key: cord-0993069-xarix8ue
authors: Ni, Dongchun; Lau, Kelvin; Turelli, Priscilla; Raclot, Charlene; Beckert, Bertrand; Nazarov, Sergey; Pojer, Florence; Myasnikov, Alexander; Stahlberg, Henning; Trono, Didier
title: Structural analysis of the Spike of the Omicron SARS-COV-2 variant by cryo-EM and implications for immune evasion
date: 2021-12-28
journal: bioRxiv
DOI: 10.1101/2021.12.27.474250
sha: 273f32ced90e6a3e56824ebbf49a5b7b4febbe7b
doc_id: 993069
cord_uid: xarix8ue

The Omicron (B.1.1.529) SARS-COV-2 was reported on November 24, 2021 and declared a variant of concern a couple of days later.1,2 With its constellation of mutations acquired by this variant on its Spike glycoprotein and the speed at which this new variant has replaced the previously dominant variant Delta in South Africa and the United Kingdom, it is crucial to have atomic structural insights to reveal the mechanism of its rapid proliferation. Here we present a high-resolution cryo-EM structure of the Spike protein of the Omicron variant.

The Omicron Spike cDNA was obtained by reverse transcription of a nasopharyngeal viral sample and cloned into an expression vector as previously described. 3 The protein was purified to homogeneity via its TwinStrep tag from the supernatant of transiently transfected ExpiCHO cells (Fig. S1) , and vitrified for cryo-electron microscopy (cryo-EM) at a concentration of 0.35 mg/ml as described in the Methods. Grids were screened with a 200kV TFS Glacios cryo-EM, and final data collection was performed on a 300kV TFS Titan Krios G4. Data was processed 'on-the-fly' with CryoSPARC Live, resulting in a first 3D map after 3 hours from the Glacios. 4 The cryo-EM maps of the Omicron Spike glycoprotein clearly revealed many of the mutations present in this variant (Fig. S3) . The overall architecture of the Spike trimer was comparable to the first structure of the wild-type Spike (PDB ID 6VSB) (RMSD Cɑ all atoms 2.4 Å) with one RBD in the up position and two RBDs down ( Fig. 1a-b ) 5 , although a minority of Spike trimers were in a closed conformation with all RBDs down. We did not observe other populations containing two or three RBDs in the up configuration. This suggests that the Omicron Spike may preferentially be in either a one RBD-up or in a closed state (Fig. S2) . The best-defined mutations are within the S2 domain intercalating between Spike monomers and within the NTD (Fig. 1c, e) . A view from the top of the Spike protein exposes a large cluster of mutations on the RBDs (Fig. 1d-e) .

The exceptional quality of the map in the S2 domain reveals potential new interactions formed by Omicron-unique mutations. The D796Y (all numbering based on wild-type positions) mutation replaces a charged surface-exposed acidic residue with tyrosine, an amino acid that contains an aromatic sidechain (Fig. 2a) . We observe that Y796 allows for potential carbohydrate-pi interactions with the N-linked glycan chain originating from N709 of the neighboring monomer chain. 6, 7 We suspect that this interaction could have a stabilizing effect for the spike trimer.

The N856K mutation is within the fusion peptide of the Spike protein. 8 The fusion peptide is required for attachment and anchoring of the virus directly to a host's cell membrane. It is exposed during proteolytic cleavage that liberates the S1 from the S2 domain followed by subsequent cleavage at the S2' site. We observe the N856K mutation with the longer side-chain of the lysine residue able to form new interactions with T572 from an adjacent monomer (Fig.   2b ). We can only speculate that this may have a functional impact on the release of the fusion peptide. In addition, the mutation from an asparagine to lysine may aid cell membrane penetration. [8] [9] [10] Other mutations located within S2 with stabilizing potential, such as N764K, can also be clearly visualized (Fig. 2c ). We can model an amine head-group of K764 forming two new interactions with Q314 and N317 from an adjacent monomer, which could stabilize the Spike trimer. In particular, Q314 and N317 precede the RBD in sequence (residues 330-520) and are just below the linker that allows for RBD transition from a down to an up position.

We observe that the RBD is predominantly in a 1-up or closed state, allowing for the Spike to occlude efficiently RBD epitopes. Thus, the many mutations present in the S2 domain present only in the Omicron Spike protein may allow for optimization of RBD positions that allow for receptor binding while concealing epitopes that are the targets of neutralizing antibodies. 11 The NTD is the target of various neutralizing antibodies and constitutes an antigenic supersite, [12] [13] [14] and was subjected to focused refinement ( Fig. 2d-h) . We observe that the overall fold of the Omicron NTD is relatively unchanged compared to either its wild-type (RMSD 1.7 Å) or Delta (RMSD 2.4 Å) counterparts ( Fig. S4a-b) . The loop encompassing deletions 143-145 (N3 loop) 12-14 has a different conformation compared to the Delta NTD. The Delta variant Spike also contains NTD mutations in the same N3 loop (E156G, Δ157-158). Superpositions of structures of Fab-bound NTDs predict that the Omicron NTD is resistant to some NTD antibodies due to steric clashes and changes in the antibody binding surface (Fig. S4c) . [12] [13] [14] We identify a second loop (residues 208-217), hereby called the EPE loop, which contains Omicron-specific mutations, deletions and inserts: deletion 211, L212I and insertion at 214 of EPE (Fig2. d-e). The locally-refined cryo-EM map in this region allowed for the de novo building of this loop with the aid of a model generated from AF2. 15 In the Omicron NTD, the EPE loop follows the Cɑ backbone as the wild-type NTD, however the chemical nature is drastically changed with the introduction of two acidic negatively charged residues (Fig. 2e ).

This insertion is not predicted to affect the binding of known NTD targeting antibodies, being localized far from the site of antibody binding (Fig. S4c) . 12, 13 The RBD is the target for a vast number of neutralizing antibodies elicited by vaccination and previous infection. All commercially available antibodies that have been approved for clinical use, target this domain of the Spike protein. Based upon previous variants of concern (VOCs) and experimental studies, some of the mutations such as N501Y and Q498R may independently or synergistically enhance binding of the Spike to its receptor, the human ACE2 protein. 16, 17 In addition, the mutations that decorate the RBD may allow for immune evasion or render monoclonal antibody therapy ineffective. [18] [19] [20] Alignment of the wild-type RBD with that of the Omicron reveals only slight changes with a RMSD of 2.6 Å (Fig. 2f) .

The locally-refined map in this region allowed us to model all the mutations located within the RBD (Fig. 2g-h and S3) . A region that encompasses residues 371 to 375 contains three serine residues mutated: S371L, S373P, S375F. These residues are on a surface exposed region of the RBD and may play a role in immune evasion of the Omicron variant. In addition, the top saddle of the RBD (Fig. 2f ) is the location of many mutations shared with others VOCs, in particular N501Y that has been shown to foster ACE2 binding, along with other Omicron-specific mutations such as Q493R and Q498R. 16, 17 Overall the Omicron RBD has a highly mutated surface that encompasses the epitopes of the known major classes of antibodies (Fig. S5a ). Many of the RBD mutations are also located along a ridge that interacts directly with the helices of the ACE2 receptor (Fig. 2f ). We performed biolayer interferometry (BLI) binding assays to evaluate the affinity of ExpiCHO-expressed dimeric Fc-human ACE2 to full-length Omicron Spike (Fig. S6 ). Compared to the wild-type Spike, hACE2 binds to the Omicron Spike with similar affinities and comparable to that of binding to the Delta Spike (4.4 nM vs. 6.8 nM vs. 6.1 nM; wild-type vs. Omicron vs. Delta). The Omicron Spike also shares many RBDspecific mutations (namely N501Y, K417N and mutation at E484) with other VOCs, namely Alpha, Beta and Gamma. These variants all show relatively stronger binding to hACE2 than Omicron (Fig. S6) . We conclude that the appearance of a large number of mutations at the ACE2 binding interface do not significantly alter the overall binding affinity of dimeric Fc-hACE2 to the Omicron Spike. In addition, we tested binding to the dimeric mouse ACE2 receptor as it has been previously reported that N501Y containing VOCs had increased ability to infect mice and bind ACE2. [28] [29] [30] The dimeric Fc-mACE2 displays robust binding to the Omicron Spike with an almost 8-fold increase in binding affinity compared to wild-type.

Mouse ACE2 also binds significantly stronger compared to other variants with the exception of gamma.

Here we present a high-resolution structure of the Spike protein of Omicron, which is predicted to soon become the dominant SARS-CoV-2 variant worldwide (as of December 2021). The overall structure of the full-length Spike, its RBD and NTD domains are similar to that of the wild-type but with specific changes in both local conformations and surfaces involved in antibody recognition. The location of the mutations in both the RBD and NTD suggests that there has been a strong evolutionary pressure on the Omicron Spike to evade the humoral immune response. The high-quality maps in parts of the S2 domain allowed us to visualize Omicron-specific mutations that form new stabilizing interactions within the Spike trimer, and which may optimize its ability to engage and bind ACE2, while occluding epitopes that are targets of neutralizing antibodies. 11 Even with a heavily mutated ACE2-binding interface, our binding experiments show that the Omicron RBD is capable of interacting with human ACE2 receptor as efficiently as that of the Delta variant. In addition, the robust binding of mouse ACE2 to the Omicron Spike suggests an increased ability to infect other animal species such as rodents.

Our structure highlights the many mutations found on the Omicron Spike, which we speculate may affect protein stability, glycan conformation diversity, membrane fusion, ACE2 and importantly evasion from antibodies from vaccination or previous infection. BLI curves of indicated Spike binding to captured dimeric human or mouse ACE2. Raw data is colored by concentration as indicated. The black lines represent the fit to the raw data using a 1:1 binding model. Spike trimers was determined to be >99% pure by SDS-PAGE analysis. (Fig. S1) The protein eluted as a highly-glycosylated trimeric species as determined by SEC-MALS analysis.

Cryo-EM grids were prepared with a Vitrobot Mark IV (Thermofisher Scientific (TFS)). Quantifoil R1.2/1.3 Au 400 holey carbon grids were glow-discharged for 120 s at 15mA using a PELCO easiGlow device (Ted Pella, Inc.). 3.0 µL of a 0.35 mg/ml protein sample was applied to the glow-discharged grids, and blotted for 6 s under blot force 10 at 100% humidity and 4 °C in the sample chamber, and the blotted grid was plunge-frozen in liquid nitrogen-cooled liquid ethane. Grids were screened for particle presence and ice quality on a TFS Glacios microscope (200kV), and the best grids were transferred to TFS Titan Krios G4. Cryo-EM data was collected using TFS Titan Krios G4 transmission electron microscope (TEM), equipped with a Cold-FEG on a Falcon IV detector in electron counting mode. Falcon IV gain references were collected just before data collection. Data was collected using TFS EPU v2.12.1 using aberration-free image shift protocol (AFIS), recording 4 micrographs per ice hole.

Movies were recorded at magnification of 96kx, corresponding to the 0.83Å pixel size at the specimen level, with defocus values ranging from -0.8 to -1.8 µm. Exposures were adjusted automatically to 60 e -/Å 2 total dose, resulting in an exposure time of approximately 3 seconds per movie. In total, 8'672 micrographs in EER format were collected.

On-the-fly processing was first performed during data acquisition for evaluating the data quality during screening by using cryoSPARC live v3.3.1. 4 The obtained ab-initio structures were used for better particle picking for template creation. Motion correction was performed on raw stacks without binning using the cryoSPARC implementation of motion correction. 32 470'735 particles were template-based automatically picked and particles were binned by a factor of 4. Two rounds of 2D classification were performed, resulting in a particle set of 144'425 particles. Selected particles resulting from the 2D classification were used for ab-initio reconstruction and hetero-refinement. After hetero-refinement, 76'450 particles contributed to an initial 3D reconstruction of 3.02 Å resolution (FSC 0.143) with C1 symmetry, while application of C3 symmetry resulted in a 3D map at 2.65 Å resolution (FSC 0.143). The same number of particles were then subjected to local 3D refinement. A soft mask volume against the RBD-NTD area was generated manually in UCSF Chimera and cryoSPARC. 33 The RBD-NTD region was subtracted from the global trimeric spike and further focused refinement was performed subsequently, which resulted in a 3D map at 3.88 Å resolution (FSC 0.143).

A model of a Spike trimer (PDB ID 6MXL) or AlphaFold2 (ColabFold implementation) models of the Spike's NTD and RBD were fit into the cryo-EM maps with UCSF Chimera. 15 These docked models were extended and rebuilt manually with refinement, using Coot and Phenix. 34, 35 Figures were prepared in UCSF Chimera and UCSF ChimeraX. 33

Running buffer was 150 mM NaCl, 10 mM HEPES 7.5 and all experiments were performed on a Gator BLI system. Dimeric mFc-human ACE2 was diluted to 10 ug/mL and captured with MFc tips (GatorBio). Loaded tips were dipped into a 3-fold serial dilution series (300 nM, 100 nM, 33.3 nM) of analyte SPIKE. Curves were processed using the Gator software with a 1:1 fit after background subtraction.

The reconstructed maps of the full Omicron Spike are available from the EMDB database, C1 symmetry, EMDB-14086. The atomic model for the full-length Omicron Spike is available from the PDB database, PDB-7QO7. The local focused-refinement map of the NTD and RBD regions are available from the EMDB database, EMDB-14087. The atomic model for the NTD and the RBD in the locally refined map is available from the PDB database, PDB-7QO9. The raw cryo-EM images will be deposited in the EMPIAR database. 

Cryo-EM data collection and initial image processing was performed at the Dubochet Center for Imaging, a common initiative of the EPFL, UNIL and UNIGE. We thank Rosa Schier, Laurence Durrer, Soraya Quinche and Michael François of the Protein Production and Structure Core Facility of EPFL for helping in the production of Spike Omicron. We thank Isabella Eckerle, Meriem Bekliz and the Virology laboratory of the HUG for the RNA sample collection. This work was supported by the Swiss National Science Foundation, grant NCCR TransCure. 

Track Omicron's spread with molecular data

Rapid epidemic expansion of the SARS-CoV-2 Omicron variant in southern Africa

A highly potent antibody effective against SARS-CoV-2 variants of concern

cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

CH/π Interactions in Carbohydrate Recognition

Carbohydrate-Aromatic Interactions in Proteins

Fusion Peptide of SARS-CoV-2 Spike Rearranges into a Wedge Inserted in Bilayered Micelles

Characterization of a Highly Conserved Domain within the Severe Acute Respiratory Syndrome Coronavirus Spike Protein S2 Domain with Characteristics of a Viral Fusion Peptide

Activation of the SARS coronavirus spike protein via sequential proteolytic cleavage at two distinct sites

The SARS-CoV-2 spike protein: balancing stability and infectivity

N-terminal domain antigenic mapping reveals a site of vulnerability for SARS-CoV-2

Potent SARS-CoV-2 neutralizing antibodies directed against spike Nterminal domain target a single supersite

Neutralizing and protective human monoclonal antibodies recognizing the N-terminal domain of the SARS-CoV-2 spike protein

ColabFold -Making protein folding accessible to all

The N501Y spike substitution enhances SARS-CoV-2 transmission

SARS-CoV-2 variant prediction and antiviral drug design are enabled by RBD in vitro evolution

Antibody resistance of SARS-CoV-2 variants B.1.351 and B.1.1.7

SARS-CoV-2 RBD antibodies that maximize breadth and resistance to escape

Resistance of SARS-CoV-2 variants to neutralization by monoclonal and serum-derived polyclonal antibodies

SARS-CoV-2 neutralizing antibody structures inform therapeutic strategies

Structures of Human Antibodies Bound to SARS-CoV-2 Spike Reveal Common Epitopes and Recurrent Features of Antibodies

Mapping Neutralizing and Immunodominant Sites on the SARS-CoV-2 Spike Receptor-Binding Domain by Structure-Guided High-Resolution Serology

Tackling COVID-19 with neutralizing monoclonal antibodies

Broad cross-reactivity across sarbecoviruses exhibited by a subset of COVID-19 donor-derived neutralizing antibodies

Mapping mutations to the SARS-CoV-2 RBD that escape binding by different classes of antibodies

Broadly-Neutralizing Antibodies Against Emerging SARS-CoV-2 Variants

The B1.351 and P.1 variants extend SARS-CoV-2 host range to mice

Characterization and structural basis of a lethal mouse-adapted SARS-CoV-2

Circulating SARS-CoV-2 variants B.1.1.7, 501Y.V2, and P.1 have gained ability to utilize rat and mouse Ace2 and altered in vitro sensitivity to neutralizing antibodies and

SARS-CoV-2 Omicron Variant: ACE2 Binding, Cryo-EM Structure of Spike Protein-ACE2 Complex and Antibody Evasion

Alignment of cryo-EM movies of individual particles by optimization of image translations

Structure visualization for researchers, educators, and developers

Features and development of Coot

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