key: cord-0269430-2n5cetxv
authors: Yang, Tzu-Jing; Yu, Pei-Yu; Chang, Yuan-Chih; Chang, Ning-En; Tsai, Yu-Xi; Liang, Kang-Hao; Draczkowski, Piotr; Lin, Bertina; Wang, Yong-Sheng; Chien, Yu-Chun; Khoo, Kay-Hooi; Wu, Han-Chung; Hsu, Shang-Te Danny
title: Structure-activity relationships of B.1.617 and other SARS-CoV-2 spike variants
date: 2021-09-13
journal: bioRxiv
DOI: 10.1101/2021.09.12.459978
sha: 25992a725fd057a016b5ddfd3e52547a6a91b74a
doc_id: 269430
cord_uid: 2n5cetxv

The surge of COVID-19 infection cases is spurred by emerging SARS-CoV-2 variants such as B.1.617. Here we report 38 cryo-EM structures, corresponding to the spike protein of the Beta (B.1.351), Gamma (P.1), Delta (B.1.617.2) and Kappa (B.1.617.1) variants in different functional states with and without its receptor, ACE2. Mutations on the N-terminal domain not only alter the conformation of the highly antigenic supersite of the Delta variant, but also remodel the glycan shield by deleting or adding N-glycans of the Delta and Gamma variants, respectively. Substantially enhanced ACE2 binding was observed for all variants, whose mutations on the receptor binding domain modulate the electrostatics of the binding interfaces. Despite their abilities to escape host immunity, all variants can be potently neutralized by three unique antibodies.

The definitions of the Beta, Gamma, Delta and Kappa variants are primarily based on the mutations of the highly glycosylated spike (S) protein that is responsible for host recognition, membrane fusion and viral entry (Fig. 1A) . The S protein can be divided into the N-terminal S1 subunit and C-terminal S2 subunit, which are responsible for host receptor binding and membrane fusion, respectively 6, 7 . The receptor binding domain (RBD) within the S1 subunit binds to the ectodomain of angiotensin converting enzyme 2 (ACE2). The RBD is the most immunogenic domain of the S protein, accounting for up to 90% of the neutralizing antibodies (nAbs) derived from convalescent sera [8] [9] [10] [11] . RBD-specific nAbs neutralize SARS-CoV-2 variants by inhibiting ACE2 binding through steric hindrance. A number of nAbs are now in clinical use to treat COVID-19 patients 12 .

However, mutations within the RBD can reduce the efficacy of nAb neutralization by 4 disrupting the respective conformational epitopes, such as N501Y in the Alpha variant, E484K in the Beta and Gamma variants, and L452R in the Delta and Kappa variants (Fig.   1, A and B) . The reduced neutralization efficacy associated with these mutations indeed lead to the Food and Drug Administration of the United States to revoke the emergency use authorization of bamlanivimab (LY-CoV555) 13 . The S1 subunit, also referred to as the N-terminal domain (NTD), harbors eight N-glycosylation sites that form a glycan shield over the protein surface rendering it less immunogenic [14] [15] [16] . Nevertheless, a supersite within the glycan shield has been identified to be targeted by a panel of nAbs 17 . The mechanism by which NTD-specific nAbs neutralize SARS-CoV-2 infection is not well understood; however, it is well established that several VOC-and VOI-associated NTD mutations significantly reduce or completely abolish nAb neutralization thus leading to immunity escape 16, [18] [19] [20] . Specifically, a recent study reports complete escape of the Delta variant from a panel of 16 NTD-specific nAbs 21 .

As global vaccination programs have expanded at an unprecedented pace, reductions in the effectiveness of vaccines against different VOCs are emerging. For instance, the Pfizer/BioNTech (BNT162b2) and Moderna (mRNA-1273) vaccines experience an astounding drop in protection (up to 12-fold) against the Beta variant while their protection against the Gamma variant is reduced by 2-to 3-fold 15, 20, 22 . While mRNA-1273 affords comparable protection against the Alpha variant compared to the wild-type (WT) Wuhan strain 20,22-24 , a 2-to 3-fold reduction in protection against the Delta variant is reported 25, 26 . In contrast, BNT162b2 generates substantially less neutralizing titers against the Delta variant compared to the Alpha variant 21, 27 . Another study shows that the Kappa variant is even more resistant to BNT162b2-induced sera compared to the Delta variant and other emerging variants 28 . Notably, one study reports that the Delta and Beta variants essentially abolish the neutralizing activity of the AstraZeneca (ChAdOx1 nCoV-19) vaccine 29 . Notably, convalescent sera from the Beta or Gamma variant-infected individuals and the vaccine-elicited antibodies show substantially reduced neutralization against the Delta and Kappa variants 11, 30 . These findings raise the questions about of the current vaccines against emerging breakthrough SARS-CoV-2 variants 31 . As current developments of COVID-19 vaccines are chiefly focused on the S protein, understanding 5 how the mutations associated with VOCs and VOIs on the structure and function of the S protein, and their interactions with ACE2 and nAbs, is therefore of paramount importance.

To examine the effect of mutations on the structures of the S protein of the Beta Table 1 and Extended Data Movie 1). To verify the authenticity of the unique 3 RBD-up conformation, we solved the cryo-EM structure of S-Beta without the stabilizing tandem proline and furin cleavage site substitutions (fm2P; see Methods).

A well-defined 3 RBD-up state could also be identified for the cleavable form of S-Beta (Extended Data Fig. 2 and Extended Data Table 2 ). While S-Gamma only populated the 1 RBD-up and 2 RBD-up states, the 1 RBD-up state of S-Gamma could be divided into two distinct substates through a robust three-dimensional variability analysis (3DVA; Table 4 and Extended Data Movie 3) 32, 33 . The separation of multiple substates within the same class (conformational basin) of the RBD-up conformation was made possible through two rounds of 3DVA (Extended Data Fig. 4 and Methods). Unexpectedly, we identified an unprecedented head-to-head dimer 6 of S-Kappa trimers, involving all three RBDs in each trimeric S protein to bind to another RBDs from the other trimeric S protein. Furthermore, S-Kappa exhibited an all RBD down conformation, a 1 RBD-up conformation with two substates, and a single 2 RBD-up conformation (Fig. 1F, Extended Data Fig. 5 , Extended Data Tables 5-6, and Extended Data Movie 4). The head-to-head dimer of S-Kappa trimers displayed rigid body motions between the two trimers that could be resolved into three substates. Collectively, our comprehensive structural analyses of the SARS-CoV-2 S variants illustrates a broad conformational space explored by each of the variants with S-Delta and S-Kappa being the most intriguing in terms of the large number of substates within each basin of the RBD conformational grouping, and the unique higher-order assembly, respectively.

Being the most prevalent COVID-19 VOC to date with increased transmissibility and immunity escape, the Delta variant harbors ten mutations in the S protein compared to the original Wuhan strain (S-WT), namely T19R, G142D, E156G, Δ157, Δ158, L452R, T478K, D614G, P681R and D950N 27 (Fig. 1, A and B) . Recent studies show that VOCs are refractory to vaccine and convalescent sera as well as nAbs that target the RBD and NTD to different degrees 21, 27, 30, 34 . The Delta variant, in particular, shows reduced sensitivity to convalescent sera from the Beta and Gamma variants, indicating that the Beta, Gamma and Delta variants are antigenically divergent, and that there is a risk of reinfection by the Delta variant for individuals who have been infected by the Beta or Gamma variants 30 . The Delta variant is also completely refractory to NTD-specific nAbs isolated from convalescent sera with prior infection of the WT strain 21 . As the mutations within the NTD are the most divergent among the VOCs and VOIs, we compared the structure of the NTD of S-Delta with respect to that of S-WT. Structural mapping of the epitope count based on the reported structures of NTD in complex with antibodies revealed a supersite encompassing a cluster of residues that are most antigenic, including Y145, H146, L249 and P251 ( Fig. 2A and Extended Data Table 7 ) 35 . The double deletions (Δ157 and Δ158) and the substitutions, G142D and E156G, resulted in major conformational rearrangements of the supersite, including the sequestering of Y145 (Fig. 2B) . The loop on which the N149 glycan is located underwent significant conformational rearrangements that greatly alters the glycan shield around the supersite ( Fig. 2B and Extended Data Fig. 6A) . Furthermore, the T19R mutation abrogates the 17 NLT 19 sequon for N-glycosylation. Indeed, we did not observe any N-glycan density in the cryo-EM maps of S-Delta (Extended Data Fig. 6A ) nor did we find evidence of glycosylation in mass spectrometry (MS) analysis (Extended Data Fig. 7) .

In the context of immunity evasion through alteration of the glycan shield, S-Gamma harbors two NTD mutations -T20N and R190S -that introduce two additional N-glycans on N20 and N188. MS analysis unambiguously established the presence of a high-mannose type N-glycan at N188 with five mannose residues (M5) being the predominant glycoform (Fig. 2C) . The remaining N-glycans on the NTD are primarily complex types, except for the N234 glycan, which is mostly high-mannose type, in line with previous reports 14 . We note that a shift from mannose-9 (M9) to mannose-8 (M8)

together with a small proportion of highly processed complex type glycan (N5) were observed for N234, suggesting that the N234 glycan is less occluded during Golgi glycosylation compared to other spike variants that show exclusively a high-mannose type glycan at N234 with the major glycoform being M8 14 (Extended Data Fig. 7) . The cryo-EM map of S-Gamma showed additional protrusion from the side-chain of N20 indicative of the presence of N-glycan (Extended Data Fig. 6B) . Molecular modeling of a fully glycosylated NTD of S-Gamma showed that the supersite is surrounded by N-glycans (Fig. 2D) , and that the additional N20 and N188 glycans can sterically hinder NTD-specific antibody binding such as nAb 4-18, which targets the supersite 19 (Fig. 2E) , and nAb P008_056, which targets an alternative epitope, requiring major conformational rearrangements in the loops on which N188 resides 18 (Fig. 2F) . These findings provided a clear structural basis underlying the immunity escape of the Gamma variant through the NTD mutations.

Our cryo-EM analysis showed that S-Kappa exhibited an ensemble of conformations with one third adopting an unprecedented head-to-head dimer of trimers ( Fig. 1F, Fig. 3A and Extended Data Movie 4). The dimeric interface involves (i) inter-RBD van der Waals contacts between F490 and the S-Kappa-specific substitution 8 E484Q, (ii) a bipartite hydrogen bonding network formed between the side chains of Q493 from the dimerizing RBDs, (iii) a bipartite hydrogen bonding network between the sidechain of N450 from one RBD to the backbone amide and side chain carboxyl group of N487 from the other RBD, and (iv) a hydrogen bond between the side chain hydroxyl group of Y489 of one RBD and side chain carboxyl of N450 from the other RBD (Fig. 3B) .

While the N450-mediated hydrogen bonding is asymmetrically formed between N487 or Y489 from the other RBD, the van der Waals contacts and the bipartite hydrogen bonding associated with Q493 are stably formed in all three pairs of RBDs. The higher order assembly of S-Kappa can be attributed to the unique E484Q substitution that is primarily S-Gamma will generate the same electrostatic repulsion to prevent the dimer formation.

We next investigated the effects of mutations on the receptor ACE2 binding by biolayer interferometry (BLI). Superfold GFP-fused ACE2 ectodomain (hereafter ACE2) was immobilized on the BLI sensor to bind to different concentrations of S-Beta, S-Gamma, S-Delta and S-Kappa, all of which showed significantly enhanced ACE2 binding with respect to that of S-D614G with an increase in the dissociation constant (Kd) by 12.4, 10.3, 7.9 and 34.4-fold, respectively ( Fig. 4A and Extended Data Table 8 ).

Among the four S variants, S-Kappa bound to ACE2 the strongest, with a Kd of 0.09 nM.

To glean structural insights into the effect of mutations on the enhanced receptor binding, we determined 13 cryo-EM structures of the S variants in complex with ACE2, aided by local refinements masked around RBD and ACE2 to improve the resolution of the ACE2-binding interface (Fig. 4B , Extended Data Figs. 8-11 and Extended Data Tables 9-12). Using the aforementioned 3DVA procedure, we determined a cryo-EM structure of S-Beta in complex with ACE2 in a 3:3 binding stoichiometry (3 ACE2-bound; Fig. 4C ).

On the one hand, a minor fraction of S-Gamma, S-Delta, and S-Kappa was found to bind to ACE2 in a 3:2 stoichiometry (2 ACE2-bound). On the other hand, two distinct 3 ACE2-bound substates were observed for S-Gamma and S-Delta. By using the most representative ACE2-RBD structure of each S variant to determine the changes of solvent accessible surface area (∆SASA), we found a strong correlation between the free energy derived from the Kd values (ΔG = -RT ln Kd) and ΔSASA (Fig. 4D) . Close examination of the binding interface between RBD and ACE2 showed highly complementary surface electrostatic potentials with RBD being positively charged and ACE2 being negatively 

In light of the increased risk of immunity escape by the SARS-CoV-2 variants, antibodies that exhibit broad-spectrum neutralization activities are imperative 34, 35 . We previously reported three potent RBD-specific monoclonal nAbs against WT, D614G and Alpha variants, namely RBD-chAb-15, -25 and -45, which have non-overlapping structural epitopes 32 . The neutralizing activity of RBD-chAb-25 against the Alpha variant is lost due to the N501Y mutation while a cocktail of RBD-chAB-15 and -45 can effectively neutralize the D614G and Alpha variants with a half-maximum inhibitory concentration (IC50) value of 0.11 and 1.18 ng ml -1 , respectively. As the Beta and Gamma variants also harbor the N501Y mutation (Fig. 1, A and B) , RBD-chAb-25 could not effectively compete with ACE2 binding (Fig. 5A) . Nevertheless, RBD-chAb-25 remained highly effective in competing ACE2 binding against S-Delta and S-Kappa. Importantly, RBD-chAb-15 and -45 robustly inhibited ACE2 binding for all four S variants. Furthermore, negative staining electron microscopy (NSEM)-based structural analyses revealed that the cocktail of RBD-chAb-15 and -45 bound to all four S variants with a 3:3:3 stoichiometry in the same poses as previously reported for S-D614G (Fig. 5B, Extended Data Fig. 13 and Extended Data Table 13 ). The atomic structure of RBD in complex with BD-chAb-15 and -45 fit well with the EM maps of all four S variants bound to the antibody cocktail ( Fig. 5B and Extended Data Fig. 13) . The same analysis showed that RBD-chAb-25 bound to 36 The other study on S-Delta focuses on the binding by the NTDand RBD-specific nAbs. 37 In this case, a highly engineered HexaPro variant is used, which may alter the propensity of RBD up/down conformations. The common finding from these studies, including ours, is the reorganization of the NTD loop structure of S-Delta, accompanied by the reposition of the N-glycan on N149 ( Fig. 2A and Extended Data Fig.   6A ). The conformational changes in NTD loops and glycan shield provide an explanation for the escape from the NTD-specific nAbs 21 and the markedly reduced efficacies of the convalescent and vaccine-elicited sera 25, 26 .

In the same vein, the S-Gamma mutations that introduce the additional N-glycans on N20 and N188 introduce significant steric hindrance not only to the well-defined NTD supersite but also to an alternative epitope in turn blocking NTD-specific nAb binding ( Fig.   2C-F, Extended Data Fig. 6B and Extended Data Fig. 7) . The N20 glycosylation is indeed observed in the full-length S-Gamma based on cryo-EM data but the N188 glycosylation is not mentioned 36 . Additionally, this study only observed two distinct 1 RBD-up conformation for S-Gamma, while anther cryo-EM study also reported a single 1 RBD-up conformation 15 . In contrast, we observed an additional 2 RBD-up conformation (Fig. 1D) . Of all the S variants investigated herein, S-Beta is the only one that exhibited a fully open 3 RBD-up conformation (Fig. 1C) . Such a conformation resembles the structure of fully open S-D614G 38 , but it is not observed in previous studies 39, 40 . The difference in the number of RBD-up domains may be attributed to the construct design and the experimental conditions. To address this issue, we further verified the authenticity of the 3 RBD-up conformation of S-Beta using the furin-cleavable form without the fm2P modification ( Fig. 1C and Extended Data Fig. 2) . Perhaps the most intriguing finding in terms of the S protein structure is the dimer of trimers of S-Kappa ( Fig. 1F and Fig. 3) .

While the biological implication of such a higher order assembly remains to be established, it is consistent with the unique E484Q mutation in the Kappa and other rare lineages.

Although the reported ACE2 binding affinities vary considerably depending on the construct designs of ACE2 (monomeric versus dimeric Fc-fusion 36 or GFP-fusion used herein) and S proteins (full-length, trimeric ectodomain or monomeric RBD 36, 37, 39, 40 , our self-consistent BLI analysis showed a strong correlation between the free energy derived from the Kd values and the size of the ACE2 binding interface area (Fig. 4D) based on the high-resolution cryo-EM structures of individual S variants (Fig. 4B, 4C, and Extended Data Figs 8-12) . Furthermore, enhanced ACE2 binding can be rationalized by changes in the isoelectrostatic potential surface around the ACE2 binding site, which is highly modulated by the individual mutations (Fig. 4E) .

NTD mutations in the Beta, Gamma, Delta and Kappa variant considerably attenuate various NTD-specific nAbs and completely abolish their neutralizing activities in many cases 15, 17, 20, 21, 27, 36, 37 . Therefore, nAbs that target the RBD remain a preferred route for success, and several RBD-specific nAbs are currently approved for emergency use to treat COVID-19 patients 12 . In this context, RBD-chAb-15, -25 and -45 that we investigated here are highly potent in neutralizing all four variants with the exception for RBD-chAb-25, which is rendered ineffective due to the N501Y mutation in S-Beta and S-Gamma (Fig. 5) . Nevertheless, all three nAbs potently neutralize the most problematic 5C ). Furthermore, RBD-chAb-15 and -45 simultaneously and non-competitively bind to all three RBDs with a 3:3:3 stoichiometry (Fig. 5B and Extended Data Fig. 13A ).

Combined with our previous studies on the same set of nAbs against the original Wuhan strain 41 and the Alpha variant 32 , our current study demonstrates the broad neutralization capacity and therapeutic potential of using a cocktail of these novel nAbs. Table 1 Funding sources 

The codon-optimized DNA sequences corresponding to residues 1-1209 of S-Beta, Gamma, Delta, and Kappa variants were individually cloned into the mammalian expression vector pcDNA3.4-TOPO (Invitrogen, U. S. A.), which contains a foldon trimerization domain based on phage T4 fibritin followed by a c-Myc epitope and a hexa-repeat histidine tag as previously described 32, 42 . All constructs contained the fm2P modification, which is defined as the tandem proline replacement (2P, 986 mM NaCl, and 0.02 % NaN3). The protein concentrations were determined using the UV absorbance at 280 nm using a UV-Vis spectrometer (Nano-photometer N60, IMPLEN, Germany).

The production of recombinant sfGFP-ACE2 (hereafter ACE2) was described (UF) and blotting by filter paper, the grids were then air-dried for one day. All images were collected by using a FEI Tecnai G2-F20 electron microscope at 200 keV (FEI, the Netherlands) with a magnification of 50,000x. The resulting pixel size was 1.732 Å. Data processing was accomplished by using cryoSPARC v2.14. After patch-CTF estimation, particle images were picked by using the function of "blob picker", and extracted with a box size of 256 pixels. Particle clean-up was achieved by iterative rounds of 2D classification. Selected particles from the best 2D classes were further used for ab-initio 3D reconstruction and homogeneous refinement. The resulting NSEM maps were visualized and analyzed by UCSF-ChimeraX 43 .

For all cryo-EM grids, three microliters of protein solution at a concentration of 1.5 mg/mL were applied onto 300-mesh Quantifoil R1. 

All 2x binned super-resolution movie files were analyzed by Relion-3.0 44 with dose-weighting and 5x5 patch-based alignment using MotionCor2 (v1.2.6) 45 . All motion-corrected micrographs were transferred to cryoSPARC v2.14 46 for further processing. Contrast transfer function (CTF) estimation was achieved by patch-based CTF. The micrographs that have the "CTF_fit_to_Res" parameters between 2.5 and 4 Å were selected for subsequent particle picking. The picked particles were Fourier-cropped to a box size of 192 pixels after the particle extraction with the 384-pixels box size and then applied to iterative rounds of 2D classification for filtering junk particles.

In general, particle images were picked and classified by ab-initio reconstruction with C1 symmetry, followed by heterogeneous refinement to generate five distinct classes.

Next, the particles from the best 3D classes were applied to non-uniform (NU) refinement to generate an initial model. The initial 3D model with a refined mask from the NU refinement was used for three-dimensional variability analysis (3DVA). Five clusters generated from the 3DVA as the templates were used for further heterogeneous refinement. Particles corresponding to three of the five clusters were re-extracted, un-binned, and non-uniformly refined with local CTF refinement to yield the final cryo-EM maps (Extended Data Figs. 2-3) .

For S-Delta and S-Kappa. For S-Delta, in addition to the above image processing procedures, another round of 3DVA and heterogeneous refinement were necessary to improve the definitions of the individual RBD maps. This resulted in a total of nine cryo-EM maps for S-Delta (Extended Data Fig. 4) . For S-Kappa, 2D classification and initial heterogeneous refinement revealed two distinct classes, with one being the canonical S trimer with the one RBD-up conformation and the other being an unprecedented dimer of S trimers. The two sets of particles were analyzed separately. For the canonical S-Kappa trimer, two rounds of 3DVA and heterogeneous refinement were performed as did for S-Delta. These steps teased out a set of particles that correspond to the dimer of S trimers so they were joined with another particle pool to determine the atomic structures of the dimer of S trimers. All particles from different 3D classes were further re-extracted with a box size of 384 pixels and applied to NU refinement to generate seven cryo-EM maps for S-Kappa, including for canonical trimeric S with different RBD conformations and three for dimer of S trimers (Extended Data Fig. 5) .

A similar workflow was applied to ACE2-bound S variants. For S-Beta: ACE2, a 3.2 Å cryo-EM map of S-Beta engaged with ACE2 in a 3:3 stoichiometry (3 ACE2-bound) was obtained (Extended Data Fig. 8) . For S-Gamma, S-Delta and S-Kappa:ACE2, a 2 ACE2-bound population was separated from the dominant 3 ACE2-bound population to generate separate sets of cryo-EM maps (Extended Data Figs. 9-11) . To improve the resolution of the binding interface between RBD and ACE2 for all S variants, another round of local refinement with a focused mask covering the interface was performed. The focus masks were created by UCSF-ChimeraX 43 . This yielded a separate set of cryo-EM maps of focus-refined ACE2 in complex with RBD with nominal resolutions ranged between 3.0 and 3.0 Å (Extended Data Figs. 8-11 ).

Initial models of S-Beta, S-Gamma, S-Delta, and S-Kappa was generated based on the previously reported structure of S-D614G (PDB ID: 7EB3) aided by Swiss-Model for in silico mutagenesis 33, 47 . The atomic coordinates were divided into individual domains and manually fitted into the cryoEM map in UCSF-ChimeraX 43 , and manually optimized by Coot 48 . The same procedure was applied to S variants in complex with ACE2. The atomic structure of ACE2 in complex with S-Alpha (PDB ID: 7EDJ) was used as the initial model 32 . After iterative refinements, the models were further processed by real-space refinement in Phenix 49 to attain convergent models. N-linked glycans based on the N-glycosylation sequon (N-X-S/T) were added onto asparagine side-chains by using the extension module "Glyco" within Coot 48 

The statistics of residue-specific epitope usage of NTD-specific nAbs was done by using the 'Protein interfaces, surfaces and assemblies' service PISA at the European Bioinformatics Institute (http://www.ebi.ac.uk/pdbe/prot_int/pistart.html) 51 . The list of cryo-EM and crystal structures of nAb-bound spike or truncated NTD was manually curated by searching the EMDB/PDB entries deposited in the Protein Data Bank and their corresponding publications as tabulated in Extended Table 7 . The nAb contacting residues within NTD were defined by the function "Interfaces" of PISA and summarized in Supplementary information Table 1 . The cumulated epitope counts were mapped onto the homology model of NTD generated by Swiss-model 47 , and rendered by using PyMol 2.4.1 (Schrodinger Inc. U. S. A.).

The pseudovirus neutralization assays were performed using 293T cells overexpressing ACE2 as described previously 32 . The pseudoviruses of SARS-CoV-2 variants were prepared the by RNAi Core of Academia Sinica. Four-fold serially dilution of chAbs were premixed with 1000 TU/well Beta, Gamma, Delta, and Kappa strains of SARS-CoV-2 pseudovirus. The mixture was incubated for 1 h at 37°C and then added to pre-seeded 293T-ACE2 cells at 100 μl/well for 24 h at 37°C. The medium was removed and refilled with 100 μl/well DMEM for additional 48-h incubation. Next, 100 μl ONE-Glo™ luciferase reagent (Promega) was added to each well for 3-min incubation at 37°C. The luciferase activity was determined using a microplate spectrophotometer (Molecular Devices). The inhibition rate was calculated by comparing the luminescence value to the negative and positive control wells. IC50 was determined by a four-parameter logistic regression using GraphPad Prism 9 (GraphPad, U. S. A.).

The in-solution proteolytic digestion, LC-MS/MS analysis, glycopeptide identification and quantification of S-Beta, S-Gamma, S-Delta, S-Kappa were carried out by using the methods described previously 52 , without searching for O-glycopeptide. Nglycopeptide identification was achieved by Byonic and the quantitation was achieved by Byos (Protein Metrics Inc., USA). The built-in N-glycan library of "132 human" for Nglycopeptide identification was used with no further modifications.

To define the representative glycoform for individual N-glycosylation sites on the NTD of S-Gamma, the most abundant glycoforms derived from the aforementioned mass spectrometry analyses were used. In cases where the glycoforms could not be defined experimentally, i.e., N17 and N149 for S-Gamma, the glycoforms used in a previously reported atomistic molecular dynamics simulation study 53 were used. In the case of N20, which is unique to S-Gamma, the same glycoform as that of N17 was used. The atomic coordinates of the individual N-glycans were manually built by using Coot 48 following the same procedure as described previously 42 .

a standardized, open-source database of COVID-19 resources and epidemiology data

World Health Organization. Tracking SARS-CoV-2 variants

Centers for Disease Control and Prevention. SARS-CoV-2 Variant Classifications and Definitions

Genomics and epidemiology of the P.1 SARS-CoV-2 lineage in

Characteristics of SARS-CoV-2 variants of concern B.1.1.7, B

1: data from seven EU/EEA countries, weeks 38/2020 to 10/2021

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

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

SARS-CoV-2 neutralizing antibody structures inform therapeutic strategies

Potent Neutralizing Antibodies against SARS-CoV-2 Identified by High-Throughput Single-Cell Sequencing of Convalescent Patients' B Cells

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

Potent neutralizing antibodies against multiple epitopes on SARS-CoV-2 spike

Tackling COVID-19 with neutralizing monoclonal antibodies

COVID-19) Update: FDA Revokes Emergency Use Authorization for Monoclonal Antibody Bamlanivimab

Site-specific glycan analysis of the SARS-CoV-2 spike

Increased resistance of SARS-CoV-2 variant P.1 to antibody neutralization

A neutralizing human antibody binds to the N-terminal domain of the Spike protein of SARS-CoV-2

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

SARS-CoV-2 can recruit a heme metabolite to evade antibody immunity

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

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

The SARS-CoV-2 Delta variant is poised to acquire complete resistance to wild-type spike vaccines. biorxiv

Emerging SARS-CoV-2 variants of concern evade humoral immune responses from infection and vaccination. medRxiv

Convalescent-Phase Sera and Vaccine-Elicited Antibodies Largely Maintain Neutralizing Titer against Global SARS-CoV-2 Variant Spikes

B.1.526 SARS-CoV-2 variants identified in New York City are neutralized by vaccine-elicited and therapeutic monoclonal antibodies. mBio

Serum Neutralizing Activity of mRNA-1273 against SARS-CoV-2

Reduced neutralizing activity of post-SARS-CoV-2 vaccination serum against variants B.1.617.2, B.1.351, B.1.1.7+E484K and a sub-variant of C.37. medRxiv

SARS-CoV-2 B.1.617.2 Delta variant replication, sensitivity to neutralising antibodies and vaccine breakthrough. bioRxiv

BNT162b2-elicited neutralization of B.1.617 and other SARS-CoV-2 variants

Reduced sensitivity of infectious SARS-CoV-2 variant B.1.617.2 to monoclonal antibodies and sera from convalescent and vaccinated individuals

Reduced neutralization of SARS-CoV-2 B.1.617 by vaccine and convalescent serum

Effectiveness of Covid-19 Vaccines against the B.1.617.2 (Delta) Variant

Effect of SARS-CoV-2 B.1.1.7 mutations on spike protein structure and function

COVID-19 dominant D614G mutation in the SARS-CoV-2 spike protein desensitizes its temperature-dependent denaturation. bioRxiv

Tackling COVID-19 with neutralizing monoclonal antibodies

SARS-CoV-2 variants, spike mutations and immune escape

Membrane fusion and immune evasion by the spike protein of SARS-CoV-2 Delta variant. bioRxiv

Molecular basis of immune evasion by the delta and kappa SARS-CoV-2 variants. bioRxiv

Structural and Functional Analysis of the D614G SARS-CoV

Effect of natural mutations of SARS-CoV-2 on spike structure, conformation, and antigenicity

Structural basis for enhanced infectivity and immune evasion of SARS-CoV-2 variants

Structure-guided Antibody Cocktail for Prevention and Treatment of COVID-19

Cryo-EM analysis of a feline coronavirus spike protein reveals a unique structure and camouflaging glycans

UCSF ChimeraX: Meeting modern challenges in visualization and analysis

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

MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy

cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination

SWISS-MODEL: homology modelling of protein structures and complexes

Features and development of Coot

PHENIX: a comprehensive Python-based system for macromolecular structure solution

MolProbity: all-atom structure validation for macromolecular crystallography

Inference of macromolecular assemblies from crystalline state

Distinct shifts in site-specific glycosylation pattern of SARS-CoV-2 spike proteins associated with arising mutations in the D614G and Alpha variants. bioRxiv

Developing a Fully Glycosylated Full-Length SARS-CoV-2 Spike Protein Model in a Viral Membrane