A Potent and Protective Human Neutralizing Antibody Against SARS-CoV-2 Variants

As severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants continue to emerge and spread around the world, antibodies and vaccines to confer broad and potent neutralizing activity are urgently needed. Through the isolation and characterization of monoclonal antibodies (mAbs) from individuals infected with SARS-CoV-2, we identified one antibody, P36-5D2, capable of neutralizing the major SARS-CoV-2 variants of concern. Crystal and electron cryo-microscopy (cryo-EM) structure analyses revealed that P36-5D2 targeted to a conserved epitope on the receptor-binding domain of the spike protein, withstanding the three key mutations—K417N, E484K, and N501Y—found in the variants that are responsible for escape from many potent neutralizing mAbs, including some already approved for emergency use authorization (EUA). A single intraperitoneal (IP) injection of P36-5D2 as a prophylactic treatment completely protected animals from challenge of infectious SARS-CoV-2 Alpha and Beta. Treated animals manifested normal body weight and were devoid of infection-associated death up to 14 days. A substantial decrease of the infectious virus in the lungs and brain, as well as reduced lung pathology, was found in these animals compared to the controls. Thus, P36-5D2 represents a new and desirable human antibody against the current and emerging SARS-CoV-2 variants.


INTRODUCTION
As the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) continues to rage around the world, multiple variants have emerged and spread rapidly. Currently, four major variants of concern (VOCs) have been designated: Alpha (B.1.1.7), initially identified in the UK; Beta (B.1.351) in South Africa; Gamma (P.1) in Brazil; and Delta (B.1.617.2) in India (1)(2)(3)(4)(5)(6)(7). These VOCs are not only rapidly displacing local SARS-CoV-2 variants but are also carrying mutations in the N-terminal domain (NTD) and receptor-binding domain (RBD) on the spike protein that are critical for interactions with the ACE2 receptor and neutralizing antibodies (8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21). Conserved among variants Alpha, Beta, and Gamma, the N501Y mutation was previously shown to enhance the binding affinity to ACE2 (22,23). The Beta and Gamma variants each has three mutation sites in common within the RBD region-K417N/T, E484K, and N501Y-which change their antigenic profiles. The Delta variant contains the unique L452R and T478K mutations while sharing a common mutation with the Alpha and Gamma variants at P681 near the furin cleavage site. Various deletion mutants in the NTD have also been identified, such as 69-70del and Y144del in Alpha and 242-244del in Beta. Mutations in other spike regions were also identified, and some are located in or proximal to major S protein antigenic sites and, therefore, may adversely affect the antibody neutralization induced by natural infection or through vaccination (12)(13)(14)24). Serum from individuals vaccinated with the Moderna or Pfizer mRNA vaccine demonstrated a nearly eightfold reduction in virus neutralization capability to the Beta variant and fivefold to the Gamma variant, while retaining similar neutralizing activity to the Alpha variant compared to the wild-type virus (11,19). The chimpanzee adenovirus-based vaccine, ChAdOx1 nCoV-19, given to individuals as a two-dose regimen failed to protect against mild to moderate coronavirus disease 2019  caused by the Beta variant (25).
We and others have previously reported the isolation and characterization of several hundred RBD-specific mAbs from individuals infected with SARS-CoV-2 (26)(27)(28)(29)(30)(31)(32)(33)(34)(35), some of which have been approved for emergency use authorization (EUA) or under active clinical development. Through structure and functional characterization, the RBD-specific antibodies were classified into four major categories (8,10): class I antibodies, which are encoded by the VH3-53 gene segment with short CDRH3 loops, block ACE2 and bind only to "up" RBDs; class II neutralizing antibodies include ACE2-blocking neutralizing antibodies that bind both "up" and "down" RBDs and can contact adjacent RBDs; class III neutralizing antibodies bind outside the ACE2 site and recognize both "up" and "down" RBDs; and class IV neutralizing antibodies, which do not block ACE2 and bind only to "up" RBDs. Most class I mAbs were largely disrupted by the K417N/T mutation and class II mAbs by the E484K mutation. Class III and IV mAbs were relatively less affected by the escape mutants (12)(13)(14). Fortunately, by screening multiple mAbs from convalescent or vaccinated individuals, a few broad and potent neutralizing antibodies directed to RBD were identified to neutralize or to in vivo protect against selected VOCs (19,20,(36)(37)(38)(39). These neutralizing antibodies may contribute to the residual serum neutralizing activity against SARS-CoV-2 VOCs in these individuals.
Here, we report on the isolation and characterization of a broad, potent, and protective antibody, P36-5D2, isolated from a SARS-CoV-2 convalescent individual and capable of neutralizing the major VOCs such as the Alpha, Beta, and Gamma variants. Through crystal and electron cryo-microscopy (cryo-EM) structure analyses, P36-5D2 was found to target a conserved epitope on the RBD of the spike protein, bypassing the three key mutations (K417N, E484K, and N501Y) responsible for immune escape from many potent neutralizing mAbs (12)(13)(14). P36-5D2 also demonstrated impressive protection in a transgenic mouse model against infection with either the SARS-CoV-2 Alpha or Beta variant. Treated animals demonstrated a substantial decrease in infectious virus load in the lungs and brain and reduced lung pathology compared to the control animals. Thus, P36-5D2 serves as an important reference for the development of next-generation antibody therapies against SARS-CoV-2 infection.
To evaluate the neutralizing breadth of the top 4 neutralizing mAbs, we used the established panel of pseudoviruses including three of the most challenging SARS-CoV-2 VOCs, Alpha, Beta, and Gamma, as well as single mutants derived from the three VOCs (20). As shown in Figure 1A, Supplementary Figure S2, P36-5D2 was the most broad and potent against all the pseudotyped VOCs and variants tested, as was the class III control REGN10987. In contrast, P36-1B7, like the class I mAb CB6, was substantially impacted by the SARS-CoV-2 Alpha, Beta, and Gamma variants, largely due to K417N/T and N501Y mutations. P74-6D2 demonstrated reduced or loss of neutralizing activities to the three VOCs, which was largely attributed to E484K, similar to the class II control mAb BD368-2 and the class I control mAb REGN10933. P36-1A3 appeared to be heavily affected by N501Y and K417-E484K-N501Y mutations. Furthermore, the neutralizing patterns against these pseudotyped VOCs and variants correlated well with their binding avidity to cell surface-expressed VOCs and the mutant spike proteins ( Figure 1B, Supplementary Figure S3). This result suggested that a compromised binding avidity is a major escape mechanism.
Subsequently, we evaluated the neutralizing ability of P36-5D2 against the infectious SARS-CoV-2 WT, Alpha, and Beta. Consistent with that from pseudoviruses, P36-5D2 showed broad neutralizing activity to all three infectious viruses, with IC 50 values of 0.060, 0.051, and 0.034 mg/ml against the SARS-CoV-2 WT, Alpha, and Beta respectively ( Figure 1C). The dissociation constant (K D ) of P36-5D2 with the SARS-CoV-2 RBD was 6.91 nM, measured using surface plasmon resonance (SPR), which was similar to those of P36-1A3, P36-1B7, and P74-6D2 (Supplementary Figure S1B). Through competition analysis measured by SPR, we found that P36-5D2 partially competed while P36-1A3, P36-1B7, and P74-6D2 completely competed with ACE2 for binding to RBD (Supplementary Figure S1C). Competition among the four mAbs further showed that P36-5D2 recognized a less overlapping epitope, while the other three strongly competed among each other (Supplementary Figures S1D, E). These results indicated that P36-5D2 recognized a different epitope from those by the other three mAbs.

P36-5D2 Binds to a Highly Conserved
Epitope on RBD and Avoids Three Key Mutant Residues (K417N, E484K, and N501Y) To reveal the molecular basis of the broad and potent neutralizing activity of P36-5D2, we determined the crystal structure of its antigen-binding fragment (Fab) bound to the RBD-3M carrying the K417N, E484K, and N501Y mutations initially identified in the SARS-CoV-2 Beta variant. At a resolution of 3.1 Å, we found that P36-5D2 recognized an epitope consisting of 11 residues (T345, R346, L441, K444, V445, G446, G447, Y449, N450, T470, and F490) on RBD, devoid of the three key mutant residues K417N, E484K, and N501Y that facilitated escape from the neutralization of many mAbs, including some approved for EUA ( Figure 2A). Over 99.5% conservation was found at all these sites among the 1.96 million spike sequences in the GISAID EpiCoV database collected from December 2019 to July 2021. The paratope of P36-5D2 consisted of nine heavy-chain residues-T30 and T31 of HCDR1; G54 and K59 of HCDR2; and R103, Q105, F106, D107, and Y108 of HCDR3-as well as six light-chain residues: W32 of LCDR1; D50 of LCDR2; and Y91, N92, G93, and Y94 of LCDR3 ( Figure 2B). At the binding interface, R346 of the RBD had salt bridge interaction with D107 of HCDR3 and D50 of LCDR2, and K444, G446, and N450 had hydrogen bond interactions with residues Y91 of LCDR3 and K59, F106, and D107 of HCDR2 and HCDR3 (Figures 2C, D and Supplementary Table S2). Based on the binding specificity, P36-5D2 was closest to the class III RBD-targeting mAbs, including REGN10987, S309, C110, and C135 (8,27,30,41). IC50 represents the half-maximal concentration and IC90 the 90% inhibitory concentration in the pseudovirus and infectious SARS-CoV-2 neutralization assay. Antibody binding to RBD was presented either by K D or by competing with ACE2, in which +++ indicates >70% competition, ++ means 50%-70%, + denotes 20%-50%, and − indicates <20% competition. The program IMGT/V-QUEST was applied to analyze the gene germline, CDR3 length, and somatic hypermutation (SHM). The CDR3 length was calculated from the amino acid sequences. The SHM frequency was calculated from the mutated nucleotides. RBD, receptor-binding domain; IGHV, immunoglobulin heavy-chain variable region gene; n.a. not available.
The P36-5D2 epitope overlapped more with that of REGN10987 and C110 than with S309 and C135 ( Figures 2E-H). All these five class III mAbs shared two residues, R346 and L441, among their epitopes. Interestingly, P36-5D2, C110, and C135 had lightchain variable regions all derived from germline IGKV 1-5, indicating a selective preference for the particular germline gene during the development of this class of mAbs from different infected individuals. We went further to determine the cryo-EM structures of the P36-5D2 Fab bound to the SARS-CoV-2 spike trimer at 3.7-Å resolution ( Figure 2I). In one model, P36-5D2 Fabs bound two "up" RBDs and one "down" RBD. In the other model, P36-5D2 Fabs bound two "down" RBDs and one "up" RBD of the spike trimer. Consistent with the crystal structure analysis, the cryo-EM structure clearly demonstrated that the mutant residues K417N, E484K, and N501Y were not involved in the P36-5D2 epitope. Furthermore, P36-5D2 and ACE2 only had two overlapping residues at G446 and Y447, which was sufficient to exert spatial hindrance in disrupting the interaction between ACE2 and RBD. These results collectively revealed the highly conserved nature of the P36-5D2 epitope and provided a molecular basis for its broad and potent neutralizing activity against all the SARS-CoV-2 VOCs and mutant strains tested here.
Next, we conducted a single-site alanine scanning mutagenesis for the 11 epitope residues within the P36-5D2 epitope to identify the key residues that mediate RBD binding. All 11 mutated spikes were successfully expressed on the surface of human embryonic kidney (HEK) 293T cells and can bind to the cellular receptor ACE2 ( Figures 3A, B, E). Among the mutated residues, R346A, K444A, G446A, and N450A resulted in a near loss of P36-5D2 binding ( Figures 3A, C). Two resides, K444A and V445A, also resulted in a substantial reduction in REGN10987 binding ( Figure 3D), consistent with the epitope residues defined by structural analysis. Furthermore, among the 11 pesudoviruses bearing alanine mutated spikes, nine were able to confer resistance to P36-5D2, while the remaining two (L441A and T470A) had a relatively weaker impact , and the respective mutations derived from the three variants of concern (VOCs) were evaluated. Values indicate the fold changes in half-maximal inhibitory concentrations (IC 50 ) (A) and the mean fluorescence intensity (MFI) relative to that of wild-type (WT) D614G (B). The IC 50 of antibodies against WT D614G are listed in (A). The minus symbol represents increased resistance and the plus sign indicates increased sensitivity. IC 50 or MFI highlighted in red indicates that resistance increased at least threefold; in blue, sensitivity increased at least threefold. BDL (below detection limit) represents the highest concentration of mAbs that failed to reach 50% potency in neutralization activity or mAbs that failed to bind the cell surface-expressed SARS-CoV-2 variants. Results are presented as the mean value from three independent experiments. (C) Neutralization of P36-5D2 against the infectious SARS-CoV-2 WT, Alpha, and Beta variants. 2G4 is the negative control antibody specific for Ebola virus glycoprotein (EBOV GP). Experiments were performed in triplicate, and all data were presented as the means ± SEM.
( Figure 3F). These results suggested that the levels of binding to the surface-expressed spikes could not be accounted for during the entire neutralization of P36-5D2.

P36-5D2 Protects K18-hACE2 Mice From Infectious SARS-CoV-2 Alpha or Beta Infection
Thereafter, we evaluated the prophylactic potential of P36-5D2 against infectious SARS-CoV-2 Alpha or Beta infection in a wellestablished hACE2 transgenic mouse model (42,43). The entire experimental protocol and assays conducted to evaluate the protection capability of P36-5D2 are outlined in Figure 4A. Briefly, 12 K18-hACE2 mice were IP administered with P36-5D2 at a dose of 10 mg/kg body weight a day before the intranasal (IN) challenge with 10 3 plaque-forming units (PFU) SARS-CoV-2 Alpha or Beta variant. The mice were monitored for 14 days to record their body weight and symptoms. Half of them were euthanized at 4 days post-infection (dpi) to obtain the right lung and brain for viral titration and the left lung for histological staining ( Figure 4A).
whereas those without treatment suffered drastic loss of body weight starting from 5 dpi and reached humane endpoint at 7 dpi ( Figures 4B, C). The viral titers found at 4 dpi for the untreated group were on average 3.3 × 10 5 ± 1.5 × 10 5 PFU/tissue and brain 7.2 × 10 5 ± 10 6 PFU/tissue. In contrast, the P36-5D2-treated group showed minimum detectable levels of viruses in the lung and brain tissues ( Figures 4F, G). To assess the extent of lung damage and pulmonary inflammation of the virus infection, we performed histopathological analysis on the lung sections using hematoxylin and eosin (H&E) staining. The untreated group infected by the SARS-CoV-2 Alpha variant revealed evidence of moderate to severe lung damage and inflammation with marked infiltration of inflammatory cells such as granulocytes. In contrast, in the P36-5D2 group, the lung tissue remained intact and well defined ( Figure 4J). For SARS-CoV-2 Beta infection, the untreated group exhibited progressive body weight loss from 3 to 5 dpi, and all succumbed to infection by 5 dpi. These results indicated that the disease progressed faster and with greater severity caused by SARS-CoV-2 Beta as compared to SARS-CoV-2 Alpha infection. Yet, P36-5D2 still conferred complete protection and the mice maintained stable body weights, except for one mouse experiencing a slight decline with mild symptoms at 6 dpi and recovering by 8 dpi (Figures 4D, E). In the untreated group, the viral titers found at 4 dpi were on average 789 ± 420 PFU/tissue and brain 7.0 × 10 4 ± 10 5 PFU/tissue. The prophylactic use of P36-5D2 against the SARS-CoV-2 Beta variant was able to reduce lung viral titers for 2 logs compared to that in the control animals. The brain titers of the P36-5D2 group were undetectable, except for one mouse ( Figures 4H, I). Evaluation of H&E staining revealed that the lung sections from the untreated group infected by the Beta variant showed similar histopathological changes to those infected by the Alpha variant, with severe inflammatory infiltrates and edema. In contrast, in the P36-5D2 group, there was no damage or pulmonary inflammation in lung tissues ( Figure 4K). These results strongly indicate the broad and potent neutralizing activity of P36-5D2 against the SARS-CoV-2 Alpha and Beta variants in vivo via prophylactic interventions.

DISCUSSION
We report here the structural and functional characterization of the human neutralizing antibody P36-5D2 isolated from SARS-CoV-2 convalescent individuals. Compared to many other isolated human mAbs, the most outstanding and unique feature of P36-5D2 is its breadth and potency against the SARS-CoV-2 Alpha, Beta, and Gamma variants, as well as mutant viruses derived from the variants. Passive delivery of P36-5D2 protected K18-hACE2 mice from infection with the SARS-CoV-2 Alpha or Beta variant and significantly reduced the viral loads in both the lungs and brain. Crystal and cryo-EM structural analyses showed that P36-5D2 recognized a highly conserved epitope on RBD, which is accessible in both "up" and "down" conformations and able to withstand the mutated residues K417N, E484K, and N501Y that compromised many neutralizing mAbs, including some approved for EUA. Recently, Hastie and colleagues conducted a comprehensive study to map the landscape of neutralizing epitopes on the RBD of SARS-CoV-2 (44). A total of 186 RBD-directed antibodies with potential therapeutic use were categorized into seven core "communities" (RBD-1 to RBD-7) based on their distinct footprints and broad competition profiles. RBD-1, RBD-2, and RBD-3 antibodies targeted the receptor-binding surface, while RBD-4 and RBD-5 antibodies bound to the outer face of RBD. RBD-6 and RBD-7 antibodies, on the other hand, bound to the inner face of RBD. In this context, P36-5D2 would fall into the community of RBD-5 antibodies together with REGN10987 and S309. Consistent with what we have found here, RBD-5 antibodies showed broad resistance to the SARS-CoV-2 variants, suggesting that this community of antibodies binds to the conserved epitopes across all variants. Similar to REGN10987 and S309, P36-5D2 barely competed with ACE2 for binding to RBD, suggesting that their mechanism of neutralization may involve immunoglobulin G (IgG)-mediated spike crosslinking on virions, steric hindrance, or aggregation of virions (41). Nevertheless, these results indicated that we have identified a broad and potent neutralizing antibody capable of overcoming the major circulating SARS-CoV-2 variants. P36-5D2 and the epitope it recognized may serve as an important reference for the development of next-generation antibody therapies and vaccines against SARS-CoV-2 infection.
A couple of points need to be highlighted here. Firstly, despite the rapid emergence and spread of the SARS-CoV-2 variants, infected and convalescent individuals can generate and produce RBD-specific antibodies with impressive breadth and potency. The number and relative proportion of such antibodies appeared to be small, but they may contribute substantially to the remaining serum neutralizing activity to the variants in recovered or vaccinated individuals, particularly after boosting with mRNA vaccines (24,45,46). Comprehensive characterization of this class of antibodies will provide us with deeper insights into their ontogeny and potential ways of inducing broader and more effective protection against the circulating and future variants. Secondly, although many of the current SARS-CoV-2 variants bear mutations in the RBD, the identification of P36-5D2 has revealed the existence of a highly conserved sequence and structure on RBD that could potentially be used to trigger more broad and potent immune responses, avoiding the current prevalent mutations in dampening the immune responses generated during natural infection and vaccination. This would require preferentially exposing the conserved regions on RBD while minimizing the receptor-binding motif that is commonly recognized during infection and vaccination. How to achieve this design and, more importantly, to realize the desired immune response will surely be a big challenge. However, recent advances in structure-based vaccine design through understanding of the antigen-antibody interaction will certainly offer new possibilities for this highly anticipated outcome. The identification of P36-5D2, as well as other similar mAbs (19,20,(36)(37)(38)(39), represents the first but important step for us to achieve the ultimate goal of developing a universal vaccine against all SARS-CoV-2 variants and beyond.

Study Approval
This study received approval from the Research Ethics

Antibody and Fab Production
Antibody and Fab production was conducted as previously described (47,48

Isolation of Spike-Specific Single B Cells by FACS
Spike-specific single B cells were sorted as previously described (28).  FBS) and resuspended in 500 ml FACS buffer before being stained with propidium iodide (PI) (eBioscience). Spike-specific single B cells were gated as live + CD3 − CD14 − CD16 − CD235a − CD20 + CD27 + IgG + Spike + and sorted into 96-well PCR plates containing 2 ml of lysis buffer (1.9 ml 0.2% Triton X-100, 0.1 ml RNase inhibitor; Clontech, Mountain View, CA, USA) per well. The plates were then snap-frozen on dry ice and stored at −80°C until the reverse transcription reaction.

Single B-Cell PCR and Construction of Antibody Genes
The IgG heavy-and light-chain variable genes were amplified by nested PCR and cloned into linear expression cassettes or expression vectors to produce full IgG1 antibodies as previously described (28). Specifically, all second-round PCR primers containing tag sequences were used to produce the linear Ig expression cassettes by overlapping PCR. Separate primer pairs containing the specific restriction enzyme cutting sites (heavy chain, 5′-AgeI/3′-SalI; kappa chain, 5′-AgeI/3′-BsiWI; and lambda chain, 5′-AgeI/3′-XhoI) were used to amplify the cloned PCR products. The PCR products were purified and cloned into the backbone of the antibody expression vectors containing the constant regions of human IgG1. Overlapping PCR products of paired heavy-and light-chain expression cassettes were co-transfected into HEK 293F cells.

Neutralization Activity of mAbs Against Pseudotyped SARS-CoV-2
WT and mutated SARS-CoV-2 pseudoviruses were generated by co-transfection of human immunodeficiency virus backbones expressing firefly luciferase (pNL43R-E-luciferase) and pcDNA3.1 (Invitrogen) expression vectors encoding the respective spike proteins into 293T cells (ATCC) (20). Viral supernatants were collected 48 h later. Viral titers were measured as the luciferase activity in relative light units (Bright-Glo Luciferase Assay Vector System; Promega Biosciences, WI, USA). Serial dilutions of mAbs were prepared with the highest concentration of 5 or 50 mg/ml. WT or mutated spike pseudoviruses were mixed with mAbs and incubated at 37°C for 1 h. HeLa-hACE2 cells (1.5 × 10 4 per well) were then added into the mixture and incubated at 37°C for 60 h before cell lysis for measuring the luciferase activity. The percent of neutralization was determined by comparing with the virus control.

Neutralization Activity of mAbs Against Infectious SARS-CoV-2
A plaque reduction neutralization test against infectious SARS-CoV-2 WT, shown in Supplementary Figure S1A, was performed in a Chinese certified BSL-3 laboratory. Neutralization assays against infectious SARS-CoV-2 WT were conducted using a clinical isolate (Beta/Shenzhen/SZTH-003/ 2020, EPI_ISL_406594 at GISAID) previously obtained from a nasopharyngeal swab of an infected patient (28). Serial dilutions of the test antibodies were conducted, mixed with 75 ml of SARS- The plaque assay against infectious SARS-CoV-2 VOCs, shown in Figure 1C, was performed in the National University of Singapore certified BSL-3 laboratory. The infectious SARS-CoV-2 used in the neutralization belongs to GISAID lineage clade L (WT, B) (accession ID: EPI_ISL_574502), GRY (Alpha, B.1.1.7) (accession ID: EPI_ISL_754083), and GH (Beta, B.1.351.3) (accession ID: EPI_ISL_1173248). Serial dilutions of the test antibodies were conducted, mixed with 50 PFU infectious SARS-CoV-2 in 12-well plates, and incubated for 1 h at 37°C. The mixtures were then transferred into 12-well plates, seeded with Vero E6 cells, and allowed absorption for 1 h at 37°C. The inocula were then removed and the wells washed once with PBS before adding the overlay media [1 ml DMEM containing 1.2% microcrystalline cellulose (MCC)]. The plates were then incubated at 37°C, 5% CO 2 for 72 h for plaque formation. Cells were fixed in 10% formalin overnight before counterstaining with crystal violet. The virus titer of each dilution was determined through the number of plaques formed and expressed in neutralization percentage in comparison to positive control.

Antibody Binding Kinetics, Competition with Receptor ACE2, Epitope Mapping Measured by SPR
The binding kinetics of mAbs to SARS-CoV-2 RBD were analyzed using SPR (Biacore 8K; GE Healthcare, Chicago, IL, USA). Specifically, recombinant protein A (Sino Biological) or the anti-His antibody (Cytiva, Marlborough, MA, USA) was covalently immobilized to a CM5 sensor chip via amine groups in 10 mM sodium acetate buffer (pH 4.5) for a final RU (response units) around 7,000. The running buffer HBS-EP was composed of 0.01 M HEPES, pH 7.4, 0.15 M NaCl, and 0.05% (v/v) Tween-20. The IgG form of mAbs was captured by the sensor chip immobilized with recombinant protein A, and then serial dilutions of SARS-CoV-2 RBD flowed through the sensor chip system. The resulting data were fitted to a 1:1 binding model using the Biacore 8K Evaluation software (GE Healthcare).
To determine competition with the human ACE2, the SARS-CoV-2 RBD was immobilized to a CM5 sensor chip via the amine group for a final RU around 250. In the first round, antibodies (1 mM) were injected onto the chip for 120 s to reach binding steady state and HBS-EP was injected for 120 s. In the second round, antibodies (1 mM) were injected onto the chip for 120 s and ACE2 (2 mM) was then injected for 120 s. In the third round, HBS-EP was injected for 120 s and ACE2 (2 mM) was then injected for 120 s. The sensorgrams of the three rounds were aligned from 120 to 240 s in Biacore 8K Evaluation software (GE Healthcare). The blocking efficacy was determined by a comparison of the response units with and without prior antibody injection.
For epitope mapping, the SARS-CoV-2 RBD was immobilized to a CM5 sensor chip via the amine group for a final RU around 250. Two different mAbs were sequentially injected and monitored for binding activity to determine whether these two mAbs recognized separate or closely situated epitopes.

Binding of mAbs to Cell Surface-Expressed WT and Mutated Spikes
The entire procedure was conducted as previously described (20). HEK 293T cells were transfected with expression plasmids encoding either WT or mutated full-length spike of SARS-CoV-2, and incubated at 37°C for 36 h. Cells were removed from the plate using trypsin and distributed into 96-well plates for individual staining. Cells were washed twice with 200 ml staining buffer (PBS with 2% FBS) between each of the following. Firstly, cells were stained with 2 mg/ml testing mAb and 5 mg/ml S2-specific monoclonal antibody (MP Biomedicals, Santa Ana, CA, USA) at 4°C for 30 min in 100 ml staining buffer. Then, PE-labeled antihuman IgG Fc (BioLegend) at a 1:100 dilution and anti-mouse IgG FITC (Thermo Fisher Scientific) at a 1:200 dilution were added into 40 ml staining buffer at room temperature for 30 min. After extensive washes, the cells were resuspended and analyzed with BD LSRFortassa (BD Biosciences) and FlowJo v10 software (FlowJo). HEK 293T cells without transfection were also stained as a background control. The RBD-specific mAbs REGN10933, CB6, BD368-2, and REGN10987 were used as positive control mAbs, while 2G4 was used as a negative control mAb. HEK 293T cells with mock transfection were stained as a background control.

Antibody Protection in hACE2 Transgenic Mice
Eight-week-old female K18-hACE2 transgenic mice (InVivos Ptd Ltd, Lim Chu Kang, Singapore) were used for this study. The mice were housed and acclimatized in an ABSL-3 facility for 72 h prior to the start of the experiment.
K18-hACE2 transgenic mice were subjected to pretreatment of mAb P36-5D2 (10 µg/kg) delivered through IP injection a day prior to infection. The inoculation of mice was conducted through IN delivery with 25 µl, 10 3 PFU of the infectious SARS-CoV-2 Alpha or Beta variant. Baseline body weights were measured prior to infection and monitored daily by two personnel post-infection for the duration of the experiment. To assess the viral load, mice from each experimental group were sacrificed 4 dpi, with brain and lung tissues harvested. Each organ was halved for the plaque assay and histology, respectively. Tissues were homogenized with 0.5 ml DMEM supplemented with antibiotic and antimycotic (Gibco, Waltham, MA, USA) and titrated in Vero E6 cells using plaque assays.
For virus titer determination, supernatants from homogenized tissues were diluted 10-fold serially in DMEM supplemented with antibiotic and antimycotic. Of each serial diluted supernatant, 250 µl was added to Vero E6 cells into 12-well plates. After 1 h of incubation for virus adsorption, the inoculum was removed and washed once with PBS. About 1.2% MCC-DMEM supplemented with antibiotic and antimycotic overlay media was added to each well and incubated at 37°C, 5% CO 2 for 72 h for plaque formation. The cells were then fixed in 10% formalin overnight and counterstained with crystal violet. The number of plaques was determined and the virus titers of individual samples were expressed in logarithm of PFU per organ.

Histopathological Analyses
Left lung lobes were fixed in 3.7% formaldehyde solution prior to removal from BSL-3 containment. The tissues were routinely processed, embedded in paraffin blocks (Leica Surgipath Paraplast), sectioned at 5-µm thickness, and stained with H&E (Thermo Scientific) following standard histological procedures. The extent and severity of lung damage was qualitatively described as no infection, mild, moderate, or severe.

Crystal Analysis and Structural Determination
The antibody P36-5D2 Fab and RBD-3M were mixed at a molar ratio of 1:1.5 and incubated at 4°C for 60 min. The mixture was purified by gel filtration pre-equilibrated by 1× HBS buffer. The complex of P36-5D2 Fab and RBD-3M was concentrated to 11 mg/ml for crystallization. Crystal of the P36-5D2 Fab and RBD-3M complex was obtained after 5 days using the sitting drop method. The well solution was 0.1 M sodium HEPES 7.5, 10% (w/v) PEG 6000, and 5% (v/v) MPD. Diffraction data were collected at SSRFBL18U1 beam line of the Shanghai Synchrotron Research Facility (SSRF). The data were processed by HKL3000 and the structure was determined using the molecular replacement method with PHASER in the CCP4 suite (50). Model building and refinement were performed using COOT v.0.9.2 and PHENIX v.1.18.2, respectively (51,52). The data processing statistics are listed in Supplementary  Table S1.
Images for complex were recorded using the FEI Titan Krios microscope (Thermo Fisher Scientific) operating at 300 kV with a Gatan K3 Summit direct electron detector (Gatan Inc., Pleasanton, CA, USA) at Tsinghua University. The automated software AutoEMation2 (53) was used to collect 4,534 movies for complex of SARS-CoV-2 spike ectodomains and P36-5D2 Fab in super-resolution mode at a nominal magnification of ×29,000 and at a defocus range between −1.5 and −1.8 mm. Each movie has a total accumulated exposure of 50 e-/Å 2 fractionated in 32 frames of 2.13-s exposures. The final image was binned twofold to a pixel size of 0.97 Å. The data collection statistics are summarized in Supplementary Table S3.

Cryo-EM Data Processing
Motion correction (MotionCor2 v.1.2.6) (54), CTF estimation (GCTF v.1.18) (55), and non-templated particle picking (Gautomatch v.0.56; http://www.mrc-lmb.cam.ac.uk/kzhang/) were automatically executed using the TsingTitan.py program. Sequential data processing was carried out on RELION-3.1 (56). Initially,~910,000 particles were subjected to 2D classification. After three additional 2D classifications, the best selected 480,469 particles were applied for the initial model and 3D classification. A subset of 242,871 particle images from state 1 (one RBD up and two RBDs down) and 225,216 particle images from state 2 (two RBDs up and one RBD down) were further subjected to 3D auto-refine and post-processing. The final resolutions for states 1 and 2 were 3.69 and 3.65 Å, respectively. The interface between the S protein of SARS-CoV-2 and Fab was subjected to focused refinement with mask on the region of the RBD-Fab complex to improve the map quality. The selected 480,469 particles were 3D classified focused on the RBD-Fab complex. Then, the good particles were selected for focused refinement and postprocessing with a final resolution of 3.8 Å. The resolution was estimated with the gold-standard Fourier shell correlation 0.143 criterion. Details of the data collection and processing are shown in Supplementary Figure S4 and Table S3.

Cryo-EM Model Building and Refinement
The initial model of complex of SARS-CoV-2 spike ectodomains and P36-5D2 Fab was generated using the models (PDB 7A94 and 7A97) and fit into the map using UCSF Chimera v.1.15 (57). Manual model rebuilding was carried out using COOT v.0.9.2 (51) and refined with PHENIX v.1.18.2 (52) real-space refinement. The quality of the final model was analyzed with PHENIX v.1.18.2 (52). The validation statistics of the structural models are summarized in Supplementary Table S3. All structural figures were generated using PyMOL 2.0 (58) and Chimera v.1.15 (57).

Quantification and Statistical Analysis
The technical and independent experiment replicates were indicated in the figure legends. The half-maximal inhibitory concentration (IC 50 ) and 90% inhibitory concentration (IC 90 ) of the mAbs were calculated using the four-parameter dose inhibition equation in Graphpad Prism 9.0. Percentages of the binding of mAbs to cell surface-expressed SARS-CoV-2 variants were calculated as the ratio between the mutated over WT mean fluorescence intensity (MFI) normalized relative to that of the S2-specific antibody. All MFI values were weighted by multiplying the number of positive cells in the selected gates.
The fold change of the mutant spike relative to WT D614G in binding or neutralization was calculated by simple division of the respective IC 50 or MFI values. In animal experiments, a twotailed unpaired t-test was used to assess statistical significance. Statistical calculations were performed in GraphPad Prism 9.0. Differences with p-values less than 0.05 were considered to be statistically significant (*p < 0.05, **p < 0.01; n.s., not significant).

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the Research Ethics Committee of Beijing Youan Hospital, China (LL-2020-039-K) and Shenzhen Third People's Hospital (2020-084). The patients/participants provided written informed consent to participate in this study. The animal study was reviewed and approved by the National University of Singapore (NUS) Institutional Animal Care and Use Committee (IACUC) and the NUS Institutional Biosafety Committee (IBC) and NUS Medicine BSL-3 Biosafety Committee (BBC).

AUTHOR CONTRIBUTIONS
LZ, JC, XW, and JW conceived, designed, and supervised the entire study. SS and JLi did antibody isolation and sequencing under the conduction of JW. SS and ZY did antibody binding analysis and pseudovirus neutralization. CM, ZA, and JC performed the antibody protection experiment in K18-hACE2 mice. SZ did the cryo-EM structure of the Fab-spike complex. JLa collected the crystal structure of the Fab-RBD complex. CM, ZA, LC, and ZZ carried out the infectious virus neutralization assays. JY provided the SARS-CoV-2 spike protein. ZZ, TZ, XS, QZ, and RW conducted the antibody evaluation, structural analysis, and animal experiment. SS, CM, SZ, JLa, JLi, and LZ had full access to the data in the study, generated the figures and tables, and took responsibility for the integrity and accuracy of the data presentation. LZ, SS, CM, XW, and ZA wrote the original draft. LZ, JC, XW, and JW reviewed and edited the manuscript. All authors contributed to the article and approved the submitted version.

ACKNOWLEDGMENTS
We acknowledge the work and contributions of all healthcare providers from Beijing Youan Hospital and Shenzhen Third People's Hospital. We also thank the affected individuals for their active participation. We thank the SSRF BL17U1 beamline for data collection and processing. We are grateful to the NUS Yong Loo Lin School of Medicine BSL-3 Core Facility for their support with this work.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2021.766821/ full#supplementary-material Supplementary Figure 1 | SARS-CoV-2 spike-specific neutralizing antibodies screening and evaluation. (A) Antibody neutralization against pseudotyped and infectious SARS-CoV-2. Antibodies included eight isolated neutralizing mAbs P33-1F1, P33-3C5, P36-1A3, P36-1B7, P36-1C5, P36-5D2, P36-8F2 and P74-6D2 from SARS-CoV-2 infected individuals, four RBD-specific representative mAbs REGN10933, CB6, BD368-2, REGN10987, one NTD-specific representative mAb 4A8 and one negative control mAb 2G4. Data are presented as the means ± SEM from three independent experiments. (B) Binding kinetics of the top four isolated neutralizing mAbs P36-1A3, P36-1B7, P36-5D2 and P74-6D2 with SARS-CoV-2 RBD measured by SPR. The black lines indicate the experimentally derived curves while the red lines represent fitted curves based on the experimental data. Results presented are representatives of two independent experiments. (C) Antibody competition with ACE2 for binding to SARS-CoV-2 RBD measured by SPR. The plots show distinct binding patterns of ACE2 to the RBD with (colored curve) or without (black curve) prior incubation with each tested mAb. The results are representative of two independent experiments and color-coded for each mAb. (D) Epitope mapping through competitive binding measured by SPR. Pairs of antibodies were sequentially applied to SARS-CoV-2 RBD immobilized sensor chip. The level of reduction in response unit comparing with or without prior antibody incubation is the criterion for determining the two antibodies recognize the separate or closely situated epitopes. (E) Summary of antibody competition in (D), in which '+++' indicates >70% competition; '++' 50-70%; '+' 20-50%; and '−' <20%.
Supplementary Figure 2 | Neutralization of SARS-CoV-2 variants by each antibody, related to Figure. 1. Pseudoviruses bearing the indicated mutations were tested against serial dilutions of each mAb. Neutralizing activity was defined as the percent reduction in luciferase activities compared to no antibody controls. Levels of resistance were calculated as the -fold change in IC 50 between each mutant and WT D614G, as presented in Figure. 1A. Results are presented as the mean value from three independent experiments.
Supplementary Figure 3 | Binding to cell surface expressed SARS-CoV-2 variants by each antibody, related to Figure. 1. Wildtype and mutated spike were expressed on the surface on HEK 293T, incubated with the mAbs, followed by staining with anti-human IgG Fc PE and analysed by FACS. The gated cell percentages are shown. The fold changes in antibody binding, as shown in Figure. 1B, was determined by comparing the total MFI in the selected gate between spike variants and WT D614G. S2 mAb is a positive control antibody used for spike expression normalization. 2G4 targeting EBOV GP is negative control antibody. NC is HEK 293T cells with mock transfection.
Supplementary Figure 4 | Cryo-EM data processing workflow. (A) Processing workflow of the P36-5D and SARS-CoV-2 spike complex cryo-EM data. Local resolution map and particle orientation distribution of the P36-5D/spike complex with 2 "down" and 1 "up" RBD (B) or with 1 "down" and 2 "up" RBD (C). (D, E) The corrected, unmasked, masked and phase randomised FSC for the cryo-EM reconstructions of the density maps of P36-5D2/spike complex with C3 symmetry. The final resolution of the P36-5D/spike complex with 2 "down" and 1 "up" RBD is 3.69 Å (D). The final resolution of the P36-5D/spike complex with 1 "down" and 2 "up" RBD is 3.65 Å (E).