A Global Map of G Protein Signaling Regulation by RGS Proteins

The control over the extent and timing of G protein signaling is provided by the regulator of G protein signaling (RGS) proteins that deactivate G protein a subunits (G a ). Mammalian genomes encode 20 canonical RGS and 16 G a genes with key roles in physiology and disease. To understand the principles governing the selectivity of G a regulation by RGS, we examine the catalytic activity of all canonical human RGS proteins and their selectivity for a complete set of G a substrates using real-time kinetic measurements in living cells. The data reveal rules governing RGS-G a recognition, the structural basis of its selectivity, and provide principles for engineering RGS proteins with deﬁned selectivity. The study also explores the evolution of RGS-G a selectivity through ancestral reconstruction and demonstrates how naturally occurring non-synonymous variants in RGS alter signaling. These results provide a blueprint for decoding signaling selectivity and advance our understanding of molecular recognition principles.


In Brief
Masuho et al. quantitatively define G protein selectivity of all of the canonical members of the regulator of G protein signaling (RGS) family. Based on this information, they determine the structural basis of selective RGS-G protein recognition and demonstrate how selectivity determinants evolved. They further show that the RGS-G protein selectivity is affected by the mutational genomic landscape and can be rationally altered.

INTRODUCTION
Heterotrimeric G proteins transduce a vast variety of extracellular stimuli, including hormones, ions, organic molecules, and light into the regulation of intracellular ''effectors'' to generate cellular responses (Neves et al., 2002). Collectively, G protein systems play a role in nearly every physiological process and in numerous pathologies (Heng et al., 2013;Kostenis et al., 2020;O'Hayre et al., 2014;Wang et al., 2018). G proteins are activated by the binding of GTP to the a subunits (Ga) that release them from inhibitory occlusion by the bg dimer (Gbg) (Glukhova et al., 2018;Lambert, 2008;Oldham et al., 2006;Syrovatkina et al., 2016). Mammalian genomes encode a conserved set of 16 Ga subunits, each possessing unique signaling properties and the ability to selectively engage a distinct set of effectors, including adenylate cyclases, phospholipase C isozymes, Rho guanine nucleotide exchange factors (GEFs), and ion channels (Hubbard and Hepler, 2006;Marinissen and Gutkind, 2001;Wettschureck and Offermanns, 2005).
The key determinant of G protein action in cells is their lifetime in an active state. Thus, the activation and deactivation of G proteins is tightly controlled and ought to occur with selectivity for individual G proteins to ensure the selectivity of downstream signaling (Siderovski and Willard, 2005;Syrovatkina et al., 2016;Wettschureck and Offermanns, 2005). Deciphering molec-ular mechanisms of this selectivity is of paramount importance for understanding how the signals are routed in the cells. A number of G protein activators have been described and demonstrated to act as GEFs on the Ga subunits with clear subtype selectivity (Cismowski et al., 1999;Garcia-Marcos et al., 2011;Tall et al., 2003). Among them, the largest class is the G protein-coupled receptor (GPCR) family (Fredriksson et al., 2003;Hilger et al., 2018;Mahoney and Sunahara, 2016). GPCRs exhibit clear preferences for activating particular Ga species, and there has been tremendous progress in understanding the molecular mechanisms in establishing this selectivity (Flock et al., 2017;Inoue et al., 2019;Masuho et al., 2015b;Okashah et al., 2019).
The opposing process of G protein deactivation occurs when G proteins hydrolyze guanosine triphosphate (GTP), a process assisted by the action of the GTPase-activating proteins (GAPs). The GAP action is essential for avoiding response saturation and for achieving temporal resolution dictated by individual physiological reactions (Ross, 2008). Most well-characterized GAPs for heterotrimeric G proteins belong to the regulator of G protein signaling (RGS) family, consisting of 20 canonical members in mammals (Dohlman and Thorner, 1997;Tesmer, 2009). RGS proteins bind to active Ga proteins and facilitate their GTPase activity, thereby accelerating the termination of G protein signaling (Berman et al., 1996b;Hunt et al., 1996; Ross  GoA  RGS3  RGS4  RGS8  RGS16  RGS5  RGS2  RGS13  RGS1  RGS21  RGS18  RGS17  RGS19  RGS20  RGS12  RGS14  RGS10  RGS6  RGS7  RGS9  RGS11 1/τ deactivation (s and Wilkie, 2000;Saitoh et al., 1997;Watson et al., 1996). It is now well established that this action of RGS proteins is crucial for achieving the physiologically relevant timing and extent of GPCR signaling (Hollinger and Hepler, 2002;Kimple et al., 2011;Neubig, 2015). Accordingly, the loss of RGS-mediated control leads to a range of pathologies observed in mouse models (Bansal et al., 2007;Gaspari et al., 2018;Lee et al., 2010;Senese et al., 2020) and is increasingly associated with human diseases (Shamseldin et al., 2016;Squires et al., 2018). Studies in several members of the RGS family indicate that they exert considerable selectivity in recognizing Ga (Heximer et al., 1997;Snow et al., 1998;Soundararajan et al., 2008;Tesmer, 2009;Wang et al., 1998). There has been significant progress documenting cases of selective RGS-Ga interactions (Hollinger and Hepler, 2002), analyzing the structural basis for this selectivity (Soundararajan et al., 2008;Taylor et al., 2016), and mapping amino acid residues involved in specific recognition (Kimple et al., 2009;Kosloff et al., 2011). Although these studies provide insights into the selectivity of RGS action for isolated cases, a comprehensive understanding of the complete landscape of Ga preferences of RGS proteins is still lacking. This study presents a map of Ga selectivity for all canonical RGS proteins. We monitored the temporal regulation of GPCRmediated G protein signaling and quantitatively characterized the GAP activity of the RGS proteins, testing nearly all of the theoretically possible Ga-RGS pairings (300 combinations). Using the functional activity as a readout in the context of a physiologically relevant cellular environment allowed us ot document the preferences of RGS proteins for Ga substrates, revealing pairings and disallowed combinations. This information led to the identification of molecular determinants involved in the selectivity of Ga-RGS recognition. Applying computational algorithms, we also show how these determinants have evolved and can be used to create designer RGS proteins with novel selectivity profiles. Analysis of human genomic data further suggests that genetic variations in RGS selectivity determinants may contribute to non-disease traits, pathological dysregulation of GPCR signaling, and variable responsiveness to drug treatments.

Assaying Activity of All Canonical RGS Proteins on Ga Deactivation with a Real-Time Kinetic Approach in Living Cells
To test their possible RGS-Ga coupling systematically, we used a cell-based system that provides a cellular environment to study the action of RGS in the context of GPCR signaling. This assay monitors RGS-induced acceleration of G protein deactivation by real-time bioluminescence resonance energy transfer (BRET) strategy tracking the kinetics of heterotrimer re-association upon antagonizing GPCR, a reaction catalyzed by RGS proteins physiologically ( Figure 1A). The key features of the assay include a ''bystander'' approach that allows the use of unmodified Ga subunits ( Figure 1B) and full-length RGS proteins ( Figure 1C).
Using a set of GPCRs with varying Ga selectivity, we recorded the deactivation kinetics of 15 Ga subunits (omitting sensory Ga t1 , Ga t2 , and Ga gust , but including the two common splice variants of Ga s and Ga o ) in the absence of exogenous RGS proteins. A combination of intrinsic differences in Ga properties and the action of endogenous RGS proteins in HEK293T/17 cells yielded characteristic baseline deactivation rates ( Figures 1D and 1E). Using a previously established approach (Masuho et al., 2013), we ensured that the deactivation kinetics were rate limited by the Ga GTPase activity. Disruption of RGS-Ga interactions by RGS-insensitive (DiBello et al., 1998;Lan et al., 1998) or GAPdeficient mutations (Druey and Kehrl, 1997;Srinivasa et al., 1998) substantially prolonged response recovery ( Figure S1). These mutations interfere with the conserved interaction of RGS proteins with the switch I region of the Ga subunits. Further controls demonstrated that (1) the exogenous expression of RGS proteins does not alter the expression of signaling molecules and sensors ( Figures S2A and S2B), (2) the different expression levels of GPCRs or the different amounts of active G proteins do not change the G protein deactivation rates (Figure S2C), and (3) deactivation rates are directly proportional to the amount of RGS ( Figure S2D). These results confirm that RGS action dictates the kinetics of G protein deactivation. Analysis of the deactivation traces for a representative Ga (Ga oA ) shows the varying impact of different exogenous RGS proteins on the kinetics of Ga termination ( Figures 1F and 1G).
To quantify the activity of RGS proteins, the baseline deactivation rates (1/t) of each Ga were subtracted from the deactivation rates in the presence of exogenous RGS proteins, yielding the k GAP parameter ( Figure 1H), a widely used metric of RGS catalytic activity (Ross, 2002). Plotting k GAP values for each of the Ga substrates provides a profile of relative activity for a given RGS protein. Analysis of the representative members of the RGS subfamilies using this strategy revealed differences in Ga preferences in a fingerprint-like fashion ( Figure 1I). These Ga selectivity fingerprints were not affected by differences in the RGS expression levels ( Figure S3A-S3D).

Principles of Ga Regulation by RGS Family
This strategy was applied to measure the activity of all 20 canonical RGS proteins on the deactivation of each of 15 Ga subunits in a total of 300 possible combinations. We optimized RGS expression levels, ensuring at least 3-fold acceleration of the (B and C) Phylogenetic trees of Ga subunits and RGS proteins. (D and E) The deactivation time course of 15 different G proteins. (F and G) The effect of RGS proteins on the deactivation of Ga oA . (H) Quantification of RGS action in G protein regulation. The rate constants in the absence (black) and presence of RGS4 (pink, left), and subtracted k GAP value for RGS4 (pink, right) are shown. (I) Ga selectivity fingerprints for representative RGS proteins. The k GAP were normalized to the largest value and plotted as corresponding vertices. The thickness of the lines represents the SEM of 3 independent experiments. Linear scale is used. (J) Heatmap of the normalized k GAP values. The black ''0'' values are assigned when no statistically significant GAP activity is detected. deactivation rate for the preferred Ga substrate to reliably assess even minor coupling. In particularly difficult cases (e.g., RGS13, RGS18), proteasomal blockade and codon optimization strategies were applied to augment RGS expression . Given the differences in the expression levels of various RGS proteins, we did not attempt to compare their absolute activities and instead focused on elucidating the relative differences in G protein preferences. Collectively, our results provide a comprehensive Ga selectivity profile for the entire RGS family ( Figures 1J and S4; Table S1).
Analysis of the RGS-Ga interaction network provided several key insights. We found that RGS proteins vary markedly in the breadth of their selectivity, with some members (e.g., RGS1) regulating all G i/o -and G q -type proteins, whereas others (e.g., RGS11) regulated only one Ga type, Ga o (Figures 2A, 2B, and S5A-S5C). The R4 and RZ subfamilies regulated the broadest range of Ga substrates (Figures 2A, 2B, and S5A-S5C). Collectively, R4 and RZ members regulated all Ga q and Ga i/o types with a spectrum of biases (Figures 1J,2A,and S4). For example, RGS3 and RGS4 preferred the Ga i/o over the Ga q , whereas RGS5 and RGS13 selected Ga q over Ga i/o . No RGS protein was shown to be specific for the Ga q subfamily. The narrowest selectivity was observed for the R7 subfamily, the members of which regulated Ga i/o proteins exclusively (but not Ga z ) with prominent selectivity for Ga o .
This analysis revealed that Ga subunits vary substantially in their sensitivity to RGS regulation ( Figures 2C and S5E). For example, we found Ga o to be the most indiscriminate Ga in that it was regulated by all of the canonical RGS proteins, whereas Ga z could be deactivated only by a limited number of RGS proteins ( Figures 2C, S5D, and S5F). We also noticed that a relatively slow rate (0.0021 ± 0.0003 s À1 ) of basal GTPase activity of Ga z possibly underestimated the selectivity of its regulation by RGS proteins when assessed by the k GAP parameter . Accordingly, we calculated a discrimination index (k dis ) defined by fold increase in the deactivation constant (1/t) upon the addition of RGS ( Figure S5J). Although considering that k dis did not change the overall picture of G protein selectivity for most RGS members, it was useful in showing the unique ability of RZ subfamily members to uniquely regulate Ga z (Figures S5K and S5L) amidst their significant activity on virtually all of the other Ga i/o and Ga q proteins based on the k GAP .
These data also revealed high selectivity in the regulation of the poorly studied Ga 15 . This G protein is activated by a wide range of GPCRs and thus likely contributes to a variety of cellular responses (Offermanns and Simon, 1995). We found that it has a very slow intrinsic deactivation rate (0.0081 ± 0.0006 s À1 ), making RGS regulation paramount for the temporal control of its signaling. Interestingly, Ga 15 can be deactivated by only a few RGS proteins ( Figure S5D), mostly Ga q -type-preferring R4 members and an RZ subfamily member, RGS17 ( Figures 2C and S5F).
These studies further revealed that no canonical RGS proteins could regulate the deactivation of Ga s , Ga olf , Ga 12 , or Ga 13 (Figure 1J). This outcome is perhaps not unexpected. Structural modeling shows that the switch I region of Ga 12/13 contains Lys-204 instead of a Thr present in all of the other Ga subfamilies in the corresponding position, rendering it incompatible with RGS binding (Figures S5M and S5N). Furthermore, the structure of the aB-C loop in the a-helical domain is also fundamentally different in Ga 12/13 , contributing to the steric occlusion of canonical RGS protein binding (Sprang et al., 2007). Similarly, the presence of Asp229 in Ga s , a position conserved as serine in all other Ga subfamilies, renders it incapable of RGS binding in Ga s family members (Natochin and Artemyev, 1998) due to collisions with the a5-a6 loop of RGS proteins ( Figures S5O and S5P). The Ga s D229S mutation restores the ability of RGS4 and RGS16 to bind and the ability of RGS16 to accelerate GTP hydrolysis on Ga s (Natochin and Artemyev, 1998).

RGS-Ga Recognition Patterns Selectively Shape Endogenous Secondary Messenger Signaling
To study how global patterns of RGS-Ga selectivity affect the processing of GPCR signals endogenously, we used striatal medium spiny neurons (MSNs) as a model ( Figure 3A). The MSNs were chosen because of their undisputed physiological importance and the critical role of several well-defined GPCRs in processing neuromodulatory inputs to these neurons (Girault, 2012; Xie and Martemyanov, 2011) ( Figure 3B). More important, several RGS proteins in the MSNs have been implicated in controlling behavioral responses to GPCR stimulation. The bestdocumented examples of these are RGS4 (Han et al., 2010;Michaelides et al., 2020), a member of the R4 subfamily, and RGS9 (Traynor et al., 2009), a member of the R7 subfamily.
We surveyed the expression landscape of RGS and Ga proteins by curating the available quantitative RNA sequencing (RNA-seq) data (Gokce et al., 2016). This analysis revealed a significant expression of 12 RGS genes, with RGS4 and RGS9 being the most abundant. Three members of the R4 subfamily (RGS4, RGS2, and RGS8) and 3 members of the R7 subfamilies (RGS9, RGS11, and RGS7) were estimated to be more highly expressed by at least an order of magnitude than other striatal RGS proteins ( Figure 3C). Interestingly, our dataset indicates that these RGS subfamilies have distinct patterns of Ga selectivity; the R7 RGS proteins are narrowly tuned for G i/o , whereas the R4 RGS members are capable of regulating a broad spectrum of Ga, including both G i/o and G q members (Figures 2B and 2C). Accordingly, transcripts encoding the members (Ga o , Ga i1-3 , Ga z , Ga q , and Ga 11 ) of the Ga i/o and Ga q subfamilies were abundantly expressed by the MSNs ( Figure 3C). Thus, we predicted that R4 RGS proteins would have a major influence on the processing of GPCR signals via both G i/o and G q pathways, whereas R7 RGS proteins would selectively affect only Ga i/o -mediated signals.
To test this prediction, we used biosensors to monitor the dynamics of second messenger pathway engagement downstream of both G i/o and G q while inactivating RGS proteins by CRISPR-Cas9 editing in the primary cultures of MSNs (Figure 3D). The G i/o activity was assessed by studying its inhibitory influence on cyclic AMP (cAMP) production in response to stimulation of the G i/o -coupled dopamine receptor D2 (D2R) by dopamine, whereas G q -type activity was monitored by Ca 2+ transients induced in response to the activation of the muscarinic M1/M3 receptors (M1/3R) by acetylcholine ( Figure 3B). Considering the intra-class similarity of RGS-Ga pairing and abundant expression of several members from each RGS class, we chose to simultaneously eliminate all MSN-expressed RGS proteins ll OPEN ACCESS belonging to the same subfamily by CRISPR-Cas9 editing. The elimination of either the R4 or the R7 subfamily resulted in a significantly enhanced cAMP response, consistent with the role of these RGS members in the deactivation of the G i/o pathway ( Figures 3E and 3F). In contrast, the elimination of R4 members but not R7 proteins augmented the Ca 2+ response, which is in line with their observed Ga selectivity profiles (Figures 3G and 3H).
We next tested the effect of overexpressing individual RGS proteins. We chose to focus on RGS2, an abundantly expressed RGS protein, widely believed to be G q selective based on biochemical measurements but able to regulate G i/o proteins (B) Ga selectivity of RGS subfamilies obtained by dividing the total GAP activity on each Ga subunit by the number of RGS proteins with statistically significant GAP activity (see Figure S5F). (C) RGS selectivity of Ga subunits obtained by dividing the total GAP activity of an RGS protein on all regulated Ga by the number of Ga subunits (see Figure S5C).  Article according to our data ( Figures 1J and S4). The overexpression of RGS2 had an opposite effect from eliminating RGS proteins and dramatically suppressed the amplitudes of both cAMP and calcium responses ( Figures 3I-3L). These observations indicate that the comprehensive RGS-Ga selectivity maps have predictive power in dissecting the logic of GPCR signal processing in an endogenous setting.

Flexibility of Ga Selectivity Encoded in the RGS Homology Domains
The analysis presented in this study revealed a wide range of Ga preferences across RGS proteins, which also feature considerable structural diversity (Riddle et al., 2005). This opens questions about the flexibility of recognition patterns across the family and the degree with which Ga selectivity is determined by the RGS domain shared by all RGS proteins. To address these questions in an unbiased way and gain insight into how the selectivity of mammalian RGS subfamilies may have evolved, we performed the reconstitution of ancestral RGS proteins ( Figure 4A). We traced the RGS family tree to reconstitute common ancestral RGS domains at three branch points before the diversification into the current four subfamilies and generated a series of chimeric RGS proteins ( Figure 4B). Examination of the Ga selectivity of the primal ancestral RGS protein (AncR4/Z/12/7) revealed that it regulated all Ga subunits that RGS proteins can regulate, except Ga 15 (Figures 4C and 4D). We next reconstructed two ancestral RGS proteins at the roots of the subfamily divisions (AncR4/Z and AncR12/7). Interestingly, AncR4/Z showed equally strong GAP activity toward Ga i/o and Ga q subfamilies, but not toward Ga z ( Figure 4D). Diversification of this precursor RGS subsequently generated various patterns of Ga i/o -and Ga q selectivity observed in current R4 and RZ subfamilies. The other ancestral RGS protein, AncR12/7, showed Ga i/o selectivity and was devoid of the ability to regulate the Ga q subfamily. This ancestral RGS gave rise to Ga i/o -selective R12 and R7 RGS proteins. These results suggest that Ga selectivity patterns of extant human RGS proteins resulted from a combination of specialization along the Ga i/o versus Ga q axis and de novo acquisition of Ga z and Ga 15 selectivity. This supports a predominantly evolutionary divergence model in which the primordial RGS precursor with balanced activity on different Ga substrates acquired various biases that followed different routes-for example, by suppressing the GAP activity toward the Ga q subfamily in R7 and R12 RGS or re-gaining the activity on Ga i/o subfamily by the R12 RGS. We thus conclude that the sequence composition of the RGS domain has considerable bearing on dictating the evolving Ga preferences of the RGS proteins, strongly suggesting that the major determinants of Ga selectivity are contained within the RGS domain.

Structural Determinants Governing the Selectivity of Ga Recognition by RGS Proteins
Elucidation of a Ga-RGS coupling map and demonstration of the crucial role of the RGS domain in determining the pairings prompted the identification of molecular determinants that govern their differential preferences. We compared the sequences of all human RGS domains, aligning them with reference to 20 available high-resolution structures that show the same conserved fold and preservation of key elements, with 9 a-helices and 10 loops ( Figure S6A; Data S1). RGS11, RGS13, RGS20, and RGS21 were not included in this analysis because their structures have not been reported. This analysis allowed us to develop a Common RGS Numbering (CRN) system for labeling amino acids relative to their structural position similar to what was previously done for Ga (Flock et al., 2015) and GPCRs (Ballesteros and Weinstein, 1995;Isberg et al., 2015) (Figures S6B and S6C). This system helps to identify the position of every residue with reference to the secondary structure. For instance, RGS4 Asn128, which directly binds to Ga i1 , is denoted as L6.10, indicating that this residue is the 10th amino acid located in loop 6 of the RGS domain ( Figure S6B). It should be noted that this nomenclature cannot be applied to the H6 region in the R12 subfamily because it is structurally distinct from other RGS subfamilies.
We further analyzed eight currently available structures of RGS/Ga complexes and found that all RGS and Ga subunits interact in a very similar manner, with low root mean square deviation (RMSD) in the range of 0.46-1.42 Å . In the RGS domain, there are 11 residues directly contacting Ga that are almost 100% conserved in all structures ( Figure S6B). In addition to these contacting positions, we found 20 residues on the RGS protein and 38 amino acids on Ga that contribute to the organization of binding interfaces based on their localization within the 5Å radius of any atom in the interface. On the RGS side, these (K) Average Ca 2+ response to acetylcholine (10 mM) in striatal neurons expressing jGCaMP7s following the overexpression of RGS2 (n = 16 neurons). (L) Quantification of maximum Ca 2+ amplitude from (K). One-way ANOVA followed by Fisher's least significant difference (LSD) (F and H). Unpaired t test (J) and (L). *p < 0.05 and **p < 0.01. Data are shown as means ± SEMs from 3-5 independent experiments. residues are distributed across 3 structural elements, 2 loops (H3-H4 and L6-H6) and 1 helix (H7-L9) ( Figures 5A and 5B). The surface on Ga is more distributed and involves both GTPase and a-helical domains.
To determine which elements most strongly contribute to the selectivity of Ga recognition, we analyzed these 31 RGS residues at the Ga-binding interface across all 20 human RGS paralogs in comparison with their orthologs from 21-65 animal species   Article (Figures 5C and 5D;Data S2). This analysis revealed 14 highly conserved positions across orthologs and paralogs, suggesting that they likely serve as invariable architectural pillars that organize Ga binding and/or GAP activity. These residues included all of the direct Ga-contacting positions found in the RGS4/Ga i1 complex ( Figure S6B). A minor fraction of the scattered residues was ortholog variable and neutrally evolving ( Figures 5C and 5E). The remaining fraction of ortholog-specific residues comprised 17 amino acids. Mapping them on the RGS domain structure showed that they are distributed at the periphery of the Ga-binding surface, surrounding the central positions of the conserved amino acids ( Figure 5D), suggesting that they may contribute to Ga selectivity by modulating the interaction. We subsequently refer to these peripheral amino acid residues that are variable among paralogs but conserved within their respective orthologs as Ga selectivity bar codes for RGS proteins.
To identify motifs in the RGS domain that contribute to establishing Ga selectivity, we reconstructed and analyzed the RGS-Ga interaction network at a single amino acid resolution (Figures 5F and 5G). This analysis confirmed that the vast majority of selectivity bar code residues are engaged in non-conserved contacts that vary between different structures of the RGS-Ga complexes ( Figure 5H). In contrast, the contacts involving the conserved residues were also predominantly conserved across RGS-Ga structures ( Figure 5H). The highest degree of conserved residue-residue contacts is observed for the H3-H4 region with G.H2 and switch I in Ga and for the L7-L9 region with switch I ( Figure 5F), indicating its crucial role as a structural backbone for RGS/Ga binding. In contrast, the interaction of the H7-L9 region with the a-helical domain showed the highest number of non-conserved contacts ( Figure 5G), suggesting that these domains could significantly contribute to the RGS/ Ga selectivity.
To better characterize the organization of the Ga-binding surface, we analyzed properties of the amino acids that form the Ga selectivity bar codes across different RGS subfamilies. This investigation revealed distinct patterns in accordance with the experimentally determined Ga selectivity patterns ( Figure 5I). For example, R4 and RZ subfamilies that are dually selective for the G i/o and G q proteins showed a similar distribution of hydrophobic and positively charged residues in the H7-L9 region; hydrophobic and positively and negatively charged residues in L6-H6; and a nucleophilic residue in H3-H4. In contrast, the G i/o -selective R12 family exhibited a different pattern featuring nucleophilic, aromatic, and amide residues in the H7-L9 region, and a unique positively charged patch in the L6-H6 lobe surrounded by the nucleophilic cluster. However, another pattern was observed in the narrowly tuned R7 proteins whose L6-H6 region is populated by small amino acids adjacent to the hydrophobic patch and a prominent positive charge in H7-L9. These findings reinforce the idea that the nature of amino acid properties at the selectivity bar code region on the Ga-binding interface of the RGS protein comprises major determinants of Ga recognition selectivity.

Design Principles for Engineering RGS Protein Selectivity
The identification of selectivity bar code residues in RGS proteins raises a question about their necessity and sufficiency in setting the selectivity of Ga recognition. This question was addressed experimentally, by transplanting the entire distributed pattern of selectivity residues ( Figure 6A). For these experiments, we chose RGS13 and RGS18, which belong to the same R4 subfamily but differ in G protein selectivity ( Figure 6C). RGS13 prefers G q members over the G i/o subfamily, whereas RGS18 equally regulates both G i/o and G q proteins. A comparison of their Ga selectivity bar codes indicates that they differ by 12 amino acid residues ( Figure S7A). All of the amino acid residues of RGS13 were replaced with the ones from RGS18, resulting in RGS13/18-F chimera ( Figure 6B). In agreement with the prediction based on our selectivity bar code model, RGS13/18-F protein exhibited RGS18-like Ga selectivity ( Figure 6C).
These experiments were then extended to RGS8 and RGS14, a pair that belongs to different subfamilies and also have markedly different Ga selectivity and composition of Ga selectivity residues ( Figures 6D and 6E). We identified 15 different amino acids within the Ga selectivity bar code different between these RGS proteins ( Figures 6D and S7B) and transplanted all of these from RGS14 into corresponding positions of RGS8, generating a ''full'' chimera (RGS8/14-F) ( Figure 6D). The RGS8/14-F chimera completely recapitulated the Ga fingerprint of RGS14 without gaining activity on G proteins not regulated by RGS8 or RGS14 ( Figure 6E). We further probed whether the change in selectivity could be achieved by mutating fewer bar code residues (i.e., by replacing only nine amino acid residues) (Figures S7B). The resulting ''partial'' RGS8/14 chimera (RGS8/14-P) had the same Ga q over Ga i/o preference as parental RGS8 ( Figure S7C). It thus failed to switch the Ga-selectivity fingerprint from the RGS8 to the RGS14 pattern, indicating that all of the bar code amino acids are required for establishing exact selectivity patterns of Ga-RGS recognition. Curiously, the RGS8/14-P mutant unexpectedly gained activity on Ga z ( Figure S7D), indicating that individual residues within the bar code can have an impact on the Ga selectivity of RGS proteins. Overall, these results indicate that identified selectivity bar codes are sufficient in dictating Ga substrate preferences.    Figures 7A-7D). On average, 13 MVs exist in each amino acid residue of RGS proteins ( Figure 7A). This density of MVs (14.8) was the highest outside of the RGS domain. In contrast, functionally important regions exhibited lower densities. The conserved and selectivity residues in RGS11 were the most variable among all of the RGS proteins ( Figures 7B  and 7C). The ratio of the MV density between selectivity and conserved residues revealed the highest MV frequency in the selectivity residues over the conserved residues in RGS17 (Figure 7D), suggesting likely extensive natural variation of Ga selectivity in RGS17.
To understand the functional implications of the observed variations, we investigated the impact of randomly chosen seven mutations across various positions in the selectivity bar code region of six RGS proteins by testing their activity on the panel of six Ga subunits ( Figure 7E). We found that all of the evaluated amino acid changes affected Ga selectivity. Notably, changes at L7.13 in RGS19 (R190W) increased the GAP activity toward Ga 15 , but decreased the activity on Ga o , Ga i1 , and Ga q without any influence on Ga z . Alterations in L6.8, H6.2, H7.6, and H7.9 selectively augmented the regulation of Ga i/o without diminishing the activity on other Ga. The balance between Ga i and Ga o regulation can also be affected by these mutations-for example, E98G (L6.8) in RGS13 preferentially increased activity toward Ga o over Ga i , while R351Q (H6.2) in RGS11 and N164S (H7.9) in RGS12 augmented Ga i regulation more than Ga o . Altering the H6.4 position in RGS9 M370K resulted in a net loss of activity across Ga regulated by this RGS.
Interestingly, variants in RGS proteins are also increasingly viewed as possibly contributing to pathological conditions due to generally disruptive effects (DiGiacomo et al., 2020;Squires et al., 2018). However, the exact mechanisms of functional alterations and implications for Ga selectivity for a vast number of cases remain unexplored. For instance, RGS16 has been recently implicated in insomnia (Hu et al., 2016;Lane et al., 2016), and knockout of this gene in mice disrupts circadian regulation (Doi et al., 2011). The genetic variation (rs1144566) in human RGS16 reported in the genome-wide association study (GWAS) catalog (Buniello et al., 2019) affects selectivity bar code residue H6.4 (Figures 7F and 7G). We experimentally evaluated the functional implication of minor allele variations in H6.4 of RGS16 prevalently occupied by arginine. Our data showed that the R137P mutation nearly completely abrogated the GAP activity of RGS16 for both of its representative preferred substrates, Ga o and Ga q , indicating a strong loss of function (Figure 7H). Curiously, the R137L substitution selectively compromised the activity of RGS16 only on Ga q without significant effects on the regulation of Ga o . These results indicate that mutations in the selectivity bar code may lead to RGS dysfunction associated not only with the complete loss of function but also with a more subtle alteration in the Ga selectivity.

DISCUSSION
In this study, we present a nearly complete map of Ga recognition selectivity for all 20 canonical human RGS proteins. The wealth of accumulated evidence in the past 2 decades since their discovery revealed that members of the RGS family exert two distinct effects on the G protein signaling. First, they accelerate G protein deactivation and thus control the duration of signaling. The slow intrinsic GTPase activity of Ga subunits rate limits the termination of the response and does not permit the rapid signaling cycles often demanded by the physiological processes (e.g., in neuronal communication and cardiac activity). By accelerating the Ga GTPase, RGS proteins speed up termination of the response and thereby increase the temporal fidelity of GPCR-initiated signaling. This function is best exemplified by studies on photoreceptors in which the loss of RGS protein in the visual cascade initiated by rhodopsin diminishes the temporal resolution of visual signals, preventing the detection of moving objects (Chen et al., 2000). Second, by deactivating G proteins and/or competing with the effector molecules, RGS proteins interfere with signal propagation, thus taming the extent of signaling (Hepler et al., 1997;Lambert et al., 2010) and allowing adjustment of the signaling volume, depending on the physiological needs. The loss of this RGS function is well noted to sensitize responses causing cellular overreactivity (Lamberts et al., 2013;Neubig, 2015;Xie et al., 2012). From this perspective, RGS proteins could be considered endogenous genetically encoded antagonists of GPCR signaling.
The results of our systematic profiling of RGS substrate preferences prompt reconsideration of the mechanisms involved in cellular signaling diversification. Despite their large numbers, GPCRs can only signal through the same limited number of G proteins that they can activate. Previous studies indicated that signaling diversity is in part dictated by a combination of G proteins activated by individual GPCRs (Inoue et al.  . The negative regulation of individual Ga by RGS proteins, if sufficiently selective, would greatly contribute to signaling diversification to allow much more refined signaling characteristics with cellular specificity depending on the available RGS and G proteins. Whereas recent large-scale efforts have provided tremendous system-level insights into the selectivity of G protein activation by GPCRs (Flock et al., 2015;Inoue et al., 2019;Masuho et al., 2015b), the information about the selectivity of RGS has been missing. We fill this gap by establishing Ga selectivity profiles for the entire family of RGS proteins. Based on this information, we propose that RGSs and GPCRs work in synergy to generate diverse cell-type-specific signaling.
Although the experiments presented in this study demonstrate the importance of the bar code residues on the Ga-interacting interface of RGS proteins in dictating Ga preferences, the sufficiency of this residue-residue contact network in dictating precise selectivity patterns across the entire RGS family remains to be tested. It appears quite likely that the secondary network of residues that make contact with the Ga-binding residues on the surface can further adjust and/or reinforce the stringency of Ga recognition. In support of this possibility, members of the R4 subfamily show more diverse functional properties than sequence similarity, suggesting contributions of additional residues within the RGS domain outside of the Ga-interacting surface in shaping Ga selectivity. This is consistent with the results of our ancestral reconstitution experiments, that shuffling wider group of the amino acid residues in the entire RGS domain can also modulate Ga selectivity. Furthermore, elements outside of the RGS domains may further contribute to the Ga recognition preferences of RGS proteins. Such a possibility is suggested by studies on complex multi-modular members of the R7 family, in which interaction partners (Gb 5 and R7BP) (Levay et al., 1999;Masuho et al., 2013) and domains (DEP, PGL) (Martemyanov et al., 2003;Skiba et al., 2001) have been shown to regulate Ga recognition. Many RGS genes also produce multiple splice isoforms that alter the structure of RGS proteins by adding or eliminating functionally important motifs without changing the RGS domain (Barker et al., 2001;Chatterjee et al., 2003;Granneman et al., 1998;Saitoh et al., 2002) and may further fine-tune Ga selectivity. Finally, several RGS proteins also interact with GPCRs, G protein effectors, and scaffold proteins (Abramow-Newerly et al., 2006), and this event may further alter Ga specificity. Although these possibilities were not addressed in this study, our experiments with shuffling determinants, mutagenesis, and ancestral reconstitutions all within the RGS domain indicate that these additional mechanisms may contribute to establishing the Ga selectivity but are unlikely to completely overwrite it.
Previous biochemical studies used purified recombinant proteins to examine the preferences of RGS proteins on Ga sub-strates selected ad hoc yielding important information that has served as a reference for RGS-Ga pairing. For example, RGS4 was shown to regulate both Ga i/o and Ga q subfamilies, but not Ga s or Ga 12 (Berman et al., 1996a;Berman et al., 1996b;Hepler et al., 1997). In contrast, RGS2 was found to have no appreciable GAP activity toward Ga i/o and to be selective for Ga q in both solution GTPase assays and pull-down experiments (Heximer et al., 1997;Kimple et al., 2009). R7 RGS family members were reported to be Ga o selective, with weaker GAP activity on Ga i (Hooks et al., 2003;Posner et al., 1999a;Snow et al., 1998). The selectivity of RGS7 for Ga o over Ga i was observed with the purified RGS domain (Lan et al., 2000), which is consistent with our conclusion that its RGS domain encodes a Ga selectivity bar code. Ga z selectivity of RZ subfamily members RGS17 (RGSZ2), RGS19 (GAIP), and RGS20 (RGSZ1) was also observed (Glick et al., 1998;Wang et al., 1998). Finally, an R12 RGS member, RGS10, has been shown to regulate Ga o , Ga i , and Ga z , but not Ga s (Hunt et al., 1996;Popov et al., 1997). Our investigation confirms many of the previously noted Ga preferences of RGS proteins, while additionally refining them to include G proteins not previously studied. However, in some cases, our results contradict previously documented coupling. One of the notable examples of this is Ga q selectivity of RGS2. Although our investigation shows that RGS2 can indeed regulate several members of the Ga q subfamily, we also find that it exhibits strong activity on the Ga i/o proteins comparable to that on Ga q . We think that the discrepancy is largely related to the choice of the assay system. Most of the previous studies used purified RGS and Ga proteins and measured GTP hydrolysis rates using biochemical assays conducted in solution. This approach has limited sensitivity and is devoid of the membrane environment where GPCRs, RGS, and G proteins normally operate under physiological context. In fact, the activity of RGS proteins has been shown to be significantly modulated by the membranes and lipid modification on Ga subunits (Tu et al., 1997). Furthermore, the proteoliposome-based assay was found to yield $100-fold higher sensitivity as compared to the solutionbased assay (Posner et al., 1999b). RGS2, in particular, was noted to act on Ga i/o in the presence of lipid bilayer (Ingi et al., 1998). Thus, the cellular BRET assay strategy that we chose provides physiologically relevant information on RGS-Ga coupling as it exploits the endogenous environment and appropriate context of RGS action.
One of the key insights provided by this work is the delineation of the determinants involved in RGS-Ga recognition. Establishing principles involved in the selectivity of protein-protein interaction has been a major goal of many investigations (Flock et al., 2017;Nooren and Thornton, 2003). Interaction between RGS and Ga provides an excellent model for interrogation of the underlying principles with possible general implications.

OPEN ACCESS
Both protein families are well represented by numerous members with clearly defined orthologs and paralogs, and conservation of the structural organization (Baltoumas et al., 2013;Tesmer, 2009). Thus, the experimental definition of the Ga preferences of all of the RGS proteins naturally prompted dissection of the underlying selectivity determinants. This study was focused on examining the contribution of the Ga-binding surface of the RGS domain. A combination of gene orthology/paralogy analysis with structural mapping identified a set of 17 variable amino acids that surround the core critical for forming direct contacts with the Ga subunits. We found that mutations in these amino acids significantly change the Ga preferences of RGS proteins. Interestingly, transplanting sets of variable amino acids from one RGS protein to another completely overwrites the Ga selectivity of the recipient. These observations support the idea that the selectivity of Ga recognition is, at least in part, encoded by the property of the amino acids that form this bar code region on the surface.
Previous studies explored the role of electrostatic interactions in specifying the selectivity of Ga recognition by several RGS proteins across all of the subfamilies Israeli et al., 2019;Kosloff et al., 2011;Salem-Mansour et al., 2018). Collectively, these studies reported 12 amino acid residues in RGS proteins that influence their ability to recognize Ga. Mutation of these residues either alone or in combination (up to 7 simultaneously) was shown to either increase or decrease the GAP activity of RGS proteins on the Ga substrates of choice. These studies examined one Ga substrate at a time, thus making it unclear whether the manipulations resulted in switching relative Ga preferences for a given RGS as opposed to overall gain or loss of substrate recognition. Nevertheless, these studies convincingly demonstrate that changes in electrostatic properties of amino acids at the RGS-Ga interface can alter the efficiency of the Ga recognition. Interestingly, all but two (H4.4 and H5.14) of these residues mapped on the Ga selectivity bar code region identified in this study, supporting the idea that electrostatic interactions play an important role in shaping the selectivity of RGS-Ga recognition. Similarly, mutations in RGS2 at the interface with the a-helical domain of Ga subunit diminished GAP activity on Ga q (Nance et al., 2013). In agreement with a large number of contacts made by the a-helical domain with the RGS domain, our analysis shows that variants mapping to this domain in several RGS proteins (H7.6, H7.9, H8.3) affect their Ga selectivity. Taken together with our observations that even single amino acid substitutions within the selectivity bar code can change the Ga preferences of RGS proteins, these results point to critical determinants of RGS-Ga recognition. Curiously, we found that altering these determinants can generate RGS proteins with novel selectivity profiles not displayed by canonical members of the family (e.g., RGS8/14-P, AncR4/Z/12/7; see Figures 4 and 6). Thus, we believe that the Ga-selectivity determinants identified here may pave the way for the de novo creation of RGS proteins with rationally designed G protein selectivity.
Our findings also have implications for pharmacogenomics and understanding disease mechanisms associated with the disruption in RGS-mediated G protein control. We uncovered a significant variation affecting nearly all of the RGS proteins.
More importantly, many of these variants occurred in selectivity bar code domains and were found experimentally to affect the Ga selectivity of RGS proteins. These genetic alterations are expected to change the profiles of signaling pathways engaged by the GPCRs, creating a situation that the same drug targeting the same receptor would produce varying effects due to RGS heterogeneity. Such a situation may be cryptic in the population if one only profiles variation within GPCRs (Hauser et al., 2018), but it may still lead to interindividual variability in drug response. Therefore, understanding the impact of RGS proteins and their genetic variability on GPCR signaling is expected to be important for individualizing drug prescriptions in the implementation of precision medicine.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Materials Availability
Plasmids generated in this study will be distributed upon request without restriction.

Data and Code Availability
The published article includes all datasets generated and analyzed during this study.

Mice
All experimental work involving mice was approved by The Scripps Research Institute's IACUC committee in accordance with NIH guidelines. Mice were housed under standard conditions in a pathogen-free facility on a 12:12 light:dark hour cycle with continuous access to food and water. Male and female CAMPER (Gt(ROSA)26Sor tm1(CAG-ECFP*/Rapgef3/Venus*)Kama and wild-type C57/Bl6 mice of both sexes aged from postnatal day 0 to postnatal day 3 were utilized in these studies and were not subjected to any prior experiments.
Live-imaging of cAMP and Ca 2+ dynamics Primary neuronal cultures were imaged under a Leica TCS SP8 confocal microscope through a 25x objective lens. Changes in cAMP were recorded from CAMPER neurons, as previously described (Doyle et al., 2019;Muntean et al., 2018). Briefly, excitation of mTurquoise FRET donor with a 442 nm diode laser was paired with simultaneous acquisition of XYZ image stacks at 10 s intervals collected through two HyD detectors tuned to 465-505 nm (mTurquoise FRET donor) and 525-600 nm (Venus FRET acceptor). Quantification of fluorescence intensity was performed on neuronal cell bodies using ImageJ (Schneider et al., 2012) to calculate FRET from the donor/acceptor ratio. The FRET ratio was converted to the concentration of cAMP using a dose-response curve to cAMP standards in permeabilized neurons. Segregated dopamine receptor subtype expression in striatal neurons enabled the identification of D2R-expressing neurons according to the directionality of cAMP response to dopamine. Dopamine was added in phasic puffs during continuous perfusion (2 mL/minute) of a pH 7.2 buffer consisting of 1.3 mM CaCl 2 , 0.5 mM MgCl 2 , 0.4 mM MgSO 4 , 0.4 mM KH 2 PO 4 , 4.2 mM NaHCO 3 , 138 mM NaCl, 0.3 mM Na 2 HPO 4 , 5.6 mM D-Glucose, and 20 mM HEPES. Changes in intracellular calcium concentration were recorded from wild-type neurons expressing jGCaMP7s. Excitation was performed with a 488 nm laser, and the acquisition of XYZ image stacks at 1 s intervals was collected through a HyD detector tuned to 494-593 nm. Quantification of fluorescence intensity was performed on neuronal cell bodies using ImageJ. Acetylcholine was added in phasic puffs during continuous perfusion (2 mL/minute) of a pH 7.3 buffer consisting of 2.2 mM CaCl 2 , 1 mM MgCl 2 , 138 mM NaCl, 11 mM D-Glucose, 10 mM HEPES, 50 mM picrotoxin, 300 nM CGP55845, and 10 mM DNQX.

Alignment of human RGS paralogs and orthologs
Whole protein sequences of human RGS proteins were downloaded from the UniProt database (https://www.uniprot.org/). The core RGS domain in each of these human RGS proteins was assigned based on HMMER searches conducted on pfam database domain profiles using human RGS proteins. Then the core RGS domains assigned in all of the human RGS paralogs were aligned using MSAProbs (Liu et al., 2010) and this alignment was termed as human RGS domain alignment (HRDA). Animal orthologs of RGS proteins were obtained from the OMA database (https://omabrowser.org/oma/home/) (Altenhoff et al., 2018) and equivalent regions to the core RGS domain of human RGS were only considered for further investigations. We aligned the core RGS domain regions in the animal orthologs with human ones. For each human RGS, i.e., RGS1 to RGS21, we constructed multiple sequence alignments of the given RGS with its corresponding animal orthologs.

RGS common numbering scheme
We developed a common RGS numbering scheme (CRN), by integrating consensus secondary structure information of available crystal structures of the RGS domain on to HRDA sequence alignment. This allowed us to uniquely assign an alignment position to a combination three types Orthology/paralogy analysis To identify the ortholog specific conserved residues and commonly conserved residues between paralogs of human RGS in the core RGS domain. We developed a strategy, by comparing assigning the CRN to each of the RGS alignments and we then categorized the residue at a given CRN position is: (a) Ortholog-specifically conserved if the normalized BLOSUM score for this CRN is 1.5 times higher in a given RGS alignment than in the equivalent CRN of HRDA alignment position and the given CRN position also displays above average normalized BLOSUM score within the RGS alignment. (b) Paralog-specifically conserved if the normalized BLOSUM score for this CRN in the HRDA alignment is 1.5 times higher than in the equivalent CRN of RGS alignment and the given position displays above average normalized BLOSUM score within the HRDA alignment. (c) Conserved in both if CRN in RGS alignment and the HRDA display comparable normalized BLOSUM scores, i.e., within 1.5 times normalized BLOSUM score of either of them. The given position displays above average normalized BLOSUM score within the HRDA and RGS alignments. (d) Neutrally evolving if the above three conditions were not met. The alignment of RGS domain from orthologs is provided as Data S1 and S2. In the datasets, the residue numbers following the accession OMA database ID and UniProt ID or Ensembl database ID are presented.

Reconstitution of recombinant ancestral RGS proteins
The reconstitution of ancestral RGS proteins based on the computational algorithm using FastML was performed (Ashkenazy et al., 2012) on different groups of RGS alignments i.e., for e.g., R4, RZ, R12, and all RGS proteins. Ancestral reconstruction methods identify most likely sequences, including indels, in a specific ancestral node in a phylogenetic tree for given multiple sequence alignment.    Increasing amount of GPCR cDNA for transient transfection increased G protein activation rates but did not alter G protein deactivation rates. Effects of increasing active G proteins on deactivation rates of G proteins (right). Increasing concentration of agonist produced more active G protein but maintain consistent G protein deactivation rates. (D) Effects of increasing RGS on G protein deactivation rates. Increasing amount of RGS cDNA for transient transfection increased deactivation rates.   Article of the alignment. The gray indicates the residues with conserved property and black indicate the conserved residues. Of note, there are two insertion/deletion regions in this alignment of the RGS domain. First, there are four amino acid residues in loop 5 in the most of RGS proteins. Instead, there are six amino acids in RGS12 and RGS14, but only three amino acid residues in all R7 RGS members in this structural element. Second, all three R12 RGS proteins are missing an amino acid residue in the H6 region. It is not possible based on existing structural alignments to say where this gap actually occurs, because the H6 region is conformationally heterogeneous in R12 structures and cannot be structurally aligned with other RGS proteins other than to say it has helical character as detected by NMR. The disorder of this region in R12 subfamily members has in fact been proposed to play a role in selecting against the Ga q family due to loss of beneficial interactions with SwIII (Taylor et al., 2016) The conserved and selectivity residues identified by ortholog/paralog analysis ( Figure 5C) are highlighted in blue and orange, respectively. The sequence alignments were generated with T-Coffee (http://tcoffee.crg.cat/apps/tcoffee/do:regular) and colored by BoxShade ( (A) Sequence pattern of the RGS13, RGS18 and RGS13/18 chimera are shown. Identical amino acid residues between RGS13 and RGS18 were indicated by asterisks. (B) Sequence pattern of the R4 and R12 subfamilies, their representative RGS proteins (RGS8 and RGS14), and mutant RGS proteins are shown. Identical amino acid residues between RGS8 and RGS14 were indicated by asterisks. (C) and (D) The Ga-selectivity fingerprints (k GAP (C) and k dis (D)) of RGS8, RGS14, and two mutants are shown. The thickness of the lines connecting each data point represents the SEM of three independent experiments. The relative values are plotted on a linear scale.