Structural Determinants of G-protein α Subunit Selectivity by Regulator of G-protein Signaling 2 (RGS2)*

“Regulator of G-protein signaling” (RGS) proteins facilitate the termination of G protein-coupled receptor (GPCR) signaling via their ability to increase the intrinsic GTP hydrolysis rate of Gα subunits (known as GTPase-accelerating protein or “GAP” activity). RGS2 is unique in its in vitro potency and selectivity as a GAP for Gαq subunits. As many vasoconstrictive hormones signal via Gq heterotrimer-coupled receptors, it is perhaps not surprising that RGS2-deficient mice exhibit constitutive hypertension. However, to date the particular structural features within RGS2 determining its selectivity for Gαq over Gαi/o substrates have not been completely characterized. Here, we examine a trio of point mutations to RGS2 that elicits Gαi-directed binding and GAP activities without perturbing its association with Gαq. Using x-ray crystallography, we determined a model of the triple mutant RGS2 in complex with a transition state mimetic form of Gαi at 2.8-Å resolution. Structural comparison with unliganded, wild type RGS2 and of other RGS domain/Gα complexes highlighted the roles of these residues in wild type RGS2 that weaken Gαi subunit association. Moreover, these three amino acids are seen to be evolutionarily conserved among organisms with modern cardiovascular systems, suggesting that RGS2 arose from the R4-subfamily of RGS proteins to have specialized activity as a potent and selective Gαq GAP that modulates cardiovascular function.

G protein-coupled receptors (GPCRs) 4 form an interface between extracellular and intracellular physiology, as they con-vert hormonal signals into changes in intracellular metabolism and ultimately cell phenotype and function (1)(2)(3). GPCRs are coupled to their underlying second messenger systems by heterotrimeric guanine nucleotide-binding protein ("G-proteins") composed of three subunits: G␣, G␤, and G␥. Four general classes of G␣ subunits have been defined based on functional couplings (in the GTP-bound state) to various effector proteins. G s subfamily G␣ subunits are stimulatory to membrane-bound adenylyl cyclases that generate the second messenger 3Ј,5Ј-cyclic adenosine monophosphate (cAMP); conversely, G i subfamily G␣ subunits are generally inhibitory to adenylyl cyclases (4). G 12/13 subfamily G␣ subunits activate the small G-protein RhoA through stimulation of the GEF subfamily of RGS proteins, namely p115-RhoGEF, LARG, and PDZ-RhoGEF (5). G q subfamily G␣ subunits are potent activators of phospholipase-C␤ enzymes that generate the second messengers diacylglycerol and inositol triphosphate (6); more recently, two additional G␣ q effector proteins have been described: the receptor kinase GRK2 and the RhoA nucleotide exchange factor p63RhoGEF (7,8).
The duration of GPCR signaling is controlled by the time G␣ remains bound to GTP before its hydrolysis to GDP. RGS proteins are key modulators of GPCR signaling by virtue of their ability to accelerate the intrinsic GTP hydrolysis activity of G␣ subunits (reviewed in Refs. 9 and 10). RGS2/G0S8, one of the first mammalian RGS proteins identified (11) and member of the R4-subfamily (10), has a critical role in the maintenance of normostatic blood pressure both in mouse models (12,13) and in humans (14,15); additionally, Rgs2-deficient mice exhibit impaired aggression and increased anxiety (16,17), behavioral phenotypes with potential human clinical correlates (18,19).
Although many RGS proteins are promiscuous and thus act on multiple G␣ substrates in vitro (e.g. Ref. 20), RGS2 exhibits exquisite specificity for G␣ q in biochemical binding assays and single turnover GTPase acceleration assays (20,21). Consistent with this in vitro selectivity, 5 mice deficient in RGS2 uniquely exhibit constitutive hypertension and prolonged responses to 5 Independent reports (e.g., Refs. [57][58][59] have demonstrated that, in membrane-reconstitution systems containing GPCRs and G-protein heterotrimers, RGS2 can affect the agonist-dependent GTPase activity of G i -coupled signaling systems. The basis for this discrepancy between RGS2 selectivity for G␣ q in binary, solution-based assays and apparent RGS2 activity on G␣ i in reconstituted systems has not yet been resolved, but it is important to note that RGS2 (like other RGS proteins) is known to interact with other components of GPCR signal transduction beyond G␣ subunits (60), including isoforms of the G␣ i effector target, adenylyl cyclase (37).
vasoconstrictors, as would be expected upon loss of a potent negative regulator of G␣ q that mediates signaling from various vasoconstrictive hormones such as angiotensin II, endothelin, thrombin, norepinephrine, and vasopressin (22). In addition, RGS2-deficient mice respond to sustained pressure overload with an accelerated time course of maladaptive cardiac remodeling (23), a pathophysiological response that evokes myocardial hypertrophy known to be critically dependent on G␣ q signaling (24,25).
To gain insight into the structural basis of the unique G␣ substrate selectivity exhibited by RGS2, a series of point mutants in RGS2 were evaluated that enable this protein to bind and accelerate GTP hydrolysis by G␣ i ; we subsequently delineated the structural determinants of the G␣ i /mutant RGS2 interaction using x-ray crystallography. Three key positions, first identified by Heximer and colleagues (21) and highlighted in our structural studies as key determinants of RGS2 substrate selection, were also found to be conserved throughout the evolution of the RGS2 protein in a manner suggestive of specialization toward cardiovascular signaling modulation.

EXPERIMENTAL PROCEDURES
Chemicals and Assay Materials-Unless otherwise noted, all chemicals were the highest grade available from Sigma or Fisher Scientific (Pittsburgh, PA).
Protein Expression and Purification-Using ligation-independent cloning, DNA encoding human RGS2 (Lys 71 -His 209 ), fused to either hexahistidine alone (His 6 ) or to His 6 -tagged enhanced yellow fluorescent protein (YFP), was hybridized into a Novagen (San Diego, CA) pET vector-based prokaryotic expression construct as previously described (26,27). Point mutations corresponding to Cys 106 to serine (C106S), Asn 184 to aspartate (N184D), Arg 188 to glutamate (R188E), and Glu 191 to lysine (E191K) were made using QuikChange site-directed mutagenesis (Stratagene, La Jolla, CA). For expression of hexahistidine-and His 6 -YFP fusion RGS2 constructs, BL21(DE3) Escherichia coli were grown to an A 600 nm of 0.7-0.8 at 37°C before induction with 0.5 mM isopropyl ␤-D-thiogalactopyranoside. After culturing for 14 -16 h at 20°C, cells were pelleted by centrifugation and frozen at Ϫ80°C. Bacterial pellets were then resuspended in N1 buffer (50 mM HEPES, pH 8.0, 400 mM NaCl, 30 mM imidazole, 5% (w/v) glycerol) and lysis of bacterial slurry was performed using an Emulsiflex (Avestin, Ottawa, Canada) according to the manufacturer's instructions. Cellular lysates were clarified by centrifugation at 100,000 ϫ g for 30 min at 4°C. The supernatant was applied to a Ni 2ϩ chelating fast protein liquid chromatography column (FF HisTrap; GE Healthcare, Piscataway, NJ), washed with 7 column volumes of N1 buffer then 3 column volumes of N1 buffer containing an additional 30 mM imidazole before elution of recombinant RGS2 protein with N1 buffer containing 300 mM imidazole. His 6 -tagged RGS2 protein was cleaved with tobacco etch virus protease overnight at 4°C and dialyzed into N1 buffer containing 5 mM dithiothreitol. To separate residual His 6 -RGS2 from untagged, cleaved RGS2, the protein was passed over a second Ni 2ϩ -chelating fast protein liquid chromatography column. The flow-through was pooled, concentrated to final volume of ϳ5 ml, and resolved using a calibrated 150-ml size exclusion column (Sephacryl S200; GE Healthcare) using S200 buffer (10 mM HEPES pH 8.0, 300 mM NaCl, 5 mM dithiothreitol, 5% (w/v) glycerol). Fractions containing monodisperse protein were then pooled and concentrated to ϳ500 M, as determined by A 280 nm measurements upon denaturation in 8 M guanidine hydrochloride. Concentration was calculated based on the predicted extinction coefficient (ProtParam, Swiss Institute for Bioinformatics). Additional details regarding protein purification for crystallography can be found online at the SGC Oxford website. Human RGS16 constructs, C-terminal biotinylated G␣ i1 , N-terminal deleted (⌬N30) G␣ i1 , CFP-G␣ i1 , and G␣ i3 were purified exactly as previously described (20,28,29).
Single Turnover GTPase Assays-Single turnover [␥-32 P]GTP hydrolysis assays were conducted using recombinant G␣ i1 and various concentrations of RGS proteins as previously described (20). Briefly, 100 nM G␣ i1 in reaction buffer (50 mM Tris pH 7.5, 0.05% C 12 E 10 , 1 mM dithiothreitol, 10 mM EDTA, 100 mM NaCl, and 5 g/ml bovine serum albumin) was incubated for 10 min at 30°C with 1 ϫ 10 6 cpm (2 nM) of [␥-32 P]GTP (specific activity of 6500 dpm/Ci). The reaction was then chilled on ice for 5 min prior to the addition of 10 mM MgCl 2 and 100 M unlabeled GTP␥S (final concentration) with or without added RGS protein. Reactions were kept on ice and 100-l aliquots were taken at the indicated time points, quenched in 900 l of charcoal slurry, and centrifuged before 600-l aliquots of supernatant were counted via liquid scintillation.
Structure Determination-Purified G␣ and RGS2(C106S, N184D,E191K) proteins were mixed at a molar ratio of 1:1.5 and incubated at 4°C for 20 min. The sample was passed through an S200 gel filtration column pre-equilibrated with 25 mM HEPES, pH 7.5, 150 mM NaCl, 5% glycerol, 2 mM dithiothreitol, 100 M AlCl 3 , 20 mM sodium fluoride, and 100 M GDP. Protein fractions that eluted as a complex were identified using SDS-PAGE and the fractions were pooled and concentrated to 23 mg/ml prior to crystallization condition screens using a 150-nl drop volume with an TTP Labtech Mosquito nanoliter liquid-handling system. The crystal of the RGS2(C106S,N184D,E191K)-G␣ i3 complex used for data collection was crystallized by vapor diffusion in sitting drops of 400 nl of protein and 200 nl of reservoir solution containing 0.1 M HEPES, pH 7.5, and 2 M ammonium sulfate (TTP Labtech Mosquito).
After cryoprotection in a solution of 2 M ammonium sulfate, 0.1 M HEPES, pH 7.5, and 20% (w/v) D-glucose, crystals were flash cooled in liquid nitrogen. A complete data set was col-lected at 100 K on a Rigaku/MSC FR-E rotating anode x-ray generator equipped with an R-AXIS HTC image plate detector. Diffraction images were evaluated with MOSFLM (31), and data were scaled using SCALA (32). The crystal belonged to the space group P3 2 21 with unit cell dimensions a ϭ 114.54 Å, b ϭ 114.54 Å, and c ϭ 99.33 Å. A molecular replacement solution was found in this space group using PHASER (33) with the RGS10/G␣ i3 complex (PDB code 2IHB) as the search model. The RGS2 coordinates from PDB code 2AF0 were superimposed onto the RGS10 coordinates of the RGS10/G␣ i3 positioned complex and rigid body refinement into the electron density was performed using REFMAC5 (34). Difference density in the GDP binding site was modeled using the higher resolution structure of G␣ i3 in the RGS8/G␣ i3 complex (PDB code 2ODE) with one molecule of GDP, a tetrafluoroaluminate ion, and a magnesium ion coordinated by two additional water molecules. Several rounds of manual rebuilding in COOT (35) and restrained refinement with REFMAC5 (34), using Translation/ Libration/Screw (TLS) groups calculated with TLSMD (36), resulted in the final structural model described in Table S1. Coordinates of the RGS2(C106S,N184D,E191K)-G␣ i3 complex were deposited in the Protein Data Bank with entry code of 2V4Z.
Cellular cAMP Signaling Assays-HEK293T cells were transfected using Lipofectamine 2000 (Invitrogen) in 6-well dishes with 4 g of total DNA including pGloSensor TM -20F cAMP biosensor plasmid (Promega Corp., Madison WI), dopamine D2 receptor, and empty vector, HA-RGS2(WT), or HA-RGS2(C106S,N184D,E191K). The RGS2 expression vectors encoded solely the RGS domain (amino acid Lys 71 -His 209 ; with an N-terminal HA epitope tag) to avoid the influence of non-RGS domain regions on adenylyl cyclase function (e.g. Ref. 37). Twenty-four hours post-transfection, cells were re-plated on poly-D-lysine-treated, clear-bottom, white 96-well plates at a density of 60,000 cells/well. Forty-eight hours post-transfection, culture medium was aspirated and cells were washed once with assay medium (Dulbecco's modified Eagle's medium with 10% fetal bovine serum (without phenol), 15 mM HEPES, pH 7.4) before being incubated for 2 h with 40 l/well of equilibration medium (assay medium with 4% GloSensor TM substrate). After 2 h, 6.6 l of 6ϫ final concentration of quinpirole (diluted in 10 M forskolin-containing assay medium) was added to each well and allowed to incubate for 10 min before GloSensor emission was read on a MicroBeta Plate Counter (PerkinElmer). Before plotting, luminescence counts were normalized to 100% maximal response for each condition to account for variability in Glo-Sensor expression, transfection efficiency, and the exact number of cells per well.

RESULTS AND DISCUSSION
Evaluating Point Mutations to RGS2 That Facilitate Interaction with G␣ i1 -RGS2 is the only member of the R4-subfamily known to bind specifically to G␣ q and not to G␣ i/o heterotrimeric G-protein subunits in vitro (20,21). Three amino acids within RGS2 were identified by Heximer and colleagues (21) as potential selectivity determinants in studies of G␣ o -directed GAP activity by RGS domain chimera derived from RGS2 and RGS4 sequences: namely, cysteine 106, asparagine 184, and glutamate 191. In the present study, we mutated these three amino acids to the highly conserved corresponding amino acids in R4-subfamily members (Cys 106 to serine, Asn 184 to aspartate, and Glu 191 to lysine; supplementary Fig. S1) to identify their respective contributions to G␣ substrate specificity.
RGS2. Although the magnitude of binding of the RGS2 double mutants was significantly less than that observed with the triple mutant, binding isotherms were nonetheless generated for all double mutants along with the triple mutant by injecting increasing concentrations of RGS2 protein over the G␣ i1 ⅐GDP⅐AlF 4 Ϫ surface. Using equilibrium binding analyses (Fig. 2), dissociation constants (K D values) for the RGS2/ G␣ i1 ⅐GDP⅐AlF 4 Ϫ interaction were estimated to be Ն5.3, Ն8.6, and Ն21.1 M, for C106S,N184D, E191K,N184D, and C106S,E191K, respectively, whereas the K D value was determined to be 1.25 M for the RGS2(C106S,N184D,E191K) triple mutant. Dissociation constants derived for the RGS2 double mutants are likely underestimated given an inability to attain saturating concentrations of these particular RGS2 analytes and thereby attain maximal binding (B max ).
To determine whether the enhanced affinity of the RGS2 triple mutant was the result of improvements to a canonical RGS domain/G␣ interaction interface, a highly conserved, surface-exposed arginine within this canonical interface (Arg 188 in the ␣VIII helix; Fig. S1) was mutated to glutamic acid. As has been shown for other RGS proteins (38), this single chargereversal point mutation (R188E) on the G␣-binding surface of the RGS2 triple mutant abolished binding to G␣ i1 ⅐GDP⅐AlF 4 Ϫ (Fig. 2B, bottom panel).
To confirm these SPR-derived results with an orthogonal technique of assessing the RGS domain/G␣ interaction, FRET measurements were performed using a YFP-RGS2 (C106S, N184D,E191K)/G␣ i1 -CFP pair, similar to the RGS4/G␣ i1 interaction FRET assay we have previously described (28). In the presence of GDP, aluminum tetrafluoride, and Mg 2ϩ ("AMF"), binding between RGS protein and the G␣ subunit is observed as an increase in YFP emission and decrease in CFP emission; in the presence of GDP alone, no binding is observed as expected (28,39) and so the ratio of YFP to CFP emission remains low. The relative affinities of wild type RGS2, RGS16, and RGS2 triple mutant were assessed by using this FRET binding assay in a competitive manner: unlabeled RGS protein was added in increasing amounts to a fixed concentration of YFP-RGS2(C106S,N184D,E191K) and G␣ i1 -CFP proteins. As expected, only unlabeled RGS2(C106S,N184D, E191K) and RGS16 proteins were able to inhibit the binding of the RGS2(C106S,N184D,E191K)/G␣ i1 FRET pair (Fig. 4), with observed IC 50 values of 526 nM (95% CI, 236 -1171 nM) and 115 nM (78 -168 nM), respectively. At no concentration tested was wild type RGS2 able to inhibit binding of the RGS2(C106S,N184D,E191K)/G␣ i1 FRET pair (Fig. 4B), consistent with the lack of affinity between wild type RGS2 and G␣ i subunits seen in our present SPR analyses and previously published studies (20,21).
Determinants of RGS2 GAP Activity on G␣ i1 in Vitro-Using SPR and FRET, we demonstrated that all three point mutations were required to facilitate high affinity binding of RGS2 to G␣ i1 . To determine whether this enhanced binding affected the ability of RGS2 to accelerate GTP hydrolysis by G␣ i1 , we performed single turnover GTPase assays with both wild type and triple mutant RGS2 proteins (Fig. 5). At no concentration tested was wild type RGS2 capable of increasing GTP hydrolysis over the intrinsic GTP hydrolysis rate of G␣ i1 (Fig.  5A). In contrast, a substoichiometric amount of RGS16 (a known G␣ i1 GAP; Ref. 40) was able to accelerate G␣ i1 GTPase activity; complete hydrolysis of bound GTP was observed in less than 15 s at 0°C. Unlike wild type RGS2, the RGS2(C106S,N184D,E191K) triple mutant was able to increase the rate of G␣ i1 GTP hydrolysis in a dosedependent manner (Fig. 5B); however, adding the R188E mutation to  the triple mutant resulted in a complete loss in GAP activity, consistent with the loss of G␣ i1 binding observed in SPR and FRET assays. To further confirm that the mechanism of action of the RGS2(C106S, N184D,E191K) triple mutant in increasing GTP hydrolysis by G␣ i1 was related to a canonical RGS domain/G␣ interaction and not the inadvertant addition of a contaminating GTPase, we assessed the effects of both RGS2(C106S, N184D,E191K) and RGS16 proteins on an RGS-insensitive G␣ i1 point mutant: specifically, G183S in the G␣ switch I region (41). Neither RGS2(C106S,N184D,E191K) nor RGS16 proteins were able to increase the intrinsic rate of GTP hydrolysis exhibited by this RGS-insensitive G␣ i1 (Fig. 5, C and D).

Determinants of RGS2 Activity on G i -coupled GPCR Signaling in
Cells-To validate in a cellular context the change in G␣ specificity exhibited in vitro by the RGS2(C106S, N184D,E191K) triple mutant, we used an intracellular cAMP biosensor to measure G i heterotrimermediated inhibition of forskolinstimulated cAMP production in HEK293T expressing the G i -coupled D2 dopamine receptor along with either wild type RGS2 or the RGS2(C106S,N184D,E191K) mutant. Upon treatment of transfected cells with forskolin, a robust increase in luminescence was observed from the cAMP sensor, reflecting direct activation of adenylyl cyclase by forskolin (4); upon administration of the dopamine D2/D3-receptor selective agonist, quinpirole, dose-dependent inhibition of this cAMP production was observed. Wild type RGS2 had no effect on the IC 50 of quinpirole (Fig. 6). However, cellular expression of the RGS2(C106S,N184D,E191K) triple mutant resulted in a significantly higher IC 50 for quinpirole (762 versus 18 nM for empty vector; Fig. 6), indicating that the gain of G␣ i -directed activity is readily apparent in a cellular context as well as in vitro for the RGS2 triple mutant.

RGS2/G␣ Complex Reveals Key Features of Selectivity
One of the three mutation sites within the RGS2 triple mutant, aspartate 184, is observed to form a double salt bridge (Fig. 8A and Fig.  S3) with the neighboring arginine 188, the latter being an ␣VIII residue completely conserved among all other R4-subfamily RGS domains (Fig. S1). Asparagine 184 of wild type RGS2, located between helix ␣VII and ␣VIII, is an aspartic acid in all other R4-subfamily RGS domains (Fig. S1). The additional terminal oxygen present in the aspartate side chain (and missing in asparagine) normally allows two salt bridges to be formed (Fig. 8A) with the conserved ␣VIII helix arginine residue (e.g. Arg 170 of RGS16, Arg 188 of RGS2). These salt bridges are not consistently observed in all unliganded RGS domain structures (20); however, this double salt bridge is present in all R4-subfamily RGS domains complexed with G␣ i/o subunits (Table S2), suggesting that their formation is important for making the RGS domain competent to bind G␣ i/o subunits. The importance of this Arg-Asn side chain interaction is supported by the loss of G␣ i binding and G␣ i -directed GAP activity when this ␣VIII helix arginine is mutated to glutamate (Figs. 2 and 5). The significance of this intramolecular interaction is further supported by observations that mutating the analogous ␣VIII helix arginine in RGS4 (Arg 167 ) and RGS12 (Arg 821 ) results in loss of G␣ i/o binding and G␣ i/o -directed GAP activity (38,43,44). Although Arg 188 of RGS2 does not make any critical contacts with G␣ i3 per se, it has a critical role in orienting Asp 184 (Fig. 8B) to form a conserved hydrogen bond with the main chain amide of a threonine residue in the G␣ switch I region (Thr 182 of G␣ i (20,42); Thr 183 of G␣ o (45)). In the structure of wild type, uncomplexed RGS2 (PDB code 2AF0; Ref. 20), asparagine at this position (Asn 184 ) FIGURE 7. Overall structural features of the RGS2(C106S,N184D,E191K)-G␣ i3 ⅐GDP⅐AlF 4 ؊ complex. A, the tertiary structure of G␣ i3 is composed of a Ras-like domain (red) and an all ␣-helical domain (blue) and is present in a transition-state mimetic form bound to a molecule of GDP (magenta) and tetrafluoroaluminate (AlF 4 Ϫ ) ion (gray/blue sticks). The three critical switch regions of G␣ (numbered Sw I to Sw III) are colored cyan. All three switch regions are engaged by the RGS2 RGS domain (yellow-green). Panel B represents the same structural model as in panel A, but rotated to highlight contacts made by residues serine 106, aspartate 184, and lysine 191 of the RGS2(C106S,N184D,E191K) triple mutant. This same orientation of the complex is presented in Fig. 8B. forms only a single hydrogen bond with terminal amine of Arg 188 and, rotated in this manner, the side chain cannot at the same time form a hydrogen bond with the Thr 182 backbone ( Fig. 8A and Table S2).
The aspartate substitution at position Asn 184 is critical to allow binding of RGS2 to G␣ i ; however, this single substitution alone is not sufficient to engender robust G␣ i binding (Fig. 1). Ser 106 is completely conserved among all R4-subfamily RGS domains except RGS2, in which this position is a cysteine residue (Fig. S1). Mutating Cys 106 to serine was also necessary to obtain high affinity binding to G␣ i subunits ( Figs. 1 and 2); whereas the Ser 106 side chain was not observed in the structural model to make any critical contacts with G␣ i3 , this residue is tightly packed among other residues (Fig. 8B). The structure of the RGS2(C106S,N184D,E191K)-G␣ i3 complex reveals that the ␤-carbon of Ser 106 is closely juxtaposed with the backbone carbonyl and ␥-hydroxyl of Thr 182 within switch I of G␣ i3 ; additionally, the ␣-carbon of Ser 106 is 3.8 Å from the terminal amine of Lys 210 within switch II of G␣ i3 . In conjunction with the SPR binding data, the observed tight packing of Ser 106 within the RGS2(C106S,N184D,E191K)-G␣ i3 complex suggests that the Cys 106 residue of wild type RGS2 prevents high affinity binding to G␣ i subunits by steric blockade of interactions with switch I and switch II of the G␣ subunit.
Although amino acid positions 106 and 184 are completely conserved among all R4-subfamily RGS domains except RGS2, the specific amino acid at position 191 is conserved only in its basic character, being either a lysine or an arginine in all R4-subfamily RGS domains (Fig.  S1). In wild type RGS2, this position is instead an acidic residue (glutamate 191). In the structural data derived from the RGS2(C106S, N184D,E191K)-G␣ i3 complex, electron density was present only for the ␣-, ␤-, and ␥-carbons of the mutated Lys 191 ; however, the final ordered carbon atom was found to be only 5.1 Å from the hydroxyl oxygen of Glu 65 in the ␣A helix of the G␣ i3 allhelical domain. Electron density was present to fit the C␣, C␤, C␥, and C␦ atoms of the Lys 191 residue (Fig. S3). The C⑀ and terminal amine were modeled by superimposing a Lys over those parts of the carbon atom chain that could be placed with electron density, revealing that this basic side chain would be less than 3.0 Å from the hydroxyl oxygen of G␣ i3 Glu 65 and thus within hydrogen bonding distance. It is possible that the high salt concentration necessary for crystallization screened the electrostatic contribution of this interaction away, resulting in a partially disordered side chain. In wild type RGS2, this salt bridge would be lacking and this position instead would create electrostatic repulsion between RGS2 Glu 191 and the allhelical domain of G␣ i3 . The importance of all-helical domain contacts to RGS protein selectivity for G␣ substrates has been previously speculated for the retinal-specific proteins RGS9-1 and G␣-transducin (46); our present finding with RGS2 provides one of the first structural insights into these interactions. These RGS domain/all-helical domain interactions, whereas typically underappreciated when considering the structural determinants of the RGS protein/G␣ interaction interface (e.g. Refs. 42 and cf. 20), may provide a unique point of interdiction to exploit with selective RGS protein inhibitors.
Unique Determinants of RGS2 G␣ q Selectivity Are Conserved among Species with Cardiovascular Systems-Current knowledge of G␣ selectivity suggests that R4-subfamily members, as well as proteins from the more ancestral RZ-subfamily (e.g. RGS17, -19, and -20), can act as GAPs for both G␣ i and G␣ q subunits (20,47), with the R4-protein RGS2 particularly attuned to G␣ q over G␣ i . Given its unique G␣ selectivity and its specialized role in cardiovascular signal transduction, RGS2 is likely to have arisen from the R4-subfamily in response to the development of cardiovascular structures and function.
In evolutionary terms, G␣ q emerged as the harbinger of a distinct and recognizable G␣ subfamily in fungi, and G␣ q subunits are present in all metazoans including sponges (48,49). Although RZ-subfamily RGS proteins are represented within the genomes of nematodes and arthropods (50), a distinct R4-subfamily does not appear until the evolution of urochor- and an additional bond with the backbone amine of the peptide bond connecting Thr 181 and Thr 182 , both located within switch I of G␣. Ser 106 of the RGS2 triple mutant is tightly packed with the backbone carbonyl and ␥-hydroxyl of G␣ Thr 182 , both being less than 3.9 Å from ␤-carbon of Ser 106 . Additionally, the G␣ switch II residue Lys 210 is 3.8 Å from the Ser 106 ␣-carbon.
dates. The genome of the urochordate Ciona intestinalis (sea squirt) encodes at least two RGS proteins (Fig. 7), an ortholog of the ancestral RZ-subfamily progenitor found in nematodes and arthropods, as well as a newly divergent R4-subfamily member (but not an RGS2 ortholog per se). With specialized tissues such as a notochord, digestive tract, single chamber heart, and gonads, C. intestinalis is commonly considered an excellent modern representative of the precursor to higher vertebrates (51,52). Agnatha (jawless fish) such as the sea lamprey Petromyzon marinus are considered the most primitive extant members of early vertebrates (53) and represent the first vertebrate to exhibit cardiac innervation (54). Although the P. marinus cardiovascular system is more advanced than the open system of C. intestinalis, it is still considered primitive in that it lacks an elastin-reinforced vasculature (55), coronary circulation, and a pericardial-contained fourth chamber (conus or bulbus arteriosus) to dampen systolic oscillations in blood pressure (54). Similar to C. intestinalis, the genome of P. marinus encodes at least two RGS proteins, the ancestral RZ member and a single R4 member (Fig. 9); however, no RGS2-like protein has yet been identified in this species.
As chordates evolved into the Gnathostomata (jawed vertebrates), the cardiovascular system rapidly developed coronary vessels, inhibitory vagal innervation, excitatory adrenergic innervation, and responses to prostaglandins, nitric oxide, and endothelin (56). This advance is marked in Danio rerio by the addition of multiple R4 proteins, specifically including a G␣ qspecific RGS2 protein (Fig. 9). This unique member of the R4-subfamily, with cysteine, asparagine, and aspartate at the three key specificity positions, is highly conserved in the extant representatives of all subsequent evolutionary steps: amphibians (e.g. Xenopus laevis and Xenopus tropicalis), avians (e.g. Gallus gallus) and mammals (Fig. 9); the three defining residues are seen to be unique among all R4-subfamily members within a given species (e.g. human R4 paralogs aligned in Fig. S1). Only amphibians (X. laevis and X. tropicalis) do not contain all three RGS2-defining amino acids (Fig. 9): whereas the RGS2 signature residue asparagine is present at position 184, serine (not cysteine) is present at position 106, and a neutral glutamine (not glutamate) is present at position 191. (Note that the latter glutamine is not seen in RGS2, RGS4, nor RGS20 paralogs.) Even though the conservation is not absolute in the amphibians, we have shown that asparagine in position 184 is sufficient on its own to significantly reduce G␣ i affinity (i.e. ϳ20-fold; compare K D of Ͼ21 M for the C106S,E191K RGS2 double mutant versus K D of 1.25 M for the C106S,N184D,E191K triple mutant in Fig.  2). In conclusion, the conservation of these three key residue positions suggests that RGS2 has indeed evolved from the R4-subfamily to be a specialized G␣ q GAP for the modern cardiovascular system by acquiring particular residues at one or more of three key positions that have been highlighted in our mutagenesis/crystallography studies.