependent kinase 5), which might not be relevant for all RTKs or other receptor types, like integrins, that rely on GIV for signaling. Here we set out to elucidate the mechanism responsible for activating the G protein regulatory function of GIV in a broader context. We reasoned that relocalization of GIV to membranes would be sufficient to trigger G protein activation based on several observations: (i) GIV is recruited to the plasma membrane in response to different stimuli (10,16,29,34,40), (ii) the G protein substrate (G␣ i ) is constitutively attached to membranes (48), and (iii) GIV without modifications (i.e. bacterially expressed) possesses intrinsic GEF activity (17,22,23). By using synthetic biology approaches and G protein activity biosensors, here we provide evidence that the spatial relocalization of GIV to membranes is a mayor mechanism responsible for the activation of its G protein regulatory function.

Results
GIV Does Not Co-fractionate with G␣ i3 on Cell Membranes-GIV has been shown to localize in different subcellular compartments under different conditions (10,22,24,42,49) and, more specifically, to become enriched at the plasma membrane in response to different stimuli (10,16,29,34,40). Acute ligand stimulation also induces the co-localization of GIV with G␣ i3 at the plasma membrane (10,29). However, the overall distribution of GIV in non-stimulated cells and how it relates to that of G␣ i subunits has not been thoroughly investigated. For this, we performed biochemical fractionation of mammalian cells under standard culture conditions. We focused on G␣ i3 for this and for subsequent parts of this study because the vast majority of prior biochemical and cell biological work in relation to GIV has been carried out with G␣ i3 . However, all three G␣ i subunits of mammals (G␣ i1 , G␣ i2 , and G␣ i3 ) share identical membrane localization signals and bind equally to GIV (17,24), suggesting that observations with G␣ i3 in this context are very likely to apply to G␣ i subunits in general. In HEK293T, endogenous G␣ i3 is almost exclusively found in the P100 fraction (Fig. 1A), which is consistent with its well documented constitutive association , or MDA-MB-231 (C) cells were homogenized in the absence of detergents, and the PNS was separated into supernatant (S100) and pellet (P100) after 100,000 ϫ g centrifugation. P100 fractions were resuspended in buffer containing Triton X-100 and centrifuged to obtain the detergent soluble (TX-S) and insoluble (TX-I) fractions. Immunoblots (IB) of equal aliquots of each fraction from one experiment representative of three is shown. TfR and tubulin were used as membrane and cytosol markers, respectively. D, detergent lysates of MDA-MB-231 cells were centrifuged in a discontinuous sucrose gradient (5-35-45%) to separate low density detergent-resistant membranes ("lipid rafts") from other cell fractions. Equal gradient fractions were immunoblotted with the indicated antibodies. Caveolin-1 was used as lipid raft marker, whereas ␤1-integrin and tubulin were membrane and cytosolic markers, respectively, of non-lipid raft fractions. One experiment representative of three is shown.
with membranes (48). In contrast, endogenous GIV shows the opposite distribution, localizing predominantly (ϳ90 -95%) in the cytosolic S100 fraction (Fig. 1A). Because the P100 is a crude particulate fraction containing not only cell membranes, we separated it into detergent-soluble (TX-S) and detergent-insoluble (TX-I) fractions to further characterize the distribution of G␣ i3 and GIV. As expected for a membrane protein, G␣ i3 was efficiently extracted by detergents from the P100 into the TX-S fraction (Fig. 1A). In contrast, all GIV present in the P100 fraction remained in the TX-I fraction (Fig. 1A), which is consistent with a previously described pool of GIV that associates with the actin cytoskeleton (40). Equivalent fractionation experiments looking at endogenous GIV and G␣ i3 were carried out in two additional cell lines: HeLa (Fig. 1B) and MDA-MB-231 (Fig.  1C). The results were similar to those obtained in HEK293T cells in that a large pool of GIV was cytosolic (S100: ϳ75% in HeLa and ϳ50% in MDA-MB-231), and the remaining pool of GIV in the P100 was detergent-insoluble ( Fig. 1, B and C). G␣ i3 distribution was almost identical in the three cell lines. These results indicate that, regardless of the relative partition of GIV between the cytosolic and detergent-insoluble fractions in the three different cell lines, GIV is excluded from the detergentsoluble P100 fraction where G␣ i3 is predominantly found. Similar results were also obtained with the three cells lines using a buffer of different composition, 3 indicating that our observations are consistent across different experimental conditions.
A possible interpretation of the TX-I fraction of GIV is that it corresponds to a pool that localizes in detergent-resistant domains of cell membranes, the so-called lipid rafts (50). If GIV localizes in lipid rafts, it could potentially interact with G␣ i3 because it has been previously described that G␣ i proteins partially localize in these membrane microdomains (51,52). To investigate this possibility, we fractionated MDA-MB-231, the cell line with the largest proportion of GIV in the TX-I fraction (Fig. 1C), using a protocol that separates detergent-insoluble lipid rafts from other cell fractions by ultracentrifugation in a sucrose density gradient. Consistent with previous reports, some G␣ i3 was found in low density, detergent-insoluble fractions enriched in the lipid raft marker caveolin-1 (Fig. 1D). However, GIV was exclusively detected in the high density fractions along with other non-lipid raft markers (Fig. 1D), ruling out the possibility that the TX-I fraction of GIV corresponds to a pool localized in lipid rafts. This supports the alternative interpretation that GIV present in the TX-I fraction corresponds to a previously described pool of the protein that associates with the actin cytoskeleton (40). Taken together, these results indicate that in the absence of an acute stimulus GIV is not localized in the cell membranes where G␣ i3 resides.
Constitutive Membrane Targeting of GIV Enhances G i3 Signaling in Yeast-The results presented above show that in unstimulated cells GIV is spatially segregated from its substrate G protein G␣ i3 that is located in cell membranes. As a first approach to investigate the impact of targeting GIV to cell membranes on G protein activation, we used a yeast-based system with a modified mating pheromone pathway (53) (Fig. 2A). Briefly, the genetically engineered yeast strain used in this system lacks pheromone-responsive GPCRs, and its endogenous yeast G␣ protein Gpa1 is replaced by mammalian G␣ i3 . Thus, FIGURE 2. Membrane association of GIV enhances its ability to activate G protein signaling in yeast. A, schematic diagram of the yeast-based assays used to monitor G protein signaling. A genetically engineered strain lacking endogenous GPCRs and with human G␣ i3 replacing the endogenous G␣ Gpa1 was used to determine the levels of G protein activation upon expression of exogenous GIV. G protein activation was determined by immunoblotting (IB) for Fus3 phosphorylation or two gene reporters controlled by the Fus1 promoter (␤-galactosidase activity or histidine auxotrophy). B, yeast cells expressing unmodified GIV (aa 1660 -1870) or fused to a membrane-targeting sequence corresponding to the first 9 aa of Gpa1 (N9Gpa1-GIV) were homogenized in the absence of detergents, and the PNS was separated into supernatant (S) and pellet (P) fractions after 100,000 ϫ g centrifugation. Immunoblots of equal aliquots of each fraction from one experiment representative of four are shown. C, membrane-targeted N9Gpa1-GIV enhances G protein-dependent ␤-galactosidase activity ϳ25-fold more than cytosolic GIV. The results are the averages of n ϭ 4, and error bars are the S.E. D, membrane-targeted N9Gpa1-GIV induces more Fus3 phosphorylation (ppERK/ppFus3; top, immunoblot panels) and more rapid cell growth in media lacking histidine (bottom, spot growth panels) than cytosolic GIV. One experiment representative of four is shown.
only an exogenous G protein activator can trigger a signaling pathway that is normally activated as a mating pheromone response, which leads to an increase in Fus3 phosphorylation and to transcriptional activation of the Fus1 promoter. For these experiments we used a construct consisting of the C-terminal region of GIV (aa 1660 -1870). This fragment expresses well in yeast, and it has been previously shown (17,23) to bind and activate G␣ i3 as efficiently as larger fragments of GIV. Moreover, a thorough characterization of GIV (aa 1660 -1870) in two studies (26,28) has validated that it recapitulates the main biological functions of the full-length protein, including the regulation of receptor-mediated signaling via heterotrimeric G proteins. This domain of GIV lacks any known membrane targeting sequence but contains the GBA motif necessary for G protein activation. We designed a modified version of this construct by fusing it to the first 9 amino acids of Gpa1 (N9Gpa1), which are sufficient to target proteins to cell membranes in yeast (54). Unmodified GIV and N9Gpa1-GIV were expressed at similar levels, and as expected, GIV was found almost exclusively in the cytosolic soluble (S) fraction, whereas N9Gpa1-GIV was predominantly recovered in the particulate insoluble (P) fraction (Fig. 2B), validating that the N9Gpa1 sequence efficiently targets GIV to membranes. Expression of membrane-anchored N9Gpa1-GIV led to an increase in G protein activation much larger (ϳ25-fold) than that observed when expressing cytosolic GIV, as determined by the ␤-galactosidase activity of a LacZ reporter under the control of the Fus1 promoter (Fig. 2C). Membrane-anchored GIV was also more potent than cytosolic GIV in two other independent assays that reflect G protein signaling in this system, i.e. Fus3 phosphorylation and cell growth in the absence of histidine (Fig. 2D). None of these G protein-dependent responses induced by cytosolic or membrane-anchored GIV were reproduced by constructs bearing a mutation in the GBA motif (F1685A, FA) that disrupts GEF activity (17,22) (Fig. 2, C and D). The FA mutation does not affect the subcellular fractionation of GIV or N9Gpa1-GIV (Fig. 2B). These findings indicate that association of GIV with membranes is sufficient to enhance G protein activation via its GBA motif.
Acute Translocation of GIV to Membranes Induces Rapid G Protein Activation in Mammalian Cells-A limitation of the experiments described above is that they reflect the long term consequences of keeping GIV constitutively anchored to membranes, which does not directly relate to the rapid kinetics (seconds to minutes) of the signaling events regulated by GIV (10,16,26,29,31). To regulate the spatial localization of GIV and mimic its rapid recruitment to cell membranes in response to receptor stimulation (16,29), we used a chemically induced dimerization (CID) system in mammalian cells (Fig. 3A) (55,56). Briefly, this system consists of two domains (FRB and FKBP) that dimerize in the presence of rapamycin. FRB was fused to the first 11 aa of Lyn (Lyn11-FRB), which targets it to the plasma membrane (56), whereas FKBP was fused to C-terminal region of GIV (aa 1660 -1870, FKBP-GIV). We chose to work with GIV (aa 1660 -1870) because (i) it recapitulates the signaling functions of full-length GIV in mammalian cells, including receptor-mediated signaling via its GBA motif (26, 28), (ii) it has been validated that bulky tags (i.e. GFP) can be tagged to its N terminus without compromising function in mammalian cells (26), (iii) it binds and activates G proteins as efficiently as larger fragments (17,23), and (iv) it provides continuity with the previous yeast experiments. Because FKBP-GIV lacks any membrane targeting sequence, FRB/FKBP dimerization by rapamycin induces the translocation of FKBP-GIV to the location of Lyn11-FRB, i.e. the plasma membrane (Fig. 3A). To monitor the effect of this translocation on G protein activation, we used a bioluminescence resonance energy transfer (BRET)-based assay (57,58) that measures G protein activation more directly than the yeast-based assays above, which rely on measuring the activity of downstream effectors/ reporters. In this BRET system, dissociation of G protein G␣␤␥ heterotrimers upon activation leads to the association of free Venus-G␤␥ (BRET acceptor) with the C-terminal domain of its effector GRK3 fused to an enhanced luciferase (Nanoluciferase, BRET donor), which leads to an increase in BRET (Fig. 3A). Expression of FKBP-GIV WT did not cause an increase in BRET in the absence of rapamycin compared with control cells expressing the GEF-deficient FA mutant (Fig. 3B). However, addition of rapamycin induced an increase in BRET in cells expressing FKBP-GIV WT but not FKBP-GIV FA (Fig. 3B), indicating that translocation of GIV to cell membranes induces G protein activation via its GBA motif. The amplitude of G protein activation was dependent on the dose of FKBP-GIV ( Fig. 3, C-E) and the kinetics were comparable with those previously described (56) for rapamycin-induced FRB/FKBP membrane recruitment (Fig. 3C), suggesting that G protein activation occurs rapidly upon CID-mediated translocation of GIV to cell membranes.
The time scale and dependence on the GIV GBA motif suggest that the increase in G protein activation is a consequence of the action of rapamycin on Lyn11-FRB/GIV-FKBP dimerization rather than on other cellular targets (e.g. mTOR). To confirm this, we carried out control experiments in which Lyn11-FRB was mutated to prevent its association with membranes (59). More specifically, we created a mutant (Lyn11 G2A -FRB) disrupting the lipidation site of Lyn required for its association with membranes ( Fig. 3F). Although cells expressing Lyn11-FRB and FKBP-GIV led to a robust increase in BRET upon rapamycin stimulation, cells expressing the same amount of Lyn11 G2A -FRB mutant and FKBP-GIV failed to show any measurable response (Fig. 3, G and H). We conclude that translocation of GIV to membranes is sufficient to promote rapid G protein activation in mammalian cells and that it requires an intact GBA motif.
GIV Translocation and GPCR Stimulation Induce G Protein Responses of Similar Magnitude-Next, we benchmarked the G protein response triggered by GIV upon membrane translocation by comparing it with the response observed after GPCR stimulation. For this, we used the M4 muscarinic receptor (M4R), a prototypical G i -coupled receptor, stimulated with saturating concentrations of its synthetic ligand carbachol. We found that the maximal amplitude of the GIV-mediated BRET response within the first minutes of stimulation was of similar magnitude to that elicited by the M4R (Fig. 4, A and B). Similar results were obtained with two other GPCRs, i.e. adenosine A 1 receptor (A 1 R) and endothelin ET A receptor (ET A R), stimu- DECEMBER 30, 2016 • VOLUME 291 • NUMBER 53 lated with physiological concentrations of their natural ligands ( Fig. 4, C and D). These results indicate that translocation of GIV to cell membranes via synthetic CID triggers a G protein response similar to the natural G protein activation induced by ligand stimulation of GPCRs.

Molecular Mechanism of GIV Action in Cells
The GBA Motif of GIV Is Sufficient to Promote G Protein Signaling upon Membrane Translocation-Our results with the F1685A mutant indicate that the GBA motif of GIV is required for the observed G protein signaling responses. However, the C-terminal region of GIV used in these experiments (i.e. amino acids 1660 -1870) contains additional domains and mediates other interactions with important signaling functions. In particular, this region contains an SH2-like domain that associates with active RTKs (25,26) and tyrosine phosphorylation sites (46); two signaling events have been shown to work coordinately with the GBA motif to mediate certain downstream responses (25,31,46). Thus, translocation of GIV to membranes may promote some of these signaling-related events, which in turn might influence how the GBA motif engages and modulates G proteins. We found that this is likely not the case because a truncated FKBP-fused GIV construct containing almost exclusively the GBA motif sequence (FKBP-GIV-GBA) induced a G protein activation response equivalent to that observed with the entire C-terminal region (aa 1660 -1870) upon CID-mediated membrane translocation (Fig. 5). This result suggests that the GBA motif is not only necessary but also sufficient to activate G protein signaling when recruited to membranes.
Membrane Recruitment of GIV Is More Potent than Phosphorylation at Ser 1674 in Activating G Proteins-It has been previously reported that after receptor stimulation, GIV is phosphorylated at Ser 1674 , a residue adjacent to the GBA motif, and that a mutant that mimics this phosphorylation (S1674D) enhances G protein binding and activation in vitro, as well as some signaling outputs in cells (47). Because the experiments reported in the current manuscript were carried out under conditions in which GIV is presumably not phosphorylated (e.g. absence of serum or receptor stimulation), Ser 1674 phosphorylation does not seem to be a requirement for G protein activation upon GIV recruitment to membranes. Next, we compared the impact of both mechanisms (Ser 1674 phosphorylation and membrane recruitment) on GIV-mediated G protein activation and also investigated their possible cooperation. For this, we first performed experiments in the yeast-based system described in Fig. 2. When the phosphomimicking S1674D mutation was introduced in the cytosolic version of GIV, it induced larger G protein activation than GIV WT, as deter-mined by the levels of Fus3 phosphorylation (Fig. 6A) and Fus1 promoter-dependent ␤-galactosidase activity (Fig. 6B). However, the enhancement mediated by S1674D was very modest compared with the effect of targeting GIV to membranes (N9Gpa1-GIV; Fig. 6, A and B); in the ␤-galactosidase reporter assay (Fig. 6B), N9Gpa1-GIV was ϳ10 times more potent that GIV S1674D (ϳ25-fold versus ϳ2.5-fold increase compared with GIV WT for N9Gpa1-GIV and GIV S1674D, respectively). Moreover, when the S1674D mutation was introduced in the membrane-anchored N9Gpa1-GIV construct, it did not cause any further increase in G protein activation in neither of the two assay readouts (Fig. 6, A and B). Although this suggests that the activation induced by GIV recruitment to membranes overcomes any possible modulatory effect of S1674D on G protein activation, it is also possible that such modulatory effect is beyond the sensitivity of these assays because of the very high responses induced by membrane-associated GIV.
To overcome this possible limitation and further compare the two mechanisms, we performed additional experiments via an alternative approach, i.e. the CID/BRET system described in Fig. 3. In this format, it is possible to titrate the amount of GIV to obtain submaximal G protein activation response and therefore be able to interrogate a possible modulatory effect of S1674D. Expression of equivalent amounts of FKBP-GIV-GBA   DECEMBER 30, 2016 • VOLUME 291 • NUMBER 53 WT or S1674D did not alter the basal BRET levels in the absence of rapamycin (Fig. 6, C and D), indicating that S1674D cannot enhance G protein signaling per se under conditions in which FKBP-GIV-GBA is not associated with membranes. As expected, rapamycin induced a rapid increase in BRET in cells expressing FKBP-GIV-GBA WT (Fig. 6E), indicative of G protein signaling activation upon membrane recruitment. When FKBP-GIV-GBA S1674D was tested in parallel, the BRET response was almost identical to WT (Fig. 6E). This cannot be attributed to a lack of sufficient dynamic range in the assay because FKBP-GIV-GBA WT and S1674D were tested at concentrations that led to approximately half of the maximal BRET response observed when larger amounts of FKBP-GIV-GBA WT are transfected (Fig. 6E). Also, the lack of difference between WT and S1674D cannot be attributed to constitutive phosphorylation of Ser 1674 in FKBP-GIV-GBA WT because the experiments were performed under conditions (i.e. absence of serum or other growth factors) in which GIV phosphorylation at Ser 1674 is nearly absent (60) and because even when GIV is phosphorylated at Ser 1674 , the stoichiometry is low (47). In summary, we conclude that membrane recruitment of GIV is a mechanism of G protein signaling activation more potent than phosphorylation at Ser 1674 and that it masks the G protein modulatory effect of this phosphoevent.

Molecular Mechanism of GIV Action in Cells
EGFR-mediated Recruitment of the GBA Motif of GIV to Membranes Is Sufficient to Trigger G Protein Signaling-Our results so far indicate that translocation of the GIV GBA motif to membranes in the proximity of its target, G␣ i3 , is sufficient to activate G protein signaling. Next we asked whether membrane translocation in response to a physiological input, instead of via synthetic CID, is also sufficient for G protein activation. For this we used activation of the prototypical RTK EGFR, which is a natural input for GIV-mediated G protein activation. It has been previously shown that full-length GIV or GIV 1660 -1870 binds to phosphotyrosines in the C-terminal tail of active EGFR at the plasma membrane (25,26,29), and such binding is required for G␣ i activation (25,26,29). However, it is unclear whether physical relocalization of GIV to membrane-bound receptors is sufficient by itself for directly activating G proteins or only required, because it might need to work in coordination with other GIV-related functions to eventually lead to G protein activation. To test for sufficiency, we generated a chimeric protein (namely Grb2-GBA) consisting of the GBA motif of GIV (thus lacking any other domain possibly involved in signaling) fused to Grb2 (as a module for binding to active EGFR) (Fig.  7A). Grb2 is a well characterized adaptor protein that translocates from the cytosol to the plasma membrane by directly binding to phosphotyrosines in the C-terminal tail of active EGFR. One of the major biological functions of this translocation is to recruit SOS (Son of Sevenless), a constitutive binding partner or Grb2 and a GEF for Ras, from the cytosol to the plasma membrane to activate membrane-bound Ras (61-63). Thus, Grb2-GBA should mimic the recruitment of native GIV in response to a stimulus that normally activates its G protein regulatory function but lacking any non-GBA contribution to the outcome. We found that EGF stimulation triggers G protein activation in cells expressing Grb2-GBA (Fig. 7B). This response was dependent on the amount of Grb2-GBA expressed and not reproduced by the expression of equivalent amounts of the GEF-deficient mutant Grb2-GBA FA or in the FIGURE 6. Membrane recruitment of GIV is a mechanism of G protein activation more potent than its phosphorylation at Ser 1674 . A and B, membrane targeting of GIV enhances G protein activation in yeast much more potently than the phosphomimicking S1674D mutation and S1674D fails to further enhance G protein activation induced by membrane-anchored GIV. G protein activation in yeast was determined by ppFus3 immunoblotting (A, IB) or ␤-galactosidase reporter activity (B) in the strain described for absence of EGFR (Fig. 7, B and C). These results indicate that relocalization of the GBA motif of GIV from the cytosol to membrane-bound receptors in response to a natural stimulus is sufficient for G protein activation.

Discussion
The main finding of this work is the identification of GIV recruitment to membranes as a major mechanism responsible for the activation of its G protein regulatory function. The data presented here together with previously published evidence (16,25,29,40) support a model in which GIV and its substrate G protein, G␣ i , are spatially segregated from each other under resting conditions but are brought in close proximity when GIV is recruited to the plasma membrane upon cell stimulation. Our results using synthetic or physiological stimuli demonstrate that such physical relocalization of GIV is sufficient to promote G protein signaling. This mechanism operates in the context of GPCR-independent G protein activation and provides novel insights into a long standing gap in the understanding of signaling cross-talk and G protein transactivation by RTKs. We believe that the mechanism described here shares striking similarity to one of the best characterized paradigms in RTK signaling, i.e. GEF-mediated activation of the small G protein Ras (Fig. 8). It has been appreciated for over two decades that the primary mechanism, albeit not the only one (64), by which the GEF SOS activates Ras is its translocation to membranes upon RTK stimulation (61)(62)(63). Much like Ras, G␣ i is constitutively associated with membranes, and much like SOS, GIV is not. Both SOS and GIV have intrinsic GEF activity without modifications, and both of them can be relocated to the proximity of their respective target G proteins on membranes by recruitment to activated RTKs. The mechanism underlying the GEF recruitment also shares similarities because in both cases it relies on phosphotyrosines on RTK tails that serve as docking sites for SH2 domains. The difference is that SOS is recruited indirectly via the SH2 domain-containing adaptor Grb2 (61)(62)(63), to which SOS binds constitutively, whereas GIV does it directly via an SH2-like domain located in its C-terminal region (25) (Fig. 8). Therefore, it appears that non-canonical activation of heterotrimeric G proteins by RTKs can be achieved with a set of signaling elements (constitutively active GEF, SH2 domains) and events (spatial segregation, recruitment via RTK-pY) analogous to those used for canonical activation of small G proteins by RTKs. Interestingly, recent evidence 4 suggests that the structural basis for the GEF action of GIV on G␣ i3 also shares similarities with the mechanism used by SOS on Ras, i.e. both GIV and SOS bind to a region that overlaps with the effector binding site on the G protein and perturb primarily structural elements responsible for binding the nucleotide phosphates. Muntean and Martemyanov (65) recently reported that the activity of a different type of G protein regulators, i.e. RGSs (regulators of G protein signaling), might also be determined by membrane localization. Much like in some of our experiments, they combined chemogenetic control of spatial localization with real time BRET-based biosensors to show that membrane association of RGS proteins of the R7 family is sufficient for their GTPase-accelerating protein activity. These observations not only emphasize the importance of subcellular localization for the activity of G protein regulators but also showcase the power of combining synthetic biology approaches and biosen-  DECEMBER 30, 2016 • VOLUME 291 • NUMBER 53 sors to simultaneously manipulate and monitor G protein signaling with high spatiotemporal resolution.

Molecular Mechanism of GIV Action in Cells
GIV translocation to membranes might also be important to regulate G proteins in a context different from receptor signaling at the plasma membrane. GIV has been reported to regulate G protein signaling in membranous compartments of the cell different from the plasma membrane, like endosomes (22), autophagosomes (41), or the Golgi apparatus (42). Because these compartments are also accessible to cytosolic GIV, it is tempting to speculate that its translocation to organelle membranes might also serve as a mechanism of G protein regulation therein. However, it is unclear how GIV localization would be regulated in these scenarios. Although it is conceivable that GIV might traffic along early steps of the endocytic pathway still associated with tyrosine phosphorylated RTKs, this could not explain the action of GIV on autophagosomes. As for the association with Golgi membranes, it has been recently reported that active Arf1, a Golgi-resident small G protein, directly recruits GIV to the Golgi apparatus, where it activates heterotrimeric G proteins to regulate trafficking (42). In light of the findings reported here, it is possible that Arf1-mediated recruitment of GIV serves as the mechanism that enables its G protein regulatory function on Golgi membranes.
Yet another mechanism of regulation for the association of GIV with membranes could be provided by its previously described phosphoinositide 4-phosphate (PI4P) binding function (40). Recent evidence indicates that, contrary to previous belief, PI4P is not only found on the Golgi apparatus but also at the plasma membrane and endosomes (66,67). Based on results presented here (Fig. 1) or previously published by others (40), full-length GIV does seem to use this PI4P-binding region to associate with membranes under unstimulated conditions. Nevertheless, it is possible that it enhances GIV membrane association achieved by other means, like receptor-mediated recruitment. An interesting speculation in this regard is raised by the fact that PI4P binding is abolished upon GIV phosphorylation at Ser 1416 by Akt (40). Because it is well established that the GEF activity of GIV leads to downstream Akt activation (16 -18, 29, 31), we speculate that GIV phosphorylation at Ser 1416 might serve as a negative feedback regulation for GIVmediated G protein activation by favoring its dissociation from membranes. However, additional work will be required to establish the impact of PI4P binding on the association of GIV with membranes and its possible role as a negative feedback regulation mechanism.
We have recently identified another non-receptor GEF of the same family as GIV, namely DAPLE (20). DAPLE contains a GBA motif that regulates heterotrimeric G proteins in the context of Wnt signaling. DAPLE shares other domains in common with GIV, including the PI4P-binding region, central coiledcoil, and N-terminal HOOK domain, but differs in the C-terminal region where the SH2-like domain of GIV is located (20). Interestingly, the corresponding C-terminal region of DAPLE is responsible for its association with Wnt receptors of the Frizzled family, which occurs only upon ligand stimulation and leads to DAPLE redistribution from the cytosol to the plasma membrane (20). This suggests that membrane translocation upon recruitment to receptors might be a mechanism of activation common for different GBA-motif containing GEFs but via a different set of molecular interactions.
Along the same lines, it is also possible that membrane translocation is a mechanism to activate GIV downstream of receptors different from RTKs. More specifically, it has been recently described that GIV also triggers G protein signaling in response to activation of integrins (16,31). Interestingly, three independent studies have shown that GIV directly binds to integrins FIGURE 8. Model depicting the parallelism between the mechanisms of activation of G␣ i and Ras by their respective cytoplasmic GEFs, GIV, and SOS, upon RTK stimulation. A, under resting conditions, SOS is primarily located in the cytosol along with Grb2, whereas its substrate G protein Ras is constitutively anchored to the plasma membrane, thereby precluding SOS action. Upon RTK stimulation, Grb2-SOS complexes translocate to the plasma membrane via binding of Grb2 SH2 domains to tyrosine phosphorylated EGFR. This change of localization brings SOS in physical proximity to Ras, thereby promoting G protein activation. B, under resting conditions, GIV is primarily located in the cytosol, whereas its substrate G protein G␣ i is constitutively anchored to the plasma membrane, thereby precluding GIV action. Upon RTK stimulation, GIV translocates to the plasma membrane via binding of its SH2-like domain to tyrosine phosphorylated EGFR. This change of localization brings GIV in physical proximity to G␣ i , thereby promoting G protein activation.
after stimulation (16,68,69). Our results presented here suggest that such recruitment to integrins might be sufficient to induce the observed GIV-dependent G protein signaling.
Although here we report membrane recruitment as a major regulatory mechanism for GIV-mediated G protein activation, we cannot rule out that other mechanisms come into play to modulate the GEF function of GIV. In particular, it will be important to address in the future how phosphorylation of Ser 1674 impinges on GIV-mediated signaling. Although our results clearly indicate that membrane translocation is more potent than Ser 1674 phosphorylation in activating GIV, previous evidence has established that this phosphorylation regulates events downstream of GIV (47). One possibility is that Ser 1674 phosphorylation fine tunes G protein activation and that the methods used in the present work cannot detect such fine alterations. Another possibility is that Ser 1674 works in coordination with other events to control the fate of the G protein regulatory function of GIV via alteration of its subcellular localization or other mechanisms. In support of the latter, it has been recently reported that sequential phosphorylation of GIV at Ser 1674 and Ser 1689 determines its preferential coupling to different G protein subtypes in different subcellular locations (60).
In summary, our findings provide novel mechanistic insights into how G protein activation by GIV is achieved in cells, which also has important implications for the understanding of G protein transactivation by non-GPCR receptors. It remains to be investigated whether this mechanism also applies to other GIVrelated GEFs, like DAPLE, and how it is integrated with other possible regulatory events, like phosphorylation or phosphoinositide binding.

Experimental Procedures
Reagents and Antibodies-Unless otherwise indicated all reagents were of analytical grade and obtained from Fisher Scientific or Sigma. Cell culture media and goat anti-rabbit and goat anti-mouse Alexa Fluor 680 antibodies were from Life Technologies. Fluorescein di-␤-D-galactopyranoside was from Marker Gene Technologies, and the protein inhibitor mixture (catalog no. S8830) from Sigma. Human EGF, leupeptin, pepstatin, and aprotinin were from Gold Biotechnology. All restriction endonucleases were from Thermo Scientific, and Escherichia coli strain DH5␣ was purchased from New England Biolabs. Pfu ultra DNA polymerase was purchased from Agilent. Rabbit antibodies raised against GIV (T-13), TfR (CD71, H68.4), G␣ i3 (C10), and G␤ (M-14) were from Santa Cruz Biotechnology (Santa Cruz, CA). Mouse monoclonal antibodies raised against Myc tag (9E10), ␣-tubulin (Sigma catalog no. T6074), and HA tag (12CA5) were from the Developmental Studies Hybridoma Bank (University of Iowa), Sigma, and Roche, respectively. Rabbit antibodies for ppERK (catalog no. 4370) and integrin ␤1 (catalog no. D2E5) and mouse antibodies for Myc (catalog no. 9B11) were from Cell Signaling. Mouse monoclonal antibodies for Caveolin-1 (catalog no. 610406) were from BD Transduction Laboratories. Rabbit polyclonal antibody for RFP was from Abcam (catalog no. 62341). Goat anti-rabbit and goat anti-mouse IRDye 800 F(abЈ) 2 were from Li-Cor Biosciences (Lincoln, NE).
Fractionation of Mammalian Cells-HEK293T, HeLa, or MDA-MB-231 cells cultured in DMEM plus 10% FBS and at ϳ90% confluency were scraped in PBS and pelleted by centrifugation. The cells corresponding to 0.5 (HEK293T) or 1.5 (HeLa or MDA-MB-231) 10-cm dishes were resuspended with 250 l of detergent-free homogenization buffer (20 mM HEPES, pH 7.2, 5 mM Mg(CH3COO) 2 , 125 mM K(CH 3 COO), 1 mM DTT, and protease inhibitor mixture) in ice. All subsequent steps were carried out in ice or at 4°C. The cells were homogenized by 30 passages though a 30-gauge needle and subsequently centrifuged at 1,000 ϫ g for 10 min. An aliquot of the supernatant was set aside (post-nuclear supernatant (PNS) fraction), and the rest was centrifuged in a fixed angle rotor TLA-55 in a Beckman Coulter Optima MAX-E tabletop centrifuge for 45 min at 100,000 ϫ g. An aliquot of the supernatant was set aside (S100 fraction). The pellet was resuspended to the starting volume with homogenization buffer, and an aliquot was set aside (P100 fraction). Triton X-100 was added to the P100 fraction (0.4% v/v final) and after a 30-min incubation DECEMBER 30, 2016 • VOLUME 291 • NUMBER 53 centrifuged for 45 min at 100,000 ϫ g as described above. The supernatant containing the Triton X-100-soluble material (TX-S fraction) was saved, and the pellet containing the Triton X-100-insoluble material resuspended to the starting volume (TX-I). Equal volumes of each fraction were added Laemmli sample buffer and boiled before protein separation by SDS-PAGE. Identical experiments were carried out with an alternative homogenization buffer (10 mM HEPES, pH 7.4, sucrose 250 mM, 10 mM KCl, 1.5 mM MgCl 2 , 1 mM DTT, and protease inhibitor mixture), obtaining similar results.

Molecular Mechanism of GIV Action in Cells
Lipid Raft Isolation-This assay was performed as previously described (70) with some modifications. MDA-MB-231 cells cultured in DMEM plus 10% FBS and at ϳ90% confluency were scraped in PBS and pelleted by centrifugation. Cells corresponding to two 10-cm dishes were resuspended with 1 ml of lysis buffer (20 mM HEPES, pH 7.2, 5 mM Mg(CH3COO) 2 , 125 mM K(CH 3 COO), 0.4% (v/v) Triton X-100, 1 mM DTT, and protease inhibitor mixture) in ice. All subsequent steps were carried out in ice or at 4°C. The lysate was incubated in ice for 30 min and subsequently mixed with an equal volume of 90% (w/v) sucrose (prepared in lysis buffer). A discontinuous sucrose gradient was prepared by layering 2 ml of 35% sucrose and 1.25 ml of 5% sucrose on top of 1.5 ml of the 45% sucrose lysate mixture. This gradient was centrifuged at 200,000g for 16h in a SW50.1 rotor using a Beckman Coulter Optima LE-80K ultracentrifuge. Fractions (0.5 ml) were carefully collected from the top. Equal volumes of each fraction were added Laemmli sample buffer and boiled before protein separation by SDS-PAGE.
Fractionation of Yeast Cells-This procedure was performed as described previously (54) with minor modifications. Pellets corresponding to 15 OD600 of exponentially growing cells resuspended in 200 l of cold homogenization buffer (40 mM triethanolamine, pH 7.2, 2 mM EDTA, 2 mM DTT, 1 M leupeptin, 2.5 M pepstatin, 0.2 M aprotinin and 1 mM PMSF). All subsequent steps were carried out in ice or at 4°C. Approximately 50 l of glass beads were added to the tubes and vortexed for 10 min. The homogenate was recovered by poking a whole in the bottom of the tubes and centrifugation onto new tubes. Beads were washed with 100 l of homogenization buffer and the recovered volume pooled with initial homogenate. After centrifugation at 500g for 10 min, an aliquot of the supernatant was set aside (PNS fraction) and the rest was centrifuged at 100,000 as described above in "Fractionation of Mammalian Cells" to obtain the "S" and "P" fractions. Equal volumes of each fraction were added Laemmli sample buffer and boiled before protein separation by SDS-PAGE.
Yeast Strains and Manipulations-The previously described (53) S. cerevisiae strain CY7967 [MAT␣ GPA1(1-41)-G␣ i3 far1⌬ fus1p-HIS3 can1 ste14:trp1:LYS2 ste3⌬ lys2 ura3 leu2 trp1 his3] (kindly provided by James Broach, Penn State University) was used for all experiments. The main features of this strain are that the only pheromone responsive GPCR (Ste3) is deleted, the endogenous G␣-subunit GPA1 is replaced by a chimeric GPA1(1-41)-human G␣ i3 (36 -354) and the cell cycle arrest-inducing protein far1 is deleted. In this strain, the pheromone response pathway can be up-regulated by the ectopic expression of activators of human G␣ i3 and does not result in the cell cycle arrest that occurs in the native pheromone response (53). Plasmid transformations were carried out using the lithium acetate method. CY7967 cells were transformed with a plasmid encoding for the P Fus1 ::LacZ reporter and subsequently with plasmids pYES2, pYES2-GIV or pYES2-N9Gpa1-GIV (see "Plasmids"). Double transformants were selected in Synthetic Defined (S.D.)-TRP-URA media. Individual colonies were inoculated into 3 ml of SDGalactose-TRP-URA and incubated overnight at 30°C to induce the expression of the proteins of interest under the control of a galactoseinducible promoter of pYES2. For the spot growth assays, the overnight cultures were washed and resuspended in water at an OD600 of 1 and equal volumes of each strain (6 l) were spotted on SDGal-TRP-URA or SDGal-TRP-URA-HIS (plus 10 mM 3-amino-triazole) agar plates and incubated at 30°C for 4 days. Plates were scanned at days 2, 3 and 4 using the same settings. For the preparation of samples for immunoblotting and ␤-galactosidase activity assays, the overnight cultures were used to inoculate 20 ml of SDGalactose-TRP-URA at 0.3 OD600. Pellets of exponentially growing cells (ϳ0.7-0.8 OD600, 4 -5 h) corresponding to 5 OD600 were washed with PBS ϩ 0.1% BSA (w/v) and used to prepare immunoblotting samples as described previously (21). Briefly, washed pellets were resuspended in 150 l of lysis buffer (10 mM Tris-HCl, pH 8.0, 10% (w/v) TCA, 25 mM NH 4 OAc, 1 mM EDTA). 100 l of glass beads were added to each tube and vortexed at 4°C for 5 min. Lysates were separated from glass beads by poking a hole in the bottom of the tubes followed by centrifugation onto a new set of tubes. The process was repeated after the addition of 50 l of lysis buffer to wash the glass beads. Proteins were precipitated by centrifugation (10 min, 20,000 ϫ g) and resuspended in 60 l of solubilization buffer (0.1 M Tris-HCl, pH 11.0, 3% SDS). Samples were boiled for 5 min, centrifuged (1 min, 20,000 ϫ g), and 50 l of the supernatant was transferred to new tubes containing 12.5 l of Laemmli sample buffer and boiled for 5 min (ϳ15-20 l were typically loaded per lane). For the ␤-galactosidase assays, pellets of exponentially growing cells corresponding to 0.5 A 600 were washed with PBS ϩ 0.1% BSA (w/v) and resuspended in 200 l assay buffer (60 mM Na 2 PO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgCl 2 , 0.25% (v/v) ␤-mercaptoethanol, 0.01% (w/v) SDS, 10% (v/v) chloroform) and vortexed. 100 l were transferred to 96-well plates, and reactions were started by the addition of 50 l of the fluorogenic ␤-galactosidase substrate fluorescein di-␤-D-galactopyranoside (100 M final). Fluorescence (excitation, 485 Ϯ 10 nm; emission, 528 Ϯ 10 nm) was measured every 2 min for 90 min at 30°C in a Biotek H1 synergy plate reader. Enzymatic activity was calculated from the slope of fluorescence (arbitrary units) versus time (min). Each independent clone was measure in duplicate, and the results are expressed as raw arbitrary fluorescent units/min or normalized (%) to the activity of strains carrying pYES2-GIV.
Cell Culture, Transfections, and BRET Assay-HEK293T, HeLa and MDA-MB-231 cells were grown at 37°C in DMEM supplemented with 10% FBS, 100 units/ml penicillin, 100 g/ml streptomycin, 1% L-glutamine and 5% CO 2 . HEK293T cells were transfected using the calcium phosphate method in gelatin-coated 6-well plates with the following plasmids (quantities in parenthesis): masGRK3ct-Nluc (0.1 g), VC-G␤ 1 (0.2 g), VN-G␥ 2 (0.2 g), A 1 R (0.2 g), ET A R (0.2 g), M4R (0.2 g), G␣ i3 (1.0 g), Lyn11-FRB-HA (3.0 g), Lyn11-FRB-Myc (3.0 g), FKBP-GIV (0.025-2.0 g, as indicated in the figures), FKBP-GIV-GBA (0.125-0.5 g, as indicated in the figures), Grb2-GBA (0.5-2 g, as indicated in the figures), and pcDNA3.1(ϩ) (to equalize the amount of DNA per well). G protein activity was monitored by BRET using a system based on a sensor for free G␤␥ that was originally described in Ref. 57 and optimized in Ref. 71 with a new generation luciferase (Nluc). BRET experiments were carried out and analyzed as described previously (58) with minor modifications. Briefly, 22-28 h after transfection, the cells were gently scrapped in PBS, centrifuged, and resuspended in Tyrode's solution (140 mM NaCl, 5 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 0.37 mM NaH 2 PO 4 , 24 mM NaHCO 3 , 10 mM HEPES, pH 7.4, 0.1% glucose) at a density of 10 6 cell/ml. Twenty-five l of cell suspensions were added to a white opaque 96-well plate (Opti-Plate; Perkin Elmer) and mixed with 3 volumes of the Nanoluciferase substrate Nano-Glo (Promega, diluted 1:150 in Tyrode's solution). After 2 min of incubation, luminescence was measured at room temperature in a Synergy H1 (Biotek) or a POLARstar Omega (BMG Labtech) plate reader at 460 Ϯ 20 and 528 Ϯ 10 nm. BRET signals were calculated as the ratio of the emission intensity at 528 Ϯ 10 nm divided by the emission intensity at 460 Ϯ 20 nm. For the single point measurements, the cells were treated with rapamycin (0.5 M) or an equal volume of vehicle (ethanol, 0.005% final concentration), and BRET was measured after 2 min. For the kinetic experiments, BRET measurements were done every second or every 0.24 s. The BRET baseline was recorded for 30 s, after which 5 l of rapamycin (0.5 M final), adenosine (0.1 M final), endothelin (ET-1, 3 M final), or EGF (50 ng/ml final) were added to the well. For clarity, most experiments are presented as increase in BRET (⌬BRET) relative to control conditions, i.e. baseline signal in kinetic experiments or unstimulated controls for single time point measurements (as indicated in the figures).
Statistical Analyses-Each experiment was performed at least three times. The data shown are presented as means with error bars representing the S.E. or as one representative result of each biological replicate. Statistical significance between various conditions was assessed with Student's t test. p Ͻ 0.05 was considered significant.