Characterization of a neutralizing antibody that recognizes a loop region adjacent to the receptor-binding interface of the SARS-CoV-2 spike receptor-binding domain

To produce monoclonal antibodies that recognize the SARS-CoV-2 spike protein, we first prepared an antigen for mouse immunization: the ectodomain of the spike from the Wuhan strain (Spike^Wuhan) with the furin cleavage site mutation and the tandem proline mutation (9) followed by the C-terminal foldon motif (25) and a His-tag, which was produced by a mammalian cell expression system (Fig. S1). Hybridomas were generated from mice immunized with purified Spike^Wuhan and then screened for clones producing anti-Spike^Wuhan antibodies, resulting in the identification of 70 clones (Fig. S2A and B). We next examined whether these antibody clones bound to the S1 domain, NTD, and/or RBD and found that 12 clones and 37 clones recognized the NTD and RBD, respectively (Fig. S2C). The remaining 21 clones did not bind to the S1 domain, NTD, or RBD, suggesting that they may recognize the S2 domain (Fig. S2C).

Next, we examined the inhibitory effect of the 70 identified clones on Spike^Wuhan and ACE2 binding by using an enzyme-linked immuno-sorbent assay (ELISA). As shown in Fig. S3A, 28 clones inhibited the binding of Spike^Wuhan with ACE2 by more than 75%. We then examined the neutralizing activity of the 70 clones by using a pseudotyped vesicular stomatitis virus (VSV) encoding a reporter luciferase gene and bearing Spike^Wuhan (VSV-SARS-CoV-2^Wuhan) (Fig. S3B). By comparing the luciferase signal observed in infected cells without antibodies (Fig. S3B; w/o mAb, 2.1 × 10⁴ RLU), we selected antibodies that reduced the signal to less than 0.5% as neutralizing antibodies, resulting in the identification of 25 neutralizing clones.

Of the 25 identified clones, CSW1-1805 had strong binding activity against Spike^Wuhan (Fig. S2), inhibitory activity against Spike^Wuhan and ACE2 binding (Fig. S3A), and neutralizing activity against VSV-SARS-CoV-2^Wuhan (Fig. S3B). We then compared the antibody epitope of CSW1-1805 with those of the other neutralizing antibodies. As shown in Fig. S3C, 22 clones inhibited the binding of CSW1-1805 with Spike^Wuhan, suggesting that they recognized almost the same epitope as CSW1-1805. In contrast, the remaining two clones, CSW2-0611 and CSW2-1353, were less able to inhibit the binding of CSW1-1805 with Spike^Wuhan, and CSW2-1353 particularly appeared to recognize a different epitope from CSW1-1805 (Fig. S3C, red bar). Therefore, we decided to characterize CSW1-1805 and CSW2-1353 in detail.

Shortly after the first appearance of SARS-CoV-2 in Wuhan, the B.1 lineage with the D614G substitution in the spike protein emerged in Europe and replaced the original Wuhan strain (26). Since the spike protein with the D614G substitution (Spike^B.1) is reported to be more stable than Spike^Wuhan (27), we used Spike^B.1 in our experiments.

First, we examined the binding ability of CSW1-1805 and CSW2-1353, which were purified from ascites fluid collected from immunized mice, with the RBD of Spike^B.1 (Fig. 1A). Although affinity values between IgG and an antigen obtained by using an ELISA (28) are apparent affinities, such values provide useful information for comparing the binding ability of antibodies. The apparent dissociation constant (K_{d, app}) values between RBD^B.1 and the antibodies were 4.53 × 10⁻¹⁰ M for CSW1-1805 and 1.18 × 10⁻¹⁰ M for CSW2-1353 (Fig. 1A; Table S1). Next, to evaluate the neutralizing activity of CSW1-1805 and CSW2-1353 against SARS-CoV-2, we performed a plaque reduction assay using the authentic SARS-CoV-2 Wuhan strain (SARS-CoV-2^Wuhan) (Fig. 1B). The 50% plaque reduction neutralization test (PRNT50) values calculated from the neutralization curve were 4.05 ng/mL for CSW1-1805 and 14.1 ng/mL for CSW2-1353 (Table S2).

We next evaluated the protective efficacy of CSW1-1805 and CSW2-1353 against SARS-CoV-2 infection in a mouse model. We first generated a recombinant SARS-CoV-2 (rSARS-CoV-2^MA-10) possessing seven mouse-adapting substitutions (i.e., T285I in nsp4; K2R in nsp7; E23G in nsp8; Q493K, Q498Y, and P499T in the spike; and F7S in orf6) identified in a previous study (29) by using a recently established reverse genetics system (30). Twelve-week-old BALB/c mice were intraperitoneally administered 500 µg of CSW1-1805, CSE2-1353, or mouse IgG as a control and then inoculated intranasally with 5 MLD₅₀ of rSARS-CoV-2^MA-10. Two days later, the mice were intraperitoneally given the same dose of CSW1-1805, CSE2-1353, or mouse IgG (Fig. 1C). Body weights were measured daily for 6 days after infection, and nasal turbinate and lung samples were collected at 6 days post-infection (dpi) for virus titration. There were no clinical signs such as weight loss in the CSW1-1805-treated group, whereas, in the CSW2-1353 and control groups, continuous weight loss was observed starting at 2 dpi, and by 6 dpi, all of the mice in this group had died (Fig. 1D). In the control group, the mean virus titers (±SD) in the nasal turbinates and lungs were 7.23 ± 0.25 log₁₀PFU/g and 7.34 ± 0.44 log₁₀PFU/g, respectively (Fig. 1E). No virus was detected in either the nasal turbinates or lungs of the CSW1-1805-treated mice, whereas the mean virus titers (±SD) in the CSW2-1353-treated group were 7.26 ± 0.36 log₁₀PFU/g in the nasal turbinates and 6.71 ± 0.23 log₁₀PFU/g in the lungs (Fig. 1E). These results indicate that CSW1-1805 completely protects mice from rSARS-CoV-2^MA-10 infection, but CSW2-1353 does not.

To define the epitope of each antibody, we attempted to generate viral escape mutants by repeating virus passages in the presence of each antibody. No mutant was detected after three passages when SARS-CoV-2^Wuhan was inoculated at a low multiplicity of infection (MOI) (MOI = 0.1). In contrast, when inoculated at a high MOI (MOI = 2), escape mutant viruses carrying the S477N and E484A/S494L substitutions in their spike protein emerged after three passages in the presence of CSW1-1805 and CSW2-1353, respectively (Fig. 2A; Fig. S4). We also confirmed that CSW1-1805 and CSW2-1353, even at the maximum concentration tested (i.e., 50 µg/mL), failed to neutralize the S477N and E484A/S494L mutants, respectively. These results suggest that S477 and E484/S494 are key residues in the epitopes of CSW1-1805 and CSW2-1353, respectively. In our animal challenge study, CSW2-1353 failed to neutralize rSARS-CoV-2^MA-10, which possessed Q493K, Q498Y, and P499T mutations in the spike (Fig. 1D and E). The loss of neutralizing activity of CSW2-1353 against SARS-CoV-2^MA-10 may be due to the Q493K and/or Q498Y mutation, located near S494 (Fig. 2B).

The global spread of SARS-CoV-2 has led to the emergence of several variants, such as Alpha, Beta, Gamma, Delta, and Omicron, which are designated as “variants of concerns (VOCs)” by the World Health Organization (WHO) (31). VOCs have one or more RBD mutation(s) that result in improved binding affinity for ACE2 (32–34) and, in some cases, resistance to neutralization by antibodies (33–37). We, therefore, examined whether CSW1-1805 and CSW2-1353 are effective against SARS-CoV-2 variants to identify key residues of the RBD recognized by the antibodies. The K_{d, app} values of CSW1-1805 against the RBD showed that CSW1-1805 retained high affinity for all VOCs tested (Alpha, 4.52 × 10⁻¹⁰ M; Beta, 4.82 × 10⁻¹⁰ M; Gamma, 4.94 × 10⁻¹⁰ M; and Delta, 2.62 × 10⁻¹⁰ M), except for Omicron BA.1 (Fig. S5A; Table S1). The loss of CSW1-1805 neutralizing activity against Omicron BA.1 is consistent with the Omicron BA.1 possessing the S477N mutation in the spike protein (Fig. S5C), which was found in the escape mutants from CSW1-1805 (Fig. 2A). In contrast, CSW2-1353 retained high affinity for the Alpha and Delta variants (Alpha, 1.52 × 10⁻¹⁰ M; Delta, 1.14 × 10⁻¹⁰ M) but lost the ability to bind to the Beta, Gamma, and Omicron BA.1 variants (Fig. S5B; Table S1). Because CSW2-1353 failed to neutralize the escape mutant possessing the E484A mutation in the spike (Fig. 2A), the substitution at E484 in the Beta, Gamma, and Omicron BA.1 variants (Fig. S5C) likely caused the loss of binding ability of CSW2-1353 to these variants.

Next, we evaluated the neutralizing activity of CSW1-1805 and CSW2-1353 against authentic SARS-CoV-2 variants. Because neither CSW1-1805 nor CSW2-1353 bound to the RBD of the Omicron variant (Fig. S5A and B), we tested antibody neutralization of previous VOCs from Alpha to Delta. The PRNT50 values of CSW1-1805 showed higher neutralizing activity against the Alpha variant (1.89 ng/mL), Beta variant (1.00 ng/mL), and Gamma variant (1.80 ng/mL) and 6.5-fold lower neutralizing activity against the Delta variant (26.4 ng/mL) compared with the Wuhan strain (Fig. 2C; Table S2). The Delta variant possesses the L452R and T478K mutations in the RBD region (Fig. S5C). Given that the neutralizing activity of CSW1-1805 was lost by the substitution of Ser for Asn at position 477, which is next to T478 (Fig. 2B), the T478K substitution was the likely cause of the reduced neutralizing activity of CSW1-1805 against the Delta variant. In contrast, the PRNT50 values of CSW2-1353 showed high neutralizing activity against the Delta variant (10.9 ng/mL) but significantly lower neutralizing activity against the Alpha variant (200 ng/mL) and, as expected from the ELISA data (Fig. S5B), no neutralizing activity against the Beta or Gamma variants (Fig. 2D; Table S2). In summary, CSW1-1805 neutralized several strains (i.e., Alpha, Beta, Gamma, and Delta), whereas CSW2-1353 neutralized Alpha and Delta but lost activity against Beta and Gamma, indicating that CSW1-1805 and CSW2-1353 recognize different epitopes.

To investigate the binding sites of CSW1-1805 and CSW2-1353, we performed cryo-EM single particle analyses of the Spike^B.1 in complex with the Fabs of CSW1-1805 and CSW2-1353 (Table S3). Many spike trimer particles were observed in the micrographs (Fig. S6A; Fig. S7A), and the 2D class averages showed various particle orientations (Fig. S6B; Fig. S7B). For CSW1-1805, the Fab bound to all three RBDs of the spike trimer, and all three RBDs were in the up conformation (3-up RBD) (Fig. 3A; Fig. S6C). The overall map resolution reached 3.62 Å (Fig. S6D). For CSW2-1353, there were two conformations: down-up-down (1-up RBD) and down-up-up (2-up RBD) in a counterclockwise direction (seen from the top), and the Fab bound to all three RBDs regardless of the conformation (Fig. 3B and C; Fig. S7C and D). The overall map resolutions of the 1-up RBD and the 2-up RBD reached 3.66 Å and 3.97 Å, respectively (Fig. S7E and F). As we could not build atomic models of CSW1-1805 and CSW2-1353 due to the low local resolution (Fig. S6C; Fig. S7C and D), homology models of CSW1-1805 and CSW2-1353 were generated based on these variable region sequences (Fig. S8) and manually fitted to the maps (Fig. S9A and B). In addition, the Fabs of CSW2-1353, which bound to the down RBD and to the counterclockwise adjacent up RBD, interacted with each other (Fig. S9B); the up RBD was shifted outward compared with the up RBD bound to the Fab of CSW1-1805 (Fig. S9C). In other words, the interaction between the Fabs of CSW2-1353 attached to the RBDs may have stabilized the conformation of the adjacent pair of RBDs.

CSW1-1805 bound to the loop region (Y473–Y489) at the RBD ridge (Fig. 3D and E). The cryo-EM structure of the CSW1-1805 Fab in complex with Spike^B.1 showed that CDR L3 and H3 of CSW1-1805 surrounded S477 and T478 of the RBD ridge, respectively (Fig. S9D and E). Based on the binding assay and viral escape mutation study (Fig. S5A; Fig. 2A), S477 makes an essential binding contribution. In contrast, T478 is part of the binding site but may not be important for binding because T478K mutations did not have much impact on binding affinity or neutralizing activity (Fig. S5A; Fig. 2C). Although this loop region also contains E484, CSW1-1805 neutralized the Beta and Gamma variants (Fig. 2C), which possess the E484K mutation (Fig. S5C). Given that the E484 sidechain faces away from the contact site of CDR L1 and H3 with the RBD ridge (Fig. S9F), E484 does not appear to be essential for the interaction with CSW1-1805.

CSW2-1353 recognized a wide area on the top of the RBD head, which overlapped extensively with the ACE2-RBD interaction interface (Fig. 3D and F). E484 of the RBD participated in the interactions with CDR H2 of CSW2-1353 (Fig. S9G). This is consistent with the complete loss of binding and neutralizing activity against the Beta and Gamma variants (Fig. S5B; Fig. 2D), which possess the E484K mutation (Fig. S5C). L452, Q493, and S494 were located between CDR H3 and L1 (Fig. S9H and I), and Q498 and N501 were positioned at the outer rim of CDR L1 (Fig. S9J). CSW2-1353 lost neutralizing activity against SARS-CoV-2^MA-10, which possessed the Q493K and Q498Y mutations (Fig. 1D and E) and the escape mutant with the E484A and S494L mutations (Fig. 2A), suggesting that Q493, S494, and Q498 contact CSW2-1353. In contrast, L452 and N501 may contribute to the interaction but are not essential because CSW2-1353 retained neutralizing activity against Alpha and Delta variants possessing N501Y and L452R mutations, respectively (Fig. 2D).

While the CSW2-1353 Fab/Spike^B.1 complex had a mixture of up and down RBD conformations (Fig. 3B and C), the CSW1-1805 Fab/Spike^B.1 complex had only the 3-up RBD conformation (Fig. 3A). These results suggest that CSW2-1353 recognizes both the up and down RBD conformations, whereas CSW1-1805 may specifically recognizes the up RBD conformation. To explore this possibility, we examined whether the antibodies could bind to a mutant spike protein (Spike^B.1 with the S383C/D985C mutations, named Spike^closed) (38–40) in which all of the RBDs were stabilized in the down conformation, or “closed state.” Closed state spike proteins have shown reduced binding affinity with ACE2 and infectivity of pseudo-typed SARS-CoV-2 (40, 41). We verified the reduced binding affinity of purified Spike^closed with ACE2, compared to Spike^B.1, by ELISA (Fig. 4A). Then, we examined the binding of CSW1-1805 and CSW2-1353 to Spike^closed by ELISA; contrary to our expectations, both antibodies bound to the Spike^closed with similar affinity to Spike^B.1 (Fig. 4B and C). Hence, the loop region, the epitope of CSW1-1805, was accessible in both the up and down RBD conformations.

Even though CSW1-1805 also bound to the down RBD conformation (Fig. 4B), Spike^B.1 in complex with CSW1-1805 eventually formed only the up RBD conformation (Fig. 3A), unlike CSW2-1353 (Fig. 3B and C). The loop region of the RBD harbors an intramolecular disulfide bond between Cys480 and Cys488 (Fig. 3E). We performed a cryo-EM analysis of the spike protein with the C480A mutation (Spike^C480A) (Fig. S10A and B) because the flexibility of the loop region should be increased due to the cleavage of the disulfide bond. The structure of Spike^C480A showed only the 2-up RBD conformation (Fig. S10C), whereas that of Spike^B.1 showed only the 1-up RBD conformation from our previous analysis (42). The overall map resolution reached 3.04 Å (Fig. S10D). Although we could not trace the main chain in the loop region (Fig. S11), these results suggest that the increased flexibility of the loop region of the RBD may weaken interactions with an adjacent protomer and subsequently induce a transition to the up RBD conformation. Perturbations to the loop region, such as antibody binding, may also affect interactions that stabilize the down RBD conformation and induce a transition to the up RBD conformation.

Cryo-EM structures showed that CSW2-1353 had an epitope that broadly overlapped the ACE2-bound interface of the RBD, whereas CSW1-1805 recognizes the narrow loop region at the RBD ridge adjacent to the ACE2-RBD interaction interface (Fig. 3D). Despite its relatively narrow epitope, CSW1-1805 had an affinity comparable to that of CSW2-1353, which has a broad epitope region (Table S1) and could completely protect against SARS-CoV-2 infection in the mouse model (Fig. 1C through E). Several antibodies that bind to the ridge of the RBD have been reported (43–48). Therefore, we compared CSW1-1805 with six previously reported antibodies. Structural alignment showed that all antibodies, including CSW1-1805, bind to the ridge-centered epitope, but CSW1-1805 exhibits a binding direction distinct from the other antibodies (Fig. 5A). We further aligned the CDR sequences of each antibody. For AZD8895, COVOX-253H55L, XGv347, and S2E12, which had similar binding modes (Fig. 5A), each CDR had similar characteristics in terms of length and amino acid properties (Fig. 5B). Antibodies 2G1, Ab159, and CSW1-1805 had different CDR lengths and amino acid properties from each other; in particular, CDR H3 and L1 of CSW1-1805 were distinctly different from those of the other antibodies (Fig. 5B), and these regions significantly contributed to the interaction with the loop in the RBD (Fig. S9F). These findings suggest that CSW1-1805 has different binding characteristics than those of previously reported antibodies.

The Expi293F cells (Thermo) were maintained in the HE400AZ medium (Gmep Inc., Japan) at 37°C in 8% CO₂. Transmembrane protease, serine 2 (TMPRSS2)-expressing Vero E6 (VeroE6/TMPRSS2) cells (58) were obtained from the Japanese Collection of Research Bioresources Cell Bank (JCRB1819) and maintained in high-glucose Dulbecco’s modified Eagle medium (DMEM) (nacalai tesque, Japan) containing 10% fetal bovine serum (FBS) (Sigma), 100 units/mL penicillin (nacalai tesque), 100 µg/mL streptomycin (nacalai tesque), and 1 mg/mL G418 (nacalai tesque) at 37°C in 5% CO₂. The LentiX-293T cells (Clontech) were maintained in DMEM containing 10% FBS, 100 units/mL penicillin (nacalai tesque), and 100 µg/mL streptomycin (nacalai tesque) at 37°C in 5% CO₂.

A Wuhan strain (strain SARS-CoV-2/Hu/DP/Kng/19-020) was kindly provided by Dr. Sakuragi at the Kanagawa Prefectural Institute of Public Health. An Alpha variant (B.1.1.7 lineage; strain hCoV-19/Japan/QHN001/2021), a Beta variant (B.1.351 lineage; strain hCoV-19/Japan/TY8-612/2021), and a Gamma variant (P.1 lineage; strain hCoV-19/Japan/TY7-503/2021) were obtained from the National Institute of Infectious Diseases, Japan. A Delta variant (B.1.617.2; strain BKVC-127) was isolated at the Research Foundation for Microbial Diseases of Osaka University (BIKEN), Japan. Previously, seven amino acid mutations (T285I in nsp4; K2R in nsp7; E23G in nsp8; Q493K, Q498Y, and P499T in the spike; F7S in orf6) have been identified in the mouse-adapted SARS-CoV-2 strain (SARS-CoV-2 MA-10) (29). In this study, SARS-CoV-2 with these mutations was generated as a recombinant mouse-adapted SARS-CoV-2 (rSARS-CoV-2^MA-10) by using our established reverse genetics method (30).

All viruses were propagated in VeroE6/TMPRSS2 cells in DMEM (nacalai tesque) containing 2% FBS (Sigma), 100 units/mL penicillin (nacalai tesque), 100 µg/mL streptomycin (nacalai tesque), and 1 mg/mL G418 (nacalai tesque) at 37°C in 5% CO₂. The virus stocks were stored at −80°C until use. The infectious virus titer of SARS-CoV-2 isolates and rSARS-CoV-2^MA-10 was determined as plaque-forming units (PFU) by use of a plaque assay and the 50% tissue culture infective doses (TCID₅₀), respectively, as described previously (30, 59).

The codon-optimized gene of the SARS-CoV-2 spike protein (Strain Wuhan-Hu-1, GenBank: QHD43416) and human angiotensin-converting enzyme 2 (ACE2; GenBank: NM_021804) was designed for expression in mammalian cells and synthesized from GeneArt DNA Synthesis (Thermo). For the expression of the recombinant spike protein, the sequence encoding the spike ectodomain (a.a. residue 1–1,208) with proline substitutions at residues 986 and 987, a “GSAS” substitution at the furin cleavage site (a.a. residues 682–685), and the C-terminal foldon trimerization motif followed by an octa-histidine tag (Fig. S1A) was cloned into a pcDNA3.1 expression vector (Invitrogen) (designated as pcDNA3.1/SARS-CoV-2_Spike). Amino acid mutations identified in the spike protein of the B.1 lineage (D614G), B.1.1.7 lineage (Alpha variant: del69-70, del144, N501Y, A570D, D614G, P681H, T716I, S982A, D1118H), B.1.351 lineage (Beta variant: D80A, D215G, K417N, E484K, N501Y, D614G, A701V), P.1 lineage (Gamma variant: L18F, T20N, P26S, D138Y, R190S, K417T, E484K, N501Y, D614G, H655Y, T1027I), B.1.617.2 lineage (Delta variant: T19R, del157-158, L452R, T478K, D614G, P681R, D950N), and B.1.1.529 lineage (Omicron variant BA.1: A67V, del69-70, T95I, del142-144, Y145D, del211, L212I, ins214EPE, G339D, S371L, S373P, S375F, S477N, T478K, E484A, Q493R, G496S, Q498R, N501Y, Y505H, T547K, D614G, H655Y, N679K, P681H, N764K, D796Y, N856K, Q954H, N969K, L981F) were introduced as described previously (60). For the expression of the recombinant S1 proteins, the sequence encoding S1 including the N-terminal native signal peptide (a.a. residue 1–660) followed by a C-terminal octa-histidine tag was cloned into a pcDNA3.1 expression vector (designated as pcDNA3.1/SARS-CoV-2_S1). For the expression of the recombinant NTD proteins, the sequence encoding the NTD including the N-terminal native signal peptide (a.a. residue 1–305) followed by a C-terminal octa-histidine tag was cloned into a pcDNA3.1 expression vector (designated as pcDNA3.1/SARS-CoV-2_NTD). For the expression of the recombinant RBD proteins, the RBD (a.a. residue 319–541) with the IL-2 signal sequence (YRMQLLSCIALSLALVTNS) at the N-terminus and an octa-histidine tag at the C-terminus was cloned into a pcDNA3.1 expression vector (designated as pcDNA3.1/SARS-CoV-2_RBD). For the expression of the recombinant ACE2 proteins, the sequence encoding the NTD including the N-terminal native signal peptide (a.a. residue 1–615) followed by a C-terminal octa-histidine tag and Strep-tag was cloned into a pcDNA3.1 expression vector [designated pcDNA3.1/hACE2(1–615)-His-Strep]. For the generation of the pseudo-typed SARS-CoV-2, the spike protein sequence with a C-terminal 19 amino acid deletion was cloned into the pCAGGS expression vector (61) (designated as pCAG/SARS-CoV-2_S^Δ19).

Expi293f cells (Thermo) were transfected with plasmids expressing the recombinant spike, S1, NTD, RBD, and ACE2 proteins by using the Gxpress 293 Transfection Kit (Gmep Inc.) and following the manufacturer’s protocol. At 5 days post-transfection, culture supernatants were harvested, and His-tagged proteins were purified by Ni-NTA affinity chromatography using Ni Sepharose 6 Fast Flow (Cytiva), followed by size exclusion chromatography using Superdex 200 Increase 10/300 Gl (Cytiva) equilibrated with 50 mM HEPES (pH 7.4) and 200 mM NaCl.

Fab fragments of IgG antibodies and the spike protein with the D614G mutation of the B.1 lineage (Spike^B.1) complexed with the Fab were prepared as described previously (62) with some modifications. Briefly, for the generation of the Fab fragments, 10 mg/mL IgG in PBS was mixed with 10 mg/mL papain (Wako, Japan) in PBS, 5 mM EDTA, and 10 mM L-cysteine at a 100:5 (IgG:papain) (wt/wt) ratio and then incubated at 37°C for 1 h. Prior to mixing the IgG and papain, the dissolved papain in PBS containing EDTA and L-cysteine was incubated at 37°C for 1 h to activate the papain. Undigested IgG and Fc fragments were removed by using HiTrap MabSelect SuRe (Cytiva) equilibrated with PBS. Cleaved Fabs were further purified by size exclusion chromatography using Superdex 200 Increase 10/300 Gl (Cytiva) equilibrated with 50 mM HEPES (pH 7.4) and 200 mM NaCl. For the preparation of the Spike^B.1-Fab complexes, purified Fab fragment was mixed with Spike^B.1 at a 2:1 (Fab:spike monomer) molar ratio and then incubated at 4°C for 1 h. After this incubation, the Spike^B.1-Fab complexes were purified by size exclusion chromatography using Superdex 200 Increase 10/300 Gl (Cytiva) equilibrated with 50 mM HEPES (pH 7.4) and 200 mM NaCl.

Monoclonal antibodies (mAbs) against the spike protein of the Wuhan strain (Spike^Wuhan) were produced by Bio Matrix Research Inc. (Japan). Seven-week-old female BALB/c mice (The Jackson Laboratory Japan, Inc., Japan) were first immunized with 10 µg of purified Spike^Wuhan conjugated with Freund’s Incomplete Adjuvant containing 30 µg of ODN-1826 (GeneDesign Inc. as Ajinomoto Bio-Pharma Services, Japan) by subcutaneous injection. Immunization was performed three times every 2 weeks using 5 µg of purified Spike^Wuhan mixed with Sigma Adjuvant System (SAS) (Sigma). Then, additional intravenous injections (5 µg each) were given two more times at weekly intervals. Three days after the final boost, spleen cells were isolated and fused with the P3-X63-Ag8.653 (63) mouse myeloma cell line in the presence of 50% polyethylene glycol (PEG, MW 3550, Sigma). Cells were seeded in 96-well plates and incubated at 37°C, with 5% CO₂, 95% humidity. After 7–10 days of culture in RPMI-1640 containing 10% heat-inactivated fetal calf serum (Hy clone, Cytiva), supplemented with hypoxanthine (1 × 10⁻⁴ M), aminopterin (4 × 10⁻⁷ M), and thymidine (1.6 × 10⁻⁵ M; HAT; Sigma), hybridomas that produced an antibody that bound with Spike^Wuhan were screened by using an immunoprecipitation method (proprietary technology of Bio Matrix Research Inc.) and an indirect ELISA method. For the indirect ELISA, briefly, 25 ng of Spike^Wuhan was coated on an ELISA plate (F96 Maxisorp Nunc Immuno plate, Thermo) and incubated with fivefold diluted hybridoma culture supernatant. Spike-binding mAb was detected by using a mixture of HRP-conjugated anti-mouse IgG antibodies (10,000-fold dilution of goat anti-mouse IgG1, #A90-105P; goat anti-mouse IgG2a, #A90-107P; and goat anti-mouse IgG2b, #A90-109P; Bethyl, USA). Selected hybridomas were cloned by limiting dilution to obtain single clones.

Anti-spike mAbs, CSW1-1805 and CSW2-1353, were purified from mouse ascites fluid by affinity chromatography. Briefly, BALB/c nude mice were primed by intraperitoneal (i.p.) inoculation with pristane for 7 days, and then hybridoma cells were administered by i.p. injection. Ten days post-administration of the hybridoma cells, the ascites fluid was collected. The antibodies in the ascites fluid were purified by using Protein A affinity matrix UNOshere (Bio-Rad). The variable regions of CSW1-1805 and CSW2-1353 were sequenced by Bio-Peak Inc. (Gunma, Japan) using hybridoma cells. The CDR of the antibodies was identified by using the AbM definition method (64).

The animal experiments were approved by the Ethics Committee of Laboratory Animal Experiment of Biomatrix Research Inc. (No. 2019-02), and all experiments were carried out according to the guidelines of this committee.

For the analysis of anti-spike mAb binding to SARS-CoV-2-related proteins, spike proteins (50 ng) dissolved in 100 mM carbonate/bicarbonate buffer at pH 9.6 were coated on 96-well plates (Nunc-Immuno Plate CII, Thermo) overnight at 4°C. After three washes with PBS containing 0.1% (vol/vol) Tween 20 (PBS-T), the plates were blocked with PBS containing 5 mg/mL BSA overnight at 4°C. After three washes with PBS-T, purified anti-spike antibodies or culture supernatants of hybridomas secreting anti-spike mAb were diluted in PBS, added to each well, and incubated for 2 h at room temperature. To detect spike-binding antibodies, HRP-conjugated anti-mouse IgG antibody (10,000-fold dilution, #115-035-003, Jackson Immuno Research) diluted in PBS was added to each well and incubated for 2 h at room temperature. O-phenylenediamine and 0.1% H₂O₂ in 100 mM sodium citrate buffer at pH 5.0 were used as the substrate solution. The absorbance was read at 492 nm using a plate reader (Multiskan FC, Thermo).

To test mAb inhibition of ACE2 binding with the spike, Spike^Wuhan was coated on the plates, treated with culture supernatants of hybridomas secreting anti-spike mAbs or medium as a control, and then incubated with 5 ng of biotinylated ACE2 (#10108-H08-B, Sino Biological). Spike^Wuhan-binding biotinylated ACE2 was detected by using HRP-conjugated Streptavidin (5,000-fold dilution, #SA-5004, VECTOR, USA).

Antibody epitope comparisons were performed by using a biotinylated anti-spike mAb, CSW1-1805. To obtain biotinylated CSW1-1805, affinity-purified IgG fractions from culture supernatants of CSW1-1805-secreting hybridomas were treated with EZ-Link Sulfo-NHS-LC-Biotin (Thermo) according to the manufacturer’s instructions. Biotinylated CSW1-1805 was added to each well after treatment of the well-coated Spike^Wuhan with culture supernatants of mAb-secreting hybridomas or medium as a control. Spike^Wuhan-binding biotinylated CSW1-1805 was detected by using HRP-conjugated Streptavidin (5,000-fold dilution, #SA-5004, VECTOR, USA).

To examine the binding of hACE2 with the spike, plates were coated with Spike^B.1 or Spike^Closed, blocked with BSA, and then incubated with serially diluted hACE2-His-Strep in PBS for 2 h at room temperature. After three washes with PBS-T, hACE2-His-Strep bound to the spike proteins was detected by using an anti-Strep-tag II antibody (1,000-fold dilution, #ab76949, Abcam) as the primary antibody and HRP-conjugated anti-rabbit IgG antibody (10,000-fold dilution, #111-035-003, Jackson Immuno Research) as the secondary antibody.

The VSV variant NC12.1, derived from the Indiana strain, was provided by M. Whitt. Pseudotyped VSV bearing the SARS-CoV-2 spike protein was generated as previously described (65) with some modifications. Briefly, about 80% confluent LentiX-293T cells in a collagen-coated tissue culture dish were transfected with the SARS-CoV-2 S expression vector pCAG/SARS-CoV-2_S^Δ19. After 24 h of incubation, the transfected cells were infected with G-complemented VSVΔG harboring the luciferase gene in place of the G gene (66). At 24 h post-infection, the culture supernatants containing pseudotyped VSVs bearing the SARS-CoV-2 spike protein (VSV-SARS-CoV-2) were collected by centrifugation and then stored at −80°C until use.

To examine the neutralization activity of antibodies against the pseudotyped viruses, 10-fold-diluted culture supernatants of hybridomas secreting anti-spike mAb were incubated with VSV-SARS-CoV-2 for 1 h at 37°C. VeroE6/TMPRSS2 cells were inoculated with the mAb-treated VSV-SARS-CoV-2 and incubated at 37°C under 5% CO₂. At 24 h post-infection, the infectivity of the VSV-SARS-CoV-2 was determined by measuring the luciferase activity using the Luciferase Assay System (Promega).

The test mAb was serially diluted with DMEM containing 2% FBS, 100 units/mL penicillin (nacalai tesque), and 100 µg/mL streptomycin (nacalai tesque). Sixty microliters of diluted antibody was mixed with an equal volume of virus (2.4 × 10³ PFU/mL). After incubation of the mixture at 37°C for 1 h, 100 µL of the mixture (containing 100 PFU of virus) was inoculated into confluent VeroE6/TMPRSS2 cells in a 6-well plate for 1 h at 37°C under 5% CO₂. After washing the cells once, the cells were overlaid with 2 mL of DMEM containing 1% Agar (Lonza), 5% FBS, and antibiotics. After incubation at 37°C under 5% CO₂ for 60 h, the cells were fixed with 10% formalin neutral buffer solution and then stained with 0.1% crystal violet.

Serially fivefold diluted mAbs starting at 100 µg/mL were prepared in 500 µL of DMEM containing 2% FBS, 100 units/mL penicillin (nacalai tesque) and 100 µg/mL streptomycin (nacalai tesque). SARS-CoV-2 at 0.1 MOI (4 × 10⁴ PFU) or 2 MOI (8 × 10⁵ PFU) in 500 µl of DMEM containing 2% FBS, 100 units/mL penicillin, and 100 µg/mL streptomycin was mixed with each antibody diluent and incubated at 37°C for 1 h. After the incubation, the mixture was added to VeroE6/TMPRSS2 cells in a 12-well plate and incubated for 96 h at 37°C under 5% CO₂. The culture supernatants were collected from the wells with the highest antibody concentration that showed an obvious cytopathic effect. For a second round of selection, collected supernatants from cells infected starting at 0.1 MOI or 2 MOI were diluted 1,000- or 250-fold, respectively, in 500 µL of DMEM containing 2% FBS, 100 units/mL penicillin, and 100 µg/mL streptomycin and passaged under the same condition as a first round. Again, the supernatants were collected from the wells with the highest antibody concentration that showed an obvious cytopathic effect. For a third-round selection, supernatants from cells infected at a starting MOI of 0.1 or 2 were diluted 1,000- or 100,000-fold, respectively, mixed with antibodies and then passaged following the same procedure as that used for the second round. After these three rounds of passaging, culture supernatants were collected from the wells with the highest antibody concentration with an obvious cytopathic effect for virus titration and extraction of the viral RNA.

The viral RNA was extracted from the culture supernatant using a QIAamp Viral RNA Kit (QIAGEN). Then, first-strand cDNA was synthesized using a PrimeScript 1st strand cDNA Synthesis Kit (Takara) with extracted RNA and random hexamer primers by following the manufacturer’s protocol. A gene fragment that covers the SARS-CoV-2 S sequence was amplified as previously described (30), and then the amplified product was directly sequenced in both directions with specific primers by using the ABI PRISM 3130 Genetic Analyzer (Applied Biosystems).

Eleven-week-old female BALB/c mice (Japan SLC Inc., Japan) were housed in a BSL-3 facility at the Research Institute for Microbial Diseases, Osaka University, for a 1-week adaptation. To determine the mouse median lethal dose (MLD₅₀) of rSARS-CoV-2^MA-10, BALB/c mice were infected with rSARS-CoV-2^MA-10 sequentially diluted 10¹–10⁵ TCID₅₀. Four mice in each group were observed for survival until 14 dpi: all mice died by 7 dpi in the 10⁴ and 10⁵ TCID₅₀ inoculated groups; three of four mice died in the 10³ TCID₅₀ inoculated group; one of four mice died in the 10² TCID₅₀ inoculated group; and all mice survived in the 10¹ TCID₅₀ inoculated group. Together, the MLD₅₀ of rSARS-CoV-2^MA-10 was determined to be 10^2.5 TCID₅₀. To evaluate the protective efficacy of the antibodies, 1 mg/mouse of mAb or isotype control (Wako, Japan) was administered to mice by i.p. injection, and then the mice were intranasally inoculated with five times the 50% mouse lethal dose (5 MLD₅₀; equivalent to 1.6 × 10³ TCID₅₀/50 µL/mouse for this experiment) of SARS-CoV-2 rMA10 virus under isoflurane anesthesia. At 2 dpi, the same dose of mAb was administered to the mice. Mice were monitored daily for survival and body weight changes. At 6 days post-inoculation, the mice were euthanized to obtain lungs and nasal turbinates for virus titration. Viral titers were determined by use of a plaque assay with Vero-E6/TMPRSS2 cells.

The animal experiments were approved by the Institutional Committee of Laboratory Animal Experimentation of the Research Institute for Microbial Diseases, Osaka University (R03-04-0), and all experiments were carried out according to the guidelines of this committee.

The solution of Spike^B.1 complexed with the CSW1-1805 Fab was diluted to 0.5 mg/mL with dilution buffer [50 mM HEPES (pH 7.4) and 200 mM NaCl]. For the CSW2-1353 Fab complexes, the solutions of Spike^B.1 and CSW2-1353 Fab were diluted to 1.0 and 0.57 mg/mL, respectively, with the same dilution buffer, mixed with the same volume to prepare 0.5 mg/mL Spike^B.1 with a five-time molar excess of CSW2-1353 Fab, and then incubated on ice for 20 min. For the Spike^C480A mutants, the protein solutions were diluted to 0.48 mg/mL with the same dilution buffer.

An epoxidized graphene grid (EG-grid) (42) was used to promote the adsorption of protein particles. Three microliters of 0.01 M NaOH and 1% (vol/vol) epichlorohydrin water solution was applied to the ClO₂-oxidized graphene grids to prepare EG-grids. Then, 3 µL of the protein solutions was applied to the EG-grids and incubated for 5 min at room temperature. The grids were blotted with a force of −3 and a time of 2 s in a Vitrobot Mark IV chamber (Thermo) equilibrated at 4°C and 100% humidity and then immediately plunged into liquid ethane. Excess ethane was removed with filter paper, and the grids were stored in liquid nitrogen.

All cryo-EM image data sets were acquired by using SerialEM (67) and a JEM-Z300FSC (CRYO ARM 300: JEOL, Japan) operated at 300 kV with a K3 direct electron detector (Gatan, Inc., USA) in CDS mode. The Ω-type in-column energy filter was operated with a slit width of 20 eV for zero-loss imaging. The nominal magnification was 60,000×. Defocus varied between –0.5 and –2.0 µm. Each movie was fractionated into 60 frames (0.0505 s each, total exposure: 3.04 s) with a total dose of 60 e⁻/Ų.

The images were processed by using RELION 3.1 (68) or 4.0 (69). Movies were motion corrected by using MotionCor2 (70), and the contrast transfer functions (CTFs) were estimated by using CTFFIND 4.1 (71). Micrographs whose CTF max resolutions were beyond 5 Å were selected. 3D template-based auto-picking was performed for all images by using a map of the spike protein trimer (from our previous data set) as a template, and the particles were extracted with 4 × binning. The particle images were subjected to several rounds of 2D and 3D classification. After the high-quality particles were selected, an initial model was generated and used as a reference for the following 3D classification without applying symmetry. The selected particles were re-extracted without binning; 3D auto-refinement without applying symmetry, soft mask generation, or postprocessing were performed. Then, focused 3D classification without alignment was performed to select best-quality particles and to separate different RBD states (only in the CSW2-1353 Fab complexes and Spike^C480A data set). CTF refinement, Bayesian polishing, 3D auto-refinement, and postprocessing were conducted with divided optics groups (500 micrographs per group). Another round of CTF refinement, 3D auto-refinement, and postprocessing were then performed. C3 symmetry was imposed in the CSW1-1805 data set. The final map resolutions (FSC = 0.143) were 3.62, 3.66, 3.97, and 3.04 in the CSW1-1805 Fab complexes, CSW2-1353 Fab complexes (1-up RBD), CSW2-1353 Fab complexes (2-up RBD), and Spike^C480A, respectively.

Homology models of CSW1-1805 and CSW2-1353 Fab complexes were generated by using SWISS-MODEL (72). The atomic model of Spike^B.1 was built previously (73), and the homology models of the Fab complexes were manually fitted into the density by using UCSF Chimera (74) and Coot (75), and one round of real space refinement was performed in PHENIX (76). Figures were prepared by using ChimeraX (77) and PyMOL (Schrödinger, LLC). The parameters are summarized in Table S3.