Report

Structure of Gαq-p63RhoGEF-RhoA Complex Reveals a Pathway for the Activation of RhoA by GPCRs

See allHide authors and affiliations

Science  21 Dec 2007:
Vol. 318, Issue 5858, pp. 1923-1927
DOI: 10.1126/science.1147554

Abstract

The guanine nucleotide exchange factor p63RhoGEF is an effector of the heterotrimeric guanine nucleotide–binding protein (G protein) Gαq and thereby links Gαq-coupled receptors (GPCRs) to the activation of the small-molecular-weight G protein RhoA. We determined the crystal structure of the Gαq-p63RhoGEF-RhoA complex, detailing the interactions of Gαq with the Dbl and pleckstrin homology (DH and PH) domains of p63RhoGEF. These interactions involve the effector-binding site and the C-terminal region of Gαq and appear to relieve autoinhibition of the catalytic DH domain by the PH domain. Trio, Duet, and p63RhoGEF are shown to constitute a family of Gαq effectors that appear to activate RhoA both in vitro and in intact cells. We propose that this structure represents the crux of an ancient signal transduction pathway that is expected to be important in an array of physiological processes.

Rho guanine nucleotide triphosphatases (GTPases) are peripheral membrane proteins that regulate essential cellular processes, including cell migration, proliferation, and contraction. RhoA, Rac1, and Cdc42 are the best-characterized members of this family, and they control the dynamics of the actin cytoskeleton and also stimulate gene transcription through several transcription factors, such as the serum response factor (SRF) or nuclear factor κB (1, 2). All Rho GTPases cycle between an inactive guanosine diphosphate (GDP)–bound and an active guanosine triphosphate (GTP)–bound state, a process accelerated by a large family of Rho guanine nucleotide exchange factors (Rho GEFs). Most Rho GEFs contain a catalytic Dbl homology (DH) domain that is immediately followed by a pleckstrin homology (PH) domain (3, 4). In some Rho GEFs, the PH domain appears to autoinhibit the intrinsic GEF activity of the DH domain (57), a constraint that is presumably released upon interaction of the PH domain with membranes or other proteins. One such autoinhibited Rho GEF is p63RhoGEF (810), a RhoA-specific enzyme predominantly expressed in the heart and brain (11). p63RhoGEF is directly activated by members of the Gαq subfamily of heterotrimeric guanine nucleotide–binding proteins (G proteins) (9, 10) and thus is regulated by G protein–coupled receptors (GPCRs). In contrast to the Gα13-regulated p115RhoGEF family (12, 13), p63RhoGEF does not contain a regulator of G protein signaling homology (RH) domain (810) to mediate interaction with Gαq.

To define the minimal elements of p63RhoGEF that mediate Gαq activation in vitro, we expressed fragments of p63RhoGEF spanning residues 149 to 477 and 149 to 580 (p63-149-477 and p63-149-580) of the total 580 amino acids in Escherichia coli (14). The p63-149-477 fragment spans a region similar to that observed in the crystal structure of the Dbl's big sister (Dbs) Rho GEF domain (15) (figs. S1 and S2), whereas p63-149-580 spans the entire region previously known to be important for Gαq binding and regulation (9). Both fragments possessed only weak GEF activity and stimulated nucleotide exchange on RhoA at less than 1/20th the rate of the isolated DH domain of p63RhoGEF (p63-149-338) (Fig. 1A and table S1). Of these fragments, only the p63-149-580 fragment could bind AlF4-activated Gαi/q (Fig. 1B, fig. S4, and table S1), a chimera in which the N-terminal helix of Gαq is replaced with that of Gαi (16) and that activates p63RhoGEF similarly to wild-type Gαq in cells (fig. S3). To precisely define the Gαq-binding core of p63RhoGEF, we tested a series of 72 p63RhoGEF fragments spanning residues 295 to the C terminus for their ability to compete with p63-149-580 for binding Gαi/q. The minimal fragment required for full inhibition corresponded to p63-295-502 (fig. S5). Like p63-149-580, p63-149-502 had low basal nucleotide exchange activity that could be activated by Gαi/q in a saturable manner by up to three- to fourfold (Fig. 1, A and C; fig. S6; and Table 1). These truncations were also functional in human embryonic kidney 293 (HEK293) cells. As with full-length p63RhoGEF (p63-FL), the p63-149-482 and p63-149-502 fragments were immunoprecipitated with the GTP hydrolysis-defective GαqRC mutant (Fig. 1D). The p63-149-482, p63-149-492, and p63-149-502 fragments also mediated GαqRC- and M3 muscarinic acetylcholine receptor (M3-R)–induced activation of RhoA and SRF (Fig. 1, E and F), with p63-149-502 being nearly as effective as wild-type p63RhoGEF.

Fig. 1.

Identification of the minimal fragment of p63RhoGEF regulated by Gαq. (A) Basal activity of p63RhoGEF fragments. Activity was monitored by the increase in fluorescence millipolarization (mP) of a fluorescent GTP analog as it bound RhoA. In this experiment, the p63RhoGEF DH domain was 25 and 35 times more effective in activating 2 μM RhoA than were the p63-149-477 and p63-149-502 fragments, respectively. Fold activation is the p63RhoGEF catalyzed nucleotide exchange rate divided by the intrinsic rate of RhoA. (B) Equilibrium binding of p63RhoGEF fragments to Gαi/q. Dissociation constants (Kd's) were determined by an equilibrium-binding flow cytometry protein interaction assay (FCPIA) (27) using various concentrations of Alexa Fluor 532–labeled p63RhoGEF fragments. In this experiment, Kd was 43 ± 4 and 36 ± 3 nM for p63-149-580 and p63-149-502, respectively. No binding was observed for p63-149-477. Binding was monitored by the median fluorescence intensity (MFI) and was corrected for nonspecific binding (MFI in the absence of AlF4). (C) Stimulation of p63-149-502 by Gαi/q. Activity was measured as in (A), but with added amounts of Gαi/q. For fold activation mediated by Gαi/q, see fig. S6. (D) Binding of p63-149-502 to activated Gαq in cell lysates. HEK293 cells were transfected with 1 μg of GαqRC, 1 μg of the respective p63RhoGEF variant plasmid, and up to 2 μg of an empty control vector. Overexpressed fragments were immunoprecipitated with an antibody to c-Myc. Antibodies to c-Myc and Gαq/11 were used for analysis of precipitated proteins and equal overexpression. (E) Activation of RhoA in cells by p63RhoGEF fragments. HEK293 cells were transfected with 1.3 μg of the respective GEF variant plasmid, 0.6 μg of the GαqRC or M3-R–encoding plasmid, and up to 2 μg of an empty control vector. Cells overexpressing M3-R were stimulated for 3 min with 1 mM carbachol. Endogenous RhoA-GTP was detected through its association with the Rho-binding domain of rhotekin. Total RhoA was detected with an antibody to RhoA. (F) SRF activation by p63RhoGEF fragments. HEK293 cells were transfected with 12.5 ng of GαqRC plasmids, up to 50 ng of p63 construct plasmids, and up to 100 ng of an empty control vector as indicated. For expression of SRF-driven firefly luciferase and constitutively expressed renilla luciferase, 21 ng of pSRE. L and 4 ng of pRL.TK were cotransfected. The given values are means ± SE (n = 12 samples) of firefly/renilla luciferase ratios relative to control-transfected cells.

Table 1.

Properties of site-directed mutations of p63RhoGEF. Ki, inhibition constant; NB, no binding.

View this table:

The molecular basis for the interaction between Gαq and p63RhoGEF was determined by the 3.5 Å crystal structure of p63-149-502 in complex with AlF4-activated Gαi/q and RhoA (Fig. 2, figs. S7 and S8, and table S2). Gαi/q engages both the DH and PH domains of p63RhoGEF, burying over 3600 Å2 of accessible surface area (Fig. 2), using residues that are highly conserved in the Trio family but not in Dbs (fig. S2). No contacts are formed between Gαi/q and RhoA. The PH domain of p63RhoGEF interacts with Gαi/q primarily using a kinked amphipathic helical extension of its PH domain (αCa and αCb helices, residues 471 to 485), which docks into the effector-binding site of Gαi/q (Fig. 3A). In addition, the α2-β4 and α3-β5 loops of Gαi/q interact with residues in the 310 helix and the β2-β3 and β4-β5 loops of the p63RhoGEF PH domain (Fig. 3B). The DH domain of p63RhoGEF interacts with the α3-β5 and α4-β6 loops and the C-terminal α5 helix of Gαi/q on the side of its helical bundle opposite to the RhoA-binding site (Figs. 2 and 3B). The extreme C terminus of Gαi/q packs against a notch formed by a break in the α2 helix of the DH domain, where the side chain of Tyr356 in Gαi/q forms extensive contacts (Fig. 3B). Thus, the C terminus of Gα can mediate interactions with effectors as well as couple to specific GPCRs (17). Overall, the interaction of Gαi/q with the PH domain of p63RhoGEF is quite similar to the interaction of Gαi/q with GRK2 (Fig. 4, A and B) (16), with the most pronounced difference in Gαi/q being the conformation of the α2-β4, α3-β5, and α4-β6 loops, which can mold themselves to form high-affinity interactions with at least two distinct effector surfaces. The conserved manner in which the PH domains of p63RhoGEF and GRK2 engage their activated heterotrimeric G protein targets (Gαi/q and Gβγ, respectively) is likewise striking (18) (Fig. 4, A and C).

Fig. 2.

Crystal structure of the Gαi/q-p63RhoGEF-RhoA complex. (A) Gαi/q interacts with both the DH and PH domains of p63RhoGEF but not with RhoA. The complex is viewed from the perspective of the expected plane of the plasma membrane. N and C denote the most N- and C-terminal residues observed for each domain. Mg2+·GDP·AlF4 is shown as spheres. The three nucleotide-dependent conformational switches of Gαi/q (SwI, SwII, and SwIII) are red. Two residues of the chimeric N terminus of Gαi/q are visible and extend toward the membrane surface, consistent with the N-terminal palmitoylation sites of Gαq engaging the lipid bilayer while it is in complex with p63RhoGEF. (B) Side view of the Gαi/q-p63RhoGEF-RhoA complex. The PH domain is modeled in its expected orientation at the plasma membrane (28), which as a consequence juxtaposes the C-terminal geranylgeranylation site of RhoA with the lipid bilayer.

Fig. 3.

Stereo views of interfaces of the Gαi/q·p63RhoGEF complex. (A) The interaction of the PH domain extension with the effector-binding site of Gαi/q. The p63RhoGEF αCa helix kinks at residue 477 to form a ∼70° elbow, with the subsequent αCb helix packing against the α3 helix of Gαi/q. The nonpolar interactions between residues in αCa and the effector-binding site are the most important for high-affinity binding (Table 1). Residues from the PH domain are violet, and those from Gαi/q are cyan. (B) Interactions of the DH and PH domains with the α-β loops and C terminus of Gαi/q. The α2-β4 loop of Gαi/q appears to stack against an Arg244-Glu385 salt bridge that bridges the DH and PH domains. The side chain of Gαq-Tyr356 at the end of α5 packs against the side of the DH domain, adjacent to p63RhoGEF residues Arg204, Ile205, Gly208, and Asn255.

Fig. 4.

Emerging themes for protein-protein interactions mediated by Gαq and PH domains and a model for p63RhoGEF activation by Gαi/q. (A) The p63RhoGEF PH domain in complex with Gαi/q. Inositol 1,4,5-trisphosphate (IP3) is modeled based on the phospholipase C–γ PH domain·IP3 complex (29) to help define the expected plane of the lipid bilayer. (B) GRK2 binds similarly to the Gαi/q effector-binding site, using exposed hydrophobic residues in its α5 helix. Only the α5 and α6 helices of the GRK2 RH domain are shown. In both the p63RhoGEF and GRK2 complexes, Gαi/q is held in an orientation in which its longest axis is roughly parallel and switch I is held relatively close to the predicted membrane surface (top). In both complexes, the switch I region appears available for the simultaneous binding of regulator of G protein signaling proteins (30). (C) The GRK2 and p63RhoGEF PH domains engage their protein targets in a similar way, using a C-terminal helical extension and the loops at one edge of the β1-β4 sheet of the PH domain to form an extensive protein interaction site (Fig. 3). (D) The DH and PH domains of p63RhoGEF adopt a conformation distinct from that of Dbs (black). The view is the same as in Fig. 2A. The bridging interactions of Gαi/q (spheres) appear to rotate the position of the p63RhoGEF PH domain away from the RhoA binding site on the DH domain, along the plane of the membrane surface.

Site-directed mutants of p63-149-502 were produced to test the role of three regions of the Gαi/q-p63RhoGEF interface: the interface of the Gαi/q effector-binding site with the PH domain extension (Fig. 3A), the interface of the Gαi/q α2-β4/α3-β5 loops with the remainder of the PH domain, and the interface of the Gαi/q α3-β5/α5 region with the DH domain (Fig. 3B). Each mutant was tested for intrinsic GEF activity as well as for its ability to bind and be activated by Gαi/q (Table 1). The alteration of residues in the PH domain extension eliminated Gαi/q binding and activation, consistent with a recent study (10). Mutations in the DH and PH domains of p63RhoGEF exhibited only minimal defects in binding Gαi/q, but some were greatly impaired in Gαi/q activation (Table 1), including Glu385→Ala385 (E385A) (19) and Q386A (in the interface with the α2-β4 and α3-β5 loops of Gαq); W216F, R244A, and R245A (in the interface with the α3-β5 and α4-β6 loops of Gαq); and R204A (in the interface with the α5 C terminus of Gαq).

The in vitro activation of p63RhoGEF thus appears to require the interaction of Gαi/q with both the DH and PH domains (Table 1). We compared our structure of p63-149-502 to those of the related DH and PH domains of Dbs and the N-terminal set of DH and PH domains of Trio (N-Trio), which are not activated by Gαq (20). The DH and PH domains of Dbs and N-Trio interact through conserved residues at their interface and adopt similar conformations in both GTPase-bound and free states (15, 2022). Although residues analogous to these are conserved in p63RhoGEF (fig. S2), the position of the PH domain of Gαi/q-bound p63RhoGEF is rotated by ∼50° around the α6 helix of the DH domain relative to those of Dbs and N-Trio (Fig. 4D). Thus, one way in which Gαq might activate p63RhoGEF is to use its domain-bridging interactions to constrain the otherwise inhibitory PH domain away from the RhoA binding site. In cells, Gαq could regulate p63RhoGEF in additional ways, such as by targeting p63RhoGEF to the plasma membrane, where constraints imposed by the interactions of both proteins with the phospholipid bilayer could in addition optimize the conformation of the DH and PH domains.

We next investigated whether Gαq binds to and stimulates the closely related set of DH and PH domains found in the C terminus of Trio (23) and Duet (24) (figs. S1 and S2). We constructed variants of Trio and Duet analogous to p63-149-502 (Trio-1894-2232 and Duet-219-558) and demonstrated that these proteins bind and are activated by Gαi/q in vitro (Fig. 5, A and B, and table S1). Incells, Trio-1894-2232 and Duet-219-558 co-immunoprecipitated with GαqRC (Fig. 5C), mediated GαqRC-as well as M3-R–induced activation of RhoA (Fig. 5D and fig. S9), and enhanced GαqRC-as well as M3-R– or H1-R–induced activation of SRF, similar to full-length Trio and Duet (Fig. 5E). As in humans (25), a splice variant of the Caenorhabditis elegans Trio ortholog exists that contains only its C-terminal set of RhoA-specific DH and PH domains (fig. S2). This variant, UNC73 E, is also activated by the constitutively active C. elegans ortholog of Gαq and was recently shown to be a critical regulator of neuronal signaling (26). Thus, Trio, Duet, and p63RhoGEF define a previously unknown class of Gαq/11-regulated Rho GEFs in higher vertebrates that diverged from a single ancient trio gene.

Fig. 5.

Regulation of Duet and Trio by Gαq. (A) Activation-dependent binding of Trio family RhoGEFs to Gαi/q measured by FCPIA. In this experiment, Kd = 39 ± 7, 160 ± 23, and 261 ± 61 nM for fluor-labeled p63-149-502, Duet-219-558, and Trio-1894-2232, respectively. (B) In vitro activation of Duet-219-558 and Trio-1894-2232 by Gαi/q. Gαi/q activation is plotted as the fold increase over the nucleotide exchange rate catalyzed by 200 nM GEF. (C) Interaction of Duet-219-558, Duet-FL, Trio-1894-2232, and Trio-FL with Gαq in cells. HEK293 cells were transfected with 1 μg of GαqRC, 1 μg of truncated GEFs, p63-FL or Duet plasmid, and up to 2 μg of an empty control vector, or alternatively with 4 μg of GαqRC, 4 μg of the Trio plasmid, and up to 8 μg of an empty control vector. Overexpressed p63RhoGEF, Duet, and Trio were immunoprecipitated with antibodies to c-Myc, Flag, and enhanced green fluorescent protein (EGFP), respectively. (D) The activation of RhoA in cells by p63-149-502, p63-FL, Duet-219-558, Duet, Trio-1894-2232, and Trio ± GαqRC. RhoA activity in transfected HEK293 cells was determined by binding to Rhotekin (Fig. 1E). Antibodies to c-Myc (p63-FL, truncated GEFs), EGFP (Trio), Flag (Duet), and Gαq/11 were used to analyze expression. (E) The influence of p63-149-502, p63-FL, Duet-219-558, Duet, Trio-1894-2232, and Trio on GαqRC-, or M3-R– or H1-R–induced SRF activation. HEK293 cells were transfected with 30 ng of GαqRC, M3-R, or H1-R plasmids, 60 ng of the GEF plasmids, and up to 90 ng of an empty control vector as indicated. For expression of SRF-driven firefly luciferase and constitutively expressed renilla luciferase, 21 ng of pSRE.L and 4 ng of pRL.TK were cotransfected. Cells overexpressing M3-R and H1-R were stimulated for 24 hours with 1 mM carbachol or 100 μM histamine, respectively. Means ± SE (n = 12 samples) of firefly/renilla luciferase ratios are given.

Our structure of the Gαi/q-p63RhoGEF-RhoA complex reveals three nodes of a signal transduction cascade caught in the act of transferring an extracellular signal from heterotrimeric to small-molecular-weight G proteins. The cornerstone of this complex is the PH domain, which engages the DH domain, Gαi/q, and also probably the membrane using conserved basic residues in its β2 and β4 strands (Fig. 4 and fig. S2). The orientation we predict for Gαq in complex with p63RhoGEF at the membrane appears similar to that predicted for its complex with GRK2-Gβγ (16) (Figs. 2B and 4)–one that is dramatically different from the expected orientation for Gαq in the inactive Gαβγ heterotrimer (17). These results suggest that a conserved signaling complex is assembled at or near activated Gq-coupled receptors, wherein activated Gαq becomes fixed in orientation with respect to the membrane and serves as a GTP-dependent docking site for structurally diverse effector enzymes.

Supporting Online Material

www.sciencemag.org/cgi/content/full/318/5858/1923/DC1

Materials and Methods

Figs. S1 to S9

Tables S1 and S2

References

References and Notes

View Abstract

Navigate This Article