Research Article

Keeping G Proteins at Bay: A Complex Between G Protein-Coupled Receptor Kinase 2 and Gßγ

See allHide authors and affiliations

Science  23 May 2003:
Vol. 300, Issue 5623, pp. 1256-1262
DOI: 10.1126/science.1082348

Abstract

The phosphorylation of heptahelical receptors by heterotrimeric guanine nucleotide–binding protein (G protein)–coupled receptor kinases (GRKs) is a universal regulatory mechanism that leads to desensitization of G protein signaling and to the activation of alternative signaling pathways.We determined the crystallographic structure of bovine GRK2 in complex with G protein β1γ2 subunits.Our results show how the three domains of GRK2–the RGS (regulator of G protein signaling) homology, protein kinase, and pleckstrin homology domains–integrate their respective activities and recruit the enzyme to the cell membrane in an orientation that not only facilitates receptor phosphorylation, but also allows for the simultaneous inhibition of signaling by Gα and Gβγ subunits.

G protein–coupled receptors (GPCRs) are integral membrane proteins that respond to specific extracellular signals by activating G proteins within the cell. They represent the largest class of receptors in the mammalian genome and play a fundamental role in the sensations of light, smell, and taste and in the regulation of heart rate, blood pressure, and glucose metabolism (1). For cells to remain responsive to their environment, activated GPCRs must be rapidly desensitized. The best characterized system for receptor desensitization is that of the GRKs and arrestins (2, 3). Activated GPCRs are first phosphorylated by GRKs and are then bound by molecules of arrestin, which block the binding of G proteins, target the receptors for clathrin-mediated endocytosis (4), and serve as adaptors that link receptors to other signaling pathways such as those of the mitogen-activated protein kinases (5). GRKs may also inhibit G protein signaling in a phosphorylation-independent manner by blocking the interactions of Gα and Gβγ subunits with their effector targets (6) or by playing a direct role in receptor sequestration and internalization (7).

GRKs, G proteins, and arrestins are the only protein families known to specifically recognize the activated form of most GPCRs. Structural studies of these proteins could therefore provide insights into the mechanism of signal transduction through activated receptors. To date, structures of several members of the G protein family have been determined, but none of any GRK. GRK2 (8, 9), the best characterized GRK, is ubiquitously expressed and can phosphorylate many different GPCRs (3). However, the enzyme plays a particularly vital role in the heart, where it regulates the force and rate of muscle contraction by phosphorylating β-adrenergic and muscarinic acetylcholine receptors (10). Despite its beneficial roles, biochemical and transgenic studies have strongly implicated GRK2 in the progression of cardiovascular disease (10). Inhibitors of GRK2 may therefore serve as useful therapeutic agents (11, 12).

GRK2 consists of three modular domains: a predominantly N-terminal RGS (regulator of G protein signaling) homology (RH) domain (13), a central protein kinase domain (9) that belongs to the AGC superfamily of kinases, and a C-terminal pleckstrin homology (PH) domain (14). The RH and kinase domains are common to all GRKs, whereas the PH domain is unique to GRK2 and GRK3 (fig. S1). Gβγ binds to the PH domain of GRK2 (15) and greatly enhances the phosphorylation of activated GPCRs, at least in part by recruiting the enzyme from the cytoplasm to the plasma membrane (16). To a lesser extent, activation of GRK2 is also dependent on the binding of anionic phospholipids to the PH domain (17, 18). Other GRKs have different domains at their C termini that are similarly involved in membrane targeting (2, 3).

We have determined the x-ray crystal structure of the complex between bovine GRK2 and Gβ1γ2 in detergent micelles. The atomic structure of GRK2 reveals how its RH, kinase, and PH domains integrate their activities to bring the enzyme to the membrane in an orientation that facilitates receptor phosphorylation. Furthermore, the GRK2-Gβγ complex serves as a model for the interaction between Gβγ and the PH domains of other important effector enzymes, including phospholipase Cβ (PLCβ) (19).

Structure determination. Bovine GRK2 and Gβ1γ2 were overexpressed in baculovirus-infected Sf9 insect cells, and their 1:1 complex was purified and crystallized as described (20, 21). The atomic structure of the complex was determined by molecular replacement and solvent flattening using anisotropic diffraction data extending to 2.5 Å in the best reciprocal space direction (Table 1) (21). The final refined model spans residues 29 to 475, 496 to 569, and 576 to 668 of GRK2 (out of 689 total), residues 2 to 340 of Gβ1 (out of 340), and residues 8 to 68 of Gγ2 (out of 68) (fig. S1).

Table 1.

Structure determination and refinement statistics.

Data set Home source (Cu Kα) ALS beamline 8.2.1
Space group C2 C2
Cell constants (Å, °) a = 187.0 a = 188.2
b = 72.1 b = 72.4
c = 122.0 c = 122.7
β = 115.2 β = 115.2
Number of crystals 1 1
Wavelength (Å) 1.542 1.000
Dmin (Å, direction 1, bView inline, direction 3)View inline 3.2 2.4, 2.8, 3.0
Mosaicity (°) 1.2 0.83
Unique reflections 23,608 38,258
Average redundancy 3.4 2.4
RsymView inline (%) 9.6 (50.3)View inline 5.3 (25.3)
Completeness (%) 98.2 (97.3) 73.6 (12.7)View inline
<I>/<σI> 12.5 (2.6) 15.3 (2.7)
Resolution range for refinement (Å)View inline 20 to 2.5
Total reflections used 36,307
Number of protein atoms 8133
Number of water molecules 26
RMSD bond lengths (Å) 0.024
RMSD bond angles (°) 2.035
RMSD bonded B factors (Å2) 1.9
RworkView inline (%) 20.2
RfreeView inline (%) 25.2
Average B factor (Å2) 27.9
  • View inline* Because of anisotropy, data beyond these limits were excluded during scaling and refinement. Direction 1 is 37.9° inclined from a* in the a*-c* plane, and direction 3 corresponds to direction 1 × b* (21).

  • View inline Rsym = ΣhΣi|I(h) - I(h)i|IΣhΣiI(h)i, where I(h) is the mean intensity after rejections.

  • View inline Numbers in parentheses correspond to the highest resolution shell of data, which was 3.3 to 3.2 Å for the Cu Kα data set and 2.59 to 2.5 Å for the ALS data set.

  • View inline§ Before ellipsoidal truncation of the data, completeness was 97.4% (90.8% in the highest shell).

  • View inline Rwork = ΣhFobs(h)| -|Fcalc(h)∥IΣh|Fobs(h)|; no I/σ cutoff was used during refinement.

  • View inline 5.1% of the truncated data set was excluded from refinement to calculate Rfree. During the last three rounds of model building, all data were included in refinement.

  • Quaternary structure of the GRK2-Gβγ complex. When viewed as in Fig. 1A, the RH, kinase, and PH domains of GRK2 fill the vertices of an equilateral triangle roughly 80 Å on a side. The RH domain contacts both the kinase and PH domains, burying 1700 and 1400 Å2 of surface area, respectively. Gβγ binds exclusively to the PH domain and buries 2200 Å2 of surface area, similar to the area buried between Gα and Gβγ in the Gαiβ1γ2 heterotrimer (22, 23).

    Fig. 1.

    Quaternary structure of the GRK2-Gβ1γ2 complex. (A) Membrane-proximal view of the GRK2-Gβγ complex. The RH (RGS homology) domain of GRK2 is colored violet. The kinase domain is depicted with yellow α helices and olive-green β strands. The PH domain is tan, Gβ1 blue, and Gγ2 green. The long axis of the RH domain is declined from the center of the enzyme into the page by about 45°. The first and last observed residues of GRK2 are labeled “N” (residue 29) and “C” (residue 668), respectively. The C-terminal residue of Gγ (Cys68) is labeled “geranyl” to indicate the site of geranylgeranylation. Connections of disordered loops in GRK2 are annotated as follows: I and I′ correspond to residues 475 and 496, respectively, of the kinase domain; II and II′ to residues 569 and 576, respectively, of the PH domain. (B) Side view of the GRK2-Gβγ complex, rotated 90° around a horizontal axis from the view in (A). The flat, membrane-proximal surface spans the top of the complex. (C) Electrostatic surface potential of the membrane-proximal surface of the complex (21). The orientation is the same as in (A). Basic regions are colored blue, acidic regions red, and neutral regions white. (D) Electrostatic surface potential of the GRK2-Gβγ complex in the same orientation as (B).

    Several lines of evidence suggest that the side of the GRK2-Gβγ complex shown in Fig. 1A packs against negatively charged membranes in the cell. This membrane-proximal surface is both positively charged and flat (Fig. 1, B to D) and thereby defines a plane that coincides with the predicted phospholipid-binding site of the PH domain, the geranylgeranylation site of Gγ, and the surface of Gβγ previously predicted to interact with the plasma membrane (23). As expected, superposition of the Gαiβ1γ2 heterotrimer (22) with Gβ1γ2 in the GRK2-Gβγ complex places the myristoylated N terminus of Gαi within this same plane. Finally, three regions in GRK2 known to be important for receptor and/or membrane targeting (7, 12) reside on the membrane-proximal side of the complex (residues 1 to 28, 476 to 495, and 571 to 575). These regions are disordered in the crystal structure.

    The GRK2 RH domain. The RH domain consists of two discontinuous segments within the primary sequence of GRK2 (Fig. 2) (fig. S1A). The first segment (residues 30 to 185) forms the characteristic nine-helix bundle (α1 to α9) found in all RH domains (24). These helices cluster into two lobes referred to as the “terminal” and “bundle” subdomains (Fig. 2, A and B). The second segment (residues 513 to 547) follows the kinase domain in the primary sequence and contributes two additional helices (α10 and α11) to the terminal subdomain. The last three turns of helix α11 (residues 538 to 547) extend away from the core of the RH domain and bear five highly conserved arginine and lysine residues (Fig. 2A) (fig. S1A) that are in close proximity to the proposed surface of the cell membrane (Fig. 1). The RH domains of p115-RhoGEF and PDZ-RhoGEF [GEF, guanine nucleotide exchange factor (25, 26)] also have two additional helices, the first of which appears homologous to helix α10 of GRK2 (Fig. 2C).

    Fig. 2.

    (A) The GRK2 RH domain. Helices that belong to the core RH domain fold are colored violet. Helices α4 to α7 constitute the “bundle” subdomain: a classic antiparallel, four-helix bundle. The N- and C-terminal helices (α1 to α3, α8, and α9) constitute the “terminal” subdomain. Helices α10 and α11 (residues 513 to 547), which follow the kinase domain in the primary sequence (fig. S1), are colored gray. Residues that bind Gαq/11 (31) are drawn with yellow side chains (fig. S1A), and basic side chains in the extended portion of α11 that are expected to interact with the plasma membrane are blue. The α1 helix is displaced by ∼13 Å from its position in RGS proteins. (B) The RGS4 RH domain. The terminal and bundle domains of GRK2 and RGS4 superimpose with RMSDs of 0.8 and 1.0 Å for equivalent Cα atoms, respectively. When their terminal subdomains are superimposed, their bundle subdomains are related by ∼20° around an axis that runs through the long axis of the domain (horizontally in the plane of this diagram). Side chains of residues that interact with Gαi are in green (fig. S1A), and the Cα backbone of the adenomatous polyposis coli (APC) peptide bound to the closely related RH domain of axin (59) is drawn as a black coil. These sites, in addition to the Gαq/11 binding site of GRK2 shown in (A), demonstrate that RH domains have evolved at least three distinct protein binding sites. (C) The RhoGEF RH domain. For this view, residues from the terminal subdomain of p115-RhoGEF (including α2, α3, α8, and portions of α9 and α10) were superimposed on analogous residues from the terminal subdomain of GRK2 (RMSD of 2.1 Å for 57 equivalent Cα atoms). The bundle subdomains of p115-RhoGEF and GRK2 are then related by ∼26° around an axis similar to that described in (B). Because the α10 helix of the RhoGEF RH domain appears homologous with that of GRK2, the GRK2 kinase domain can be thought of as an insertion within the α9-α10 loop of an ancestral RH domain, which consisted of at least 10 helices.

    The best characterized RH domains are those of the RGS proteins (27, 28), which bind to activated [i.e., guanosine triphosphate (GTP)–bound] Gαi or Gαq/11 subunits and serve either as GTPase activating proteins (29) or as competitive inhibitors versus downstream effectors of Gα (30). The RH domains of GRK2 and GRK3 likewise bind activated Gαq/11 (6, 31). However, the Gα-interaction surface of the GRK2 RH domain is remarkably different from those of RGS proteins (Fig. 2, A and B) (fig. S1A) and maps almost exclusively to the length of the α5 helix (31).

    The GRK2 kinase domain. GRK2 is roughly 32% identical in sequence to other AGC kinases of known structure (fig. S1B), including protein kinase A (PKA) (32), protein kinase B (PKB/Akt) (33, 34), and 3-phosphoinositide–dependent kinase 1 (PDK1) (35). Upon comparison with these other kinases, it is clear that the GRK2 kinase domain is in an inactive conformation (Fig. 3). The two lobes of the GRK2 kinase domain are splayed 6° farther apart than in the most open conformations observed for PKA (36) but to a similar extent as that described for the inactive conformation of PKB (33). In addition, the “nucleotide gate” (residues 476 to 495) of the GRK2 kinase domain is disordered, as it is in structures of PKA that adopt an open conformation (37, 38). However, unlike in other inactive kinases (33, 39), the αC helix of GRK2 is completely ordered and properly oriented with respect to the active site.

    Fig. 3.

    Stereoview of the GRK2 kinase domain. The view is roughly from the proposed plane of the plasma membrane (Fig. 1B). Like other protein kinases, the GRK2 kinase domain is composed of small and large lobes. The small lobe (residues 186 to 272 and 496 to 513) consists of a six-stranded antiparallel β sheet (β1 to β5, β10) and three α helices (αB, αC, and αK). The β10 strand (residues 511 to 513) has no structural equivalent in PKA, whose primary sequence ends two residues before the strand at Phe350 (equivalent to Phe509 in GRK2). The large lobe of GRK2 (residues 273 to 475) is predominantly α-helical except for four antiparallel β strands. AMP-PNP (adenylyl-imidodiphosphate; ball-and-stick model) and the GSK3 peptide (black coil with serine residue) are modeled from the structure of PKB (34) and indicate the locations of the active site and the polypeptide binding cleft, respectively. Elements of secondary structure discussed in the text are labeled. Insertions, deletions (fig. S1B), or elements that have a different conformation in GRK2 from their counterparts in mouse PKA are drawn in red and are labeled as follows: I, residues 246 to 252 (insertion of three residues in the αC-β4 loop, which form part of the interface with the RH domain); II, residues 279 to 287 (helix αD rolls ∼2.5 Å to the left in this view compared to PKA); III, residues 342 to 351 (one-residue deletion); IV, residues 364 to 366 (one-residue insertion); V, residues 392 to 399 (two-residue insertion); and VI, residues 495 to 513 (end of nucleotide gate, αK and β10). The residues on either end of the disordered nucleotide gate of GRK2 are labeled with their amino acid numbers (475 and 496).

    A striking difference between GRK2 and other AGC kinases is that its catalytic activity is solely regulated by the binding of substrates and ligands, not by phosphorylation of its kinase domain. Other AGC kinases achieve full activity only after phosphorylation of their activation loop and at one or two additional sites termed the “turn” and “hydrophobic” motifs (40). GRK2 lacks serines, threonines, or even acidic side chains that could mimic phosphorylation at equivalent positions within these three sites (fig. S1B). Even so, the GRK hydrophobic motif could still play an important role in regulating enzymatic activity (Fig. 4A).

    Fig. 4.

    The RH domain–kinase domain interface of GRK2. (A) The hydrophobic motif of GRK2 and the interface between the α10 helix and the kinase domain. The kinase domain is colored as described in Fig. 1, and residues from each domain have side-chain carbon atoms colored the same as their respective domains. Amino acids with orange labels belong to the hydrophobic motif of GRK2, which consists of a serine, threonine, or acidic residue preceded by two aromatic residues (Tyr506 and Phe509 in GRK2) and immediately followed by a hydrophobic residue (Leu511, the first residue of the β10 strand). In other AGC kinases, but not GRKs, phosphorylation of the residue equivalent to Pro510 is required for full activation (40). The most striking feature of the interface between the α10 helix of the RH domain (violet Cα trace) and the kinase domain is a buried interdomain salt bridge between Glu520 and Lys210. The side chain of each residue is fully coordinated with hydrogen bonds. The N-terminal loop of the RH domain (residues 29 to 35) and the two linkers that join the RH and kinase domains also contribute several hydrophobic residues to the RH domain–kinase domain interface (fig. S1, A and B). Arg516 forms a salt bridge with Asp272 in the kinase domain, whose backbone carbonyl will accept a hydrogen bond from the purine ring of adenosine triphosphate. Arg516 also forms extensive contacts with the αC-β4 loop. Just before this loop within αC are several essential active-site residues. Conformational changes at this domain interface could therefore influence the catalytic activity of the kinase domain through the hydrophobic motif and/or Arg516. (B) Comparison of two modular kinases, GRK2 and c-Src. Both kinases are shown as Cα traces. The phosphotyrosine residue (pTyr) in the C-terminal tail of c-Src is shown as a ball-and-stick model. Domains that are expected to play analogous functional roles in each kinase are colored similarly. Gray coils correspond to regions outside the conventional boundaries of each signaling domain. The short green segment in each kinase corresponds to residues 511 to 513 in GRK2 and residues 260 and 261 in c-Src. Both segments form short, superimposable β strands in the small lobe (although their orientations are antiparallel and parallel, respectively), and both are expected to play regulatory roles within their respective small lobes. The interaction of the GRK2 α10 helix with the kinase small lobe is analogous to that of the SH2–catalytic domain linker (41, 42). Side chains of residues that form the interface between the bundle subdomain and the large lobe of the kinase domain are drawn as ball-and-stick models.

    The RH and kinase domain core. The RH and kinase domains represent the conserved functional core found in all GRKs, and they are intimately associated with each other in the GRK2-Gβγ structure (Fig. 4) (fig. S1). The most highly conserved portion of their interface (1500 Å2 of buried surface area) is formed between the α10 helix of the RH terminal subdomain and the β2-β3 and αC-β4 loops of the kinase small lobe (Fig. 4A). The second section of the interface (∼250 Å2 of buried surface area) is formed between the α4-α5 loop of the RH bundle subdomain and the αJ helix of the kinase large lobe (Fig. 4B). This interface consists of one ion pair and several hydrophobic contacts.

    The RH and kinase domain core of GRK2 has intriguing similarities with the inactive structure of Src (41, 42). When the small lobes of the GRK2 and Src kinase domains are superimposed, the RH terminal and bundle subdomains of GRK2 roughly align with the SH3 and SH2 domains of Src, respectively, and make analogous contacts with the small and large lobes of their kinase domains (Fig. 4B). Like the SH3 and SH2 domains, the terminal and bundle subdomains of GRK2 interact with other proteins and/or domains that could modulate their interactions with the kinase domain. Interestingly, the α10 helix of GRK2 appears structurally analogous to the SH2–kinase domain linker, which is thought to play an important role in Src activation (43, 44). Both elements are sandwiched between the same surface of the kinase small lobe and a regulatory domain (terminal subdomain in GRK2; SH3 domain in Src). The α10 helix is therefore in prime position to similarly regulate the activity of GRK2. Although much less extensive, the interface between the RH bundle subdomain and the kinase large lobe could modulate GRK2 activity by influencing the relative orientation of the kinase small and large lobes, as does the analogous contact in Src (39).

    The GRK2 PH domain. The PH domain of GRK2 (residues 553 to 661) is a flattened, seven-stranded antiparallel β barrel that is capped on one end with a C-terminal helix (αCT) (Fig. 5A). The β1-β2, β3-β4, and β5-β6 loops circumscribe the open end of the barrel. As in other PH domains (45), residues from GRK2 known to be involved in phospholipid binding (14, 46) map primarily to the inner face of the β1-β2 loop (Fig. 5A). Accordingly, the β1-β2 loop in the GRK2-Gβγ structure is juxtaposed with the proposed plane of the plasma membrane.

    Fig. 5.

    The GRK2 PH domain and its interface with Gβγ. (A) The GRK2 PH domain. Residues implicated in the binding of anionic phospholipids (14, 46) are drawn with blue side chains. For perspective, inositol 1,4,5-trisphosphate [Ins(1,4,5)P3] is modeled from the structure of PLCδ (60). Two residues from GRK2, Lys567 and Arg579, are in position to coordinate the phosphates of the anionic head group, as do equivalent residues in other PH domains (45). The four regions within the primary sequence of the GRK2 PH domain that contact Gβγ are each drawn witha different color for reference [see (B) and (D)] (21). These regions form a continuous surface that includes the β1-β4 sheet of the PH domain, the extended portion of the αCT helix, and the C-terminal tail. The conformation of the β1-β4 sheet is highly conserved among PH domains of known structure (table S1). Therefore, many PH domains have a surface that is complementary in shape to the effector-binding surface of Gβγ. (B) Specific interactions between the PH domain and Gβ. The location of each interacting residue within the tertiary structure of the PH domain is indicated alongside each amino acid. In addition to the interactions shown, Arg689 of GRK2, which was not modeled because of weak electron density, is also expected to contact the surface of Gβγ. (C) Comparison of the surfaces of Gβγ that bind Gα subunits and GRK2. The molecular surface of Gβγ was colored according to its contacts with Gα (blue), GRK2 (red), or neither (white). Common binding surfaces are colored purple. The footprint of Gα on the surface of Gβγ overlaps extensively with that of the GRK2 PH domain, and thus their binding is mutually exclusive. Positions of various residues from Gβ that contact the PH domain of GRK2 are labeled for reference. (D) Stereoview of the PH domain–Gβ1 interface. Residues from Gβ are drawn with gray carbons, residues from the PH domain with tan carbons. Hydrogen bonds or salt bridges between residues are indicated with dashed lines. The side chain of Met664 from the RH domain binds within a hydrophobic pocket, one wall of which is formed by Leu117 of Gβ (omitted for clarity). Trp99 of Gβ docks into a hydrophobic groove at the interface such that its indole nitrogen is oriented toward solvent. Although electron density for the side chain of Lys663 is not observed beyond Cβ, it could extend far enough into the central channel of Gβ to allow interaction of its Nξ atom witha ring of seven carbonyl oxygens donated by the innermost strand from each blade of Gβ. Figure S4 details additional interactions between the PH domain C-terminal tail and Gβ.

    Adjacent to the phospholipid binding site is the RH domain–PH domain interface, formed primarily between a groove between the β1 strand and αCT helix of the PH domain and the α1 and α9 helices of the RH terminal subdomain (figs. S1 and S2). The interface has a substantial hydrophobic core in addition to a salt bridge and two ion pairs. Site-directed mutagenesis of one of these ion pairs leads to a deficiency in phospholipidmediated activation of GRK2 (46), suggesting a possible route for allosteric communication between the PH and RH domains. Residues within the α1 helix and preceding loop (residues 19 to 29) are not conserved in GRKs that lack PH domains (fig. S1A). However, given the close structural proximity of the N and C termini of the conserved RH and kinase domain core, it seems likely that in each GRK analogous interactions will occur between the α1 region and the C-terminal membrane-targeting domain.

    The conformation of the PH domain in the GRK2-Gβγ complex is distinct from its nuclear magnetic resonance solution structure (14). The binding of Gβγ appears to order the C-terminal region of the domain and to mold the β3 and β4 strands into a conformation more consistent with the core secondary structure observed in other PH domains (table S1). In some other PH domains (47) and possibly in GRK2 (14), the β3 and β4 strands contribute residues to the phospholipid binding site. Therefore, the conformational change induced by Gβγ could explain the cooperative binding of Gβγ and phospholipids to GRK2 (48).

    The structure of Gβ1γ2 in complex with GRK2. In prior studies, structures of Gβγ have been determined alone (49) and in complex with Gα subunits (22, 23) or phosducin (50, 51). The structure of Gβ1 in the GRK2-Gβγ complex agrees well with these structures [root mean square deviation (RMSD) of 0.85 Å for equivalent Cα atoms] except for those of the phosducin complexes (RMSD ∼1.2 Å). The β5-β6 loop and the side chains of Tyr59 and Gln75 of Gβ1, which make extensive contacts with the PH domain (Fig. 5), adopt distinct conformations from their positions in other Gβ complexes (52). Compared with the structure of the Gαiβ1γ2 heterotrimer (22), seven additional residues are observed at the C terminus of Gγ2, including the terminal residue Cys68, whose α-carboxylate and sulfhydryl are methylated and geranylgeranylated, respectively, both in vivo and in our structure (21) (fig. S3). In addition, residues 52 to 61 of Gγ2 adopt a distinct conformation, perhaps as a result of the ordered C terminus.

    The PH domain–Gβγ interface. Many PH domains have been shown to bind Gβγ, some with dissociation constants on the order of 20 to 50 nM (53, 54). The complementarity, sequence conservation, and extent of the PH domain–Gβγ interface observed in our crystal structure strongly support its physiological relevance. As predicted (55, 56), the footprint of the GRK2 PH domain on Gβγ overlaps extensively with the binding sites for Gα subunits and other Gβγ effectors, including PLCβ and ion channels (Fig. 5, B and C). Therefore, GRK2 will compete with these proteins for binding to Gβγ. Four regions within the primary sequence of the PH domain contribute to the Gβγ interface (fig. S1C). These regions form a continuous surface that includes strands β1 to β4 and the extended portion of the αCT helix (Fig. 5A) (21).

    Contact residues within the GRK2 PH domain are not well conserved in other PH domains known to bind Gβγ (fig. S1C) (21). The specific details of the interface of each PH domain with Gβγ will obviously differ, although some general features are apparent. Superposition of the core secondary structures of PH domains known to bind Gβγ with high affinity (table S1 and fig. S1C) shows that the conformation of their respective β strands (β1 to β4) is highly conserved and forms a concave surface complementary to the effector-binding surface of Gβ. In addition, the top of the Gβ propeller responsible for binding PH domains and Gα subunits is extremely acidic (far right of Fig. 1D). Accordingly, regions analogous to the C-terminal tail of the GRK2 PH domain in other Gβγ-binding PH domains tend to harbor basic residues (figs. S1C and S4).

    Implications. The GRK-Gβγ structure reveals how three common, modular signaling domains successfully coordinate their activities to perform several complex tasks. The first is to control the intrinsic activity of the kinase domain in lieu of phosphorylation. The RH terminal subdomain clearly contributes to this regulation (Fig. 6A). The subdomain is not only intimately associated with both the PH domain, which binds two synergistic activators (Gβγ and anionic phospholipids), and the bundle subdomain, which binds Gαq/11·GTP; it also contacts the small lobe of the kinase domain in multiple regions, each known by analogy with Src, PKA, and PKB to be critical for the regulation of catalytic activity (Fig. 4A). Therefore, a change in the conformation of the terminal subdomain, induced by the binding of Gβγ, phospholipids, or Gαq/11·GTP, could directly lead to changes in catalytic activity via its interface with the kinase domain.

    Fig. 6.

    Nonexclusive binding of GRK2 to Gβγ, GPCRs, and Gαq. (A) Membrane-proximal view. Models of a GPCR (maroon) and Gαq (dark blue with tan β strands) were docked with GRK2-Gβγ in a manner consistent with the expected orientation of each individual protein at the cell membrane (21). The kinase domain of GRK2 is modeled in a closed, “active” conformation, and the six C-terminal residues of the receptor have been repositioned to follow the same path as peptides bound to other AGC kinases (33, 58). GRK2 is known to contact GPCRs at two distinct locations: within the polypeptide binding cleft of the kinase domain, and at a noncatalytic, allosteric “docking site” that discriminates between inactive and stimulated receptors (3, 57). The proposed GPCR docking site is on the kinase domain, near two disordered regions of GRK2 (residues 1 to 28 and 476 to 495) that not only are believed to play important roles in receptor and/or phospholipid binding, but also contain external regulatory sites for Ca2+-calmodulin, PKC, and clathrin (12, 60). The switch regions of Gαq (red) pack against the Gαq-binding residues identified in the RH bundle subdomain of GRK2 (31), rendering their interaction dependent on the signaling state of the G protein. The N and C termini of the GPCR and Gαq are indicated. GDP·AlF4, bound to Gαq, is drawn as a ball-and-stick model. (B) Side view of the complex, rotated 90° around a horizontal axis with respect to (A). The gray bar represents the proposed membrane bilayer, as defined by the top surface of the GRK2-Gβγ complex and the belt of hydrophobic residues presented by the transmembrane helices of the GPCR.

    Another task is to recruit GRK2 from the cytosol to the cell membrane close to activated receptors. This is accomplished by the synergistic binding of Gβγ and anionic phospholipids to the PH domain. Because free Gβγ subunits are generated close to activated receptors, GRK2 will be targeted primarily to specific sites on the membrane where active signal transduction is taking place (16).

    A third task is to orient GRK2 at the membrane in a manner that facilitates interaction with and phosphorylation of activated receptors. This is achieved by the domain interfaces of GRK2, which fix each of its three domains in an optimal orientation with respect to the cell membrane (and therefore with respect to GPCRs). With the plane of the membrane as defined in Fig. 1, there is a substantial gap between the membrane and the large lobe of the GRK2 kinase domain (Fig. 6B). The cytosolic domain of a GPCR can be docked into this gap such that receptor residues known to be important for binding GRKs (57) are in close proximity to both lobes, the nucleotide gate, and the active site of the kinase domain. In this model, either an extended third cytoplasmic loop or the C-terminal tail of the receptor can enter the polypeptide binding cleft of GRK2 in the same orientation observed for peptides cocrystallized with PKA (58).

    The proposed orientation of GRK2 at the membrane is also consistent with the formation of a complex between its RH domain and Gαq/11·GTP (Fig. 6). The resulting interface consists of a wedge-like intrusion of the α5 and α6 helices of the RH domain into the cleft formed between the Ras-like and α-helical domains of Gαq/11, making extensive contacts with the switch regions of Gα (6, 31).

    Remarkably, the Gβγ and the proposed GPCR and Gαq binding sites are each on a different vertex of the “triangle” formed by the three domains of GRK2 (Fig. 6A). GRK2 could therefore bind all three proteins simultaneously. This would represent an extremely efficient way to attenuate G protein signaling. The kinase domain would bind and phosphorylate activated receptors, recruiting arrestin and thereby blocking access of G proteins and tagging the receptor for endocytosis. The RH and PH domains would bind Gαq/11·GTP and Gβγ, respectively, blocking the G proteins from their downstream targets–effectively keeping them at bay until receptor signaling is terminated.

    Supporting Online Material

    www.sciencemag.org/cgi/content/full/300/5623/1256/DC1

    Materials and Methods

    Figs. S1 to S4

    Table S1

    References and Notes

    View Abstract

    Stay Connected to Science

    Navigate This Article