Structural Basis of a Human Neutralizing Antibody Specific to the SARS-CoV-2 Spike Protein Receptor-Binding Domain

To identify S protein-directed human neutralizing antibodies, we collected two longitudinal convalescent blood samples from a recovered patient at 17 and 25 days after the onset of the disease symptoms. The patient recovered from COVID-19 during the outbreak in Zhuhai, Guangdong Province, China. Plasma samples and peripheral blood mononuclear cells (PBMCs) were isolated for serological analysis and antibody isolation. Serum antibody titers to SARS-CoV-2 recombinant S truncated proteins (including the extracellular domain of S protein [residues 1 to 1,142, S-ECD], the S1 subunit [residues 1 to 678, S1], and the receptor-binding domain [residues 319 to 533, S-RBD]) were measured by enzyme-linked immunosorbent assays (ELISA). The serologic analysis demonstrated that serum antibody titers to the different versions of S protein were substantially high in the serum (Fig. 1A).

To take advantage of the high percentage of antigen-specific plasma cells, single plasma cells (Fig. 1B) with the phenotype CD3⁻/CD14-/CD16⁻/CD235a-/CD19⁺/CD20^low-neg/CD27^hi/CD38^hi, as well as antigen-specific memory B cells with the phenotypes CD20⁺/CD27⁺ and S-ECD-BV421⁺/S-ECD-PE-Cy7⁺ (Fig. 1C), were sorted from PBMCs by fluorescence-activating cell sorter (FACS). The variable region of immunoglobulin (Ig) heavy- and light-chain gene segment (V_H and V_L, respectively) pairs from the single sorted B cells were amplified by reverse transcriptase PCR (RT-PCR), sequenced, annotated, and expressed as recombinant MAbs using methods described previously (39). Recombinant MAbs were screened against SARS-CoV-2 recombinant S-ECD protein. In total, we identified 28 MAbs that reacted with SARS-CoV-2 recombinant S-ECD protein, including 19 MAbs from plasmocytes and 9 MAbs from memory B cells (Fig. 1D).

We found that IgG1 was the predominant isotype, at 64.3%, followed by IgA1 (17.9%), IgG2 (10.7%), IgG4 (3.6%), and IgM (3.6%) (see Table S1 in the supplemental material). V_H gene family usage in SARS-COV2 S protein-reactive antibodies was 21.4% V_H1, 57.1% V_H3, 14.3% V_H4, 3.6% V_H5, and 3.6% V_H6 (Table S1), which was similar to the distribution of V_H families collected in the NCBI database. Out of 28 SARS-CoV-2 S protein-reactive antibodies, 4 had no mutation from their germ line V_H segments (Table S1). The chains of nCoV617 are encoded by the immunoglobulin (IG) genes IGHV3-30, IGHD3-3, IGHJ4, and IGKV1-44. IGHV of nCoV617 is 1.04% somatically mutated in its nucleotide sequence from the germ line gene, whereas its IGKV is 0.7% somatically mutated.

To screen potent antibodies against SARS-CoV-2 S proteins, we employed binding assays by using antibody dilutions on ELISA plates coated with either recombinant trimeric spike protein ECD, spike protein S1 subunit, or recombinant S-RBD (Fig. 2A). The MAbs nCoV531 (green curves), nCoV608 (orange curves), nCoV617 (red curves), and nCoV653 (blue curves) showed relatively stable binding. Dissociation constants with the SARS-CoV-2 S-RBD were obtained by surface plasmon resonance (SPR). For the above-described antibodies, the dissociation constant (K_D) ranged from 1.72 to 36.2 nM (Fig. 2B to F).

Authentic virus neutralization assay results showed that among the five antibodies that were screened, nCoV617 had the most potent inhibitory ability, with a 90% inhibitory efficiency at 0.01 μg/ml (Fig. 2G). Next, we examined the potential competitive binding of nCoV617 and ACE2 with SARS-CoV-2 S-RBD. The recombinant SARS-CoV-2 S-RBD is covalently immobilized on the CM5 sensor chip and saturated with nCoV617 or ACE2, and soluble nCoV617 or ACE2 is injected and flows through the CM5 sensor. As shown in Fig. 2H and I, the S-RBD-binding capacity of ACE2 is strongly competitive by nCoV617. Taken together, these results indicate that the antibody nCoV617 is a potent neutralizing antibody to SARS-CoV-2, with classic receptor-binding site (RBS) epitopes competing with ACE2 binding.

We further tested nCoV617 in vivo in SARS-CoV-2-infected ACE2-HU mice (a southern model organism) in a preventive and therapeutic environment. SARS-CoV-2 (a Chinese epidemic strain) incubated intratracheally in mice (6 to 8 weeks old) was challenged with a 1 × 10⁴ 50% tissue culture infection dose (TCID₅₀). In the treatment group, nCoV617 (20 mg/kg) was injected intravenously 2 h postinfection. The control group (n = 3) was given the irrelevant IgG1 antibody, Ag005 (20 mg/kg), 2 h after infection, and the prevention group was given the same standard nCoV617 24 h before infection (Fig. 3A). After challenge, the mice were weighed every day, and the lungs and trachea of the mice were collected at 3 days postinfection (dpi). The weight of the mice in the control group decreased every day from the first day after infection (Fig. 3B). In contrast, the nCoV617 treatment group and the prevention group maintained the weight of the mice unchanged after administration (Fig. 3B). Evaluation of the virus titers in the lungs and trachea found that, compared with the control group, nCoV617 reduced the virus titers in the lungs by approximately 6∗log10 of copies in 200 mg tissue and approximately 3∗log10 of copies in 200 mg tissue in the trachea by 3 dpi (Fig. 3C and D). In addition, in the prevention group of three mice, a single dose of nCoV617 (20 mg/kg) before challenge with SARS-CoV-2 protected mice from SARS-CoV-2 infection. The lowest level of virus was detected in the lungs of this group. Compared with the control group, the viral load in the trachea was significantly lower, similar to that of the treatment group (Fig. 3C and D). These results indicate that the nCoV617 antibody has a powerful ability to prevent SARS-CoV-2 infection.

In addition to reducing the virus titer, we also investigated whether nCoV617 also inhibited pathological lung injury in mice challenged with SARS-CoV-2. One mouse in each group was euthanized and necropsied at 3 dpi. Control mice showed interstitial pneumonia, which was characterized by thickening of the alveolar septum and proliferation and fibrosis of fibroblasts, as well as intense infiltration of monocytes and lymphocytes. In some alveolar cavities, cellulose exudation was observed to form a hyaline membrane and pulmonary hemorrhage. There was also apparent thrombosis in the pulmonary capillary lumen, bronchiole necrosis, and the accumulation of shed epithelial cells (Fig. 3E). In contrast, after prophylaxis or treatment, mice showed limited pathological lung damage. Compared with the control mice, the autopsied mice from the treatment group and the prevention group had complete alveolar structure, reduced edema and hyaline membrane formation, less fibrosis, and less leukocyte infiltration (Fig. 3F and G). In addition, no severe bronchial or pulmonary capillary lesions were observed (Fig. 3F and G). Therefore, in terms of prevention and treatment, nCoV617 inhibits the SARS-CoV-2 virus titer and reduces infection-related lung damage.

To explore the molecular recognition mechanism of the MAb nCoV617 with S-RBD, we next solved the complex structure at 2.5-Å resolution by X-ray crystallography. Briefly, the complex structure was determined by molecular replacement using the S-RBD structure (PDB: 6M0J) and monoclonal antibody C135 (PDB: 7K8Z) as the search models. The asymmetric unit of the crystal contains only one complex of SARS-CoV-2 S-RBD with nCoV617. The complete statistics for data collection, phasing, and refinement are presented in Table S2. The complex structure is deposited in the Protein Data Bank under the access code 7E3O.

The overall structure demonstrated that complementarity determining regions (CDRs), including H-CDR1, H-CDR3, L-CDR1, L-CDR2, and L-CDR3, of nCoV617 recognize the ridge region of S-RBD with an interface area of 672.2 Ų to V_H and 446.2 Ų to V_L, respectively (Fig. 4A and Fig. S1). With the buried surface criterion (BSA, >10 Ų), 10 residues of the RBD were involved in the interaction with nCoV617 by analysis with the PISA server (https://www.ebi.ac.uk/pdbe/pisa/). Briefly, the Y449, L452, and F490 residues of S-RBD are located in the adjacent area of the ridge two-strand antiparallel β-sheet, interacting with both chains of the NAb nCoV617 mainly via a hydrophobic network (Fig. 4B). Additionally, the tip of the ridge region of S-RBD forms multiple hydrogen bonds (H-bonds) with residues from H-CDR1 and H-CDR3 of nCoV617 V_H. Among these interactions, backbones of F486 (S-RBD) form weak interactions with G26 and T28 from H-CDR1, while the δ-amide group of Q493 (S-RBD) is bound by the backbone of L101 and L104 from H-CDR3. Notably, the δ-carboxy group of E484 (S-RBD) forms strong interactions with the ɛ-amide group of R97 from H-CDR3 (Fig. 4C). In contrast, the interactions between the V_L of nCoV617 and the S-RBD are mainly mediated by long-distance H-bonds (Fig. 4D). Taken together, nCoV617 primarily recognizes several critical epitopes on the ridge region of the S-RBD via various hydrophilic and hydrophobic interactions, excluding the Y501 residue found in the newly emerged SARS-CoV-2 strain B.1.1.7 (Fig. S1).

Previous evidence has suggested that the S-RBD fluctuates between up and down conformations in the context of the S trimer (16). In the S-RBD up conformations, the ridge region is completely exposed to the solvent region and accessible by nCoV617. To examine whether nCoV617 binds to the S-RBD down conformation, we next superimposed the nCoV617-S-RBD complex with trimeric S protein cryo-electron microscopy (cryo-EM) models (PDB: 6VXX and 7DF4). As shown in Fig. S2, alignment of the nCoV617-S-RBD complex structure into the “all-down” RBD conformation of the S-ECD cryo-EM structure (PDB: 6VXX) demonstrated that nCoV617 is tolerated to be fitted into the model without any clash (Fig. S2A). However, a slight clash to the up state S-RBD was observed when the nCoV617-S-RBD was aligned to the down RBD in the “one-up/two-down” S-ECD model (PDB: 7DF4) (Fig. S2B). These alignment results reflect that nCoV617 potentially binds to both up and down RBD states in the context of full-length S protein.

To examine the capacity of nCoV617 to block the interaction between S-RBD and ACE2, the nCoV617-S-RBD complex structure was superimposed with the S-RBD-ACE2 complex structure. As shown in Fig. 5A, the approaching angle of nCoV617 is dramatically different from that of ACE2, with an approximately 106° deviation measured by the Cα atoms of residues E214 (nCoV617-V_L), Y495 (S-RBD), and P258 (ACE2). However, the V_H of nCoV617 sterically blocked ACE2 binding, mediated by overlapping residues Y449, F486, and Q493 shared with ACE2 binding sites (Fig. 5A). Notably, the ridge region of the S-RBD interacts with nCoV617 mainly on the front side, whereas ACE2 binds to its back side (Fig. 5A). To explore the spatial relationship of the epitope sites, the nCoV617 epitope is highlighted along with the ACE2 binding sites. As shown in Fig. 5B, only limited overlapping residues were found at the interacting surfaces. As a result, the total volume of clashed residues was only ∼1,026 ų, much less than the previously reported RBS-A and RBS-B subgroup antibodies (Fig. 5B). Subsequently, epitope alignment suggested that the nCoV617 epitope belongs to the RBS-C subgroup (Fig. 5C).

To assess the structural basis of the nCoV617-S-RBD complex and whether natural mutants affect the antibody-binding affinity, we next conducted single-site alanine mutagenesis SPR assays for the key residues in the S-RBD. Among the hydrophobic interaction residues, the F490A mutations resulted in complete loss in the binding of nCoV617 compared with wild-type S-RBD (with a K_D value of 4.23 nM; Fig. 6A to C). Meanwhile, the other hydrophobic interaction residue mutant, L452A, showed an approximately 50-fold reduction (with a K_D value of 195 nM; Fig. 6D). The remaining mutant in this interaction network, Y449A, resulted in a moderate reduction (with a K_D value of 47.2 nM; Fig. 6E). The V_H recognition mode suggested that E484 is pivotal in nCoV617 binding, and F486 and Q493 show weak binding to the NAb. As expected, the E484K mutant on the V_H recognition surface resulted in complete loss of nCoV617 binding (Fig. 6F), whereas the natural F486A and Q493A mutants only slightly affected the binding (with a K_D value of 11.0 nM and 5.16 nM; Fig. 6G and H). Furthermore, the unbound residue mutants F456A and F489A were slightly affected by nCoV617 binding (with K_D values of 26.5 nM and 51.1 nM, respectively; Fig. 6I and J). It is worth noting that the newly emerged mutations N501Y and K417N have almost negligible influence on the binding affinity of nCoV617 (with K_D values of 35.0 nM and 4.72 nM, respectively; Fig. 6K and L).

In summary, the complex structure of nCoV617-S-RBD demonstrates that hydrophobic interactions and E484 hydrophilic interactions are the main driving forces.

For cocrystallization, recombinant SARS-CoV-2 S-RBD (residues 319 to 527) was expressed using the Bac-to-Bac baculovirus system. Specifically, the SARS-CoV-2 RBD containing the gp67 secretion signal peptide and a C-terminal octahistidine was inserted into pFastBac-HTB vectors and transformed into DH10Bac component cells. The recombinant bacmid was extracted and further transfected into Sf9 cells using Cellfectin II reagents (Invitrogen). The recombinant viruses were harvested from the transfected supernatant and amplified to generate a high-titer virus stock. Viruses were then used to infect Sf9 cells for RBD expression. Briefly, secreted RBD was harvested from cell culture medium and purified sequentially on a Ni-NTA column and Superdex 200 size exclusion column (GE Healthcare) (42). For the enzyme-linked immunosorbent assay (ELISA), the recombinant SARS-CoV-2 spike protein (S-ECD, His and Flag tag ZO3481) and S1 protein (His tag Z03507) were purchased from GenScript. S-RBD protein (amino acids [aa] 319 to 533) with a 6 histidine tag at the C-terminal end was designed and codon-optimized for Sf9 cell expression. Expression and purification of the human angiotensin-converting enzyme ectodomain (ACE2, residues 1 to 614) (GenBank accession no. NM_001389402.1) fused to the Fc region of human IgG (hFc) was performed as previously described (43), and cleavage of the Fc fragment was carried out with thrombin (15).

Blood samples were collected 17 and 25 days after the onset of the disease from a patient who recovered from COVID-19 infection. Peripheral blood mononuclear cells (PBMCs) and plasma were isolated from blood by Ficoll-Paque PLUS (GE, 17-1440-02) density gradient centrifugation. All work related to human subjects was carried out in compliance with institutional review board protocols approved by the institutional review board of the Fifth Affiliated Hospital of Sun Yat-sen University. Single plasma cells with the surface markers CD3⁻, CD14⁻, CD16⁻, CD235a⁻, CD19⁺, CD20^low-neg, CD27^hi, and CD38^hi and memory B cells with the surface markers CD3⁻, CD14⁻, CD16⁻, CD235a⁻, CD20⁺, CD19⁺, CD27⁺, and SARS-CoV-2 S-ECD-BV421⁺/S-ECD-PE-Cy7⁺ (BD Biosciences and Invitrogen) were sorted into individual wells in 96-well microtiter plates containing cell lysis and RT buffer for Ig gene amplification by fluorescence activated cell sorting (FACS) as previously described (44) on a BD FACS Aria SORP. Data were analyzed using BD FACS Diva 8.0.1 software.

Genes encoding the variable region of Ig heavy and light chains (V_HDJ_H and V_L) were amplified by reverse transcription (RT) and nested PCR using a previously described method (45). PCR products of Ig V_HDJ_H and V_L genes were purified using a PCR purification kit (Qiagen), sequenced in forward and reverse directions (Thermo Fisher Scientific) and annotated by using IMGT/V QUEST (www.imgt.org/IMGT_vquest). Functional V_HDJ_H and V_L genes were used for assembling full-length Ig heavy- and light-chain linear expression cassettes by one-step overlapping PCR (45). HEK-293T cells in 12-well plates were transfected with the assembled Ig heavy- and light-chain pairs derived from the same single individual plasma cells using Effectene (Qiagen) as the transfection reagent (45).

For production of purified full-length IgG1 antibodies, the V_HDJ_H and V_L genes were cloned into the pCDNA3.1⁺ (Invitrogen) mammalian expression vector containing either the human IgG1 constant region gene, the human lambda light-chain constant region, or the lambda light-chain constant region gene using standard recombinant DNA technology (45). For the production of purified nCoV617Fab antibody, a stop codon TGA was introduced after the sequence (5′-TCTTGTGACAAA-3′) encoding amino acid residues, SCDK, just before the hinge of the human IgG1 constant region (46). Recombinant IgG1 antibodies and the nCoV617Fab antibody were produced in 293F cells cultured in serum-free medium by cotransfection with the generated full-length IgG1 or Fab heavy- and light-chain gene expression plasmid pairs using polyethylenimine (47). Full-length IgG1 antibodies were purified by using protein A column chromatography as described previously (45). The nCoV617Fab antibody used for the crystal structure was purified by Lambda FabSelect, an affinity resin designed for the purification of human Fab with a lambda light chain (GE Healthcare) (46).

We collected plasma from a patient at 17 and 25 days after the onset of the disease symptoms and measured plasma antibody titers using recombinant SARS-CoV-2 S-ECD, S1, and S-RBD proteins as antigens to coat ELISA plates. Antibodies in the supernatant of the transfected HEK293T cultures harvested 3 days after transfection were screened by ELISA as described previously (45). The binding of purified antibodies to the S-ECD, S1, or S-RBD protein was also analyzed by ELISA. Briefly, all protein antigens (S-ECD, S1, and S-RBD) were used at 200 ng/well to coat 96-well high-binding ELISA plates (Nunc 442404) using carbonate-bicarbonate buffer at pH 9.6. Plates were incubated overnight at 4°C and blocked at room temperature for 2 h with phosphate-buffered saline (PBS) blocking buffer containing 5% wt/vol goat serum and 0.05% Tween 20. Plasma or supernatant of transfected 293T cell cultures or purified MAbs in serial dilutions in PBS were incubated at 37°C for 1 h. The secondary antibody, goat-anti-human IgG-HRP (1:10,000 dilution) (Promega, W4031), diluted in blocking buffer was added and incubated at 37°C for 1 h. These plates were then washed 5 times with PBS and developed with 100 μl/well tetramethylbenzidine (TMB) substrate (Solarbio PR1200). The reaction of the mixtures in the plates was stopped with 50 μl/well 2 M H₂SO₄ and read at a wavelength of an optical density at 450 nm (OD₄₅₀) by an ELISA reader.

The binding affinity (K_D), association rate (k_a) and dissociation rate (k_d) of purified MAbs to S-ECD or RBD protein were determined by SPR using a Biacore X100 System (GE Healthcare). Anti-human Fc IgG antibody was first immobilized on a CM5 chip to approximately 6,000 response units (RUs) by covalent amine coupling using a human antibody capture kit (GE Healthcare). Purified MAbs were captured on channel 2 of the CM5 chip to approximately 200 RUs. Five 2-fold serial dilutions of the S-ECD protein starting at 40 μg/ml were injected at a rate of 20 μl/min for 90 s with a 300-s dissociation. The chip was regenerated by injection of 3 M MgCl₂ for 30 s. All experiments were performed at room temperature, and data were analyzed using Biacore X100 evaluation software (version 2.0.1). Curves were fitted to a 1:1 binding model to determine kinetic rate constants (k_a and k_d). K_D values were calculated from these rate constants. To test the nCoV617 and ACE2 competition for binding to RBD, the RBD was covalently immobilized on a CM5 sensor chip and first saturated with antibody and then flowed through with soluble ACE2.

Authentic neutralization assays were performed using Vero cells in 96-well cell culture plates (Corning; no. 3988). In the antibody screening assay, various concentrations of MAbs (3-fold serial dilution) were mixed with 50-μl 30-fold TCID₅₀ authentic SARS-CoV-2 (Chinese endemic strains) and incubated at 37°C and 5% CO₂ for 1 h. After incubation, 100 μl of Vero cell suspension (density, 1.5 to 2.5 × 10⁴ cells/ml) was added to the mixture for each well, and then the plates were placed in an incubator at 37°C with 5% CO₂ for 3 to 4 days. All experiments were performed in a biosafety level 3 facility of the Institutional Biosafety Committee of the Institute of Medical Biology (IMB) in the Kunming National High-Level Biosafety Primate Research Center.

The human ACE2 gene transgenic mouse model in C57BL/6J mouse background was developed by the Shanghai Model Organisms Center, Inc. ACE2-HU mice, aged 6 to 8 weeks (southern model organism, female), were randomly divided into three groups—the prophylactic group, the therapeutic group, and the irrelevant negative-antibody group. The antibody was administered at an intramuscular dose of 20 mg/kg. The prophylactic group was given nCoV617 antibody 24 h before challenge, the therapeutic group was given nCoV617 antibody 2 h after challenge, and the negative-antibody group was given irrelevant IgG1 antibody Ag005 2 h after challenge. All animals were challenged by nasal infusion of 1 × 10⁴ TCID₅₀ SARS-CoV-2 (Chinese epidemic strain; GenBank accession no. MT226610). The animals were weighed every day after challenge, and the lungs and trachea of the animals were collected on the third day after infection. Tissues were taken for paraffin embedding and sectioning, and sections were stained by hematoxylin-eosin staining. Total RNA of the lung and trachea was extracted with TRIzol universal total RNA extraction reagent (Tiangen; DP424), and the copy number of the SARS-CoV-2 gene was detected by one-step quantitative PCR using a real-time fluorescence quantitative PCR instrument (Bio-Rad; CFX96). Primers were as follows: ORF1ab-F, 5′-CCCTGTGGGTTTTACACTTA-3′; ORF1ab-R, 5′-ACGATTGTGCATCAGCTG-3′; and ORF1ab-P, 5′-CCGTCTGCGGTATGTGGAAAGGTTATGG-3′. All animal experiments were conducted with prior approval from the Animal Ethics Committee of the Institute of Medical Biology, IMBCAMS, permit number DWSP202009002, according to the National Guidelines on Animal Work in China. Female hACE2 transgenic mice were purchased from Shanghai Model Organisms. All work with infectious SARS-CoV-2 was performed with approval under biosafety level 3 (BSL3) and animal biosafety level 3 (ABSL3) conditions by the Institutional Biosafety Committee of the Institute of Medical Biology (IMB) in the Kunming National High-Level Biosafety Primate Research Center. To alleviate pain, all experiments were carried out in strict accordance with national guidelines for animal welfare. In accordance with the principle of animal ethics, humane euthanasia was performed.

Fab fragments were each mixed with SARS-CoV-2 RBD at a molar ratio of 1:1, incubated for 2 h at 4°C, and further purified by gel-filtration chromatography. The purified complex was concentrated to 11.55 mg/ml in buffer (20 mM Tris-HCl, pH 8.0, 150 mM NaCl [Sigma-Aldrich]) for crystallization. RBD-nCoV617 complex crystal seedings were grown in 0.2 M NH₄Cl and 30% PEG4000 (Sigma-Aldrich). Screening trials were performed at 16°C. The hanging drop vapor diffusion method was used by mixing 1 μl of protein with 1 μl of reservoir solution and then adding 0.2 μl of 1/1,000 seedings. Crystals of RBD-Fab complexes were successfully obtained in 1/1,000 seedings in 152 mM NH₄Cl and 22.8% PEG8000 (Sigma-Aldrich). Crystals were frozen in liquid nitrogen in reservoir solutions.

X-ray diffraction data were collected at the Shanghai Synchrotron Facility BL19U1 at a wavelength of 0.979 Å and a temperature of 100 K. The complex structure of NAb nCoV617 with S-RBD was determined by molecular replacement using the S-RBD structure (PDB: 6M0J) and monoclonal antibody C135 (PDB: 7K8Z) as the search model with the PHENIX software suite.

The clash between neutralizing antibodies and SARS-CoV-2 RBD was calculated with Chimera. Once the RBD-ACE2 (PDB ID: 6M0J) was aligned to the RBD-NAb complexes, any residues that clashed were isolated and saved in a new coordinate file for volume calculation using the UCSF Chimera “measure volume” command.