A Potently Neutralizing Antibody Protects Mice against SARS-CoV-2 Infection

This article is featured in In This Issue, p.875

Most members of the Coronaviridae family infect the respiratory tract of mammals, causing mild respiratory disease (1). In the past two decades, however, two highly pathogenic coronaviruses, severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV), crossed the species barrier and led to epidemics with high morbidity and mortality in humans (2–4). In December 2019, a third highly pathogenic human coronavirus, SARS-CoV 2 (SARS-CoV-2) emerged in Wuhan, Hubei province of China (5–7). Compared with SARS-CoV and MERS-CoV, SARS-CoV-2 is more readily transmitted among humans, spreading to multiple continents and leading to the World Health Organization’s declaration of a coronavirus disease 2019 (COVID-19) pandemic (8, 9). As of June 16, 2020, SARS-CoV-2 caused >8 million confirmed cases globally, leading to at least 440,000 deaths (10). Currently, there are no widely available licensed therapeutics to prevent or treat COVID-19. This underlines the need for immediate development of prophylactic and therapeutic reagents to combat SARS-CoV-2 virus infection.

Betacoronavirus entry into host cells is mediated by a densely glycosylated spike (S) protein that forms homotrimers protruding from the viral envelope (11). The S protein is composed of an N-terminal S1 subunit responsible for virus–receptor binding and a C-terminal S2 subunit that mediates virus–cell membrane fusion (12). SARS-CoV-2 gains entry into host cells initially through the interaction between the receptor-binding domain (RBD) within its S1 subunit with the cellular receptor, human angiotensin-converting enzyme 2 (hACE2), and subsequently by fusion between the viral envelope and the host cell lipid bilayer mediated by the S2 subunit (13, 14). This points to the RBD as a critical target for Ab-based treatments to prevent SARS-CoV-2 virus infection and limit its spread. Indeed, several preclinical studies demonstrated that polyclonal Abs induced against SARS-CoV and MERS-CoV RBD can inhibit viral entry (15, 16). Such critical proof-of-concept findings suggest that SARS-CoV-2 RBD could be used as an immunogen to elicit potently neutralizing Abs that block SARS-CoV-2 entry.

Passive administration of mAbs has become one of the essential tools in treating many human diseases, including those caused by emerging viruses (17). Indeed, in the face of the West African ebolavirus outbreak of 2013–2016, two therapeutic recombinant mAb preparations, REGN-EB3 and MAb114, showed significant efficacy in preventing death (18–20). Based on the likelihood of protective Abs being induced by natural infection, sera from convalescent patients are currently being used as an experimental treatment for COVID-19. However, it remains unclear how nonneutralizing or weakly neutralizing Abs, which are also likely elicited by infection, alter viral infectivity and disease progression (15). Therefore, there is an urgent need to develop and fully characterize potently neutralizing mAbs that can be quickly harnessed for the prevention and treatment of SARS-CoV-2 infection.

Expi293F cells (Life Technologies) were cultured at 37°C in Expi293 Expression medium (Life Technologies). Vero E6 cells (CRL-1586, American Type Culture Collection), Vero CCL81 (American Type Culture Collection), and HEK293 were cultured at 37°C in DMEM supplemented with 10% FBS, 10 mM HEPES pH 7.3, 1 mM sodium pyruvate, 1× nonessential amino acids, and 100 U/ml of penicillin–streptomycin. SARS-CoV-2 strain 2019 n-CoV/USA_WA1/2020 was obtained from the Centers for Disease Control and Prevention (gift of Natalie Thornburg) (21). A p3 stock was passaged once in CCL81-Vero cells and titrated by focus-forming assay on Vero E6 cells.

The adenoviral AdV-hACE2-GFP construct and defective virus preparation has been reported previously (22). AdV-hACE2-GFP was propagated in 293T cells and purified by cesium chloride density-gradient ultracentrifugation. The number of virus particles was determined using OD (260 nm) measurement and plaque assay, as previously described (23). The viral stock titer was determined to be 10¹¹ PFU/ml.

DNA fragments encoding ectodomain of S from SARS-CoV (residues 14–1193, GenBank: AY278488.2), SARS-CoV-2 (residues 14–1211, GenBank: MN908947.3), and MERS-CoV (residues 19–1294, GenBank: JX869059.2) were synthesized and placed into the mammalian expression vector pFM1.2 with N-terminal μ-phosphatase signal peptide. The C terminus of all DNAs were engineered with an HRV3C protease cleavage site (GSTLEVLFQGP) linked by a foldon trimerization motif (YIPEAPRDGQAYVRKDGEWVLLSTFL) and an 8XHis Tag. The S1/S2 furin cleavage sites were mutated in both SARS-CoV-2 and MERS-CoV S, and all three S proteins were stabilized with the 2P mutations (24). The plasmids were transiently transfected in Expi293F cells using FectoPRO reagent (Poluplus), and cell supernatants containing target protein were harvested 96 h after transfection. The soluble S proteins were recovered using 2 ml of cobalt-charged resin (G-Biosciences). Mammalian SARS-CoV-2 RBD (residues 331−524) was cloned into vector pFM1.2 with N-terminal μ-phosphatase signal peptide and C-terminal 6XHis Tag. The protein was expressed as S protein and recovered by nickel agarose beads (GoldBio), further purified by passage over S75i Superdex (GE Healthcare). The bacterial version of RBD was cloned into the pET21a vector (Novagen) and expressed as inclusion bodies in Escherichia coli BL21(DE3) and purified as previously described for ZIKV DIII (25).

All procedures involving animals were performed in accordance with guidelines of the Institutional Animal Care and Use Committee of Washington University in Saint Louis.

Female C57BL/6J mice (The Jackson Laboratory) were immunized i.m. with 10 μg of SARS-CoV-2 RBD resuspended in PBS emulsified with AddaVax (InvivoGen). Two weeks later, mice were boosted with 5 μg of SARS-CoV-2 S protein twice, at 10-d intervals. One control mouse received PBS emulsified with AddaVax according to the same schedule. Sera were collected 5 d after the final boost and stored at −20°C before use. Draining iliac and inguinal lymph nodes were also harvested on day 5 after the final boost for plasmablast (PB) sorting.

Staining for sorting was performed using fresh lymph node single-cell suspensions in PBS supplemented with 2% FBS and 1 mM EDTA (P2). Cells were stained for 30 min on ice with biotinylated recombinant SARS-CoV-2 RBD diluted in P2, washed twice, and then stained for 30 min at 4°C with Fas-PE (Jo2; BD Pharmingen), CD4-eFluor 780 (GK1.5; eBioscience), CD138-BV421 (281-2), IgD-FITC (11-26c.2a), GL7-PerCP-Cy5.5, CD38-PE-Cy7 (90), CD19-APC (1D3), and Zombie Aqua (all BioLegend) diluted in P2. Cells were washed twice, and single SARS-CoV-2 RBD-specific PBs (live singlet CD19⁺ CD4⁻ IgD^lo Fas⁺ CD38^lo CD138⁺ RBD⁺) and total PBs (live singlet CD19⁺ CD4⁻ IgD^lo Fas⁺ CD38^lo CD138⁺) were sorted using a FACSAria II into 96-well plates containing 2 μl of Lysis Buffer (Clontech) supplemented with 1 U/μl RNase inhibitor (New England BioLabs) and immediately frozen on dry ice or bulk sorted into PBS supplemented with 0.05% BSA and processed for single-cell RNA-sequencing (scRNA-seq).

Abs were cloned as previously described (26). In brief, VH, Vκ, and Vλ genes were amplified by RT-PCR and nested PCR from singly sorted SARS-CoV-2 RBD⁺ PBs using mixtures of primers specific for IgG, IgM/A, Igκ, and Igλ using first-round and nested primer sets (26–28) (Supplemental Table II) and then sequenced. Clonally related cells were identified by the same length and composition of IGHV, IGHJ, and H chain CDR3 and shared somatic hypermutation at the nucleotide level. To generate recombinant Abs, H chain V-D-J and L chain V-J fragments were PCR amplified from first-round PCR products with mouse variable gene forward primers and joining gene reverse primers having 5′ extensions for cloning by Gibson assembly as previously described (29) (Supplemental Table II) and were cloned into pABVec6W Ab expression vectors (30) in frame with either human IgG, IgK, or IgL constant domain. Plasmids were cotransfected at a 1:2 H chain/L chain ratio into Expi293F cells using the Expifectamine 293 Expression Kit (Thermo Fisher Scientific), and Abs were purified with protein A agarose (Invitrogen).

Ninety-six–well microtiter plates (Nunc MaxiSorp; Thermo Fisher Scientific) were coated with 100 μl of recombinant SARS-CoV-2 S or RBD at a concentration of 0.5 and 1 μg/ml, respectively, in 1× PBS (Life Technologies) at 4°C overnight; negative control wells were coated with 1 μg/ml BSA (Sigma-Aldrich). Plates were blocked for 1.5 h at room temperature with 280 μl of blocking solution (1× PBS supplemented with 0.05% Tween-20 [Sigma-Aldrich] and 10% FBS [Corning]). The mAbs were diluted to a starting concentration of 10 μg/ml, serially diluted 1:3, and incubated for 1 h at room temperature. The plates were washed three times with 1× PBS supplemented with 0.05% Tween-20, and 100 μl of anti-human IgG HRP Ab (goat polyclonal; Jackson ImmunoResearch) diluted 1:2500 in blocking solution was added to all wells and incubated for 1 h at room temperature. Plates were washed three times with 1× PBS supplemented with 0.05% Tween-20 and three times with 1× PBS, and 100 μl of peroxidase substrate (SigmaFast o-phenylenediamine dihydrochloride; Sigma-Aldrich) was added to all wells. The reaction was stopped after 5 min using 100 μl of 1 M hydrochloric acid, and the plates were read at a wavelength of 490 nm using a microtiter plate reader (BioTek). The data were analyzed using Prism v8 (GraphPad). The minimum positive concentration was defined as having OD at least 3-fold above background.

Mouse serum ELISAs were performed similarly. Plates were coated and blocked as above. The sera were prediluted 1:30 and then serially diluted 1:3. Anti-mouse IgG HRP Ab (goat polyclonal; Southern Biotech) diluted 1:1000 in blocking solution was used as secondary Ab.

Libraries were prepared using the following 10x Genomics kits: Chromium Single Cell 5′ Library and Gel Bead Kit v2 (PN-1000006); Chromium Single Cell A Chip Kit (PN-1000152); Chromium Single Cell V(D)J Enrichment Kit, Mouse Bcell (96rxns) (PN- 1000072); and Single Index Kit T (PN-1000213). The cDNAs were prepared after the gel bead in emulsion generation and barcoding, followed by gel bead in emulsion reverse transcriptase reaction and bead cleanup steps. Purified cDNA was amplified for 10–14 cycles before being cleaned up using SPRIselect beads. Samples were then run on a Bioanalyzer to determine cDNA concentration. BCR target enrichments were done on the full-length cDNA. Gene expression and enriched BCR libraries were prepared as recommended by 10x Genomics Chromium Single Cell V(D)J Reagent Kits (v1 Chemistry) user guide with appropriate modifications to the PCR cycles based on the calculated cDNA concentration. The cDNA Libraries were sequenced on Novaseq S4 (Illumina), targeting a median sequencing depth of 50,000 and 5000 read pairs per cell for gene expression and BCR libraries, respectively.

We obtained a list of 262 annotated Ig genes with “IG_*_gene” for their “gene_biotype” from the Ensembl 93 gene annotation (31) for the current genome assembly for the C57BL/6 strain of Mus musculus (GRCm38, or mm10) (32). The genes Ighv1-13, Ighv5-8, and Iglc4 were removed as a result of being annotated as pseudogenes by both Mouse Genome Informatics (MGI) and National Center for Biotechnology Information Gene and having biotype conflicts with Ensembl. The final list of 259 mm10 Ig genes included 113 Ighv genes, 17 Ighd genes, 4 Ighj genes, 100 Igkv genes, 5 Igkj genes, 3 Iglv genes, 5 Iglj genes, 8 Ighc genes, 1 Igkc gene, and 3 Iglc genes. Genomic sequences for these genes were retrieved based on their Ensembl identifiers via the Ensembl REST API (release 13.0) (33).

Ig reference alleles (release 202011-3) for mouse were downloaded from the ImMunoGeneTics information system (IMGT) on 2020-04-02 under the “F+ORF+in frame P” configuration (34). Alleles annotated as Mus spretus were removed, leaving only alleles annotated as Mus musculus. The final list of IMGT alleles for Mus musculus included 406 IGHV alleles, 38 IGHD alleles, 9 IGHJ alleles, 150 IGKV alleles, 10 IGKJ alleles, 14 IGLV alleles, 5 IGLJ alleles, 106 IGHC alleles, 3 IGKC alleles, and 3 IGLC alleles.

To identify the closest IMGT allele, each mm10 Ig gene was aligned against the IMGT alleles for its corresponding gene segment using blastn (v2.9.0) (35). For Ighd genes, blastn-short was also used to accommodate short sequence lengths. For each mm10 Ig gene, a search for the IMGT allele with 100% match for the full length of the allele was conducted (Supplemental Table III).

For each of 247 out of the 259 mm10 genes, one or more matching alleles were identified. For 18 of these 247 mm10 genes, two IMGT alleles were identified with identical nucleotide sequences and full-length 100% matches. Where possible (16 out of 18), the allele with the name matching that of the mm10 Ig gene was designated as the corresponding IMGT allele. For example, the identical IMGT alleles IGKV4-54*01 and IGKV4-52*01 both matched with mm10 gene Igkv4-54; in this case, IGKV4-54*01 was noted as the corresponding C57BL/6 IMGT allele. Where this was not possible (two out of 18), an allele was chosen based on the locus representation map. For Ighd5-7, which matched with IGHD6-1*01 and IGHD6-3*01, IGHD6-3*01 was chosen. For Ighd5-8, which matched with IGHD6-1*02 and IGHD6-4*01, IGHD6-4*01 was chosen. For the 247 mm10 genes with full-length 100% matches with IMGT alleles, the corresponding IMGT alleles were used as the curated reference alleles.

For mm10 genes Ighv1-62-1, Ighv12-3, Ighv2-3, Ighv8-2, and Ighv8-4, length discrepancies were noted at the 3′ end in the form of additional nucleotides in the closest matching IMGT alleles: IGHV1-62-1*01, 2 bp; IGHV12-3*01, 1 bp; IGHV2-3*01, 3 bp; IGHV8-2*01, 1 bp; and IGHV8-4*01, 7 bp. In each case, the sequence immediately downstream of the mm10 gene was examined in the Ensembl Genome Browser (36) for identification of candidate heptamer-spacer-nonamer recombination signal sequence (RSS) motif under the 12/23 rule (37). For Ighv1-62-1, an RSS motif was observed immediately adjacent to the final nucleotide annotated in mm10. In this case, the additional nucleotide in the IMGT allele was not included for the curated reference allele. For Ighv12-3, Ighv2-3, Ighv8-2, and Ighv8-4, evidence for putative RSS motifs were observed adjacent to the final nucleotides of the IMGT alleles. In these cases, the additional nucleotides in the IMGT alleles were included for the curated reference alleles.

For mm10 genes Igkv3-7, Igkv9-120, Ighg2b, and Ighg3, IGKV3-7*01, IGKV9-120*01, IGHG2B*02, and IGHG3*01 were identified as the closest IMGT alleles with, respectively, 1, 1, 3, and 3 nucleotide mismatches. For the mismatched positions, the curated reference alleles deferred to the nucleotides found in the corresponding mm10 genomic sequences.

For mm10 genes Igkc, Iglc1, and Iglc3, length discrepancies were noted at the 5′ end, where the mm10 genomic sequences begin, in the form of one additional nucleotide each in the closest matching IMGT alleles: IGKC*01, IGLC*01, and IGLC*03. The curated reference alleles deferred to the mm10 genomic sequences and did not include the additional nucleotide found in IMGT alleles.

The final curated set of C57BL/6 reference alleles included 113 IGHV alleles, 17 IGHD alleles, 4 IGHJ alleles, 100 IGKV alleles, 5 IGKJ alleles, 3 IGLV alleles, 5 IGLJ alleles, 8 IGHC alleles, 1 IGKC allele, and 3 IGLC alleles (Supplemental Table III).

Demultiplexed pair-end FASTQ reads from 10x Genomics single-cell V(D)J profiling were preprocessed using the “cellranger vdj” command from Cell Ranger v3.1.0 for alignment against the GRCm38 mouse reference v3.1.0 (refdata-cellranger-vdj-GRCm38-alts-ensembl-3.1.0), generating 15,270 assembled high-confidence BCR sequences for 6635 cells. Primers were removed from paired H and L chain mAb sequences from 34 cells using the “MaskPrimers” command from pRESTO v0.5.11 (38). The 10x Genomics and mAb sequences were combined with paired H and L chain nested PCR sequences from 100 cells. Germline V(D)J gene annotation was performed for all sequences using IgBLAST v1.15.0 (39), with a curated set of Ig reference alleles specific for the C57BL/6 strain of Mus musculus (see above section). IgBLAST output was parsed using Change-O v0.4.6 (40). Additional quality control required sequences to be productively rearranged and have valid V and J gene annotations, consistent chain annotation (excluding sequences annotated with H chain V gene and L chain J gene), and a junction length that is a multiple of 3. Furthermore, only cells with exactly one H chain sequence paired with at least one L chain sequence were kept. After processing, there were 6262 cells with paired H and L chains, including 83 cells with nested PCR sequences, 34 cells with mAb sequences, and 6145 cells with 10x Genomics BCR sequences.

B cell clonal lineages were inferred using hierarchical clustering with single linkage (41). Cells were first partitioned based on common H and L chain V and J gene annotations and junction region lengths, in which junction was defined to be from IMGT codon 104 encoding the conserved cysteine to codon 118 encoding phenylalanine or tryptophan (42). Within each partition, cells whose H chain junction regions were within 0.1 normalized Hamming distance from each other were clustered as clones. This distance threshold was determined by manual inspection in conjunction with kernel density estimates, to identify the local minimum between the two modes of the bimodal distance-to-nearest distribution (Supplemental Fig. 2A). Following clonal clustering, full-length clonal consensus germline sequences were reconstructed for the H chains in each clone with D-segment and N/P regions masked with Ns, resolving any ambiguous gene assignments by majority rule.

Mutation frequency was calculated for cells with 10x Genomics BCRs by counting the number of nucleotide mismatches from the germline sequence in the H chain variable segment leading up to the CDR3. Calculation was performed using the calcObservedMutations function from SHazaM v0.2.3 (40).

Demultiplexed pair-end FASTQ reads were preprocessed using the “cellranger count” command from 10x Genomics’ Cell Ranger v3.1.0 for alignment against the GRCm38 mouse reference v3.0.0 (refdata-cellranger-mm10-3.0.0). A feature unique molecular identifier (UMI) count matrix containing 7485 cells and 31,053 features was generated. The biotypes of the features were retrieved from the gene annotation of Ensembl release 93 (31). Additional quality control was performed as follows. 1) To remove presumably lysed cells, cells with mitochondrial content >15% of all transcripts were removed. 2) To remove likely doublets, cells with >5000 features or 80,000 total UMIs were removed. 3) To remove cells with no detectable expression of common housekeeping mouse genes, cells with no transcript for any of Actb, Gapdh, B2m, Hsp90ab1, Gusb, Ppih, Pgk1, Tbp, Tfrc, Sdha, Ldha, Eef2, Rpl37, Rpl38, Leng8, Heatr3, Eif3f, Chmp2a, Psmd4, Puf60, and Ppia were removed (43, 44). 4) The feature matrix was subset, based on their biotypes, to protein-coding, Ig, and TCR genes that were expressed in at least 0.1% of the cells. 5) Cells with detectable expression of <200 genes were removed. After quality control, the final feature matrix contained 7264 cells and 11,507 genes.

Single-cell gene expression analysis was performed using Seurat v3.1.1 (45). UMI counts measuring gene expression were log normalized. The top 2000 highly variable genes (HVGs) were identified using the “FindVariableFeatures” function with the “vst” method. Mouse homologs for a set of 293 immune-related, “immunoStates” human genes (46) were added to the HVG list, whereas Ig and TCR genes were removed. The mouse homologs were obtained by first looking up the Human and Mouse Homology Class report from MGI (47), accessed on April 6, 2020, and then manually searching the National Center for Biotechnology Information Gene for the human genes for which MGI reported no mouse homolog. The data were then scaled and centered, and principal component analysis was performed based on the expression of the HVGs. Principal component analysis–guided t-distributed stochastic neighbor embedding was performed using the top 20 principal components.

Gene expression–based clusters were identified using the “FindClusters” function with resolution 0.05. Differentially expressed genes for each cluster were identified via the “FindAllMarkers” function using Wilcoxon Rank Sum tests, followed by Bonferroni correction for multiple testing. The identities of the clusters were assigned by examining the expression of canonical marker genes and differentially expressed genes. The PB clusters were based on high expression of Cd79a, Cd79b, Xbp1, Sdc1, and Fkbp11. One of the PB clusters was highly proliferating based on high expression of Mki67, Top2a, Cdk1, Ccna2, and Cdca3. The T cell cluster was based on high expression of Cd8b1, Ms4a4b, Cd3d, Cd3e, Ccr7, and Il7r.

Three-fold serial dilutions of mouse sera and mAbs were incubated with 10² focus-forming units of SARS-CoV-2 at 37°C for 1 h. Ab–virus mixtures were added to Vero E6 cell monolayers in 96-well plates and incubated at 37°C for 1 h. After incubation, cells were overlaid with 1% (w/v) methylcellulose in MEM supplemented with 2% FBS. Plates were harvested 30 h later by removing overlays and were fixed with 4% paraformaldehyde (PFA) in PBS for 20 min at room temperature. Plates were washed six times with PBS and sequentially incubated with 1 μg/ml CR3022 anti-S protein Ab (48) and HRP-conjugated goat anti-human IgG in PBS supplemented with 0.1% saponin and 0.1% BSA. SARS-CoV-2 foci were visualized by incubating monolayers with TrueBlue peroxidase substrate (SeraCare) for 20 min at room temperature and quantitated using an ImmunoSpot microanalyzer (Cellular Technologies). Data were processed and neutralization curves were generated using Prism v8 (GraphPad).

The ACE2 competition binding assay was performed at 25°C on an Octet Red bilayer interferometry instrument (ForteBio) using anti-human IgG Fc biosensors to capture target Ab. Briefly, Abs were loaded onto anti-human IgG Fc pins for 3 min at 10 μg/ml in assay buffer (10 mM HEPES, 150 mM NaCl, 3 mM EDTA, and 0.005% P20 surfactant with 3% BSA). Unbound Abs were washed away, and the IgG-loaded tips were dipped into RBD-containing wells for 1 or 3 min, followed by immersion into wells containing 1 μM ACE2 protein. The mAbs were considered competing if no additional bilayer interferometry signal was observed compared with control mAb hE16 (humanized West Nile virus–specific mAb), whereas increased signal indicated inability of mAbs to block RBD binding to ACE2.

Eight-week-old male BALB/cJ mice (The Jackson Laboratory) were administered 2 mg of anti-IFNAR1 (MAR1-5A3; Leinco) (49) via i.p. injection 24 h prior to intranasal (i.n.) administration of 2.5 × 10⁸ PFU of AdV-hACE2. Five days later, mice were inoculated i.n. with 4 × 10⁵ focus-forming units of SARS-CoV-2. Weight was monitored daily, animals were euthanized 4 or 6 d postinfection and perfused with 20 ml of PBS, and tissues were harvested. For histological analysis, the right lung was inflated with ∼1.2 ml of 10% neutral buffered formalin using a 3-ml syringe and catheter inserted into the trachea. For fixation postinfection, inflated lungs were kept in a 40-ml suspension of neutral buffered formalin for 7 d before further processing. Lungs were embedded in paraffin, and sections were stained with H&E. Tissue sections were visualized using a Nikon Eclipse microscope equipped with an Olympus DP71 color camera or a Leica DM6B microscope equipped with a Leica DFC7000T camera. For viral load analyses, collected tissues were weighed and homogenized with zirconia beads in a MagNA Lyser instrument (Roche Life Science) in 1 ml of DMEM media supplemented with 2% heat-inactivated FBS. Tissue homogenates were clarified by centrifugation at 10,000 rpm for 5 min and stored at −80°C. RNA was extracted using MagMax mirVana Total RNA isolation kit (Thermo Fisher Scientific) and a Kingfisher duo prime extraction machine (Thermo Fisher Scientific). Viral burden was determined by reverse transcriptase quantitative PCR (forward primer: 5′-ATGCTGCAATCGTGCTACAA-3′; reverse primer: 5′-GACTGCCGCCTCTGCTC-3′; probe: /56-FAM/TCAAGGAAC/ZEN/AACATTGCCAA/3IABkFQ/), and for lung samples, plaque assay on Vero E6 cells was used. Briefly, homogenates were serially diluted 10-fold and applied to Vero E6 cell monolayers in 12-well plates. Plates were incubated at 37°C for 1 h with rocking every 15 min. Subsequently, cells were overlaid with 1% (w/v) methylcellulose in MEM supplemented with 2% FBS. Plates were harvested 72 h later by removing overlays and fixed with 4% PFA in PBS for 20 min at room temperature. After removing the 4% PFA, plaques were visualized by adding 1 ml/well 0.05% crystal violet in 20% methanol for 20 min at room temperature. Excess crystal violet was washed away with PBS, and plaques were counted.

We immunized two mice i.m. with 10 μg of recombinant SARS-CoV-2 RBD in squalene-based adjuvant. Fourteen days after primary immunization, mice were boosted twice with 5 μg of recombinant SARS-CoV-2 S protein at a 10-d interval (Fig. 1A). Serum Ab binding to SARS-CoV-2 recombinant trimeric S protein or RBD was measured by ELISA 5 d after the final booster immunization. Serum from both mice demonstrated potent binding to both SARS-CoV-2 RBD and S protein (Fig. 1B). Serum samples from the immunized mice were also evaluated for neutralization of an SARS-CoV-2 isolate (2019 n-CoV/USA_WA1/2020) (21). Potent neutralizing activities (half-maximal inhibitory dilutions of 1:473 and 1:1476) against SARS-CoV-2 were found for both mice in a focus reduction neutralization test (FRNT) (Fig. 1C). These results suggest that our immunization strategy successfully induced RBD and S protein–specific and –neutralizing Ab responses.

To further characterize the Ab response, PBs were sorted from draining lymph nodes pooled from both mice 5 d after the final boost immunization. We sorted single RBD-binding PBs for cloning and evaluation of the Ab response and total PBs in bulk for scRNA-seq (Fig. 1D, Supplemental Fig. 1A). For mAb generation, Ig H, κ, and λ L chain variable genes were cloned into a human IgG1 expression vector and expressed as murine/human chimeric mAbs as previously described (26–28). Thirty-four mAbs were expressed and screened for binding to recombinant SARS-CoV-2 RBD expressed in mammalian cells, of which 26 were positive (Fig. 1E).

A total of 117 IGHV sequences were cloned from two 96-well plates of singly sorted PBs, of which 47 were clonally distinct (Fig. 2A, Supplemental Fig. 2A). Nineteen clonal lineages comprised the 26 mAbs that bound to SARS-CoV-2 RBD. We selected a representative mAb from each clonal lineage (Supplemental Table I) and verified that all 19 mAbs bound to the recombinant SARS-CoV-2 RBD with minimum positive concentrations ≤5 μg/ml (Fig. 2B). To more comprehensively characterize the transcriptional profile, isotype distribution, and somatic hypermutation among responding PBs, we analyzed bulk-sorted total PBs using scRNA-seq. Gene expression–based clustering of PBs revealed two populations, Ki67^hi and Ki67^low, corresponding to proliferation states among responding PBs (Fig. 2C, Supplemental Fig. 2B). We then identified the BCR sequences from the scRNA-seq data that were clonally related to those encoding the RBD-specific mAbs and found that these RBD-specific clones were distributed homogenously between both PB populations (Fig. 2D, Supplemental Fig. 2C). The RBD-specific clones were mostly isotype switched, with IgG⁺ cells comprising the vast majority (640 of 657) of RBD⁺ cells (Fig. 2E). Additionally, the mutation frequency of RBD-specific clones was higher compared with RBD-negative clones (Fig. 2F, Supplemental Fig. 2D), indicating that our immunization strategy resulted in selective enrichment of a more-mature and isotype-switched RBD-specific PB response among the total S protein–induced B cell response.

Multiple amino acid variations exist between SARS-CoV-2 and SARS-CoV RBDs and to a much larger extent between SARS-CoV-2 and MERS-CoV RBDs (13, 50). To determine whether our mAbs recognize distinct or conserved epitopes, we tested their binding to SARS-CoV-2, SARS-CoV, and MERS-CoV S proteins. The 19 mAbs bound recombinant SARS-CoV-2 S protein, with five (2C02, 2E06, 1C05, 1C07, and 2E10) recognizing SARS-CoV, but none binding to MERS-CoV S protein (Fig. 3A–C). The five cross-reactive mAbs recognized the SARS-CoV RBD (Supplemental Fig. 3A–C). Despite binding SARS-CoV-2 RBD, 1A12 and 2H04 weakly bound SARS-CoV S protein but not RBD, and 2B04 weakly bound SARS-CoV and MERS-CoV RBD. Because binding is not an indicator for antiviral capacity, we tested whether any of the mAbs had neutralizing activity against SARS-CoV-2 strain 2019 n-CoV/USA_WA1/2020 using a Vero E6 cell FRNT. Five of the mAbs (1B10, 2B04, 1B07, 1E07, and 2H04) displayed strong neutralizing activity against SARS-CoV-2. Among these, mAb 2B04 displayed the most potent neutralizing activity against SARS-CoV-2, with a remarkable IC₅₀ value of 1.46 ng/ml (Fig. 3D, Supplemental Fig. 3D). All neutralizing mAbs except 2H04 competed with hACE2 for binding to RBD (Supplemental Fig. 3E).

To assess the protective capacities of 2B04 and 2H04 in vivo, we used a mouse model of SARS-CoV-2 infection in which hACE2 is transiently expressed via a nonreplicating adenoviral vector (hACE2-AdV) (22). BALB/c mice were transduced with hACE2-AdV via i.n. administration to establish receptor expression in lung tissues. Animals then received 10 mg/kg 2B04, 2H04, or isotype control via i.p. injection 1 d before infection with the SARS-CoV-2 strain 2019 n-CoV/USA_WA1/2020 (Fig. 4A). Mice receiving 2B04 lost significantly less body weight compared with those receiving isotype control mAb, and those receiving 2H04 trended toward decreased weight loss (Fig. 4B, Supplemental Fig. 4A). Viral load was measured at the peak of viral burden in this model, 4 d postinfection, in the lung and spleen (51). Compared with the isotype control mAb-treated mice, animals receiving 2B04 had 31- and 11-fold lower median levels of viral RNA, and those receiving 2H04 had 6- and 5-fold lower median levels of viral RNA in the lung and spleen, respectively (Fig. 4C, Supplemental Fig. 4B). Furthermore, those receiving 2B04 had no detectable infectious virus in the lungs by plaque assay (Fig. 4D). Consistent with the reduction of infectious virus titers in lungs from animals treated with 2B04, infiltration of inflammatory cells was substantially decreased within the alveolar spaces in 2B04-treated animals compared with those treated with an isotype control mAb (Supplemental Fig. 4C, 4D). Altogether, these data indicate that 2B04 can limit SARS-CoV-2 disease and reduce viral dissemination.

In this article, we describe a panel of RBD-binding mAbs generated from mice immunized with recombinant SARS-CoV-2 RBD and boosted with S protein. Consistent with a recent report (52), none of the neutralizing mAbs strongly cross-reacted with SARS-CoV and MERS-CoV RBD, although one neutralizing SARS-CoV mAb that cross-reacts with SARS-CoV-2 has been recently described (53). Notably, all neutralizing mAbs except 2H04 competed with hACE2 for binding to RBD. 2H04 activity is reminiscent of CR3022, an mAb that recognizes an epitope within the RBD that does not overlap with the hACE2 binding site (48). This result suggests that the isolated anti-RBD mAbs described in this article bind nonoverlapping epitopes and can efficiently neutralize the virus via potentially distinct mechanisms and therefore may demonstrate enhanced protective capacity if used in combination. The latter point is critical as it would decrease the chance of escape mutants emerging following the use of each mAb alone. Several mAbs recognized epitopes that apparently overlapped with hACE2 binding site based on the hACE2 competition assay but did not show substantial neutralizing activity. The basis for this remains unknown but could be attributed to low binding affinity or steric hindrance that impedes engagement of the RBD on the virion surface.

One mAb, 2B04, was particularly potent in the in vitro FRNT with an IC₅₀ of 1.46 ng/ml and was protective in vivo. Other SARS-CoV-2–neutralizing mAbs that bind RBD and are protective in vivo in both prophylactic and therapeutic settings have recently been described (54, 55), although their in vitro neutralizing concentrations are at least 10-fold higher. A critical question for any candidate therapeutic mAb is whether a suboptimal dose could enhance viral infection (56, 57). Further studies could best address this question using native models of infection, such as hamsters (58).

In summary, we isolated an array of 19 PB-derived clonally distinct murine mAbs that are directed against the RBD within the S protein of the SARS-CoV-2 virus. Five of these mAbs have strong neutralizing activity (IC₅₀ < 0.5 μg/ml) against wild-type infectious SARS-CoV-2. One mAb, 2B04, showed highly potent neutralizing activity, protected mice against weight loss, and reduced viral burden, making it a promising candidate for therapeutic development.

We thank Erica Lantelme for facilitating sorting, the Genome Technology Access Center in the Department of Genetics at Washington University School of Medicine for scRNA-seq library preparation and sequencing, and the Yale Center for Research Computing for use of high-performance computing infrastructure.

A.H.E. is a consultant for Inbios and Fimbrion Therapeutics. S.H.K. receives consulting fees from Northrop Grumman. M.S.D. is a consultant for Inbios, Eli Lilly, Vir Biotechnology, and NGM Biopharmaceuticals and is on the Scientific Advisory Board of Moderna. The Ellebedy laboratory received funding under sponsored research agreements from Emergent BioSolutions. The Diamond laboratory at Washington University School of Medicine has received sponsored research agreements from Moderna and Emergent BioSolutions. All other authors declare no conflict of interest.