Snapshot of Activated G Proteins at the Membrane: The Gαq-GRK2-Gßγ Complex

See allHide authors and affiliations

Science  09 Dec 2005:
Vol. 310, Issue 5754, pp. 1686-1690
DOI: 10.1126/science.1118890


G protein–coupled receptor kinase 2 (GRK2) plays a key role in the desensitization of G protein–coupled receptor signaling by phosphorylating activated heptahelical receptors and by sequestering heterotrimeric G proteins. We report the atomic structure of GRK2 in complex with Gαq and Gβγ, in which the activated Gα subunit of Gq is fully dissociated from Gβγ and dramatically reoriented from its position in the inactive Gαβγ heterotrimer. Gαq forms an effector-like interaction with the GRK2 regulator of G protein signaling (RGS) homology domain that is distinct from and does not overlap with that used to bind RGS proteins such as RGS4.

G protein–coupled receptors (GPCRs) are involved in a vast array of physiological processes, and the molecular basis for how signals are passed from activated receptors, through heterotrimeric G proteins (Gαβγ), and then to downstream effectors has been the subject of intense investigation (1, 2). Crystal structures of inactive rhodopsin (3, 4) and the Gαβγ heterotrimer (5, 6) have been determined, as have structures of activated Gα and Gβγ subunits bound to various effector targets (710). These atomic models provide the first and last frames, respectively, of a molecular signaling movie that describes the course of heterotrimeric G protein signaling. The three switch regions of the Gα subunit play key roles, changing conformation depending on whether guanosine diphosphate (GDP) or guanosine triphosphate (GTP) is bound. In the Gαβγ heterotrimer, Gα is bound to GDP, and switch II is sequestered by Gβγ (Fig. 1, A and B). On activation of Gα, GTP is bound; switch II dissociates from Gβγ; and switches I, II, and III adopt a conformation appropriate for binding effectors and RGS proteins (7, 11, 12). The events that occur between the first and last frames of this molecular signaling movie are not well understood. Although receptor recognition of Gαβγ appears to be mediated primarily by the C-terminal region of the Gα subunit (13, 14), fundamental issues remain unresolved, including how activated GPCRs manipulate Gαβγ to mediate nucleotide exchange on Gα (15, 16); whether Gα, Gβγ, and GPCRs remain associated after activation (1719); and how G protein subunits and their effector complexes are arranged at the membrane during signal transduction.

Fig. 1.

Comparison of the inactive Gαβγ heterotrimer and the Gαi/q-GRK2-Gβγ complex. (A) Side view of Gαqβγ. Gαqβγ was homology modeled by using the structure of Gαiβ1γ2 (5). The expected membrane surface is modeled as a gray rectangle that extends out from the plane of the figure (31), and the heterotrimer is oriented as proposed in (6). Gαq is cyan with orange β-strands, Gβ is blue, and Gγ is green. The three switch regions (labeled I, II, and III) and the N-terminal helix of Gαq are red and yellow, respectively. GDP and Gαq-Cys9 and Cys10, which can be palmitoylated, are shown as ball-and-stick models. (B) Top view of Gαqβγ from the perspective of the modeled membrane surface. (C) Side view of the Gαi/q-GRK2-Gβγ complex. For purposes of comparison, GRK2-bound Gβγ was centered in the same position as Gβγ in panel (A). The chimeric N-terminal helix of GRK2-bound Gαi/q is disordered in the crystal structure. The kinase domain of GRK2 is yellow with olive β strands, the RH domain is purple, and the PH domain is tan. Mg2+ (black sphere) and AIF4 (green and magenta) are bound in the active site of Gαi/q. (D) Top view of the Gαi/q-GRK2-Gβγ complex from the same orientation as (B). Residues 114 to 121 in α5 of GRK2 (shaded pink) alter their conformation upon docking with the effector-binding pocket of Gαi/q (see SOM text).

G protein–coupled receptor kinase 2 (GRK2) initiates phosphorylation-dependent desensitization of GPCRs (20, 21) by phosphorylating the C-terminal tail or third intracellular loop of activated GPCRs (22). GRK2 also can inhibit GPCR signaling via phosphorylation-independent mechanisms (23, 24), including sequestration of Gαq/11/14 subunits with its RGS homology (RH) domain (2528) and Gβγ with its pleckstrin homology (PH) domain (29, 30). The crystallographic structure of GRK2 in complex with Gβγ suggested that the arrangement of its kinase, RH, and PH domains is compatible with the simultaneous recognition of activated receptor, Gαq, and Gβγ, respectively (9). The structure of a Gαq-GRK2-Gβγ complex should therefore reveal the configuration of Gα and Gβγ subunits as they engage a single protein target and provides another snapshot of the events that unfold after GPCR activation.

q was overexpressed in insect cells as a soluble chimera (henceforth referred to as Gαi/q) in which the wild-type N-terminal helix was replaced with that of Gαi1 (31, 32). Gαi/q bound GRK2 in an Math manner (fig. S1), and the resulting Math complex could be crystallized in the presence of a soluble mutant of Gβ1γ2 (fig. S2). The resulting Gαi/q-GRK2-Gβγ complex was solved by molecular replacement with the use of x-ray diffraction data extending to 3.1 and 4.5 Å spacings in the best and worst reciprocal lattice directions, respectively (table S1) (31).

In the Gαi/q-GRK2-Gβγ complex, GRK2 serves as a scaffold for the activated heterotrimeric G proteins, with Math bound to the RH domain and Gβγ bound to the PH domain (Fig. 1, C and D). The switch regions of Gαi/q adopt a conformation typical of other activated Gα subunits (fig. S3), and Gαi/q-bound GRK2-Gβγ differs only subtly from GRK2-Gβγ alone (see supporting online text). The Gα subunit, however, undergoes a dramatic ∼105° rotation from its position in the Gαβγ heterotrimer to engage GRK2 (Figs. 1 and 2; movies S1 and S2). In doing so, the regions of Gα believed to be adjacent to the membrane in Gαβγ (i.e., the N and C termini) are rotated away, such that switch I, switch II, linker 1, and the αB-αC loop are closest to the predicted membrane surface, although ∼30 Å removed (Fig. 2). It is not clear whether this reorientation of Gαi/q is GRK2-specific or if it could also represent the position of other activated, effector-bound Gα subunits at the membrane. We note that whereas structures of Gαt in complex with phosphodiesterase-γ (PDEγ) and Gα13 in complex with p115-Rho guanine nucleotide exchange factor (p115RhoGEF) are compatible with the predicted membrane surface when superimposed on GRK2-bound Gαi/q, the Gαs–adenylyl cyclase complex is not (7, 8, 10). Gβγ also undergoes an apparent ∼22° rotation from its position in the Gαβγ heterotrimer (compare Fig. 1, A and C), which was also evident in the GRK2-Gβγ structure (9).

Fig. 2.

Changes in the orientation of Gαq on activation and binding of GRK2. The model of Gβγ-bound Gαq·GDP and the structure of GRK2-bound Embedded Image are viewed from a direction roughly 90° around a vertical axis from those of Fig. 1, A and C, respectively. The Gα subunits were positioned by translationally centering the Gβγ subunits of their respective complexes along the plane of the modeled membrane (Fig. 1). On binding GRK2, Gαi/q rotates by ∼105° such that Gly188 in switch I of GRK2-bound Gαi/q becomes the closest residue to the modeled membrane surface (∼25 Å below). The most N-terminal residue observed in GRK2-bound Gαi/q, Arg38, which is expected to be adjacent to the membrane in Gαβγ, is displaced by ∼30 Å from the membrane. However, the native N-terminal helix of Gαq is sufficiently long (37 residues, ∼55 Å long) to allow the palmitoylation sites at Cys9 and Cys10 to be adjacent to the membrane. If one assumes that Gα and Gβγ derive from a single heterotrimer, Gαq-Arg38 also translates ∼80 Å away from its position in the Gαβγ heterotrimer (Fig. 2), and the fully extended wild-type N-terminal helix of Gαq would fall short of contacting Gβγ. Therefore, activated Gαq dissociates partially, if not completely, from the membrane and entirely from Gβγ, at least when in complex with GRK2.

In Gαβγ, the N terminus of Gα forms a single, extended α helix that interacts with Gβ and, presumably, with the membrane via basic residues and/or lipid modifications (Fig. 1, A and B) (5, 6, 33). Interpreting the role of this helix in our structure is problematic because it is both chimeric and disordered. However, the first observed residue of Gαi/q, corresponding to Gαq-Arg38, is sufficiently removed from the predicted membrane surface (∼30 Å) and from its position in the Gαβγ heterotrimer (∼80 Å) to suggest that the N-terminal helix is at least partially dissociated from the membrane and completely dissociated from Gβγ (Fig. 2).

The Gαi/q-GRK2 interface buries ∼1700 Å2 of accessible surface area and involves α2 (switch II), α3, and the α3-β5 loop of Gαi/q, as well as the α5 and α6 helices of the GRK2 RH domain (Figs. 1D and 3A). Within the interface, hydrogen bonds are formed between the hydroxyl of Gαq-Tyr261 and the side chains of GRK2-Asp110 and -Arg106, as well as between the side chains of Gαq-Thr260 and GRK2-Gln133. The primary nonpolar interactions are made by the side chains of GRK2-Met114, Leu117, Leu118, and Cys120, which dock into a cleft formed between the α2 (switch II) and α3 helices of Gαi/q. The residues of GRK2 that form the interface with Gαi/q are essentially the same as those identified in previous studies, wherein mutation of Asp110, Arg106, and Leu118 of the GRK2 RH domain eliminated Gαq binding (34, 35). Furthermore, the D110A mutation in GRK2 abrogates its ability to mediate phosphorylation-independent desensitization in vivo (36, 37). The GRK2-binding residues of Gαi/q are analogous to those in Gαs and Gαt that bind adenylyl cyclase and PDEγ, respectively (7, 8), and are among those previously implicated in the binding of phospholipase C–β (PLC-β) (38, 39). Thus, the GRK2 RH domain binds Gαi/q more like an effector than an RGS protein (8, 12) (Fig. 3B), a result that is consistent with the facts that GRK2 efficiently binds Gαq·GTPγS and does not exhibit significant guanosine triphosphatase (GTPase)–activating protein (GAP) activity toward Gαq (25). Residues in switches I and III of Gαq previously implicated in binding GRK2 (35) appear to play only an indirect role, perhaps by altering the structure or dynamics of switch II.

Fig. 3.

The GRK2-binding surface of Gαq. (A) Stereoview of the interface. The switch II and α3 helices from Gαi/q are shown as Cα traces; the α5 and α6 helices from GRK2 are shown as cartoon ribbons. Side chains of interfacial residues are shown as ball-and-stick models, with carbon atoms from Gαi/q and GRK2 colored cyan and yellow, respectively. Hydrogen bonds are shown as dashed black lines. Residues targeted by site-directed mutagenesis in this study are underlined. (B) Sequence alignment of the switch regions and the α3/β5 sequence for representative members of all four Gα subfamilies. Switch regions (I to III) are outlined in black and are assigned on the basis of comparison of the active and deactivated structures of Gαi1. Secondary structure is represented by cylinders and arrows for α helices and β strands, respectively. Gα residues that contact effectors are green, those that bind GAPs are red, and those that contact both are purple. Contacting residues that were chimeric (i.e., nonnative) in the crystal structures of the Gαt and Gα13 effector complexes are shown in a lighter shade of the appropriate color. Green boxes outline Gαi residues proposed to interact with adenylyl cyclase (50), and asterisks indicate conserved residues that contribute to the hydrophobic effector-binding pocket. The crystal structures used for these assignments are those of Gαi/q-GRK2-Gβγ (this study), Gαi-RGS4 [Protein Data Bank (PDB) code 1AGR] (12), Gαt-PDEγ-RGS9 (1FQJ) (8), Gα13-p115RhoGEF (1SHZ) (10), and Gαs-adenylyl cyclase (1AZS) (7). The sequences are those of mouse Gαq (M55412), mouse Gα11 (NP_034431), mouse Gα14 (NP_032163), human Gα16 (M63904), rat Gαi1 (M17527), bovine Gαt (P04695), mouse Gα13 (NP_034433), and bovine Gαs (M13006). (C) Mutational analysis of Gαq residues that directly interact with GRK2. Lysates of HEK293 cells expressing Gαq mutants were subjected to limited trypsin digestion in the presence and absence (shown only for wild type) of Embedded Image and immunoblotted with Gαq-specific antibody (upper left) (31). The I217D mutation could not be protected from trypsin digestion and was judged nonfunctional. All Gαq mutants expressed at a similar level compared with wild-type Gαq (lower left). Pull-down assays were performed by incubating lysates with 40 nM glutathione S-transferase (GST) fusion protein of either GRK2-RH, RGS3 (amino acids 313 to 519), or RGS4 either in the presence or absence of Embedded Image and then detecting bound Gαq with Gαq-specific antibody (right).

The R214A, I217D, T257E, Y261F, and W263D mutants of Gαq (40) were generated to test the importance of these positions for binding GRK2 (Fig. 3C). The Gαq-T257E, Gαq-Y261F, and Gαq-W263D mutants completely abrogated binding, whereas the Gαq-R214A mutant retained its interaction with the GRK2 RH domain, and the I217D mutant was nonfunctional (Fig. 3C). The complete loss of binding caused by the subtle Y261F mutation emphasizes the importance of the hydrogen bonds formed by the hydroxyl of Gαq-Tyr261. Previously, it was shown that the Gαq-I259A/T260A/Y261A mutant stimulates PLC-β similarly to wild type and that the Gαq-R256A/T257A mutant is deficient (39). Therefore, whereas GRK2 and PLC-β bind overlapping regions on Gαq, the residues of Gαq most critical for binding differ.

Next, the Gαq-P262K, R256G, and Y261L mutants were created to test the role of these positions in dictating specificity of Gαq for GRK2 (Fig. 3B). In other Gα subfamilies, the residue equivalent to Gαq-Pro262, which packs between Gαq-Trp263, GRK2-Leu136, and GRK2-Val137 (Fig. 3A), is replaced by either arginine or lysine. As expected, the Gαq-P262K mutation abolished GRK2 binding (Fig. 3C). The Gαq-R256G and Gαq-Y261L mutants represent conversions of these residues to their equivalents in Gα16 (Fig. 3B), which does not bind GRK2 (28). The R256G mutation significantly reduced binding, whereas the Gαq-Y261L substitution eliminated binding (Fig. 3C). Therefore, residues 261 to 263 of Gαq, and their equivalents in Gα11 and Gα14, appear sufficient to dictate the Gα specificity of GRK2.

The Gαq-R214A mutation in switch II completely abolished binding to RGS3 and RGS4, but not to GRK2, emphasizing the importance of Gαq-Arg214 in Gα-RGS protein recognition (8, 12). Strikingly, none of the mutations in Gαq that affected GRK2 binding interfered with the binding of RGS proteins (Fig. 3C). Therefore, Gαq binds the GRK2 RH domain using a surface distinct from that used for binding RGS proteins. Indeed, when RGS4 is modeled in complex with Gαi/q, there is no obvious steric overlap between RGS4 and GRK2 (Fig. 4A), which implies that Gαq could bind two different RH domains at the same time: one as an effector (GRK2) and the other as a GAP (RGS protein). This model also predicts that the α-helical domain of Gαq will form substantial contacts with RGS4 (and presumably RGS2) that are not possible in Gαi or Gαt owing to substitutions in αA and differences in the structure of the αB-αC loop (Fig. 4B). This novel interaction may help dictate the relative specificity of RGS4 and RGS2 for Gαq (41).

Fig. 4.

Comparison of effector and GAP-binding sites among the four Gα subfamilies. (A) Model of RGS4 bound to the Gαi/q-GRK2-Gβγ complex. RGS4 was positioned by superimposing Gαi of the Gαi-RGS4 complex (12) with Gαi/q. The docked RGS4 has no obvious steric overlaps with GRK2. The α4-α5 loop and the N-terminal region of RGS4, which are both believed to interact with the cell membrane (51, 52), are juxtaposed with the membrane surface modeled for the Gαi/q-GRK2-Gβγ complex. The Gαi/q-GRK2-Gβγ complex is shown as a molecular surface with the same colors as in Fig. 1, except that the switch regions of Gαi/q are not highlighted. RGS4 is colored in a spectrum from blue to red from its observed N and C termini (residues 51 and 178, respectively). The side chains of basic residues in its α4-α5 loop believed to interact with phosphatidylinositol 3,4,5-trisphosphate (PIP3) (51), are shown as ball-and-stick models. (B) Structural alignment of Ras-like domains of activated Gα subunits. The region of Gα that encompasses the three switch regions was used for the alignment: Gαq (this study, green), residues 183 to 261; Gαt (PDB code 1TAD, red), residues 174 to 252 (53); Gαs (1AZS, blue), residues 201 to 279 (7); and Gα12 (1ZCA, yellow), residues 203 to 281 (32). The structure of activated Gα12 is used to represent the Gα12/13 family, because the Gα13 protein used in the p115RhoGEF complex is a Gαi1 chimera within the effector-binding region (10). Overall, Gαq is most similar to Gαi and Gαt (root mean square deviation of 1.0 Å for 303 analogous Cα positions). The most structurally heterogeneous regions of Gαq are the β5-α4 and α4-β6 loops in the Ras-like domain and, in the α-helical domain, the αB-αC loop. The distinct structures of the α4-β6 and αB-αC loops may allow for specific recognition of Gα subunits by receptors or guanine-nucleotide exchange inhibitors, respectively (2, 54). In contrast, the tertiary structures of the switch regions, which dictate effector and GAP protein interactions, are well conserved. (C) Footprints of effector and GAP-binding sites on the molecular surface of Gαi/q. Colors are assigned as in Fig. 3B. The yellow asterisk indicates the position of the hydrophobic pocket used by all characterized Gα effectors. As originally proposed on inspection of the Gαs-adenylyl cyclase complex (7, 55), effectors and GAPs have apparently evolved to bind to distinct and generally nonoverlapping regions of the Gα subunit. Although the residues colored purple imply steric overlap, different surfaces of these residues are used to bind effectors and GAPs.

Together with the Gαs–adenylyl cyclase, Gαt-PDEγ-RGS9, and Gα13-p115RhoGEF complexes (7, 8, 10), the Gαi/q-GRK2-Gβγ structure completes a survey of effector complexes representing the four Gα protein subfamilies (Fig. 4B). Comparison of these structures demonstrates that structurally diverse effectors recognize a highly localized region on each Gα subunit in a manner that does not necessarily exclude the binding of GAP domains (Figs. 3B and 4C). In each case, solvent-exposed hydrophobic side chains from the effector dock into a nearly invariant pocket formed between the N termini of the switch II (α2) and α3 helices of Gα (Fig. 3B; Fig. 4, B and C). Additional specificity-determining contacts are made with residues at the C-terminal ends of these helices and within the α2-β4 and α3-β5 loops (Fig. 3B). With the exception of the α3-β5 loop in Gαs, the tertiary structures of the effector interacting regions are well conserved (Fig. 4B), which implies that effector specificity in most Gα subunits is dictated by primary sequence and, at least in some cases, differences in electrostatic potential (fig. S4).

The physiological consequence and/or necessity of GRK2 binding both Gαq and Gβγ is not known, but the nanomolar affinity of these interactions (25, 42, 43) and the expected close proximity of these proteins to each other while associated with the membrane suggest that a Gαq-GRK2-Gβγ complex can form soon after a Gq-coupled receptor is activated. Simultaneous engagement of Gα and Gβγ is a characteristic shared among GRK2 and classic effectors like adenylyl cyclase and PLC-β. This, along with the observed effector-like interaction between GRK2 and Gαq and the fact that heptahelical receptors directly stimulate the kinase activity of GRK2 (22), invokes the question of whether GRK2 can instigate its own signaling cascade. Potential downstream targets include insulin receptor substrate–1 (IRS-1) (44) and the cytoskeletal regulator ezrin (45), which can be phosphorylated by GRK2 in response to activation of Gq-coupled receptors.

An increasing body of evidence suggests that GPCR signaling systems can function as preassembled complexes, which should allow for efficient transmission and desensitization of extracellular signals (19). The Gαi/q-GRK2-Gβγ structure strongly supports this hypothesis, at least in the case of Gq-coupled receptors, with GRK2 harboring at least one additional protein-binding site for an activated receptor. The potential coassembly of this complex with RGS proteins like RGS4 and RGS2 (Fig. 4A) is intriguing in light of their reported association with receptor complexes (4649). Defining the molecular basis for the interactions among GPCRs, RGS proteins, heterotrimeric G proteins, and GRK2 will be the focus of future studies.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S4

Table S1

Movies S1 and S2

References and Notes

View Abstract

Navigate This Article