Cryo-EM Structure of the 2019-nCoV Spike in the Prefusion Conformation

The outbreak of a novel betacoronavirus (2019-nCov) represents a pandemic threat that has been declared a public health emergency of international concern. The CoV spike (S) glycoprotein is a key target for urgently needed vaccines, therapeutic antibodies, and diagnostics. To facilitate medical countermeasure (MCM) development we determined a 3.5 Å-resolution cryo-EM structure of the 2019-nCoV S trimer in the prefusion conformation. The predominant state of the trimer has one of the three receptor-binding domains (RBDs) rotated up in a receptor-accessible conformation. We also show biophysical and structural evidence that the 2019-nCoV S binds ACE2 with higher affinity than SARS-CoV S. Additionally we tested several published SARS-CoV RBD-specific monoclonal antibodies and found that they do not have appreciable binding to nCoV-2019 S, suggesting antibody cross-reactivity may be limited between the two virus RBDs. The atomic-resolution structure of 2019-nCoV S should enable rapid development and evaluation of MCMs to address the ongoing public health crisis.

Based on the reported genome sequence of 2019-nCoV(4), we expressed ectodomain 37 residues 1−1208 of 2019-nCoV S ( Figure 1A, Supplementary Figure 1), adding two stabilizing 38 proline mutations in the C-terminal S2 fusion machinery based on a previous stabilization 39 strategy which proved highly effective for betacoronavirus S proteins (11,14). We obtained 40 roughly 0.5 mg/L of the recombinant prefusion-stabilized S ectodomain from FreeStyle 293 41 cells, and the protein was purified to homogeneity by affinity chromatography and size-exclusion 42 chromatography (Supplementary Figure 1). Cryo-EM grids were prepared using this purified, 43 fully glycosylated S protein and preliminary screening revealed a high particle density with little 44 aggregation near the edges of the holes. 45 After collecting and processing 3,207 micrograph movies, we obtained a 3.5 Å-resolution 46 3D reconstruction of an asymmetrical trimer in which a single RBD was observed in the "up" 47 conformation. (Figure 1B, Supplementary Figure 2). Due to the small size of the RBD (~21 48 kDa), the asymmetry of this conformation was not readily apparent until ab initio 3D 49 reconstruction and 3D classification were performed ( Figure 1B, Supplementary Figure 3). By 50 using the 3D variability feature in cryoSPARC v2 (15), we were able to observe breathing of the 51 S1 subunits as the RBD underwent a hinge-like movement, which likely contributed to the 52 relatively poor local resolution of S1 compared to the more stable S2 subunit (Supplementary 53 Movies 1 and 2). This seemingly stochastic RBD movement has been captured during structural 54 characterization of the closely related betacoronaviruses SARS-CoV and MERS-CoV, as well as 55 the more distantly related alphacoronavirus porcine epidemic diarrhea virus (PEDV) (10, 11, 13, 56 16). The observation of this phenomenon in 2019-nCoV S suggests that it shares the same 57 mechanism of triggering that is thought to be conserved among the Coronaviridae, wherein 58 receptor-binding to exposed RBDs leads to an unstable 3 RBD-up conformation that results in 59 shedding of S1 and refolding of S2 (11,12). 60 Because the S2 subunit appeared to be a symmetric trimer, we performed a 3D 61 refinement imposing C3 symmetry, resulting in a 3.2 Å-resolution map, with excellent density 62 for the S2 subunit. Using both maps we built the vast majority of the 2019-nCoV S ectodomain,   Figure 4A) (13,(16)(17)(18). 68 The overall structure of 2019-nCoV S resembles that of SARS-CoV S, with a root mean 69 square deviation (RMSD) of 3.8 Å over 959 Cα atoms. The largest discrepancy between these 70 two structures is a conformational difference between the positions of the RBDs in their 71 respective "down" conformations ( Figure 2A). Whereas the SARS-CoV RBD in the "down" 72 conformation packs tightly against the N-terminal domain (NTD) of the neighboring protomer, 73 the 2019-nCoV RBD in the "down" conformation is angled closer to the central cavity of the 74 homotrimer. Despite this observed conformational difference, when the individual structural 75 domains of 2019-nCoV S are aligned to their counterparts from SARS-CoV S, they reflect the 76 high degree of structural homology between the two proteins, with the NTDs, RBDs, 77 subdomains 1 and 2 (SD1 and SD2) and S2 subunits yielding RMSD values of 2.6 Å, 3.0 Å, 2.7 78 Å and 2.0 Å, respectively ( Figure 2B). 79 2019-nCoV S shares roughly 96% sequence identity with the S protein from the bat 80 coronavirus RaTG13, with the most notable variation arising from an insertion in the S1/S2 81 . CC-BY 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.11.944462 doi: bioRxiv preprint protease cleavage site that results in an "RRAR" furin recognition site in 2019-nCoV, rather than 82 the single arginine in SARS-CoV (Supplementary Figure 5) (19)(20)(21)(22). A similar phenomenon 83 has been observed for influenza viruses, where amino acid insertions that create a polybasic furin 84 site in a related position in influenza hemagglutinin proteins are often found in highly virulent 85 avian and human influenza viruses (23). In addition to this insertion of residues in the S1/S2 86 junction, 29 variant residues exist between 2019-nCoV S and RaTG13 S, with 17 of these 87 positions mapping to the RBD (Supplementary Figures 5 and 6). We also analyzed the 61 88  (14). We also formed a complex of ACE2 bound to the 2019-nCoV S 98 ectodomain and observed it by negative-stain EM, where it strongly resembled the complex 99 formed between SARS-CoV S and ACE2, which has been observed at high-resolution by cryo-100 EM ( Figure 3B) (14,27). The high affinity of 2019-nCoV S for human ACE2 may contribute to 101 the apparent ease with which 2019-nCoV can spread from human-to-human(1), however 102 additional studies are needed to investigate this possibility. 103 . CC-BY 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.11.944462 doi: bioRxiv preprint The overall structural homology and shared receptor usage between SARS-CoV S and 104

2019-nCoV S prompted us to test published SARS-CoV RBD-directed monoclonal antibodies 105
(mAbs) for cross-reactivity to the 2019-nCoV RBD ( Figure 4A). A 2019-nCoV RBD-SD1 106 fragment (S residues 319-591) was recombinantly expressed, and appropriate folding of this 107 construct was validated by measuring ACE2 binding using biolayer interferometry (BLI) 108 ( Figure 4B). Cross-reactivity of the SARS-CoV RBD-directed mAbs S230, m396 and 80R was 109 then evaluated by BLI (12,(28)(29)(30). Despite the relatively high degree of structural homology suggests that SARS-directed mAbs will not necessarily be cross-reactive and that future antibody 115 isolation and therapeutic design efforts will benefit from using 2019-nCoV S proteins as probes. 116 The rapid global spread of 2019-nCoV, prompting the PHEIC declaration by WHO 117 signals the urgent need for coronavirus vaccines and therapeutics. Knowing the atomic-level 118 structure of the spike will support precision vaccine design and discovery of antivirals, 119 facilitating medical countermeasure development. 120 . CC-BY 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.11.944462 doi: bioRxiv preprint

Protein expression and purification 122
To express the prefusion S ectodomain, a gene encoding residues 1−1208 of 2019-nCoV S 123 The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.11.944462 doi: bioRxiv preprint Plasmids encoding the heavy and light chains of S230, 80R and m396 IgG were transiently 144 transfected into Expi293 (Thermo Fischer) using polyethylenimine. Antibodies were purified 145 from cell supernatants using Protein A resin before being used for biolayer interferometry.  Table 1. 156

Cryo-EM data processing 157
Motion correction, CTF-estimation and non-templated particle picking were performed in Warp 158 (32). Extracted particles were imported into cryoSPARC v2.12.4 (15) for 2D classification, 3D 159 classification and non-uniform 3D refinement. The C1 RBD "up" reconstruction was sharpened 160 in cryoSPARC, and the 3D reconstruction with C3 symmetry was subjected to local B-factor 161 sharpening using LocalDeBlur (33). Models were built in Coot, before being iteratively refined 162 in both Phenix and ISOLDE (34)(35)(36). Some of the data processing and refinement software was 163 curated by SBGrid (37). The full cryo-EM data processing workflow is described in

Surface plasmon resonance 167
His-tagged 2019-nCoV S was immobilized to an NiNTA sensorchip (GE Healthcare) to a level 168 of ~800 response units (RUs) using a Biacore X100 (GE Healthcare) and a running buffer 169 composed of 10 mM HEPES pH 8.0, 150 mM NaCl and 0.05% Tween 20. Serial dilutions of 170 purified and untagged ACE2 were injected ranging in concentration from 250 to 15.6 nM. The 171 resulting data were fit to a 1:1 binding model using Biacore Evaluation Software (GE 172 Healthcare). His-tagged SARS-CoV RBD-SD1 was immobilized to an NiNTA sensorchip to a 173 level of ~350 RUs using a Biacore X100 and the same running buffer listed above. Serial 174 dilutions of purified and untagged ACE2 were injected ranging in concentration from 500 to 31.3 175 nM. The resulting data were fit to a 1:1 binding model using Biacore Evaluation Software. 176

Negative stain EM 177
Purified 2019-nCoV S was diluted to a concentration of 0.032 mg/mL in 2 mM Tris pH 8.0, 200 178 mM NaCl and 0.02% NaN3. Diluted S protein was mixed with a 1.5-fold molar excess of ACE2 179 and the mixture was incubated on ice for 1 minute before 4.8 uL of the protein mixture was 180 deposited on a CF400-Cu grid (Electron Microscopy Sciences) before being stained with 181 methylamine tungstate (Nanoprobes). This grid was imaged in an FEI Talos TEM (Thermo 182 Scientific) equipped with a Ceta 16M detector. Micrographs were collected manually using TIA 183 v4.14 software at a magnification of x92,000, corresponding to a pixel size of 1.63 Å/pixel. CTF 184 estimation, particle picking and 2D class averaging were performed in cisTEM (38). 185

Biolayer interferometry 186
Fc-tagged 2019-nCoV RBD-SD1 was immobilized to an anti-human capture (AHC) sensortip 187 (FortéBio) using an Octet RED96e (FortéBio). The sensortip was then dipped into 100 nM 188 ACE2 to measure association before being dipped into a well containing only running buffer 189 . CC-BY 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.11.944462 doi: bioRxiv preprint The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.11.944462 doi: bioRxiv preprint    The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.11.944462 doi: bioRxiv preprint class averages of 2019-nCoV S bound by ACE2. Averages have been rotated so that ACE2 is positioned above the 2019-nCoV S protein with respect to the viral membrane. A cartoon depicting the ACE2-bound 2019-nCoV S protein is shown (right) with ACE2 in blue and S protein monomers colored tan, pink and green. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.11.944462 doi: bioRxiv preprint  subdomain 2, S1/S2= S1/S2 protease cleavage site, S2′= S2′ protease cleavage site, FP= fusion peptide, HR1= heptad repeat 1, CH= central helix, CD= connector domain, HR2= heptad repeat 2, TM= transmembrane domain, CT= cytoplasmic tail. Arrows denote protease cleavage sites. (B) Select 2D class averages of the particles that were used to calculate the 2019-nCoV S reconstruction (left). Side and top views of the prefusion structure of the 2019-nCoV S protein with a single RBD in the "up" conformation (right). The two RBD "down" protomers are shown as cryo-EM density in either white or gray and the RBD "up" protomer is shown in ribbons, colored corresponding to the schematic in Fig 1A. . CC-BY 4.0 International license author/funder. It is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.11.944462 doi: bioRxiv preprint  The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.11.944462 doi: bioRxiv preprint  The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.02.11.944462 doi: bioRxiv preprint