RGS proteins and signaling by heterotrimeric G proteins.

A ubiquitously employed mechanism for signal transduction involves ligand binding to a cell surface receptor coupled to a heterotrimeric guanine nucleotide-binding protein (G protein). Receptor activation stimulates nucleotide exchange and dissociation of the G protein, releasing the Ga subunit in its GTP-bound state from the Gbg complex. The released subunits can stimulate a variety of target (effector) enzymes (1), thereby eliciting biochemical responses and changes in cellular physiology. Hundreds of G proteincoupled receptors have been identified (2, 3). These receptors share a common architecture containing seven membrane-spanning segments (4, 5). G proteins also comprise a superfamily that includes at least 17 distinct Ga (6), 5 Gb, and 6 Gg isoforms (1), allowing many combinatorial possibilities. Three-dimensional structures of several Ga subunits and two different Gabg heterotrimers (7, 8) have been determined, providing insights about how these molecular “switches” operate. How are the strength and duration of signaling adjusted to achieve an appropriate response? Attention in this regard has been devoted primarily to receptors, where phosphorylation by protein kinases (9) and receptor-binding proteins, like arrestins (10, 11), contribute to signal desensitization. However, additional proteins participate in signal attenuation at other levels, including phosducins (which act on Gbg) (12) and recoverins (13, 14). Here we focus on discovery of another superfamily of evolutionarily conserved proteins, dubbed RGS proteins, for “regulators of G protein signaling.” RGS proteins act as negative regulators of G proteindependent signaling, at least in part, because they stimulate hydrolysis of the GTP bound to activated Ga subunits.

theless, the importance of GTP hydrolysis provoked a search for factors that accelerate the GTPase activity of G␣ subunits, termed GAPs (for "GTPase-activating proteins") ( Fig. 1). Effector enzymes can serve this role: phospholipase C␤ stimulates the GTPase activity of G q ␣ in vitro (18); and ␥ subunit of cGMP phosphodiesterase stimulates GTP hydrolysis by G t ␣ (19). On the other hand, recent findings have converged on the conclusion that RGS proteins are GAPs for G i ␣ and G o ␣. Identification of the RGS proteins provides an instructive example of how model organisms, like the yeast Saccharomyces cerevisiae and the nematode Caenorhabditis elegans, can reveal new mechanisms of signal regulation applicable to more complex organisms, including humans.

Establishing a Paradigm: Identification of Yeast SST2
In yeast, signaling processes can be dissected genetically. Because the cells can grow as haploids, recessive mutations can be isolated and characterized. Haploid cells undergo mating in response to a pheromone signal by transcribing new genes, by changing morphology, and by transiently arresting in the G 1 phase of the cell cycle. Genes necessary for pheromone response were identified through isolation of mutations that prevent mating, so-called sterile (ste) mutations (20). If normal haploids fail to mate, they become refractory to pheromone and resume cell division. Thus, yeast cells display adaptation and recovery, as observed in mammalian desensitization. In 1982, to identify gene products involved in this downregulation, Chan and Otte (21,22) screened for mutant haploids hypersensitive to pheromone-induced cell cycle arrest. These supersensitive (sst) mutations fell into two classes. One (sst1) was allelic to a gene (BAR1) encoding a secreted protease responsible for inactivation of pheromone (␣-factor). The sst2 mutations defined a novel gene and the first RGS family member.
In addition to responding to doses of pheromone 2 orders of magnitude lower than normal cells (23), sst2 mutants also fail to emerge from pheromone-imposed cell cycle arrest (22). The SST2 gene was isolated (24) by selecting for genomic DNA clones that allowed sst2 mutants to recover from pheromone-induced growth arrest. At the time, however, knowledge about G protein-dependent signaling was in its infancy, sequence data bases were rudimentary, and relevance of signaling events in a unicellular eukaryote to those in humans was not widely appreciated. By the late 1980s, however, it became clear that signaling pathways in yeast and mammalian cells bear considerable similarity (20). For example, the yeast pheromone receptors have a topology resembling other G protein-coupled receptors (2,3), and the yeast G␣ subunit shares 45% amino acid identity with mammalian G i ␣ (6). This evolutionary conservation even allows some mammalian receptors and G proteins to function in yeast (25,26).
Because high-copy vectors were used to isolate SST2, other genes that could rescue, when overexpressed, the sst2 mutation were also obtained. One such dosage suppressor was the KSS1 gene (for "kinase suppressor of sst2") (27), the first MAPK cloned from any organism, thus indicating a connection between G protein signaling and MAPKs. Another dosage suppressor was the GPA1 gene (28,29), which encodes the ␣ subunit of the pheromone receptor-coupled G protein. Cells lacking GPA1 display constitutive activation of the pheromone response pathway (28,29), suggesting that G␤␥ transmits the signal and providing the first in vivo evidence that a G␤␥ complex has a positive and direct role in regulating downstream effectors.
The fact that overproduction of GPA1 (G␣) overcame the need for functional SST2 implied, in a formal genetic sense, that G␣ operates in the same pathway, but downstream of, SST2. G␣ overexpression helps attenuate signaling through sequestration of G␤␥. Other genetic observations suggested a direct interaction between SST2 and GPA1. First, certain mutations in the G␣ gene could bypass the requirement for a fully active SST2 gene (30,31). Second, the recovery-promoting effect of G␣ overexpression was more efficient if residual SST2 function was present (32,33). Third, dominant gain-of-function mutations in SST2 confer resistance to receptor-mediated activation of the pathway but not to activation by mutations that permit constitutive release of G␤␥ (33). Fourth, stimulating the pheromone response pathway induces SST2 production; yet, resumption of growth does not occur in cells that lack G␣, indicating that SST2 cannot stimulate adaptation when G␣ is absent (34). Indeed, it was proposed in 1992 that Sst2 serves as a GAP for G␣ (20). An alternative model, that Sst2 promotes G␣ degradation (35), is inconsistent with the genetic findings and with a subsequent study showing that GPA1 half-life is the same in sst2⌬ and SST2 cells (33).
Our recent biochemical and cytological studies (34) demonstrate that SST2 and G␣ colocalize in the cell. Both proteins are associated with the plasma membrane and, to a lesser extent, with the Golgi body (34). Membrane localization of SST2 is due, at least in part, to direct binding to G␣ because SST2 can be isolated from cell lysates in a complex with a GPA1-glutathione S-transferase fusion, even after detergent and salt extraction of membranes (34), but binding seems unaffected by the nature of the nucleotide added (GDP or GTP␥S). Membrane association of Sst2 is striking because its deduced sequence is hydrophilic, lacks any significant stretches of hydrophobic residues, and is devoid of myristoylation or isoprenylation sites. Overexpression of a GTPase-deficient allele of G␣ (gpa1 R297H ) in wild-type cells enhanced their pheromone sensitivity, resembling the loss of SST2 (34). This G␣ mutant presumably remains in the GTP-bound state, cannot bind G␤␥, but still forms a complex with SST2. By sequestering SST2, less is available to act on the normal G␣, thereby explaining the observed enhancement in pheromone sensitivity.

RGS Proteins in Other Model Organisms
The closest counterpart to SST2 in another organism is the flbA gene product of the filamentous fungus, A. nidulans (36). Under conditions that should cause sporulation, flbA mutants proliferate, remain undifferentiated, and eventually lyse, a phenotype called "fluffy autolysis." Conversely, overexpression of normal flbA permits sporulation under conditions that would otherwise prevent it. A dominant fluffy autolysis mutation is a substitution (G42R) in a G␣ gene, fadA (37). Alteration of the analogous Gly in mammalian G s ␣ slows its rate of GTP hydrolysis and impairs its ability to release G␤␥ (38). On the other hand, certain other fadA (G␣) mutations, including a substitution predicted to block G␤␥ disso-ciation, actually suppress a flbA mutation (37). One interpretation of these findings is that free G␤␥ somehow blocks sporulation and that flbA is required for sporulation because it promotes the ability of G␣ to remain associated with G␤␥.
Another SST2-related gene was identified in the nematode, C. elegans, during a genetic screen for mutants that alter certain neuronal activities (39). When placed on a lawn of bacterial prey, C. elegans adjusts several of its behaviors, including frequency of egg laying. Egg laying is controlled by serotonergic motor neurons that innervate the vulval and uterine muscle cells and is suited to genetic analysis because the number of laid and unlaid eggs (which are clearly visible inside the adult) can be readily compared. A mutation (egl-10) that decreased the frequency of egg laying was identified and the corresponding gene cloned (39). The C-terminal portion of the 555-residue EGL-10 product bore similarity to the C-terminal segment of the 698-residue SST2 protein (Fig. 2). While egl-10 mutants rarely lay eggs, overexpression of normal EGL-10 causes animals to lay eggs more frequently (39). These phenotypes suggest that EGL-10 is required for serotonin-stimulated egg laying. In other animals, serotonin acts through G protein-coupled receptors (2,3). Indeed, the goa-1 mutation, which resides in a gene homologous to mammalian G o ␣ (40,41), results in elevated egg laying, and conversely, overexpression of normal GOA-1 causes reduced egg laying (40,41). The fact that egl-10 and goa-1 mutations have related (but opposite) phenotypes and the fact that an egl-10 mutation has no effect if a goa-1 mutation is present suggest that the normal role of EGL-10 is to down-modulate the activity of GOA-1.

A Superfamily of RGS Proteins
Multiple homologs of SST2 and EGL-10 are present in higher eukaryotes. Using the yeast two-hybrid system, a human protein, GAIP, that interacts with human G i3 ␣ was identified (42). G i3 ␣ (prepared by in vitro translation) binds to a GAIP-glutathione S-transferase fusion protein. GAIP is expressed in many tissues and is similar over most of its 217-residue length to the C-terminal portions of SST2, FlbA, and EGL-10. GAIP is more related to products of two other short mammalian cDNAs, GOS8 and BL34/ 1R20 (42). In these proteins, similarity to EGL-10 extends over ϳ130 contiguous amino acids, whereas in SST2 and FlbA, similarity is divided into three discontinuous blocks (Fig. 2). This ϳ130residue core domain defines the RGS superfamily and, in GAIP, is both necessary and sufficient for interaction with G i3 ␣ (42). GAIP also interacts more weakly with G i2 ␣, but not with G q ␣.
BL34/1R20 cDNA was identified because the corresponding mRNA is elevated in chronic lymphocytic leukemia (43). Expression is specific to B lymphocytes and induced by mitogenic stimuli (44). The 196-residue BL34/1R20 product has been renamed RGS1, in light of its presumed function. Similarly, GOS8 cDNA, encoding a 211-residue protein (RGS2), was isolated because expression of its corresponding mRNA in human blood lymphocytes is induced by concanavalin A (a T-cell mitogen) in combination with cycloheximide (45). However, GOS8 mRNA is induced by cycloheximide alone (46). Yet another homolog, RGS3, was identified by screening a B-cell cDNA library with an oligonucleotide corresponding to sequences conserved between RGS1 and RGS2 (47). RGS3 (519 residues) is much longer than RGS1 or RGS2.
Because of the evidence that yeast SST2 and nematode EGL-10 regulate G protein signaling, it was presumed that mammalian homologs would act analogously. Indeed, RGS4 was identified by screening for rat brain cDNAs that, when expressed in a yeast sst2⌬ mutant, could stimulate recovery from pheromone-induced growth arrest and partially block pheromone-induced gene transcription (47). In an independent study, RGS2 was also able to confer pheromone resistance to yeast sst2 cells (48). The 205-residue RGS4 product is expressed exclusively in the brain (47). 2 RGS1, RGS2, RGS3, and RGS4 regulate G protein signaling in mammalian cells. Elevation of RGS1 by transient transfection attenuated MAPK activation in response to PAF and diminished the Ca 2ϩ flux provoked by either PAF or lysophosphatidic acid; similarly, expression of each of the four RGS proteins attenuated MAPK activation by interleukin 8 (47). Likewise, RGS4 expression 2 K. Druey and J. Kehrl, personal communication.

FIG. 1. The cycle of G protein activation and inactivation. Top panel,
when GDP-bound, G␣ is inactive and associated with G␤␥. Agonist binding to a receptor promotes guanine nucleotide exchange; G␣ releases GDP, binds GTP, and dissociates from G␤␥. Dissociated subunits activate target proteins (effectors). When GTP is hydrolyzed, subunits reassociate. G␤␥ antagonizes receptor action by inhibiting guanine nucleotide exchange. RGS proteins bind to G␣, stimulate GTP hydrolysis, and thereby reverse G protein activation. Bottom panel, the roles of a receptor, G␤␥, and an RGS are completely analogous to the GDSs, GDIs, and GAPs that regulate small monomeric G proteins like Ras.
attenuates MAPK activation in response to agonist stimulation of M2 muscarinic acetylcholine receptors. 2 In contrast, RGS3 had no effect on MAPK stimulation by two post-G protein activators, phorbol ester and activated Raf-1 kinase (47).
Searches of expressed sequence tag (EST) data bases and screening by polymerase chain reaction amplification (39,47,48) have revealed more RGS homologs in mammals and in the C. elegans genome, including one with two tandem RGS domains (Fig. 2).

Evidence for Bifunctional RGS Proteins
Some RGS proteins (including SST2, FlbA, and RGS3) possess long N-terminal extensions. This number may increase as RGS clones identified only as EST fragments are fully characterized. In this regard, it is striking that the identity between RGS7 and EGL-10 (39) is 75% across their N-termini versus 46% within their RGS cores, suggesting that the N-terminal segment is needed for function. Yeast SST2 provides a precedent for such a situation.
Whereas the C-terminal 230 residues of SST2 constitute its RGS domain, the N-terminal 300 residues bear weak similarity to the catalytic domain of mammalian Ras-GAP (34). Both the GAP-like and RGS segments are required, since truncations at either the Nor C-terminal ends completely eliminate biological activity (34). A dominant gain-of-function allele, SST2(P20L), maps within the GAP-like region, supporting its functional importance (33). The GAP-like segment could stimulate the GTPase activity of G␣ (in the manner that GAP acts on Ras) after SST2 associates (via its RGS domain) with G␣. However, as discussed above, RGS proteins lacking a GAP-like domain can partially substitute for loss of SST2 in yeast and, as discussed below, are presumably able to stimulate the GTPase activity of G␣ subunits in situ. Hence, the GAP-like domain of SST2 may contribute to adaptation through another mechanism.
Another potentially multifunctional RGS is the mouse fused gene product (49), which contains an N-terminal extension homologous to proteins that bind to phosphoprotein phosphatase, PP2A. 3 Establishing the functions of other domains is as important as sorting out specificity determinants in the core RGS domain.

RGS Proteins Act as GAPs for G␣ Subunits
Effects of purified GAIP and RGS4 on purified G i1 ␣, G i2 ␣, G i3 ␣, G o ␣, and G s ␣ have been examined in vitro (50). Neither affected the steady-state rate of GTP hydrolysis by these G␣ subtypes. Under such conditions, however, GTP hydrolysis is limited by GDP dissociation. Hence, the steady-state assay actually measures the rate of guanine nucleotide exchange and suggests that RGS proteins do not function as either GDIs or GDSs. When a single round of GTP hydrolysis was measured, either RGS4 or GAIP stimulated the rate of hydrolysis more than 40-fold for all of the G␣ subtypes, except G s ␣. Thus, RGS proteins act as GAPs (at least for the G i ␣ subfamily). RGS4 partially restored GTPase activity to G i1 ␣(R178C) and to G i1 ␣(S47N) but not to G i1 ␣(Q204L) (50). Arg-178 stabilizes the developing negative charge on the ␥-phosphate leaving group in the transition state during hydrolysis; Ser-47 contributes to Mg 2ϩ binding, which is required for nucleotide hydrolysis and subunit dissociation; and Gln-204 is essential for orienting the attacking water molecule and for transition-state stabilization (51,52). Restoration of function by RGS4 suggests that RGS proteins may accelerate GTP hydrolysis by stabilizing G␣ proteins in their active conformation. Indeed, RGS1 (53) (also RGS4 4 ) interact weakly with G o ␣ in the presence of GDP or GTP␥S but strongly in the presence of both GDP and AlF 4 Ϫ , a combination that mimics the transitionstate of the nucleotide during hydrolysis.
Selective binding to the transition-state conformation of G␣ may not be a general feature of RGS proteins, however. A human 173residue RGS10, isolated via its interaction with a GTPase-deficient G␣ mutant, G i3 ␣(Q204L) (54), can co-immunoprecipitate with either G i3 ␣(Q204L) or G z ␣(Q204L), both presumably GTP-bound, but not with wild-type (presumably GDP-bound) G i3 ␣ or G z ␣ or with either wild-type or mutationally activated G s ␣. Like RGS4 (50), GAIP (50), and RGS1 (53), RGS10 (54) stimulates the GTPase activity of several members of the G i ␣ subfamily but is ineffective against G s ␣. 3 F. Costantini, personal communication. 4