key: cord-327711-welf0eb1
authors: Zhou, Daming; Duyvesteyn, Helen ME; Chen, Cheng-Pin; Huang, Chung-Guei; Chen, Ting-Hua; Shih, Shin-Ru; Lin, Yi-Chun; Cheng, Chien-Yu; Cheng, Shu-Hsing; Huang, Yhu-Chering; Lin, Tzou-Yien; Ma, Che; Huo, Jiandong; Carrique, Loic; Malinauskas, Tomas; Ruza, Reinis R; Shah, Pranav NM; Tan, Tiong Kit; Rijal, Pramila; Donat, Robert F.; Godwin, Kerry; Buttigieg, Karen; Tree, Julia; Radecke, Julika; Paterson, Neil G; Supasa, Piyasa; Mongkolsapaya, Juthathip; Screaton, Gavin R; Carroll, Miles W.; Jaramillo, Javier G.; Knight, Michael; James, William; Owens, Raymond J; Naismith, James H.; Townsend, Alain; Fry, Elizabeth E; Zhao, Yuguang; Ren, Jingshan; Stuart, David I; Huang, Kuan-Ying A.
title: Structural basis for the neutralization of SARS-CoV-2 by an antibody from a convalescent patient
date: 2020-06-13
journal: bioRxiv
DOI: 10.1101/2020.06.12.148387
sha: 
doc_id: 327711
cord_uid: welf0eb1

The COVID-19 pandemic has had unprecedented health and economic impact, but currently there are no approved therapies. We have isolated an antibody, EY6A, from a late-stage COVID-19 patient and show it neutralises SARS-CoV-2 and cross-reacts with SARS-CoV-1. EY6A Fab binds tightly (KD of 2 nM) the receptor binding domain (RBD) of the viral Spike glycoprotein and a 2.6Å crystal structure of an RBD/EY6A Fab complex identifies the highly conserved epitope, away from the ACE2 receptor binding site. Residues of this epitope are key to stabilising the pre-fusion Spike. Cryo-EM analyses of the pre-fusion Spike incubated with EY6A Fab reveal a complex of the intact trimer with three Fabs bound and two further multimeric forms comprising destabilized Spike attached to Fab. EY6A binds what is probably a major neutralising epitope, making it a candidate therapeutic for COVID-19.

conversion to the post-fusion form where the S2 subunit engages the host membrane whilst dispensing with S1 2,3 .

Neutralising human monoclonal antibodies that recognise the ACE2 receptor binding site for SARS-CoV-1 and SARS-CoV-2 are generally not cross-reactive between the two viruses and are susceptible to escape mutation 6-10 (indeed a natural mutation Y495N has already been identified at this site (GISAID 11 : Accession ID: EPI_ISL_429783 Wienecke-Baldacchino et al.)). In contrast CR3022 (derived from a SARS-CoV-1 patient) cross-reacts strongly with SARS-CoV-2 (Methods, Fig. 1 ) and has been shown to recognise a cryptic, conserved epitope on the RBD distinct from the binding epitope of ACE2 7, [12] [13] [14] . That this is not uncommon for SARS-CoV-1 antibodies is suggested by similar observations for 47D11 15 .

To isolate SARS-CoV-2 Spike-reactive monoclonal antibodies, we cloned antibody genes from blood-derived plasmablasts of a COVID-19 patient in the convalescent phase. One of these, EY6A was shown by ELISA to bind S1 of SARS-CoV-2 and cross react with SARS-CoV-1 (Fig. 1) . Binding of EY6A to SARS-CoV-2-infected cells was detected by immunofluorescence ( Fig. 1 ). Surface plasmon resonance (SPR) measurements for EY6A Fab showed high affinity binding to immobilised SARS-CoV-2 RBD (KD = 2 nM, although the value for immobilised EY6A IgG was somewhat higher) as derived from the kinetic data (Methods, Extended Data Fig. 1 , Extended Data Table 1 ). SPR studies showed that there was some interdependence of EY6A and CR3022 binding, which varied depending on which component was immobilised on to the sensor chip; ACE2 blocking assays confirmed a somewhat asymmetric blocking effect (Extended Data Fig. 2 ). With RBD stably expressed on 5 MDCK-SIAT1 cells (MDCK-RBD), EY6A did not block binding of ACE2 to the RBD, whereas with ACE2 stably expressed on MDCK-SIAT1 cells (MDCK-ACE2) EY6A blocked the interaction of RBD with ACE2. In this assay, EY6A exhibits around 7 times stronger ACE2 blocking than CR3022 13 (EY6A, IC50 = 54 nM; CR3022, IC50 = 347 nM) and has equivalent ACE2 inhibition compared to ACE2-Fc (IC50 = 54 nM) and VHH72-Fc (IC50 = 69 nM) 2 . These observations are suggestive of an indirect effect by EY6A once bound to the RBD, consistent with an allosteric or weak direct interaction. This was supported by an SPR competition assay with immobilised CR3022, which binds distant from the ACE2 binding site (Extended Data Fig. 2 ) 12 . This showed complete competition with EY6A for RBD binding suggesting they recognise the same or overlapping epitopes, and indicated that EY6A binds the SARS-CoV-2 RBD more tightly.

Two independent neutralisation tests, both using live wild type SARS-CoV-2 showed strong neutralisation. A neutralisation test for EY6A based on quantitative PCR detection of virus in the supernatant bathing infected Vero E6 cells after 5 days of culture, showed a ~1000-fold reduction in virus signal (Methods, Extended Data Fig. 3 ) indicating that it is highly neutralising. This was further corroborated by a plaque reduction neutralisation test (PRNT) at PHE Porton Down (Methods and Extended Data Table 2 ) using SARS-CoV-2 virus and EY6A which gave an ND50 of ~10.8 µg/mL (70 nM) (calculated according to Grist 16 ) . A separate PRNT implementation at Oxford gave a slightly higher ND50 of ~30 µg/mL, consistent with a shorter incubation time of antibody with virus at lower temperature (Extended Data Fig. 4 ).

To elucidate the epitope of EY6A, we determined the crystal structures of the deglycosylated SARS-CoV-2 RBD in complex with EY6A Fab alone and in a ternary complex incorporating a nanobody (Nb) which has been shown to compete with ACE2 (for data on a closely related Nb see Huo 2020, submitted), as a crystallisation chaperone. The crystals of the binary complex diffracted to 3.8 Å resolution (Methods, Extended Data Table 3 ) and those of the ternary complex to 2.65 Å. The interaction between EY6A and the RBD was identical in both complexes (Extended Data Fig. 5 ). The higher resolution ternary complex, which showed that there was no interaction between EY6A and the Nb, permitted a full interpretation of the detailed interactions (Figs. 2 and 3) and has been refined to give an Rwork/R-free of 0.216/0.262 and good stereochemistry (Methods, Extended Data Table 3 ).

Residues 333-527 of the RBD, 1-136 and 141-224 of the heavy chain and 1-215 of the light chain of EY6A and 2-126 of the Nb are well defined Fig. 2a ,b. The Nb recognises an epitope adjacent to and slightly overlapping the ACE2 receptor binding site and binds the RBD orthogonally to EY6A (Fig. 2b,c) . EY6A binds essentially the same epitope as CR3022 12, 13 but with a different pose corresponding to a rotation of 73° about an axis perpendicular to the RBD α3 helix (central to both epitopes) (Fig. 2d,e) . The Fab complex interface buries 564 and 361 Å 2 of surface area for the CDRs of the heavy and light chains respectively. The EY6A interaction is mediated by the CDR loops H1, H2, H3, L1 and L3 contacting predominantly α3 but also α2 and the β2-α3, α4-α5 and α5-β4 loops of the RBD (Fig. 3 and Extended Data Fig. 6 ). A total of 16 residues from the heavy chain and 11 from the light chain participate in the interface together with 31 residues from the RBD. For the heavy chain these form potentially 6 hydrogen bonds and a single salt bridge between D99 (of H3) and K386 of the RBD and the light chain interface residues contribute an additional 6 hydrogen bonds. Hydrophobic interactions further increase the binding affinity (Fig. 3) . Of the 31 residues involved in the interaction 21 are conserved between the CR3022 and EY6A epitopes ( Fig. 3 and Extended Data Fig. 6 ). Conformational changes are introduced into the RBD by binding to EY6A at the α2 (residues 365-371) and α3 (residues 384-388) helices (Extended Data Fig. 6 ), similar to those seen for the CR3022 complex 12 . Comparison of the epitope residues for EY6A, CR3022 12 and VHH-72 17 shows that there is a very substantial overlap (Extended Data Fig. 6 ), although the bulk of the molecules extend in different directions, such that VHH-72 directly blocks ACE-2 binding 17 .

In the first pre-fusion Spike structures (PDB IDs: 6VSB 2 , 6VXX, 6VYB 4 ), where residues 986 and 987 in the linker between two helices in S2 were mutated to a Pro-Pro sequence to prevent the conversion to the post-fusion helical conformation, the RBDs were found in either one 'up' two 'down' or all three 'down' configuration, and in both cases the epitope is inaccessible. In the 'down' position it is packed against another RBD of the trimer and the Nterminal domain (NTD) of the neighbouring protomer. A recent publication for the wild type Spike identifies a more closed form 18 where the S1 portion of the Spike is tightened up. The structure is not yet deposited however, and so we have looked at the role of the epitope in the down rather than fully closed form, which will be broadly similar. Here the EY6A epitope packs down tightly against the S2 'knuckle' bearing the Pro-Pro mutations, forming a buried protein-protein interface and making the epitope completely inaccessible. We assume that in the closed form this interaction will be even tighter and is probably responsible for maintaining the Spike in the pre-fusion state. Even when the RBD is in the 'up' configuration, the epitope remains largely inaccessible and substantial further movement of the RBD would be required to permit interaction unless more than one RBD was in the up conformation 12 .

To investigate how the Fab insinuates itself into the Spike, we performed cryo-EM analysis.

Spike ectodomain was mixed with a 6-fold molar excess of EY6A Fab and incubated at room temperature (21 °C) with an aliquot taken at 5 hours, applied to cryo-EM grids and frozen (Methods). Unbiased 2D class averages revealed three major particle classes with over onethird comprising a trimeric Spike/EY6A complex (some of which are self-associated) (Methods, Extended Data Table 4 and 121°. In addition, the orientations of the Vh domains relative to their associated RBDs differ slightly from that of the crystal structure (by 5°, 2° and 7°, respectively). The quality of the density suggests that these likely samples selected from a continuous distribution (Extended Data Fig. 10 ).

The majority of the remaining particles form either a roughly 2-fold symmetric structure or a triangular association (Methods, Extended Data Table 4 and Figs. 7-11). Reconstructions of these particles were anisotropic due to a preferential orientation of the particles on the grid which was somewhat mitigated by collecting data with 30° tilt to yield reconstructions at 4.4 Å and 4.7 Å, respectively, in the plane of the grid but significantly worse resolution perpendicular to the grid (Extended Data Fig. 10 ). The reconstructions were sufficiently clear to allow the unambiguous fitting of EY6A-RBD complexes (Extended Data Fig. 11 ), however the density for what we assume are the N-terminal domains is poor in both reconstructions and we did not attempt to fit a model. These structures likely represent a residual well-structured fragment from the unfolding of the pre-fusion state of the Spike (SDS PAGE analysis shows that the Spike polypeptide remains largely uncleaved, Extended Data Fig. 12 ). The 'dimeric' and 'trimeric' structures are formed by different lateral associations and these also differ from that seen for similarly structurally degraded Spike-CR3022 complexes 12 (Extended Data Fig. 11 ).

Convalescent serum has shown promise in patients severely ill with COVID-19 19, 20 , thus immunotherapeutics have potential for treating COVID-19 even at a relatively late stage in the disease. To this end, it is desirable to find a combination of antibodies that neutralise the virus by different mechanisms to mitigate against immune evasion and antibody dependent enhancement. One neutralisation mechanism is blocking receptor attachment. We propose that attachment at the EY6A epitope is a further major neutralisation mechanism. In support of this, the epitope recognised by EY6A has been reported for several antibodies 12, 13, 21, 22 and nanobodies 17,23 raised against SARS-CoV-2, SARS-CoV-1 and MERS. For SARS-CoV-1, CR3022 has also been shown to neutralise synergistically with ACE2 blocking antibodies 7 .

Despite the spatial separation of the EY6A and ACE2 epitopes we find some cross-talk between the two binding events.

The EY6A epitope is extremely unusual, since it is completely inaccessible in the pre-fusion Spike trimer. This raises the question of what the mechanism of neutralisation might be. In the pre-fusion state the EY6A/CR3022 epitope rests down upon the upper end of the helixturn-helix between heptad repeat 1 (HR1) and the central helix (CH) of S2, essentially putting a lid on the spring-loaded extension of the helix which occurs on conversion to the postfusion state in the vicinity of the mutations designed to prevent conversion between the preand post-fusion conformation 24 (Fig. 5 ). The residues of the epitope are crucial to these protein-protein interactions, and therefore highly conserved, explaining why it has, to date, proved impossible to generate mutations that escape binding of the antibody 7,12 . EY6A binding to the isolated RBD is tight (at ~2 nM it is roughly an order of magnitude tighter than CR3022) and, remarkably, the binding pose on top of the Spike allows three Fabs to bind simultaneously around the central axis (whereas CR3022 Fab cannot be accommodated). In line with this, a major portion of Spike molecules incubated for 5 h with EY6A are still in the intact pre-fusion state, with only about 1/3 being converted. Simple modelling suggests that a similar packing could occur for intact antibodies (Extended Data Fig. 13 ). In general, we would expect binders at this epitope to neutralise by displacing the 'lid' on the HR1/CH turn, reducing the stability of the pre-fusion state and therefore reducing the barrier to conversion to the more stable post-fusion trimer. This conversion is hindered in the construct we have used by the presence of the proline mutations at the turn between the helices. Premature conversion would prevent later attachment to the cell and block infectivity. The kinetics of this process will determine the effectiveness of the antibody in neutralisation and ultimately protection. Since the RBD is a relatively small domain there might also be an interplay between separate epitopes, thus we saw allosteric effects between EY6A and ACE2 binding and similarly VHH-72, which binds an overlapping epitope to EY6A, strongly inhibits ACE-2 binding by virtue of its different angle of attack 17 . The reason for the cross-talk between

This study was designed to isolate SARS-CoV-2 antigen-specific human MAbs from peripheral plasmablasts in humans with natural SARS-CoV-2 infection, to characterize the antigenic specificity and phenotypic activity of SARS-CoV-2 Spike-reactive MAb, and to determine the structure of antibody in complex with viral antigen. 

Fresh peripheral blood mononuclear cells (PBMCs) were separated from whole blood by density gradient centrifugation and cryopreserved PBMCs were thawed. PBMCs were stained with a mix of fluorescent-labelled antibodies to cellular surface markers, including anti-CD3 (BD Biosciences, USA), anti-CD19 (BD Biosciences, USA), anti-CD27 (BD Biosciences, USA), anti-CD20 (BD Biosciences, USA), anti-CD38 (BD Biosciences, USA), anti-IgG (BD Biosciences, USA) and anti-IgM (BD Biosciences, USA). Plasmablasts were selected by gating on CD3-CD20-CD19+CD27hiCD38hiIgG+IgM-events and were isolated in chamber as single cell as previously described 26 . Sorted single cells were used to produce human IgG monoclonal antibodies as previously described 26 . Expression vectors that carry variable domains of heavy and light chains were transfected into the 293T cell line for expression of recombinant full-length human IgG monoclonal antibodies in serum-free transfection medium.

To determine the individual gene segments employed by VDJ and VJ rearrangements and the number of nucleotide mutations and amino acid replacements, the variable domain sequences were aligned with germline gene segments using the international ImMunoGeneTics (IMGT) alignment tool (http://www.imgt.org/IMGT_vquest/input).

EY6A IgG used for neutralisation and making Fab: Antibody was expressed using Nanobody: This was derived from a naïve library followed by affinity maturation as described Deglycosylation of RBD:10 µL of Endoglycosidase F1 (~1mg/mL) was added to RBD (~2 mg/mL, 3 mL) and incubated at room temperature for two hours. RBD was then loaded to a Superdex 75 HiLoad 16/60 gel filtration column (GE Healthcare) for further purification using buffer 10 mM HEPES pH 7.4, 150 mM NaCl. Purified RBD was concentrated using a 10 kDa ultrafiltration tube (Amicon) to 12 mg/mL.

The 

Neutralization activity of monoclonal antibody-containing supernatant was measured using a SARS-CoV-2 (strain CDC-4) infection of Vero E6 cells 27 . Briefly, Vero E6 cells were preseeded in a 96 well plate at a concentration of 2 x 10 4 cells per well. On the following day, monoclonal antibody-containing supernatant were mixed with an equal volume of 100 TCID50 virus preparation and incubated at 37 °C for 1 hour. The mixture was added into seeded Vero E6 cells and incubated at 37 °C for 5 days. The cell control, virus control, and virus back-titration were setup for each experiment. At day 5, the culture supernatant was harvested from each well and the viral RNA was extracted by the automatic LabTurbo system (Taigen, Taiwan) following the manufacturer's instructions. for the most part, except that the specimen was pretreated with Proteinase K prior to RNA extraction. Real-time reverse transcription polymerase chain reaction was performed in a 25-µL reaction containing 5 µL of RNA 28 

SARS-CoV-2 (Australia/VIC01/2020) 29 

Plaque reduction neutralization tests were performed using passage 4 of SARS-CoV-2 Victoria/01/2020 29 . Virus suspension at appropriate concentrations in Dulbecco's Modification of Eagle's Medium containing 1 % FBS (D1; 100 µL) was mixed antibody (100 µL) diluted in D1 at a final concentration of 50 µg/mL, 25 µg/mL, 12.5 ug/mL or 6.125 µg/mL, in triplicate, in wells of a 24 well tissue culture plate, and incubated at room temperature for 30 minutes. Thereafter, 0.5 mL of a single cell suspension of Vero E6 cells in D1 at 5 x 105/mL was added, and incubated for 2 h at 37 o C before being overlain with 0.5 mL of D1 supplemented with carboxymethyl cellulose (1.5 %). Cultures were incubated for a further 4 days at 37 o C before plaques were revealed by staining the cell monolayers with amido black in acetic acid/methanol.

Purified and deglycosylated RBD and EY6A Fab were combined in an approximate molar ratio of 1:1 at a concentration of 6.5 mg/mL. Nb was also combined with EY6A-6His Fab and RBD in a 1:1:1 molar ratio with a final concentration of 5.7 mg/mL. These two complexes were separately incubated at room temperature for one hour. Initial screening of crystals was performed in Crystalquick 96-well X plates (Greiner Bio-One) with a Cartesian Robot using the nanolitre sitting-drop vapour diffusion method as previously described 31, 32 . Crystals were soaked in a solution containing 25% glycerol and 75% reservoir solution for a few seconds and then mounted in loops and frozen in liquid nitrogen prior to data collection. Diffraction data were collected at 100 K at beamline I03 of Diamond Light Source, UK.

Diffraction images of 0.1° rotation were recorded on an Eiger2 XE 16M detector with an exposure time of 0.01 s per frame, beam size 80×20 µm and 100% beam transmission. Data were indexed, integrated and scaled with the automated data processing program Xia2-dials 33, 34 . The data set for the binary complex of 360° was collected from a single frozen crystal to 3.8 Å resolution with 20-fold redundancy. The crystal belongs to space group P3121 with unit cell dimensions a = b = 166.6 Å and c = 270.8 Å. The structure was determined by molecular replacement with PHASER 35 using search models of antibody CR3022 Fab and the RBD of the RBD/CR3022 Fab complex (PDB ID 6YLA; 12 ). There are three RBD/EY6A complexes in the crystal asymmetric unit, resulting in a crystal solvent content of ~75%. For the ternary complex, a data set of 360° rotation with data extending to 2.6 Å was collected on beamline I03 of Diamond with exposure time 0.008 s per 0.1° frame, beam size 80×20 µm and 100% beam transmission). The crystal also belongs to space group R3 but with unit cell dimensions (a = b = 178.1 Å and c = 87.8 Å). There is one RBD/EY6A/Nb complex in the asymmetric unit and a solvent content of ~61%.

One cycle of REFMAC5 36 was used to refine atomic coordinates after manual correction in COOT 37 to the protein sequence from the search model. For both the binary and ternary complexes the final refinement used PHENIX 38 resulting in Rwork = 0.219 and Rfree = 0.259 for all data to 3.8 Å resolution for the binary complex and to Rwork = 0.216 and Rfree = 0.262 for all data to 2.64 Å resolution for the ternary complex. There is well ordered density for a single glycan at the glycosylation site N343 in the RBD.

Data collection and structure refinement statistics are given in Extended Data Table 3 .

Structural comparisons used SHP 39 , residues forming the RBD/Fab interface were identified with PISA 40 , figures were prepared with PyMOL (The PyMOL Molecular Graphics System, Version 1.2r3pre, Schrödinger, LLC).

Spike protein, following SEC purification, was buffer exchanged into 2 mM Tris pH 8.0, 200 mM NaCl, 0.02 % NaN3 buffer using a desalting column (Zeba, Thermo Fisher). A final concentration of 0.18 mg/mL was incubated with EY6A Fab (in the same buffer) in a 6:1 molar ratio (Fab to trimeric Spike) at room temperature for 5 hrs. Control grids of Spike alone after incubation at room temperature for 5 hrs were also prepared.

Each grid was prepared using 3 µL sample applied to a freshly glow-discharged on high for 20 Grids were screened on a Titan Krios microscope using SerialEM operating at 300 kV (Thermo Fisher). Movies were collected on a K3 detector on a Titan Krios operating at 300 kV in super resolution mode, with a calibrated Super Resolution pixel size of 0.415 A/pix at both 0° and 30° tilt. To compensate for the poorer contrast with tilted data, it was necessary to use a higher dose rate for the latter dataset.

Alignment and motion correction was performed using Relion3.1's implementation of motion correction 41 , with a 5-by-5 patch-based alignment. All frames were binned by two, resulting in a final calibrated pixel size of 0.83 Å/pixel. Contrast-transfer-function (CTF) of full-dose and non-weighted micrographs was estimated within a CryoSPARC wrapper for Gctf-v1.06 42 . Images were then manually inspected and those with poor CTF-fits were discarded.

Particles were then picked by unbiased blob picking in CryoSPARC v.2.14.1 43 and subjected to rounds of 2D classification.

For the Spike-EY6A dataset (structure A), 2,096,246 Spike-like particles were used to make a template to pick particles from the untilted dataset, which were then filtered by 2D classification to 110,096 particles and then further refined by 3D classification with an ab initio model set. For the 30 ° dataset, 124,194 particles were used as a template, and filtered by 2D classification to a set of 84,230 particles and then, as before, further refined by unbiased 3D classification. The two particle sets were then refined together, with a final set of 144,680 particles.

For B and C (triangular ring and 'dimeric' form), particles from both the zero and 30° datasets were combined in a similar manner to the Spike-EY6A dataset using the 'Exposure Group Utilities' module in CryoSPARC. Both particle sets (B, 41372 particles and C, 119,343 particles) were then reclassified and the best class refined with non-uniform refinement. For B, C3 symmetry was imposed at this final refinement stage, resulting in an appreciable improvement in resolution, as indicated by inspection and gold-standard FSC = 0.143 (4.7 versus 5.9 Å, see Extended Data Table 4 ).

The EM density of Spike/EY6A was fitted with the structure of a closed form of Spike (PDB ID 6VXX) apart from the RBDs and EY6A Fab which were fitted with RBD/EY6A of the ternary crystal structure using COOT 37 . Due to the lower resolution, RBD and EY6A are only fitted to the 'dimeric' and 'trimeric' EM density. The Spike/EY6A structure was refined with PHENIX 38 real space refinement, first as a rigid body and then by positional and Bfactor refinements. Only rigid body refinement was applied to the 'dimeric' and the 'trimeric'

complexes. The statistics of EM data collection and structure refinement are shown in Extended Data Table 4 

These authors contributed equally: D.Z., HMED, C.-P.C. EY6A binds the S1 subunit of SARS-CoV-2 and cross react with S1 of SARS-CoV-1. b, Antibody CR3022 similarly binds the S1 subunit of SARS-CoV-2 and cross react with SARS-CoV-1 S1, but with lower affinity. c, Convalescent serum from a COVID-19 patient was also included as a control and in this case binding to MERS and OC43 Spike proteins also investigated. d, Binding of EY6A on the SARS-CoV-2 infected cells in immunofluorescence assay. Anti-influenza H3 MAb BS 1A was included as a control.

SARS-CoV-2 Spike SARS-CoV-2 S1 SARS-CoV-1 S1 MERS Spike OC43 Spike 

performed cryo-EM sample preparation, screening and processing and J.Raedecke performed cryo-EM data collection, and J.Ren refined the cryo-EM structures

helped prepare materials, perform experiments and analysed data. All authors read and approved the manuscript

Real estimates of mortality following COVID-19 infection

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation. Science (80-. )

Cryo-EM structure of the SARS coronavirus spike glycoprotein in complex with its host cell receptor ACE2

Structure, Function, and Antigenicity of the SARS-CoV-2 Spike Glycoprotein

Dynamical asymmetry exposes 2019-nCoV prefusion spike

Human monoclonal antibodies block the binding of SARS-CoV-2 spike protein to angiotensin converting enzyme 2 receptor

Human monoclonal antibody combination against SARS coronavirus: Synergy and coverage of escape mutants

Potent neutralization of severe acute respiratory syndrome (SARS) coronavirus by a human mAb to S1 protein that blocks receptor association

Potent cross-reactive neutralization of SARS coronavirus isolates by human monoclonal antibodies

Human neutralizing antibodies elicited by SARS-CoV-2 infection

Global initiative on sharing all influenza data -from vision to reality

Neutralization of SARS-CoV-2 by destruction of the prefusion Spike

A highly conserved cryptic epitope in the receptor-binding domains of SARS-CoV-2 and SARS-CoV. Science (80-. )

Potent binding of 2019 novel coronavirus spike protein by a SARS coronavirus-specific human monoclonal antibody

A human monoclonal antibody blocking SARS-CoV-2 infection

Diagnostic methods in clinical virology. x

Article Structural Basis for Potent Neutralization of Betacoronaviruses by Single-Domain Camelid Antibodies

Distinct conformational states of SARS-CoV-2 spike protein

Deployment of convalescent plasma for the prevention and treatment of COVID-19

Treatment of 5 Critically Ill Patients with COVID-19 with Convalescent Plasma

Structural basis for neutralization of SARS-CoV-2 and SARS-CoV by a potent therapeutic antibody

Early Release-Severe Acute Respiratory Syndrome Coronavirus 2−Specific Antibody Responses in Coronavirus Disease

Identification of Human Single-Domain Antibodies against SARS-CoV-2

Immunogenicity and structures of a rationally designed prefusion MERS-CoV spike antigen

The production of glycoproteins by transient expression in mammalian cells

Epitope-associated and specificity-focused features of EV71-neutralizing antibody repertoires from plasmablasts of infected children

Sequence variation among SARS-CoV-2 isolates in Taiwan

Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR

Isolation and rapid sharing of the 2019 novel coronavirus (SAR-CoV-2) from the first patient diagnosed with COVID-19 in Australia

Beitrag zur kollektiven Behandlung pharmakologischer Reihenversuche

A procedure for setting up high-throughput nanolitre crystallization experiments. I. Protocol design and validation

A procedure for setting up high-throughput nanolitre crystallization experiments. Crystallization workflow for initial screening, automated storage, imaging and optimization

Xia2: An expert system for macromolecular crystallography data reduction

DIALS: Implementation and evaluation of a new integration package

Phaser crystallographic software

REFMAC5 for the refinement of macromolecular crystal structures

Coot: Model-building tools for molecular graphics

Macromolecular structure determination using X-rays, neutrons and electrons: Recent developments in Phenix

Crystal structure of cat muscle pyruvate kinase at a resolution of 2.6 Å

Inference of Macromolecular Assemblies from Crystalline State

New tools for automated high-resolution cryo-EM structure determination in RELION-3

Real-time CTF determination and correction

Algorithms for rapid unsupervised cryo-EM structure determination

Multiple ligand-protein interaction diagrams for drug discovery

We acknowledge the BD FACSAria™ cell sorter service provided by the Core Instrument 

The authors declare no competing interests.

Correspondence to David I. Stuart or Kuan-Ying A. Huang.

The coordinates and structure factors of the SARS-CoV-2 RBD/EY6A crystallographic complexes are available from the PDB with accession codes XXX and VVV respectively. EM maps and structure models are deposited in EMDB and PDB with accession codes XXX and YYY for the pre-fusion Spike, and XXXXX and yyyy for the dimeric complex respectively. The data that support the findings of this study are available from the corresponding authors on request.