Potent neutralizing antibodies from COVID-19 patients define multiple targets of vulnerability

Abstract

The rapid spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has had a large impact on global health, travel, and economy. Therefore, preventative and therapeutic measures are urgently needed. Here, we isolated monoclonal antibodies from three convalescent coronavirus disease 2019 (COVID-19) patients using a SARS-CoV-2 stabilized prefusion spike protein. These antibodies had low levels of somatic hypermutation and showed a strong enrichment in VH1-69, VH3-30-3, and VH1-24 gene usage. A subset of the antibodies was able to potently inhibit authentic SARS-CoV-2 infection at a concentration as low as 0.007 micrograms per milliliter. Competition and electron microscopy studies illustrate that the SARS-CoV-2 spike protein contains multiple distinct antigenic sites, including several receptor-binding domain (RBD) epitopes as well as non-RBD epitopes. In addition to providing guidance for vaccine design, the antibodies described here are promising candidates for COVID-19 treatment and prevention.

Fig. 1. Design of SARS-CoV-2 S protein and serology of COSCA1, COSCA2, and COSCA3.

(A) (Top) Schematic overview of the authentic SARS-CoV-2 S protein with the signal peptide shown in blue and the S1 (red) and S2 (yellow) domains separated by a furin-cleavage site (RRAR; top). (Bottom) Schematic overview of the stabilized prefusion SARS-CoV-2 S ectodomain, where the furin cleavage site is replaced with a glycine linker (GGGG), two proline mutations are introduced (K986P and V987P), and a trimerization domain (cyan) preceded by a linker (GSGG) is attached. (B) Binding of sera from COSCA1, COSCA2, and COSCA3 to prefusion SARS-CoV-2 S protein as determined by ELISA. The mean values and SDs of two technical replicates are shown. (C) Neutralization of SARS-CoV-2 pseudovirus by heat-inactivated sera from COSCA1, COSCA2, and COSCA3. The mean and SEM of at least three technical replicates are shown. The dotted line indicates 50% neutralization.

Fig. 2. Characterization of SARS-CoV-2 S protein–specific B cells derived from COSCA1, COSCA2, and COSCA3.

(A) Representative gates of SARS-CoV-2 S protein–specific B cells shown for a naïve donor (left panel) or COSCA1 (middle left panel). Each dot represents a B cell. The gating strategy to identify B cells is shown in fig. S2. From the total pool of SARS-CoV-2 S protein–specific B cells, CD27⁺CD38^– memory B cells (Mem B cells; blue gate) and CD27⁺CD38⁺ B cells were identified (middle panel). From the latter gate, PBs/PCs (CD20^–; red gate) could be identified (middle right panel). SARS-CoV-2 S protein–specific B cells were also analyzed for their IgG or IgM isotype (right panel). (B) Frequency of SARS-CoV-2 S protein–specific B cells in total B cells, Mem B cells, and PBs/PCs. Symbols represent individual patients, as shown in (D). (C) Comparison of the frequency of Mem B cells (CD27⁺CD38^–) and PB/PC cells (CD27⁺CD38⁺CD20^–) between the specific (SARS-CoV2 S⁺⁺) and nonspecific B cells (gating strategy is shown in fig. S2). Symbols represent individual patients, as shown in (D). Statistical differences between two groups were determined using paired t test (*P = 0.034). (D) Comparison of the frequency of IgM⁺, IgG⁺, and IgM^–IgG^– B cells in specific and nonspecific compartments. Bars represent means; symbols represent individual patients.

Fig. 3. Genotypic characterization of SARS-CoV-2 S protein–specific B cell receptors.

(A) Maximum-likelihood phylogenetic tree of 409 isolated paired B cell receptor HCs. Each color represents sequences isolated from different patients (COSCA1, COSCA2, and COSCA3). (B) Violin plot showing SHM levels (%; nucleotides) per patient. The dot represents the median SHM percentage. (C) Distribution of CDRH3 lengths in B cells from COSCA1, COSCA2, and COSCA3 (purple, n = 323) versus a representative naïve population from three donors (cyan, n = 9.791.115) (37). (D) Bar graphs showing the mean (± SEM) VH gene usage (%) in COSCA1, COSCA2, and COSCA3 (purple, n = 323) versus a representative naïve population (cyan, n = 363,506,788). The error bars represent the variation between different patients (COSCA1, COSCA2, and COSCA3) or naïve donors (37). Statistical differences between two groups were determined using unpaired t tests (with Holm–Sidak correction for multiple comparisons, adjusted P values: *P < 0.05; **P < 0.01; ***P < 0.001).

Fig. 4. Phenotypic characterization of SARS-CoV-2 S protein–specific mAbs.

(A) Bar graph depicting the binding of mAbs from COSCA1 (blue), COSCA2 (red), and COSCA3 (yellow) to SARS-CoV-2 S protein (dark shading) and SARS-CoV-2 RBD (light shading) as determined by ELISA. Each bar indicates the representative area under the curve (AUC) of the mAb indicated below from two experiments. The gray area represents the cutoff for binding (AUC = 1). The maximum concentration of mAb tested was 10 μg/ml. (B) Scatter plot depicting the binding of mAbs from COSCA1, COSCA2, and COSCA3 [see (C) for color coding] to SARS-CoV-2 S protein and SARS-CoV-2 RBD as determined by ELISA. Each dot indicates the representative AUC of a mAb from two experiments. (C) Midpoint neutralization concentrations (IC₅₀) of SARS-CoV-2 pseudovirus (left) or authentic SARS-CoV-2 virus (right). Each symbol represents the IC₅₀ of a single mAb. For comparability, the highest concentration was set to 10 μg/ml, although the actual start concentration for the authentic virus neutralization assay was 20 μg/ml. The IC₅₀s for pseudotyped and authentic SARS-CoV-2 virus of a selection of potently neutralizing RBD and non-RBD–specific mAbs (with asterisk) are shown in the adjacent table. Colored shading indicates the most potent mAbs from COSCA1, COSCA2, and COSCA3.

Fig. 5. Antigenic clustering of SARS-CoV-2 S protein–specific mAbs.

(A) Dendrogram showing hierarchical clustering of the SPR-based cross-competition heat map (table S2). Clusters are numbered I to XI and are depicted with color shading. ELISA binding to SARS-CoV-2 S protein, SARS-CoV S protein, and SARS-CoV-2 RBD as presented by AUC and neutralization IC₅₀ (μg/ml) of SARS-CoV-2 is shown in the columns on the left. ELISA AUCs are shown in gray (AUC < 1) or blue (AUC > 1), and neutralization IC₅₀ is shown in gray (>10 μg/ml), blue (1 to 10 μg/ml), violet (0.1 to 1 μg/ml), or purple (0.001 to 0.1 μg/ml). Asterisks indicate antibodies that cross-neutralize SARS-CoV pseudovirus. (B) Composite figure demonstrating binding of NTD-mAb COVA1-22 (blue) and RBD mAbs COVA2-07 (green), COVA2-39 (orange), COVA1-12 (yellow), COVA2-15 (salmon), and COVA2-04 (purple) to SARS-CoV-2 spike (gray). The spike model (PDB 6VYB) is fit into the density. (C) Magnification of SARS-CoV-2 spike comparing epitopes of RBD mAbs with the ACE2-binding site (red) and the epitope of mAb CR3022 (blue). (D) Side (left) and top (right) views of the 3D reconstruction of COVA2-15 bound to SARS-CoV-2 S protein. COVA2-15 binds to both the down (magenta) and up (salmon) conformations of the RBD. The RBDs are colored blue in the down conformation and black in the up conformation. The angle of approach for COVA2-15 enables this broader recognition of the RBD while also partially overlapping with the ACE2-binding site and therefore blocking receptor engagement.