Structural Determinants Involved in the Formation and Activation of G Protein βγ Dimers

Abstract

Heterotrimeric G proteins, composed of an α, β and γ subunit, represent one of the most important and dynamic families of signaling proteins. As a testament to the significance of G protein signaling, the hundreds of seven-transmembrane-spanning receptors that interact with G proteins are estimated to occupy 1–2% of the human genome. This broad diversity of receptors is echoed in the number of potential heterotrimer combinations that can arise from the 23 α subunit, 7 β subunit and 12 γ subunit isoforms that have been identified. The potential for such vast complexity implies that the receptor G protein interface is the site of much regulation. The historical model for the activation of a G protein holds that activated receptor catalyzes the exchange of GDP for GTP on the α subunit, inducing a conformational change that substantially lowers the affinity of α for βγ. This decreased affinity enables dissociation of βγ from α and receptor. The free form of βγ is thought to activate effectors, until the hydrolysis of GTP by G α (aided by RGS proteins) allows the subunits to re-associate, effectively deactivating the G protein until another interaction with activated receptor.

Introduction

Although there are many facets of G protein signaling via the βγ dimer, this review will primarily focus on how the structure of a βγ dimer participates in the transfer of signal from activated receptor to the heterotrimeric G protein. Structural heterogeneity of βγ combinations will be discussed regarding how specificity of interactions of βγ with G α and receptor determine which βγ dimer isoforms are activated. A discussion of specific interactions of βγ dimers with effector molecules is beyond the scope of this review, but is addressed elsewhere in this issue. However, some references to interactions with effectors are included, as the activation of effectors is by definition one of the primary measures of activity of a G protein βγ dimer. Beyond the brief descriptions of the G protein activation cycle that have appeared in thousands of papers over the years, it is clear that βγ dimer ‘activity’ may not simply be synonymous with dissociation from α. Thus, an important question is whether dissociation of α from βγ is a requisite step in the activation of βγ. Also, is a βγ active if it does not properly localize with its effector molecule? Does a βγ need a receptor, or even an α subunit, to be activated? Does a βγ always exist as dimer? The record in the literature may be limited for some of these later questions, but they represent intriguing ideas that will help refine the model of G protein signaling.

G Protein βγ Heterogeneity

At least 16 α genes, 5 β genes and 12 γ genes have been identified in the human genome [1,2,3,4,5,6]. G α isoforms can be separated into four subfamilies, G_i, G_s, G_q and G₁₂; when alternative splicing and posttranslational processing are taken into account, there are at least 23 α isoforms, which are reviewed elsewhere [5]. The first four β isoforms discovered, β_1–4, are highly homologous (80–90% identical) 36-kDa proteins; G β₅, a 40-kDa protein, is only 50% identical to the first four β isoforms. Several truncated splice variants of β₃ have been characterized, β₃s [7], β₃s2 [8] and β₃v [9]; a splice variant of β₅ which has an N-terminal extension, β_5L[10], has also been characterized. All 12 γ isoforms are between 7 and 8.5 kDa in size, but are much more divergent than the β isoforms. Since β and γ are believed to form a functional dimer in vivo, heterogeneity of βγ defined as the product of the β and γ genes is likely more diverse than G α [11] even though not every possible βγ combination can form. Posttranslational processing of β and γ further contribute to the structural diversity of these proteins [12], with γ isoforms receiving more extensive study than β isoforms to date.

Posttranslational Modifications

β Subunit

Several posttranslational modifications of G β have been characterized. The N-terminus of β₁ was reported to undergo removal of the methionine at position 1, followed by N-acetlyation of serine at postion 2 [13], the functional implications of which were unclear. Phosphorylation of β has also been reported, but interestingly, at a histidine residue [14] instead of the traditional serine, threonine or tyrosine; nucleoside diphosphate kinase (NDPK) was identified as responsible for phosphorylating histidine 266 of β₁[15]. This phosphorylation event was predicted as a mechanism for G protein activation, which will be discussed further below. In contrast, the reversible mono-ADP-ribosylation of arginine-129 of activated or ‘free’ G β was demonstrated to reduce activity at effectors such as type 1 adenylyl cyclase [16], phosphoinositide 3-kinase-γ and phospholipase C-β₂[17]. Mono-ADP ribosylation of β increased upon activation of a variety of cell surface receptors, including the G_q-linked thrombin receptor and the G_i-linked 5-HT serotonin receptor, and thus was predicted to be a regulatory mechanism to inhibit βγ signaling. Inhibition of βγ activity by posttranslational modification has parallels to the GTPase activity of the G α subunit, and may represent an unappreciated regulatory mechanism in G protein signaling.

γ Subunit

C-Terminal Processing

The G γ subunit contains posttranslational modifications at the N- and C-terminus, and much of this processing has been demonstrated to be critical to G protein function. Most attention has focused on covalent modification at the C-terminus of γ with either of two distinct isoprenoid moieties [for review, [18]]. For γ isoforms ending in amino acid CAAX, where X is serine, glutamine or methionine, CVIS in the case of γ₁, and additionally γ₈, and γ₁₁, the enzyme farnesyltransferase (FTase) covalently attaches a 15-carbon farnesyl group via a thioether bond to the cysteine in the CAAX motif [19]. If the X in the CAAX sequence is a leucine such as CAIL for γ₂, or the C-terminal sequences in γ₃, γ₄, γ₅, γ₇, γ₉, γ₁₀, γ₁₂ and γ₁₃, the enzyme geranylgeranyltransferase type I (GGTase-I) attaches a larger 20-carbon geranylgeranyl group to the cysteine [20, 21] via the same thioether bond. After modification with either prenyl moiety, processing is similar for all γ isoforms. An endoprotease residing in the microsomal membranes cleaves the C-terminal-AAX residues [22], and the new prenylated C-terminal cysteine is carboxy methylated by a methyltransferase [23]. Both assembly of βγ dimers, which occurs in the cytosol [24], and prenylation of γ are required to target βγ to membranes [25]; there is some debate over which membranes, as βγ has been observed in endoplasmic reticulum membranes from biochemical and immunofluorescence studies of βγ expressed in mammalian cells [26]. Alternatively, live cell imaging demonstrated localization of βγ predominantly to the plasma membrane [27]. Although posttranslational processing of G γ as generalized above is well documented and well accepted, it is also evident that processing exceptions may provide insights into how G βγ functions in vivo.

For example, it is believed that the cleavage of the C-terminal-AAX amino acids is a necessary step in protein maturation prior to βγ assembly [28]. However, this is not always the case; a geranylgeranylated γ₅ isoform that had not undergone cleavage of the C-terminal-AAX amino acids was characterized by mass spectrometry from a purified preparation of G protein from bovine brain [29]. This unprocessed form was discovered to predominate over the cleaved form, and be dependent on an aromatic phenylalanine residue in the CSFL C-terminal sequence of γ₅[30]. A recent study noting a physical interaction between proteins containing PDZ domains, which are important in the construction of elaborate scaffolding networks, and Gγ₁₃[31] made the story of this processing pattern more interesting. Gγ₅, with its C-terminus ending in CSFL, is one of only four γ isoforms other than γ₁₃ with a C-terminal target sequence (CT/SXX) for class I PDZ domain containing proteins. The fact that the C-terminal sequence of γ₅ can remain after maturation of a βγ dimer suggests that γ₅ isoforms that do not undergo C-terminal proteolytic cleavage are differentially targeted compared to other βγ isoforms. Since this form of processing appears to be unique for γ₅, there may be a specific signaling role for βγ dimers containing γ₅ in signaling complexes containing PDZ domains.

Unprenylated γ

The absence of prenylation in Gγ is also associated with unexpected signaling properties. After the observation that a fraction of β₂γ₂ could localize to the nucleus and regulate transcriptional activity [32], a study by Kino et al. [33] demonstrated that lack of prenylation of γ₂, either by mutation of the C-terminal cysteine to serine or by pharmacological inhibition, resulted in increased nuclear localization of βγ and increased ability to regulate transcription. Further, non-prenylated γ₅ expressed in bacteria was shown to regulate transcription by binding to the adipocyte enhancer-binding protein (AEBP1) transcriptional repressor [34]. Although prenylation of γ is not thought to be reversible, fully processed isoforms, missing a prenyl group, have been characterized by mass spectrometry in γ2 (<1% of γ2 observed) and γ7 (1–5% of γ7 observed) from G protein purified from bovine brain [12]. These exceptions in the prenylation pathway, although apparently low in occurrence, are surprising in that conventional wisdom assumes that, although prenylation is not required for assembly [28], it is a prerequisite for βγ activity. Moreover, since regulation of transcription is not a classical signaling function for βγ, the prenylation status of βγ may represent a significant point in modulation of activity of βγ dimers, with respect to identity and localization of effector targets.

N-Terminal Processing

Other regions of γ that are sites for covalent modification include the N-terminus, which were initially observed to be refractory to Edman degradation in the case of γ₂, γ₅ and γ₇; alternatively, γ₁[35], γ₃[36] and γ₁₁[37] were not found to be N-terminally blocked. Mass spectrometry was used to identify structural modifications at the N-terminus of the γ₂ isoform, which were revealed to be cleavage of the N-terminal methionine followed by N-acetylation of alanine formerly at position 2 [38]. Yet another variant of γ₂ was characterized by Edman degradation, in which a novel N-terminal sequence was determined to be a substrate for the N-end rule ubiquitylation pathway [39]. Modification by ubiquitin was found likely to occur at lysine residues in the C-terminal region of γ; although ubiquitylation is generally regarded as a signal for protein degradation, the authors suggest that this modification may also modulate membrane binding of βγ.

Phosphorylation

Interestingly, phosphorylation has only been observed in the γ₁₂ isoform, in which protein kinase C was shown to phosphorylate a serine at position one in vitro [40]. Phosphorylation appeared to increase the affinity of β₁γ₁₂ for Go α [40], and increase the ability of β₁γ₁₂ to couple to receptor [41]; however, phosphorylation diminished the activity of β₁γ₁₂ at adenylyl cyclase type II, but not phospholipase C-β [41]. Phorbol 12-myristate 13-acetate (PMA) induced in vivo phosphorylation of γ₁₂ in cultured Swiss 3T3 cells was augmented by activation of the lysophosphatidic acid receptor, and inhibited by pertussis toxin, suggesting a role for G_i proteins in the regulation of γ₁₂ by protein kinase C [42]. The increase in γ₁₂ phosphorylation after receptor activation is also in agreement with the observation that βγ dimer, rather than heterotrimer, is the preferred substrate for protein kinase C [40]. It is possible that two of the potential signaling effects of this modification, increased G protein cycling, and targeting of specific effectors, combine to regulate signaling pathways within a cell. These examples demonstrate that posttranslational modifications of G protein β and γ subunit isoforms, while often overlooked, can be dynamic and thus represent points of signaling modulation that could affect intracellular targeting of a βγ dimer, and regulation of interactions with G α, receptor, and effector molecules.

Structure of βγ

G protein β subunits belong to a family of WD40 repeat proteins, characterized by a repeating motif of 27–45 amino acids punctuated at the C-terminal end by the Trp-Asp (WD) dipeptide sequence [43]. The solutions to the crystal structures of a G protein and its constituent subunits were watershed events in the G protein field, providing for the first time a three-dimensional model of the heterotrimeric signaling molecule. This review will begin with those structures, and also discuss biochemical data that have served to complement the crystal structures of the G protein βγ dimer. The structure of the β₁γ₁ dimer of transducin was published by Sondek et al. [44] in 1996. At about the same time, the crystal structures of the β₁γ₂ dimer associated with the G_i1 α subunit, and the β₁γ₁ dimer associated with a chimera of the G_t α and G_i1 α subunits, were published by Wall et al. [45] and Lambright et al. [46], respectively. According to the structures, the 340-amino-acid β subunit forms a toroidal structure defined by seven propeller blades (fig. 1), with each blade comprised of a series of β-sheets. The N-terminus of the β subunit is an α-helix which interacts with the α-helical structure of the N-terminus of the approximately 70-amino-acid γ subunit to form a coiled-coil domain (fig. 1), which had been predicted from earlier biochemical studies [47]. The remainder of the γ subunit exists in an α helical structure that makes extensive contacts with the blades of the β torus (fig. 1). Crystal structures of β₁γ₁ and β₁γ₂ were highly homologous; further, a crystal structure of the more divergent β₅ complexed with an RGS9 protein [48] revealed that β₁ and β₅ form very similar toroidal structures. This suggests that, even considering the high degree of diversity among the β and γ subunit isoforms, the overall structure of a βγ dimer is highly conserved.

Fig. 1.

Clam shell model of G protein opening after activation using the crystal structure of Wall et al. [45] . Left structure : Relationship of inactive GDP-bound form of G_i1 αβ₁γ₂ to the plasma membrane, with stylized lipids added to the structure. Right structure: βγ is in same conformation, but G_i1 α has been rotated approximately 90° counterclockwise; note that although lipids maintain their proximity, the face of the β propeller and switch I and II regions of G_i1 α, both occluded at left, are now sterically free to interact with effectors, and are thus ‘active’. (Although the same structure was used for simplicity, a truly active G_i1 α would have conformational changes in the switch regions, and GTPγS bound instead of GDP.)

Although the β₁γ₂[45] and β₁γ₁[46] dimers had been crystallized with G α subunits, revealing important sites of interaction, data elucidating the dynamic nature of these interactions was derived from the comparison of the crystal structures of G_t α in the inactive GDP-bound form [49] and the active GTPγS-bound form [50]. The general architecture of the G α subunit consists of a GTPase domain that is homologous to the monomeric GTP-binding proteins (such as ras), and a helical domain that is only found in heterotrimeric G protein α subunits. Two regions in the GTPase domain that interact with the γ-phosphate of GTP, coined switch I and switch II (fig. 1), stand out in the G_t α crystal structures in that they undergo a conformational change depending on the nucleotide bound. A third switch region found in the helical domain, switch III, also undergoes a conformational change, apparently dependent upon the conformational change of switch II [49]. These conformational changes, reviewed in more detail elsewhere [51], were critical to understanding the mechanism of activation of G proteins, as the crystal structures of two heterotrimeric G proteins revealed that although no contacts were observed between α and γ, many of the contact sites between α and β reside in the switch I and switch II regions (fig. 1) [46, 45]. Thus, the logical conclusion based on the crystal structures, and also predicted from earlier biochemical experiments that reported decreased affinity between activated α and βγ [52], was that nucleotide exchange in a G protein could induce conformational changes in the βγ-binding site of G α, resulting in subunit dissociation and activation of both α and βγ.

The model of G protein activation based on the crystal structures will continue to benefit from biochemical studies for a number of reasons. For one, purified G proteins could only be crystallized after removal of posttranslational farnesyl or geranylgeranyl lipid modifications [45, 46]. The three-dimensional structure of the G protein did, however, strongly suggest that the lipid modified N- and C-termini of α and γ, respectively, were proximal to one another [46]. Even after crystallization with an intact γ including the prenylated C-terminus, conformational instability of the C-terminal region precluded assignment of a static structure to this region of γ [53]. In addition, no crystal structure exists for a G protein in the empty state, that is, with no nucleotide bound; this state represents the transition between receptor-dependent release of GDP and binding GTP. Furthermore, a seven-transmembrane-spanning receptor has not been crystallized with a G protein; what is known of receptor:G protein interactions has been derived from biochemical data. Putative points of contact between G protein and receptor on the heterotrimer include the N-terminus [54] and C-terminus [55] of G α, the C-terminus of γ [56,57,58] and the C-terminus of β [59]. Thus, the mechanism of signal transfer from activated receptor in a membrane environment to G protein continues to be a point of conjecture based on empirical evidence.

Mechanism of Activation

Based largely on the crystal structures of heterotrimeric G proteins and monomeric GTP-binding proteins, several studies from Henry Bourne's laboratory, Onrust et al. [60] in 1997 and Iiri et al. [61] in 1998, sought to explain the molecular mechanism of activation of G protein by receptor through the ‘lever’ hypothesis. Such a hypothesis was necessary because the distance between the intracellular loops of a receptor and GDP were thought to be too far for direct interaction (fig. 1), thus the requirement of receptor to ‘act at a distance’ [62]. Integral to the hypothesis was the interaction between activated receptor and the β6 strand/α5 helix region of G α, which contains a loop to which GDP binds. Further, the switch I region and the β3/α2 loop of G α, in addition to containing βγ-binding sites, also form a lip that provides a secure binding site for GDP. Thus, two important events in receptor activation of G protein occur when: (1) receptor induces a conformational change in the β6 strand/α5 helix region of Gα, which would subsequently alter the GDP-binding site on the β6/α5 loop, and (2) the postulated insertion of the intracellular loops of the receptor into a crevice between α and βγ, which by employing βγ as a lever to pry α and βγ apart, could lead to an allosterically induced conformational change in the switch I/β3α2 loop ensconcing GDP, allowing release of nucleotide.

This hypothesis was tested biochemically by the creation of a mutant G_s α subunit that bound the βγ dimer in a conformation that mimicked the G protein structure described above in the absence of receptor [63]. Transfecting cells with βγ and the mutant G_s α increased the activity of the mutant G_s α at adenylyl cyclase, compared with the mutant G_s α alone, further supporting the hypotheses that βγ acts as a lever to affect GDP release in the course of G protein activation. An iteration of the lever hypothesis is the ‘gear-shift’ hypothesis [64], which holds that the N-terminus of the γ subunit acts as a gear-shift by interacting with the helical domain of the α subunit, thereby stabilizing the transitory nucleotide free empty state, and facilitating nucleotide exchange.

Studies with rhodopsin and synthetic G protein peptides suggest that the C-terminus of γ and the C-terminus of α interact with the activated receptor in a sequential, interdependent manner, and this binding facilitates the conformational change that allows release of GDP [65]. A later study refined this model using synthetic peptides and time-resolved near infrared lightscattering, indicating that the prenylated C-terminus of γ was the first point of contact with activated receptor, followed by the C-terminus of α in a brief transitory state that enabled GDP release [66]. A conformational switch in βγ, first proposed in 1995 [67], was envisioned as a mechanism for this sequence of events. Such a switch was predicted because in both detergent solubilized heterotrimeric transducin and β₁γ₁, the C-terminus of γ was resistant to carboxypeptidase Y, suggesting inaccessibility to other proteins; however, a 12-amino-acid C-terminal farnesylated peptide from γ₁ was able to bind to and stabilize light-activated rhodopsin [67]. The details of the switch, characterized by mutational analysis and NMR structural studies, describe an unstructured C-terminus of γ that transforms into an amphipathic helix upon interaction with activated receptor [68, 69]. Three residues conserved across γ subunit isoforms, Asn62, Pro63 and Phe64, were proposed to be critical for the receptor-dependent conformational change in γ.

More recent experiments with protein in solution have begun to yield structural data on the nucleotide free empty state of a G protein activated by receptor. NMR studies with transducin and light-activated rhodopsin suggest that the empty state of transducin α is conformationally dynamic, a condition not attained in the absence of activated receptor [70]. The use of site-directed spin-labeling methods with G_i α_GDP, G_i α_GDPβγ and G_i α_GTPγ_S has allowed even more precise characterization of structural changes in solution; experiments examining G protein activation by rhodopsin indicate that structural changes in G_i α are propagated via the switch I region to the αF-helix, in concert with movement of the α5 helix, which form part of the nucleotide-binding pocket [71]. While this conformational change facilitates GDP release, another interesting structural change resulting from the empty state is the formation of new contacts between α and βγ, suggested by the crystal structure. The presence of these new contacts between α and βγ in the empty state further support a receptor-dependent conformation of the empty heterotrimer distinct from G_i α_GDPβγ.

Subunit Dissociation

Although receptor-dependent nucleotide exchange has been well established, the issue of subunit dissociation has been somewhat more controversial. In early experiments characterizing the biochemical nature of G proteins, purified transducin, G_i and G_s proteins activated with various combinations of aluminum, magnesium, fluoride, GDP, or GTP analogues and detergent could be dissociated into their constituent α and βγ dimers using a number of separation techniques [72,73,74]. The extrapolation that this phenomenon occurred in vivo was attractive for several reasons. For one, subunit dissociation provided a bifurcation of signal after activation of receptor; both α and βγ were free to regulate downstream effectors independently. The potential for α and βγ dimers to form new G protein combinations as a result of activation of several different receptors was yet another point where a cell could regulate signaling specificity. The molecular underpinnings to support the idea that subunit dissociation represented the mechanism of βγ activation, and reassociation of G protein served to inhibit βγ signaling, came from mutational studies that concluded that regions of βγ that activated effectors were found to overlap with G α-binding sites (fig.1) [75, 76]. These findings suggested that a βγ dimer in a heterotrimeric complex was in fact inactive, and that activation of βγ was synonymous with release from G α upon activation by receptor.

Interestingly, a study by Bonacci et al. [77] found that the region of the β subunit that interacts with the switch II region of G α could be targeted with peptides and small organic molecules to selectively disrupt interactions between βγ dimers and effectors; this strategy has been used inhibit inflammation in vivo by blocking βγ-mediated activation of PI3-kinase γ [78], and thus represents a promising area for pharmacological intervention.

Further evidence for subunit dissociation was revealed in the crystal structure of a G_q α-p63RhoGEF complex [79], in which p63RhoGEF interacted with the α2/β4 region, containing switch II, and the α3/β5 regions of G_q α. In addition, the crystal structure of activated G_s α in a complex with the catalytic domains of adenylyl cyclase also revealed effector-binding sites on G_s α to be the switch II region and the α3-β5 loop [80]. Since βγ binds the switch II region, the binding of p63RhoGEF and adenylyl cyclase to G_q α and G_s α, respectively, appears to be mutually exclusive to the binding of βγ, and supports the notion that subunit dissociation is a consequence of G protein activation. However, biochemical and kinetic arguments have been advanced to suggest that subunit dissociation is not necessary for G protein activation [81]. For example, expression of a fusion protein of G protein α and β subunits in yeast signaled as well as α and β subunits co-expressed individually [82]; since the fusion protein did not allow complete physical separation of α and β, a conformational change was proposed as a means for activation of the heterotrimer leading to signaling.

Fortunately, in the last several years sophisticated imaging techniques such as FRET (fluorescence resonance energy transfer) have emerged that allow the question of subunit dissociation to be more fully evaluated in live cells. FRET studies with receptor-dependent activation of fluorescently tagged G protein in Dictyostelium discoideum concluded that subunits did in fact dissociate upon activation [83], although a conformational change in the heterotrimer could not be completely ruled out. A similar FRET study with fluorescently tagged G_i α and β₁γ₂ in HEK cells concluded that the heterotrimer underwent a molecular rearrangement upon activation by receptor, but did not dissociate [84]. A comparison of heterotrimers consisting of β₁γ₂ and other members of the G_i subfamily using FRET found G_O α appeared to dissociate from β₁γ₂ upon receptor stimulation, whereas G_i1,2,3 α or G_z α did not [85]; these differences in subunit dissociation were traced to several distinct regions in G_i1 α. Because the FRET signal is dependent on both proximity and orientation between two fluorophores, it became clear that the distinction between conformational change and physical dissociation of G protein subunits was not to be unequivocally resolved by FRET.

Another imaging technique called FRAP (fluorescence recovery after photobleaching) was more suited to address complete physical dissociation of G protein subunits, as fluorescence recovery is dependent on the movement of fluorescently tagged proteins into and out of bleached regions of the plasma membrane. In one FRAP study which examined G protein subunit dissociation by evaluation of the ability of fluorescently tagged, immobile G_OA α, G_i3 α and G_s α subunits to constrain fluorescently tagged β₁γ₂[86], results were similar to the FRET studies described above. Consistent with the FRET experiments, receptor activation resulted in dissociation of G_OA α from β₁γ₂; however, in these experiments, G_i3 α was also shown to dissociate from β₁γ₂. In contrast, the G_s α β₁γ₂ heterotrimer did not dissociate upon receptor stimulation [86]. Further evidence that subunit dissociation is not always necessary for G protein activation was provided by the discovery that the stability of complexes of effectors and heterotrimeric G proteins persisted even after activation of receptor [87]. Moreover, effector activity of G_q α was demonstrated to be augmented by βγ, independent of receptor activation [88].

If subunit dissociation does not occur, one striking paradox is how a βγ dimer can signal with effector-binding sites occluded by G α. Part of the answer is likely related to the three switch regions of G α (fig. 1) which undergo a conformational change upon exchange of GDP for GTP (GTPγS in the crystal structure) that leads to a decreased affinity of G α for βγ. However, an additional contact site suggested by the crystal structure between α and γ is the N-terminus of α and the C-terminus of γ, both modified by lipid. Biochemical evidence also supports this interaction, as myristoylated G_O α has a higher affinity for βγ than G_O α with an unmodified N-terminus [89]; interactions between the γ C-terminal prenyl group and a N-terminal lipid of α may also be strengthened if both modifications are inserted into the lipid membrane (fig. 1).

Thus, one model, often referred to as the ‘clam shell’, predicts that upon activation, α and βγ pull apart, perhaps with the lipid moieties inserted into the membrane serving as a hinge in the heterotrimer, enough for effector molecules to interact with the binding regions of βγ or α (fig. 1). This activation without complete subunit dissociation may be manifested in subtle conformational changes that may be difficult to detect using cellular imaging techniques. The rearrangement, but not dissociation of G_i α:βγ upon activation [84] discussed earlier may be an example of the clam shell model. Lack of, or more likely incomplete subunit dissociation implies that particular βγ dimers from heterotrimers such as G_s may not readily reassociate with other G α isoforms after activation by receptor. In a model where a receptor has unlimited access to all G proteins expressed in a cell [90], the question of preserving signaling specificity may be related to the extent to which a G protein dissociates upon activation, and subsequently, the potential to re-association with other subunit combinations.

Specificity of β/γ Interactions

Years of experimental data have led to the consensus that β and γ isoforms form a functional dimer that does not dissociate under physiological conditions. Thus, the activity of a βγ dimer is derived by the identity of both β and γ isoforms. To examine the functional diversity of βγ dimers, many experiments have aimed to discover which βγ dimers can physically form. In cell types with a restricted assortment of β and γ isoforms, such as expression of β₁ and γ₁ in rods and β₃ and γ₈ in cones, this approach was sufficient to estimate probable G protein subunit interactions. A more difficult question, especially in the context of expression of several β and γ isoforms in a single cell, is which βγ dimers actually form? Useful answers have come from studies dissecting receptor signaling pathways, discussed below; however, the most direct answer is revealed by the purification of G protein βγ isoforms from a specific tissue or cell type. Unfortunately, these are not practical experiments to undertake for the multitude of cell types and tissues that constitute an organism.

β₁ and β₂ Interactions with γ Isoforms

Several systems have been used to study β and γ interactions, including the yeast two-hybrid [91], in vitrotranslation [92,93,94] and transfected cell assays [95]. The conclusions generally held that for the first two β isoforms, β₁ was the least restricted in its interactions with other γ isoforms, and β₂ failed to interact with γ₁ and γ₁₁. The region of γ₁ that inhibited dimer formation with β₂ was ultimately determined to be the 3 amino acids at positions 38–40 of γ₁[96]; conversely, replacement of these residues in γ₂ with the analogous residues of γ₁ also inhibited dimer formation with β₂. More recent live cell-imaging techniques have confirmed much of the earlier literature, although in a noteworthy finding, β₁ appeared to display the strongest preference for γ₁₂[97].

β₃ Interactions with γ Isoforms

Interpretation of results with β₃ is more complex. Unlike dimers containing β₁ or β₂, dimers containing β₃ are much less resistant to complete proteolysis by trypsin [7], and the β₃ subunit has displayed weak or absent capacity to interact with γ subunits, including γ₁ and γ₂[92]. The β₃ isoform does contain an additional tryptic cleavage site at lysine 177; however, in experiments with the more specific protease Arg-C, complete digestion of β₃γ dimers was also observed [93]. On the other hand, β₃ has been shown to co-precipitate with γ₅, γ_8cone or γ₁₂[7], and a β₃γ₂ dimer with activity at receptor and effector has been purified using a baculovirus expression system [98]. The anomalous results obtained from studies with β₃ suggest that the crystal structure of a G protein βγ dimer based on the β₁ isoform may not necessarily predict subtle structural variations of βγ dimers containing other even highly homologous β isoforms.

β₃ Splice Variant Interactions with γ Isoforms

Several β₃ splice variants have been characterized: β₃s, β₃s2 and β₃v. β₃s, distinguished from β₃ by deletion of 41 amino acids, or one entire WD-40 domain affecting blades three and four of the torus, is similar to β₃ in γ-binding specificity [7]. However, purification of the β₃s isoform from mammalian or insect cell expression systems has not been reported; in contrast to other βγ combinations, β₃s was poorly extracted with 1% (v/v) Genapol, 1% (w/v) CHAPS, or 1% (w/v) cholate when expressed in Sf9 insect cells with either the γ₂, γ₅ or γ₇ isoforms [98]. The β₃s2 splice variant, characterized by a similar deletion that affects blades five and six, was also not found to be substantially different in γ-binding preference from β₃[8]. However, the β₃v splice variant, which lacks the last three blades of the β torus and has a unique C-terminal region, dimerizes with only the γ₃ and γ₁₂ isoforms [9]. The β₃s and β₃s2 splice variants have been found to be associated with a C825T polymorphism in the β₃ gene that is associated with increased risk of hypertension [99]. Functional studies on the β₃s protein suggest that the loss of 41 amino acids results in enhanced receptor-dependent G protein signaling [7], which may contribute to the etiology of hypertension.

β₄ Interactions with γ Isoforms

The β₄ isoform has been shown to interact similarly with all γ isoforms in precipitation experiments from in vitro expression systems [94, 100]. However, when βγ dimers were precipitated from bovine lung using specific γ antibodies, β₄ clearly showed a preference for association with γ₅ and γ₁₂ over γ₂ and γ₃[101]. This is a good example of the difference between what βγ dimers can form in vitro and which dimers do actually form in vivo. Factors that influence such specificity in βγ formation may include chaperone proteins, which have recently been suggested to play a role in dimer formation between specific β and γ isoforms. For example, the Cytosolic Chaperonin Complex (CCT), which has been demonstrated to bind G β during βγ dimer biosynthesis, interacts most strongly with the β₁ and β₄ subunits, weakly with β₅ and β₃s, and intermediately with β₂ and β₃[102]. In addition, the Dopamine Receptor-interacting Protein 78 (DRiP78) has been characterized as a chaperone for G γ, and also exhibits specificity in binding with highest affinity to γ₂ and γ₃, compared to γ₁, γ₇ and γ₁₁[103]. More detail on this emerging field as it relates to βγ dimer formation can be found in a recent review by Willardson and Howlett [104].

β₅ Interactions with γ Isoforms

β₅ occupies a special niche in the G β family both structurally and functionally. Early in its characterization, it was noted that dimers containing β₅ and γ₂[105], or any of several other γ isoforms [M.B. Jones, unpubl. observation] were highly unstable in certain detergents; this observation will be discussed further below. It was also observed that in addition to G γ subunits, the β₅ isoform could bind to members of the R7 subfamily of Regulators of G Protein Signaling (RGS) proteins, and the crystal structure of a complex of β₅ and RGS9 was recently solved [48]. The β₅RGS literature is beyond the scope of this review, but see Berman and Gilman [4] and De Vries et al. [106] for other reviews. In terms of forming a conventional βγ dimer, β₅ was shown to interact preferentially with γ₄ by the yeast two-hybrid system [91], and live cell-imaging techniques suggested β₅γ₂ and β₅γ₇ are favored dimer combinations [107], although in the same system, β₅ could interact with an RGS7 protein as well. The specificity observed in formation of βγ dimers with different β and γ isoforms suggests the capacity for extensive modulation of G protein signaling. Further, the fact that the β₅ isoform can bind γ subunits, or alternatively members of the R7 subfamily of RGS proteins, indicates that other signaling pathways may be highly integrated into receptor-dependent βγ signaling.

Specificity of α/βγ Interactions

The G α isoform is necessary for the activation of a G βγ dimer, since specific localization of βγ to the plasma membrane has been reported to require both prenylation and heterotrimer formation [26]. Further, all three subunits are required for transfer of signal from activated receptor to G protein [108]. In addition, the identity of the Gα subunit may also determine the cellular mobility of an activated βγ dimer via its propensity to undergo subunit dissociation upon activation (discussed above). Considering the large number of α:β:γ subunit combinations that could potentially form, relatively little is known as to which heterotrimers can form, not to mention which actually do form in vivo. Limited subunit expression has provided information on probable heterotrimer combinations in different physiological systems. For example, in the visual system, G_t αβ₁γ₁ is likely to predominate in rods, whereas G_t αβ₃γ₈ is most prevalent in cones; in taste receptor cells, the most abundant isoforms are α-gustducin, β₁, β₃ and γ₁₃[109].

Regulation of Heterotrimer Composition

From empirical observations, many studies characterizing various G α isoforms have used β₁γ₂ as the archetypal βγ dimer. The β₄γ₂ dimer was also observed to form a heterotrimer with G_OA α [110]. Specificity has been reported in the βγ dimers associated with different G_O isoforms purified from bovine brain, with the G_OA heterotrimers containing much more γ₇ than the G_OC heterotrimers [111]; G_OC α is distinguished structurally from G_OA α by deamidation of Asn346 and Asn347 at the C-terminus [112]. However, βγ purified from G_OA interacted equally well with the α subunits purified from G_OA and G_OC, as judged by γ₇ immunoreactivity [113], suggesting that in this case, differences in the βγ composition observed between G_OA and G_OC were due to restricted expression of isoforms within cells or tissues. Transcriptional regulation is another mechanism that may regulate combinations of G α and βγ dimers; one study noted that G β₄ mRNA was regulated by expression of G_s α, G_i3 α and G₁₁ α [114]. Reduction of G_olf α protein was also correlated with deletion of the G γ₇ gene in mice [115]. Furthermore, genetic deletion of the γ₁ gene in mice resulted in a greater than 25-fold reduction in G_t α and β₁ protein levels in the retina, although interestingly, mRNA levels for G_t α and β₁ were similar to wild-type mice [116]. These studies suggest that heterotrimer composition and formation are highly regulated at the levels of transcription, translation and posttranslational processing.

In one study that used immunofluorescence microscopy to compare the ability of a panel of βγ dimers to target mutant G_s α and G_q α to plasma membranes as an indication of heterotrimer formation, βγ dimers containing β₁ or β₂ were equally effective at interacting with either G_s or G_q α isoforms [117]; βγ dimers containing β₃ did not interact well with G_s α or G_q α, and βγ dimers containing β₄ were able to interact with G_s α, but not G_q α. Interactions between G_s α or G_q α and βγ dimers containing G β₅ were not observed [117]. On the other hand, a study that used live cell-imaging techniques observed that both G_O α and G_q α could target β₅γ₂ to the plasma membrane [107]. Moreover, there is also a report of G_q α from brain extract binding a β₅γ₂ affinity column [118]. The presence of a receptor may also facilitate interactions between specific G α and βγ isoforms. For example, in contrast to co-localization studies with only G protein subunits [117], purified β₄γ₂ was able to couple G_q α to the M₁ muscarinic receptor [119], and β₃γ₂ was able to couple G_s α to the β₁-adrenergic and adenosine A_2A receptors [98]. Furthermore, the β₅γ₂ dimer has also been demonstrated to couple G_q α to the M₁ muscarinic receptor [120], and weakly couple G_s α to the β₁-adrenergic receptor [98].

Receptor-Dependent Translocation of βγ

The βγ dimer has been shown to translocate upon receptor stimulation. In perfused rat hearts, stimulation of the β₁-adrenergic receptor induced the translocation of the β₃ subunit from cytosol to membranes [121]; no such effect was observed for the β₁ or β₂ subunits, which were predominantly in the membrane fraction. It should be noted that the γ isoforms identified with the β₁, β₂ and β₃ subunits were not characterized, and thus it is possible that both β and γ isoforms contributed to the translocation of β₃γ.

The γ subunit is also a determinant of which βγ dimers undergo receptor-dependent translocation. Live cell-imaging experiments revealed that βγ dimers containing γ₁, γ₁₁ and γ₉ translocate rapidly to the Golgi membranes upon receptor stimulation, βγ dimers containing γ₅ and γ₁₀ translocate slowly, and βγ dimers containing γ₂, γ₃, γ₄, γ₇, γ₈ and γ₁₂ do not translocate [122, 123]; the translocation was observed to be reversed upon addition of a receptor antagonist. Interestingly, although γ₁, γ₁₁ and γ₉ are all farnesylated, the geranylgeranylated γ₁₃ also translocated rapidly to the endoplasmic reticulum [123]. This behavior of γ₁₃ reflects the study's finding that βγ translocation occurs as a reversible, diffusion-mediated process that is related to the amino acid sequence of the γ isoform, and not the identity of the prenyl group [123]. Within the family of βγ dimers that translocate, the α subunit also has influence on the rate of translocation, primarily related to the nucleotide exchange rates of the α subunits [124]. One important point to make with respect to receptor-dependent βγ translocation is that it can occur independently of G α [122], and thus represents an example of complete G protein subunit dissociation (see discussion above) determined by the nature of the γ isoform in a heterotrimer. Subsequent studies have observed that even in the absence of receptor activation, there is a basal level of heterotrimer shuttling between plasma and intracellular membranes [125]. The discovery of receptor-dependent βγ translocation is important in that it increases the complexity of the spatial dimension to βγ signaling, and suggests that cellular localization of βγ dimers after G protein activation is tightly controlled by the identity of the β and γ isoforms in a G protein.

Exceptions to the Rule

Receptor-Independent G Protein Activation

The review to this point has discussed the conventional wisdom regarding the molecular determinants that influence activation of a βγ dimer, which is usually preceded by partial or complete dissociation from G α. As discussed above, the βγ dimer is capable of dynamic translocation in the absence of G α, suggesting that activity may not be limited to the plasma membrane. Activation of a βγ dimer has been proposed to occur in the absence of receptor, or nucleotide exchange, via direct interaction by an Activator of G protein Signaling protein, AGS8 [126]; this activation was also suggested to not require subunit dissociation [127]. Another mechanism proposed for G protein activation is the absence of receptor involves phosphorylation of histidine 266 of β by NDPK; the phosphate is subsequently transferred onto the GDP bound to G α, effectively producing a GTP-bound activated α subunit without the requirement of receptor catalyzed release of GDP [15]. The discovery of G βγ in the endoplasmic reticulum (ER) of Arabidopsis led one researcher to speculate that βγ has signaling functions in the ER, such as regulation of PLC or IP₃ receptors [128], independent of heterotrimer formation with G α [129]. This is an intriguing hypothesis, considering that RACK1, a G β-like scaffolding protein, has been shown to bind both βγ [130] and IP₃ receptors [131]. Moreover, the ability of RACK1 to regulate βγ signaling, such as attenuation of PLC-β₂ activation [132], provides a potential mechanism for G βγ signaling to occur without G α or activated receptor.

Stability of βγ Dimers: Unconventional Roles for β…γ

Deviating further still from conventional wisdom is the biological significance of instability or low affinity between particular combinations of β and γ. The model of the tightly associated βγ dimer that couples a G α subunit to an activated receptor is based largely on studies with β₁γ₁ or β₁γ₂. However, other less stable combinations such as β₅γ₂ and β₄γ₁₁[119] are less amenable to biochemical studies, and thus, instability could easily be interpreted as incompatibility between particular combinations of β and γ isoforms. The high degree of heterogeneity inherent in the potential combinations of βγ dimers, and the spectrum of biophysical properties suggest that unconventional signaling paradigms for β and γ exist. It may be useful in this speculative exercise to first consider examples of potential dissociation between β and γ that have been reported in the literature.

βγ Instability in vitro

Examples of low affinity of β for γ can be found with many of the β isoforms. Lower affinity of β₃ for γ subunits has been reported in several systems examining βγ formation, such as in vitro translation and yeast two-hybrid [91, 92, 94], although after purification from Sf9 cells with γ₂, the β₃γ₂ dimer was completely stable in several different detergents [105]. More evidence exists for the instability of β₅ with γ subunits, and the biochemical properties of a β₅ monomer. Attempts to purify the β₅γ₂ dimer revealed that it was sensitive to detergent, and high concentrations of CHAPS or cholate induced subunit dissociation [105]. Other β₅γ dimers containing γ₁, γ₇, γ₁₀ and γ₁₂ could not be purified even at low (0.1% Genapol) detergent concentration [Miller B. Jones, doct. thesis]. Whereas the β₁ subunit forms unstable high-molecular-weight aggregates in the absence of γ [133], the β₅ subunit can exist as a monomer which is highly resistant tryptic cleavage [134].

Instances of βγ instability related to γ isoforms have also been reported. One early study characterizing the purification of G protein from bovine brain reported monomeric γ₃ under non-denaturing conditions [36]. The γ₁₁ isoform has also been noted for its propensity to dissociate from the β during purification [37], indicating a weak interaction. Parameters affecting G βγ₁₁ affinity were further explored with recombinant heterotrimers consisting of G_i1 α β₁γ₁₁ and G_i1 α β₄γ₁₁, both of which were stable upon purification; however, activation of the heterotrimers with GTPγS resulted in a significant dissociation of β₄ and γ₁₁, but not β₁ and γ₁₁[119]. Immunoprecipitation experiments examining βγ dimers containing the γ₁₃ isoform found that β₁, β₃ and β₄ could be immunoprecipitated with a hemagglutinin tagged γ₁₃ subunit, whereas the β₂ and β₅ subunits could not [135]; the authors speculated that the βγ instability was due to detergents used in the immunoprecipitation. This instability of β₂γ₁₃ and β₅γ₁₃ was not indicative of incompatibility between subunits, as these dimers were found to be effective in the activation and inhibition, respectively, of GIRK1/4 channels in transfection experiments [135]. Further, live cell-imaging techniques were used to observe that β₂ and γ₁₃ could effectively translocate together following receptor activation [123].

Potential Biological Activity of β and γ Monomers

Biochemical evidence of instability in particular combinations of βγ dimers in vitro logically leads to questions of whether monomeric β or γ subunits exist, and what they may be doing in a cell. Overexpression of two of the β₃ splice variants, β₃s and β₃s2, was shown to markedly stimulate MAP-kinase activation [8]; interestingly, co-expression of known γ partners for these β₃ splice variants had no further effect on MAP-kinase activity. These data suggest that the β₃ splice variants may have biological activity in the absence of γ. Moreover, the authors of the study offer the possibility that another β₃ splice variant with restricted γ-binding partners, β₃v, may exist as a monomer [9]. One study established that purified β₅ and γ₂ monomers can be reconstituted to form a dimer with the ability to active PLC-β₂ in vitro [134]; the study further demonstrated that monomeric β₅ was able to functionally interact with G_i α and G_O α. Another report characterized the ability γ₂ expressed in Sf9 cells to bind to a G_O α column, and protect the α subunit from tryptic cleavage [136]. Functional activity of a monomeric γ₅ subunit was also suggested in a study that demonstrated the ability of a bacterially expressed non-prenylated γ₅ subunit to regulate transcription by binding to the adipocyte enhancer-binding protein (AEBP1) transcriptional repressor [34]. Although β protein was also observed in co-immunoprecipitation studies with γ₅ and AEBP1 in mammalian cells, the data support the idea of transcriptional activity of a γ₅ monomer, perhaps in addition to βγ₅ dimers in the nucleus. Absence of prenylation, as discussed above, may not be the only signal to direct particular βγ dimers to the nucleus; a study examining the localization of fluorescently tagged β₅ co-expressed in HEK-293 cells with various fluorescently tagged γ isoforms concluded that the γ₁, γ₅, γ₁₀ and γ₁₁ isoforms targeted β₅ to the nucleus, whereas γ₂ and γ₇ did not [107]. Since most of the γ isoforms that targeted β₅ to the nucleus could not be purified as a complex with β₅ even under conditions of low detergent stringency (see above), the observed differences among β₅γ dimers in the context of localization suggests that dimer instability is related to βγ signaling roles.

Receptor:γ Interactions

Another potential biological role for γ subunits can be inferred from studies characterizing C-terminal γ peptides with activated receptor. A C-terminal farnesylated γ₁ peptide was reported to stabilize the active form of rhodopsin [56], and a C-terminal geranylgeranylated peptide corresponding to γ₅, but not γ₇ or γ₁₂, was able to inhibit M₂ muscarinic receptor signaling [137]. Further receptor-binding experiments revealed that the γ₅ peptide was able to stabilize a novel state of the M2 muscarinic receptor [138]. These studies suggest an interaction between the C-terminal tail of γ and receptors. However, the notion that receptor may not interact with the C-terminal tail of γ during G protein signaling was supported by reconstitution studies that found that a β₁γ₁ dimer engineered with a photoreactive farnesyl analogue was cross-linked to phospholipid, but not receptor, after reconstitution of transducin with rhodopsin in membranes [139]. The discrepancy between peptide and whole protein studies was also highlighted by the fact that the M₂ muscarinic receptor more efficiently stimulated GTP hydrolysis in a G_Oαβ₁γ₇ heterotrimer compared to a G_Oαβ₁γ₅ heterotrimer [140]. This raises the possibility that γ peptides may have properties distinct from the analogous βγ dimer at receptor, and more importantly, begs the question regarding the biological activity of a monomeric γ subunit: If a γ peptide can functionally interact with receptor, would a γ subunit, which has also been reported to exist as a monomer in the absence of β or detergent [141], have similar binding and regulatory properties at receptor?

Receptor:β Interactions

Less complex biological systems may also inform the prospective roles of β and γ in G protein signaling. For example, in fission yeast Schizosaccharomyces pombe, the β subunit, which lacks an amino terminal coiled-coil domain, retains some activity when its γ partner is deleted [142]; further, when the C-terminal CAAX box from the γ subunit is fused onto the C-terminus of the β subunit, some activity is recovered, suggesting more of a targeting role for γ in that system. This result was mirrored in studies with mutant yeast γ subunits in Saccharomyces cerevisiae, which found that a C-terminal domain preceding the CAAX box was not required for βγ coupling to receptor in vivo [143]. Recently, another study in S. cerevisiae also found that the RACK1 ortholog Asc1 could bind and influence the nucleotide-binding properties of Gα, and essentially perform the role of a β subunit, presumably without the presence of γ [144].

There are several instances that suggest that interactions between β and γ and receptor in higher vertebrates may be more dynamic than previously thought. Expression of β₃s, but not β₃, was shown to be able to activate Gα subunits in the presence of mastoparan-7 in digitonin-permeabilized COS-7 cells [7]; the authors reasoned that the β₃s protein dimerized with endogenous γ subunits. Such an experiment suggests the existence of monomeric β and γ subunits in vivo, because either there is biological activity of an expressed β₃ splice variant, or alternatively, a pool of γ isoforms ready to dimerize with an ectopically expressed β subunit. Moreover, purification of βγ from rod outer segments of Bufo marinus yielded a small population of free β with no γ partner [145]; interestingly, the β subunit, noted for its high homology with bovine β₁, was as effective as the purified βγ at stimulating GTPγS binding to Gα after reconstitution with illuminated rod outer segment disc membranes. Possibly related to the results of β₁ activity in B. marinus, a γ₁ knockout mouse retained some ability of rhodopsin to signal through G_t in rod outer segments with only residual amount of G_t α and β₁ remaining [116]. Since no upregulation of other β or γ isoforms was observed, and no other γ isoforms were detected in immunoprecipitations of β₁, one possible explanation (not advanced by the authors) is that the residual β₁ may be able to substitute on some level for the β₁γ₁ dimer. Although examples of β coupling G α to receptor likely represent a small minority of signaling paradigms in higher vertebrates, they may shed light on the interactions of unstable βγ dimer combinations with receptors and effectors, and prove useful in providing a way of dissecting the contribution of β and γ to G protein signaling.

Receptor/βγ Interactions Determine Signaling Specificity

Specificity of receptor:G protein interactions can be influenced by the identity of each of the subunit isoforms. Thus, the assembly of a βγ dimer with an α subunit to form a heterotrimer of defined composition within a cell functions to impart a degree of specificity by limiting the number of signaling pathways in which the G protein can directly participate. The first evidence suggesting that the identity of the β and γ isoforms in a heterotrimer can influence receptor coupling were based on experiments using antisense oligonucleotides to attenuate expression of specific subunit isoforms. Inhibition of voltage-sensitive Ca²⁺ channels in rat GH₃ cells was reported to be mediated via coupling of the G_O1 αβ₃γ₄ heterotrimer to the muscarinic receptor and coupling of the G_O2αβ₁γ₃ heterotrimer to the somatostatin receptor [146,147,148]. The antisense approach was further employed to suggest that in rat portal vein myocytes, the angiotensin AT_1A receptor couples to G₁₃αβ₁γ₃[149], and the ET_A receptor couples to G₁₁ αβ₃γ₅[150].

Prenyl Status

The type of prenyl moiety on the γ subunit has also been shown to be critical for receptor G protein interactions. Experiments involving rhodopsin have generated the most conflicting reports; on the one hand, the farnesyl group was observed to interact more efficiently than the geranylgeranyl group with activated rhodopsin [151, 152]. Alternatively, a mutated β₁γ₁ dimer that incorporated the geranylgeranyl group instead of the farnesyl group was found to have a 3-fold higher affinity for rhodopsin than wild-type β₁γ₁[153]. These results were mirrored in other studies that characterized β₁γ₁ and β₁γ₂ dimers that were mutated to alter the specificity of prenylation; farneslyation of β₁γ₂ reduced its affinity for the adenosine A₁ receptor, and conversely, geranylgeranylation of β₁γ₁ increased its ability to couple to the same receptor [154]. Farnesylation, however, does not always result in a decreased affinity of βγ for receptor. In the case of β₁γ₁₁, which like β₁γ₁ is farnesylated, high receptor coupling efficiency was observed with the α_2A-adrenergic receptor [155], adenosine A₁ and 5-HT_1A receptors [156] and the M₁ muscarinic receptor [119]. These results suggest that the differences between farnesylation and geranylgeranylation of βγ dimers reflect more than degrees of hydrophobicity of a membrane anchor, and the identity of the prenyl group is intimately related to activation of a G protein, and hence βγ, by receptor.

γ Isoform Specificity

Thus, the primary sequence of γ is also critical for determining the efficiency of receptor:G protein interactions. This point was borne out in studies with chimeras of γ₁ and γ₂, which concluded that the C-terminal third of γ, along with the type of prenyl group, is particularly important at determining the affinity of G protein receptor interactions [58, 153]. Studies with peptides constructed with the C-terminal sequence of a geranylgeranylated γ₅ subunit reached similar conclusions after these peptides were found to inhibit M₂ muscarinic receptor signaling [137]. The γ₅ C-terminal sequence apparently confers specificity in receptor:G protein interactions, as neither γ₇ nor γ₁₂ C-terminal peptides were able to inhibit muscarinic receptor signaling [137]. A specific signaling role for γ₇ has however, been assigned to the β-adrenergic receptor using a ribozyme strategy to attenuate γ₇ mRNA and protein levels [157, 158]; the β₁ isoform was also suggested as the partner for γ₇ in this signaling cascade, as β₁ protein levels fell in response to loss of γ₇ protein [158]. The ribozyme approach was also used to demonstrate that the β₁ and γ₇ isoforms are involved in activation of adenylyl cyclase via coupling to the D₁ dopamine receptor, but not the D₅ dopamine receptor [159]. The identity of γ also influences how a β isoform can interact with receptors. For example, the β₁γ₅ dimer couples poorly to the α_2A-adrenergic receptor, whereas the β₃γ₅ dimer couples effectively; however, when the γ isoform is changed to γ₁₁, the resulting β₁γ₁₁ dimer couples almost as well as β₃γ₅[155], suggesting that the receptor coupling efficiency of β or γ isoforms can be interrelated.

β Isoform Specificity

The identity of the β isoform has also been shown to influence interactions between receptor and G proteins. Attenuation of specific β isoforms using RNAi has indicated that the β₂ isoform, but not β₁, is involved in C5a-mediated chemotaxis in mouse macrophages [160]. In reconstitution experiments, the M₂ muscarinic receptor was more efficient in catalyzing nucleotide exchange with an G_O αβ₄γ₂ heterotrimer compared to a G_O αβ₁γ₂ heterotrimer [161]. The β₄ subunit as a dimer with γ₂ was also noted to have a greater ability to couple G_s α to the adenosine A_2A receptor than β₁γ₂[98], and shift a larger percentage of A_2A receptors into the high affinity binding state than β₁γ₂[162]. Interestingly, the inflammatory cytokines IL-1 and TNF-α were demonstrated to upregulate G β₄, but not G β₁ mRNA levels in human dermal microvascular endothelialcells [163]. Figure 2 illustrates how this regulation of G β₄, and the observed differences in adenosine A_2A receptor coupling between β₁γ₂ and β₄γ₂ may affect second messenger levels in cells. If the β₁γ₂ dimer predominates over β₄γ₂, G_s α may couple poorly to the adenosine A_2A receptor; however, an inflammatory stimulus could produce cytokines that elevate G β₄ levels, thus increasing the availability of β₄γ₂ and enhancing the ability of G_s α to couple to the adenosine A_2A receptor, activate adenylyl cyclase and raise cAMP levels.

Fig. 2.

Effect of different βγ isoforms on A_2a adenosine receptor signaling. β₁γ₂ couples G_s α poorly to the A_2a receptor, leading to lower activation of adenylyl cyclase; increasing levels of β₄γ₂, which couples G_s α more efficiently to the A_2a receptor, increases adenylyl cyclase activation and intracellular cAMP levels.

Conclusion

In conclusion, in the scope of this review, there are multiple factors that determine the activation of a particular βγ dimer within a cell: (1) Transcription, translation and posttranslational processing of specific β and γ isoforms, together with protein:protein interactions determine the constellation and potential localization of βγ dimer combinations. (2) Expression of specific G α subunit isoforms, likely integrated to specific β and γ isoform expression, determines the identity of heterotrimer combinations, thus affecting interactions with receptors, and the degree to which a G protein may dissociate upon activation, and potentially reform with different G α isoforms. (3) Expression and activation of specific receptors, which through preferences for binding particular heterotrimer combinations, ultimately starts the signal which leads to activation of specific βγ dimers. The crystallization of a receptor G protein complex would greatly enhance our knowledge of the specificity of receptor:α:βγ interactions.