Structural organization of a major neuronal G protein regulator, the RGS7-Gβ5-R7BP complex

Signaling by the G-protein-coupled receptors (GPCRs) plays fundamental role in a vast number of essential physiological functions. Precise control of GPCR signaling requires action of regulators of G protein signaling (RGS) proteins that deactivate heterotrimeric G proteins. RGS proteins are elaborately regulated and comprise multiple domains and subunits, yet structural organization of these assemblies is poorly understood. Here, we report a crystal structure and dynamics analyses of the multisubunit complex of RGS7, a major regulator of neuronal signaling with key roles in controlling a number of drug target GPCRs and links to neuropsychiatric disease, metabolism, and cancer. The crystal structure in combination with molecular dynamics and mass spectrometry analyses reveals unique organizational features of the complex and long-range conformational changes imposed by its constituent subunits during allosteric modulation. Notably, several intermolecular interfaces in the complex work in synergy to provide coordinated modulation of this key GPCR regulator.


Introduction
G-protein-coupled receptors (GPCRs) constitute the largest family of cell surface receptors that endow cells with an ability to detect and respond to a vast array of extracellular stimuli including neurotransmitters, hormones, ions, and light (Shoichet and Kobilka, 2012). GPCR signaling systems are involved in virtually all essential physiological processes and their dysregulation is firmly linked to a number of human diseases (Gutierrez, 2018;Insel et al., 2007;Iyinikkel and Murray, 2018). Accordingly, an estimated one-third of all FDA approved and marketed drugs target GPCRs for therapeutic benefits (Hauser et al., 2017). Therefore, understanding how GPCRs transmit their signal and how this process is regulated is of paramount importance.
Ligand-bound GPCRs transmit signals by catalyzing the exchange of GDP for GTP on the heterotrimeric G proteins thereby releasing active GaGTP and free Gbg subunits, which in turn engage a wide range of the intracellular effectors to produce a cellular response (Pierce et al., 2002). A wealth of accumulated evidence indicates that the key role in regulating this process belongs to a conserved family of regulators of G-protein signaling (RGS) proteins (Hollinger and Hepler, 2002;Ross and Wilkie, 2000;Kimple et al., 2011). RGS proteins act as GTPase-activating proteins (GAPs) by interacting directly with activated Ga and accelerating GTP hydrolysis thereby promoting G-protein inactivation upon reassembly of the inactive Gabg heterotrimer. The action of the RGS proteins is essential in providing control over both the extent and duration of signaling (Hollinger and Hepler, 2002;Ross and Wilkie, 2000) and deficits in this control results in severe dysregulation of GPCR signaling causing a number of human pathophysiological conditions (Woodard et al., 2015;Salaga et al., 2016;Druey, 2017;Sjö gren, 2017).
More than 30 RGS proteins have been identified and grouped into six subfamilies (Zheng et al., 1999). Among them, the R7 family (RGS6, RGS7, RGS9, and RGS11) stands out for its evolutionary conservation in all animals from worm to man and crucial roles in multiple processes and organ systems including nervous and cardiovascular system function, vision, movement control, and cellular proliferation with ever growing causal connection to many diseases from blindness to cancer (Ahlers et al., 2016;Qutob et al., 2018;Yang et al., 2013;Anderson et al., 2009b). A unique hallmark of R7 RGS proteins is their modular architecture encompassing several domains and subunits. The catalytic RGS domain at the C-terminus is the site for the interaction with the Ga subunits (Hooks et al., 2003). The central GGL domain forms an obligatory heterodimer with type 5 G-protein b-subunit (Gb5) which stabilizes the complex and docks it onto effectors (Snow et al., 1998;Xie et al., 2010). In addition, the molecule contains the N-terminal interwoven DEP (Dishevelled, Egl10, Pleckstrin) and DHEX (DEP helical extension) domains, a site for major allosteric regulation of the complex via association with GPCRs (Orlandi et al., 2012;Sandiford and Slepak, 2009;Kovoor et al., 2005) and small SNARE-like membrane proteins: the R7-binding protein (R7BP) and/ or RGS9-anchor protein (R9AP) (Hu and Wensel, 2002;Drenan et al., 2005;Song et al., 2006;Martemyanov et al., 2005). This complexity of organization is thought to reflect regulatory precision and flexibility adapting R7 RGS proteins to universally meet diverse needs of multiple GPCR pathways in changing molecular landscape of different cells. Yet, our understanding of the structural basis underlying functional diversity is limited. The only structure reported to date is that of the RGS9-Gb5 dimer, a specialized R7 RGS member with expression limited to striatal neurons and photoreceptors. Furthermore, no structural information exists for the regulation of RGS proteins by R7BP/R9AP subunits.
RGS7 is a prototypic member of the family, broadly expressed in the nervous system and periphery and involved in regulation of multiple GPCR systems with key roles in physiology. A few documented examples include regulation of mGluR in retina bipolar (Cao et al., 2012), GABA(B) in hippocampal CA1 (Fajardo-Serrano et al., 2013;Ostrovskaya et al., 2014), m-opioid in striatal neurons (Sutton et al., 2016;Anderson et al., 2010;Masuho et al., 2013), and M3 muscarinic receptors in the pancreatic b-cells (Wang et al., 2017). Given these roles, perhaps not surprisingly, the RGS7 complex has been found to play a critical role in vision, learning and memory, drug addiction, insulin production, and cancer progression (Qutob et al., 2018;Sutton et al., 2016;Wang et al., 2017). Furthermore, there is a growing appreciation that the RGS7 complex can be exploited as a target for drug development with the hopes of increasing efficacy, selectivity, and the side effect profile of existing GPCR medications (Muntean et al., 2018). Thus, so far the lack of structural information on organization and regulation of the RGS7 complex has been a major bottleneck in understanding basic biology of GPCR signaling regulation and rational design of small molecule modulators.
Here, we report the high-resolution crystal structure of the RGS7-Gb5 dimer and the structural basis for its association with R7BP revealed by hydrogen-deuterium exchange as detected by mass spectrometry (HDX-MS) analyses. In combination with molecular dynamics simulations, our data reveal unique structural features of the RGS7 complex that make it suitable for the regulation of diverse GPCR target pathways. These include the long-range rearrangement in the complex induced by R7BP binding and of the complex providing structural basis for its allosteric regulation. Together our findings reconcile a wealth of biochemical observations pertaining to the function of RGS7 thereby providing crucial insights into the biology of this essential GPCR signaling regulator.

Results
Overall structural organization of the RGS7-Gb5 heterodimer To gain structural insights into the specificity of the RGS7-Gb5 interaction, we solved the 2.13 Å crystal structure of the RGS7-Gb5 dimer complex which required buffer optimization using thermofluor ( Figure 1, Tables 1 and 2). The structure contains all domains common to the R7 RGS family members: DEP-DHEX, GGL and RGS ( Figure 1). The core of the molecule is formed by the Gb5 subunit, which has a typical b-propeller fold commonly found in domains with WD40 repeats. Notably, Gb5 is engulfed by all RGS7 domains, making extensive contacts at four main interfaces: the DEP domain, the DHEX-GGL linker, the Gg-like (GGL) domain, and the RGS domain (Figure 1-figure supplement 1). Two major distinctive features of RGS7-Gb5 include a unique arrangement of the DEP-DHEX and GGL-Gb5 modules, discussed in detail below. The total buried solvent-accessible surface area of Gb5 is 29% by RGS7 domains and involves 108 out of 353 Gb5 amino acids. Electron density is visible for residues 19-218 and 252-450 of chain A as well as 19-218 and 252-450 of chain B of RGS7 and for residues 2-353 of chain C and 4-353 of chain D of Gb5. Overall, the structure shows extensive coordination between multiple RGS structural features integrated around the central 'scaffolding' element provided by the Gb subunit.

Distinct conformational arrangement of the DEP-DHEX module
The DEP domain forms the N-terminus of RGS7 protein. DEP domains are present in a number of signaling proteins and play diverse functions with the common role of targeting signaling proteins to specific membranous subcellular compartments (Kovoor et al., 2005;Consonni et al., 2014;Paclíková et al., 2017;Chen and Hamm, 2006;Martemyanov et al., 2003;Ballon et al., 2006;Civera et al., 2005). Despite limited amino acid sequence identity, the DEP domain of RGS7 shows high structural similarity with the DEP domains of other proteins with known structures featuring a characteristic a/b fold (Figure 2-figure supplement 1a). It features a three helix bundle covered by an antiparallel b-sheet and a flexible loop (the Da1Da2 loop). In several available atomic structures of DEP domains, this flexible loop forms a protruding b-hairpin arm. Superposition of DEP domains of RGS7, RGS9, DVL, EPAC, and pleckstrin shows considerable structural diversity in the organization of this loop known for its involvement in molecular interactions ( Figure 2a). Interestingly, in RGS7, the Da1Da2 and Ea1Ea2 loops adopt a unique conformation in both molecules within asymmetric unit documenting yet another alternative organization and intrinsic flexibility of this region ( Figure 2-figure supplement 1b). In the structure, the DEP domain is tightly integrated with the DHEX domain, a unique structural module present only in R7 RGS proteins. Overall, it resembles the four-helix bundle organization reported for RGS9 (Cheever et al., 2008). In both DHEX domains, three antiparallel a-helices wrap around one a-helix at the center forming the characteristic helix wrap. The fourth a-helix of DHEX interfaces with the DEP domain via contacts with the b-sheet and flexible loops of the DEP domain.
Despite the overall similarity, the organization of the DEP-DHEX module in RGS7 is strikingly different from that seen in RGS9 (Figure 2b). The most distinctive feature is the different orientation of the DHEX domain relative to DEP. In RGS7, the DHEX domain is lifted upward and away from Gb5 thereby creating a wider groove between DHEX and Gb5 when compared to RGS9. Moreover, the organization of all four a-helices is distinct. The DHEX domain of RGS7 has shorter a-helices and longer connecting loops compared to RGS9. Most prominently, this includes much longer Ea1Ea2 and DEP-DHEX loops that adopt different conformations ( Figure 2b). Together, this distinct organization of the Ea1Ea2 and Da1Da2 loops positions them close to the invariant Ea3Ea4 loop to assemble a strong cluster of basic residues on the surface of the DEP-DHEX interface (Figure 2c). Such a structural bow is absent in RGS9 where basic residues are scattered without forming organized basic patch ( Figure 2c). Thus, both relative orientation of DEP and DHEX domains and organization of individual constituent structural elements contribute to distinct features of the RGS7.  (Weiss and Hilgenfeld, 1997;Weiss, 2001). † Pearson correlation coefficient calculated between the average intensities of each random half of measurements of unique reflections (Karplus and Diederichs, 2012).

Unique conformational state and interactions of the central scaffolding subunit Gb5
Gb5 contains identical seven bladed b-propellers (WD40 domains), as seen for the canonical Gb subunit, Gb1. Comparison of the RGS7 and RGS9-containing dimers reveals that the same Gb5 protein adopts different conformations indicating flexibility and that the conformation of Gb5 could be influenced by the identity of the associated RGS protein (Figure 3a). One prominent difference includes the organization of the N-terminal a-helix. In RGS9, it displays the same conformation as observed in Gb1 but in RGS7 it protrudes outward from the b-barrel core. Another difference lies in the organization of the S5b1b2 and S3b3b4 loops which are oriented in a distinct fashion ( Figure 3a). Interestingly, these two loops have low sequence conservation with other Gb subunits each containing two amino acids insertions (Figure 3-figure supplement 1). Such unique organization likely underlies specialization in interactions with the RGS proteins, and possibly other partners, as discussed in detail below. Notably, Gb5 forms extensive contact interfaces with all RGS7 modules, playing a critical role in their spatial arrangement. The top surface of Gb5 contacts two sites in RGS7: the DHEX-GGL linker and the DEP domain ( Figure 3b). The DHEX-GGL linker interacts with the 'hotspot' region in Gb5, which in canonical Gb subunits is engaged in interactions with various effector molecules. In fact, there is a significant overlap between the hotspot surface in Gb1 and the corresponding residues in Gb5 involved in binding to the DHEX-GGL linker. Out of the 12 hotspot residues that engage in shared contacts with Ga, GRK2, and phosducin in Gb1, seven in Gb5 engage in interactions with the amino acids of the DHEX-GGL linker (Figure 3c; Figure 3- figure supplements 1 and 2). However, only a single residue in the Gb5 'hotspot' overlaps with the surface that Gb1g2 uses to interact with GIRK (W107 in Gb5, W99 in Gb1), which binds to both Gb1 and Gb5 complexes, suggesting a unique organization of the effector binding interfaces in Gb5 relative to canonical Gb subunits (  Several features also distinguish the organization of the interface that the top surface of Gb5 forms with RGS7 compared to previously observed analogous interactions with RGS9. In RGS7, the DHEX-GGL linker forms a 3 10 -helix embedded into the b-propeller ring at the hotspot region of Gb5 ( Figure 3b). In particular, three residues (D213, K215, and K216) insert deep into the ring, forming extensive contacts with surrounding residues of Gb5 (  PDB entry 2byv), pleckstrin (brown; PDB entry 1w4m), and Dvl (grey; PDB entry 1fsh) reveals diversity in the conformational organization of the b-hairpin arm. (b) Comparison of the DEP-DHEX domains of RGS7 and RGS9 shows differences in several loops and in the orientation of the DHEX domain. In RGS7, the DHEX domain is lifted upward thus generating a wide space between the domains. (c) Evaluation of the electrostatic surface potentials of the DEP-DHEX domains of RGS7 (left) and RGS9 (right). A strong basic patch (blue) is formed by three loops in the RGS7 structure but differences in the organization of these loops result in a more scattered distribution of basic charges in the RGS9 structure. The online version of this article includes the following figure supplement(s) for figure 2:   with RGS9 and Gb1. The differences between the loop regions are highlighted in red boxes. The N-terminal ahelix of Gb5 from RGS7 also shows variation compared to Gb5 from the RGS9 complex and Gb1 from the Gabg complex. (b) Surface representation of Gb5 top face (grey) highlighting the RGS7 contact residues (red). Two contact interfaces that are formed by the DHEX-GGL linker and the DEP domain are shown as a carton in green and pink, respectively. (c) The hotspot region (green) mapped onto the Gb5 surface based on contacts formed by known Gb1 binding partners. RGS7 (red) overlaps with the hotspot footprint. The DEP domain forms a unique contact surface on the top face of Gb5. (d) Distinct interacting residues of the DEP domain of RGS7 (pink) that forms direct contacts with Gb5 (cyan) are shown along with equivalent residues of RGS9. A unique electrostatic interaction is formed by Da1 residue R33 of the RGS7 DEP domain with E280. Its organizational equivalent in RGS9 is formed by interactions with R62 from a-helix Da2. The Da2-Da3 loop of the DEP domain that is involved in interactions with Gb5 has a different organization in RGS7 and RGS9 and is indicated by an arrow. The online version of this article includes the following figure supplement(s) for figure 3:     (c) The Ga1 a-helix of RGS7 is located 9.6 Å away from its position in RGS9, Gg1, and Gg2 diminishing the contacts of the Ga1 a-helix with Gb5. (d) Schematic view of detailed comparison of the direct contact residues of GGL of RGS7 and RGS9 with Gb5 residues. Most of the contact residues of Figure 4 continued on next page accommodates Gb5 with most extensive interactions. As noted before (Snow et al., 1998), it is structurally similar to the canonical G protein Gg1 and Gg2 subunits as well as GGL domain of RGS9 superimposing with r.m.s.d. values of 1.6 Å , 1.7 Å , and 1.8 Å , respectively ( Figure 4b). Strikingly, the N-terminal Ga1 a-helix of the RGS7 GGL domain adopts a substantially different conformation. It is tilted about 9.6 Å relative to the position occupied by the corresponding a-helices in all previously reported crystal structures of the Gg1, Gg2, and GGL domains of RGS9 ( Figure 4b). This increases its distance to Gb5 thus loosing all interactions with the b1 sheet of the WD5 propeller in Gb5, prominently maintained in RGS9 ( Figure 4c).
The core of contacts with Gb5 are formed by residues of the Ga2-Ga5 a-helices of RGS7 GGL ( These additional contacts are unique to RGS7 and the region involved shows poor sequence conservation with RGS9 GGL (35%), Gg1 (23%), and Gg2 (33%) (Figure 4-figure supplement 1b). We further detect the 'selectivity filter' within Ga4-Ga5 a-helices that makes the GGL domain incompatible with canonical Gb binding, similarly to what was seen in RGS9. Overall, these features point to a unique conformational arrangement of Gb5 positioning it at the heart of interactions within the RGS complex.

Organization of the RGS domain and its interface with the Gb5
The C-terminus of the RGS7 molecule contains the hallmark feature of the entire RGS family, namely the RGS homology domain. It is structurally similar to RGS domains of other proteins with known structures composed of nine a-helices (a1-a9) arranged into two bundles. Furthermore, it is nearly identical to the solution structure of the isolated RGS domain of human RGS7 (PDB entry 2d9j) with r.m.s.d. of 0.93 Å , suggesting that the conformation of the RGS domain does not change upon embedding into the context of the entire RGS7-Gb5 complex ( Figure 4-figure supplement 2a).
An interesting feature of the RGS domain is its integration with the rest of the complex which occurs through its interactions with the 'bottom' surface of the Gb5 and GGL domain. The RGS-Gb5 interface involves two loops (Ra4a5 and Ra6a7) and 3 a-helices (a4, a5, and a7) of RGS7 and 4 loops (S4b1b2, S4b3b4, S5b1b2, and S5b3b4) of Gb5. They form an extensive network of hydrogen bonds along with hydrophobic/van der Waals interactions that bury 1316 Å of solvent accessible surface area ( . Comparison of the RGS domain-Gb5 interface in RGS7 and the RGS9 dimer structures shows many similarities as well as prominent differences. Most interestingly, electrostatic interactions formed by R375 of a5 and D421 of a7 in RGS7 with E183 and R227 in Gb5 contribute to a more extensive interface stronger compared to RGS9 (Figure 4f and Figure 4-figure supplement 2b). The unique rearrangement of these electrostatic interactions in RGS7 likely results in locking the interface at that side thus limiting the flexibility of the RGS domain. Interestingly, the Ra4a5 loop that forms many Gb5 are conserved (cyan); however, the RGS7 Ga1 a-helix does not form any contacts. The Ga1-Ga2 and Ga2 a-helices contribute to forming additional contacts (pink) that consist of two electrostatic interactions which are unique to RGS7. The RGS9-specific residues of Gb5 are shown in green color. (e) Cartoon representation of the contact interface of RGS domain (orange) with Gb5 (cyan) and GGL (blue) are shown. The a-helices of the RGS domain are numbered as a1-a9. Loops are indicated by black arrows. (f) Schematic view of the comparison of the direct contact residues of the RGS domains of RGS7 and RGS9 with Gb5. Common contact residues of Gb5 are shown in cyan, RGS-specific residues in green, and RGS9 in olive. (g) The S5b1b2 loop of Gb5 (cyan and green for the two molecules in the asymmetric unit for the RGS7 complex and blue for the RGS9 complex) contacts different residues in the RGS domains of RGS7 (orange) and RGS9 (pink). (h) Additional contacts of the RGS7 GGL domain with the RGS domain contribute to a more extensive interface compared to RGS9. The residue differences in RGS9 which cannot form hydrogen bonds are shown in red. The online version of this article includes the following figure supplement(s) for figure 4:  direct contacts with Gb5 is not conserved in the R7 RGS family. A particularly interesting contact point in the RGS domain-Gb5 interface is formed by the S5b1b2 loop of Gb5. This loop is longer in Gb5 compared to other Gb subunits and shows considerable remodeling of interactions with residues in the RGS domain ( Figure 4g). It further adopts different conformations between two Gb5 molecules in the asymmetric unit, resulting in variable interactions with residues of the RGS domain. In particular, T208 in Gb5 toggles its contact with the RGS7 RGS domain on and off. The corresponding region in RGS9 engages a neighboring residue, E207, which is absent in RGS7. This organization suggests a possibility of this region serving as a flexible conformational switch for the remodeling of RGS domain by interactions with Gb5.
Finally, the RGS domain of RGS7 also forms interactions with the GGL domain in a manner distinct from that seen in the RGS9 complex structure. Notably, the GGL-RGS interface in RGS7 is built by a network of hydrogen bonds, a feature that is absent in the corresponding interface of RGS9 ( Figure 4h). Additionally, the connecting loop between the GGL and RGS domains is longer in RGS9 due to two additional amino acids, providing more flexibility for the rearrangement of the RGS domain ( Figure 4-figure supplement 2c). Collectively, interactions of the RGS domain with the Gb5 and GGL domains might lead to more intricate integration of RGS domain in the RGS7 complex as compared to RGS9.

Conformational dynamics of the RGS7-Gb5 complex
To obtain insights into the conformational dynamics of the RGS7-Gb5 complex, we first assessed the distribution of crystallographic temperature factors that reflect movement. We found that the temperature values to be at their lowest throughout the Gb5 structure indicating its conformational rigidity. The distribution varied across structural elements of RGS7 with modest scores in the RGS domains and higher thermal motions in the DEP-DHEX module and the GGL-linker interfacing with the RGS domain suggesting greater conformational flexibility of these regions (Figure 5a).
To experimentally validate our findings, we performed HDX-MS measuring the exchange of amide hydrogens with solvent deuterium in the RGS7-Gb5 complex in solution. The resultant data were plotted onto the crystal structure of the RGS7-Gb5 dimer complex as a heat map that reflects its conformational dynamics at the baseline (Figure 5b and Figure 5-figure supplement 1 for sequence coverage). We found that the HDX-MS measurements agreed well with the results of our temperature factor analysis while providing additional detail on the conformational flexibility within the RGS7-Gb5 complex. Rapid exchange with solvent was observed in the DEP-DHEX module indicating its high flexibility. Two loops, Da1Da2 (b-hairpin) and Ea1Ea2, showed the most rapid deuterium exchange consistent with the dynamic nature of these regions also captured by the alternate conformations of the two molecules in the asymmetric unit in the crystal (Figure 2-figure supplement 1b). The second prominent region with high flexibility included the DHEX-GGL linker as well as a1 and a2 a-helices of GGL that interface with the RGS domain. Flexibility was also noted for the N-terminus of Gb5. In contrast, both the RGS domain and the core b-propeller fold of Gb5 exhibited low exchange rates supporting the notion of their conformational rigidity.
Finally, to better understand intra-molecular transitions in RGS7-Gb5 complex we conducted molecular dynamics (MD) simulations and performed dynamical network analysis (Eargle and Luthey-Schulten, 2012;Sethi et al., 2009). This analysis partitioned the RGS7-Gb5 complex into various communities (Figure 5c, Figure 5-source datas 1 and 2), which represent structural 'minidomains' or 'substructures' that are highly intraconnected, loosely interconnected, and defined by shared dynamical features. The analysis also identified critical nodes, which are key interactions between the different communities that facilitate dynamical communication between the communities. It revealed that the DHEX and DEP domains each form a distinct community linked by several critical nodes that facilitate communication between the domains. The DEP domain community also includes several loop regions from the Gb5 b-propeller fold. Gb5, which is engulfed between the DHEX/DEP domains and the GGL-linker/RGS domain, was partitioned into four communities that split the b-propeller fold WD40 repeat into four communities of nearly equivalent size. The DEP domain and Gb5 loop community forms a hotspot of critical node interactions with the b-propeller core, indicating this region is important for allosteric dynamical communication between DEP domain and Gb5. The GGL-linker is part of one of the four Gb5 b-propeller communities that form critical nodes with the DEP domain and importantly a critical communication node was identified that links the GGL-linker to RGS domain. Thus, the dynamical network analysis revealed a potential To determine if the putative allosteric pathway linking the DHEX/DEP domains and the GGLlinker/RGS domains through Gb5 may be shared by other RGS family members, we performed MD simulations and the network community analysis using the previously reported crystal structure of the RGS9-Gb5 complex (Cheever et al., 2008). The computed network communities and critical nodes between communities for RGS9-Gb5 (Figure 5d, Figure 5-source datas 3 and 4) are similar to that observed for RGS7-Gb5. There are some small differences between the two sets of communities; for example the RGS7 DEP community (Figure 5c; displayed in yellow) slightly extends into the loop regions of Gb5, but in the RGS9 complex these loops are part of a core Gb5 community that slightly extends into the DEP domain (Figure 5d; displayed in pink). As such, residues comprising the critical nodes between the DEP domain and the loops connect the Gb5 b-propeller domain do not involve residues in the exact same position based on a structural alignment. However, the same surfaces defining the critical node connections observed in the RGS7-Gb5 analysis (Figure 5c) are conserved in the RGS9-Gb5 analysis (Figure 5d), including the DEP/Gb5 hotspot and Gb5/RGS domain hotspot, suggesting that the potential allosteric pathway may be conserved among other members of RGS family, at least within this analysis of the only two available crystal structures of RGS-Gb5 complexes.

Structural insights into modulation of the RGS7-Gb5 complex by R7BP
One of the key modulators of the RGS7-Gb5 complex function with critical implications for its physiological regulation is the small SNARE-like molecule R7BP. Therefore, we sought to determine how organization and dynamics of the RGS7-Gb5 complex is influenced by its association with R7BP. For this purpose, we identified the structural determinants in RGS7-Gb5 influenced by R7BP binding by differential HDX-MS. In these experiments, we compared the HDX rates and mapped differences in deuterium exchange between the RGS7-Gb5 dimer and the RGS7-Gb5-R7BP trimer (Figure 6a). Consolidated differential HDX-MS data were mapped onto our structure of the RGS7-Gb5 dimer and color coded according to the extent of protection that reflects decreased solvent accessibility typically observed upon ligand binding or deprotection associated with increases in protein flexibility (Figure 6b and Figure 6-figure supplement 1 for differential coverage mapping). This differential HDX-MS data revealed robust and highly focused patterns of stabilization induced by the R7BP association. Most prominently, protected regions included an entire DEP domain and adjacent parts of the DHEX (Figure 6b), consistent with their involvement in R7BP binding (Martemyanov et al., 2005;Anderson et al., 2007). Furthermore, more subtle changes were observed in the N-terminus of the Gb5, the DHEX-GGL linker, and two a-helices of the GGL domain involved in interactions with the RGS domain. Interestingly, the S7b1 sheet and a part of the S6b4-S7b1 loop of Gb5 that together form a 'conduit' linking the DEP-DHEX and the GGL domains on opposite sides of Gb5 were also significantly protected from deuterium exchange by R7BP binding, likely by indirect stabilization via a network of interacting regions. No significant changes in deuterium exchange were seen in the RGS domain and throughout most of the Gb5 core. Importantly, these regions of Gb5 and RGS7 showing protection from HDX upon binding R7BP comprise critical node regions identified in the molecular dynamics simulations via dynamical network analysis (Figure 5c), features which are Figure 5 continued RGS9-Gb5 reveals conserved communities or structural 'mini-domains' that have shared dynamical features (color coded using the same colors). And critical interaction nodes calculated between the various communities, which form dynamical signaling interfaces between communities (shown as spheres for Ca atoms, which are color coded by community linked by black lines). Red circles highlight critical nodes that form key hotspots for allosteric communication between the RGS DEP domain and Gb5; and Gb5 and the RGS GGL-linker and RGS domain. The online version of this article includes the following source data and figure supplement(s) for figure 5: Source data 1. RGS7-Gb5 communities. Source data 2. RGS7-Gb5 critical nodes. Source data 3. RGS9-Gb5 communities. Source data 4. RGS9-Gb5 critical nodes. shared with the RGS9-Gb5 complex (Figure 5d), validating the allosteric communication pathway between the DEP domain and GGL-linker/RGS domain through the Gb5 b-propeller core. Together, these findings suggest that binding to R7BP also induces long-distance conformational rearrangements of the RGS7 complex.

Discussion
The structure of the RGS7 complex and analysis of its conformational dynamics presented here provide critical insights into its mechanisms in regulating GPCR signaling and explains a wealth of biochemical observations. Overall, the structure supports a model of modular organization of the R7 RGS proteins highlighting distinct interfaces that integrate separate elements into a coordinated complex. Our RGS7 structure displays a distinct organization compared to the RGS9 structure.
One of the most striking features of the R7 RGS proteins is their constitutive recruitment of the outlier G-protein b subunit, Gb5, a central organizational hallmark of the complexes. The present structure shows a unique mode of Gb5 integration into the complex providing the first example of Gbg like complexes in which N-terminal a-helices of both Gb subunit and GGL are protruding away from b-propeller fold of Gb subunit. These features point to the importance of the RGS-Gb5 interactions in controlling conformational transitions in the complex. One of the key interfaces formed by Gb5 involves interaction of its top 'hot-spot' surface with the DEP-DHEX module, which forms a major regulatory hub of the molecule. The DEP domain is found in many signaling proteins and has been shown to be involved in membrane anchoring and protein-protein interactions with regulatory proteins (Consonni et al., 2014;Paclíková et al., 2017;Ballon et al., 2006) while the DHEX domain is a unique feature of the R7 family. In the R7 RGS proteins, both DEP and DHEX domains are indispensable for binding to membrane anchors R7BP and R9AP (Anderson et al., 2007). Our HDX data reveal that the DEP domain indeed forms a major binding site for R7BP that further engages the Gb5 and the GGL domains. These observations are in good agreement with site-directed mutagenesis data that indicate requirement of both DEP-DHEX and Gb5 for R7BP/R9AP binding (Masuho et al., 2011). Interestingly, organization of the DEP-DHEX module in the RGS7 complex shows marked differences from that seen in RGS9-Gb5 including a distinct conformation of the b-hairpin and Ea1Ea2 loops, the DHEX orientation relative to DEP and clustering of basic residues on the surface. These differences likely underlie selectivity and strength of the RGS interaction with membrane anchors restricting RGS7 to interact only with R7BP while increasing the affinity of RGS9 for R7BP and allowing it to additionally bind R9AP. One particularly interesting structural feature in this regard is the b-hairpin known for its involvement in protein-protein interactions sensitive to mutations that affect its conformation (Figure 2-figure supplement 2; Wong et al., 2000;Yu et al., 2010). Reminiscent of this, we find that the dynamic b-hairpin of RGS7 is stabilized upon R7BP binding, suggesting it role in binding and likely undergoing conformational rearrangements. Curiously, a pathogenic mutation (R44C) in the b-hairpin of RGS7 that causes melanoma reduces the stability of RGS7 and its catalytic activity, further supporting the importance of conformational transitions in this region in regulating RGS7 function (Qutob et al., 2018).
Our MD simulations, HDX data, and comparison of the RGS7 and RGS9 complexes structures suggest that the interaction of the DEP domain with Gb5 forms a critical hotspot for communication in the complex, which propagates the conformational signal to the GGL-linker and RGS domain. The DEP/Gb5 interface undergoes large changes in HDX upon R7BP binding that result in shielding constituent residues from solvent accessibility consistent with the stabilization of this region. These observations are consistent with the dynamic nature of the DEP-DHEX-Gb5 interface organization which was proposed to exist in 'open' and 'closed' states (Narayanan et al., 2007). There are several biological implications of this regulation that also explain noted differences between RGS9 and RGS7 complexes.
First, we believe it serves as a major determinant of proteolytic stability of the complexes, which is a major point for controlling their expression levels and resultant negative pressure on GPCR signaling. RGS9 is notably more susceptible to proteolytic degradation than RGS7 and relies to a greater extent on R7BP binding for stability (Anderson et al., 2009a). Accordingly, we find that the DEP-Gb5 interface of RGS7 is more extensive, and thus less dynamic, than that of RGS9. The mutations that weaken the DEP-Gb5 interface increase susceptibility of the complex to proteolytic degradation (Porter et al., 2010). Taken together, this suggests a model whereby R7BP acts to promote the 'closed' conformation, stabilizing the DEP-DHEX interactions with Gb5 to hide degradation determinants and increase proteolytic stability of the complex. Second, our model supports a role of the DEP-DHEX-Gb5 interface in regulating the association of RGS complexes with components of G-protein signaling cascades. The 'hot spot' of the G-protein b subunits is well known to engage in interactions with effectors, modulatory proteins, and Ga subunits. Interestingly, residues involved in effector contacts show high levels of conservation (80-90%) in Gb5; thus, it is possible that it serves a similar function in docking signaling molecules. While the exact structural determinants mediating this binding are not known, the occlusion of the 'hotspot' interface by the DEP-DHEX domain suggests that the recruitment of R7 RGS complexes to effectors via Gb5 could be impossible without re-arrangement of the DEP-DHEX module opening up this interface. Thus, modulation of this interface by R7BP/R9AP may play a role in controlling effector association. One notable exception is the Gb5-mediated association of the RGS7 complex with the GIRK channel, that is promoted by R7BP (Ostrovskaya et al., 2014;Zhou et al., 2012). Interestingly, based on the crystal structure of the Gb1-GIRK complex, the analogous surface in Gb5 possibly involved in GIRK interaction is not occluded by the DEP-DHEX module and RGS7/Gb5 binding to GIRK could be accommodated with minimal repositioning of cytoplasmic domain of GIRK channel (Figure 7a). Comparison of interfaces of RGS7 DEP-Gb5 and GIRK2-Gb1g2 further shows poor conservation of amino acids with only one shared residue (W107 in Gb5, W99 in Gb1) (Figure 3-figure  supplement 3 and Figure 3-figure supplement 1). This suggests the distinct binding mode of RGS7 and GIRK2 on Gb subunits and likely explains why binding of RGS7-Gb5 does not activate GIRK the way Gbg association does. The hotspot surface of Gb5 further features conservation of many residues which in canonical Gb subunits mediate binding to Ga (Figure 3-figure supplement 1). Indeed, Gb5 has been shown to be interacting with Ga subunits when complexed with Gg instead of RGS (Yoshikawa et al., 2000). Furthermore, RGS9-Gb5-R9AP complex co-purifies with tightly bound Ga-GDP from endogenous tissues (Hu and Wensel, 2002). However, the extensive shielding of the putative Ga binding interface by the DEP-DHEX module makes such association unlikely without 'uncapping' requiring conformational re-arrangement that would swing the DEP-DHEX module away from Gb5 (Cheever et al., 2008). Given our finding that R7BP stabilizes this interface in the RGS7-Gb5 complex, we suggest that such uncapping would further require the severing of the association with R7BP. Based on these considerations, our model suggests that at the ground state, the hotspot binding surface of Gb5 in the context of a stable complex with R7BP and RGS7 is inaccessible to Ga-GDP. Thus, association of R7 RGS proteins with Ga-GDP at the hotspot interface, if it occurs, likely requires extensive rearrangement of the complex possibly triggered by yet undiscovered events.
Another intriguing consideration involves the role of the DEP-DHEX-Gb5 interface in the association with GPCRs. One of the key roles of conventional Gb subunits is its interfacing with GPCRs and recent crystal structures reveal the mode of the Gb association with receptors (Liang et al., 2017;Liang et al., 2018;Zhang et al., 2017). The RGS7-Gb5 complex is also well documented to associate with several GPCRs including the M3 muscarinic receptor and the orphan receptors GPR158 and GPR179. This association is impacted by R7BP (Orlandi et al., 2012;Sandiford and Slepak, 2009) suggesting involvement of the Gb5 and DEP-DHEX interface. Interestingly, the amino acids of Gb subunit involved in GPCR contacts are also reasonably well preserved in Gb5 (3 out of 5 are conserved) and RGS7-Gb5 can be confidently docked to accommodate GPCR interactions (Figure 7b). However, as in the case of Ga-GDP binding to the Gb5, docking of RGS7-Gb5 onto GPCR based on interactions of the canonical Gbg complexes requires rearranging of the DEP-DHEX module away from the hotspot surface of Gb5. This fuels a tempting speculation that interactions with GPCR may in fact provide the conformational trigger for the re-arrangement in RGS7-Gb5-R7BP complex that leads to the opening of the interface between DEP-DHEX and the hotspot surface of Gb5 much in the same way GPCRs promote swinging of a-helical domain in Ga (Rasmussen et al., 2011).
Third, a major implication of dynamic interfacing of Gb5 with multiple domains of RGS7 concerns regulation of the catalytic activity on its Ga-GTP substrates. To better understand the possible mechanisms, we modeled the association of RGS7-Gb5 with activated Ga by superposing the structures of RGS7-Gb5 and the RGS9 RGS domain-Gat/i1 complex (PDB entry 1fqk; Figure 7-figure supplement 1a). The Ga subunit docks well and supports the canonical interaction of Ra3-a4, Ra5-a6, and Ra7-a8 of RGS7 directly with switch I and II residues of Ga observed in several many RGS-Ga complexes (Figure 7-figure supplement 1b,  c). These loops of the RGS domain move closer to switch regions to facilitate the stabilization of the Ga transition state and increases GTP hydrolysis (Slep et al., 2001). Comparison of Ga binding residues of RGS9 with RGS7 shows conservation of residues with minor variations (Figure 7-figure supplement 1d). However, the aB-aC residues of the helical domain sterically clash with residues of the GGL aÀhelices and the N-terminus of Gb5 (Figure 7-figure supplement 1a). These clashes are consistent with noted inhibitory effects of Gb5 on binding of the RGS domain to activated Ga (Levay et al., 1999) and suggest that recognition of Ga-GTP by RGS7-Gb5 could be improved by structural rearrangement of the RGS domain relative to Gb5. The HDX-MS analysis shows stabilization of the region (GGL a-helices and the N-terminus of Gb5) clashing with Ga residues (Figure 6b and Figure 7-figure supplement 1a) suggesting that conformational rearrangement of this region upon R7BP binding could alleviate this steric hindrance and improve Ga binding. When considered with our MD simulations, these data suggest that R7BP binding exerts its effects via remodeling contacts initiated at the top surface of Gb5 which are translated by the S7b1 blade of the WD6 b-propeller to the opposite side and impact the network of the RGS domain contacts with the Gb5 to allow changes in their relative orientations. Thus, Gb5 may serve as a conduit of the conformational changes triggered at the N-terminus of the molecule to impact the organization of the RGS domain at the C-terminus of the RGS7 molecule. This model could explain biochemical observations that the GAP activity of RGS7-Gb5 can be markedly enhanced upon association with R7BP (Masuho et al., 2013;Drenan et al., 2006). Although membrane recruitment of RGS7-Gb5 by itself plays a major role in augmenting the catalytic activity of RGS7-Gb5 (Muntean and Martemyanov, 2016), rearrangement of the RGS domain triggered by R7BP likely further contributes to fine tuning this process, possibly in coordination with GPCRs.
To obtain further insights, we modeled R7BP based on its homology with SNARE proteins (PDB entries 1fio, 2xhe, 1hs7). An iterative structural assembly simulation by the I-TASSER algorithm predicts R7BP to form a four-helix bundle (Figure 7-figure supplement 2). The resulting model was docked onto the RGS7-Gb5 structure guided by our HDX data (Figure 7c). We further docked the GaÀGTP onto the RGS domain guided by the structure of the RGS domain of RGS9 in complex with the Gat/i1 chimera (PDB entry 1fqk). The resulting model illustrates the possible orientation of the tetramer on the membrane constrained by the positively charged patch in DEP-DHEX, C-terminal palmitoylation on R7BP, and N-terminal lipidation of Ga (Figure 7c). Notably, the orientation of the complex involving R7BP on the membrane is also compatible with the predicted macromolecular assemblies involving the GPCR and the GIRK.
In summary, our study reveals the versatile nature underlying the organization of the macromolecular complex that exploits several interfaces and mechanisms for interaction with GPCR signaling Figure 7 continued GPCR-RGS7-Gb5 assembly. The RGS7-Gb5 dimer is docked onto GPCR similar to Gbg in a typical GPCR-Gabg complex (here we used calcitonin receptorheterotrimeric Gas protein complex structure; PDB entry 5uz7). The DEP-DHEX domain (shown as arrow) is repositioned as it capped the probable GPCR binding interface of the Gb5 subunit. (c) A model representing the tetrameric assembly complex of RGS7-Gb5-R7BP-GaGTP. The complex is oriented relative to membrane according to the Ga position and its interaction with the RGS domain as well as docked to R7BP near its binding site based on HDX analysis. Structural rearrangement is required to avoid clashes of the helical domain of Ga with RGS7 GGL and Gb5 (shown in arrow). The online version of this article includes the following figure supplement(s) for figure 7:  components undergoing conformational transitions that adjust its function allowing it to serve its critical needs in precise regulation of key neuro-modulatory reactions in the nervous system.

Expression and purification of the RGS7-Gb5 complex
The RGS7-Gb5 complex was co-expressed and purified as described previously (Martemyanov et al., 2005;Muntean et al., 2018) with modifications. Briefly, RGS7 was coexpressed with Gb5 in Sf9 insect cells via the baculovirus-mediated expression system. Insect cells containing the recombinant RGS7-Gb5 complex were lysed in buffer A (20 mM HEPES pH 8, 300 mM NaCl, 10 mM imidazole (pH 8), 5 mM b-mercaptoethanol, 1% (v/v) glycerol) containing EDTAfree complete protease inhibitor tablets (Roche) by sonication. The recombinant RGS7-Gb5 complex containing supernatant was clarified by centrifugation at 32,000 rpm for 30 min and loaded onto pre-equilibrated HisTALON Superflow Cartridge (Clontech Laboratories, Inc) with buffer A. The protein complex was eluted over a 250 mM imidazole gradient. The protein complex was analyzed by SDS-PAGE and fractions containing the RGS7-Gb5 complex were pooled, dialyzed, and loaded onto a MonoQ column and eluted over a 1 M NaCl gradient. The eluted complex was further purified by using a Hiload 26/60 Superdex 200 column (GE Healthcare), which was pre-equilibrated with buffer B (20 mM PIPES pH 6.5, 200 mM NaCl, 2 mM TCEP-HCl). Peak fractions were analyzed by SDS-PAGE, pooled, and concentrated to 18 mg/ml.

Crystallization of the complex
We performed extensive crystallization trials for the RGS7-Gb5 dimer complex using several commercially available crystallization screens (~1000 conditions) by vapor diffusion hanging drop at 20˚C and 4˚C. We obtained only poorly diffracting crystals that we further optimized at 4˚C and obtained crystals that diffracted X-rays beyond 2.0 Å Bragg spacings from a reservoir solution of 0.1M sodium malonate (pH 6.25) and 10% w/v Polyethylene glycol 3350 over 3-4 days. Crystals were harvested by flash-freezing directly into liquid nitrogen by including 30% ethylene glycol as a cryoprotectant.
X-ray structure determination, model building, and crystallographic refinement X-ray diffraction data were collected on beamline 22-ID of SER-CAT at the Advanced Photon Source at Argonne National Laboratory. All X-ray diffraction data were processed and integrated and scaled using XDS and AIMLESS as implemented in autoPROC package (Vonrhein et al., 2011). A 2.13 Å dataset was obtained. The space group was determined to be P2 1 (space group number 4). Molecular replacement was performed using the program Phaser in the CCP4 package (McCoy et al., 2007) with the RGS9-Gb5 structure (PDB entry 2pbi) as the search model. Two molecules in the asymmetric unit were identified. Iterative model-building was performed using COOT (Emsley and Cowtan, 2004) and refined by performing maximum likelihood as implemented in autoBUSTER (Bricogne, 2011) and PHENIX (Adams et al., 2010). The quality of the structure was assessed using MolProbity . Data processing and structure refinement statistics for isotropic (to 2.81 Å ) and anisotropic (to 2.13 Å ) are provided in Table 1. Structural figures and electrostatic potential were generated using PyMOL (https://pymol.org/2/). Accessible surface area (ASA) were obtained using 'Protein interfaces, surfaces and assemblies' service PISA at the European Bioinformatics Institute (http://www.ebi.ac.uk/pdbe/prot_int/pistart.html) (Krissinel and Henrick, 2007).

Expression and purification of R7BP
The coding sequence of Mus musculus R7BP (residues 47-231) was amplified by PCR. A glutathione S-transferase (GST) tag followed by PreScission Protease site is present at the N-terminus of R7BP. The construct was generated using the baculodirect system (Invitrogen) and expressed into the baculovirus expression system. Insect cells containing recombinant R7BP were harvested, resuspended in buffer C (20 mM Tris pH 8, 300 mM NaCl, 1 mM DTT, 1% (v/v) glycerol) having complete protease inhibitor tablets (Roche), disrupted by sonication and clarified by centrifugation at 32,000 rpm for 30 min. The cell lysate was loaded onto a pre-equilibrated GSTPrep FF 16/10 (GE Healthcare) with buffer C and eluted with 15 mM glutathione. The eluted protein was analyzed by SDS-PAGE and fractions containing R7BP were pooled, dialyzed, and treated by PreScission Protease to cleave off the GST tag. The cleaved tag was removed by further passing the protein over the GST column. The flow-through fraction containing R7BP was concentrated and further purified using a Hiload 26/60 Superdex 75 column (GE Healthcare), which was pre-equilibrated with buffer B. The purity of the protein was analyzed by SDS-PAGE, pooled, and concentrated to 10 mg/ml.

Formation of the RGS7-Gb5-R7BP trimer complex
The RGS7-Gb5 dimer complex and R7BP were mixed together at a 1:1.2 molar ratio and incubated on ice at up to 2 hr. The trimer was loaded onto a Hiload 26/60 Superdex 200 column (GE Healthcare) which was pre-equilibrated with buffer B. The peak fraction containing the trimer complex was analyzed by SDS-PAGE, pooled, and concentrated for crystallization.

Thermoflour/DSF analysis
A thermofluor shift assay was used to optimize the buffer compositions that could stabilize the complex for structural study. The stability is determined by the T m (melting temperature) of the protein.
The assay was carried out in a 96-well thin-wall PCR plate (BioRad), and the plates were sealed with Optical-Quality Sealing Tape (Bio-Rad). The standardization of the assay was performed using various concentrations of the ratio of the protein complex and the Sypro Orange dye to get ideal peaks to calculate T m . Buffers (pH range 4.0-9.0), NaCl concentration (0-500 mM), reducing agent (DTT and TCEP-HCl), cheating agent (EDTA), stabilizer (glycerol, NDSB-256) and other additives and salts (MgCl 2 , CaCl 2 , KCl) were screened. The total reaction volume used was 20 ml and the reactions were heated in a real-time PCR machine (Roche lightcycler 480) from 20˚C to 90˚C with an acquisition mode of continuous, and 10 acquisitions per degree C. The final protein concentration of protein used was 1.5 mM in a buffer containing 20 M HEPES pH 8.0, 300 mM NaCl, 1 mM DTT, and 3X Sypro Orange dye. We ran the protocol as per LightCycler 480 instrument instructions which is created for the SYPRO Orange format. We obtained reasonably shaped curves with a good amount of fluorescence for calculation of T m and used the T m calling analysis feature to obtain negative first derivative drawings of the curves. Optimized buffers for RGS7 complexes were used for crystallization trials and biophysical characterizations. We have observed that pH of buffer has greatest effect on stability of the complex followed by stronger reducing agent TCEP-HCl.
Size exclusion chromatography with multi-angle light scattering (SEC-

MALS)
The purified RGS7 trimer complex in optimized buffer was analyzed by SEC-MALS using an HPLC system (Agilent Technologies 1260 Infinity) equipped with a MALS system (Wyatt DAWN HELEOS II Ambient with Optilab TrEX HC differential refractive index detector). The SEC-MALS system was calibrated with bovine serum album prior to RGS7 complex. The gel filtration purified complex was loaded onto pre-equilibrated analytical superdex200 (GE Healthcare) with buffer B. An aliquot of 100 ml sample of RGS7 trimer complex at a concentration of 1.2 mg/ml was injected with a flow rate 0.5 ml/min. ASTRA (Wyatt) Windows-based software was used for data collection and analysis.

Hydrogen-deuterium exchange (HDX) detected by mass spectrometry (MS)
Differential HDX-MS experiments were conducted as previously described with a few modifications (Chalmers et al., 2006). Peptide Identification: Peptides were identified using tandem MS (MS/MS) with an Orbitrap mass spectrometer (Q Exactive, ThermoFisher). Product ion spectra were acquired in data-dependent mode with the top five most abundant ions selected for the product ion analysis per scan event. The MS/MS data files were submitted to Mascot (Matrix Science) for peptide identification. Peptides included in the HDX analysis peptide set had a MASCOT score greater than 20 and the MS/MS spectra were verified by manual inspection. The MASCOT search was repeated against a decoy (reverse) sequence and ambiguous identifications were ruled out and not included in the HDX peptide set.

HDX-MS analysis
5 ml of each protein complex (10 mM) was diluted into 20 ml D2O buffer (20 mM Tris-HCl, pH 7.4; 150 mM NaCl; 2 mM DTT) and incubated for various time points (0, 10, 60, 300, and 900 s) at 4˚C. The deuterium exchange was then slowed by mixing with 25 ml of cold (4˚C) 3 M urea and 1% trifluoroacetic acid. Quenched samples were immediately injected into the HDX platform. Upon injection, samples were passed through an immobilized pepsin column (2 mm Â 2 cm) at 200 ml minÀone and the digested peptides were captured on a 2 mm Â 1cm C8 trap column (Agilent) and desalted. Peptides were separated across a 2.1 mm Â 5cm C18 column (Hypersil Gold, ThermoFisher) with a linear gradient of 4-40% CH3CN and 0.3% formic acid, over 5 min. Sample handling, protein digestion and peptide separation were conducted at 4˚C. Mass spectrometric data were acquired using an Orbitrap mass spectrometer (Exactive, ThermoFisher). The intensity weighted mean m/z centroid value of each peptide envelope was calculated and subsequently converted into a percentage of deuterium incorporation. This was accomplished by determining the observed averages of the undeuterated and fully deuterated spectra using the conventional formula described elsewhere (Zhang and Smith, 1993). Corrections for back-exchange were made on the basis of an estimated 70% deuterium recovery, and accounting for the known 80% deuterium content of the deuterium exchange buffer.
Data Analysis: HDX analyses were performed in triplicate, with single preparations of each purified protein/complex. Statistical significance for the differential HDX data is determined by t test for each time point, and is integrated into the HDX Workbench software (Pascal et al., 2012).
Data Rendering: Deuterium uptake for each peptide is calculated as the average of % D for all on-exchange time points and the difference in average %D values between the apo and ligand bound samples is presented as a heat map with a color code given at the bottom of each figure (warm colors for deprotection and cool colors for protection). Peptides are colored by the software automatically to display significant differences, determined either by a > 5% difference (less or more protection) in average deuterium uptake between the two states, or by using the results of unpaired t-tests at each time point (p-value < 0.05 for any two time points or a p-value < 0.01 for any single time point). Peptides with non-significant changes between the two states are colored grey. The exchange at the first two residues for any given peptide is not colored.

Homology modeling of R7BP
The three-dimensional structure of mouse R7BP (residues 47-231) was generated using the automated I-TASSER web service (http://zhanglab.ccmb.med.umich.edu/I-TASSER). I-TASSER generates 3D atomic models from multiple threading alignments and iterative structural assembly simulations. It uses multiple PDBs and accuracy of the models was estimated using parameters such as C-score, TM-score, and RMSD (Roy et al., 2010;Yang and Zhang, 2015). The R7BP homology model is used for docking with RGS7-Gb5 dimer structure based on HDX-MS data and documented the involvement of both DEP-DHEX and Gb5 in the trimer complex formation.

Molecular dynamics simulations
The Modeller (Eswar et al., 2006) extension within UCSF Chimera (Pettersen et al., 2004) was used to fill in the missing DHEX-GGL linker (R201-P253) of the RGS7-Gb5 complex structure (chains A and C) to prepare input files for all-atom molecular simulations using AMBER 16. The RGS9-Gb5 complex structure (PDB: 2PBI) was used without additional modification. The resulting complexes were submitted to H ++ server (Gordon et al., 2005) to determine the protonation states of titratable residues at pH 7.4. AMBER names were assigned to different protonation states of histidine using pdb4amber provided in AmberTools 16. Tleap was used to generate topology and coordinate files using the ff14SB force field following a method provided in the AMBER tutorial (http:// ambermd.org/tutorials/basic/tutorial5/). The resulting structures were solvated in a truncated octahedral box of TIP3P water molecules with the 10 Å spacing between the protein and the boundary, and neutralized with K + and Clions added to 50 mM (the number of iones to add was calculated using a molarity.perl script provided by Prof. Thomas Cheatham III; http://archive.ambermd.org/ 200907/0332.html). The system was minimized and equilibrated in nine steps at 310 K with nonbonded cutoff of 8 Å . In the first step the heavy protein atoms were restrained by a spring constant of 5 kcal mol -1 Å -for 2000 steps, followed by 15 ps simulation under NVT conditions with shake, then two rounds of 2000 cycles of steepest descent minimization with 2 and 0.1 kcal mol À1 Årestraints were performed. After one round without restraints, three rounds of simulations with shake were conducted for five ps, 10 ps and 10 ps under NPT conditions and restraints of 1, 0.5 and 0.5 kcal mol À1 Å À2 on heavy atoms. Finally, an unrestrained NPT simulation was performed for 200 ps. Production runs were carried out with hydrogen mass repartitioned (Hopkins et al., 2015) parameter files to enable 4 fs time steps. Constant pressure replicate production runs were carried out with independent randomized starting velocities. Pressure was controlled with a Monte Carlo barostat and a pressure relaxation time (taup) of 2 ps. Temperature was kept constant at 310 K with Langevin dynamics utilizing a collision frequency (gamma_ln) of 3 ps À1 . The particle mesh ewald method was used to calculate non-bonded atom interactions with a cutoff (cut) of 8.0 Å . SHAKE (Ryckaert et al., 1977) and hydrogen mass repartitioning was used to allow longer time steps. Production simulations were run for 2 ms each for the RGS7-Gb5 and RGS9-Gb5 complexes, which were initially analyzed using cpptraj (Roe and Cheatham, 2013).
Dynamic network analysis was performed on the last half (1 ms) of the production simulations following methods described in the 'Dynamical Network Analysis' tutorial developed by the Luthey-Schlten Group (http://faculty.scs.illinois.edu/schulten/tutorials/), which is provided as a PDF file (http://faculty.scs.illinois.edu/schulten/tutorials/network/network_tutorial.pdf). Briefly, the analysis required installation of the following software: the molecular visualization program VMD (Humphrey et al., 1996) (http://www.ks.uiuc.edu/Research/vmd/), which includes the NetworkView (Eargle and Luthey-Schulten, 2012) extension; the molecular dynamics analysis program Carma (http://utopia.duth.gr/glykos/Carma.html) (Glykos, 2006); a program that concatenates DCD trajectory files into a single DCD file called CatDCD (http://www.ks.uiuc.edu/Development/MDTools/ catdcd/); and the program gncommunities (http://faculty.scs.illinois.edu/schulten/software/network-Tools/index.html), which generates network communities from the simulation trajectory using methods described previously by the authors of the program (Sethi et al., 2009;Alexander et al., 2010;Black Pyrkosz et al., 2010). AMBER trajectory and prmtop files from the RGS7-Gb5 and RGS9-Gb5 complex simulations were converted to DCD and PSF format, respectively, using the commands 'animate write dcd' and 'writepsf' within VMD, respectively. The dynamic network models were created using the 'networkSetup' command within VMD, which invokes CatDCD and Carma and takes as input PSF and DCD files, information about node selection (Ca atoms were specified), and restrictions that specify constraints on the network generation (notSameResidue and notNeighboringCAlpha were specified); otherwise, default parameters were used for the calculation, that is the contact cutoff distance (distanceCutoff, = 4.5 Å ) for at least 75% (requiredOccupancy = 0.75) of the MD trajectory. The resulting output file (contact.dat) was read into VMD using the NetworkView plugin to visualize the computed network. The dynamical network generated by the networkSetup command in VMD was used as input into the program gncommunities, which was invoked within VMD and subsequently read using the NetworkView plugin for initial analysis. The gncommunities program uses a Girvan-Newman algorithm (Girvan and Newman, 2002) to partition the dynamical network into subnetworks or 'communities', identifies critical nodes (i.e. residue contacts) that lie in the interface between two different communities, and determines a 'betweenness' score representing how important the contact is to the entire dynamical network. The resulting output file (communities.out) from the gncommunities calculation contains the list of residues in each community, the critical nodes between communities, and the betweenness score; this information was used to visualize the communities (color coded similarly for the RGS7-Gb5 and RGS9-Gb5 complexes) and display the computed critical nodes (we displayed nodes with the largest betweenness scores comprising the top two-thirds of all betweenness values computed, which represent the most critical connections) as spheres with black lines linking two Ca atoms in the PyMOL Molecular Graphics System (Version 2.0, Schrö dinger, LLC).