Cryo-EM structure of the rhodopsin-Gαi-βγ complex reveals binding of the rhodopsin C-terminal tail to the gβ subunit

One of the largest membrane protein families in eukaryotes are G protein-coupled receptors (GPCRs). GPCRs modulate cell physiology by activating diverse intracellular transducers, prominently heterotrimeric G proteins. The recent surge in structural data has expanded our understanding of GPCR-mediated signal transduction. However, many aspects, including the existence of transient interactions, remain elusive. We present the cryo-EM structure of the light-sensitive GPCR rhodopsin in complex with heterotrimeric Gi. Our density map reveals the receptor C-terminal tail bound to the Gβ subunit of the G protein, providing a structural foundation for the role of the C-terminal tail in GPCR signaling, and of Gβ as scaffold for recruiting Gα subunits and G protein-receptor kinases. By comparing available complexes, we found a small set of common anchoring points that are G protein-subtype specific. Taken together, our structure and analysis provide new structural basis for the molecular events of the GPCR signaling pathway.

Abstract 23 G protein-coupled receptors (GPCRs) are the largest class of integral membrane proteins and 24 represent key targets for pharmacological research. GPCRs modulate cell physiology by 25 engaging and activating a diversity of intracellular transducers, prominently heterotrimeric G 26 proteins, but also G protein-receptor kinases (GRKs) and arrestins. The recent surge in the 27 number of structures of GPCR-G protein complexes has expanded our understanding of G 28 protein recognition and GPCR-mediated signal transduction. However, many aspects of these 29 mechanisms, including the existence of transient interactions with transducers, have remained 30 elusive. 31 Here, we present the cryo-EM structure of the light-sensitive GPCR rhodopsin in complex with 32 heterotrimeric Gi. In contrast to all reported structures, our density map reveals the receptor 33 C-terminal tail bound to the Gβ subunit of the G protein heterotrimer. This observation provides 34 a structural foundation for the role of the C-terminal tail in GPCR signaling, and of Gβ as 35 scaffold for recruiting Gα subunits and GRKs. By comparing all available complex structures, 36 we found a small set of common anchoring points that are G protein-subtype specific. Taken 37 together, our structure and analysis provide new structural basis for the molecular events of 38 the GPCR signaling pathway .  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64 Introduction 70 G protein-coupled receptors (GPCRs) are the most diverse class of integral membrane 71 proteins with almost 800 members in humans. GPCRs are activated by a great diversity of 72 extracellular stimuli including photons, neurotransmitters, ions, proteins, and hormones 73 (Glukhova et al., 2018). Upon activation, GPCRs couple to intracellular transducers, including 74 four subtypes of G proteins (Gαs, Gαi/o, Gαq/11, Gα12/13) (Milligan & Kostenis, 2006), seven 75 subtypes of GPCR kinases (GRKs) (Gurevich, Tesmer, Mushegian, & Gurevich, 2012), and 76 four subtypes of arrestins (Smith & Rajagopal, 2016) (Figure 1A), among many other partners 77 (Magalhaes, Dunn, & Ferguson, 2012). While most GPCRs are promiscuous and can couple 78 to more than one G protein subtype (Flock et al., 2017), the molecular determinants of G 79 protein recognition are not yet fully understood. Understanding the molecular basis for G 80 protein coupling and selectivity could lead to the design of drugs that promote specific 81 signaling pathways and avoid unwanted side effects (Hauser, Attwood, Rask-Andersen, 82 Schiöth, & Gloriam, 2017). 83 The recent surge in the number of structures of GPCR-G protein complexes has greatly 84 expanded our understanding of G protein recognition and GPCR-mediated signal 85 transduction. Out of the 13 structures of GPCR-G protein complexes available, six contain a 86 Gi/o subtype: μ-opioid receptor bound to Gi (Koehl et (Goodman Jr et al., 1996). Despite this 107 biochemical evidence, a direct interaction between rhodopsin and Gβγ could not be observed 108 in the existing complex (Kang et al., 2018). 109 Here, we present the cryo-EM structure of bovine rhodopsin in complex with a heterotrimeric 110 Gi. Overall, our structure agrees well with a currently published structures (Kang et al., 2018;111 Tsai et al., 2018). Remarkably, the EM density map provides a first structural evidence for the 112 interaction between the C-tail of the receptor and the Gβ subunit. The density map also shows 113 that intracellular loops (ICL) 2 and 3 of rhodopsin are at contact distance to Gα. This prompted 114 us to perform a comparison of all available structures of GPCR-G protein complexes to 115 generate a comprehensive contact map of this region. We then extended this analysis to the 116 binding interface formed by the C-terminal helix α5 of Gα and found that only a few G protein 117 subtype-specific residues consistently bind to the receptors. These contacts are ubiquitous 118 anchoring points that may be also involved in the selective engagement and activation of G 119 proteins. 120 121

122
To obtain a rhodopsin-Gi complex suitable for structural studies, we expressed in HEK cells  Figure 1), as cryo-EM screening revealed 130 that the complex without Fab could not be refined to high resolution (Suppl. Figure 2, Suppl. 131 Table 1). During unsupervised 3D classification of the complex bound to Fab16 (Suppl. 132 Figure 3), we observed that the density corresponding to Gα is heterogenous (Suppl. Figure  133 4), particularly at flexible regions such as the α-helical (AH) domain. The AH domain was then 134 excluded by using a soft mask during refinement, resulting in a map with a nominal global 135 resolution of 4.38 Å (Suppl. Figure 3D, E). The EM map was used to build a model of the 136 complex (Suppl. Figure 5). 137

138
Architecture of the rhodopsin-G protein complex 139 The structure of rhodopsin-Gi-Fab16 reveals the features observed in previously reported 140 GPCR-G protein complexes (Figure 1B, C; Suppl. Figure 6A). In particular, our cryo-EM 141 structure is in excellent agreement with the crystal structure of the same rhodopsin mutant 142 bound to a mini-Go protein (Tsai et al., 2018), with a nearly identical orientation of the C-143 terminal α5 helix (Suppl. Figure 6B), which contacts transmembrane helices (TM) 2, 3, 5-7 144 and the TM7/H8 turn of the receptor (Suppl. Figure 5B, and Suppl. Table 2).

Interaction between the C-terminal tail of rhodopsin and Gβ 164
The EM map reveals a density on the Gβ subunit as continuation of H8 of the receptor ( Figure  165 2A), which corresponds to the C-tail of rhodopsin. This feature does not exist in any other 166 GPCR-G protein complex, including the human rhodopsin-Gi complex bound to Fab_G50 167 (Kang et al., 2018). We modeled half of the C-tail of rhodopsin (12 out of 25 residues; 324-168 335) into this density as a continuous stretched peptide with residues G324, P327 and G329 169 serving as flexible hinges ( Figure 2B). This allows us to compare the structure of this region 170 in three signaling states of the receptor: inactive (Okada et al., 2004), G protein-bound (this 171 structure), and arrestin-bound (Zhou et al., 2017) ( Figure 2B). In the inactive state, the C-tail 172 folds around the cytoplasmic side of rhodopsin, although this is likely due to crystal packing 173 as this region is intrinsically disordered in the absence of a binding partner (Jaakola, Prilusky, 174 Sussman, & Goldman, 2005; Venkatakrishnan et al., 2014). In our G protein complex, the C-175 tail stretches over the cleft between Gα and Gβ, interacting with both subunits ( Figure 2C). In 176 the rhodopsin-arrestin complex part of the proximal segment of the C-terminus (residues 325-177 329) is not resolved, but the distal part could be modeled up to S343, eight residues longer 178 than in our G protein complex and including two phosphorylated sites. In the presence of 179 arrestin, the C-tail stretches further, with residues K339-T342 forming a short β-strand 180 antiparallel to the N-terminal β-strand of arrestin ( Figure 2B). Thus, it appears that distinct 181 portions of the C-tail are responsible for contacting different intracellular partners. The central 182 part

Comparison of the GPCR-G protein binding interface 201
As in the other existing GPCR-G protein complex structures, the C-terminal α5 helix of Gα 202 forms the major contact interface to rhodopsin. We aligned the structures of available complexes using the Cα atoms of H5.11-26. This 207 alignment reduces apparent differences in the binding poses and provides the "viewpoint of 208 the G protein" (Figure 3A). We compiled an exhaustive list of the residue-residue contacts 209 between the receptor and the α5 helix in all the available GPCR/G protein complexes, and 210 observed that the main contacts to the α5 helix are formed by TM5 and TM6, followed by TM3 211 and TM7/8 (Figure 3B, Suppl. GPCRs, and with Lys 5.64 in class B GPCRs ( Figure 3E, Suppl. Table 3). We suggest that a 218 tight interaction between Arg H5.17 and TM5 might be one of the main determinants for the Gs-219 specific relocation of TM6. 220 Besides the canonical interaction with the α5 helix, our complex shows that ICL2 and ICL3 of 221 the receptor are at contact distance to Gi (Suppl. Figure 7A). In all analyzed structures, we 222 found that ICL2 lies near αN/β1 and β2/β3 of Gα, while ICL3 is close to α4/β6 (Suppl. Figure  223 7 and Suppl. Figure 8). Interestingly, ICL2 folds into an α-helical structure in all the class A 224 receptors except rhodopsin (Suppl. Figure 9). 225 ICL2 does not contribute to binding Gα in the structures of 5HT1B-mini-Go and A1AR-Gi. 226 Nevertheless, the contact between Gα and ICL2 seems to discriminate between class A 227 GPCRs -which interact via the αN/β1 loop-and class B GPCRs -which instead the region 228 around β1 and β2/β3 (Suppl. Fig. 8  In this work, we present the cryo-EM structure of the signaling complex between bovine 265 rhodopsin and a Gi protein heterotrimer, stabilized using an antibody Fab fragment (Fab16) 266 . Overall, this structure agrees very well with existing complexes. In 267 particular, we found that the binding mode of the G protein Ras domain -including the key C-268 terminal α5 helix-is virtually identical to our previously reported X-ray structure of the same 269 receptor bound to a mini-Go protein (Tsai et al., 2018) (Suppl. Fig. 6B). 270 However, our EM map shows a density on the Gβ subunit that extends from H8 of the receptor, 271 constituting the proximal segment of the C-terminus (residues G324 to T335) (Figure 2). One 272 explanation of why the C-tail is observed in our density map and not in other structures may 273 rely on the nature of the components used to reconstitute the reported GPCR complexes 274 (Suppl.  Fig. 7A). In particular, we found that ICL2 is at contact 317 distance to the αN helix, the αN/β1 and the β2/β3 Gα in most structures (Suppl. Fig. 7, Suppl. 318 Fig. 8, and Suppl. Fig. 9). While ICL2 contributes to the binding interface, there are no 319 apparent conserved contacts among complexes (Suppl. Fig. 8) human Gαi subunit (Gαi1) with an N-terminal TEV protease-cleavable deca-histidine tag was 343 expressed and purified as described . The transducin heterotrimer was 344 isolated from the rod outer segment of bovine retina and Gβ1γ1 was separated from Gαt with 345 Blue Sepharose 6 Fast Flow (GE Healthcare) as described (Maeda et al., 2014). The 346 Gαi1β1γ1 heterotrimer (Gi) was prepared by mixing equimolar amounts of Gαi1 with or without 347 10xHis-tag and Gβ1γ1 and incubated at 4°C for 1 hour shortly before use for rhodopsin-Gi 348 complex formation on the 1D4 immunoaffinity column. 349 350 Fab16 production 351 The monoclonal mouse antibody IgG16 was generated as described . 352 Large scale production of IgG16 was performed using adherent hybridoma cell culture grown 353 in DMEM medium supplemented with 10% ultra-low IgG fetal bovine serum (FBS) (Gibco, 354 #16250078) and 25 U/ml of mouse interleukin-6 (Invitrogen, #PMC0064) at 37°C and 5% CO2. 355 Antibody expression was increased by stepwise dilution of FBS concentration down to 2%. 356 After incubation for 10-14 days, ~500-ml cell suspension containing the secreted IgG was 357 clarified by centrifugation and subsequent filtration through a 0.45 µm HAWP membrane 358 (Merck Millipore). The filtrate was mixed with an equal volume of binding buffer (20 mM 359 Na2HPO4/NaH2PO4, pH 7.0) and loaded to a 1-ml HiTrap Protein G Sepharose FF column 360 (GE Healthcare). The column was washed with binding buffer until the UV280 absorbance 361 dropped to a stable baseline, and IgG was eluted with 0.1 M glycine-HCl (pH 2.7). Fractions 362 were immediately neutralized with 1 M Tris-HCl (pH 9.6). Fractions containing IgG16 were 363 combined and dialyzed against 20 mM Na2HPO4/NaH2PO4 (pH 7.0), 1.5 mM NaN3 using a 364 slide-A-lyzer dialysis cassette (12-14 kDa MWCO, Thermo Scientific) at 4 °C for 15 h. The 365 dialysate was collected and mixed with the immobilized papain resin (0.05 ml resin for 1 mg 366 IgG) (Thermo Scientific, #20341). Papain was activated by the addition of L-cysteine and 367 EDTA to a final concentration of 20 mM and 10 mM respectively. IgG was digested overnight 368 at 37°C with gentle mixing. Afterwards the immobilized papain resin was removed and the 369 digested fraction was mixed with Protein A Sepharose (0.2 ml resin for 1 mg digested IgG, GE 370 Healthcare, #17078001) for 1 hour at RT. Resins were washed with two column volumes (CV) 371 of wash buffer 10 mM Tris (pH 7.5), 2.5 M NaCl. The flow-through and washing fractions 372 containing Fab16 were collected and dialyzed against PBS supplemented with 1.5 mM NaN3 373 using a slide-A-lyzer dialysis cassette (12-14 kDa MWCO) at 4 °C. The initial models of rhodopsin and the Ras-like domain of Gαi protein were adapted from the 466 structure of rhodopsin-mini-Go complex (PDB id: 6FUF). The initial model of Gβγ was obtained 467 from the crystal structure of guanosine 5'-diphosphate-bound transducin (PDB id: 1GOT). The 468 models were docked into the 3D map as rigid bodies in Chimera (Pettersen et al., 2004). The 469 coordinates of the structure were manually adjusted and the C-terminus of rhodopsin was built 470 in Coot (Emsley & Cowtan, 2004), which was used to visualize the EM map and the models. 471 The models were refined using the phenix.real_space_refine in the Phenix suite (Adams et  472 al