Atypical activation of the G protein Gαq by the oncogenic mutation Q209P

The causative role of G protein–coupled receptor (GPCR) pathway mutations in uveal melanoma (UM) has been well-established. Nearly all UMs bear an activating mutation in a GPCR pathway mediated by G proteins of the Gq/11 family, driving tumor initiation and possibly metastatic progression. Thus, targeting this pathway holds therapeutic promise for managing UM. However, direct targeting of oncogenic Gαq/11 mutants, present in ∼90% of UMs, is complicated by the belief that these mutants structurally resemble active Gαq/11 WT. This notion is solidly founded on previous studies characterizing Gα mutants in which a conserved catalytic glutamine (Gln-209 in Gαq) is replaced by leucine, which leads to GTPase function deficiency and constitutive activation. Whereas Q209L accounts for approximately half of GNAQ mutations in UM, Q209P is as frequent as Q209L and also promotes oncogenesis, but has not been characterized at the molecular level. Here, we characterized the biochemical and signaling properties of Gαq Q209P and found that it is also GTPase-deficient and activates downstream signaling as efficiently as Gαq Q209L. However, Gαq Q209P had distinct molecular and functional features, including in the switch II region of Gαq Q209P, which adopted a conformation different from that of Gαq Q209L or active WT Gαq, resulting in altered binding to effectors, Gβγ, and regulators of G-protein signaling (RGS) proteins. Our findings reveal that the molecular properties of Gαq Q209P are fundamentally different from those in other active Gαq proteins and could be leveraged as a specific vulnerability for the ∼20% of UMs bearing this mutation.

toring and detection, survival rates have not improved over the past few decades (1,2). Regardless of tumor classification, UMs are treated by eye enucleation, plaque radiation, or tumor resection, which result in loss of vision (3,4). For the ϳ50% of UMs that are metastatic (5,6), these treatments do not reduce the probability of mortality. Patients with metastatic UM have a grim median survival of less than 1 year (15% 5-year survival) (7)(8)(9). UM is largely insensitive to chemotherapeutics or immune checkpoint inhibitors that have shown efficacy in cutaneous melanomas (7,9). This is easily explained by the fact that the oncogenic drivers of UM are different from those of cutaneous melanoma (i.e. ϳ90% of UMs are caused by activating mutations in GNAQ or GNA11, the genes encoding the G proteins G␣ q and G␣ 11 , and not by mutations in BRAF) (10 -14).
Heterotrimeric G proteins, such as G q and G 11 , are composed of a nucleotide-binding G␣ subunit and an obligatory G␤␥ heterodimer, which form a tight complex in the GDPbound resting state (15). Upon ligand binding, GPCRs promote their activation by accelerating the exchange of GDP for GTP. In turn, G␣-GTP dissociates from G␤␥, allowing them both to regulate numerous downstream effectors. Cancer-associated mutations in G␣ q and G␣ 11 affect residue Gln-209 in ϳ95% of the cases, whereas mutations in Arg-183 are less frequent (11,12). These residues are critical for the intrinsic GTPase activity of G␣, and their mutation is known to result in a constitutively active protein (16 -18). Active G␣ q/11 can engage with multiple downstream effectors, which trigger various signaling pathways implicated in cell growth and oncogenic behavior. For instance, active G␣ q binds to some PLC␤ isoforms, which ultimately promotes activation of the MAPK/ERK pathway via Ca 2ϩ /diacylglycerol/protein kinase C (19 -21). Another major G␣ q -dependent mechanism relies on its direct binding and activation of a subfamily of RhoGEFs (22,23), such as Trio, which ultimately activate TAZ/YAP to promote oncogenic transformation (24,25). Recent evidence indicates that nearly all UMs bear one mutation within this GPCR-driven pathway (13,14,26). In addition to mutations in GNAQ or GNA11 found in ϳ90% of the cases, there are also mutually exclusive mutations in CYSLTR2 (encoding the GPCR cysteinyl leukotriene receptor 2, CysLT2R) (27) and PLCB4 (encoding the G protein effector PLC␤4) (28), which operate directly upstream or downstream, respectively, of G␣ q/11 . Interestingly, a similar pattern of mutually exclusive mutations in GNAQ, GNA11, CYSLTR2, and PLCB4 has been reported to occur in leptomeningeal melanocytic tumors (29 -31), another type of noncutaneous melanoma that afflicts the central nervous system.
Given the insensitivity of UM to therapies used for other types of melanoma, targeting the signaling mechanisms triggered by G␣ q/11 has been in the limelight for the development of novel therapeutics for this type of cancer (32). Their suitability as targets is supported by many lines of evidence. For example, expression of the active G protein mutants in nontransformed cells is oncogenic (11,12,33). Similarly, mouse models in which activated G␣ q or G␣ 11 are expressed in melanocytes develop metastatic UM (34,35). Importantly from the standpoint of therapeutic targeting, genetic disruption of mutant G␣ q/11 or downstream signaling effectors in UM cells impairs proliferation and/or tumor growth in mice (24, 25, 36 -38). Unfortunately, attempts to pharmacologically target signaling pathways activated downstream of mutant G␣ q/11 have not been successful, even when using multiple drugs in combination (7). A likely explanation for the inefficiency of these approaches in blunting UM is that G␣ q/11 activates a complex network of signaling effectors (14), such that targeting individual nodes of this network is insufficient to achieve therapeutic effects. Thus, direct inhibition of mutant G␣ q/11 may be required to completely inhibit all the network components required to promote UM and achieve sufficient efficacy and therapeutic benefit.
Direct targeting of oncogenic G␣ q/11 mutants is a reasonable therapeutic approach, although it presents challenges, as these mutants are predicted to closely resemble active G␣ q/11 WT. If so, strategies to inhibit mutant G␣ q/11 would be expected to also cause inhibition of G␣ q/11 WT, which could result in undesired side effects related to the major physiological functions of these G proteins (e.g. double G␣ q /G␣ 11 knockout mice are nonviable (39)). Here, we present evidence that one of the most frequent G␣ q mutants in UM, Q209P, displays properties different from that of active G␣ q WT that could be leveraged to achieve specific targeting at the molecular level. Approximately 40 -45% of UMs have mutations in residue Gln-209 of G␣ q , which are split evenly between Q209L and Q209P (11)(12)(13). Whereas G␣ q Q209L has been extensively characterized and used as a tool mutant to study G q signaling for decades, G␣ q Q209P has not been adequately studied. G␣ q Q209P is not only as frequent as G␣ q Q209L in tumors, but it is also the driver mutation of many of the UM cell lines commonly used for cell biological and pharmacological experimentation, such as Mel270, OMM1.3 (also known as OMM2.3), OMM2.2, OMM2.5, and UPMM3 (25,40). In fact, it has been shown that depletion of G␣ q Q209P and/or signaling components downstream of it from UM cells decreases proliferation and/or tumor growth (24,36,37). Motivated by the fundamental gap in knowledge about the molecular properties of G␣ q Q209P, we characterized this mutant by direct comparison with G␣ q Q209L, to discover that while leading to signaling hyperactivation, as expected, it possessed structural features different from other active G␣ q species, including active G␣ q WT and the G␣ q Q209L mutant. From a broader perspective, our findings reveal novel mechanistic insights into how G protein mutants lead to oncogenesis and into their possible suitability as direct targets for pharmacological intervention.

Binding of signaling effectors to G␣ q Q209P is weaker than to G␣ q Q209L or GTP-bound G␣ q WT
To start characterizing the properties of G␣ q Q209P, we compared its ability to bind effectors with that of G␣ q Q209L or active G␣ q WT. First, we investigated binding to GRK2. GRK2 was the first protein co-crystallized with active G␣ q (41). Based on the atomic resolution structure of this complex, it was proposed, and subsequently confirmed, that GRK2 binding to G␣ q has effector-like properties. We expressed G␣ q WT, G␣ q Q209L, and G␣ q Q209P in HEK293T cells and carried out pulldowns with the RGS homology (RH) domain of GRK2 (aa 45-178) fused to GST. Cell lysis and pulldowns were carried out rapidly (ϳ1.5 h from lysis to protein complex elution) at 4°C in a buffer without Mg 2ϩ supplemented with EDTA to minimize GTP hydrolysis during the assay. As a positive control, we also included a condition in which lysates of cells expressing G␣ q WT were supplemented with Mg⅐AlF 4 Ϫ , which mimics the GTP-bound transition state of the G protein that binds with high affinity to effectors. As expected, G␣ q WT loaded with Mg⅐AlF 4 Ϫ bound robustly to GRK2, whereas G␣ q WT in the absence of Mg⅐AlF 4 Ϫ , presumably in its GDP-bound inactive conformation, did not (Fig. 1A). Also, as expected, G␣ q Q209L bound to GRK2 as much as G␣ q WT supplemented with Mg⅐AlF 4 Ϫ , which is consistent with its predicted constitutive GTP-bound status. Surprisingly, G␣ q Q209P binding to GRK2 was weaker than G␣ q Q209L or active G␣ q WT, although consistently stronger than G␣ q WT in the absence of Mg⅐AlF 4 Ϫ (Fig. 1A). Prompted by these surprising results, we set out to investigate the binding properties of G␣ q Q209P with other G q effectors. G␣ q works predominantly on two classes of effectors: a subfamily of PLC isoforms (PLC␤s), and a subfamily of Rho-GEFs composed of p63RhoGEF (also known as ARHGEF25), Trio (also known as ARHGEF23), and kalirin (also known as ARHGEF24). Both PLC␤3 and p63RhoGEF have been co-crystallized with G␣ q (42)(43)(44). The atomic resolution structures of these complexes reveal similarities with the GRK2-G␣ q complex. More specifically, one of the primary contacts of GRK2, p63RhoGEF, and PLC␤3 with G␣ q is mediated by a helix-turnhelix element that docks onto a conserved pocket between the ␣3 helix and switch II (SwII) region of the G protein (Fig. 1B). Consistent with these structural similarities, we found that the binding pattern of p63RhoGEF to different G␣ q species closely resembled that of GRK2 (i.e. G␣ q Q209P binding to p63RhoGEF (DH/PHext domain (45), aa 155-493) was weaker than that of G␣ q Q209L or Mg⅐AlF 4 Ϫ -loaded G␣ q WT, although stronger than G␣ q WT in the absence of Mg⅐AlF 4 Ϫ ) (Fig. 1C). Similar results were obtained in pulldowns with the other two

Atypical properties of GNAQ Q209P
members of the p63RhoGEF subfamily, Trio and kalirin (Fig. 1,  D and E), and in co-immunoprecipitations with PLC␤3 (Fig.  1F). The defect in G␣ q Q209P binding to effectors was not res-cued by the addition of Mg⅐AlF 4 Ϫ (Fig. S1), suggesting that the diminished binding is not because the G protein is in a GDPbound state. does not bind to effectors as efficiently as G␣ q Q209L or active G␣ q WT. A, G␣ q Q209P binds to GRK2 less than G␣ q Q209L or active G␣ q WT (ϩMg⅐AlF 4 Ϫ ) but more than inactive G␣ q WT (ϪMg⅐AlF 4 Ϫ ). Lysates of HEK293T cells transfected with HA-G␣ q (WT or mutants, as indicated) were incubated with GST-GRK2 RH immobilized on GSH-agarose beads in the presence or absence of 10 mM MgCl 2 ,30M AlCl 3 , and 10 mM NaF (Mg⅐AlF 4 Ϫ ) as indicated. Resin-bound proteins (top panels) and aliquots of the lysates (bottom panels) were analyzed by Ponceau S staining and immunoblotting (IB) as indicated. One representative result of at least three independent experiments is shown. B, a helix-turn-helix in either GRK2 (purple), PLC␤3 (cyan) or p63RhoGEF (red) docks onto a groove of active G␣ q (gray) formed between the SwII region (green) and the ␣3 helix (orange). The structures of G␣ q bound to either GRK2 (PDB code 2BCJ), PLC␤3 (PDB code 3OHM), or p63RhoGEF (PDB code 2RGN) were superimposed using G␣ q , and the protein-protein interfaces are shown enlarged on the left. Only a single G␣ q is shown for clarity. C-E, G␣ q Q209P binds to RhoGEFs p63RhoGEF (C), Trio (D), or kalirin (E) less than G␣ q Q209L or active G␣ q WT (ϩMg⅐AlF 4 Ϫ ) but more than inactive G␣ q WT (ϪMg⅐AlF 4 Ϫ ). Pulldowns were carried out as in A using GST fusions of the DH/PHext domain of p63RhoGEF (aa 155-493), Trio (aa 1967-2296), or kalirin (aa 1901-2223). One representative result of at least three independent experiments is shown. F, G␣ q Q209P binds to PLC␤3 less than G␣ q Q209L or active G␣ q WT (ϩMg⅐AlF 4 Ϫ ) but more than inactive G␣ q WT (ϪMg⅐AlF 4 Ϫ ). Lysates of HEK293T cells co-transfected with HA-G␣ q (WT or mutants, as indicated) and FLAG-PLC␤3 were immunoprecipitated (IP) with FLAG antibodies in the presence or absence of 10 mM MgCl 2 , 30 M AlCl 3 , and 10 mM NaF (Mg⅐AlF 4 Ϫ ) as indicated. Resin-bound proteins (top panels) and aliquots of the lysates (bottom panels) were analyzed by Ponceau S staining and immunoblotting (IB) as indicated. One representative result of at least three independent experiments is shown. G, binding of G␣ q Q209P to GRK2 RH in cells is diminished compared with G␣ q Q209L. Top left, schematic of the BRET assay used to monitor G␣ q binding to GRK2 RH . Nanoluciferase-fused GRK2 RH (GRK2 RH -Nluc, BRET donor) binds to active but not inactive G␣ q -V (BRET acceptor), which results in high or low BRET, respectively. Bottom left, representative immunoblots of lysates of HEK293T cells used in the BRET experiments shown on the right. Right, HEK293T cells were co-transfected with plasmids encoding GRK2 RH -Nluc and G␣ q -V (WT or mutants, as indicated) and BRET measured 24 h later. Results are mean Ϯ S.E. (error bars) (n ϭ 5; **, p Ͻ 0.01, Student's t test).

Atypical properties of GNAQ Q209P
To rule out that the weaker binding observed for G␣ q Q209P was due to the experimental conditions of our biochemical assays in vitro, we carried out bioluminescence resonance energy transfer (BRET) experiments in live cells to monitor the interaction between GRK2 and G␣ q (Fig. 1G). Briefly, the same domain of GRK2 used in the pulldowns above (RH domain) was fused to nanoluciferase to generate a BRET donor molecule (GRK2 RH -Nluc). The BRET acceptor consisted of a previously validated G␣ q construct internally tagged with Venus (G␣ q -V) (46). When co-expressed with GRK2 RH -Nluc in HEK293T cells, G␣ q -V Q209L led to a higher BRET signal than G␣ q -V WT, whereas G␣ q -V Q209P did not (Fig. 1G). This result confirms that G␣ q Q209P binding to GRK2 is weaker than Q209L in cells, much like what is observed in the in vitro pulldown experiments. In contrast to the results in pulldown experiments, the BRET assay failed to detect increased binding of G␣ q Q209P compared with G␣ q WT (Fig. 1G). A likely explanation for this is that nonspecific BRET due to random collision of donor and acceptor molecules restricted to the plane of a lipid membrane masks small signal increases due to weak interactions. Taken together, these findings demonstrate that the effector-binding properties of G␣ q Q209P are different from those of other active G␣ q proteins, showing diminished binding compared with GTP-bound G␣ q WT or the constitutively active mutant Q209L.

G␣ q Q209P activates downstream signaling in cells as efficiently as G␣ q Q209L
The findings above are puzzling because previous evidence indicates that G q -dependent signaling in UM cells bearing the Q209P mutation is required to drive cell proliferation and/or tumor growth in mice. More specifically, it has been demonstrated that genetic disruption of G q , of its signaling effectors like the RhoGEF Trio, or of other downstream components of the signaling cascade, such as YAP or RasGRP3, diminishes the proliferation and/or growth of tumor xenografts of UM cells with Q209P mutation, such as OMM1.3 (also known as OMM2.3) or Mel270 (24,36,37). These previous findings argue strongly that the G␣ q Q209P mutant is hyperactive and promotes oncogenic signaling in UM cancer cells.
To further substantiate this point, we investigated the ability of G␣ q Q209P to activate signaling in cells by comparing it side by side with the more thoroughly characterized G␣ q Q209L mutant. First, we looked at MAPK/ERK activation, which occurs downstream of G q -mediated activation of PLC. We found that expression of G␣ q Q209L or G␣ q Q209P in HEK293T cells led to similar increases in phospho-ERK1/2 compared with cells expressing G␣ q WT or a vector control ( Fig. 2A). Next, we investigated whether G␣ q Q209P was also capable of activating a different G q -mediated pathway triggered by activation of RhoGEFs, as it has been previously shown that UM cells bearing the Q209P mutation also rely on this pathway to maintain their oncogenic properties (24,37). For this, we used a luminescent reporter based on the serum response element (SRE) that monitors RhoGEF-dependent activity downstream of active G q (42). We found that both G␣ q Q209L and G␣ q Q209P led to a robust and similar increase of SRE reporter activity compared with G␣ q WT or a vector control when expressed in HEK293T cells (Fig. 2B). One possible explanation for this result is that the signaling response is saturated at high levels of active G␣ q expression, thereby masking possible differences between the two mutants. However, we found that this is not the case because expression of lower amounts of G␣ q Q209L and G␣ q Q209P led to smaller increases of SRE reporter activity, but they were still similar for the two G protein mutants (Fig. S2). Taken together, these results demonstrate that G␣ q Q209P is constitutively active, leading to a hyperactivation of downstream signaling comparable with that observed for the other oncogenic mutant Q209L.

G␣ q Q209P has impaired GTPase activity in vitro like G␣ q Q209L
To start dissecting the possible causes of the apparent discrepancy between signaling activity (Fig. 2, A and B) and effector binding ( Fig. 1) observed when comparing the Q209P and Q209L mutants, we investigated their ability to hydrolyze nucleotides. Gln-209 in G␣ q corresponds to a residue with a conserved function in G protein-mediated GTP hydrolysis, and the oncogenicity of mutations at this residue is generally Figure 2. G␣ q Q209P activates downstream signaling in cells like G␣ q Q209L. A, expression of G␣ q Q209L or G␣ q Q209P leads to similar levels of ERK1/2 phosphorylation. Lysates of HEK293T cells transfected with HA-G␣ q (WT or mutants, as indicated) were immunoblotted (IB) with the indicated antibodies. One representative result of at least three independent experiments is shown. B, expression of G␣ q Q209L or G␣ q Q209P leads to similar levels of SRE reporter activation. HEK293T cells were co-transfected with HA-G␣ q (WT or mutants, as indicated) and a firefly luciferase reporter driven by the SRE.L promoter and cultured overnight in medium with a low concentration of serum (0.5% FBS) before measuring luminescence. Results are mean Ϯ S.E. (error bars) (n ϭ 4).

Atypical properties of GNAQ Q209P
ascribed to constitutive G protein activity due to GTPase deficiency (16,17). Whereas prior research indicates that G␣ q Q209L is GTPase-deficient, the GTPase function of G␣ q Q209P has not been studied before. To directly investigate GTPase activity in vitro, we purified a novel G␣ q variant, hereafter named G␣ q *, that expresses well in Escherichia coli. G␣ q * is a G␣ q /G␣ i1 chimera in which some residues of G␣ q not involved in interacting with G q -specific binding partners or in nucleotide binding/hydrolysis have been replaced by the corresponding residues in G␣ i1 (Fig. S3A). His-tagged G␣ q * purified from E. coli binds GTP␥S with a level of nucleotide occupancy similar to that previously reported for G␣ q purified from insect cells (47) and also binds robustly to several known G␣ q binding partners, such as effectors p63RhoGEF and GRK2, the GEF/chaperone Ric-8A, or the GAP GAIP (also known as RGS19), with the expected activation state dependence (Fig. S3). These results validate that His-G␣ q * recapitulates many of the properties of G␣ q . Next, we compared the GTPase activity of G␣ q * Q209P with that of G␣ q * WT and G␣ q * Q209L. Because the slow rate of nucleotide exchange of G␣ q precludes the efficient loading of nucleotide required to carry out GTPase assays under single-turnover conditions (48), we investigated the GTPase activity under multiple-turnover conditions (i.e. at steady state). As expected, G␣ q * Q209L displayed a lower rate of GTP hydrolysis than G␣ q * WT (Fig. 3). Moreover, G␣ q * Q209P also had diminished GTPase activity, which was comparable with that observed in G␣ q * Q209L (Fig. 3). These results are consistent with G␣ q * Q209P having a defect in GTPase activity comparable with that of G␣ q * Q209L.

The switch II region of G␣ q Q209P adopts a conformation different from other active G␣ q proteins
The results presented so far indicate that although G␣ q Q209P is GTPase-deficient and leads to signaling hyperactivation like G␣ q Q209L, it fails to recapitulate the strong binding to effectors observed for other active G␣ q proteins. We reasoned that whereas mutation of Gln-209 to any residue (including leucine) might be deleterious for GTPase activity, mutation to proline in particular might have a unique impact on the shape of the effector binding site and thereby account for the observed differences in protein-protein binding. There is a strong rationale for this reasoning based on available structural information. First, Gln-209 is located at the end of the SwII region (Fig. 4A). As its name indicates, the SwII alternates between different conformations, depending on the G protein activation status (49). Upon activation, the SwII region adopts a helical conformation that allows high-affinity binding of effectors to a newly formed pocket between SwII and ␣3 helix (Fig. 4A). Thus, we hypothesized that mutation of Gln-209 to proline, an amino acid with a cyclical side chain that imposes rotational constraints, might affect the conformation of the SwII region and distort the effector-binding site. To test this hypothesis, we carried out a well-established trypsin proteolysis assay that reports the conformation of the SwII (50). Briefly, when SwII adopts a helical conformation like that observed in active, GTP-bound G␣ subunits, it becomes insensitive to trypsin digestion. This results in the formation of a trypsin-resistant species in which only an N-terminal fragment of the protein is cleaved off by trypsin, whereas G proteins in which the SwII region does not adopt such helical conformation, like GDP-bound G␣, are readily digested to low-molecular weight products (Fig. 4B). As expected, when lysates of HEK293T cells expressing G␣ q WT were incubated with trypsin, the G protein was readily digested, whereas supplementing the reaction with Mg⅐AlF 4 Ϫ to mimic the GTP-bound state resulted in the formation of a trypsinresistant species (Fig. 4C). Trypsinization of lysates of cells expressing G␣ q Q209L also resulted in the formation of the trypsin-resistant species, consistent with the idea that this mutant is constitutively bound to GTP and mimics the canonical active conformation of G␣ q WT (Fig. 4C). In contrast, the G␣ q Q209P mutant was not trypsin-resistant under the same conditions (Fig. 4C), indicating that its SwII does not adopt the helical conformation that confers trypsin resistance to active G␣ q WT. Because SwII is one of the structural elements of G␣ q that makes direct contact with effectors and shapes the effector binding pocket, these findings help explain the diminished binding of effectors to G␣ q Q209P mutant compared with G␣ q Q209L or active G␣ q WT (Fig. 1).

G␣ q Q209P is constitutively dissociated from G␤␥ in cells
A possible interpretation for the results obtained in the trypsin protection experiments described above is that G␣ q Q209P adopts a conformation analogous to that of GDP-bound G␣ q . This is unlikely because G␣ q Q209P binds to effectors better than inactive G␣ q (Fig. 1) and can also trigger signaling in cells (Fig. 2), which is largely incompatible with the properties of GDP-bound G␣. Moreover, G␣ q Q209P did not become trypsin-resistant upon the addition of Mg⅐AlF 4 Ϫ (Fig. S4), suggesting that the trypsin sensitivity is not because the G protein is in a GDP-bound state. Instead, it is more likely that the SwII of G␣ q Q209P adopts a unique conformation that represents neither the active nor the inactive conformation of G␣ q WT. To further substantiate this idea, we investigated the association of G␣ q Q209P with G␤␥. G␤␥ binds preferentially to G␣ subunits in the GDP-bound state by making extensive contacts with the SwII (Fig. 5A). For this, the SwII must adopt an inactive conformation, as formation of the characteristic helical conformation of GTP-bound G␣ leads to G␤␥ dissociation. Thus, G␤␥ bind-

Atypical properties of GNAQ Q209P
ing is sensitive to the conformational status of the SwII region of G␣. To investigate the association between G␣ q and G␤␥ in cells, we used a previously described BRET-based assay (51,52). Briefly, a fusion of the C-terminal domain of GRK3 fused to nanoluciferase (GRK3ct-Nluc) is used as the BRET donor, and G␤␥ fused to Venus (V-G␤␥) serves as the BRET acceptor. Because GRK3ct and G␣ q share overlapping binding sites on G␤␥ and can bind only in a mutually exclusive manner, BRET due to V-G␤␥ binding to GRK3ct-Nluc reflects the dissociation of G␣-G␤␥ heterotrimers (Fig. 5B, top).
Consistent with G␣-G␤␥ heterotrimer dissociation upon G protein activation and previous observations using this assay system (51-53), agonist-induced activation of the G q -coupled GPCR M3 muscarinic acetylcholine receptor (M3R) led to a rapid increase in BRET in HEK293T cells co-expressing the donor/acceptor pair and G␣ q WT. The subsequent addition of atropine, an antagonist of the M3 muscarinic receptor, caused the rapid decline of the BRET signal to basal levels, consistent with reassociation of G␣ q and V-G␤␥ (Fig. 5B). When the same experiment was done in cells expressing G␣ q Q209L or G␣ q Q209P, the basal levels of BRET before agonist stimulation were as high as that observed for agonist-stimulated BRET in cells expressing G␣ q WT (Fig. 5B), indicating that G␣ q Q209L and G␣ q Q209P are constitutively dissociated from G␤␥. Consistent with constitutive dissociation of G␣ and G␤␥, GPCR stimulation of cells expressing G␣ q Q209L and G␣ q Q209P failed to elicit a BRET response (Fig. 5B). The amount of G␣ q WT, G␣ q Q209L, and G␣ q Q209P in the cells used in these experiments was equal, ruling out the possibility that the BRET differences observed between G␣ q WT and the mutants were due to differential expression (Fig. 5B). There are two important points to be drawn from these observations. One is that the SwII of G␣ q Q209P adopts a unique conformation different from that of inactive G␣ q WT, as it cannot bind to G␤␥ (Fig.   5B), and from that of other active G␣ proteins, as it does not become trypsin-resistant (Fig. 4). The second point is that G␣ q Q209P exists preferentially in a monomeric, G␤␥-free form in cells.

Binding of RGS proteins to G␣ q Q209P is weaker than to G␣ q Q209L
We reasoned that binding of G␣ q Q209P to RGS proteins might be impaired because they also utilize the SwII as an important contact point. In fact, crystal structures of RGS8 and RGS2 in complex with G␣ q (54,55) have revealed that, despite adopting different poses, both RGS proteins make contact with Gln-209 and several adjacent residues in the SwII (Fig. 5, C and  D). Consistent with our expectation, we found that both GAIP (also known as RGS19) and RGS2, representing the two G protein binding poses of RGS proteins (54), bound to G␣ q Q209P less than to G␣ q Q209L or G␣ q WT activated with Mg⅐AlF 4 Ϫ , but slightly more than inactive G␣ q WT (Fig. 5, C and D). Once again, adding Mg⅐AlF 4 Ϫ did not rescue the defect in RGS binding of G␣ q Q209P (Fig. S5), suggesting that the diminished binding is not because the G protein is in a GDP-bound state. These results indicate that G␣ q Q209P does not bind to RGS proteins, a family of negative regulators of G q signaling.

Mechanistic model and discussion
Our findings using in vitro and cell-based reconstitution systems reveal that the oncogenic G␣ q Q209P mutant is an active G protein with atypical properties. Although we can only speculate about the mechanism by which this mutant leads to signaling hyperactivation in the context of uveal melanoma, our findings provide a framework to try to explain how two mutants with different biochemical properties, G␣ q Q209L and G␣ q Q209P, result in similar oncogenicity (Fig. 6). Both G␣ q Q209L and G␣ q Q209P are GTPase-deficient and trigger constitutive Ϫ . Right panels, surface view of the structure of G␣ q with the p63RhoGEF or PLC␤3 binding regions in cyan or purple, respectively, which overlap with the SwII. Red circle, position of Gln-209. B, diagram of limited proteolysis assay used to assess the conformational status of the SwII. The SwII becomes trypsin-resistant when it adopts a helical conformation, such as in active G␣ q WT, which results in the formation of a trypsin-resistant fragment of G␣ in which only an N-terminal fragment of the protein has been cleaved off by trypsin. Mock gel depicts potential outcomes, depending on SwII conformation. C, the SwII of G␣ q Q209L but not of G␣ q Q209P adopts a conformation equivalent to that of active G␣ q WT (ϩMg⅐AlF 4 Ϫ ) as determined by limited proteolysis with trypsin. Lysates of HEK293T cells transfected with HA-G␣ q (WT or mutants, as indicated) were incubated with trypsin in the presence or absence of 10 mM MgCl 2 , 30 M AlCl 3 , and 10 mM NaF (Mg⅐AlF 4 Ϫ ) as indicated. The four lanes on the left are controls not treated with trypsin. The arrowhead indicates the position of full-length G␣ q , and the asterisk indicates the position of the N-terminally cleaved trypsin-resistant fragment of active G␣ q . One representative result of at least three independent experiments is shown.

Figure 5. G␣ q Q209P is constitutively dissociated from G␤␥ in cells and binds weakly to RGS proteins.
A, G␤␥ binds to the SwII region of G␣ q . Top, surface view of the structure of G␣ q with the SwII in green. Bottom, surface view of the structure of G␣ q with the G␤␥ binding area in blue. Red circle, position of Gln-209. B, G␣ q Q209L and G␣ q Q209P are both constitutively dissociated from G␤␥, as observed for GPCR-activated G␣ q WT. Top, schematic of the assay used to monitor G protein activity with a BRET reporter of free G␤␥. Under resting conditions, Venus-tagged G␤␥ (V-G␤␥, BRET acceptor) associates with inactive G␣ q , and BRET signal is low. Upon GPCR stimulation, V-G␤␥ dissociates from active G␣ and interacts with the C-terminal domain of GRK3 fused to nanoluciferase (GRK3ct-Nluc, BRET donor) inducing a BRET increase. Bottom, G␤␥ dissociates from G␣ q WT upon GPCR stimulation, whereas G␣ q Q209L and G␣ q Q209P fail to associate efficiently with G␤␥ under resting conditions, as indicated by the high BRET before agonist stimulation, and fail to undergo further dissociation upon GPCR stimulation. HEK293T cells transfected with plasmids encoding M3R, Venus(1-155)-G␥ 2 (VN-G␥ 2 ) and Venus(155-239)-G␤ 1 (VC-G␤ 1 ), HA-G␣ q (WT or mutants), and GRK3ct-Nluc were treated with carbachol (100 M) and atropine (100 M) as indicated. Twenty-four h after transfection, cells were harvested, and BRET was measured every 0.24 s. Results are mean Ϯ S.E. (error bars) (shown only at 5-s intervals for clarity) of n ϭ 3. Immunoblots of lysates of HEK293T cells, used to confirm the equal expression of G␣ q proteins, are shown below the graph. C and D, G␣ q Q209P binds to GAIP (C) or RGS2 (D) less than G␣ q Q209L or active G␣ q WT (ϩMg⅐AlF 4 Ϫ ). Left panels, surface view of the structure of G␣ q with the RGS8 (C) or RGS2 (D) binding area in yellow or purple, respectively. Red circle, position of Gln-209. Right panels, lysates of HEK293T cells transfected with HA-G␣ q (WT or mutants, as indicated) were incubated with GST-GAIP (C) or GST RGS2 (D) immobilized on GSH-agarose beads in the presence or absence of 10 mM MgCl 2 , 30 M AlCl 3 , and 10 mM NaF (Mg⅐AlF 4 Ϫ ) as indicated. Resin-bound proteins (Pulldown) and aliquots of the lysates were analyzed by Ponceau S staining and immunoblotting (IB) as indicated. One representative result of three independent experiments is shown.

Atypical properties of GNAQ Q209P
activation of downstream signaling, which explains well their ability to promote cell phenotypic traits that support oncogenic transformation. The fact that the frequency of Q209P mutations in uveal melanoma tumors is similar to that of Q209L mutations supports the idea that both of them provide a similar adaptive advantage during cancer development through abnormal signaling hyperactivation. This is in good agreement with previous observations indicating that both G␣ q Q209L-and G␣ q Q209P-dependent signaling in UM-derived cell lines are required for cell proliferation and/or tumor growth in mice (24, 25, 36 -38). The main difference between G␣ q Q209P and other active G␣ q species (i.e. G␣ q Q209L and GTP-bound G␣ q WT) is that it has unique structural properties that impact its ability to bind different interacting partners. More specifically, our results show that the SwII of G␣ q Q209P adopts a conformation different from that of G␣ q Q209L or G␣ q WT that results in diminished binding to effectors as well as to G␤␥ or RGS proteins, the two main protein classes that function as negative regulators of G␣ q activity in cells. It is unlikely that the prooncogenic properties of G␣ q Q209P arise from engagement to an effector different from those investigated in Fig. 1 for two reasons. One is that the effectors investigated in Fig. 1 represent not only the most widely and best characterized direct effectors of G␣ q , but also the ones that lead to pro-oncogenic signaling in UM cell lines (24,25,36,37). The second one is that the mode of engagement of effectors on G␣ subunits is highly conserved, invariably involving the ␣3 helix/SwII groove (43,56). This is consistent with our finding that five different effectors engage similarly with G␣ q Q209P (Fig. 1), making it very unlikely that another putative effector would engage G␣ q Q209P in a completely different manner.
Then, how can G␣ q Q209P lead to hyperactive signaling and oncogenic transformation if its binding to effectors is diminished compared with GTP-bound G␣ q WT? First, G␣ q Q209P binding to effectors is stronger than that of inactive G␣ q WT, so it still retains some ability to engage and activate effectors. Second, G␣ q Q209P is constitutively dissociated from G␤␥ and has diminished binding to RGS proteins, both of which are negative regulators of G␣ q in cells. A frequently overlooked function of G␤␥ is to prevent the spurious action of G␣ on its effectors by competitive binding (57,58). The importance of this function is further highlighted by evidence that G␤␥ can bind to GTPbound G␣ q at physiological concentrations and antagonize binding of effectors such as PLC␤ (48,59). The role of RGS proteins as negative regulators of G protein signaling is welldocumented. This function is not only mediated by their GAP activity, to which GTPase-deficient Gln-209 mutants are presumably insensitive (60), but also by their ability to antagonize the binding of effectors to G proteins. In fact, it has been reported that RGS proteins are potent inhibitors of signaling activation by the G␣ q Q209L oncogenic mutant (61,62). Thus, the moderately higher affinity of G␣ q Q209P for effectors compared with inactive G␣ q might work concurrently with the relief from G␤␥-and RGS-mediated antagonism to attain effective levels of downstream signaling (Fig. 6).
This model shares similarities with the recently proposed mode of action for another G␣ oncogenic mutant (i.e. G␣ s R201C). Hu and Shokat (63) have recently reported that this mutant might exist in a GDP-bound yet active state that accounts for its pro-oncogenic properties in cells. In the presence of G␤␥, only GDP-bound G␣ s R201C and not GDP-bound G␣ s WT can activate a downstream effector (i.e. adenylyl cyclase) (63). Thus, although GDP-bound G␣ s R201C is a weaker activator of adenylyl cyclase than GTP-bound G␣ s , it leads to efficient signaling due to its inability to bind G␤␥. We propose that insensitivity to G␤␥ antagonism might contribute similarly to enhancing G␣ q Q209P-dependent signaling.
In summary, our findings reveal a fundamental difference in the molecular properties of G␣ q Q209P compared with other active G␣ q proteins, including the other most frequent G␣ q mutation in uveal melanoma, Q209L. The unique structural features of G␣ q Q209P could be leveraged as a specific vulnerability of uveal melanomas bearing this mutation, which account for ϳ20% (11)(12)(13), and to overcome current limitations for the treatment of this type of cancer. To date, targeting signaling pathways downstream of G q in uveal melanoma has not been effective for therapeutic purposes (63). On the other hand, although it is logical to think that targeting mutant G␣ q directly might be more efficacious than targeting downstream signaling nodes, it has been an area underexplored due to the assumption that it could lead to concurrent blockade of G␣ q WT function and subsequent undesired side effects. Thus, the safety of cell-permeable peptide-like inhibitors of G q recently shown to effectively blunt proliferation of uveal melanoma cells bearing activating mutations in G␣ q remains to be established (64). However, we show that contrary to other mutants like G␣ q Q209L, G␣ q Q209P could be specifically inhibited without affecting G␣ q WT if targeting the unique structural features of the mutant proves to be feasible. Based on our results, it is likely Figure 6. Working model for the mechanisms leading to the constitutive signaling activity of G␣ q Q209L and G␣ q Q209P in cancer. In the absence of GPCR stimulation, the low signaling activity of G␣ q WT (top) is ensured by two mechanisms: 1) low affinity for effectors (blue arrow) and 2) the action of negative regulators, such as G␤␥ and RGS protein, which diminishes coupling to effectors. In G␣ q Q209L (middle), the SwII (dark green) adopts an active conformation equivalent to that of GTP-bound G␣ q that causes 1) high-affinity binding to effectors (red arrow) and 2) dissociation from G␤␥, which lead to constitutive signaling activity. Nevertheless, this mutant still binds with high affinity to RGS proteins, which are potent inhibitors of effector binding. In G␣ q Q209P (bottom), the SwII (dark green) adopts a conformation different from that of G␣ q Q209L or GTP-bound G␣ q that causes 1) moderate affinity binding to effectors (orange arrow), 2) dissociation from G␤␥, and 3) impaired binding to RGS proteins. We speculate that the moderate affinity binding to effectors combined with the loss of G␤␥-mediated blockade of the effector-binding site and the lack of RGS-mediated antagonism for effector binding might be sufficient to account for the signaling hyperactivation reported in cells bearing this mutation (11,24,25,37).

Atypical properties of GNAQ Q209P
that the ␣3 helix/SwII groove of G␣ q in the Q209P mutant is different than in other active G␣ q species. Identifying molecules that specifically bind to this putative pocket would disrupt the binding and activation of effectors downstream of the Q209P mutant. Although disruption of protein-protein interactions is challenging, there are now dozens of different protein-protein interactions that have been targeted by small molecules, and many of them have shown promising therapeutic effects, even entering clinical trials (65)(66)(67)(68)(69). An atomic resolution structure of G␣ q Q209P might provide additional information on the molecular basis for its unique properties while establishing a framework to explore its druggability.

Reagents and antibodies
Unless otherwise indicated, all chemical reagents were obtained from Sigma or Fisher Scientific. E. coli DH5␣ strain was purchased from New England Biolabs, and the BL21(DE3) strain was purchased from Life Technologies. PfuUltra DNA polymerase was purchased from Agilent. Carbachol (catalog no. AC-10824) was obtained from Acros Organics, and atropine (catalog no. A10236) was from Alfa Aesar. Leupeptin (catalog no. L-010), pepstatin A (catalog no. P-020), and aprotinin (catalog no. A-655) were from Gold Biotechnology. [␥-32 P]GTP was from PerkinElmer Life Sciences. Mouse mAb raised against hemagglutinin (HA) tag (clone 12CA5) was obtained from Roche Applied Science. Mouse monoclonal antibodies raised against ␣-tubulin (T6074) and FLAG tag (F1804) were from Sigma. Rabbit polyclonal antibody raised against nanoluciferase (Nluc) was kindly provided by Dr. Lance Encell (Promega). Rabbit polyclonal antibodies raised against G␣ q (E-17) were purchased from Santa Cruz Biotechnology, Inc. Rabbit antibodies for phospho-ERK1/2 (Thr-202/Tyr-204) (catalog no. 4370) and total ERK1/2 (catalog no. 9102) were obtained from Cell Signaling. Goat anti-rabbit Alexa Fluor 680 and goat anti-mouse or IRDye 800 secondary antibodies were from Life Technologies and LI-COR, respectively.
His-G␣ q * was purified by adapting a protocol described by Waldo et al. (43). His-G␣ q * expression in BL21(DE3) E. coli was induced overnight with 1 mM isopropyl-␤-D-1-thio-galactopyranoside at 23°C. Bacteria were pelleted, lysed by sonication, and cleared from insoluble material as described above but using a different lysis buffer (20 mM HEPES, pH 8.0, 300 mM NaCl, 5 mM MgCl 2 , 10 mM ␤-mercaptoethanol, 15 mM imidazole, 10% glycerol, 50 M GDP, 30 M AlCl 3 , and 10 mM NaF supplemented with protease inhibitor mixture (1 M leupeptin, 2.5 M pepstatin, 0.2 M aprotinin, 1 mM phenylmethylsulfonyl fluoride)). The soluble fraction of the lysate was incubated with HisPur cobalt resins (Pierce, catalog no. 89964) for 90 -120 min at 4°C with rotation, followed by four cycles of washes with ϳ20 resin volumes of lysis buffer and centrifugation (2,000 ϫ g, 2 min), and eluted with lysis buffer supplemented with imidazole to obtain a final concentration of 500 mM. The eluate was supplemented with 100 mM EDTA and incubated on ice for 20

Atypical properties of GNAQ Q209P
min before passing it over a Superdex 200 HR10/30 gel filtration column using a running buffer consisting of 20 mM HEPES, pH 8.0, 100 mM NaCl, 1 mM MgCl 2 , 2 mM DTT, 2% glycerol, and 10 M GDP. In some cases, proteins were concentrated using an Amicon centrifugal filter with a molecular mass cut-off of 10 kDa. The typical yield was 0.5 mg of protein/liter of culture. Purified His-G␣ q * was aliquoted and stored at Ϫ80°C.
For pulldowns of purified His-G␣ q * (Fig. S1), the following GST-fused proteins were immobilized on GSH-agarose beads by incubation at room temperature for 90 min in PBS (amounts indicated in parenthesis): GST-GRK2 RH (10 g), GST-GAIP (10 g), GST-p63RhoGEF DH/PHext (10 g), and GST-Ric-8A Ϫ condition) as indicated. Resin-bound proteins were eluted with Laemmli sample buffer by incubation at 37°C for 10 min. Proteins were separated by SDS-PAGE, transferred to PVDF membranes, stained with Ponceau S, and immunoblotted with the corresponding antibodies.

Atypical properties of GNAQ Q209P
PerkinElmer Life Sciences) and mixed with the nanoluciferase substrate Nano-Glo (Promega, final dilution 1:200) for 2 min before measuring. A POLARstar OMEGA plate reader (BMG Labtech) was used to measure luminescence signals at 460 Ϯ 20 and 528 Ϯ 10 nm at 28°C, and BRET signals were calculated as the ratio between the emission intensity at 528 Ϯ 10 nm over the emission intensity at 460 Ϯ 20 nm. For the kinetic experiments, BRET measurements were performed every 0.24 s. An aliquot of cells from each experiment was processed for subsequent immunoblot analysis as follows. Cells were centrifuged (1 min at 14,000 ϫ g) and resuspended on ice with lysis buffer (20 mM Hepes, pH 7.2, 5 mM Mg(CH 3 COO) 2 , 125 mM K(CH 3 COO), 0.4% (v/v) Triton X-100, 1 mM DTT, 10 mM ␤-glycerophosphate, and 0.5 mM Na 3 VO 4 supplemented with a protease inhibitor mixture (SigmaFAST, catalog no. S8830)). Lysates were cleared by centrifugation (14,000 ϫ g, 10 min, 4°C) and boiled for 5 min in Laemmli sample buffer before protein separation by SDS-PAGE and immunoblotting.

SRE reporter activation
These experiments were performed using a luciferase reporter assay as described by Lutz et al. (42) with minor modifications. HEK293T cells were seeded on 6-well plates (ϳ350,000 cells/well) coated with gelatin and after 1 day were transfected using the calcium phosphate method with plasmids encoding G␣ q -HA (WT or mutants, 1 g/well in Fig. 2 or 0.1 g/well in Fig. S2) or with an empty vector, along with the luciferase reporter plasmids pGL3-SRE.L and pRL-TK (0.5 g of each per well). Cell medium was changed 6 h after transfection by medium containing a reduced concentration of FBS (0.5% (v/v) final). Approximately 16 -24 h after transfection, cells were harvested for measurement of firefly and Renilla luciferase activity using the Dual-Glo Luciferase Assay System (Promega, catalog no. E2920). Approximately one-sixth of the transfected cells were used for each 96-well condition in this assay using the Passive Lysis Buffer (Promega, catalog no. E1941). SRE activity-related counts (firefly) were normalized by the counts obtained for the Renilla luciferase under the control of a constitutive promoter, and then results were expressed as -fold activation compared with control cells transfected with an empty plasmid instead of G␣ q .

Limited proteolysis
HEK293T cells were seeded on 100-mm dishes (ϳ2 ϫ 10 6 cells/dish) coated with gelatin and after 1 day were transfected using the calcium phosphate method with plasmids encoding G␣ q -HA WT or mutants (3 g/dish). Cell medium was changed 6 h after transfection. Approximately 16 -24 h after transfection, cells were scraped in PBS, centrifuged (5 min at 550 ϫ g) and resuspended on ice with 2 ml of lysis buffer (20 mM Hepes, pH 7.2, 5 mM Mg(CH 3 COO) 2 , 125 mM K(CH 3 COO), 0.4% (v/v) Triton X-100, 30 M GDP, 1 mM DTT, 10 mM ␤-glycerophosphate, and 0.5 mM Na 3 VO 4 supplemented with a protease inhibitor mixture (SigmaFAST, catalog no. S8830)). For some conditions, the lysis buffer was supplemented with 30 M AlCl 3 , 10 mM NaF, and 10 mM MgCl 2 (Mg⅐AlF 4 Ϫ condition). Ten l of the cleared lysate (14,000 ϫ g, 10 min) were used for each condition. Digestions were started by the addition of 100 ng of trypsin (or an equivalent volume of PBS) to each tube and incubated for 20 min at 30°C. Reactions were stopped by the addition of Laemmli sample buffer and boiling followed by protein separation by SDS-PAGE and immunoblotting.

Statistical analyses
Each experiment was performed at least three times unless otherwise indicated. The data shown are presented as means with error bars representing the S.E. or as one representative

Atypical properties of GNAQ Q209P
result of each biological replicate (as indicated in the figure legends). Statistical significance between various conditions was assessed with Student's t test. p Ͻ 0.05 was considered significant.