A potent neutralizing human antibody reveals the N-terminal domain of the Spike protein of SARS-CoV-2 as a site of vulnerability

The pandemic of coronavirus disease 2019 (COVID-19) caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) presents a global public health threat. Most research on therapeutics against SARS-CoV-2 focused on the receptor binding domain (RBD) of the Spike (S) protein, whereas the vulnerable epitopes and functional mechanism of non-RBD regions are poorly understood. Here we isolated and characterized monoclonal antibodies (mAbs) derived from convalescent COVID-19 patients. An mAb targeting the N-terminal domain (NTD) of the SARS-CoV-2 S protein, named 4A8, exhibits high neutralization potency against both authentic and pseudotyped SARS-CoV-2, although it does not block the interaction between angiotensin-converting enzyme 2 (ACE2) receptor and S protein. The cryo-EM structure of the SARS-CoV-2 S protein in complex with 4A8 has been determined to an overall resolution of 3.1 Angstrom and local resolution of 3.4 Angstrom for the 4A8-NTD interface, revealing detailed interactions between the NTD and 4A8. Our functional and structural characterizations discover a new vulnerable epitope of the S protein and identify promising neutralizing mAbs as potential clinical therapy for COVID-19.

The trimeric S protein covers the surface of coronavirus and plays a pivotal role during viral entry (20,21). The S protein is cleaved into the N-terminal S1 subunit and C-terminal S2 subunit 50 by host proteases such as TMPRSS2 (21,22) and undergoes conformational change from prefusion to postfusion state during infection (23). S1 and S2 mediate receptor binding and membrane fusion, respectively (18). S1, consisting of the N-terminal domain (NTD) and the receptor binding domain (RBD), is critical in determining tissue tropism and host ranges (24,25). The RBD is responsible for binding to ACE2, while the function of NTD is not well 55 understood. In some coronaviruses, the NTD may recognize specific sugar moieties upon initial attachment and might play an important role in the prefusion to postfusion transition of the S protein (26)(27)(28)(29). The NTD of the MERS-CoV S protein can serve as a critical epitope for antibody neutralizing (29).
The SARS-CoV-2 S protein-targeting monoclonal antibodies (mAbs) with potent neutralizing activity have become the focus of therapeutic interventions for COVID-19 (30)(31)(32). Most SARS-CoV-2 antibodies reported previously were sorted to target the RBD in order to inhibit the association between the S protein and ACE2 (31)(32)(33)(34). The RBD-targeting antibodies, when applied individually, may induce viral escaping mutations, thus fostering evolution of new viral 65 strains that are insensitive to these antibodies (29). Antibodies targeting non-RBD regions may serve as "cocktail" therapeutics for SARS-CoV-2.
Here we report the isolation and characterization of S protein-specific monoclonal antibodies derived from memory B or plasma B cells of COVID-19 survivors. We also determined the 70 structure of the complex between the SARS-CoV-2 S protein and 4A8, one of the isolated mAbs, at an overall resolution of 3.1 Å and local resolution of 3.4 Å for the interfaces between 4A8 and the S protein. 4A8 targets the S-NTD with potent neutralizing activity. These findings reveal a new epitope of the SARS-CoV-2 S protein for antibody neutralizing, which may lay the foundation for new prophylactic and therapeutic interventions for SARS-CoV-2.

Isolation of human mAbs from memory B cells and plasma B cells
To isolate monoclonal antibodies and analyze the humoral antibody responses to SARS-CoV-2, we collected plasma and peripheral blood mononuclear cells (PBMCs) from 10 patients 80 recovered from SARS-CoV-2 infection. The age of donors ranges from 25 to 53 years. The interval from blood collection date to disease confirmation date ranged from 23 to 29 days for No. 1-5 patients and 10 to 15 days for No. 6-10 patients (Table S1). We evaluated the titers of binding antibodies in plasma to different fragments of the SARS-CoV-2 S protein, including the extracellular domain (S-ECD), S1, S2, and RBD, and the Nucleocapsid (N) Protein. Plasma from 85 all the patients except donor No. 2 bound to all 5 SARS-CoV-2 protein segments, while that from donor No. 2 recognized S-ECD and S2 only (Fig. 1A). The neutralizing capacities of plasma against live SARS-CoV-2 and HIV-vectored pseudotyped SARS-CoV-2 are correlated (r=0.6868, p<0.05) (Fig. 1B). These results indicate that humoral immune responses were specifically elicited for all of the 10 patients during the natural infection of SARS-CoV-2. 90 To isolate S protein-specific monoclonal antibodies, we first sorted the IgG + memory B cells from peripheral blood mononuclear cells (PBMCs) of the No. 1-5 convalescent patients with flow cytometry using S-ECD as probe (Fig. 1C). The percentage of S-ECD-reactive IgG + B cells ranges from 0.56% to 11% as revealed by fluorescence activating cell sorter (FACS). To avoid 95 bias introduced by S-ECD, we sorted plasma B cells from mixed PBMCs derived from another five convalescent patients (No. 6-10) without any antigen-specific probes. The percentage of plasma B cells in CD3-CD19 + B cells was 12.8%, higher than that of memory B cells (Fig. 1C).
From the sorted B cells, we identified 9, 286, 43, 12 and 26 clones of single B cell from patients 100 No. 1 to 5, respectively, and 23 clones of single B cell from the mixed PBMCs of patients No. 6 to 10 (Fig. 1D). The distribution of the sequenced heavy (IgH) gene families was comparable among the 10 donors, with VH3 being the most commonly used VH gene, while different donors displayed variable preferences for the light chain (IgL) gene families (Fig. 1D). The combination of V3 and J4, V3 and D3, and D3 and J4 were the most common usage for the IgH gene family

Binding profiles of SARS-CoV-2 S protein-specific human mAbs
To screen S protein-specific antibodies, we determined the binding specificity using ELISA for the 399 human mAbs sorted above. 1, 16, 1, 3, and 9 S-ECD-specific mAbs were identified from donors No. 1-5, respectively. A total of 5 mAbs were identified from donors No. 6 -10 ( Fig. 2A). 115 The sequence identities of CDRH3 in the 35 S-ECD-specific mAb ranged from 40.9% to 97.6% ( Fig. S2 and Table S2). We further characterized domain specificities of the 35 mAbs with different fragments of the S protein, including S1, S2 and RBD ( Fig. 2A). The S-reactive mAbs are classified into 4 major groups based on their EC 50 values ( Fig. 2A). Group 1 recognizes only S-ECD. Group 2 recognizes S-ECD and S1, with subgroup 2A binding S-ECD and S1 and 120 subgroup 2B binding S-ECD, S1, and RBD. Group 3 interacts with both S1 and S2, with subgroup 3A targeting RBD and subgroup 3B fails to bind RBD. Group 4 recognizes S-ECD and S2. Remarkably, only 4 mAbs recognize RBD among the 35 S-specific mAbs (Figure 2A and A competition-binding assay using ELISA were performed for several representative mAbs to determine if there are any overlapping antigenic sites within distinct binding groups or between different mAbs, with CR3022 being used as a control that cross-reacts with the RBD of both SARS-CoV and SARS-CoV-2 (35) (Fig. 2C). Among these mAbs, 4A8 competed with 1D2.
Another RBD-reactive mAb, 2M-10B11, competed with CR3022, suggesting overlapped 130 epitopes on RBD for these two mAbs. These results indicate that the S-specific antibodies elicited by SARS-CoV-2 infection target at least four antigenic regions on the S protein of SARS-CoV-2.
To characterize the diversity in gene usage and affinity maturation, the phylogenetic trees of 135 these S-specific mAbs were analyzed based on the amino acid sequences of VHDJH and VLJL using a neighbor-joining method in MEGA7 Software. Results indicate that the VH gene usage is very diverse among the 35 mAbs from 10 donors, with VH 3-30 being the most frequently used germline gene. Moreover, there was no particularly favored VH gene identified among S1, S2, or RBD-reactive mAbs (Fig. 2D).

Neutralizing activities of SARS-CoV-2 S-specific human mAbs
We first performed in vitro neutralization studies of the 35 S-specific mAbs using live SARS-CoV-2 virus in Vero-E6 cells (Fig. 3A). 1M-1D2 and 4A8 exhibited medium to high neutralizing capacity with EC 50 of 23.99 and 0.607 µg/ml, respectively, while the neutralizing curves of 145 0304-3H3 cannot be fit well. As expected, the RBD-targeting control mAb, CR3022, failed to neutralize live SARS-CoV-2 (35). Moreover, while the CR3022-competing mAb, 10B11, bound to the SARS-CoV-2 RBD with EC 50 of 5 ng/ml, it also failed to neutralize live SARS-CoV-2.
These results suggest that binding affinities of mAbs against RBD did not correlate fully with the neutralizing abilities of mAbs. To further investigate the inhibitory activity of 4A8, 0304-3H3, 150 and 1M-1D2 to live virus, we tested the RNA load of live viruses in Vero-E6 cells treated with each mAb using real time qPCR (Fig. 3B). Consistent with the cytopathic effect ( CPE) assay results (Fig. 3A), 4A8 displayed higher inhibitory capacities than 1M-1D2.
We next performed luciferase reporter gene assays for all 35 S-binding mAbs using HIV-155 vectored pseudotyped SARS-CoV-2, among which 3 mAbs exhibited neutralizing activity against the pseudotyped virus (Fig. 3C). 4A8 protected ACE2-293T cells with EC 50 of 52.03 µg/ml, while neutralization of 0304-3H3 and 1M-1D2 were not observed. 10B11 and 9A1 showed a minor level of inhibition. These findings suggest that the results of pseudotyped SARS-CoV-2 were not entirely consistent with that of live SARS-CoV-2. This difference is 160 likely explained by the insusceptibility of pseudotyped SARS-CoV-2 to some mAbs with specific neutralizing mechanisms. 4A8 is likely to be a potential candidate for the treatment of SARS-CoV-2, since 4A8 displayed high levels neutralizing capacities against both authentic and pseudotyped SARS-CoV-2. 165

Binding characterization of candidate mAbs
To determine the possible neutralizing mechanism of the mAbs, we first determined the binding affinities of mAbs to different segments of the S protein, including S-ECD, S1, S2, and RBD using bio-layer interferometry. All 5 tested mAbs bound to S-ECD with high affinity, with the equilibrium dissociation constant (K D ) less than 2.14 nM (Fig. 4A). 4A8, 1M-1D2, and 10B11 170 bound to S1 with similar K D ranging from 92.7 to 159 nM, whereas 0304-3H3 and 9A1 targeted S2. Additionally, 10B11 were shown to bind RBD with K D of 26.8 nM.
To further investigate whether these mAbs block the binding of S protein to ACE2, we performed flow cytometry using HEK 293T cells expressing human ACE2. As expected, only 175 10B11 among the 5 mAbs prevented S protein from binding to ACE2, with percentage of 0.52% for IgG and S protein double-positive cell ( Fig 4B). However, CR3022, which competes with 10B11, did not interfere with the binding of S to ACE2 with double positive cell percentage of 31.7%. 4A8 also failed to interfere with binding of the S protein to ACE2.

Cryo-EM structure of the complex between 4A8 and S-ECD
The mAb 4A8 was overexpressed and purified by Protein A resin from the B memory cells and S-ECD of SARS-CoV-2 was purified through M2 affinity resin and size exclusion chromatography. 4A8 and S-ECD protein were mixed and incubated at a stoichiometric ratio of ~ 1.2 to 1 for 1 hour and applied to size exclusion chromatography to remove excess proteins. 185 The fraction containing the complex was concentrated for cryo-EM sample preparation.
To investigate the interactions between 4A8 and the S protein, we solved the cryo-EM structure of the complex at an overall resolution of 3.1 Å (Fig. 5, Movie S1). Details of cryo-EM sample preparation, data collection and processing, and model building can be found in Materials and 190 Methods and supplementary materials (Figs. S3-S5). The S protein exhibits asymmetric conformations similar to the previously reported structures (24, 25), with one RBD "up" and the other two "down" (Fig. 5, fig. S3I).

Recognition of the NTD by 4A8
195 Three 4A8 molecules bind to one trimeric S protein, each interacting with one NTD domain.
Despite the different conformations of the three S protein protomers, the interface between 4A8 and each NTD is identical (Fig. 5). The map quality at the NTD-4A8 region was improved by focused refinement with a local resolution of 3.4 Å, enabling reliable analysis of the interactions between the NTD and the 4A8. 200 Association with 4A8 appears to stabilize the NTD epitope, which is invisible in the reported S protein structure alone (24, 25). Supported by the high resolution of NTD, we were able to build the structural model for three new loops for NTD, designated N1 (residues 67-79), N2 (residues 141-156), and N3 (residues 246-260), among which N2 and N3 loops mediate the interaction 205 with 4A8.
Only the heavy chain of 4A8 participates in binding to the NTD mainly through three loops, named L1, L2 and L3 (Fig. 6A). The interface is constituted by extensive hydrophilic interaction network. R246 on the N3 loop of the NTD represents one docking site, simultaneously 210 interacting with Glu1, Tyr27 and Glu31 of 4A8 (Fig. 6B). On the N2 loop of the NTD, Lys150 and Lys147 respectively form salt bridge with Glu54 and Glu72 of 4A8 (Fig. 6C). Lys150 is also hydrogen bonded (H-bond) with 4A8-Tyr111, while His146 forms a H-bond with 4A8-Thr30 (Fig. 6C). In addition to the hydrophilic interactions, Trp152 and Tyr145 on the N2 loop of the NTD also interact with Val102, Pro106, and Phe109 on the L3 loop of the mAb through 215 hydrophobic and/or π -π interactions (Fig. 6D).

Discussion
There The S1-targeting mAb 4A8 does not block the interaction between ACE2 and S protein, but 230 exhibits high levels of neutralization against both authentic and pseudotyped SARS-CoV-2 in vitro. Many neutralizing antibodies against the SARS-CoV-2 were reported to target the RBD of the S protein and block the binding between RBD and ACE2. Our results show that 4A8 bind to the NTD of S protein with potent neutralizing activity. Previous study showed that mAb 7D10 could bind to the NTD of S protein of MERS-CoV probably by inhibiting the RBD-ACE2 235 binding and the prefusion to postfusion conformational change of S protein (29). We aligned the crystal structure of 7D10 in complex with the NTD of S protein of MERS-CoV with our complex structure and found that the interfaces between the mAb and the NTDs are partially overlapped (Fig. S6). 7D10 may inhibit the interaction between MERS-CoV and ACE2 through its light chain that is close to the RBD. In our complex, the light chain of 4A8 is away from the 240 RBD (Fig. S6). Therefore, we speculate that 4A8 may neutralize SARS-CoV-2 by restraining the conformational changes of the S protein. Furthermore, sequences alignment of the S proteins from SARS-CoV-2, SARS-CoV, and MERS-CoV revealed varied NTD surface sequences that are respectively recognized by different mAbs (Fig. S7). 245 Overall, this work reports a fully human neutralizing mAb recognizing a vulnerable epitope of NTD on S protein of SARS-CoV-2, functioning with a mechanism that is independent of receptor binding inhibition. Combination of 4A8 with RBD-targeting antibodies may avoid the escaping mutations of virus and serve as promising "cocktail" therapeutics. The information obtained from these studies can be used for development of the structure-based vaccine design

Materials and Methods
Figures S1-S7