Structure of the Regulator of G Protein Signaling 8 (RGS8)-Gαq Complex

Regulator of G protein signaling (RGS) proteins interact with activated Gα subunits via their RGS domains and accelerate the hydrolysis of GTP. Although the R4 subfamily of RGS proteins generally accepts both Gαi/o and Gαq/11 subunits as substrates, the R7 and R12 subfamilies select against Gαq/11. In contrast, only one RGS protein, RGS2, is known to be selective for Gαq/11. The molecular basis for this selectivity is not clear. Previously, the crystal structure of RGS2 in complex with Gαq revealed a non-canonical interaction that could be due to interfacial differences imposed by RGS2, the Gα subunit, or both. To resolve this ambiguity, the 2.6 Å crystal structure of RGS8, an R4 subfamily member, was determined in complex with Gαq. RGS8 adopts the same pose on Gαq as it does when bound to Gαi3, indicating that the non-canonical interaction of RGS2 with Gαq is due to unique features of RGS2. Based on the RGS8-Gαq structure, residues in RGS8 that contact a unique α-helical domain loop of Gαq were converted to those typically found in R12 subfamily members, and the reverse substitutions were introduced into RGS10, an R12 subfamily member. Although these substitutions perturbed their ability to stimulate GTP hydrolysis, they did not reverse selectivity. Instead, selectivity for Gαq seems more likely determined by whether strong contacts can be maintained between α6 of the RGS domain and Switch III of Gαq, regions of high sequence and conformational diversity in both protein families.

discrepancy helped lead to the discovery of a family of GTPaseactivating proteins (GAPs), 2 now known as regulator of G protein signaling (RGS) proteins (4 -6). RGS proteins contain a conserved helical domain called the RGS domain that directly binds to the three switch regions (SwI-III) of the G␣ subunit and stabilizes them in a transition state conformation (7).
RGS domains are divided into four subfamilies based on sequence homology and substrate preference: RZ, R4, R7, and R12 (8). All utilize G␣ i/o subunits as substrates, although some RZ members seem selective for G␣ z subunits (9). A recent study using surface plasmon resonance indicated that the RGS domains that belong to the R7 and R12 subfamilies bind weakly or not at all to G␣ q , whereas the RZ and R4 subfamilies tend to interact with both G␣ i/o and G␣ q/11 (10). The exception is RGS2, an R4 subfamily member that is uniquely selective for G␣ q/11 (11). The underlying molecular mechanism dictating RGS selectivity has not been fully answered, and past research has focused mainly on RGS2 in part because of its strong link to hypertension and cardiac hypertrophy via its regulation of G␣ q signaling in vivo (12)(13)(14)(15)(16). RGS2 can be altered to enhance its selectivity for G␣ i/o by making mutations at three positions unique to RGS2 (Cys 106 , Asn 184 , and Glu 191 ) that interact with the G protein, primarily near the SwI region (11). These residues are conserved as Ser, Asp, and Lys, respectively, in other RGS proteins. The crystal structure of an RGS2 mutant with conversion of these three residues to their equivalents in other RGS domains (RGS2 SDK ) in complex with G␣ i confirmed that the mutant binds in the same canonical orientation as observed in other RGS-G␣ i/o complexes (17).
Subsequently, the structure of wild-type RGS2 in complex with G␣ q revealed a distinct binding mode (18). Because this was also the first RGS domain-G␣ q complex to be structurally characterized, it was unclear whether the significant tilt in the orientation of the RGS domain with respect to G␣ was due to sequence differences in either RGS2 or G␣ q . Therefore, before one can address the molecular basis for why some RGS proteins select against G␣ q/11 subfamily members, structures of conventional R4 subfamily members in complex with G␣ q need to be determined.
Herein, we show that RGS8, an R4 subfamily member, binds to G␣ q in a manner similar to how other RGS proteins bind * This work was supported in part by National Institutes of Health Grants HL086865 and HL122416 (to J. J. G. T.) and by National Institutes of Health Training Grant T32GM008270 (to V. G. T.) The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the author and does not necessarily represent the official views of the National Institutes of Health. G␣ i/o , indicating that the distinct pose of RGS2 is driven by its unique switch binding residues. We investigate unique contacts formed between RGS8 and the G␣ ␣-helical domain and demonstrate that although residues in the ␣-helical domain modulate GAP activity, the chief determinant of selectivity is found elsewhere, most likely SwIII. These results clarify the molecular determinants of RGS domain selectivity and, by extension, how RGS proteins impact signaling pathways in vivo.

Experimental Procedures
Protein Expression and Purification-All constructs encoding RGS variants were confirmed by sequencing on both strands. Residues 42-173 spanning the RGS domain of RGS8 were expressed using the pQTEV vector (a kind gift from Dr. R. Neubig, Michigan State University). Residues 22-157 spanning the RGS domain of RGS10 were expressed using a pLIC-SGC1 vector obtained from the Structural Genomics Consortium (10). After cleaving the N-terminal His 6 tag with tobacco etch virus protease from RGS8 and RGS10, the exogenous sequence QSM is left on the N terminus.
For RGS8 variants, 1 liter of BL21 Rosetta cells grown in Terrific Broth was induced with 100 M isopropyl-1-thio-␤-Dgalactopyranoside at 20°C. The cells were pelleted by centrifugation at 3500 ϫ g for 15 min and then resuspended in Buffer A (20 mM HEPES, pH 8.0, 500 mM NaCl, and 10 mM ␤-mercaptoethanol), supplemented with 7.6 M leupeptin, 360 nM lima bean trypsin inhibitor, 1 mM PMSF, and 0.1 mM EDTA before being homogenized with a Dounce. Cells were then lysed using an EmulsiFlex-C3 homogenizer (Avestin). Cell debris was pelleted by centrifugation at 40,000 rpm (185,500 g) in a Type 45 Ti fixed angle rotor (Beckman-Coulter). The supernatant was passed over a 5-ml Ni-NTA affinity column pre-equilibrated with Buffer A. The column was then washed with 100 ml of Buffer A, followed by 100 ml of Buffer A with 10 mM imidazole, pH 8.0. RGS8 was eluted from the affinity column using 25 ml of Buffer A with 150 mM imidazole, pH 8.0, and then dialyzed into 20 mM sodium acetate, pH 5.5, 0.5 M NaCl, and 10 mM ␤-mercaptoethanol. The salt concentration was reduced by diluting 4-fold into Buffer B (20 mM sodium acetate, pH 5.5, and 2 mM DTT), and the protein was loaded onto an UnoS ion-exchange column (Bio-Rad) and eluted using Buffer B in an NaCl gradient increasing from 125 mM to 1 M. Fractions absorbing at 280 nm were verified using SDS-PAGE and pooled and concentrated to 4 -9.5 mg/ml. For cleavage of the His tag, 2% (w/w) tobacco etch virus protease was added during a final dialysis into Buffer A.
RGS10 variants were expressed and purified similarly to RGS8 until after elution from the Ni-NTA affinity column. Cleavage of the His 6 tag was then performed as described above for RGS8, followed by passage over a second Ni-NTA affinity column to remove the cleaved tag and His 6 -tagged protease. The flow-through was collected, concentrated to ϳ7.5 mg/ml, and then buffer-exchanged on tandem Superdex 200 10/300 GL (GE Life Sciences) gel filtration columns into Buffer A with 5 mM DTT instead of ␤-mercaptoethanol. Fractions absorbing at 280 nm were verified using SDS-PAGE and then pooled and concentrated to 7.5-9 mg/ml. The insect cell vector pFastBacHT expressing an N-terminally truncated variant of Mus musculus G␣ q spanning residues 35-359 (⌬N-G␣ q ) was described previously (18). For purification, 6 liters of High Five TM cells (BTI-TN-5B1-4) expressing ⌬N-G␣ q were pelleted at 3000 ϫ g for 20 min. The pellet was then resuspended in Buffer A (20 mM HEPES, pH 8.0, 100 mM NaCl, 10 mM ␤-mercaptoethanol, and 10 M GDP, pH 8.0), 7.6 M leupeptin, 360 nM lima bean trypsin inhibitor, 1 mM PMSF, 0.1 mM EDTA, and 3 mM MgCl 2 . Cells were then homogenized, lysed, and pelleted as described for RGS8. The supernatant was then passed through a Ni-NTA agarose affinity column pre-equilibrated with Buffer A supplemented with 1 mM MgCl 2 . The column was washed with 100 ml of Buffer A plus 1 mM MgCl 2 , followed by 100 ml of Buffer A plus 1 mM MgCl 2 and 10 mM imidazole, pH 8.0, and then eluted with 25 ml of Elution Buffer (Buffer A with 1 mM MgCl 2 and 150 mM imidazole, pH 8.0). G␣ i/q -R183C was produced as described previously (19).
The Escherichia coli vector pQE60 expressing a C-terminal, His 6 -tagged G␣ i1 spanning residues 1-354 was provided courtesy of Dr. Barry Kreutz (University of Illinois at Chicago). Expression was carried out as described previously (20). Purification was similar to RGS8, with the following exceptions. The Lysis Buffer was 50 mM HEPES, pH 8.0, 1 mM EDTA, 2 mM DTT, 0.1 mM PMSF, 7.6 M leupeptin, and 360 nM lima bean trypsin inhibitor. Buffer A was 50 mM HEPES, pH 8.0, and 2 mM DTT. After washing the Ni-NTA affinity column with Buffer A, the elution step was performed using Elution Buffer (50 mM HEPES, pH 8.0, 2 mM DTT, and 150 mM imidazole, pH 8.0). The eluate was then loaded onto a UnoQ anion exchange chromatography column (Bio-Rad) pre-equilibrated with Buffer A and eluted using a gradient of 0 -250 mM NaCl in Buffer A. The integrity of G␣ i1 was confirmed by visualizing trypsin digests on SDS-PAGE as described previously (21).
Purification of the RGS8-G␣ q Complex-Purified ⌬N-G␣ q was incubated with 30 M AlCl 3 , 10 mM NaF, and 1 mM MgCl 2 in a buffer also containing 10 M GDP, pH 8.0, 20 mM HEPES, pH 8, 100 mM NaCl, and 2 mM DTT. It was then mixed with purified RGS8 in a 1:1 molar ratio based on the RGS8 concentration determined using a NanoDrop TM ND-1000 spectrophotometer, and the ⌬N-G␣ q concentration was determined using Bradford reagent. The proteins were incubated together for 30 min on ice before loading onto tandem Superdex 200 10/300 GL (GE Life Sciences) gel filtration columns equilibrated with 20 mM HEPES, pH 8.0, 100 mM NaCl, 2 mM DTT, 10 M GDP, pH 8.0, and 1 mM MgCl 2 . Fractions shown to contain 1:1 complex by SDS-PAGE were then concentrated to 5-7 mg/ml.
Crystallization and Cryoprotection-Crystals were grown in VDX plates (Hampton Research) using hanging drop vapor diffusion on glass cover slides. The RGS8-G␣ q complex (6.6 mg/ml) was mixed 1:1 with well solution to a final volume of 1 l and suspended over 1 ml of well solution. Octahedral crystals grew in 2 weeks at 4°C using a well solution containing 0.1 M ammonium acetate, 0.1 M Bis-Tris, pH 5.5, and 11% PEG 8000. Crystals were harvested by adding several l of cryoprotectant (20 mM HEPES, pH 8.0, 100 mM Bis-Tris, pH 5.4, 200 mM ammonium acetate, 15% PEG 8000, 200 mM NaCl, 1 mM DTT, 50 M GDP, pH 8.0, 20 M AlCl 3 , 10 mM NaF, and 5 mM MgCl 2 ) in 0.5-l increments to the drop containing the crystal. The crystal was then transferred into 100% cryoprotectant and moved stepwise through mixtures of cryoprotectant plus glycerol until a final glycerol concentration of 24% (v/v). The crystal was then suspended in a nylon loop and frozen in liquid nitrogen.
Data Collection, Processing, and Model Building-X-ray diffraction data were collected at the Life Sciences Collaborative Access Team (LS-CAT) beamline 21-ID-D at the Advanced Photon Source (APS). Reflection intensities were integrated and scaled using HKL2000, and initial phases were determined by molecular replacement using PHASER and structures of ⌬N-G␣ q from Protein Data Bank (PDB) entry 4EKD and RGS8 from PDB entry 2ODE as search models. Manual model building in Coot was alternated with TLS refinement with local non-crystallographic symmetry restraints in REFMAC5. Coordinates and structure factors were deposited with the Protein Data Bank as entry 5DO9. Figures were generated using The PyMOL Molecular Graphics System, Version 1.5.0.4 (Schrödinger, LLC).
GAP Assays-3 and 4 mg/ml stocks of G␣ i/q -R183C and G␣ i1 , respectively, were incubated for 10 min with 10 mM EDTA, and then diluted to 0.3 M final concentration in 300 l of Incubation Buffer (50 mM HEPES, pH 8.0, 1 mM DTT, 1 mM EDTA, 100 g/ml albumin, 5.5 mM CHAPS, 5% glycerol, and 37.5 M ammonium sulfate) plus 33.3 Ci/ml [␥-32 P]GTP (PerkinElmer, EasyTide) and enough cold GTP, pH 8.0, to make the total GTP concentration 6.25 M. The reaction was then incubated at room temperature for 3 h (G␣ i/q -R183C) or 30 min (G␣ i1 ). Samples were buffer-exchanged into fresh Incubation Buffer using a pre-equilibrated Micro Bio-Spin TM chromatography column (Bio-Rad) and stored on ice for the duration of the assay. Each assay was initiated by adding 20 l of the bufferexchanged G␣ subunit to a tube containing 100 nM RGS protein in 180 l of Assay Buffer (20 mM HEPES, pH 8.0, 80 mM NaCl, 1 mM DTT, 1 mM EDTA, 0.9 mM MgCl 2 , 1 mM cold GTP, pH 8.0, 10 g/ml albumin, and 0.20% w/v cholate) on ice. 40-l aliquots of the reaction were quenched at various time points by vortexing with 750 l of ice-cold Quenching Buffer (10 mM sodium phosphate, pH 2.0, and 5% (w/v) activated charcoal). The quenched reaction was spun for 25 min at 6500 ϫ g at 4°C. Afterward, 200 l of the supernatant was added to 3 ml of MicroScint TM 40 scintillation mixture (Perkin-Elmer) and read on a liquid scintillation counter instrument measuring 32 P cpm. Each RGS variant was tested in triplicate in three separate experiments. Data were processed in Prism 6 using a non-linear exponential fit with a time lag for G␣ i1 or a straight line fit for G␣ i/q -R183C.
Sequence Alignment and Structure Comparisons-Human RGS sequences from UniProt were aligned using Clustal Omega at the European Molecular Biology Laboratory-European Bioinformatics Institute (22). RMSD calculations were performed using Superpose from the CCP4 software suite (23,24). Calculation of buried surface area for complexes was performed using PISA (25).

Results
Crystal Structure of the RGS8-G␣ q Complex-To determine whether the altered pose of RGS2 on G␣ q was due to the unique switch interface residues Cys 106 , Asn 184 , and Glu 191 of RGS2 or to unique residues in the switch regions of G␣ q relative to G␣ i , the crystal structure of the RGS domain of RGS8, an R4 subfamily member selective for both G␣ q and G␣ i , was determined in complex with N-terminally truncated G␣ q (⌬N-G q ). RGS8 was used because it readily crystallizes, and the structure of its complex with G␣ i3 was previously reported (10). The final structure was refined to 2.6 Å spacings (Table 1, Fig. 1A). Residues 42-173 of RGS8 with two N-terminal exogenous residues are visible in all complexes, as well as residues 38 -350 of ⌬N-G␣ q . Three complexes of RGS8-G␣ q exist in the asymmetric unit, with their overall RMSD in C␣ positions varying by less than 0.6 Å. Comparisons with the RGS8-G␣ i3 and RGS2 SDK -G␣ i3 complexes give RMSD values of 0.9 Å for 432 and 426 C␣ atoms, respectively, whereas comparison with the RGS2-G␣ q complex gives an RMSD of 1.2 Å for 439 C␣ atoms. This indicates that RGS8 binds G␣ q in a manner most similar to how RGS proteins have previously been shown to bind G␣ i/o subfamily members (Fig. 1B). Thus, the unique substitutions ( Fig.  2A) in the G␣-binding interface of RGS2 are primarily responsible for its altered pose when bound to G␣ q .
Molecular Basis for RGS Subfamily Selectivity-Next, the structures of the RGS8-G␣ i and RGS8-G␣ q complexes were compared to identify RGS domain contacts with G␣ that are distinct between the G␣ i/o and G␣ q/11 subfamilies. The structural element that differs most is the ␣B-␣C loop in the ␣-helical domain. In the RGS8-G␣ q complex, the ␣B-␣C loop is less ordered when compared with the RGS8-G␣ i complex based on temperature factors, but extends closer to the RGS protein, which creates additional buried surface area (Fig. 1A). In fact, RGS2 seems to exploit this surface to maintain greater buried accessible surface area with G␣ q (2050 Å 2 ) than does RGS8 (1900 Å 2 ), or than does RGS8 in complex with G␣ i (1650 Å 2 ). RGS residues that would contact this loop exhibit sequence heterogeneity among the various RGS subfamilies (Fig. 2A).
The R4 family has a conserved Glu-Lys dyad in the ␣7 helix (RGS8 residues 155-156), whereas RGS10, an R12 member, has Lys-Tyr (residues 131-132) ( Figs. 2A and 3, A and B). Superpo-sition of the G␣ subunits in the RGS8-G␣ q (Fig. 3A) and RGS10-G␣ i (Fig. 3B) complexes suggests that charge repulsion and/or steric hindrance by this dyad could discourage binding of R12 family members to G␣ q , as there is a charge reversal in the first position and introduction of a bulkier Tyr residue for Lys in the second (Fig. 3C). In comparison, modeling Glu-Lys for the Lys-Tyr dyad of RGS10 anticipates no overt issues with G␣ i binding (Fig. 3D). R7 subfamily members instead have a Lys-(Lys/Ser) dyad ( Fig. 2A). The subfamily-specific sequences of these dyads could therefore contribute to G␣ selectivity. In support of this hypothesis, a previous study found that mutating these positions contributes to differences in GAP activity of various R4 family members on G␣ o (26). Functional Analysis of the ␣-Helical Domain Interface-The aforementioned ␣-helical domain interface was tested by sitedirected mutagenesis followed by single turnover GTPase assays using [␥-32 P]GTP ( Table 2, Fig. 4). RGS8-Glu 155 and/or Lys 156 were converted to their analogous residues in RGS10 (Lys and Tyr, respectively). Complementary mutations were introduced in RGS10, mutating Lys 131 and/or Tyr 132 to Glu and Lys, respectively. If RGS selectivity for G␣ subunits was achieved via ionic repulsion with the ␣-helical domain of G␣ q , then mutation at the first position (E155K in RGS8, K131E in RGS10) would result in a selectivity switch. Selectivity achieved through steric pressure would be potentially altered by mutation at the second position (K156Y in RGS8, Y132K in RGS10). If both sterics and charge were necessary to affect a selectivity switch, then both point mutations (E155K/K156Y in RGS8 and K131E/Y132K in RGS10) would be required.
Wild-type G␣ i1 and the slow-hydrolyzing mutant G␣ i/q -R183C were used as substrates for each RGS variant. As expected, wild-type RGS8 showed robust GAP activity on both G␣ i1 (Fig. 4A) and G␣ i/q -R183C (Fig. 4B), whereas wild-type RGS10 only showed GAP activity on G␣ i1 (Fig. 4C, Table 2). All three mutants of RGS8 retained their activity on G␣ i1 , but also retained wild-type, if not higher, activity on G␣ i/q -R183C. The RGS10 double mutant and K131E single mutant also retained activity on G␣ i3 . Interestingly, the Y132K mutant did not. None of the RGS10 mutants showed G〈P activity on G␣ i/q -R183C. These results indicate that the RGS domain dyad that contacts the ␣B-␣C loop is not responsible for RGS10 being inactive on G␣ q , as RGS10 and RGS8 mutants had no increase or loss, respectively, in selectivity for G␣ i/q -R183C.

Discussion
RGS proteins range from being relatively small proteins with little more than the RGS domain, to complex multi-domain entities with multiple signaling domains. However, even in simple RGS proteins such as RGS2, RGS4, and RGS8, the regions outside the RGS domain can play important roles such as targeting these enzymes to membranes, G protein-coupled receptors, or effector enzymes (27)(28)(29)(30). Thus, when one considers the selectivity of an RGS protein for a particular G␣ signaling pathway, there are many levels at which this can occur. However, the most fundamental aspect of selectivity is imposed by the direct interaction of the RGS domain with G␣ to promote acceleration of GTP hydrolysis. Consequently, this study focused solely on the interaction of the RGS domain found in FIGURE 1. Structure of RGS8 in complex with G␣ q reveals a canonical tilt. A, the 2.6 Å crystal structure of the RGS8 GAP domain in complex with ⌬N-G␣ q . The ␣B-␣C loop exhibits structural differences between G␣ i and G␣ q that could dictate the selectivity of RGS proteins. RGS8 is cyan, G␣ q is yellow, the three switch regions are red, GDP is black, AlF 4 Ϫ is green, Mg 2ϩ is orange, and the ␣B-␣C loop of G␣ i is shown in pink (PDB code 2ODE). B, RGS2 adopts a unique tilt when bound to G␣ q . The G␣ subunits of the RGS2-G␣ q complex (PDB code 4EKD), RGS8-G␣ i complex (PDB code 2ODE), and RGS8-G␣ q complex (PDB code 5DO9) were superimposed to compare the position of the RGS domain in each complex. G␣ q is shown in yellow, G␣ q -bound RGS8 is shown in pale cyan, G␣ ibound RGS8 is shown in orange, and RGS2 is shown in dark blue.
RGS proteins with G␣ i and G␣ q subunits. Moreover, previous studies have shown that isolated RGS domains exhibit selectivity for individual G␣ subfamilies (10).
Previous structural analysis of the RGS2 complex with G␣ q suggested that RGS2 has a distinct tilt relative to the G␣ subunit (Fig. 1B) that allows it to bury more surface area with G␣ q than it could with G␣ i . Moreover, the conformationally flexible ␣6 helix of RGS2 allows it to maintain optimal contacts with SwIII, despite the unique pose of the RGS domain (18). When the Cys 106 , Asn 184 , and Glu 191 interface residues are mutated to their equivalents in other RGS proteins in RGS2 SDK , it can bind to G␣ i in a canonical fashion (17), but does not lose activity against G␣ q , suggesting that interactions with SwI are not directly responsible for G␣ q selectivity (11,17). It was further demonstrated that altering interactions between the ␣7 helix of the RGS2 domain and the ␣-helical domain of G␣ can dramatically promote or inhibit GAP activity (18), but the molecular basis for selectivity against G␣ q observed in other RGS subfamilies remained unclear.
In this work, it was shown that an R4 family member, RGS8, binds to G␣ q in a canonical fashion, permitting a more precise comparison of the interactions between RGS proteins and these two G␣ subfamilies. The tilt of RGS2 in complex with G␣ q can thus be attributed to interfacial differences dictated by unique interfacial residues in RGS2. Two regions, in particular the ␣B-␣C loop of the ␣-helical domain, stand out as a potential selectivity determinant. The G protein ␣-helical domain has previously been shown in some instances to be a major determinant of GAP activity, and there are sequence signatures unique to each RGS subfamily that interact with this domain (18,31). Although a selectivity switch was not achieved in our study, the GAP assay results are consistent with RGS activity being enhanced or inhibited by interactions with the ␣-helical domain. Interestingly, the RGS10 Y132K mutant, creating a Lys-Lys dyad in ␣7, did not retain activity for G␣ i1 , but could be rescued by the addition of the K131E mutation. The disadvantage of having a Lys-Lys dyad may be due to electrostatic repulsion between the adjacent positions or with the ␣-helical domain. However, it seems clear that the ␣-helical domain is not a major G␣ selectivity determinant because no substitution in this interface could promote activity on G␣ i/q -R183C by RGS10 (Table 2).
Instead, the evidence now points toward SwIII, which interacts with the N-terminal end of the RGS ␣6 helix, as being the

RGS10 RGS8
Gα i Gα q α6 B FIGURE 2. Sequence conservation of RGS8 residues in ␣6 and ␣7 suggests selectivity mechanisms. A, sequence alignment of the ␣6 -9 regions of the R4, R7, and R12 subfamilies. Alignments were performed with Clustal Omega using human sequences. Residue positions important in G␣ i interactions are in dark gray, discussed SwIII interacting residues are in purple, and the dyad that interacts with the ␣-helical domain is in green. B, SwIII interface for the RGS8-G␣ q complex (PDB code 5DO9). RGS8 is in cyan, and G␣ q is in yellow. C, SwIII interface for the RGS10-G␣ i complex (PDB code 2IHB). RGS10 is in blue, and G␣ i is in pink. The disordered ␣6 region of RGS10 is depicted as a dashed line.
primary determinant of selectivity, as suggested in Ref. 10. In SwIII, the side chain of G␣ q -Asp 243 stacks with the side chain of RGS8-Phe 125 . The analogous residue in G␣ i , Glu 238 , cannot make this interaction because the backbone of its SwIII is positioned differently. Phe is shared by several other R4 family members at this position, but not by any R7 or R12 members ( Fig. 2A). R4 RGS domains typically also have a basic residue, e.g. RGS8-Arg 128 , that is positioned to form a hydrogen bond with a backbone carbonyl of another SwIII residue in G␣ q , Glu 241 . Members of the R7 and R12 subfamilies typically lack this basic residue. Notably, although the RGS region that comes in contact with SwIII is well ordered in complexes involving R4 family members (Fig. 2, B and C), the RGS10 ␣6 helix is disordered in its complex with G␣ i , and thus contacts with SwIII are nearly entirely lost (Fig. 2C). It is therefore quite possible that loss of pro-ductive interactions with SwIII mediated by the ␣6 region are responsible for the inability of some RGS subfamilies to recognize G␣ q . However, due to the poor sequence conservation and conformational heterogeneity of this region in R12 family members (R12 subfamily members also have a 1-residue deletion in the ␣6 helical region), it is not possible to easily test this hypothesis because conversion of SwIII contacts in RGS8 to those found in RGS10, and vice versa, is not possible by simple substitution.
Regardless, this model does not explain how RGS proteins in the R7 and R12 subfamilies retain activity for G␣ i if they fail to make productive interactions with SwIII. These subfamilies may have optimized interactions in other contact regions (e.g. ␣A and other regions in the G␣ ␣-helical domain), as has been shown for some R4 family members in determining their relative activity on members of the G␣ i subfamily (26). This    . Single turnover GAP assays of RGS variants reveal modulation by contacts with the ␣-helical domain. A, representative data from three experiments performed in triplicate of RGS8 variants using G␣ i as a substrate. Points were fit with a non-linear regression using one-phase association with a time lag. B, representative data from RGS8 variants using G␣ q as a substrate. Points were fit with a single steady state rate. C, representative data from RGS10 variants using G␣ i as a substrate. Error bars correspond to standard deviations. GTPase activity of the G␣ subunit alone is indicated in pink, wild-type RGS is indicated in blue, RGS8 E155K or RGS10 K131E is indicated in green, RGS8 K156Y or RGS10 Y132K is indicated in orange, and RGS8 E155K/K156Y or RGS10 K131E/Y132K is indicated in red. hypothesis is supported by the increased GAP activity observed for RGS8 variants that have RGS10 substitutions in ␣7, and decreased GAP activity for RGS10 variants with RGS8 substitutions (Table 2). This result is consistent with prior studies that have likewise probed positions in ␣7 and shown them to modulate RGS domain interactions and GAP activity (18,26,32). Another possible explanation might be found in the SwI interface. G␣ q has Pro 185 , whereas G␣ i has Lys 180 . The G␣ i -Lys 180 side chain buries more surface area with the RGS protein when compared with Pro 185 in G␣ q . Hence, if the interactions with SwIII are not strong, RGS proteins may be less active against G␣ q as a result of less buried surface area with SwI. Indeed the specific activity of RGS4 is ϳ10 times lower than wild type when using the G␣ i -K180P variant as a substrate (33). Inversely, AlF 4 Ϫ -activated G␣ q -P185K can be pulled down by RGS2 in significantly greater amounts than wild-type G␣ q (32).
In summary, the rules that dictate RGS domain selectivity for a given G␣ subunit are complex. They involve leveraging beneficial versus negative interactions at different points of contact with the G␣ subunits, as well as the ability of individual RGS proteins to undergo induced fit when required (18). However, the structure-function analysis reported here still points to the SwI and SwIII interactions as being the key determinants of selectivity for G␣ i versus G␣ q . Interactions with the ␣-helical domain can tune GAP activity (such as for RGS9 in the G␣ i/o subfamily) or can even rescue RGS2 from loss of activity when it binds to G␣ q (18). How is selectivity achieved for or against other G␣ subfamilies? G␣ s is not a substrate for any known RGS protein due to presence of Asp 229 in SwII (Ser in G␣ i and G␣ q ) (34). G␣ z , a G␣ i/o subfamily member, is the preferred substrate for RZ subfamily member RGS20, but the mechanism for this selectivity is unknown (35). A full understanding of the intricacies of how RGS proteins interact with G␣ subunits is required if one seeks to design an RGS domain specific to a particular G␣ subunit, or, conversely, a G␣ subunit that is a specific substrate for a select subgroup of RGS proteins. These would serve as useful tools to decipher the roles of individual RGS proteins in cellular signaling.
Author Contributions-V. G. T. and J. J. G. T. designed the experiments, determined the x-ray structure, and prepared the manuscript. V. G. T. and P. A. B. purified the complex. P. A. B. crystallized the complex. V. G. T. purified proteins for, and performed and analyzed the GAP assays. All authors approved the final paper.