Structural basis for neutralization of SARS-CoV-2 and SARS-CoV by a potent therapeutic antibody

1The coronavirus disease 2019 (COVID-19) pandemic had afflicted >7 million people in more than 200 countries/regions, resulting in 400,000 deaths as of 8 June, according to the World Health Organization. The etiological agent of this pandemic is the newly emerging coronavirus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which, together with the closely related SARS-CoV belongs to the lineage B of the genus Betacoronavirus in the Coronaviridae family. Sharing an amino acid sequence identity of ~80% in the envelope-located spike (S) glycoprotein, both SARS-CoV-2 and SARS-CoV utilize human angiotensin converting enzyme 2 (hACE2) to enter into host cells. Cellular entry is achieved by the homotrimeric S mediated virus-receptor engagement through the receptor binding domain (RBD) followed by virus-host membrane fusion. Abrogation of this crucial role played by the S-protein in the establishment of an infection is the main goal of neutralizing antibodies and the focus of therapeutic interventions as well as vaccine design. Several previously characterized SARS-CoV neutralizing antibodies (NAbs) were demonstrated to exhibit very limited neutralization activities against SARS-CoV-2. Among these, CR3022, a weakly neutralizing antibody against SARS-CoV, is tight binding, but non-neutralizing for SARS-CoV-2, indicative of possible conformational differences in the neutralizing epitopes. More recent studies have reported two SARS-CoV neutralizing antibodies, 47D11 and S309, that have been shown to neutralize SARS-CoV-2 as well, suggesting that broad cross-neutralizing epitopes exist within lineage B. Convalescent plasma containing SARS-CoV-2 neutralizing antibodies (NAbs) have been shown to confer clear protection in COVID-19 patients, yet gaps in our knowledge concerning the immunogenic features and key epitopes of SARS-CoV-2, have hampered the development of effective immuno-therapeutics against the virus.

2The RBDs of SARS-CoV and SARS-CoV-2 have an amino-acid sequence identity of around 75%, raising the possibility that RBD-targeting cross-neutralizing NAbs could be possibly identified. Using phage display technique, we constructed an antibody library which was generated from RNAs extracted from peripheral lymphocytes of mice immunized with recombinant SARS-CoV RBD. SARS-CoV-2 RBD was used as the target for screening the phage antibody library for potential hits. Antibodies showing tight binding for SARS-CoV-2 RBD were further propagated as chimeric antibodies and tested for neutralizing activities using a vesicular stomatitis virus (VSV) based pseudotyping system (fig. S1). Among the antibodies tested, clone 014 which showed potent neutralizing activity against SARS-CoV-2 pseudovirus was humanized, named as H014. To evaluate the binding affinities, real-time association and dissociation of H014 binding to either SARS-CoV-2 RBD or SARS-CoV RBD were monitored using the OCTET system (Fortebio). Both H014 Ig G and Fab fragments exhibited tight binding to both RBDs with comparable binding affinities at sub-nM levels for SARS-CoV-2 RBD and SARS-CoV RBD, respectively (Fig. 1A and fig. S2). Pseudovirus neutralization assays revealed that H014 has potent neutralizing activities: a 50% neutralizing concentration value (IC50) of 3 nM and 1 nM against SARS-CoV-2 and SARS-CoV pseudoviruses, respectively (Fig. 1B). Plaque-reduction neutralization test (PRNT) conducted against an authentic SARS-CoV-2 strain (BetaCoV/Beijing/AMMS01/2020) verified the neutralizing activities with an IC50 of 38 nM, 10-fold lower than those observed in the pseudotyping system (Fig. 1C). We next sought to assess in vivo protection efficacy of H014 in our previously established human ACE2 (hACE2) humanized mouse model that was sensitized to SARS-CoV-2 infection. In this model, as a result of hACE2 expression on lung cells, SARS-CoV-2 gains entry into the lungs and replicates as in a human disease, exhibiting lung pathology at 5 days post infection (dpi). hACE2-humanized mice were treated by intraperitoneal injection of H014 at 50 mg/kg either 4 hours after (1 dose, therapeutic) or 12 hours before and 4 hours after (2 doses, prophylactic plus therapeutic) intranasal infection with 5 × 105 PFU of SARS-CoV-2 (BetaCoV/Beijing /AMMS01/2020). All challenged animals were sacrificed at day 5. While the viral loads in the lungs of the PBS group (control) surged to ~107 RNA copies/g at day 5 (Fig. 1D), remarkably, in the prophylactic and prophylactic plus therapeutic groups, H014 treatment resulted in a ~10-fold and 100-fold reduction of viral titers in the lungs at day 5, respectively (Fig. 1D). Lung pathology analysis showed that SARS-CoV-2 caused mild interstitial pneumonia characterized by inflammatory cell infiltration, alveolar septal thickening and distinctive vascular system injury upon PBS treatment. In contrast, no obvious lesions of alveolar epithelial cells or focal hemorrhage were observed in the lung sections from mice that received H014 treatment (Fig. 1E), indicative of a potential therapeutic role for H014 in curing COVID-19.

3The overall structure of SARS-CoV-2 S trimer resembles those of SARS-CoV and other coronaviruses. Each monomer of the S protein is composed of two functional subunits. The S1 subunit binds the host cell receptor, while the S2 subunit mediates fusion of the viral membrane with the host cell membrane. The four domains within S1 include N-terminal domain (NTD), RBD and two subdomains (SD1 and SD2) where the latter are positioned adjacent to the S1/S2 cleavage site. Hinge-like movements of the RBD give rise to two distinct conformational states referred to as the “close” and “open”, where close corresponds to the receptor-inaccessible state and open corresponds to the receptor-accessible state, which is supposed to be metastable. Cryo-EM characterization of the stabilized SARS-CoV-2 S ectodomain in complex with the H014 Fab fragment revealed that the complex adopts three distinct conformational states, corresponding to one RBD open + two RBDs closed (state 1), two RBDs open + one RBD closed (state 2) and all three open RBDs (state 3) (Fig. 2A). Interestingly, structure of the completely closed (state 4) SARS-CoV-2 S trimer without any Fab bound was also observed during 3D classification of the cryo-EM data, albeit in the presence of excessive Fab, suggesting that the binding sites of H014 are exposed through protein “breathing” followed by a stochastic RBD movement (Fig. 2A). We determined asymmetric cryo-EM reconstructions of the 4 states at 3.4 – 3.6 Å (figs. S3 to S6 and table S1). However, the electron potential maps for binding interface between RBD and H014 are relatively weak due to conformational heterogeneity. To solve this problem, focusing classification and refinement by using a “block-based” reconstruction approach were performed to further improve the local resolution up to 3.9 Å, enabling reliable analysis of the interaction mode (fig. S5 to S6). Detailed analysis of the interactions between H014 and S was done using the binding interface structure.

4H014 recognizes a conformational epitope on one side of the open RBD, only involving protein/protein contacts, distinct from the receptor-binding motif (RBM) (Fig. 2B). The H014 paratope constitutes all six complementary determining region (CDR) loops (CDRL1-3 and CDRH1-3) and unusual the heavy-chain framework (HF-R, residues 58-65) that forge tight interactions with the RBD, resulting in a buried area of ~1,000 Å2 (Fig. 2C). Variable domains of the light-chain and heavy-chain contribute ~32% and 68% of the buried surface area, respectively, through hydrophobic and hydrophilic contacts. The H014 epitope is composed of 21 residues, primarily locating in the α2-β2-η2 (residues 368-386), η3 (residues 405-408, 411-413), α4 (residue 439) and η4 (residues 503) regions, which construct a cavity on one side of the RBD (Fig. 2, B and D, and fig. S7). The 12-residue long CDRH3 inserts into this cavity and the hydrophobic residues (YDPYYVM) enriched CDRH3 contacts the η3 and edge of the five-stranded β-sheet (β2) region of the RBD (Fig. 2D). Tight bindings between the RBD and H014 are primarily due to extensive hydrophobic interactions contributed by two patches: one formed by F54 from CDRH2, Y101 from CDRH3 and A411, P412 and Y508 of the RBD and the other composed of Y49 from CDRL2, P103, Y104, Y105 from CDRH3 and V407, V503 and Y508 of the RBD (Fig. 2E and table S2). Additionally, hydrophilic contacts from CDRH1 and HF-R further enhance the RBD-H014 interactions, leading to an extremely high binding affinity at sub nM level at temperatures of 25°C or 37°C (Fig. 2E and fig. S8). Residues comprising the epitope are mostly conserved with 3 single-site mutants (R408I, N439K and V503F) in this region among currently circulating SARS-CoV-2 strains reported (fig. S9). In addition, a number of SARS-CoV-2 isolates bear a common mutation V367F in the RBD, which lies adjacent to the major epitope patch α2-β2-η2. Constructs of the recombinant RBD harboring point mutations of above residues and other reported substitutions exhibited an indistinguishable binding affinity to H014 (fig. S10), suggesting that H014 may exhibit broad neutralization activities against SARS-CoV-2 strains currently circulating worldwide. Out of the 21 residues in the H014 epitope, 17 (81%) are identical between SARS-CoV-2 and SARS-CoV (fig. S9), which explains the cross-reactivity and comparable binding affinities.

5To investigate whether H014 interferes with the binding of RBDs of SARS-CoV-2 or SARS-CoV to ACE2, we performed competitive binding assays at both protein and cellular levels. The enzyme-linked immunosorbent assay (ELISA) indicated that H014 was able to compete with recombinant ACE2 for binding to RBDs of SARS-CoV-2 and SARS-CoV with EC50 values of 0.7 nM and 5 nM, respectively (Fig. 3A). Additionally, H014 efficiently blocked both the attachment of SARS-CoV-2 RBD to ACE2 expressing 293T cells and the binding of recombinant ACE2 to SARS-CoV-2 S expressing 293T cells (Fig. 3B). To verify its potential full occlusion on trimeric S, we conducted two sets of surface plasmon resonance (SPR) assays by exposing the trimeric S to H014 first and then to ACE2 or the other way around. As expected, binding of H014 completely blocked the attachment of ACE2 to trimeric S. Moreover, ACE2 could be displaced from trimeric S and replaced by H014 (Fig. 3C). To further verify these results in a cell-based viral infection model, real time RT-PCR analysis was carried out to quantify the amount of virus remaining on the host cell surface, which were exposed to antibodies pre- or post-virus attachment to cells at 4°C. H014 efficiently prevented attachment of SARS-CoV-2 to the cell surface in a dose-dependent manner and the viral particles that had already bound to the cell surface could be partially stripped by H014 (Fig. 3D). Superimposition of the structure of the H014-SARS-CoV-2 trimeric S complex over the ACE2-SARS-CoV-2 RBD complex structure revealed clashes between the ACE2 and H014, arising out of an overlap of the regions belonging to the binding sites located at the apical helix (η4) of the RBD (Fig. 3E). This observation differs substantially from those of most known SARS RBD-targeted antibody complexes, where the antibodies directly recognize the RBM. Thus, the ability of H014 to prevent SARS-CoV-2 from attaching to host cells can be attributed to steric clashes with ACE2.

6Similar to the RBM, the H014 epitope is only accessible in the open state, indicative of a role akin to the RBM – involving dynamic interferences in interactions with host cells. Our structures together with previously reported coronavirus S structures, not including human coronavirus HKU1, have observed the breathing of the S1 subunit, that mediates the transition between “close” and “open” conformation (Fig. 4A). In contrast to the hinge-like movement of the RBD observed in most structures, the conformational transition from “close” to “open” observed in our structures mainly involves two steps of rotations: 1) counterclockwise movement of SD1 by ~25° encircling the hinge point (at residue 320); 2) the RBD rotates itself counterclockwise by ~60° (Fig. 4B). These conformational rotations of the SD1 and RBD at proximal points relay an amplified alteration at the distal end, leading to opening up of adequate space for the binding of H014 or ACE2 (Fig. 4B). Ambiguously, the special conformational transition observed in our complex structures results from the engagement of H014 or a synergistic movement of the SD1 and RBD.

7Humoral immunity is essential for protection against coronavirus infections and passive immunization of convalescent plasma has been demonstrated to be effective in treating SARS and COVID-19. A number of SARS-CoV specific neutralizing antibodies including m396, 80R and F26G19, have been previously reported to engage primarily the RBM core (residues 484-492 in SARS-CoV) provided that the RBD is in open conformation (Fig. 4, C and D). Unexpectedly, none of these antibodies has so far demonstrated impressive neutralizing activity against SARS-CoV-2. S230, a SARS-CoV neutralizing antibody that functionally mimics receptor attachment and promotes membrane fusion, despite requiring an open state of the RBD as well, recognizes a small patch at the top of the RBM (Fig. 4, C and D). Recently, two other antibodies (CR3022 and S309) originating from the memory B cells of SARS survivors in 2003, demonstrated strong binding to SARS-CoV-2. Surprisingly, CR3022 failed to neutralize SARS-CoV-2 in vitro, which could be attributed to the fact that CR3022 binds a cryptic epitope, which is distal from the RBD and is only accessible when at least two RBDs are in the open state. Furthermore, CR3022 does not compete with ACE2 for binding to the RBD (Fig. 4, C and D). By contrast, S309, a more recently reported neutralizing mAb against both SARS-CoV and SARS-CoV-2, recognizes a conserved glycan-containing epitope accessible in both the open and closed states of SARS-CoV and SARS-CoV-2 RBDs. The neutralization mechanism of S309 does not depend on direct blocking of receptor binding, but it induces antibody-dependent cell cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP). We reported a potent therapeutic antibody H014, which can cross-neutralize SARS-CoV-2 and SARS-CoV infections by blocking attachment of the virus to the host cell (Fig. 3). Distinct from other antibodies, H014 binds one full side of the RBD, spanning from the proximal η2 to the distal edge of the RBM and possesses partially overlapping epitopes with SARS-CoV specific antibodies like m392, 80R, F26G19, CR3022 and S309 (Fig. 4, C and D). Although both SARS-CoV and SARS-CoV-2 utilize ACE2 to enter host cells, high sequence variations in the RBM and local conformational rearrangements at the distal end of the RBM from the two viruses have been observed. These probably lead to the failure of some antibodies, especially those targeting the RBM of SARS-CoV, in cross-binding and cross-neutralizing SARS-CoV-2. Three cross-reactive antibodies (CR3022, S309 and H014) recognize more conserved epitopes located beyond the RBM, amongst which the CR3022 epitope is most distant from the RBM (Fig. 4, C and D), suggesting a possible reason for its inability to neutralize SARS-CoV-2. H014 binds more conserved epitopes, some of which are in close proximity to the RBM, rendering it effective in cross-neutralizing lineage B viruses. Additionally, antibodies recognizing more conserved patches beyond the RBM would function synergistically with antibodies targeting the RBM to enhance neutralization activities, which could mitigate the risk of viral escape. The molecular features of H014 epitopes unveiled in this study facilitate the discovery of broad cross-neutralizing epitopes within lineage B and pose interesting targets for structure-based rational vaccine design. Our studies also highlight the promise of antibody-based therapeutic interventions for the treatment of COVID-19.