I. INTRODUCTION

Top Previous Next References

More than a decade has passed since the highly cited review of G protein signaling by Dr. Alfred G. Gliman was published (188). Although the central elements of G protein signaling remain the cell-surface receptors coupled to G proteins, the family of nearly 20 heterotrimeric G proteins, and the ever-expanding, diverse groups of effector units [e.g., adenylyl cyclases (AC), phospholipases, various ion channels], detailing the physiological aspects of signaling through this pathways continues to stimulate and challenge us. A Medline search of the topic of G protein signaling including publications since the Gilman review reveals more than 10,000 citations relevant to the structure and function of members of the three cardinal elements of G protein signaling and how they relate to physiology as well as pathophysiology. Needless to say, the daunting challenge of covering adequately this volume of highly regarded scientific literature is surpassed only by the challenge of limiting the number of citations to 400-500 articles. Let us be explicit in stating at the outset that this review is not meant as a comprehensive analysis of all that has been published on G protein signaling in the last decade, but rather as a snapshot of a complex and large, work in progress. We shall try to highlight the nature of the basic elements constituting G protein-linked signaling and to construct a consensus for understanding how physiological regulation occurs using transcriptional, posttranscriptional, and posttranslational mechanisms well known to the molecular and cell biologist.

In an effort to anticipate criticism of our inability to include all the most relevant articles, we have created many figures and tables, richly annotated with references. Careful analysis of the citations will reveal the inclusion of many references as the most complete, and perhaps up-to-date review of the specific topic. We make no excuses that the systems adopted as "typical" for this review are highly familiar to the authors. It is true that many of the central themes described in the review could be expressed in a multitude of contexts. Because the superfamily of G protein-linked receptors will probably rise to somewhere between 500 and 1,000, few will agree with our forced selection of a given type or even subfamily of receptor for more detailed consideration of its physiological regulation. Perhaps more telling is that the challenging task of reviewing a broad field of literature does result in the emergence of "sound" from "noise," "pinnacles" from "peaks," and "timeless" from "timely" with regard to discoveries in this field. Wherever possible, these seminal discoveries and changes in how we understand or approach the study of physiological regulation of signaling via heterotrimeric G proteins are highlighted. The review is structured to illuminate the formidable tasks that lay ahead in exploitation of all that we have learned from studies of the structure and function of individual members of the troika that transduces cell membrane signals into a diverse downstream network of pathways involved in cell proliferation, differentiation, and apoptosis to a more unified understanding of the integration of these inputs.

	II. ELEMENTS OF G PROTEIN-LINKED SIGNALING

Top Previous Next References

Evolution of our understanding of G protein-linked receptors (GPLR) increased rapidly after the first molecular cloning of a well-known member of the GPLR that bind hormones. In 1986, the primary sequences of the hamster (139) and turkey (626) -adrenergic receptors were reported, providing long-awaited information that proved critical to understanding, classifying, and probing GPLR. Before 1986, what we knew about GPLR was garnered first from pharmacological studies, then radioligand binding studies, and later through heroic efforts in protein purification of these low-abundance membrane proteins (357, 529). The identification of seven hydrophobic sequences similar in length and in organization to that of bacteriorhodopsin permitted modeling of GPLR and the molecular tools with which to test many details of emerging hypotheses (Fig. 1). By the end of 1998, more than 300 members of the superfamily of hormone receptors linked to G proteins have been entered in the GenBank. If one also includes in the listing of GPLR the expanding subfamily of odorant receptors (69, 97), it is likely that ~1% of the mammalian genome encodes GPLR.

View larger version (78K): [in this window] [in a new window]   Fig. 1. Schematic of hypothetical organization a G protein-linked receptor, using mammalian 2-adrenergic receptor as a model. All G protein-linked receptors (GPLR) display 7 heptihelical domains that are hydrophobic and span lipid bilayer. Seven transmembrane segments are displayed, each identified with a roman numeral starting with most NH2 terminal designated as I and most COOH terminal as VII. Salient features of GPLR are exofacial NH2 terminus (usually N-glycosylated as shown), 7 transmembrane segments with intervening intracellular and extracellular "loops" (of variable length), and a cytoplasmic COOH terminus (often referred to as "COOH-terminal tail" of variable length). Extracellular domains and core of bundle of 7 transmembrane segments act in signal discrimination and ligand binding. Intracellular domains function in signal propagation to heterotrimeric G proteins and are substrates for phosphorylation by protein kinases.

Several useful approaches have been reported to the classification of GPLR. Molecular cloning provides a wealth of information, best managed as a database with automated data collection, structured flatfiles, and data query mechanisms. One such database was reported within the National Center for Biotechnology Information, first available in 1994 (307). Such databases are extremely useful tools, providing means to search homologies, to segregate paralogues (variants of receptors created by gene duplication) and orthologues (same GPLR, but different species), and to discern family, subfamilies, and other relationships among the members of this superfamily of receptors.

Using 10 residues conserved among more than a hundred different GPLR and analysis of sequence alignment, BIN maps have been created that can discriminate among the families based on the length of consecutive segments, termed "partitions" (387). Patterns of relationships develop that are provocative. The key residues are the NH₂-terminal Met termed "," the last nonconserved residue termed "," and the following, intervening conserved residues (in transmembrane-spanning regions, see Fig. 1): Asn (TMSR1), Asp (TMSR2), Cys (top of TMR3), Arg (bottom of TMSR3), Try (TMSR4), Cys (bottom loop between TMSR4 and TMSR5), Pro (TMSR5), Pro (TMSR6), and Pro (TMSR7). Although only a proposed method of analysis, BIN mapping may prove valuable in assigning identities to orphan receptors (387).

Based simply on the chemical nature of the ligand, GPLR may be classified into families and subfamilies (537). A simple, nonexhaustive listing of the prominent members of families is provided in Table 1. Table 1 reveals the marked diversity of the ligands/stimulants toward which GPLR apparently have evolved. Visual excitation operates via GPLR signaling, relying on the capture of a photon by the 9-cis-retinal covalently coupled via Lys-296 to the opsin photopigment rhodopsin (7, 86, 134, 455, 542). The physiology and pathophysiology of vision can be explored elsewhere (422). Gustatory signals (taste) and odorant signals (smell) likewise operate via GPLR (1, 60, 365), the odorant family likely representing the most populous of all GPLR families with membership expected in the range of 1,000-2,000 receptors (22). A genetic analysis of mammalian olfaction reveals many intriguing aspects of this GPLR signaling-based sensory physiology (457). Small molecule receptors in the classification include the most well-characterized receptors for biogenic amines, such as dopamine, epinephrine, and serotonin. Included in this group are both receptors for ATP found ubiquitously in mammals (350, 356) and for cAMP, a molecule that is central to the sensory biology of Dictyostelium discoidum, the slime mold (165, 254, 380). Thus GPLR mediate the actions of a chemically diverse family of ligands (including photons) and do so in a broad spectrum of organisms, from human to mold.

The adrenergic receptors typify the relationship that can exist among subfamilies in the GPLR superfamily (413). At least nine adrenergic receptor paralogues have been identified. In the Ahlquist era, a discrimination between - and -adrenergic agonists was discovered. Although pharmacology created both agonists and antagonists that collectively drove the subclassification further for both the - and -adrenergic ligands, few envisioned the revelations emerging from molecular cloning that provided no less than nine members to the adrenergic receptor gene subfamilies (535, 536). The -adrenergic receptor subfamily, initially composed of the ₁- and ₂-members, saw the number expand to three most recently, with the discovery of the ₃-adrenergic receptor (152). The -adrenergic subfamilies are composed of ₁- and ₂-adrenergic groups, with three gene products for each (₁ with A, B, and D; ₂ with A, B, and C) (440). Molecular cloning has revealed the existence of other subfamilies, such as the muscarinic acetylcholine receptor with five members (343, 406, 609) and the dopamine receptor subfamily with at least seven members (6, 88, 124, 274, 382).

The receptor subfamily for 5-hydroxytryptamine (5-HT or serotonin) provides an impressive display of diversity (Fig. 2) (88, 94, 103, 127, 218-220). There exists at least 13 members of the rat serotonin receptor subfamily, each the product of a separate gene (161). In some cases (5-HT_1B/D, 5-HT₂, and 5-HT₃) serotonin receptors were identified pharmacologically, before analysis by molecular cloning (219). Many, however, were generated by molecular screening of DNA libraries under low-stringency hybridization conditions. Although the sequence homology chart provides information on the extent to which various members of the subfamily of serotonin receptors display homology at the level of protein sequence, they do not reveal information on the nature of the signal propagation. Combined expression and characterization of the cloned genes will enable a methodical approach to understanding the basis and need for the many members of this subfamily of GPLR.

View larger version (19K): [in this window] [in a new window]   Fig. 2. Analysis of relative sequence homologies among members of family of GPLR that bind 5-hydroxytryptamine (5-HT). Note diversity that exists among family member groupings. Highest homology levels are observed among members of subgroups.

The ₂-adrenergic receptor provides an excellent model from which to discuss the general structure of GPLR (530, 531, 535). The primary sequence information deduced from the molecular cloning (139) revealed the presence of seven domains of the sequence, rich in hydrophobic residues and spanning 22-28 residues in length (Fig. 1). On the basis of the hydrodynamic and detergent solubility data, it was speculated that these hydrophobic domains of the receptor spanned the lipid bilayer of the membrane, organizing the receptor in a fashion that resembled the deduced structure of bacteriorhodopsin (164, 226, 469). Subsequent biochemical analysis and topographical analysis by site-directed, antipeptide antibodies established the NH₂-terminal domain and three segments intervening between the membrane-spanning domains as exofacial, whereas three additional intervening segments as well as the COOH-terminal tail of the receptor were localized to the cytoplasmic face of the lipid bilayer (Fig. 3) (17, 589-591). Studies of N-glycosylation sites of the ₂-adrenergic receptor confirmed the location of the NH₂-terminal sequence as exofacial (181, 182). In a similar manner, study of the protein phosphorylation of GPLR confirmed the COOH-terminal tail and the segment between transmembrane-spanning regions 3 and 4, 5 and 6 display sites for phosphorylation by a variety of protein kinases, including protein kinase (PK) A (222), PKC (456), and members of the GPLR kinases, termed GRK (292, 348). The NH₂-terminal segments of GPLR vary in size from 7 to 595 amino acids and the COOH-terminal region from 12 to 359 amino acids. Loops vary in size from 12 to 359 amino acids. Nearly one-half of the known GPLR that act as receptors for hormones, autacoids, and neurotransmitters also possess a conserved protein motif, either Asp-Arg-Tyr or Glu-Arg-Trp located in the NH₂-terminal portion of the intracellular loop between transmembrane-spanning regions 3 and 4. G protein-linked receptors for calcitonin, CRH, growth hormone-releasing hormone, glucagon, and parathyroid hormone lack this motif, while retaining all of the other features discussed above (536).

View larger version (38K): [in this window] [in a new window]   Fig. 3. Topographical model of 2-adrenergic receptor, based on use of site-directed, antipeptide antibodies prepared to each of hydrophilic domains of receptor. Antibodies were prepared to synthetic peptides of hydrophilic domains and employed with indirect immunofluorescence in intact versus detergent-permeabilized A 431 epidermoid carcinoma cells in culture. Positive staining in intact cells was interpreted as extracellular, and positive stained confined only to permeabilized cells was interpreted as intracellular.

Although commonly displayed in the "serpentine" arrangement for the sake of simplicity, GPLR have been shown by a number of indirect but compelling data to be organized much like a basket composed of the transmembrane-spanning domains (187, 235, 235, 529, 590). Photoaffinity labeling with high-affinity antagonist ligands have revealed the close proximity of transmembrane segments 6 and 7 with segments 1 and 2 (Fig. 4). These receptors for monoamines invariably possess an acidic side chain of Asp in transmembrane-spanning segment 3 necessary for formation of a salt bridge with the amino group of the ligand (531). A hydrophobic pocket critical to binding is created by aromatic residues of transmembrane-spanning segments 6 and 7. The hydroxyl groups of the ligand family are believed to be hydrogen bonded through Ser and Thr residues in transmembrane-spanning segments 4 and 5. At least for the ₂-adrenergic receptor, these data and the overall similarity to bacteriorhodopsin (385) suggest the existence of a binding "pocket" in which ligands interact. For the photopigment rhodopsin, the retinal chromophore attached covalently to Lys-296 of transmembrane-spanning segment 7 appears to be buried ~20 Å into the lipid bilayer. The hydrophobic, tryptic core of the ₂-adrenergic receptor retains the ability to activate G_s in response to agonists (477). Fluorescence energy transfer measurements with the ₂-adrenergic receptor, likewise, suggest that the binding of antagonists occurs within the dimensions of the lipid bilayer (98, 167, 302, 338, 531), although establishing the depth to which ligands penetrate the lipid bilayer will require more precise analysis.

View larger version (94K): [in this window] [in a new window]   Fig. 4. Analysis of 2-adrenergic receptor ligand binding domain by use of various photoaffinity, radiolabeled antagonist ligands. Top: schematic of hypothetical interactions between transmembrane segments (TM) (1-7, in blue shown as -helixes) and iodoazidobenzylpindolol (IABP) (shown in green). Distances according to this model are displayed in green. Model accommodates photoaffinity labeling data that display interaction of Ser-207 with antagonist ligand. Bottom: model accommodates photoaffinity labeling data that display aryloxy oxygen of ligand and Asn-312. Initial alignment was based on that of bacteriorhodopsin, with transmembrane segments displayed as -helices. Molecular dynamic simulations were selected which allowed for alkylamine was within bonding distance of Asp-113, and indole amine was within bonding distance of Ser-204 and Ser-207.

Many GPLR bind protein ligands. The size and complexity of protein ligands for GPLR span from small peptides [thyrotropin-releasing hormone (TRH)] to large glycoproteins [luteinizing hormone (LH) and thyrotropin-stimulating hormone (TSH)] (619). Whereas the small molecule ligands may bind and activate their GPLR via a domain embedded in the bilayer, GPLR that bind protein ligands often display large NH₂-terminal, exofacial domains (100, 150, 225, 265, 381, 619). There is a positive correlation between the ligand size and the GPLR NH₂ terminus (266). Even more intriguing is the thrombin receptor, a substrate for the thrombin protesers carrying an NH₂-terminal sequence in which its ligand is embedded (8, 57, 123, 129, 585). Cleavage of the exofacial domain of the receptors yields an activating peptide ligand, tethered to the receptor in its basal state (57, 585). Other members of this novel class of protease-activated receptors (PAR) have been cloned (57), providing an additional dimension to our understanding of ways in which GPLR can be activated. This mechanism of activation, irreversible by nature, poses some interesting possibilities about the manner in which thrombin receptors and other members of the PAR family are activated, desensitized, and downregulated (56, 57, 129, 428, 559, 561).

By definition, all GPLR not only provide the discriminator activity for ligand binding but must propagate the activation to a G protein(s) via protein-protein interactions (187). First analyzed for the ₁- and ₂-adrenergic receptors, the NH₂-terminal and COOH-terminal extremes of the segment intervening between transmembrane-spanning regions 5 and 6 appear to play a prominent role in G protein interaction. Proteolytic cleavage products of native receptor (433, 477, 615) and forms of the receptor with mutations in these regions (530, 531) provided model systems in which to ascertain important G protein contact sites for GPLR. Mutagenesis not only identifies regions critical to G protein coupling by GPLR, but also reveal some residues that alter the very nature of the G proteins to which a GPLR couples (614, 615). The early work in the -adrenergic receptors has been since confirmed by similar studies in a wide variety of GPLR. Taken together, the data provide a compelling picture of receptor-G protein contact sites that transduce agonist binding and transformation to a receptor with high affinity for agonist into changes in G protein binding of and activation by GTP.

B.  Heterotrimeric G Proteins

1.  Family classification

G proteins are members of a superfamily of GTPases that are fundamentally conserved from bacteria to mammals and play diverse roles in many aspects of cell regulation (190). The family of receptor-coupled G proteins has a unique heterotrimeric composition, and there is structural and functional diversity among each of the three polypeptide components of a G protein heterotrimer (228, 533). In general, G proteins are classified by reference to their -subunits, although the newly appreciated regulatory roles for the tightly associated -dimers suggest that this nomenclature should be revised to account for the potentially complex variety of G protein heterotrimers that can be assembled from these distinct -, -, and -monomeric gene products (105, 107, 108, 189).

Agonist-liganded receptors associate with and promote activation of G proteins by stimulating release of GDP bound to the guanine nucleotide-binding site of the -subunit. Release of GDP is followed by GTP binding, causing dissociation of the heterotrimer into derivative substrate - and -dimer. The intrinsic GTPase activity of the -subunit determines the lifetime of this active (dissociated) state of the G protein. Hydrolysis of bound GTP to GDP allows the -subunit to reassociate with the -dimer, ready for another round of receptor-regulated activation (48). Both the - and - subunits can regulate G protein-coupled effectors in a selective manner that can be either independent, synergistic, or antagonistic (408). The GTPase activity of G and its association with -subunits are both regulated by accessory proteins. The accessory proteins involved, "regulators of G protein signaling" have been afforded the acronym RGS proteins. The RGS proteins may bind to G_i subunits, for example, accelerating the rate of GTP hydrolysis (30, 32, 33, 135). The G subunits are regulated by the protein phosducin, which binds G tightly, preventing their interaction with G and effectors (39). Twenty years of biochemical and genetic studies of the heterotrimeric G proteins have recently culminated in determination of the three-dimensional structure of several G protein -subunits in both their GDP- and GTP-liganded states and of a G protein -heterotrimer (30, 111, 113, 298, 323, 324, 384, 515, 516, 522, 540, 557, 587). These new structural data are treated in a later segment.

G subunits are extrinsic membrane proteins. Lipid modification generally involving myristoylation or palmitoylation at the NH₂ terminus is essential for membrane anchoring of these proteins (76, 84). Posttranslational features of G subunit function are detailed in section V.

Structural and functional classification of G protein oligomers has been defined by the -subunits (see Table 2). As might be expected of proteins that perform certain highly conserved functions (for example, association with activated hormone receptors, GTP binding, and hydrolysis as well as association with -dimers), the primary sequence of all known G subunits contains ~20% invariant conserved amino acids. Outside of these regions, the sequences of the G proteins diverge. Four families of these proteins, termed s, i, q, and 12/13, have been proposed based on amino acid sequence comparisons (228, 611).

View this table: [in this window] [in a new window]   Table 2. G protein -subunits

The G_s class contains G_s and G_olf, which are 88% identical (273). Both proteins activate AC and are substrates for ADP-ribosylation catalyzed by the A₁ protomer of a toxin elaborated by Vibrio cholera (i.e., cholera toxin). This posttranslational modification inhibits the intrinsic GTPase activity of the G proteins (273) (see sect. VC).

The G_i class contains G_i-1, G_i-2, G_i-3, the retinal G, G_t, two forms of the brain-specific G subunit G_o-1 and G_o-2, as well as G_z. All members of this class (with the exception of G_z) contain a conserved COOH-terminal cysteine residue that is the site of ADP-ribosylation catalyzed by a toxin elaborated by Bortadella pertussis (i.e., pertussis toxin). This irreversible, covalent modification uncouples the G protein from its activating receptor (see sect. VC). Blockade of cellular responses to stimulation by pertussis toxin treatment has been an effective experimental procedure employed to implicate this class of G subunits in specific cellular signaling processes. The G_t subunit activates retinal cGMP phosphodiesterase, the major effector in vertebrate phototransduction. Members of the G_i and G_o subfamilies are implicated in the regulation of ion channel activity and regulation of phospholipase (PL) C, whereas the function of G_z is not known (85, 349).

The G_q class contains five family members, G₁₁, G₁₄, G₁₅, G₁₆, and G_q. These closely related proteins are substrates for neither cholera toxin- nor pertussis toxin-catalyzed ADP-ribosylation. The G_q subunits are notable regulators of the -class of phosphoinositide-specific PLC (PLC-). G_q and G₁₁ are widely expressed in mammalian tissues. The expression other members of the G_q class, in contrast, is restricted to stromal and epithelial cells as well as to cells of the hematopoietic lineage. These G subunits also activate PLC- isoforms and may exhibit a preference for members of the PLC-₂ family that also display a similarly restricted pattern of expression (4, 10, 200, 327, 328, 416, 512, 513).

The final class of pertussis toxin- and cholera toxin-resistant G subunits contains two proteins, G₁₂ and G₁₃. These -subunits are widely expressed (510). The functions of G₁₂ and G₁₃ have not been clearly defined. Overexpression of activated forms of these proteins transforms fibroblasts (312). Expression and activation of G₁₂ and G₁₃ occurs in differentiation of P19 embryonic stem cells in response to retinoic acid (264). Activation of G₁₃ leads to selective activation of mitogen-activated protein kinases, especially jun NH₂-terminal kinases (263). Other data implicate G₁₂ and G₁₃ in regulation of the Na⁺/Cl antiporter activity (9, 414, 534).

The realization that the G subunits play direct roles in regulation of effectors has focused attention on these tightly associated -dimers (see Table 3). Five distinct members of the G family have been identified (595). Each -subunit displays 340 amino acid residues and a molecular mass of ~35,000 Da. The linear sequence of these proteins consists of seven or eight tandem-repeats with a central conserved Trp-Asp sequence that has been termed a "WD-40" motif (178). The G family are more divergent, displaying seven family members to date, at least six of which are mammalian isoforms. These -subunits vary in molecular masses from 7.3 to 8.5 kDa and are considerably more diverse in primary sequence than the G subunits (30, 107, 253, 294, 408, 491). The predicted COOH-terminal sequences of all G subunits contain the sequence CysAAX, where A is any aliphatic amino acids. The G subunits are modified by prenylation of the Cys residue by removal of the final three amino acids by proteolysis, and then by carboxymethylation of the newly generated COOH-terminal Cys. The retinal G₁ subunit is modified by farnesylation, whereas the other mammalian G subunits all are modified by geranylgeranylation (see sect. VB). This acyl chain modification is necessary for association of the G dimer with the lipid bilayer. G and G subunits are tightly associated and cannot be dissociated except under strongly denaturing conditions. Acylation of G is not required for assembly of the G dimer, although the mechanisms by which these proteins are assembled posttranslationally are not known. Acylation does appear to play critical roles both in membrane association of the G dimer, in the association of G with G, and in interactions with AC (175, 195, 253, 279, 569).

View this table: [in this window] [in a new window]   Table 3. G subunits

Structures of G_t and G_i have been solved in their GTP-, GDP- and AlF₄-liganded states (30, 111, 113, 298, 323, 516). A mutant of G_i-1 in which Gly-203 is substituted with Ala has been solved in its GDP and phosphate-bound state. The structures of G complexes with G_i-1 and phosducin have been described (454), and a very recent study reports the structure of the heterotrimeric catalytic core of AC complexed with G_s (558) (Fig. 6). These structures have provided invaluable insight into the mechanisms by which G proteins interact with and are oriented relative to cell membranes and their cognate receptors, as well as the processes of receptor-catalyzed guanine nucleotide exchange, GTP hydrolysis, and effector activation.

G subunits are composed of essentially two distinct domains, a Ras-like GTPase domain and a predominantly helical domain that is unique to the G subunits. The bound guanine nucleotide is held at the interface of these domains. Three switch regions within the GTPase domain (switches I, II, and III) change conformation in response to the guanine nucleotide-liganded state of the G subunit, and significantly, all three switch regions form significant contacts with G and effectors. In the GTP-bound state, the switch regions are held in place by contacts to the terminal -phosphate of the nucleotide, whereas these regions appear to be less ordered (more flexible) in crystals of the GDP-liganded G proteins (Fig. 5).

View larger version (56K): [in this window] [in a new window]   Fig. 5. Crystal structure of GTP and GDP-liganded forms of a heterotrimeric G protein -subunit. Structures of guanosine 5'-O-(3-thiotriphosphate) (GTPS)- (A) and GDP-liganded Gi-1 (B) are shown with 3 switch regions (S1, S2, and S3) arrowed. [Data from Coleman et al. (112) and Mixon et al. (384).]

Structures of two G dimers have been determined. G₁₁ (the retinal G species) has been crystallized alone and in combination with either a G subunit or the regulator phosducin (515). A second G dimer, G₁₂, has been solved in the G-bound state (587). The G structure is dominated by a -propeller, which is a structural motif found in a number of different proteins (522) (Fig. 6). This motif is composed of seven repeats of four-stranded -sheets that are arranged around a small central hole. This prominent symmetry arises from the internally repeated WD-40 motif found in the G subunits and a functionally diverse group of proteins. G subunits contain a membrane-anchoring COOH-terminal farnesyl group. G associates tightly with G through a coiled-coil structure. The presence of a region of sequence enriched in basic amino acids on the side of the -propeller abutting this coiled-coil region suggests a possible role in an electrostatic interaction with membrane lipids. It is noteworthy that the G₁₁ and G₁₂ structures differ in the relative orientation of the coiled-coil, which may reflect differences in interactions either with receptors or with effectors, or perhaps with both.

View larger version (38K): [in this window] [in a new window]   Fig. 6. Crystal structure of a G protein heterotrimer. Structure of G protein heterotrimer Gi-1·GDP12 is shown with NH2-terminal -helix that forms major interface with -dimer highlighted. [Data from Wall et al. (587).]

The structures of the individual components of a G protein heterotrimer have been augmented by descriptions of the structures of two G complexes. The major interaction between G and the -complex involves the NH₂ terminus. These structures indicate that association with G produces a significant change in conformation of the switch I and II regions of the G protein (Fig. 6). The structure of G is not appreciably changed by association with G. This important observation suggests that receptor-promoted guanine nucleotide exchange alters the conformation of the switch II region of G, disrupting interactions that are critical for G binding (587). The structure of guanosine 5'-O-(3-thiotriphosphate) (GTPS)-liganded G_s in complex the catalytic domains of AC reveals a unique interaction between the switch II region of the G protein and its effector protein (Fig. 7) (558). A number of excellent reviews discuss these structures and the implications of the information they reveal for understanding G protein function (215, 522).

View larger version (39K): [in this window] [in a new window]   Fig. 7. Crystal structure of a G protein -subunit (Gs) in association with domains of effector adenylyl cyclase. Structure of G·GTPS complexed with C1 and C2 adenylyl cyclase catalytic domains. Structure is oriented to highlight major interaction interface between G protein and its effector. [Data from Tesmer et al. (558).]

C.  G Protein-Linked Effectors

1.  Classification

G protein - and -subunits regulate the activities of a structurally diverse group of effector molecules. These include enzymes engaged in the synthesis and degradation of intracellular second messengers, as well as ion-selective channels (see Table 4). In some cases, direct regulation of these effectors by G protein subunits has been demonstrated unequivocally by in vitro reconstitution, whereas in other instances, a timely example being the mitogen-activated protein kinases (MAP kinase) cascade, G protein regulation of the effectors may be indirect. We consider these classes of G protein effectors further below. It is also noteworthy that members of at least two classes of G protein effectors, the PLC- enzymes and the cGMP-phosphodiesterases, function as GTPase-activating proteins (RGS-like functions) for their G protein regulators.

Adenylyl cyclase catalyzes the formation of cAMP from the substrate Mg²⁺-ATP. These enzymes are expressed in a wide range of species from bacteria, yeasts, and slime molds to mammals. In general, AC are membrane-bound proteins, although certain of the bacterial and possibly mammalian enzymes are cytosolic. The yeast and bacterial AC are peripheral membrane proteins, whereas the Dictyostelium enzyme is an integral membrane protein with a single transmembrane spanning domain (79). The G protein-regulated AC isoforms identified in higher eukaryotes share a common motif (159, 511). These enzymes are large, single polypeptides (molecular masses in the range of 120 kDa). In general, mammalian AC appear to be proteins embedded in the plasma membrane. The cloning of the first of these enzymes revealed a deduced topology that comprises an apparently duplicated structure consisting of a short NH₂ terminus, two membrane-spanning regions of six -helices which link ~40-kDa cytoplasmic domains (termed domains C1 and C2). The helical portions of the transmembrane domains are linked by short extramembranous regions. This topological organization is unique among enzymes but common to several ATP-dependent transport proteins including the P-glycoprotein and cystic fibrosis transmembrane conductance regulator (18, 313). Whether or not these transmembrane AC isoforms function as membrane transport proteins remains a provocative question.

Complementary DNA encoding nine distinct mammalian AC isoforms have been cloned and expressed (95, 159, 176, 258, 549). These proteins are designated types IIX with evidence for existence of alternatively spliced transcripts of unknown significance, further adding to the complexity. Homologs have been identified both in Dictyostelium and Drosophila. The C1 and C2 domains of the mammalian AC are well conserved (50-90% identity). It is also noteworthy that the C1 and C2 domains are homologous to each other and to the catalytic domains of a number of membrane-bound guanylyl cyclases (603). Catalytic activity requires both the C1 and C2 domains, and point mutations in either of these two domains can inhibit catalytic activity. Soluble guanylyl cyclase isoforms function as heterodimers, and this also appears to be the case for the AC. Expression of either the NH₂- and COOH-terminal halves of the molecule or, remarkably, of the individual C1 and C2 domains devoid of their membrane anchors reconstitutes AC activity. These soluble AC are activated both by G_s and by the plant diterpene forskolin (547). Although neither the C1 nor C2 domains display sequences homologous to those involved in binding nucleotides, it is clear that the enzyme must interact with its substrates as well as with so-called "P-site" nucleotide regulators (272). That binding sites for these substrates and regulators are formed through the interaction between the two domains seems obvious based on such considerations. The recently described structure of a C2 domain homodimer revealed that forskolin and nucleotide substrates bind at the C1/C2 domain interface (558).

Where examined, the mammalian AC isoforms are generally widely expressed, although types I and II appear to be primarily neuronal, and type III is restricted to olfactory epithelium (258). All forms of AC identified in mammalian systems are stimulated by GTP-liganded G_s. The enzymes differ in their susceptibilities to regulation by -subunits, by members of the G_i class, by Ca²⁺/calmodulin, and by PKC. -Subunits are effective inhibitors of the type I enzyme but stimulate activity of the type II and type IV enzymes in a manner that is highly conditional on costimulation by G_s. It is noteworthy that as with other -subunit-dependent phenomena, stimulation of type II AC requires considerably higher concentrations of -subunits than activating concentrations of -subunits (548, 551, 552). Thus abundant G_i heterotrimers are likely to be the physiologically important source of -subunits for this mode of regulation of AC.

Although the G_i family was initially identified as the G proteins responsible for inhibition of AC activity, mechanisms proposed for the inhibitory mode of regulation have been the focus of intense debate. A failure to observe inhibition of adenylyl cyclase activity by isolated G_i led to the proposition that sequestration of G_s by -subunits might be the mechanism underpinning the inhibitory response. More recent investigations reveal that all three isoforms of G_i are equally effective inhibitors of the types V and VI enzymes. Type I AC is selectively inhibited by G_o, whereas the types I and V enzymes can be inhibited by G_z (550, 551, 554).

Intracellular Ca²⁺ is an important regulator of AC activity. The types I, VIII, and to a lesser extent III enzymes are potently activated by Ca²⁺/calmodulin, whereas the other known isoenzymes are rather insensitive to this activator (598). Studies using intact cells have implicated a number of phosphorylation-dependent mechanisms in control of AC activity. Of these, control by cAMP-dependent PKA and by PKC has been most widely studied. Although both protein kinases can phosphorylate certain AC isoforms in vitro, the effects on adenylyl cyclase activity observed are modest in comparison with those observed in intact cells. It seems likely that PKA and PKC control AC activity by an indirect mechanism(s) (604).

The distinct patterns of regulation of the AC isoforms suggest that these enzymes may function as integrators of G protein-mediated signaling pathways. Regulatory input from G_q-coupled receptors can control AC activity by Ca²⁺ or PKC-dependent processes. G_i can directly inhibit the types I, V, and VI enzymes, whereas G_i-derived -subunits can activate the type II enzyme in a conditional manner. All of the AC isoforms described to date continue to share the ability to be stimulated by GTP-loaded G_s.

The cGMP phosphodiesterase (PDE) plays a central role in visual excitation in vertebrate rod photoreceptor cells. Absorption of a photon by the photopigment rhodopsin leads to activation of the PDE which, in turn, catalyzes the hydrolysis of cGMP, causing closure of cGMP-gated cation channels and in hyperpolarization of the cell membrane (214). The cGMP PDE exists in both membrane-bound and soluble forms. Irrespective of localization, the protein is a heterotetramer composed of three distinct polypeptides P, P, and P with molecular masses of 88, 84, and 14 kDa, respectively, found in a molar ratio of 1:1:2. The P and P polypeptides both contain catalytic sites, whereas the P subunit is an inhibitor of their enzymatic activities (52). The retinal G_t protein mediates regulation of the PDE by rhodopsin. GTP-loaded G_t binds the P subunit, releasing inhibition of the P and P catalytic subunits. P is a GTPase-activating protein (GAP) for G_t, increasing its rate of GTP hydrolysis (12). Guanosine 5'-triphosphate hydrolysis releases the P subunit to reassociate with and to inhibit the catalytic subunits of the enzyme, completing the activation/deactivation cycle.

Three families of PLC enzymes classified as -, -, and - have been identified in mammalian systems. There are multiple isoenzymes within each class (464). The PLC enzymes share some common structural features including two regions of conserved sequence, termed "X" and "Y" domains, that contain residues important for substrate binding and catalysis. Outside of these sequences, the proteins diverge extensively. The four PLC- isoenzymes are targets for activation by G- and G subunits. Members of the G_q family of G protein -subunits couple receptors to activation of the PLC- enzymes in a pertussis toxin-insensitive manner (512, 555, 556). G protein regulation of the PLC- enzymes has been studied using both reconsititution assays with purified proteins and exogenously provided substrates as well as by transient transfection of COS-7 cells with vectors for the expression of the PLC enzymes and G protein subunits. Experiments using transient expression of G subunits and PLC- isoenzymes in COS-7 cells suggest that there are differences in susceptibility of the individual PLC- enzymes to activation by the various G_q family members (10, 416, 617, 618). However, experiments with purified PLC-₁, -₂, and -₃ show little difference in activation by purified G_q, G₁₁, and G₁₆ (229, 311, 513).

In many systems, receptor regulation of inositol lipid hydrolysis is sensitive to pertussis toxin (216). The observation that G protein -subunits activate the PLC- enzymes both in vitro and in transient transfection assays provides a biochemical explanation for this finding (54, 55, 78, 137, 247). -Subunits can activate all of the PLC- isoenzymes. Phospholipase C-₂ and PLC-₃ appear more sensitive to stimulation by -subunits than PLC-₁ and PLC-₄ (430, 463). In the case of PLC-₃, G_q and -subunits are approximately equally effective activators on a molar basis. For other members of the PLC- family, the _q-subunits are considerably more potent (50- to 100-fold) than G subunits.

The PLC- enzymes appear to function as GAP for their G protein regulators. The G_q subunits used in these activation experiments were liganded with nonhydrolyzable guanine nucleotide analogs (35). Under these conditions with GTPase-resistant GTP analogs, it is possible that the apparent potency of -subunits is truely an overestimate. In general, such reconstitution studies employ G subunits purified from bovine brain. Since a large number of -dimers can be assembled from individual members of the - and -subunit multigene families, it is possible that individual G dimers not tested may prove to be more effective PLC activators. A limited number of studies have been performed to test this possibility (53, 253). With the exception of G₁₂, G dimers of various combinations tested appear equipotent and effective as PLC activators. In fact, the reduced potency of G subunits (compared with -subunits) as activators of the PLC- enzymes may provide a degree of selectivity in receptor regulation of the PLC- enzymes. Accordingly, only activation of abundant G protein heterotrimers (for example, members of the pertussis toxin-sensitive G_o and G_i families) would produce sufficient -subunits for activation for PLC- by this manner.

One unique characteristic of the G_q family of heterotrimeric G proteins is that, when purified and reconstituted with appropriate receptors which promote the guanine nucleotide exchange step in the G protein activation cycle, their intrinsic steady-state GTPase activities are much lower than those of members of the other heterotrimeric G protein families (229, 311, 426). This finding was paradoxical, since direct measurement of PLC-mediated increases in intracellular Ca²⁺ revealed that upon addition of an agonist, the G_q/PLC- system was activated extremely rapidly (216). Ross and co-workers (34) studied regulation of PLC-₁ in a reconstituted system containing purified M₁ muscarinic cholinergic receptors and a purified mixture of G_q and G₁₁. In this system, receptor-promoted binding of GTPS to the G protein and PLC-catalyzed phosphatidylinositol 4,5-bisphosphate (PIP₂) hydrolysis are tightly coupled. When the nonhydrolyzable GTPS was replaced by hydrolyzable GTP, PLC activation was much reduced (35). This apparent uncoupling between G protein and effector in the presence of GTP suggests that PLC-₁ is a GAP for its G_q/11 activator (35). Promotion of GTP hydrolysis accelerates the deactivation rate of the G_q/11 subunits and, in turn, accelerates the deactivation of PLC-₁. The GAP activity of PLC-₁ enables the inositol lipid signaling system to respond rapidly to receptor deactivation. In this way, inositol signaling is regulated not only by the rate of receptor-catalyzed GDP/GTP exchange of the -subunit, but also by acceleration of the rate of GTP hydrolysis accelerated by the effector. More detailed studies of the time courses of PIP₂ hydrolysis by PLC-₁ in this reconstitution system suggest that, in the presence of saturating agonist, receptor-G_q complexes can remain stable over multiple GTPase cycles (436).

All PLC enzymes contain two highly conserved X and Y domains. The three-dimensional structure of a portion of PLC- containing the X and Y domains has been reported recently (153). The other regions of the PLC enzymes display divergent sequences, focusing attention on the unique regions of the PLC- enzymes as possible sites of G protein association. Although the precise structural nature of the interaction between the PLC- enzymes and G protein - and -subunits is not known, several lines of evidence support the idea that these regions which comprise the NH₂ terminus, the inter X-Y region, as well as the COOH terminus play important roles in the PLC- G protein coupling.

A number of independent experiments suggest that the COOH terminus of the PLC- enzymes contains sites for interaction with G protein -subunits. Rhee and co-workers (431) reported that truncation of PLC-₁ and PLC-₂ immediately after the Y domain produces catalytically active proteins that can be activated by G protein -subunits but not by G subunits. Simon and colleagues (616) found that overexpression of the COOH terminus of PLC-₁ in COS-7 cells blocked activation of cotransfected PLC-₁ by muscarinic cholinergic receptors and G protein -subunits. Two short peptides corresponding to a portion of the PLC-₁ amino acid sequence displayed the activity to block activation of PLC-₁ by G_q in vitro, whereas a peptide of identical amino acid composition but different sequence did not (616). Ross and co-workers (34, 35) found that several recombinantly expressed fragments of the COOH terminus of PLC-₁ functioned as GAP for purified G_q when reconstituted with purified muscarinic cholinergic receptors. The PLC- "tail" with the most effective GAP activity did not correspond to the region identified as a site of G_q interaction with PLC-₁ using the synthetic peptide approach described above (616). It is possible that the G protein PLC- interactions that provide the basis for stimulation of GTPase activity and activation of PLC catalytic activity involve different parts of the PLC- COOH terminus. Interestingly, some of the PLC- COOH-terminal tails inhibited basal PLC-₁ activity in a manner that could be overcome by G_q. Perhaps PLC-₁ can form oligomers, and this self-association may have some importance for the mechanism by which G protein -subunits activate the enzymes (436).

The site of interaction between the PLC- enzymes and G protein -subunits is less well defined. Because removal of the COOH terminus of PLC-₁ and PLC-₂ does not diminish the capacity of -subunits to activate the enzymes, it is reasonable to presume that the remaining portion of the enzymes contain the -subunit interaction site. The X and Y domains are common to all PLC enzymes. Therefore, the most likely sites for -subunit interaction are the NH₂ terminus and the inter X-Y region. There exists experimental evidence suggesting that both of these regions of the proteins are important for activation by -subunits. The PLC- and - isoforms contain a pleckstrin homology (PH) domain at their NH₂ terminus. PH domains have been found in a diverse group of proteins including guanine nucleotide exchange factors for small G proteins, as well as for some protein kinases (334). It is clear that a subset of PH domains mediate selective interactions of proteins with inositol lipids and phosphates (334). The PLC- PH domain has been studied intensely. Both intact PLC- and the isolated PLC- PH domain bind PIP₂ and inositol 1,4,5-trisphosphate (IP₃). Binding of PIP₂ to the PLC- PH domain appears to anchor the enzyme to the lipid bilayer, allowing it to function in a "scooting" mode of catalysis (177). The PLC-₁ and PLC-₂ PH domains do not appear to serve an analogous function, since these enzymes bind to membranes with high affinity in a manner that does not depend on PIP₂. The structure of a complex between the PLC- PH domain and IP₃ identifies amino acid residues that participate in formation of ionic and hydrogen bonds with the inositide headgroup. The PH domains of the PLC- isoforms have substitutions of key residues that interact with IP₃ in the PH domains of the -isoform, which presumably accounts for their inability to bind to PIP₂ (480).

Studies with various other proteins suggest that some PH domains may mediate protein interactions with G subunits. The -adrenergic receptor kinase (-ARK) is recruited to the plasma membrane and activated by G subunits (330). -Adrenergic receptor kinase has an NH₂-terminal PH domain, and several lines of evidence suggest that amino acid residues in the COOH-terminal portion of this motif mediate interactions with -subunits (29). As discussed above, the finding that removal of the COOH terminus of the PLC- proteins does not impair activation by -subunits would be consistent with a role for the NH₂ terminus in this process. Removal of the NH₂ terminus of PLC-₂ produced a catalytically inactive protein, and there has been no direct investigation of a role for the NH₂-terminal PH domain of the PLC- enzymes in activation by G subunits. Other work implicates the highly charged residues of the inter X-Y region in regulation of PLC-₂ by -subunits. Peptide fragments from this region of PLC-