High-resolution structure of RGS17 suggests a role for Ca2+ in promoting the GTPase-activating protein activity by RZ subfamily members

Regulator of G protein signaling (RGS) proteins are negative regulators of G protein–coupled receptor (GPCR) signaling through their ability to act as GTPase-activating proteins (GAPs) for activated Gα subunits. Members of the RZ subfamily of RGS proteins bind to activated Gαo, Gαz, and Gαi1–3 proteins in the nervous system and thereby inhibit downstream pathways, including those involved in Ca2+-dependent signaling. In contrast to other RGS proteins, little is known about RZ subfamily structure and regulation. Herein, we present the 1.5-Å crystal structure of RGS17, the most complete and highest-resolution structure of an RZ subfamily member to date. RGS17 cocrystallized with Ca2+ bound to conserved positions on the predicted Gα-binding surface of the protein. Using NMR chemical shift perturbations, we confirmed that Ca2+ binds in solution to the same site. Furthermore, RGS17 had greater than 55-fold higher affinity for Ca2+ than for Mg2+. Finally, we found that Ca2+ promotes interactions between RGS17 and activated Gα and decreases the Km for GTP hydrolysis, potentially by altering the binding mechanism between these proteins. Taken together, these findings suggest that Ca2+ positively regulates RGS17, which may represent a general mechanism by which increased Ca2+ concentration promotes the GAP activity of the RZ subfamily, leading to RZ-mediated inhibition of Ca2+ signaling.

G protein-coupled receptors (GPCRs) 3 regulate many physiological processes in response to the binding of an extracellular ligand, leading to the activation of diverse pathways, including vision and hormonal signaling. The intracellular response to ligand binding is mediated by heterotrimeric G proteins, which consist of G␣ and G␤␥ subunits. In the inactive state, G␣ is bound to GDP and stably associated with G␤␥. The activated GPCR is a guanine nucleotide exchange factor for G␣, catalyzing the exchange of GDP for GTP. G␣⅐GTP and G␤␥ dissociate from one another and bind downstream effector enzymes to stimulate second messenger production (1). G␣ subunits are deactivated upon GTP hydrolysis and reassociate with G␤␥, terminating downstream signaling. However, the intrinsic rate of GTP hydrolysis for many G␣ subunits is too slow to be physiologically relevant. This discrepancy led to the discovery of the regulator of G protein signaling (RGS) proteins, which are GTPase-activating proteins (GAPs) for some classes of G␣ subunits (2)(3)(4)(5). RGS proteins increase the rate of GTP hydrolysis by binding to the switch regions of G␣⅐GTP and stabilizing the transition-state conformation (6).
Over 20 RGS proteins have been identified and are subdivided into four families (R4, R7, R12, and RZ) based on sequence conservation and G␣ subunit preference. All RGS proteins share the highly conserved RGS homology (RH) domain that is required for G␣⅐GTP binding and hydrolysis (7,8). This domain is composed of terminal and bundle subdomains, which consist of the N and C termini and a four-helix bundle, respectively. Additional subfamily-specific domains or regions flanking the RH domain contribute to the subcellular localization of the RGS protein, the specificity and affinity for the G␣⅐GTP subunit, and/or participate in protein-protein interactions (9). The majority of RGS proteins act as GAPs for G␣ i/o and G␣ q subunits. However, some RGS proteins have narrower substrate specificity, such as RGS2, which preferentially binds G␣ q ⅐GTP, or broader substrate specificity, like the RZ subfamily, which can also bind G␣ z ⅐GTP (9,10).
The RZ subfamily, comprised of RGS17 (RGSZ2), RGS19 (GAIP, G␣-interacting protein), and RGS20 (RGSZ1 or Ret-RGS) are among the simplest RGS proteins. They consist of the RH domain flanked by short N and C termini. The N terminus contains a highly conserved cysteine string that can be palmitoylated, as has been reported for RGS19, potentially allowing these proteins to be localized to the plasma membrane (11,12). The defining feature of the RZ subfamily is the residue implicated in GAP activity. The R4, R7, and R12 subfamilies use a highly conserved asparagine residue (e.g. Asn-128 in RGS4) to engage the switch regions of G␣ (6,13,14). In contrast, the RZ family contains a serine at this position (9,13,14). Whether or how this serine recapitulates the interactions of the canonical asparagine residue in promoting GAP activity is not understood.
RGS17 is a potent regulator of cAMP and Ca 2ϩ signaling and is expressed at highest levels in the cerebellum (10,15,16). It was first identified as a GAP for G␣ o and subsequently also found to interact with G␣ z and G␣ i1-3 . Thus, RGS17 increases cAMP accumulation by inhibition of G␣ i/o . RGS17 has also been reported to negatively regulate Ca 2ϩ through a G␣ q -dependent mechanism (15). The preferred substrate of RGS17 is therefore likely dependent upon the cellular context (12). More recently, RGS17 has emerged as a promising therapeutic target in several cancers. RGS17 expression in nonneuronal tissues is linked to lung (17,18) and breast cancer (19) as well as hepatocellular carcinoma (20). The mechanistic role of RGS17 in these diseases is attributed to the increased cAMP driven by the inhibition of G␣ i signaling pathways (17). However, how RGS17 itself is regulated is unclear.
Herein, we present the 1.53-Å structure of RGS17, the most complete structure of an RZ family member and the highest-resolution RGS structure reported to date. RGS17 crystallized as a dimer with strong electron density observed for two Ca 2ϩ atoms in each chain. One site is formed by the side chain of Glu-109 and the backbone of Tyr-106, which are situated on the face of RGS17 that is predicted to bind the switch regions of G␣⅐GTP. This places the site in close proximity to Ser-150, the RZ subfamily residue thought to be responsible for GAP activity and which is analogous to RGS4 residue Asn-128 (13,14). We confirmed that RGS17 binds Ca 2ϩ at the Tyr-106 -Glu-109 site in solution and at a concentration significantly less than that used to obtain the crystal structure. Furthermore, we found that Ca 2ϩ binds to RGS17 with over 50-fold higher affinity than Mg 2ϩ . Finally, we found that although Ca 2ϩ has no effect on the stability of RGS17 itself, Ca 2ϩ enhances the ability of RGS17 to bind G␣ o ⅐GTP. These findings suggest that Ca 2ϩ is a novel potentiator of RGS17 activity. As RGS17 is known to regulate Ca 2ϩ signaling (10,15,16), this could represent a mechanism of feedback inhibition wherein elevated Ca 2ϩ promotes RGS17-G␣ interactions to terminate G␣ signaling.

Crystal structure of RGS17 bound to calcium
RGS17 (residues 70 -206) crystallized as an asymmetric dimer with the final structure refined to 1.53-Å spacings (Table  1 and Fig. 1). The dimer is a crystallographic dimer as RGS17 is monomeric in solution as determined by both size-exclusion chromatography and nuclear magnetic resonance (NMR). Continuous electron density was observed for residues 72-203 in Chain A and 72-206 in Chain X, but the two chains were otherwise essentially identical (r.m.s.d. of 0.11 Å for C␣ atoms in residues 72-203). Overall, the RGS17 RH domain is similar to that of other RGS proteins (7) and a previously determined

Regulation of RGS17 by Ca 2؉
structure of RGS17 (Protein Data Bank (PDB) code 1ZV4 (21)). However, the two RGS17 structures differ from one another by 1.08 Å (C␣ atoms of residues 72-204), which is unexpected given their identical sequence. This arises primarily due to differences within the terminal subdomain and the orientation between the terminal and bundle subdomains of the RH domain (Fig. 2). Superimposing the bundle subdomains from this structure and 1ZV4 confirms that the bundle subdomains are highly similar, with an r.m.s.d. of 0.52 Å for the C␣ atoms of residues 106 -186. In contrast, superimposing the RGS17 structures over their terminal subdomains results in an r.m.s.d. of 0.86 Å for the C␣ atoms of residues 72-105 and 187-206. In addition, the terminal subdomain of our RGS17 structure is rotated by ϳ17°with respect to its orientation in the 1ZV4 structure ( Fig. 2) (22). Each chain of RGS17 in the crystal structure also contained strong electron density consistent with bound Ca 2ϩ , which was present in the crystallization conditions as 200 mM CaCl 2 . Each RGS17 chain coordinated two Ca 2ϩ atoms, but only one site is conserved between the two chains. This site, located in the loop connecting helices ␣3 and ␣4, coordinates Ca 2ϩ through the side chain of Glu-109 and the backbone carbonyl oxygen of Tyr-106 with the rest of the Ca 2ϩ coordination sites occupied by five water molecules (Fig. 3, A and B). These Ca 2ϩ are tightly bound as electron density is still observed when the ͉F o ͉ Ϫ ͉F c ͉ omit map is contoured to 20 (Fig. 1). This Ca 2ϩ ion is on the same face of RGS17 that is predicted to interact with the switch regions of G␣⅐GTP and is ϳ11 Å from Ser-150 (3,23,24). Thus, the Ca 2ϩ ion is positioned to potentially modulate the interactions between RGS17 and its cognate G␣⅐GTP.
Each RGS17 chain in the crystal structure also contains strong electron density for a second Ca 2ϩ ion. However, these secondary sites are not conserved between the two chains. In Chain A, a Ca 2ϩ ion is coordinated by the side chain of Glu-148 and the carbonyl oxygen of Ile-143 with the remaining coordination sites occupied by three water molecules (Fig. 3B). This site is located in the loop connecting helices ␣5 and ␣6, on the predicted G␣-binding surface, and in close proximity to both Ser-150 (ϳ10 Å) and the Ca 2ϩ site formed by Tyr-106 and Glu-109 (Fig. 3B). In Chain X, a Ca 2ϩ is coordinated by the backbone carbonyl oxygen of Gln-124, located in the loop con-necting helices ␣4 and ␣5, along with three water molecules that complete the coordination of Ca 2ϩ (Fig. 3C). Although this Ca 2ϩ is near the dimer interface and the N and C termini of Chain A, it does not contribute to the dimer interface or interact with any residues in Chain A. For both of these sites (Ile-143-Glu-148 and Gln124), the Ca 2ϩ ions are tightly bound as electron density is visible when the ͉F o ͉ Ϫ ͉F c ͉ omit map is contoured to 20 .

RGS17 binds Ca 2؉ and Mg 2؉ in solution
RGS17 and other RGS proteins have not previously been reported to directly bind Ca 2ϩ or other divalent cations. One possible explanation for the presence of bound Ca 2ϩ in our RGS17 structure is that it is an artifact due to the presence of 200 mM CaCl 2 in the crystallization conditions. To establish whether RGS17 binds Ca 2ϩ in solution, we used NMR to monitor changes in the 1 H-15 N 2D HSQC spectrum of RGS17 in the presence of 15 mM CaCl 2 . In this spectrum, protons that are  RGS17 chains are colored as in Fig. 1. Ca 2ϩ ions are shown as black spheres, waters are shown as red spheres, and the distance between the Ca 2ϩ ion and the coordinating atoms are shown in dashed yellow lines. All coordination distances are between 2.3 and 2.6 Å. A, the backbone carbonyl oxygen of Tyr-106 and side chain of Glu-109 in the ␣3-␣4 loop coordinate one Ca 2ϩ ion in Chain X. This is in close proximity to Ser-150, the GAP residue in the RZ subfamily. B, as observed in A, RGS17 Chain A also coordinates Ca 2ϩ via Tyr-106 and Glu-109. A second Ca 2ϩ is bound by the backbone carbonyl of Ile-143 and side chain of Glu-148 in the ␣5-␣6 loop. C, the carbonyl oxygen of Gln-124, located in the ␣4 -␣5 loop, coordinates a Ca 2ϩ ion in Chain X.

Regulation of RGS17 by Ca 2؉
directly bound to 15 N are detected, providing a "fingerprint" of the amide backbone of the protein. To assign each peak in the 1 H-15 N HSQC spectra, the protein backbone and C␤ carbons were first assigned using 13 C-and 15 N-labeled RGS17 and standard triple-resonance experiments.
If RGS17 binds Ca 2ϩ in solution, Ca 2ϩ will alter the local chemical environment, causing perturbations in the chemical shifts of the amide proton and/or nitrogen of residues in close proximity to the bound ion. In contrast, residues that are distant from the site of binding and/or that are unaffected by Ca 2ϩ addition will not have significantly perturbed chemical shifts relative to the 1 H-15 N HSQC spectrum of the protein alone. Addition of 15 mM CaCl 2 to 15 N-labeled RGS17 caused significant chemical shift perturbations (CSPs) for residues Ser-107 (0.178 ppm), Glu-108 (0.160 ppm), Glu-109 (0.065 ppm), and Asn-110 (0.096 ppm) (Ͼ2 S.D. from the average CSP of all residues, which is 0.059 ppm) (Fig. 4). These residues are adjacent to Tyr-106 and Glu-109, the residues that directly bind Ca 2ϩ in both chains of the crystal structure. A modest shift perturbation was also observed for Tyr-106 (0.019 ppm).
The other Ca 2ϩ ions observed in the crystal structure are coordinated by the backbone carbonyl of Ile-143 and side chain of Glu-148 in Chain A or by the backbone carbonyl of Gln-124 in Chain X. Val-149 displays significant perturbation (0.141 ppm), and a CSP for Ser-145 was obtained that was just below the 2 S.D. cutoff (0.058 ppm). However, other residues adjacent to this location could not be definitively assigned in the triple-resonance experiments. Similarly, definitive measurement of the CSP for Gln-124 was not possible due to spectral overlap. However, minimal shift perturbations were observed for the adjacent residues Glu-123 and Asn-125 (Fig. 4). Taken together, these findings suggests that the sites formed by Tyr-106/Glu-109 and Ile-143/Glu-148 are most likely the preferred binding sites for Ca 2ϩ , even at concentrations an order of magnitude lower than that used in crystallization (15 versus 200 mM CaCl 2 ).
We next tested whether another divalent cation could bind to the same site(s) on RGS17 or whether the interaction is unique to Ca 2ϩ . We assessed the ability of Mg 2ϩ to bind RGS17 given its relative abundance in cells. Addition of 15 mM MgCl 2 to 15 N-labeled RGS17 resulted in a significant CSP (Ͼ2 S.D. from average; 0.017 ppm) for Ser-108 (0.024 ppm) (Fig. 5), similar to the CSP observed in the presence of Ca 2ϩ . Ser-107 (0.022 ppm) and Glu-109 (0.072 ppm) also displayed substantial shifts, consistent with Mg 2ϩ binding to this site in RGS17. 15 mM MgCl 2 addition induced CSPs in a second group of residues, including Val-149 (0.031 ppm), Tyr-140 0.024 ppm), and Ser-150 (Ͼ0.023 ppm), located in helix ␣5 and the ␣5-␣6 loop (Fig.  5). These residues are adjacent to or in close proximity to Ile-143 and Glu-148, which coordinate a second Ca 2ϩ site in Chain A of the crystal structure (Figs. 1 and 3) and which display altered chemical shifts in the presence of Ca 2ϩ (Fig. 4).
These results demonstrate that both Ca 2ϩ and Mg 2ϩ bind directly to RGS17 in solution through a site formed by the backbone carbonyl of Tyr-106 and the side chain of Glu-109. This site is also the only Ca 2ϩ -binding site observed in both chains in the crystal structure. Although secondary Ca 2ϩ -binding sites are observed within each chain of the crystal structure, the only other site in which significant CSPs could be reliably observed was the site formed by the backbone carbonyl of Ile-143 and the side chain of Glu-148. This second site may be less favorable for cation binding given that only one RGS17 chain has Ca 2ϩ bound in this site in the crystal structure. Finally, we also found no evidence that the binding of divalent cations impacts the thermal stability of RGS17 as determined by differential scanning fluorimetry ( Fig. S1 and Table S1).

RGS17 has higher affinity for Ca 2؉ than Mg 2؉
Having demonstrated that RGS17 can bind both Ca 2ϩ and Mg 2ϩ at multiple sites, we next determined the affinity of RGS17 for these ions. Using NMR, we observed concentrationdependent changes in CSP for both ions, and the magnitude of CSPs induced by Ca 2ϩ binding was greater than that observed upon Mg 2ϩ binding (Fig. 6). We determined the dissociation constant (K D ) for Ca 2ϩ and Mg 2ϩ for each residue that displayed a significant CSP upon the addition of the divalent cat-

RGS17 binds cations with higher affinity than RGS4 or RGS2
The RH domain is highly conserved across the RGS family, including the residues we have shown to bind cations in RGS17. Additionally, as some of the interactions are mediated by the peptide backbone, it is possible that all RGS proteins bind cations in solution. We used NMR spectroscopy to determine whether RGS4 and RGS2 bind Ca 2ϩ and/or Mg 2ϩ in solution. RGS4 and RGS2 share 40 and 38% identity with RGS17 across the RH domain and are well-characterized with respect to their structure and regulation (7) (Fig. 7 and Fig. S2).
The spectrum of RGS4 has been fully assigned (25), allowing us to identify amino acids that show CSPs upon the addition of Ca 2ϩ or Mg 2ϩ . Using the same approach as described for  . RGS17 binds Ca 2؉ with higher affinity than Mg 2؉ . The K D,avg for Ca 2ϩ and Mg 2ϩ binding to RGS17 was determined for each amino acid that displayed a significant CSP upon the addition of divalent cation. Two cation-binding sites were identified on RGS17, one formed by Tyr-106 and Glu-109 and a secondary site formed by Ile-143 and Glu-148. Residues adjacent to these binding sites that displayed CSPs Ͼ2 S.D. greater than the average CSP were used to calculate the K D,avg for each site by fitting the CSP as a function of ion concentration to a one-site binding model. A, CSP as a function of increasing Ca 2ϩ concentration. The K D,avg for residues Ser-107, Glu-108, Glu-109, and Asn-110 is 132 Ϯ 35 M, and the K D,avg for residues Ser-145 and Val-149 is 91 Ϯ 6 M. B, CSP as a function of increasing Mg 2ϩ concentration. The K D,avg for residues Ser-107, Glu-108, Glu-109, and Asn-110 is 34 Ϯ 23 mM, and the K D,avg for residues Ser-145 and Val-149 is 20 Ϯ 4 mM.  056 ppm). The K D,avg for these Ca 2ϩ binding sites was calculated as a function of ion concentration using a one-site binding model. For the site encompassing residues Tyr-84, Ser-85, Glu-86, Glu-87, Asn-88, and Ile-89, the calculated K D,avg was 9.6 Ϯ 3 mM, whereas the K D,avg for the site associated with residues Ala-123, Lys-125, and Val-127 was 6.1 Ϯ 1.6 mM. Thus, the RGS4 K D,avg values for Ca 2ϩ at the two sites are 72-and 67-fold lower than those calculated for RGS17 (Fig. 7B).
The ability of RGS2 to bind Ca 2ϩ and Mg 2ϩ in solution was then tested. The NMR spectrum of RSG2 has not been assigned, and thus the 1 H-15 N HSQC spectra of RGS2 were compared with spectra obtained with increasing concentrations of CaCl 2 or MgCl 2 (Fig. S2). Addition of a 20-or 250-fold molar excess of CaCl 2 to 15 N-labeled RGS2 caused few changes in the 1 H-15 N HSQC spectra (Fig. S2A). This is consistent with RGS2 binding weakly to Ca 2ϩ in solution. Similarly, addition of up to a 500fold molar excess of MgCl 2 to 15 N-labeled RGS2 had essentially no impact on the 1 H-15 N HSQC spectra (Fig. S2B), demonstrating that RGS2 binds Mg 2ϩ very weakly under these experimental conditions.

Calcium enhances interactions between RGS17 and activated G␣ o
The binding sites for Ca 2ϩ and Mg 2ϩ on RGS17 are located on the same face of the protein predicted to interact with activated G␣ subunits (6, 14, 15) and in close proximity

Regulation of RGS17 by Ca 2؉
to the putative GAP residue, Ser-150. To determine whether Ca 2ϩ impacts the ability of RGS17 to bind activated G␣ o , an AlphaScreen protein interaction assay (26) was used to quantify association. Briefly, RGS17 was biotinylated and immobilized on streptavidin-coated donor beads, whereas activated GST-tagged G␣ o ⅐GDP⅐AlF 4 was immobilized on anti-GST acceptor beads. RGS17-G␣ o binding was quantified as an increase in bead-based fluorescence. RGS17 binds activated G␣ o in a concentration-dependent manner, with saturation occurring at ϳ5 nM G␣ o ⅐GDP⅐AlF 4 (Fig. 8). Addition of 5 mM CaCl 2 had no impact on the apparent affinity of the interaction but did cause a significant increase (p Ͻ 0.01) in the amount of bead-based fluorescence relative to the control, suggesting increased binding between RGS17 and G␣ o ⅐GDP⅐AlF 4 (Fig. 8, A  and B). In contrast, addition of 10 mM EGTA, which preferentially chelates free Ca 2ϩ , had no significant effect on the amount of bead-based fluorescence, suggesting that the increased signal depends upon the presence of Ca 2ϩ . This increase in binding in the presence of Ca 2ϩ was only observed between RGS17 and G␣ o ⅐GDP⅐AlF 4 as Ca 2ϩ had no impact on the interaction between G␣ o ⅐GDP⅐AlF 4 and the closely related RGS4 protein (Fig. 8, C and D).
Mg 2ϩ binds the same sites on RGS17 as Ca 2ϩ in solution and thus could also impact the RGS17-G␣ o ⅐GDP⅐AlF 4 interaction. However, because Mg 2ϩ is required to stabilize the transition state of G␣ o ⅐GDP⅐AlF 4 , its role in the protein-protein interaction cannot be directly assessed. To indirectly probe the role of Mg 2ϩ in binding, 10 mM EDTA was used to chelate free Mg 2ϩ in the binding reaction. Under these conditions, bead-based fluorescence decreased relative to the control. However, it is not possible to determine how much of the decrease is due to perturbation of the RGS17-G␣ o ⅐GDP⅐AlF 4 interaction versus destabilization of activated G␣ o . It is possible that Ca 2ϩ alters the affinity between RGS17 and G␣ o ⅐GDP; however, the affinity of RGS17 for inactive G␣ subunits is too low to accurately determine.

Ca 2؉ increases the GTPase activity of RGS17
Ca 2ϩ selectively binds to RGS17 and increases its interactions with G␣ o ⅐GDP⅐AlF 4 Ϫ . Therefore, Ca 2ϩ may alter the abil-

Regulation of RGS17 by Ca 2؉
ity of RGS17 to stimulate GTP hydrolysis. A GTPase-Glo TM assay was therefore used to measure the rate of GTP hydrolysis on a previously reported rate-altered G␣ i1 mutant (G␣ i1 R178M/A236S) (27) with or without Ca 2ϩ . RGS17 stimulated GTPase activity on G␣ i1 with statistically similar k cat and k cat /K m values regardless of the presence of saturating [Ca 2ϩ ] ( Fig. 9A and Table 2). However, the K m for RGS17-stimulated GTP hydrolysis decreased from 1.49 Ϯ 0.3 to 0.56 Ϯ 0.1 M (p ϭ 0.018) in the presence of Ca 2ϩ .
To determine whether the decrease in K m upon Ca 2ϩ addition was specific for RGS17, the rate of GTP hydrolysis was also measured for RGS4 and the rate-altered G␣ i1 mutant (G␣ i1 R178M/A236S) with or without Ca 2ϩ . RGS4 increased the rate of GTP hydrolysis, but no significant change in the kinetic parameters was observed in the presence of Ca 2ϩ (Fig. 9B and Table 2). Thus, Ca 2ϩ appears to selectively enhance the GTPase activity of RGS17. The impact of Ca 2ϩ on GTP hydrolysis was also investigated using G␣ q . However, G␣ q hydrolyzed GTP too quickly for accurate measurement in this assay.

Calcium alters the binding mechanism between RGS17 and activated G␣ o
Isothermal titration calorimetry (ITC) was utilized to further characterize the RGS17-G␣ o binding interaction in the presence and absence of saturating concentrations of Ca 2ϩ (Fig. 10). We observed no significant difference in the K D of the RGS17-G␣ o interaction in the presence (596 Ϯ 257 nM) or absence (611 Ϯ 128.5 nM) of Ca 2ϩ . The stoichiometry of the G␣ o -RGS17 complex was also found to be unchanged in the presence or absence of Ca 2ϩ . However, the binding enthalpy for the interaction in the presence and absence of Ca 2ϩ was found to be significantly different at Ϫ2.76 Ϯ 0.74 and Ϫ7.33 Ϯ 0.72 kcal/ mol respectively. The presence of Ca 2ϩ in the experiment increased the enthalpy by 4.57 kcal/mol, which is consistent with Ca 2ϩ changing the binding mechanism between RGS17 and G␣ o . This change in binding is most likely due to a decrease in the electrostatic interactions between RGS17 and G␣ o and/or alteration of the local structure at the protein-protein interface (28,29).

Discussion
RGS proteins are critical negative regulators of GPCR signaling through their ability to act as GAPs for activated G␣ subunits. The RZ subfamily inhibits GPCR signaling in the nervous system where they inactivate G␣ z , G␣ o , and G␣ i1-3 , thereby preventing G i -dependent inhibition of adenylyl cyclase (9, 10). RGS17, a member of the RZ family, has also been reported to negatively regulate Ca 2ϩ signaling, suggesting that it also contributes to the regulation of G␣ q -dependent processes (15). RGS17 has emerged as a driver in cancer, in particular lung and breast cancers (17)(18)(19), where its overexpression results in increased inhibition of G␣ i -dependent signaling, thereby increasing cAMP and PKA activity (17). However, how RGS17 itself is regulated is not well characterized. In this study, we report a high-resolution crystal structure of RGS17 revealing that this protein binds Ca 2ϩ and provide support for a mechanism wherein Ca 2ϩ binding to RGS17 enhances its interactions with activated G␣.
RGS17 crystallized as a dimer, and both chains in the asymmetric unit preserve the canonical RH fold (Fig. 1) (7). Despite being identical in sequence to a previously published RGS17 structure (PDB code 1ZV4 (21)), our RGS17 structure differs from 1ZV4 in two major ways. First, the terminal subdomain is rotated with respect to the bundle subdomain by ϳ17°relative to their orientation in the 1ZV4 structure (Fig. 2) (21,22). This suggests that the terminal and bundle subdomains may be flexible in their relative orientation, which could potentially facilitate binding to activated G␣ subunits. Second, our structure of RGS17 shows strong electron density for four well-resolved Ca 2ϩ ions, with each chain in the asymmetric unit binding two Ca 2ϩ ions (Fig. 1). One Ca 2ϩ site, formed by the backbone carbonyl of Tyr-106 and the side chain of Glu-109, is observed in both chains of the crystal structure (Fig. 3, A and B). We confirmed that RGS17 binds Ca 2ϩ or Mg 2ϩ in solution by monitoring CSPs in the NMR spectra of 15 N-labeled RGS17 in the presence or absence of CaCl 2 or MgCl 2 (Figs. 4 and 5). Each RGS17 chain in the crystal structure also bound a second Ca 2ϩ Figure 9. Ca 2؉ increases RGS17-stimulated GTP hydrolysis. A GTPase-Glo assay was used to detect and quantify RGS-stimulated GTP hydrolysis on a rate-altered G␣ i1 mutant, G␣ i1 R178M/A326S. A, RGS17 increases the rate of GTP hydrolysis in the presence (red squares) or absence (blue circles) of saturating CaCl 2 . Addition of CaCl 2 significantly decreased the K m (Table 2). B, RGS4 stimulates GTP hydrolysis on G␣ i1 but is insensitive to the presence of CaCl 2 . Data represents the mean of four independent experiments ϮS.E. (error bars).

Table 2 Steady-state kinetic parameters for RGS17 and RGS4 GTPase activity
The GTPase activity of RGS17 and RGS4 was measured using a rate-altered G␣ i1 variant (26,45) in the presence or absence of saturating CaCl 2 (100 M). Data represents the mean of four independent experiments ϮS.E. *, p ϭ 0.018.

Regulation of RGS17 by Ca 2؉
through the backbone carbonyl of Gln-124 or via the backbone carbonyl of Ile-143 and the side chain of Glu-148. We demonstrated that this latter site also displayed CSPs upon addition of MgCl 2 and CaCl 2 and therefore likely represents a secondary cation-binding site (Fig. 5). Subsequent NMR experiments examined the affinities of these two metal-binding sites for Mg 2ϩ and Ca 2ϩ . RGS17 was found to preferentially bind Ca 2ϩ in solution as the site formed by Tyr-106 and Glu-109 exhibited a K D,avg of 132 Ϯ 35 M, a 257-fold increase in the affinity relative to Mg 2ϩ . The metal-binding site formed by the carbonyl of Ile-143 and the Glu-148 side chain showed a similar trend in the affinity where the K D,avg was found to be 91 Ϯ 6 M for Ca 2ϩ , a 219-fold higher affinity than that observed for Mg 2ϩ (Fig. 6).
RGS17 and potentially the RZ subfamily bind Ca 2ϩ and Mg 2ϩ to a greater extent than other RGS proteins. For example, RGS4 is able to bind Ca 2ϩ in solution through sites equivalent to those observed in RGS17 (Fig. 7) but with ϳ70-fold lower affinity. RGS4 bound even more weakly to Mg 2ϩ with 10-fold lower affinity than Ca 2ϩ . However, the calculated K D,avg values for the two cation-binding sites were within the error range of the experiment (93 Ϯ 193 mM for the site including residues Tyr-84 and Glu-86 and 90 Ϯ 75 mM for the site formed by Val-121 and Glu-126; Fig. 7). Thus, RGS4 does not appreciably bind Mg 2ϩ under these conditions. Similar results were obtained with RGS2, which had only minimal changes in its 1 H-15 N spectra at Ca 2ϩ concentrations up to 250-fold excess and in the spectra observed at 500-fold excess Mg 2ϩ (Fig. S2). Thus, despite the sequence conservation of the RH domain and the residues involved in cation binding, RGS4 and RGS2 interact only very weakly with Ca 2ϩ in solution.
Ca 2ϩ and Mg 2ϩ bind the same face of RGS17 that is predicted to interact with the switch regions of activated G␣ subunits to promote GTP hydrolysis (Fig. S3). Comparison of the RGS17-Ca 2ϩ structure with RGS-G␣ complexes (6, 30 -33) would suggest that Ca 2ϩ could inhibit G␣ binding. To test this hypothesis, an AlphaScreen protein interaction assay was used to measure binding between RGS17 and G␣ o . Addition of Ca 2ϩ increased the observed binding between RGS17 and G␣ o , whereas addition of EGTA had no effect on the interaction (Fig.  8). The increased interactions between RGS17 and G␣ may also contribute to changes in GTP hydrolysis. We found that Ca 2ϩ decreased the K m for RGS17-stimulated GTP hydrolysis on G␣ i1 but had no impact on RGS4-stimulated GTP hydrolysis ( Fig. 9 and Table 2). Finally, isothermal titration calorimetry revealed a significant difference in the binding enthalpy for the RGS17-G␣ o interaction brought about by the presence of Ca 2ϩ (Fig. 10). Ca 2ϩ was found to increase the binding enthalpy by 4.57 kcal/mol, demonstrating that Ca 2ϩ changes the binding mechanism for RGS17-G␣ o . Polar interactions and/or conformational changes are typically reflected in the enthalpy of binding (28,29). Thus, Ca 2ϩ may alter binding by decreasing the polar contacts between RGS17 and G␣ o and/or altering the local structure at the protein interfaces.
Taken together, our data suggest Ca 2ϩ regulates RGS17 activity. Although Ca 2ϩ binding directly to other RGS proteins has not been reported previously, there is precedent for Ca 2ϩmediated regulation of RGS activity. For example, Ca 2ϩ /calmodulin is a known regulator of some RZ and R4 subfamily members. These RGS proteins are inhibited when bound to phosphatidylinositol-3,4,5-triphosphate at the cell membrane, and this inhibition is relieved upon binding Ca 2ϩ /calmodulin  (9, 34 -36). Thus, these proteins are activated following G q -dependent Ca 2ϩ release, providing a feedback mechanism to inhibit further G q signaling (34,35). Additional studies will be required to determine whether Ca 2ϩ , alone or in combination with calmodulin, may be a general regulator of RZ RGS function.

RGS17 expression and purification
A construct for the RGS17 protein, encoding residues 70 -206, was obtained from Addgene (catalog number 39141) and was a gift from Nicola Burgess-Brown (The Structural Genomics Consortium). RGS17 was purified largely as described previously (21). Briefly, RGS17 was expressed in BL21-CodonPlus(DE3)-RIPL cells and grown at 37°C and 275-300 rpm until an OD 600 of 2.0 was reached. Protein production was induced with 0.5-1 mM isopropyl 1-thio-␤-D-galactopyranoside, and the culture was incubated for an additional 16 h at 18°C while shaking at 275-300 rpm. Bacterial cells were then pelleted and resuspended in 50 mM HEPES, 500 mM NaCl, 1 mM ␤-mercaptoethanol, 10 mM imidazole at pH 8 (Buffer A) at 4°C. Cells were lysed with lysozyme (1 mg ml Ϫ1 cell pellet), and DNase I (ϳ2 mg) was added. Lysate was then subjected to multiple freeze-thaw cycles in liquid N 2 , and the soluble lysate fraction was separated by centrifugation at 100,000 ϫ g. Histagged RGS17 was then separated from the supernatant using an Ä KTA FPLC (GE Healthcare) equipped with an immobilized metal affinity chromatography column (Ni-Sepharose 6 Fast Flow, GE Healthcare). Eluted fractions containing RGS protein were then treated with His-tagged tobacco etch virus protease at a molar ratio of ϳ1:20 tobacco etch virus:RGS and dialyzed overnight at 4°C against 5 liters of Buffer A to cleave the His 6 tag. Samples were again subjected to the immobilized metal affinity chromatography column, and the flow-through was collected. Size-exclusion chromatography (10 mM borate, 500 mM NaCl, and 1 mM DTT at pH 7.0) was then used to obtain 99ϩ% pure RGS17 as determined by SDS-PAGE.
Isotope-labeled ( 15 N and 13 C, 15 N) RGS17 was purified largely as above with the exception that when culture OD 600 reached 1.5, cells were pelleted at 3,500 ϫ g at 4°C for 15 min and resuspended in an equal volume of M9 minimal medium supplemented with 2 g liter Ϫ1 D-[ 13 C 6 ]glucose and 1 g liter Ϫ1 15 NH 4 Cl for 13 C, 15 N-labeled sample or 1 g liter Ϫ1 15 NH 4 Cl for 15 N-labeled sample. Isotope-labeled samples were concentrated to Ͼ1 mM in 20 mM K 2 HPO 4 buffer with 100 mM NaCl, 0.5 mM ␤-mercaptoethanol, and 2 mM NaN 3 at pH 7.6. Prior to all NMR experiments, RGS17 was exhaustively dialyzed against 55 mM HEPES, 110 mM NaCl, and 0.55 mM ␤-mercaptoethanol at pH 7.6 to remove phosphate buffer.

Crystallization of RGS17
Initial crystallization conditions were determined using commercially available screens. Hanging-drop vapor-diffusion experiments were set up in 96-well polystyrene microplates (Greiner Bio-one) using a Mosquito LCP crystallization robot (TTP Labtech) at 25°C. The drops contained an equal volume (200 nl) of RGS17 (20.6 mg/ml in 10 mM borate, pH 7.0, 500 mM NaCl, and 1 mM DTT) and reservoir solution (0.2 M CaCl 2 and 20% PEG 3350, pH 5.1) of the PEG/Ion Screen (Hampton Research) suspended over 50 l of reservoir. Conditions were optimized in-house using 24-well SuperClear Pregreased plates (Crystalgen) with drops containing equal volumes (0.5 l) of RGS17 (16 mg/ml) and precipitant solution. Final crystals were obtained from reservoirs containing 0.2 M CaCl 2 , 22% (w/v) PEG 3350, and 0.1 M MES, pH 6.0, at 12°C using streak seeding. Crystals were harvested in 0.2 M calcium chloride dihydrate, 40% PEG 3350, and 0.1 M MES, pH 6.0, and frozen on nylon loops in liquid N 2 .

Data collection, processing, and refinement
Diffraction data were collected at 100 K using an Eiger detector at the Advanced Photon Source at LS-CAT 21-ID-D. HKL2000 was used to integrate and scale the data, and Phaser in CCP4 (37) was used to solve the structure by molecular replacement with the prior structure of RGS17 (PDB code 1ZV4 (21)) as a starting model. The structure was built by manual model building in Coot (38) alternating with translation/libration/screw (TLS) refinement in REFMAC5 (39). The correctness of the structure was assessed using MolProbity (40). Structure figures were generated using PyMOL 1.8.6.2 (Schrödinger, LLC).

Nuclear magnetic resonance
The following triple-resonance experiments were performed at 25°C using a 600-MHz Varian INOVA NMR spectrometer equipped with a triple-resonance gradient probe to assign RGS17-RH backbone (and C␤) chemical shifts: HNCACB, CBCA(CO)NH, HNCO, and HN(CA)CO). Data were processed and analyzed using NMRPipe (41) and CCPNAnalysis (42), respectively. 260 M RGS17 in 50 mM HEPES, 100 mM NaCl, and 0.5 mM ␤-mercaptoethanol, pH 7.6, in 10% D 2 O was incubated with or without the indicated concentration of CaCl 2 or MgCl 2 . 1 H-15 N HSQC spectra were then acquired at 25°C using either a 500-MHz Bruker Avance II or a 600-MHz Varian INOVA NMR spectrometer, each equipped with a triple-resonance gradient probe.
NMR experiments with RGS4 and RGS2 were carried out as described for RSG17. 1 H-15 N HSQC spectra of 375 M RGS4 in 50 mM HEPES, 100 mM NaCl, and 10 mM ␤-mercaptoethanol at pH 7.6 and 10% D 2 O or 300 M RGS2 in 50 mM HEPES, 100 mM NaCl, and 11 mM ␤-mercaptoethanol at pH 7.6 in 10% D 2 O were acquired followed by addition of increasing concentrations of CaCl 2 or MgCl 2 . Chemical shift assignments for RGS4 were confirmed using prior assignments (25), whereas the spectrum of RGS2 has not been assigned.
CSPs between control and metal-treated samples were calculated by measuring the distance between the centers of the peaks using the following equation, where ⌬␦ H and ⌬␦ N are the difference in chemical shift in the absence and presence of cation in the indicated dimension.
K D values for individual residues with CSP Ն2 S.D. from the mean were obtained using GraphPad Prism 7 by fitting CSP titration data to a one-site binding model with correction for ligand depletion as follows,

Regulation of RGS17 by Ca 2؉
where CSP max is the maximum CSP value observed, [M 2ϩ ] is the concentration of metal ion, [P] is the concentration of protein, and K D is the dissociation constant. Note that 2 ϫ [P] was used to account for the two metal-binding sites on each molecule of RGS17. K D values obtained for residues were averaged to determine the K D,avg of metal-binding sites. Identical residues were used for Mg 2ϩ K D determinations.

Differential scanning fluorimetry
The impact of Ca 2ϩ or Mg 2ϩ on the thermal stability of RGS17 was determined by measuring the change in fluorescence of SYPRO Orange (Molecular Probes, Eugene, OR) due to protein denaturation as a function of temperature (43). RGS17 was exchanged into buffer containing 10 mM borate, pH 9.0, 500 mM NaCl, and 1 mM DTT. RGS17 at a final concentration of 0.9 mg/ml was incubated with 5ϫ SYPRO Orange dye and increasing concentrations of CaCl 2 or MgCl 2 (0.2 nM-200 mM) for 30 min on ice in a final volume of 20 l. Samples were assayed in triplicate in a MicroAmp Optical 96-well plate, sealed with MicroAmp Optical Adhesive Film (Applied Biosystems), and centrifuged for 1 min. Differential scanning fluorimetry assays were carried out on a ViiA7 qPCR instrument (Thermo Fisher). The change in fluorescence was measured at 0.2°C intervals from 25 to 95°C. The T m was calculated by fitting the increase in fluorescence as a function of temperature to a Boltzmann sigmoid (GraphPad Prism 7.0). Data represent the mean of at least three experiments performed in triplicate ϮS.E.

AlphaScreen method for the RGS17-G␣ interaction
RGS17-G␣ binding was assessed as described previously (44). Biotinylated RGS17 was conjugated to streptavidin donor beads in ALPHA Buffer (20 mM HEPES, 100 mM NaCl, 1% BSA, and 1% Lubrol, pH 8). GST-G␣ , purified as described previously (27), was conjugated to anti-GST acceptor beads in ALPHA Buffer. The G␣ mixture was supplemented with 5 M AlCl 3 , 5 mM MgCl 2 , and 5 mM NaF (AMF) and 2.5 mM GDP. Final concentrations were 100 nM RGS17, the indicated concentration of G␣ o , and 15 ng/l for each bead. The assay was incubated for 1 h at ambient temperature, and then fluorescence was measured using a PerkinElmer Life Sciences Envision plate reader. Wells lacking AMF represented negative control and were normalized to 0%, and wells containing RGS17 in ALPHA Buffer alone were normalized to 100%. Data analysis was performed using GraphPad Prism 7.

GTP hydrolysis assays
RGS-stimulated GTP hydrolysis was measured using a ratealtered G␣ i variant, G␣ i1 R178M/A326S, which was expressed and purified as described previously (27,45). GTPase activity was measured using the GTPase-Glo assay (Promega, Madison, WI) as described (46) but with some modifications. Briefly, 1 M G␣ i1 R178M/A326S and 1 M RGS17 or RGS4 were incubated in GTPase/GAP reaction buffer (50 mM Tris-HCl, pH 7.5, 50 mM NaCl, and 10 MgCl 2 ) in the presence or absence of 100 M CaCl 2 for 30 min. The reaction was initiated by the addition of 2.5 M GTP (10-l final volume) and allowed to proceed for increasing time points before the addition of an equal volume of GTPase-Glo reagent and a 30-min incubation. The GTPase-Glo reagent uses a nucleoside-diphosphate kinase and ADP to convert remaining GTP to ATP and GDP. The GTPase activity inversely correlates with ATP production and is measured with a detection reagent containing a luciferase/luciferin mixture. 20 l of this detection reagent was added to 20 l of the reaction mixture and incubated in the dark for 10 min. Assay plates (Corning, 3572; 384-well) were read on a Synergy 2 plate reader (BioTek, Winooski, VT) in luminescence mode. Time points at 0.5, 1, 2, 3, 4, 5, and 10 min post-GTP addition were taken. Wells without GTP were used in normalization of values for data analysis to represent 100% GTP hydrolysis, whereas wells without G␣ i1 R178M/ A326S represented 0% GTP hydrolysis. Data analysis was performed using GraphPad Prism 7.