key: cord-0766031-i8jkkk5h authors: Toelzer, Christine; Gupta, Kapil; Yadav, Sathish K.N.; Hodgson, Lorna; Williamson, Maia Kavanagh; Buzas, Dora; Borucu, Ufuk; Powers, Kyle; Stenner, Richard; Vasileiou, Kate; Garzoni, Frederic; Fitzgerald, Daniel; Payré, Christine; Lambeau, Gérard; Davidson, Andrew D.; Verkade, Paul; Frank, Martin; Berger, Imre; Schaffitzel, Christiane title: The Free Fatty Acid-Binding Pocket is a Conserved Hallmark in Pathogenic β-Coronavirus Spike Proteins from SARS-CoV to Omicron date: 2022-04-22 journal: bioRxiv DOI: 10.1101/2022.04.22.489083 sha: a9d5d83e3ec6bb637c78c0d9699701a3bd192927 doc_id: 766031 cord_uid: i8jkkk5h As COVID-19 persists, severe acquired respiratory syndrome coronavirus-2 (SARS-CoV-2) Variants of Concern (VOCs) emerge, accumulating spike (S) glycoprotein mutations. S receptor-binding domain (RBD) comprises a free fatty acid (FFA)-binding pocket. FFA-binding stabilizes a locked S conformation, interfering with virus infectivity. We provide evidence that the pocket is conserved in pathogenic β-coronaviruses (β-CoVs) infecting humans. SARS-CoV, MERS-CoV, SARS-CoV-2 and VOCs bind the essential FFA linoleic acid (LA), while binding is abolished by one mutation in common cold-causing HCoV-HKU1. In the SARS-CoV S structure, LA stabilizes the locked conformation while the open, infectious conformation is LA-free. Electron tomography of SARS-CoV-2 infected cells reveals that LA-treatment inhibits viral replication, resulting in fewer, deformed virions. Our results establish FFA-binding as a hallmark of pathogenic β-CoV infection and replication, highlighting potential antiviral strategies. One-Sentence Summary Free fatty acid-binding is conserved in pathogenic β-coronavirus S proteins and suppresses viral infection and replication. The trimeric spike glycoprotein (S) decorates the surface of coronaviruses and mediates entry into host cells. S is the major antigen recognized by neutralizing antibodies and the main 15 target for vaccine development (3). SARS-CoV S and SARS-CoV-2 S both bind to human angiotensin-converting enzyme 2 (ACE2) receptor on the host cell surface (4) (5) (6) , MERS-CoV S binds to dipeptidyl-peptidase-4 (DPP4) (5, 7) while the HCoV-HKU1 and HCoV-OC43 S proteins bind to the N-acetyl-9-O-acetylneuraminic acid receptor (8). S is cleaved by host cell proteases into the receptor-binding fragment S1 and the partially buried fusion fragment S2 (4). 20 S1 is comprised of the N-terminal domain (NTD), the RBD with a receptor-binding motif (RBM), and two C-terminal domains (CTDs). S2 mediates fusion of the viral envelope with host cell membranes and is comprised of the fusion peptide, heptad repeats, transmembrane domain and cytoplasmic C-terminus (9) . In the prefusion conformation, the RBDs in the S trimer can alternate between closed ('down') and open ('up') conformations. SARS-CoV and SARS-CoV-2 25 S require RBD 'up' conformations for interaction with ACE2 (6, 9, 10) for cell entry. In our previous SARS-CoV-2 S structure, we discovered a free fatty acid (FFA) bound to a hydrophobic pocket in the RBD (11) . Mass spectroscopy identified this ligand as LA, an essential omega-6 poly-unsaturated fatty acid (PUFA) the human body cannot synthesize (11, 12) . LA-bound S is stabilized in a compact, locked conformation which is incompatible with 30 ACE2 receptor binding (11) . In immunofluorescence assays, synthetic mini-virus particles 4 decorated with LA-bound S showed reduced docking to ACE2-expressing host cells as compared to mini-virus with LA-free S (13) , confirming that LA interferes with receptor binding. S protein sequence alignments suggest conservation of the hydrophobic pocket in the RBDs of SARS-CoV, SARS-CoV-2, MERS-CoV, hCoV-OC43 and hCoV-HKU1 (11) , indicating that the pocket may be a hallmark shared by all human β-CoVs. Intriguingly, all SARS-CoV-2 VOCs stringently 5 maintain this pocket, notably including omicron which accumulated a wide range of mutations in S elsewhere, suggesting that the pocket provides a selective advantage. Here, we investigate if LA-binding, and the functional consequences of LA-binding, is conserved in S glycoproteins of pathogenic SARS-CoV, MERS-CoV, HCoV-HKU1 as well as the SARS-CoV-2 VOCs alpha, beta, gamma, delta and omicron. We demonstrate that the 10 respective RBDs all comprise a hydrophobic pocket capable of binding LA, except common cold-causing HCoV-HKU1. However, we also demonstrate that a single amino acid substitution of a residue lining the entrance of the hydrophobic pocket in the HCoV-HKU1 RBD is sufficient to restore LA-binding. We analyze SARS-CoV S by cryo-EM showing that LA-bound SARS-CoV S adopts a hitherto elusive locked structure akin to LA-bound locked SARS-CoV-2 S (11), Using correlative light-electron microscopy (CLEM) followed by electron tomography of SARS- 20 CoV-2 infected cells, we provide evidence that LA, beyond counteracting infection at the S protein level, also interferes with viral replication inside infected cells. This likely occurs through inhibition of cytoplasmic phospholipase A2 (cPLA2), a key enzyme implicated in viral replication via formation of intracellular replication compartments (14) and in the cytokine storm causing systemic inflammation in COVID-19 (15-17). 25 LA-binding to S RBD can be analyzed by surface plasmon resonance (SPR). We previously determined a binding constant of ~41 nM for LA to SARS-CoV-2 S RBD (11) . To corroborate our hypothesis that a functional hydrophobic pocket is evolutionarily conserved in β-CoV RBDs, we tested if RBDs from other β-CoVs also are capable of LA-binding ( Fig. 1, table S1 ). Based on sequence alignments, the RBDs of SARS-CoV, MERS-CoV, SARS-CoV-2 and 30 VOCs alpha, beta, gamma, delta and omicron all maintain the hydrophobic pocket, at least since 5 the emergence of SARS-CoV in 2002 ( Fig. 1A-C) . In SPR experiments, LA bound with a high affinity (~72 nM) to immobilized SARS-CoV RBD (Fig. 1D ), comparable to SARS-CoV-2 (11). Moreover, we observed a slow off-rate in agreement with tight LA-binding. LA bound with ~96 nM affinity to MERS-CoV RBD (Fig. 1E ). In contrast, the RBD of HCoV-HKU1 S did not bind LA despite high sequence similarity (Fig. 1A,F) . HCoV-HKU1 S comprises a bulky glutamate 5 E375 located directly in front of the hydrophobic pocket (18) obstructing the pocket entrance ( Fig 1F) . We mutated HCoV-HKU1 E375 to alanine and restored LA-binding, albeit with a reduced affinity of ~178 nM (Fig. 1F ). This indicates that the pocket function, while structurally conserved, may have been abrogated in HVoC-HKU1, a -CoV which causes mild disease. The RBDs of SARS-CoV-2 VOCs bound LA with affinities between 50 nM and 87 nM, virtually 10 identical to SARS-CoV-2, confirming that LA-binding is conserved (Fig. 1G) and not affected by the mutations in S which cluster away from the pocket (Fig. 1H ). Taken together, we confirmed full conservation of LA-binding in highly pathogenic β-CoV S proteins, but not in S of HCoV-HKU1. Interestingly, HCoV-OC43 which likewise causes common cold, appears to also comprise a hydrophobic pocket (Fig. 1A) , as seen in an earlier HCoV-OC43 S cryo-EM 15 structure which displays unassigned density in the RBD (fig. S1) (19) . It remains unclear what exactly this unassigned density corresponds to, which appears too small to accommodate the C18 hydrocarbon chain of LA ( fig. S1 ). To elucidate LA-binding by SARS-CoV that emerged 2002, we determined the S cryo-EM structure. The S ectodomain was produced as a secreted trimer using MultiBac (20) 20 identically as described for SARS-CoV-2 S (fig. S2) (11) . As before, we did not supplement LA during expression or subsequent sample purification and preparation steps. Cryo-EM data Next, we scrutinized LA-binding to S RBDs of β-CoVs by extensive MD simulations 25 (Fig. 3) . As proof-of-principle, unbiased and spontaneous LA-binding to the SARS-CoV-2 RBD was simulated. Using the distance of the -carbon atoms of residues N370 (gating helix) and (11)), the system was equilibrated for 30 ns, and then LA was pushed out of the pocket by applying a small force, which is indicated by a sharp increase of the distance of LA from the center of the pocket (blue curve, Fig. 3B ). Interestingly, without LA inside the pocket the distance D_pocket fluctuates between open and closed state (grey curve 5 changes from 15 Å to 10 Å distance; Fig. 3B ) until LA (randomly) approaches the entrance after about 600 ns, binds back to the pocket and stabilizes the open pocket (Fig.3B , movie S1). It should be noted, that during the MD simulation, LA-rebinding to the pocket was not immediate. Instead, LA transiently interacted with residues on the surface of the SARS-CoV-2 RBD (simulation time from 40 ns to 600 ns in Fig. 3B , movie S1). We identified hotspots of LA 10 interactions on the RBD surface which include the RBM and residue N343 which is glycosylated and located close to the pocket entrance (Fig. 3A) . However, spontaneous (re)binding of LA to the hydrophobic pocket was observed once LA approached the pocket entrance (Fig. 3B, fig. S5C ), and LA subsequently remained stably bound in the pocket. Three different outcomes emerged in our MD simulations of LA (re)binding to the 15 SARS-CoV RBD (Fig. 3C) . i) LA did not (re)enter the pocket during a 1 s simulation but interacted closely with the RBM over a significant time period (A in Fig. 3C) ; ii) LA rebound to the pocket after removal (B in Fig. 3C , movie S2), and iii) LA entered the pocket but could seemingly dissociate again (C in Fig. 3C ). Additional simulations show that LA-binding to the RBD is dynamic because the contacts formed between LA and the residues lining the pocket 20 vary over time and between experiments, indicating diverse binding modes ( fig. S6A-D) . However, when LA-binding was analyzed for the complete SARS-CoV S in the MD simulations, LA was stably bound to the three pockets formed by adjacent RBDs within the S trimer with minimal dynamics (fig. S6E,F) . After validating the simulation method with experimentally derived LA-bound RBD structures (Fig. 3A-C) , we applied the same MD simulation protocol to 25 analyze LA-binding to MERS-CoV RBD (from PDB ID 6q05). Spontaneous binding of LA to the pocket of MERS-CoV RBD was observed in 10 out of 11 independent simulations (Fig. 3D,E, fig. S7A ). Further analyses suggested a prevalent LA-binding mode where LA does not entirely enter the hydrophobic pocket ( fig. S7B ), while demonstrating significant dynamics of the portion of LA within the pocket similar to SARS-CoV ( fig. S7C,D) . Notably, LA-binding 30 and pocket opening occurred simultaneously in the MERS-CoV RBD suggesting that LA can 8 bind to the entrance of the closed pocket and pry/force the gate open (Fig. 3D, movie S3) . As a control, we analyzed LA-binding to the HCoV-HKU1 RBD pocket: LA was found to transiently interact with hydrophobic residues at the pocket entrance of HCoV-HKU1 RBD but it did not enter the pocket in our simulations (Fig. 3F) , reproducing our LA-binding SPR experiments (Fig. 1C ). In contrast, tight LA-binding to the RBD pocket of SARS-CoV-2 VOC omicron was 5 observed, consistent with the SPR data (Fig. 3G, movie S4) . In order to evaluate the impact of LA-treatment on SARS-CoV-2 infected cells, we Viral replication and cell viability were monitored by brightfield and fluorescence microscopy. 10 We used CLEM followed by electron tomography to analyse SARS-CoV-2 infected Caco-2-ACE2 cells 35 hrs after LA-supplementation (Fig. 4 , figs. S8-S10). Despite analysing only strong GFP-expressing cells, we detected significantly more virions in infected cells that were not LA-treated (~25 virions per micrograph) than in infected cells that had been treated with LA after infection (~9 virions per micrograph) ( Fig. 4A-C, fig. S10 ). LA treatment also resulted in 15 the emergence of lipid droplets in the cytoplasm of cells which appear dark in the EM micrographs (Fig. 4B, fig. S9 ). These droplets appear independent of whether the cells were SARS-CoV-2 infected or uninfected (Fig. 4B , figs. S9,S10), and occur in many cells including adipocytes. Moreover, we notice significant membrane remodeling in SARS-CoV-2 infected Caco-2-ACE2 cells compared to non-infected cells, as reported previously for β-CoV infections 20 (22, 23) (Fig. 4, fig. S10 ). In addition to their reduced number (Fig. 4C) , virions in SARS-CoV-2 infected, LAtreated cells appeared irregular in size and shape as compared to the virions in untreated cells which adopt a characteristic regular spherical form (Fig. 4A,B, fig. S10 , movies S5, S6). Closer analysis of virions derived from LA-treated cells confirmed a statistically significant increase in 25 size and deformation (ellipticity) (Fig. 4C) . The average diameter of virions from untreated cells (n=427) was calculated as 86 nm (SD=13 nm), in agreement with previous reports (14) . In contrast, virions derived from LA-treated cells (n=319) had a larger average diameter of 96 nm (SD=15 nm). Consistently, the average area of virions increased from 4756 nm 2 to 5614 nm 2 (Fig. 4C) . Moreover, the average ellipticity of virions increased from 0.55 (SD=0.14) in 30 untreated cells to 0.61 (SD=0.15) in LA-treated cells, confirming deformation. Taken together, 9 we observed that LA treatment after infection leads to a lower viral load in SARS-CoV-2 infected cells with the virions being larger in size and deformed as compared to untreated, infected cells (Fig. 4A-C) indicating that their integrity, and potentially their infectivity, may be compromised. During early stages of coronavirus infection, phospholipase A2s (PLA2s) are activated as Previously, it was shown that cPLA2 is tightly regulated by PUFAs, including LA, which are potent competitive inhibitors of the enzyme (27) . We find that cPLA2 inhibition is half-maximal in the presence of ~100 µM LA in vitro (Fig. 4D) . We note that the S proteins of pathogenic β- 15 CoVs have orders of magnitude higher affinity for LA (Fig. 1) . It is thus likely that, by utilizing (Fig. 2) . We observe a high correlation between LA-binding in the pocket and the locked conformation which is a noninfectious form of S, incompatible with ACE2 receptor binding (Fig. 2B) . MD simulations 25 consistently show transient interactions of LA with hydrophobic patches on the surface of the RBDs before binding the pocket, suggesting a conserved mechanism of LA approaching the pocket entrance (Fig. 3) 30 scenarios in the different simulations, but an opening of the pocket while interacting with LA (induced fit) occurred more frequently (Fig. 3B,D) . Previous studies showed that LA renders the virus less infectious by stabilizing a locked form of S inhibiting receptor-binding (11), virion attachment and entry into cells (13) . Here we show that LA treatment of infected cells significantly reduces the production of virions, and the 5 few virions produced are markedly deformed (Fig. 4A-C, fig. S10 ). These modes of action will likely synergize to significantly reduce, or even abrogate, viral infectivity and transmission. Therefore, FFA-binding by S can be conceived as an 'Achilles heel' of pathogenic β-CoVs, making this feature a highly attractive target for LA-or LA-mimetic based antiviral interventions against SARS-CoV-2. It is noteworthy that fatty acids have been used prolifically as excipients 10 in unrelated medications with established safety in multiple administration routes (nasal, pulmonary, oral and intra-venous) (28) . The evolutionary conservation of the FFA-pocket implies significant selection advantage for the virus itself. We can conceive of several such advantages. For instance, the LA-bound S form is more stable than the open S forms (29) . LA stabilizes the S protein in a prefusion 15 conformation which is particularly useful when the protein is still in the host cell, preventing premature proteolytic cleavage by host proteases. Moreover, LA-bound, locked S buries key epitopes of the RBM and of core RBD parts, potentially hiding these from attack by neutralizing antibodies (1, 13) . We speculate that LA levels can be sensed by the virus, allowing the S protein to switch from the 'stealth' locked form to the infectious open form that mediates ACE2-binding 20 and cell entry. Lipid metabolome remodeling is a central element of viral infection (25, 30, 31) . LA levels are markedly perturbed during COVID-19 disease progression, with serum levels significantly decreased in COVID-19 patients (32) . Conversely, intracellular levels of LA are elevated (25) . This correlates with β-CoV-induced membrane remodeling to generate new membrane compartments for viral replication (22, 23) . cPLA2 activation is a central mediator of 25 lipidome remodeling in β-CoVs, and of various additional +RNA virus families, and is therefore a validated target for broad-spectrum antiviral drug development (24) . We propose that during β-CoV infection when cPLA2 is activated, the virus can circumvent feedback inhibition of cPLA2 by sequestering LA (27) (Fig. 4D) . This would keep cPLA2 in a hyperactivated state. In this model, inhibition of cPLA2 by supplementing excess LA will downregulate membrane 30 remodeling, thus interfering with a mechanism required for viral replication. Indeed, we 11 demonstrate that supplementing excess LA to SARS-CoV-2 infected cells interferes with viral replication (Fig. 4) . Previous reports show that LA also interferes with MERS-CoV replication (25) . In conclusion, our results convey that the conserved FFA-S interaction, while affording selective advantages to the virus, renders it vulnerable to antiviral intervention exploiting this highly conserved feature. This could be achieved by supplementing LA or a related molecule, 5 ideally during early stages of infection when the virus resides in the upper respiratory tract where it can be conveniently targeted. This could be accomplished by delivering FFA formulations, e.g. via a nasal or inhaled spray, to suppress viral replication and spreading within a patient, concomitantly reducing transmission (33) and protecting others from infection. Figs. S1 to S10 Tables S1 to S3 The biological and clinical significance of emerging SARS-CoV-2 variants Addendum: A dynamic nomenclature proposal for SARS-CoV-2 lineages to assist genomic epidemiology The spike protein of SARS-CoV--a target for vaccine and therapeutic development SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor Functional assessment of cell entry and receptor usage for 10 SARS-CoV-2 and other lineage B betacoronaviruses Structure, Function, and Antigenicity of the SARS-CoV-2 Spike Glycoprotein MERS-CoV spike protein: a key target for antivirals Human coronaviruses OC43 and HKU1 bind to 9-O-acetylated sialic acids via a conserved receptor-binding site in spike protein domain A Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation Cryo-EM structures of MERS-CoV and SARS-CoV spike glycoproteins reveal the dynamic receptor binding domains Free fatty acid binding pocket in the locked structure of SARS-CoV-2 spike protein Structural insights in cell-type specific evolution of intra-host diversity by 25 SARS-CoV-2 Synthetic virions reveal fatty acid-coupled adaptive immunogenicity of SARS-CoV-2 spike glycoprotein SARS-CoV-2 structure and replication characterized by in situ cryo-electron tomography High levels of eicosanoids and docosanoids in the lungs of intubated COVID-19 patients Group IIA secreted phospholipase A2 is associated with the pathobiology leading to COVID-19 mortality Large-Scale Plasma Analysis Revealed New Mechanisms and Molecules Associated with the Host Response to SARS-CoV-2 Crystal structure of the receptor binding domain of the spike glycoprotein of 5 human betacoronavirus HKU1 Structural basis for human coronavirus attachment to sialic acid receptors Protein complex expression by using multigene baculoviral vectors Cryo-EM structure of the SARS coronavirus spike glycoprotein in complex with its host cell receptor ACE2 SARS-coronavirus replication is supported by a reticulovesicular network of modified endoplasmic reticulum Correlative multi-scale cryo-imaging unveils SARS-CoV-2 assembly and egress Inhibition of Cytosolic Phospholipase A2alpha Impairs an Early Step of Coronavirus Replication in Cell Culture Cells: Implications for Lipid Metabolism Remodeling upon Coronavirus Replication What do secreted phospholipases A2 have to offer in combat against different viruses up to SARS-CoV-2? Inhibition of human platelet phospholipase A2 activity by 25 unsaturated fatty acids Fatty acids as therapeutic auxiliaries for oral and parenteral formulations The SARS-CoV-2 spike protein: balancing stability and infectivity Stealing the Keys to the Kitchen: Viral Manipulation of the Host Cell Metabolic Network Dissecting lipid metabolism alterations in SARS-CoV-2 Proteomic and Metabolomic Characterization of COVID-19 Patient Sera The effect of omega-3 fatty acid supplementation on clinical and biochemical parameters of critically ill patients with COVID-19: a randomized clinical trial ConSurf 2016: an improved methodology to estimate and visualize 10 evolutionary conservation in macromolecules Site-specific biotinylation of purified proteins using MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy Gctf: Real-time CTF determination and correction RELION: implementation of a Bayesian approach to cryo-EM structure determination Visualizing density maps with UCSF Chimera Namdinator -automatic molecular dynamics flexible fitting of structural models into cryo-EM and crystallography experimental maps Features and development of Coot Automated map sharpening by maximization of detail and connectivity Electronic Ligand Builder and Optimization Workbench (eLBOW): a tool for ligand coordinate and restraint generation MolProbity: all-atom structure validation for macromolecular crystallography EMRinger: side chain-directed model and map validation for 3D cryoelectron microscopy YASARA View -molecular graphics for all devices -from smartphones to workstations Structures of MERS-CoV spike glycoprotein in complex with sialoside 10 attachment receptors Crystal structure of the receptor-binding domain from newly emerged Middle East respiratory syndrome coronavirus New ways to boost molecular dynamics simulations Quantitative Characterization of the Binding and Unbinding of Millimolar Drug Fragments with Molecular Dynamics Simulations VMD: visual molecular dynamics Rapid reconstruction of SARS-CoV-2 using a synthetic genomics platform Morphometry of SARS-CoV and SARS-CoV-2 particles in ultrathin plastic sections of infected Vero cell cultures Computer visualization of three-25 dimensional image data using IMOD Identification of phosphorylation sites of human 85-kDa cytosolic phospholipase A2 expressed in insect cells and present in human monocytes Preparation of the Full Set of Recombinant Mouse-and Human-30 Secreted Phospholipases A2 SARS-CoV-2 and bat RaTG13 spike glycoprotein structures inform on virus evolution and furin-cleavage effects A potent SARS-CoV-2 neutralising nanobody shows therapeutic efficacy in the Syrian golden hamster model of COVID-19 Movies S1 to S6 SARS-CoV-2 and Variants of Concern (VOCs) alpha, beta, gamma, delta and omicron. Residues lining the FFA-pocket are boxed in grey. Residues in the 5 hydrophilic lid that interacts with the polar head group of bound FFA are highlighted in green ) LA-binding to RBDs from all five VOCs analyzed by SPR. KD values are indicated. LA concentrations ranging from 4M to 10M were used for all SPR experiments, 10 except for HCoV-HKU1 wildtype where only 10M was tested. (H) Mutations found in the VOCs were mapped on the 3D structure of SARS-CoV-2 omicron RBD. Residue numbers and mutations are indicated. The FFA-binding A) Cryo-EM density of SARS-CoV S trimer in the locked (left) and open (right) conformation in a front and top view. Monomers are shown in cyan, blue and magenta, respectively. Magenta arrows point to the RBD in the up conformation. (B) Top view of the 5 locked S conformation in a cartoon representation The red boxes highlight the composite LA-binding pocket formed by adjacent S RBDs. Left: the acidic LA headgroup interacts with R395 and Q396. Tube-shaped EM density is shown as mesh. Right: EM density is shown as mesh for selected hydrophobic residues lining the LA-binding pocket PDB ID 6ACC (21)). (D) Overlay of S RBD trimers in the open conformation (cyan, blue, magenta) and the closed conformation without LA ligand (grey) (PDB ID 6ACC (21)). (E) Top view of the open conformation with one LAbinding pocket boxed in red. Right: Close-up view of the unoccupied hydrophobic pocket. EM 10 density is shown for selected hydrophobic residues lining the entrance of the pocket (mesh). (F) Overlay of one subunit of LA-bound and unbound S RBD from the locked (blue) and open (grey) conformation, respectively. Red arrows show the movement of the gating helix and of Y352 and Y356 upon LA-binding. (G) Side view of the locked conformation Y819 (cyan) from the neighboring subunit. (H) Side view of the open conformation. The black boxes highlight the region around R1021, forming the center of an H-bond cluster, cation-π interactions with F1024 and a salt bridge to E1013. Right: Short-range interactions formed by R1021 of S subunits have 3-fold symmetry in the locked conformation We thank all members of the Berger and Schaffitzel laboratories for their contributions and suggestions, and David Veesler (University of Washington, Seattle) and Roman Laskowski (EMBL-EBI) for helpful discussions. We acknowledge support and assistance by the Wolfson 10