I. Introduction

Top Previous Next References

Our understanding of G protein signaling has undergone fundamental changes in recent years. Established models based on information gathered over the last quarter century suggest that most hormones, neurotransmitters, and sensory input rely upon a G protein-coupled receptor (GPCR¹), a heterotrimeric guanine nucleotide-binding regulatory protein (G protein), and a limited number of well described downstream effector proteins (e.g., adenylyl cyclases, phospholipases) and chemical second messengers to transmit their signals across the plasma membrane (Bourne et al., 1990; Simon et al., 1991; Hepler and Gilman, 1992; Hamm, 1998). However, recent studies indicate that GPCRs and G proteins engage a growing list of newly appreciated proteins and linked signaling pathways to carry out their cellular functions (Bockaert and Pin, 1999; Hall et al., 1999). Prominent among these new binding partners are the regulators of G protein signaling (RGS proteins). RGS proteins are a large family of highly diverse, multifunctional signaling proteins, which share a conserved signature domain (RGS domain) that binds directly to activated G subunits to modulate G protein signaling. RGS proteins differ widely in their overall size and amino acid identity, and many family members possess a remarkable variety of structural domains and motifs that regulate their actions and/or enable them to interact with protein binding partners with diverse cellular roles (Hepler, 1999; Siderovski et al., 1999). Several comprehensive reviews have appeared recently, which examine RGS biochemistry and cellular functions from different perspectives (Burchett, 2000; De Vries et al., 2000; Ross and Wilkie, 2000; Zhong and Neubig, 2001).

Although considerable information is now available describing the biochemical and cellular properties of RGS proteins as blockers of G protein signaling, less is known about cellular mechanisms that regulate RGS functions per se. In addition, although early evidence suggested that RGS proteins acted primarily as negative regulators of G protein signaling, recent findings indicate that these proteins act as tightly regulated modulators and/or as multifunctional integrators of G protein signaling. This review will highlight emerging concepts regarding RGS proteins as modulators and integrators of multiple signaling pathways and focus on new information regarding cellular mechanisms that regulate RGS functions. A brief discussion will also center on roles of RGS proteins in physiology and their potential as therapeutic targets.

	II. RGS Proteins Directly Regulate G Protein Activity

Top Previous Next References

G proteins consist of G, G, and G subunits, and G subunits bind and hydrolyze GTP to act as molecular switches. Agonist occupancy of GPCRs stimulates the exchange of GTP for GDP on G subunits and subunit dissociation, and the amplitude and lifetime of G protein-directed signaling events are dictated by the lifetime of GTP on G. Purified G subunits in solution are inefficient GTPases with intrinsic rates of GTP hydrolysis in the range of 0.1 to 0.3 Pi/mol G/min for most G (Gilman, 1987). In the absence of receptor, GTP hydrolysis is limited by the rate of GDP release from G following hydrolysis. Agonist-occupied GPCRs stimulate the release of GDP from G, opening the guanine nucleotide binding pocket for rapid binding of abundant intracellular GTP. Thus, in the presence of receptors and agonist using reconstituted systems, the observed rates of G-directed GTP hydrolysis are enhanced 10-fold or more and are reflective of intrinsic rates of GTP hydrolysis (Gilman, 1987; Ross and Wilkie, 2000).

However, for many G protein-regulated signaling responses, their rates of deactivation in a cellular context are much faster (100- to 300-fold) than is predicted from observed rates of G-GTP hydrolysis using purified components. G protein signaling events in the retina, brain, and heart proceed on a much faster time scale than is the case with other non-electrically excitable tissues. For example, G protein-directed (i.e., Gt or transducin) phototransduction in intact rod outer segments begins and ends within milliseconds, with recovery times of less than 200 ms. These rates are much faster than is expected from known rates of rhodopsin and G_t-mediated GTP hydrolysis in light-stimulated rod outer segment membranes (for review, see Arshavsky and Pugh, 1998). Similar discrepancies between in vitro and in vivo data were also observed for rates of deactivation of G protein-regulated potassium and calcium channels (for review, see Zerangue and Jan, 1998). These observations predicted the existence of unidentified factors or proteins that regulate the rates of G-GTP hydrolysis to fine-tune G protein activation state and signaling responses.

Initial evidence for cellular regulators of G protein signaling came from genetic studies of lower eukaryotes (Dohlman and Thorner, 1997; for review, see Koelle, 1997). Studies in yeast nearly two decades ago recognized a gene product (Sst2p) that, when mutated, presented a phenotype that was supersensitive to G protein-directed pheromone responses (Chan and Otte, 1982; Weiner et al., 1993; Dohlman et al., 1995). A similar gene (flbA) was identified as a negative regulator of G protein signaling responses in the fungal organism Aspergillus nidulans (Lee and Adams, 1994). Other investigators studying mammalian systems independently identified a novel gene (GOS8) that was rapidly up-regulated in stimulated monocytes (Siderovski et al., 1994), and a new protein that bound activated G_i3 in yeast two-hybrid screens, which was termed G alpha interacting protein (GAIP, or later RGS-GAIP) (De Vries et al., 1995). Although cellular roles for GOS8 and RGS-GAIP remained obscure at that time, GOS8 was recognized to share a novel conserved domain with other mammalian proteins (Siderovski et al., 1996). Full appreciation that each of these proteins belonged to a larger superfamily of signaling proteins came from subsequent genetic studies of Caenorhabditis elegans describing a gene (egl-10) that negatively regulated G_o-directed locomotion and egg-laying behavior (Koelle and Horvitz, 1996). Partial nucleotide sequences for 15 mammalian genes were identified from a brain cDNA library that shared a conserved 130-amino acid core domain with the egl-10, sst2p, flbA, and GOS8 genes. These proteins were termed regulators of G protein signaling and numbered consecutively (RGS1-RGS15), and the conserved domain was termed the RGS domain, henceforth recognized as the protein family hallmark. The previously discovered GOS8 was identical with one sequence and was renamed RGS2, and RGS7 was recognized as the mammalian homolog of Egl-10. Separate studies showed that mammalian RGS4 could substitute for Sst2p as an inhibitor of pheromone responses in yeast, demonstrating a conservation of RGS function across species (Druey et al., 1996). Since that time, full-length cDNA for these and other mammalian RGS proteins have been reported to reveal a large family of highly divergent, multifunctional proteins (Fig. 1; Table 1).

View larger version (64K): [in this window] [in a new window]   Fig. 1.   Structures and classification of mammalian RGS and RGS-like proteins. RGS and RGS-like proteins are classified into subfamilies based on alignment of RGS domain amino acid sequences (Zheng et al., 1999; Ross and Wilkie, 2000). Proteins are oriented with their N termini on the left and their C termini on the right. See the text for a description of the domains and motifs.

Although RGS proteins were first identified as negative regulators of G protein signaling, the biochemical mechanisms whereby these proteins regulated G signaling were unknown. G proteins act as molecular switches, and RGS proteins could block G signaling by preventing GTP binding to G or by limiting the lifetime of GTP bound to G. Following the discovery of RGS proteins, a series of studies demonstrated that various RGS proteins act as GTPase-activating proteins (GAPs) to greatly accelerate (up to 1000-fold) the rate of G-GTP hydrolysis and limit the lifetime of the active G-GTP species (Berman et al., 1996b; Hunt et al., 1996; Watson et al., 1996; Hepler et al., 1997; Kozasa et al., 1998). These and other studies demonstrated that RGS proteins bind directly and preferentially to the active GTP bound forms of G_i/o, G_q, G_12/13, or G_s and that RGS domains exhibit highest affinity for the GDP-Mg²⁺-AlF4 bound G, which mimics the transition state during GTP hydrolysis (for review, see Berman and Gilman, 1998 or Ross and Wilkie, 2000). Additional studies demonstrated that RGS proteins can also bind tightly to active G to block effector activation independent of GAP activity by acting as effector antagonists (Hepler et al., 1997; Carman et al., 1999). For a comprehensive discussion of the biochemical properties of RGS proteins as GTPase-activating proteins, see Ross and Wilkie (2000).

Completion of the human genome project has confirmed the existence of more than 30 distinct proteins that contain an RGS or an RGS-like (RL) domain (see below; Fig. 1 and Table 1). Based on amino acid identities within the conserved RGS domain, two independent research groups have classified RGS proteins into six distinct subfamilies (Zheng et al., 1999; Ross and Wilkie, 2000). These groupings also correlate well with overall structure and identified functions within subfamilies. In one classification, subfamily names are arbitrarily designated A-F (Zheng et al., 1999) whereas in the other classification (Ross and Wilkie, 2000), subfamily names are derived from a prototypical RGS protein member (e.g., the RZ subfamily is typified by RGSZ). Both classifications are consistent except for the assignment of RGS10, which in one case is not assigned to a subfamily (Zheng et al., 1999) and in the other is grouped in the R12 subfamily based on RGS domain similarities (Ross and Wilkie, 2000). The six groupings (Fig. 1; Table 1) include the A or RZ subfamily (prototype RGSZ); the B or R4 subfamily (prototype RGS4); the C or R7 (prototype RGS7); the D or R12 subfamily (prototype RGS12); the E or RA subfamily (prototype Axin); and the F or RL subfamily (containing proteins with RL domains). The RL domains of this subfamily are only distantly related to each other and to other RGS domains both in amino acid sequence identities and in G recognition. Proteins containing RL domains include D-AKAPs (dual specificity A kinase anchoring proteins), p115RhoGEFs, RGS-PX1, and G protein receptor kinases (GRKs). Unlike the other RGS subfamilies, proteins within the RL category are a collection of miscellaneous proteins, classified together only because they each contain a weakly homologous RGS domain. Members of the RZ and R4 subfamilies, with two exceptions (RGS3 and RET-RGS1), are small 20- to 30-kDa proteins that contain short N- and C-terminal regions flanking the RGS domain. In contrast, members of the R7, R12, RA, and RL subfamilies, with one exception (RGS10), are much larger proteins (up to 160 kDa) that possess longer N and C termini encoding various binding domains and motifs for other proteins.

Aside from a shared RGS domain, RGS proteins differ widely in their overall size and amino acid identity, and possess a remarkable variety of structural domains and motifs (Fig. 1). Unlike members of the A/RZ and B/R4 subfamily, which are simple proteins with little more than an RGS domain, members of the C/R7, D/R12, E/RA, and F/RL subfamilies each have additional subfamily-specific domains. All members of the C/R7 subfamily (RGS6, RGS7, RGS9, RGS11) contain a DEP (disheveled, Egl-10, pleckstrin) domain, a previously unknown conserved region (Sondek and Siderovski, 2001), which we call the R7 homology or R7H domain, and a GGL (G protein gamma subunit-like) domain. Members of the D/R12 family (RGS12, RGS14) share a RBD (Rap1/2-binding domain) and a GoLoco motif. Members of the E/RA family (Axin, Conductin) each contain a glycogen synthase kinase 3-binding domain (GSK3), a -catenin binding site (Cat), a protein phosphatase 2A (PP2A) homology region, and a dimerization domain (DIX). Within the RL subfamily, the RhoGEF proteins each contain DH (dbl homology) and PH (pleckstrin homology) domains, the GRK proteins each contain a Ser/Thr kinase catalytic domain, whereas RGS-PX1 contains Phox homology (PX) and Phox-associated (PXA) domains. Individual RGS proteins within subfamilies also contain additional protein-specific domains (Fig. 1). These structural features impart to RGS proteins the capacity to interact with a growing list of protein binding partners that mediate RGS signaling, RGS subcellular targeting, and regulation of RGS functions. Reported non-G protein binding partners for RGS proteins are illustrated in Fig. 4, and their roles as regulators or mediators of RGS signaling functions are discussed elsewhere within the text.

Given the extraordinary diversity of RGS proteins and their functions, emerging ideas suggest that the smaller simple RGS proteins (primarily those of the A/RZ and B/R4 subfamilies) serve almost exclusively as negative regulators of G protein signaling. However, the functions of these proteins are tightly regulated such that they act as modulators rather than dedicated inhibitors of G protein signaling. In contrast, the larger RGS family members (C/R7, D/R12, E/RA, and F/RL subfamilies) are multifunctional proteins that have the capacity to bind both G proteins and other signaling proteins. As such, these complex RGS proteins act as integrators of G protein signaling, possibly as novel G protein effectors or as scaffolding proteins. In this regard, the complex RGS proteins are similar to the growing list of signaling proteins that share a single modular domain but are functionally dissimilar, e.g., proteins that contain SH2 or SH3 domains (see Zhong and Neubig, 2001). We will discuss examples of simple RGS proteins that serve as modulators of G protein signaling, examples of complex RGS proteins as integrators of G protein signaling and then examine mechanisms that regulate their cellular functions.

	III. RGS Proteins Modulate G Protein Signaling

Top Previous Next References

RGS4 was among the first RGS proteins discovered (Druey et al., 1996; Koelle and Horvitz, 1996) and its biochemical and cellular properties have been studied more extensively than those of any other family member. Thus, we have chosen to focus our discussion on RGS4 as the best developed (although still incomplete) case study for understanding mechanisms that underlie RGS modulation of G protein signaling. RGS4 is the prototypical member of the B/R4 family, of which nearly all members are relatively simple proteins composed of an RGS domain flanked by minimal N and C termini lacking prominent modular domains (Fig. 1). Because of their relative simplicity, emerging ideas suggest that the principle cellular role for these proteins is to modulate G protein signaling through their RGS domains, while their N and C termini dictate RGS subcellular localization and signaling capacity. Consistent with this idea, a large body of literature demonstrates that heterologous expression of recombinant forms of RGS4 and other simple B/R4 family members blocks receptor and G protein signaling (for review, see Burchett, 2000; De Vries et al., 2000; Zhong and Neubig, 2001). However, growing evidence now suggests that endogenous RGS proteins act not as simple inhibitors of G protein signaling but, instead, as tightly regulated modulators that fine tune G protein signaling events in a cell- and context-dependent manner. To illustrate this point, we will focus on emerging models of cellular roles for RGS4 as a modulator of G_q/11 and Ca²⁺ signaling.

Recombinant RGS4 is an effective GAP for both G_i/o family members and G_q, and its heterologous high-level expression blocks both G_i/o-mediated signaling events and G_q/11-directed inositol lipid/Ca²⁺ signaling in mammalian cells. However, in the case of RGS4 and other simple RGS proteins, where examined, native protein levels are typically low in host cells even when their mRNA levels are high. When RGS4 and other simple B/R4 family members are introduced into permeabilized cells at low levels reflective of their physiological concentrations, these proteins do not block receptor and G_q/11-directed Ca²⁺ signaling. Instead, these RGS proteins quite unexpectedly elicit rhythmic Ca²⁺ oscillations (Xu et al., 1999; Luo et al., 2001), suggesting that the observed complete blockade of G protein responses by RGS in other circumstances may be a consequence of overexpressing the protein.

Sufficient information is now available to propose a working model that describes cellular roles and regulation of RGS4 as a modulator of the rhythmic Ca²⁺ oscillations elicited by many hormones and neurotransmitters (Thomas et al., 1996). This model (illustrated in Fig. 2) is derived in large part from a recently proposed hypothesis (Sierra et al., 2000; Luo et al., 2001), and as is true of all models, the supporting findings are open to other interpretations. The chief limitation of this hypothesis is that many of the supporting observations have not yet been independently confirmed by other laboratories. In addition, some of the cellular components of the model have been extrapolated from in vitro studies, and other supporting studies ignore the fact that B/R4 RGS family members also modulate G_i/o signaling in parallel. Thus, this model should be considered not as fact but, instead, as a provocative and testable scenario for describing cellular mechanisms whereby simple RGS proteins may act as highly regulated modulators of GPCR signaling. We will discuss each step of the model in some depth to better understand possible mechanisms that regulate the contribution of RGS4 to this process.

View larger version (44K): [in this window] [in a new window]   Fig. 2.   Proposed model depicting RGS4 modulation of Ca2+ oscillations in mammalian cells. The following model is based on recently proposed ideas (Popov et al., 2000; Luo et al., 2001) and supporting data (see text and references therein). Top panel, schematic diagram illustrating peak amplitudes (1 and 3) and refractive periods (2 and 4) of rhythmic, oscillatory Ca2+ spikes in cells. Corresponding diagrams illustrating the proposed role of RGS4 in modulating each stage of the Ca2+ oscillations (1-4) are presented in panels 1-4. Panel 1, hormone (H) activation of GPCR stimulates the Gq/IP3/Ca2+ pathway, resulting in RGS4 membrane recruitment and receptor association. Depicted are associations of the amphipathic helix on the N terminus of RGS4 (+++) with anionic lipids in the plasma membrane (- - -). Panel 2, RGS4 forms a complex with GPCR, Gq-GTP, and PLC and exerts its GAP effects on Gq to shut off Ca2+ mobilization. G activates PI3K, which synthesizes PIP3. Panel 3, PIP3 binding to RGS4 inhibits its GAP activity toward Gq, allowing resumption of IP3 production and Ca2+ mobilization. Ca2+ () activated CaM competes with PIP3 for binding to RGS4. Panel 4, Ca2+/CaM binding uncouples RGS4 from PIP3 but does not inhibit RGS4 GAP activity. RGS4 reassociates with the Gq/GPCR complex to shut off IP3/Ca2+ production. As Ca2+ levels fall, CaM is deactivated and dissociates from RGS4, allowing rebinding of PIP3. See text for further discussion.

1. Cellular Mechanisms That Influence RGS4 Membrane Recruitment and Attachment. RGS4 (as well as other RGS proteins) is predicted to exist as a soluble hydrophilic protein, but it is found both in the cytosol and tightly bound to membranes (Srinivasa et al., 1998; Bernstein et al., 2000). To modulate G_q/11 and G_i signaling events, RGS4 needs to be present at the cytoplasmic face of the plasma membrane. Indeed, recombinant RGS4 and other B/R4 family members are recruited from cytosol to membranes by activated forms of these G subunits (Srinivasa et al., 1998; Druey et al., 1999; Heximer et al., 2001). However, in reconstituted systems using purified proteins, the isolated N terminus of RGS4 can associate rapidly and irreversibly with anionic lipid vesicles independent of whether receptor and/or G and G subunits were present (Tu et al., 2001). Apparently, G and G subunits can enhance but are not necessary for constitutive RGS4 membrane association (Dowal et al., 2001). Structural features on RGS4 responsible for its membrane attachment have been identified. The N terminus of RGS4 contains a 33-amino acid cationic amphipathic  helix that drives RGS4 membrane attachment (Bernstein et al., 2000; Tu et al., 2001). RGS4 is also reversibly palmitoylated near its N terminus at Cys2 and, to a lesser extent, Cys12 (Bernstein et al., 2000). However, acylation does not seem to be essential for RGS4 binding to anionic lipid vesicles but does appear to accelerate the process, likely due to its hydrophobic contributions (Tu et al., 2001). N-terminal palmitoylation also targets RGS4 to specialized cholesterol and glycosphingolipid-rich vesicles in vitro, and it has been suggested that reversible acylation may target RGS4 and other RGS proteins to specialized lipid rafts within the plasma membrane (Moffett et al., 2000).

2. Once Bound to Membranes, What Factors Influence RGS4 Specificity for Target G in Cells? In solution-based reconstituted systems using purified proteins, RGS4 is quite promiscuous (as are most other simple RGS proteins) and can block G_i and G_q signaling functions in vitro (Berman et al., 1996a; Hepler et al., 1997) and when introduced into intact cells (Druey et al., 1996; Huang et al., 1997a; Yan et al., 1997; Heximer et al., 1999). However, under experimental conditions where protein can be introduced directly to cells at defined concentrations, RGS4 appears to regulate G function based on recognition of receptors rather than association with G. In pancreatic acinar cells, carbachol, bombesin, and CCK each stimulate Ca²⁺ signaling to similar extents by activating G_q/11-linked to their respective receptors. Wilkie, Muallem, and coworkers demonstrated that introduction of purified RGS4 directly into these cells selectively inhibited inositol lipid/Ca²⁺ signaling by carbachol at concentrations that were 4- and 33-fold more potent than required to block bombesin- and CCK-directed Ca²⁺ signaling, respectively (Xu et al., 1999). The B/R4 family members RGS1 and RGS16 also displayed receptor selectivity, and RGS1 was nearly 1000-fold more potent at blocking carbachol- than CCK-directed Ca²⁺ signaling (Xu et al., 1999). In stark contrast, RGS2 displayed no preference between the three receptors. This receptor selectivity of RGS4 is conferred by its N terminus since truncated RGS4 lacking this domain exhibited reduced potency and no receptor selectivity. Furthermore, RGS4 potency and selectivity for affecting muscarinic receptor-directed Ca²⁺ signals was restored by combined addition of the N terminus and the RGS core domain (Zeng et al., 1998). Considering all of these findings together, a plausible explanation forwarded by the authors is that a feature associated with the N terminus of RGS4 selectively recognizes certain receptors but not necessarily the linked G. Although there is no direct evidence that RGS4 physically contacts receptors, this possibility cannot be ruled out. Although receptor and G protein subunits do not appear to be required for RGS binding to anionic vesicles, they may conspire to help localize and orient RGS4 (and other RGS) to optimize their GAP activities toward G (Tu et al., 2001) (Fig. 2, panel 2).

3. Once RGS4 Is Bound to Membranes and Functionally Linked to Receptor and G Protein, What Factors Regulate Its Effects on Ca²⁺ Signaling? RGS4 binds very selectively to the anionic lipid PIP₃ (Popov et al., 2000). PIP₃ is formed transiently from PIP₂ by the actions of phosphatidylinositol 3-kinase (PI3K), certain isoforms of which are directly stimulated by G subunits (Fig. 2, panel 3) (Stephens et al., 1997). PIP₃ binds RGS4 at a site within the RGS domain (distinct from and opposite to the RGS/G contact face), and PIP₃ binding inhibits RGS4 GAP activity toward G (Popov et al., 2000). When complexed with Ca²⁺, activated calmodulin (Ca²⁺/CaM) apparently binds to this same site. Although Ca²⁺/CaM competes with PIP₃ for binding at this site, it has no effect on G_q GAP activity (Popov et al., 2000) (Fig. 2, panel 4).

4. What Factors Contribute to Turning Off This Signaling Loop? The simplest means for shutting off Ca²⁺ oscillations would involve withdrawal of agonist. However, in the continued presence of agonist, several cellular mechanisms may contribute to turning off the RGS signal. At the level of the RGS protein, regulated post-translational modification of sites within the RGS domain may block further RGS/G interactions. Consistent with this idea, addition of palmitate to a conserved Cys residue on helix 4 of the RGS domain blocks RGS4 interactions with G (Tu et al., 1999). Alternatively, palmitoylation at other sites within the N terminus may target RGS4 to specialized "lipid rafts" within the plasma membrane and thereby limit its availability (Druey et al., 1999; Moffett et al., 2000). Other post-translational modifications may also contribute to feedback and inhibit RGS functions. Phosphorylation of several B/R4 family members modulates their capacity to interact with G. For example, phosphorylation of RGS2 by PKC (Cunningham et al., 2001) or RGS16 by undefined kinases (Chen et al., 2001a) blocks their interaction with G. Another possibility is that phosphorylation of certain B/R4 family members at sites within the RGS domain promotes their binding with the cytosolic scaffolding protein 14-3-3 to prevent their interactions with G (Benzing et al., 2000). Taken together, these findings suggest that mechanisms for uncoupling RGS from G could proceed in conjunction with well defined classical mechanisms that desensitize receptor and G protein actions such as receptor phosphorylation and internalization (Ferguson, 2001).

Experimental evidence supporting the model of RGS4 regulation of Ca²⁺ oscillations (Fig. 2) also suggests that RGS4 and certain other RGS proteins may bind directly to GPCRs to form a stable ternary complex between the GPCR, G_q, PLC, and CaM (for discussion, see Sierra et al., 2000). Although no direct evidence has been reported demonstrating RGS4 physically binding to receptors, several lines of indirect evidence support the idea that RGS4 assembles related signaling proteins, perhaps as a stable complex with receptors. As discussed above, a synthetic peptide corresponding to the N terminus of RGS4 selectively blocks certain GPCR signals (Zeng et al., 1998; Xu et al., 1999). RGS4 also binds Ca²⁺/CaM at a regulatory site within the RGS domain (Popov et al., 2000), G_q at the G/RGS interface (Hepler et al., 1997), and PLC1 at an undefined site (Dowal et al., 2001). RGS4 also displays relatively weak but significant affinity for binding G (Wang et al., 1998; Dowal et al., 2001). These data support the idea that RGS4 could potentially act as a multifunctional scaffold to assemble related proteins in a shared pathway in concert, specifically IP₃-mediated Ca²⁺ signaling. In this regard, RGS4 and other RGS proteins may act in a manner similar to the -arrestins, which form a stable signaling complex with GPCR and serve as a scaffold to assemble related kinases (JNK, Ask1, and MKK) to facilitate MAPK signaling at the plasma membrane (Miller and Lefkowitz, 2001). Further studies will be necessary to determine whether RGS4 and other simple RGS proteins make direct physical contact with GPCR and, if so, whether RGS and GPCRs form a stable complex. The relative role of G proteins in this process is currently unknown. Even so, other studies (discussed elsewhere in the text) provide compelling evidence that certain of the larger complex RGS proteins bind directly to GPCRs by PDZ domain interactions to serve as multifunctional integrators of receptor and G protein signaling.

Unlike inositol lipid/Ca²⁺ signaling and other slow-acting G protein-regulated processes (e.g., MAPK cascades), ion conductances in electrically excitable cells respond to signals on a subsecond time scale. GPCR and linked G protein subunits directly regulate several important fast-acting signaling events in the brain, retina, and heart (Arshavsky and Pugh, 1998; for review, see Zerangue and Jan, 1998). Most notable among these are phototransduction in the retina (mediated by G_t), G protein-regulated inwardly rectifying potassium channels (GIRK) in brain and heart, and the voltage-dependent N-type Ca²⁺ channels in the brain (both channels are directly mediated by G). The onset and deactivation of these signaling events are very rapid, and compelling evidence now demonstrates that RGS proteins modulate the kinetics of these responses. In atrial myocytes, GIRK currents deactivate within 600 ms. However, when GIRKs are exogenously expressed in Xenopus oocytes lacking RGS proteins, their deactivation occurs at rates markedly slower than the intrinsic GTPase rates of G_i/o (Higashijima et al., 1987). Introduction of RGS4 and other simple RGS proteins of the B/R4 family (RGS1, RGS3, and RGS8) markedly accelerates GIRK activation and deactivation rates, and RGS4 restores GIRK kinetics in mammalian cell lines to levels similar to those observed in heart and brain (Doupnik et al., 1997; Saitoh et al., 1997, 1999). The RGS in question had no effect on the amplitude of GIRK currents, indicating that they are not dedicated inhibitors, but rather modulators that fine-tune these signal responses.

Mechanisms underlying RGS modulation of activation and deactivation rates for GIRK are unclear, although several possibilities have been proposed (Zerangue and Jan, 1998). Since GIRK and N-type Ca²⁺ channels are directly regulated by G and not G-GTP, rates of channel activation and deactivation presumably reflect availability of free G. RGS may stabilize an active GPCR/G protein/channel complex to limit the diffusion time required for activation and deactivation. To affect deactivation rates, RGS proteins may modulate the lifetime of free G. RGS and G compete for the same face of G, and binding to G is predicted to be mutually exclusive (Tesmer et al., 1997). Unstable and transient RGS/G-GDP interactions following GTP hydrolysis would promote G-GDP and G reassociation and increase deactivation rates. In support of this idea, RGS proteins block slow-acting G-mediated responses such as MAPK signaling in mammalian cells (Yan et al., 1997) and pheromone responses in yeast (Druey et al., 1996). Alternatively, formation of a stable RGS/G-GDP complex would prolong G availability thereby slowing deactivation rates and possibly enhancing the amplitude of response. Consistent with this idea, coexpression of the B/R4 family members RGS4 or RGS3 with GIRK markedly enhances basal current in a G-dependent manner (Bunemann and Hosey, 1998). Both processes may be at work in a cell-dependent context, depending on the signaling responses and proteins involved. Other studies clarify this point further by demonstrating that low levels of expressed RGS proteins modulate the rate of GIRK channel deactivation whereas higher levels of protein enhance the current amplitude (Keren-Raifman et al., 2001).

RGS proteins also modulate the kinetics of Ca²⁺ currents carried by voltage-dependent N-type Ca²⁺ channels. Stimulation of G_i/o-linked GPCR by various neurotransmitters inhibits N-type Ca²⁺ channels in neurons, and this inhibition is mediated by free G. In mammalian cells, overexpression of B/R4 RGS family members RGS3, RGS4, or RGS8 accelerates the rate of recovery of Ca²⁺ currents from neurotransmitter (G-mediated) inhibition, as well as decreasing the potency of agonist required (Jeong and Ikeda, 1998; Melliti et al., 1999). Consistent with these findings, expression of a mutant form of G_oA that is insensitive to endogenous RGS (DiBello et al., 1998; Lan et al., 1998) resulted in an increase in agonist potency, a marked reduction in recovery time, and an increase in the time to reach steady state after application of agonist (Jeong and Ikeda, 2000). Increased agonist potency and slow recovery times are consistent with RGS-mediated GAP effects on G and hence availability of free G. Reasons for RGS effects on the delayed rate to steady state are unclear, but could be explained if RGS promoted formation of a stable GPCR/G protein/channel complex that limited the diffusion time required for G (for further discussion, see Jeong and Ikeda, 2000).

RGS proteins also modulate the time course of phototransduction (Arshavsky and Pugh, 1998; He et al., 1998). In rod outer segments, elevated levels of cytosolic cGMP bind to and open channels to maintain resting membrane potential. In response to light, the photon-activated GPCR rhodopsin stimulates GTP binding to G_t, which in turn, binds to the inhibitory -subunit of cGMP-phosphodiesterase (-PDE). G_t-GTP sequesters -PDE and disinhibits the catalytic -subunit of PDE (-PDE), which becomes free to rapidly hydrolyze cGMP. In turn, a reduction in cytosolic cGMP levels causes channels to close and initiates electrical signaling pulses to the visual cortex. Remarkably, this entire signaling cascade proceeds within 150 ms with a recovery time of 200 ms while providing single photon reliability and maximal signal amplification. Intrinsic rates of rhodopsin-stimulated G_t GTPase activity are 100-fold too slow to account for the onset and deactivation of this signaling event, predicting the existence of a regulatory factor. It was recognized that an essential unidentified cellular factor, in concerted action with -PDE, served as a GAP for G_t to limit the lifetime of the signaling response (Angleson and Wensel, 1994; Arshavsky et al., 1994). Wensel and coworkers identified a membrane bound protein that is a GAP for G_t (Angleson and Wensel, 1993), and later showed that this protein was the RGS protein, RGS9 (He et al., 1998). In these studies, RGS9-1 was found to meet all of the requirements of the essential factor, i.e., it is expressed exclusively in rod outer segment, is tightly membrane-bound, and is observed to act synergistically with -PDE as a potent GAP for G_t. Thus, RGS9-1 is capable of determining the lifetime of active G_t-GTP/-PDE complex. To achieve this, RGS9-1 forms a high-affinity ternary complex with active G_t and -PDE (Slep et al., 2001) thereby eliminating the slow binding constants typical of protein-protein diffusion. Confirmation of the importance of RGS9-1 as the key modulator of the recovery step in phototransduction comes from recent studies of mouse retinas derived from homozygous RGS9(/) knockouts (Chen et al., 2000; Lyubarsky et al., 2001). Unlike RGS9, other RGS proteins (RGS4 and RGS-GAIP) fail to act cooperatively with -PDE as GAPs for G_t, even though each is an effective GAP for G_t in isolation (Nekrasova et al., 1997). In fact, these proteins inhibit rather than enhance -PDE-directed GAP activity toward G_t, suggesting a high degree of specificity for the proteins involved in this signaling response.

	IV. RGS Proteins Integrate G Protein Signals

Top Previous Next References

We have discussed examples of RGS proteins as modulators of both slow- and fast-acting G protein signals. Evidence suggests that the simple RGS proteins (members of the B/R4 and perhaps the A/RZ subfamilies) serve as highly regulated modulators of G protein signaling responses rather than as dedicated inhibitors. However, the larger more complex RGS proteins (members of the C/R7, D/R12, E/RA, and F/RL subfamilies) in many cases likely perform additional cellular functions. For example, although RGS9 is clearly an essential modulator of G_t-directed phototransduction, it differs from the simple RGS proteins in that it is larger and exists as multiple splice variants (RGS9-1 and RGS9-2). Both forms of RGS9 bind the G5 subunit and the longer variant, RGS9-2, may also bind other signaling proteins (Chen et al., 2001b) at its C-terminal extension (Figs. 1 and 4; Table 1). The extended N and C termini confer additional regulatory and/or signaling functions to RGS9 and other complex RGS proteins. Emerging concepts suggest that these complex RGS proteins link active G subunits to other signaling pathways to serve as multifunctional integrators of G protein signaling.

Members of the C/R7 family (RGS6, 7, 9, and 11) could represent important contributors to signal integration from multiple receptors. In addition to their RGS domain, C/R7 subfamily proteins contain DEP, R7H, and GGL domains (Fig. 1). GGL domains bind the G subunit G5 specifically and with high affinity (Snow et al., 1999). G5 is unique among G subunits in that it has reduced sequence homology with other family members (53%), is expressed almost exclusively in the nervous system, and does not bind well to most G subunits (Watson et al., 1994). In studies designed to identify binding partners for G5, RGS7 was the main binding partner to copurify out of retinal extracts (Cabrera et al., 1998). Further studies showed that the other members of the C/R7 family also bind G5 with high affinity, both in vitro and in vivo (Snow et al., 1998b; Posner et al., 1999; Zhang and Simonds, 2000). This interaction, taken together with evidence from genetic studies in lower eukaryotes (described below), has led to working models that propose that RGS/G5 complexes can potentially substitute for G in G heterotrimers, although no direct evidence for this has been reported.

Genetic evidence from C. elegans supports this model, showing that C/R7 subfamily members couple competing G protein-regulated behaviors (Hajdu-Cronin et al., 1999; Chase et al., 2001). In worms, the interplay between G_q- and G_o-linked signals controls egg laying and locomotion. Proposed models suggest that serotonin (G_o-linked), and acetylcholine (G_q-linked), cross-regulate these behaviors (Hajdu-Cronin et al., 1999). Activation of G_o causes lethargic movements, delayed egg laying, and reduced mating, whereas activation of G_q has the opposite effect. Egl-10 and Eat-16, C. elegans C/R7-like proteins which bind the G5 homolog GPB-2, cross-regulate these signals. Loss-of-function mutations in Eat-16 suppress the constitutively active G_o phenotype, indicating that Eat-16 acts downstream of G_o. However, reducing the levels of G_q reverses the Eat-16 loss-of-function phenotype, indicating that Eat-16 acts downstream of G_o by limiting G_q activity (Chase et al., 2001).

The genetic evidence can be interpreted in a number of ways. In the simplest model, the RGS/GPB-2 dimers act only as negative regulators of their respective G homologs (van der Linden et al., 2001). However, this interpretation does not fully take into account the interplay between the different G signaling pathways. A more comprehensive model is derived from scenarios proposed by several research groups (Guan and Han, 1999; Hajdu-Cronin et al., 1999; Sierra et al., 2000) (depicted in Fig. 3A). In this model, Eat-16, GPB-2, and G_o-GDP exist as a heterotrimer at rest whereas Egl-10 and GPB-2 complex with G_q-GDP. In this case, Eat-16 acts as a G subunit for G_o whereas Egl-10 is a G for G_q. Neurotransmitter activation of either receptor releases the RGS protein. When the Eat-16/GPB-2 complex disengages from G_o-GTP, Eat-16 is free to act as a GAP on G_q via its RGS domain. In parallel, activation of G_q by competing neurotransmitters releases Egl-10/GPB-2 from G_q-GTP and allows Egl-10 to limit G_o signals. While the physiological purpose of this cross-talk is uncertain, reciprocal inhibition among neurotransmitters allows the worms a tighter level of control over reproduction and locomotion (van der Linden et al., 2001). Many aspects of this model remain to be tested. For example, there is no direct evidence that the RGS-G5 dimers bind the inactive G homologs. Even given the need for further testing, this scenario provides a testable model to explain the physiological role of RGS-G5 interactions.

View larger version (79K): [in this window] [in a new window]   Fig. 3.   RGS proteins as integrators of G protein signaling. Panel A, speculative model for RGS-directed integration of G protein signaling in the worm C. elegans. The worm RGS proteins Egl-10 and Eat-16 each form a heterotrimeric complex with the C. elegans equivalent of G5 (GPB-2) and G subunits to couple to GPCR (left panel). Following receptor activation by either acetylcholine (Ach) (middle panel) or serotonin (5-HT) (right panel), the corresponding G protein signaling pathway is stimulated and the linked RGS/G5-like complex is released to inhibit the opposing G protein signaling pathway. Gq activation of the Egl-8 (PLC) and formation of diacylglycerol stimulates locomotion and egg laying, whereas Go activation of GDK-1 (diacylglycerol kinase) opposes this behavior. See text for further details. Panel B, a hypothetical model for RGS-directed integration of G protein signaling in the mammalian central nervous system. Left, an R7 subfamily protein (containing DEP, R7H, GGL, and RGS domains) exists as a complex with G5 (RGS/G5) in the cytosol and at the membrane. RGS/G5 forms a heterotrimeric complex with Gq and couples to a GPCR (R1). Right, activation of receptor (R1) with neurotransmitter (NT1) stimulates Gq signaling and releases RGS/G5, which is available to block subsequent signaling initiated by a second neurotransmitter (NT2) that activates a GPCR (R2) linked to Go. See text for a more detailed description of the model and appropriate references. Panel C, PDZ-RGS3 mediates reverse signaling by Ephrin-B. In response to EphB receptor binding to EphB on granule cells, the PDZ domain of PDZ-RGS3 binds the C-terminal tail of Ephrin-B at the plasma membrane. This positions PDZ-RGS3 to block G protein-stimulated granule cell migration initiated by SDF-1 (guidance factor) binding to the CXCR4 GPCR. See text for further details about the model and appropriate references. Panel D, the RGS-like protein p115RhoGEF mediates G13 stimulation of RhoA. Following activation of a G13-linked GPCR by hormone or neurotransmitter (H/NT), p115RhoGEF (containing RGS, DH, and PH domains) is recruited from the cytosol and associates with G13 at the plasma membrane. The DH domain of p115RhoGEF recruits and stimulates GTP binding to RhoA, which activates Rho kinase and downstream changes in cell morphology, adhesion, and motility. The GAP activity of the RGS domain feeds back to shut-off G13 signaling. See text for a more detailed description of the model and appropriate references.

Similar pathways may exist in mammalian cells, although in vitro studies using purified mammalian RGS proteins have failed to show association of G-GDP/RGS/G5 heterotrimers, and RGS/G5 dimers do not mimic conventional G signals such as modulation of adenylyl cyclase or activation of PLC (Posner et al., 1999). However, based on recent unpublished studies involving RGS9/G5 dimers and their involvement in M2AchR signaling, Sondek and Siderovski (2001) have proposed a model in which members of the C/R7 subfamily, complexed with G5, act as a G. A variation of this model is diagramed in Fig. 3B. If RGS/G5 dimers can substitute for G, several questions remain to be addressed. These include how G subunits discriminate between conventional G subunits and RGS/G5 complexes in cells, how RGS/G5 target specific G subunits, and what role RGS GAP activity plays in their signaling functions. Unlike other G dimers, RGS/G5 complexes, particularly RGS7/G5, are both membrane bound and cytosolic (Cabrera et al., 1998; Rose et al., 2000). Signaling roles for the cytosolic subpopulation are unclear but may relate to interactions with cytosolic-binding proteins such as 14-3-3 (discussed below).

In addition to the RGS and GGL domains, all C/R7 family members also contain several other conserved regions that may influence their function, including the DEP domain. DEP domains exist in many unrelated signaling proteins and may influence membrane interactions or possibly influence RGS/G interactions. C/R7 family members also share a conserved region of residues identified on the PfamB data base (Fig. 1) (Sondek and Siderovski, 2001), which we term the R7H domain (Fig. 1). Although roles for this domain are unknown, one possibility is that it may be involved in regulated membrane attachment. This attachment could occur through post-translational lipid modifications, which, in at least one C/R7 family member, RGS7, localize to this region (Rose et al., 2000). This conserved region may also have independent signaling functions through interactions with as yet unidentified binding partners.

Models proposing that C/R7 family members act as both a G and a GAP can help account for signal integration at multiple levels. Such a mechanism could help focus signals from multiple receptors activated by a single ligand, integrate signals from competing ligands, or auto-inhibit a G for which the RGS acts as both a G and a GAP. Because all C/R7 family members, as well as G5, are found exclusively in brain, fine-tuning of neuronal transmission is a hypothetical function of these complexes. For example in the striatum, where the C/R7 family member RGS9-2 is highly enriched, this protein could help fine-tune glutamate signals and present a therapeutic target to help regulate striatal activity in diseases such as Parkinson's. In the striatum, glutamate is released onto cells expressing multiple subtypes of metabotropic receptors including G_q-linked group I mGluR (mGluRI), as well as G_i/o-linked group II mGluR (mGluRII). In a hypothetical model (Fig. 3B), the RGS9-2/G5 complex associates with the mGluRI subpopulation of postsynaptic receptors, acting as a G subunit. When glutamate is released, both mGluRI and mGluRII receptors would be activated and mGluRI-linked G_q would release RGS9-2/G5. The RGS domain of RGS9-2 would then be free to act as a GAP for the G_i/o subunits activated by mGluRII receptors. Through this interaction, RGS9-2 expression could limit ion channel modulation through G_i/o-linked receptors while enhancing Ca²⁺ and PKC signals from G_q.

Additional binding partners may modulate RGS/G5 function. For example, RGS9-2 interaction with the protein evectin could affect its localization or interactions (Chen et al., 2001b). Although little is known about the signaling properties of evectins, they are membrane-anchored proteins with an N-terminal PH domain and may therefore secure RGS9-2 to the membrane. Other C/R7 family members also recruit a variety of additional binding partners including -PDE, polycystin, and 14-3-3, as discussed elsewhere in the text. Interactions between C/R7 family members and these proteins could recruit the RGS to membrane compartments, direct interactions with G or G5, or alter RGS/G5 activity, and thereby regulate the novel signaling functions of this RGS subfamily.

A compelling example of RGS proteins directly linking different receptor systems is found in mouse cerebellar granule cells (Lu et al., 2001). Ephrin B (EphB), a single transmembrane-spanning cell surface ligand for a tyrosine kinase receptor, is implicated in a variety of developmental processes. Although membrane-spanning ligands were traditionally thought to signal exclusively through their receptors, evidence now indicates that many also relay signals through their own C termini. In the case of EphB, this "reverse signaling" mediates several developmental processes including axon guidance and vascularization and allows proper migration of granule cells in the developing cerebellum. The EphB tail binds directly to the PDZ domain of PDZ-RGS3, a potential new splice variant of RGS3 identified by yeast two-hybrid screening using the cytoplasmic tail of EphB as bait. In migration assays, EphB inhibits chemoattraction by the chemokine receptor CXCR4, a G protein-coupled receptor. PDZ-RGS3 mediates this inhibition, which requires both the PDZ and RGS domains of the protein. The proposed model for EphB reverse signaling is depicted in Fig. 3C (Lu et al., 2001). At birth, granule cells are retained at the pia by chemoattraction through activation of CXCR4. At approximately postnatal day 3, EphB is up-regulated and interacts with its receptor, which promotes binding of PDZ-RGS3 to the cytoplasmic tail of EphB. With the PDZ domain recruited to EphB, the RGS domain is free to inhibit CXCR4 signals and allow cells to begin migrating through the cerebellum. A caveat to this model is that EphB and CXCR4 must be in close proximity, since they directly link through PDZ-RGS3. In this model, PDZ-RGS3 is an EphB effector, directly integrating the tyrosine kinase ligand with a GPCR signaling cascade.

One of the most exciting areas of research in the RGS field is in the newfound appreciation that RGS proteins can directly link G to nontraditional signaling cascades, particularly to regulation of monomeric GTPases. The first example of this was the RhoA exchange factor p115RhoGEF (Hart et al., 1998; Kozasa et al., 1998). The structure of RhoGEFs consists of an N-terminal RGS domain, a more C-terminal Rho guanine nucleotide exchange factor (DH) domain, and a PH domain (Fig. 1). In reconstituted systems using purified proteins, the RGS domain specifically interacts with and acts as a GAP for G_12/13 family members whereas the DH domain exchanges GTP for GDP on RhoA. A model illustrating G_12/13 regulation of Rho signaling is pictured in Fig. 3D. At rest, the RGS and DH domains of the cytosolic RhoGEF inhibit one another. After receptor activation, G₁₃-GTP recruits RhoGEF through interaction with the RGS domain, releasing the DH domain that then binds RhoA, draws it to the membrane and initiates nucleotide exchange. These proteins therefore directly link GPCR activation to the cytoskeletal changes initiated by activated RhoA. The discovery of the p1