A global collaboration for systematic analysis of broad-ranging antibodies against the SARS-CoV-2 spike protein

Abstract

The Coronavirus Immunotherapeutic Consortium (CoVIC) conducted side-by-side comparisons of over 400 anti-SARS-CoV-2 spike therapeutic antibody candidates contributed by large and small companies as well as academic groups on multiple continents. Nine reference labs analyzed antibody features, including in vivo protection in a mouse model of infection, spike protein affinity, high-resolution epitope binning, ACE-2 binding blockage, structures, and neutralization of pseudovirus and authentic virus infection, to build a publicly accessible dataset in the database CoVIC-DB. High-throughput, high-resolution binning of CoVIC antibodies defines a broad and predictive landscape of antibody epitopes on the SARS-CoV-2 spike protein and identifies features associated with durable potency against multiple SARS-CoV-2 variants of concern and high in vivo efficacy. Results of the CoVIC studies provide a guide for selecting effective and durable antibody therapeutics and for immunogen design as well as providing a framework for rapid response to future viral disease outbreaks.

Keywords: COVID-19; CP: Immunology; SARS-CoV-2 spike protein; coronavirus; high-resolution epitope binning; neutralization; therapeutic antibodies.

Graphical abstract

Figure 1

High-throughput surface plasmon resonance (HT-SPR), with the soluble receptor binding domain (RBD) and full-length spike as antigens, defines a broad range of epitope communities (A and B) Dendrograms based on comparisons of competition profile similarity with (A) soluble RBD and (B) full-length spike antigens. Cutoffs were applied to create clusters having highly related profiles. The inner and outer faces, as well as the receptor binding motif (RBM) of RBD are shown. (C) Stacked histogram with full-length epitope communities on the y axis. Individual boxes representing each antibody are shaded according to the RBD epitope community. (D) Location of epitope communities on the RBD. A space-filling diagram of the full-length spike ectodomain with one RBD “up” is shown with relative positions of RBD communities and S1, NTD, and S2 domains indicated. A top view is shown on the lower left with the RBM shaded red. The three monomers of spike are shaded white and light and dark gray. The center shows a space-filling model of the RBD (adapted from PDB: 7A94¹¹) with the RBM shaded red. Colored ovals correspond to the general regions targeted by the epitope communities. Vertical and horizontal dashed lines roughly divide the RBD into upper and lower quadrants and an inner and outer face, respectively. Classes defined by Barnes et al. are shown. The far right shows side views of the RBD.

Figure 2

Epitope communities have characteristic degrees of ACE-2 blocking activity and competition (A) Boxplot of the percentage blockage of ACE-2 binding to spike by CoVIC antibodies. The mean value is at the intersection of the darker- and lighter-shaded regions, which represent the lower and upper quartile, respectively. Whiskers extend to 1.5 times the interquartile range. Circles correspond to individual CoVIC antibodies. (B) Competition matrix of CoVIC antibodies, with dark and light blue boxes indicating competition and no competition, respectively, between the antibody pair. Columns and rows represent antibodies as analytes and ligands, respectively. Orange circles represent RBD-2 sub-communities; circles with red lines indicate unfavorable pairing based on competition. (C) Location of epitope communities on the RBD explain, in part, predicted competition between RBD-2 sub-communities and other main communities. The images show top views of the four RBD-2 sub-communities, with dashed ovals indicating epitope communities predicted to compete for binding.

Figure 3

Correlation of results for neutralization assays using pseudovirus and authentic virus and effect of VoCs on CoVIC antibody neutralization activity (A) Neutralization activity for CoVIC antibodies tested with authentic virus carrying mNeonGreen (Auth-M) or luciferase reporter (Auth-L) or rVSV pseudovirus with either luciferase (PNV-L) or GFP (PNV-G) reporters. Neutralization for nearly all antibodies was measured against authentic or pseudotyped Wuhan virus (W). Activity against Beta (B.1.351) was measured for the indicated antibodies using authentic virus with a luciferase reporter and with both pseudovirus platforms. Neutralization toward the Delta variant (B.1.617.2) was tested using both pseudovirus platforms, while neutralization of Omicron (BA.1) and two Omicron sub-variants (BA1.1 and BA.2) was measured using GFP pseudovirus. Shading corresponds to IC₅₀ (ng/mL) for Auth-M and -L and for PNV-G. Neutralization of Omicron was measured with 25 μg/mL and 250 ng/mL antibody; the percentage of infected cells using 250 ng/mL is shown. Antibodies are grouped by RBD community and sub-grouped by full-length epitope community with colors corresponding to those in Figure 1. Black dots indicate antibodies that neutralized all VoCs tested. (B) Locations of epitope communities on RBD. Colored ovals correspond to the region targeted by the epitope community. In the table, green boxes highlight the presence of mutations at the indicated residue in Beta (B), Delta (Δ), and Omicron (O) VoCs.

Figure 4

Protective efficacy is related to neutralization and affinity (A) Relationship between RBD community and affinity. Each antibody that was tested in the K18-hACE2 transgenic mouse model of SARS-CoV-2 infection and that had K_D < 1.0 × 10⁻⁹ M is plotted. (B) Relationship between neutralization of authentic virus with an mNeonGreen reporter (Authentic-M) and protective efficacy. IC₅₀ values (ng/mL) are shown on the y axis. Circles in (A) and (B) correspond to individual CoVIC antibodies, with shading corresponding to survival in the K18-hACE2 transgenic mouse model of infection. Dark teal represents the highest protective efficacy. Antibodies were tested using groups of 10 mice with daily monitoring of body weight. Survival is expressed as the percentage of mice surviving at 10 days post-infection. Epitope communities are sorted by the percentage of antibodies within the community that offered at least 60% survival. In the box plots, the mean value is at the intersection of the darker and lighter-shaded regions that represent the lower and upper quartile, respectively. Whiskers extend to 1.5-times the interquartile range.

Figure 5

Representative negative-stain electron microscopy (nsEM) structures show the variety of epitope footprints and binding mechanisms (A) nsEM structures of CoVIC panel antibodies in complex with the full-length spike ectodomain were determined. The epitope footprint is shaded by RBD epitope community, and the full-length epitope community is in the upper right of each square. Side (left) and top (right) views of the spike protein are shown. Structures were determined using full-length IgG, except for those with an asterisk by the CoVIC ID, for which Fab or ScFv was used. Black circles indicate antibodies that exhibit bivalent binding. (B) Side and top views of the CoVIC-154-spike complex. The CoVIC-154 variable domain is colored green. (C) Side view of CoVIC-154 Fab bound to spike. Two of the three Fab variable domains are modeled as green ribbon diagrams. The third binding site is illustrated as the antibody footprint on the spike surface. The light chain footprint is illustrated in light green (upper left side), the heavy chain footprint is in dark green, and residues contacted by both chains are in blue. (D) Hydrogen bonding between spike monomers and (i) heavy and (ii and iii) light chains of CoVIC-154. Residues participating in hydrogen bonding are labeled. Spike monomers 1 and 2 are shown in light and dark gray, respectively, and the heavy and light chains are shown in dark and light green, respectively.

Figure 6

Escape-mutation assay shows regions on spike vulnerable to antigenic escape To identify escape mutations, antibodies were incubated with authentic virus, and the mixtures were added to Vero E6/TMPRSS2 cells. Virus was harvested and the spike gene was sequenced. (A) The top row lists all the amino acid positions at which mutations were detected. Amino acids present in the Wuhan strain (G614, B.1) are shown in the second row, and the rows below list the indicated variants. “X” indicates that the residue was mutated in a variant. Squares indicate that the detected mutation was within one or two residues of an amino acid mutated in a variant. Rows at the bottom show mutations reported in public databases (e.g., GSAID). (B) Bubble plot of detected escape mutations. Each circle shows the residue that was mutated (some residues had more than one amino acid change detected). Circle size corresponds to the number of antibodies affected by the mutation. Circle color indicates RBD epitope community; some mutations affected multiple epitope communities and the circle is divided according to the percentage of each epitope group affected. Residue numbers radiate outward, with the lowest residue number in the center. (C) Detected escape mutations.

Figure 7

Five-fold cross-validation using different combinations of antibody features to predict in vivo protection The analysis was carried out on a sub-set of antibodies for which in vivo data and all seven features considered were available. Mean Spearman’s correlation coefficients over three replicates are shown; error bars indicate SEM. (A) Plot of the ability of individual antibody features to predict in vivo protection. Then, (B) two, (C) three, or (D) four features were combined to determine whether the predictive performance was enhanced. Auth-M and Auth-L indicate neutralization of authentic virus with mNeonGreen and luciferase reporters, respectively. PNV-L and PNV-G indicate neutralization of pseudovirus with luciferase and GFP reporters, respectively. The PNV-G assay tested a single concentration (250 ng/mL), and the other neutralization assays reported IC₅₀ values determined from eight-point curves. RBD and FL are epitope communities determined from binning with soluble RBD and trimeric full-length spike ectodomain. Affinity is the K_D (M) for the D614G full-length ectodomain and ACE-2 block is the percentage blockage in the presence of antibody.