Multivalency transforms SARS-CoV-2 antibodies into ultrapotent neutralizers

Abstract

SARS-CoV-2, the virus responsible for COVID-19, has caused a global pandemic. Antibodies can be powerful biotherapeutics to fight viral infections. Here, we use the human apoferritin protomer as a modular subunit to drive oligomerization of antibody fragments and transform antibodies targeting SARS-CoV-2 into exceptionally potent neutralizers. Using this platform, half-maximal inhibitory concentration (IC₅₀) values as low as 9 × 10^-14 M are achieved as a result of up to 10,000-fold potency enhancements compared to corresponding IgGs. Combination of three different antibody specificities and the fragment crystallizable (Fc) domain on a single multivalent molecule conferred the ability to overcome viral sequence variability together with outstanding potency and IgG-like bioavailability. The MULTi-specific, multi-Affinity antiBODY (Multabody or MB) platform thus uniquely leverages binding avidity together with multi-specificity to deliver ultrapotent and broad neutralizers against SARS-CoV-2. The modularity of the platform also makes it relevant for rapid evaluation against other infectious diseases of global health importance. Neutralizing antibodies are a promising therapeutic for SARS-CoV-2.

Fig. 1. Avidity enhances binding and neutralization of VHH against SARS-CoV-2.

a Schematic representation of a monomeric VHH domain and its multimerization using a conventional Fc (dark red) scaffold or human apoferritin (gray). b Size exclusion chromatography and SDS-PAGE of apoferritin alone (gray) and VHH-72 apoferritin particles (gold). c Negative stain electron microscopy of VHH-72 apoferritin particles. (Scale bar 50 nm, representative of two independent experiments). d Comparison of the binding avidity (apparent K_D) of VHH-72 to SARS-CoV-2 S protein when displayed in a bivalent (dark red) or 24-mer (gold) format. Bars indicate the mean values of n = 2 biologically independent experiments. Apparent K_D lower than 10⁻¹² M (dash line) is beyond the instrument detection limit. e Neutralization potency against SARS-CoV-2 PsV (color coding is as in (d)). One representative out of two biologically independent replicates with similar results is shown. Mean values ± SD of two technical replicates is represented in the plot. Median IC₅₀ values of the two biologically independent replicates are shown. Source data are provided as a Source Data file.

Fig. 2. Bioavailability, biodistribution, and immunogenicity of a mouse surrogate Multabody.

a Binding kinetics of WT and Fc-modified (LALAP mutation) MB to mouse FcγRI (left) and mouse FcRn at endosomal (middle) and physiological (right) pH in comparison to the parental IgG. Two-fold dilution series from 100 to 3 nM (IgG) and 10 to 0.3 nM (MB) were used. Red lines represent raw data; black lines represent global fits. b Five male C57BL/6 mice per group were used to assess the serum concentration of a surrogate mouse MB, a Fc-modified MB (LALAP mutation), and parental mouse IgGs (IgG1 and IgG2a subtypes) after subcutaneous administration of 5 mg/kg. c MB and IgG2a samples were labeled with Alexa-647 for visualization of their biodistribution post subcutaneous injection into three male BALB/c mice/group via live noninvasive 2D whole body imaging. 15 nm fluorescently-labeled gold nanoparticles (GNP), which have a similar Rh value as the Multabody are shown as a comparator. d Five male C57BL/6 mice per group were used to assess any anti-drug-antibody response induced by the mouse surrogate Multabody in comparison to parental IgG and a species-mismatched malaria PfCSP peptide fused to Helicobacter pylori ferritin (HpFerr). Mean values ± SD of n =  5 mice is shown in (b) and (d). Source data of panels b and d are provided as a Source Data file.

Fig. 3. Protein engineering to multimerize IgG-like particles against SARS-CoV-2.

a Schematic representation of the human apoferritin split design. b Negative stain electron micrograph of the MB. (Scale bar 50 nm, representative of two independent experiments). c Hydrodynamic radius (Rh) of the MB. d Avidity effect on the binding (apparent K_D) of 4A8 (purple) and BD23 (gray) to the SARS-CoV-2 Spike. e Sensograms of BD23 IgG and MB with different Fc sequence variants binding to FcγRI (top row), FcRn at endosomal pH (middle row) and FcRn at physiological pH (bottom row). Red lines represent raw data whereas black lines represent global fits. f Neutralization of SARS-CoV-2 PsV by 4A8 and BD23 IgGs and MBs. Representative data of three biologically independent samples. The mean values ± SD for two technical replicates is shown in each neutralization plot. Median IC₅₀ values of the three biologically independent replicates are indicated. Source data are provided as a Source Data file.

Fig. 4. The Multabody enhances the potency of human mAbs from phage display.

a Work flow for the identification of potent anti-SARS-CoV-2 neutralizers using the MB technology. Created with Biorender. b Comparison of neutralization potency between IgGs (cyan) and MBs (pink) that display the same human Fab sequences derived from phage display. c IC₅₀ values fold increase upon multimerization. d Apparent affinity (K_D), association (k_on), and dissociation (k_off) rates of the most potent neutralizing MBs (pink) compared to their IgG counterparts (cyan) for binding the SARS-CoV-2 S protein. Three biological replicates and their mean are shown for IC₅₀ values in (b) and (c). Source data of panels b–d are provided as a Source Data file.

Fig. 5. Epitope delineation of the most potent mAb specificities.

a Surface and cartoon representation of RBD (light green for the core and dark green for RBM) and ACE2 (light brown) binding. Heat map showing binding competition experiments. High signal responses (red) represent low competition while low signal responses (white) correspond to high competition. Epitope bins are highlighted by dashed-line boxes. b 15.0 Å filtered cryo-EM reconstruction of the Spike (gray) in complex with Fab 80 (yellow), 298 (orange), and 324 (red). The RBD and NTD are shown in green and blue, respectively. c Cryo-EM reconstruction of the Fab 46 (pink) and RBD (green) complex. A RBD secondary structure cartoon is fitted into the partial density observed for the RBD. d Crystal structure of the ternary complex formed by Fab 52 (purple), Fab 298 (orange), and RBD (green). e Composite image depicting the side and top view of the unliganded (PDB 6XM4) and the antibody-bound SARS-CoV-2 spike with available PDB or EMD entries^,,,,,,,–. Inset: close up view of antibodies targeting different antigenic sites on the RBD. The mAb with the lowest reported IC₅₀ value against SARS-CoV-2 PsV was selected as a representative antibody of the bin (highlighted in bold) and those antibodies with similar binding epitopes are listed in the same color below (color coding of Spike, NTD and RBD as in (b)). Individual protomers in the unliganded spike are shown in white, pink, and purple.

Fig. 6. Multabodies overcome SARS-CoV-2 sequence diversity.

a Cartoon representation of the RBD showing four naturally occurring mutations as spheres. The epitopes of mAbs 52 (light pink) and 298 (yellow) are shown as representative epitopes of each bin. b Affinity and c IC₅₀ fold-change comparison between WT and mutated RBD and PsV, respectively. d Neutralization potency of IgG (gray bars) vs MB (dark red bars) against SARS-CoV-2 PsV variants in comparison to WT PsV. e Neutralization potency comparison of two IgG cocktails (three IgGs), monospecific MB cocktails (three MBs) and tri-specific MBs against WT SARS-CoV-2 PsV and variants. mAbs sensitive to one or more PsV variants (d) were selected to generate the cocktails and the tri-specific MBs. f Neutralization potency of the tri-specific 298-80-52 MB against SARS-CoV-2 B.1.351 PsV variant. g IC₅₀ values in PsV (y-axis) and replication competent SARS-CoV-2 virus (SB2-P4-PB: x-axis) demonstrating the ability of tri-specific MBs (red) to enhance potency across a wide range of mAb characteristics (blue and black). h IC₅₀ values fold increase upon multimerization. The mean of three biological replicates is shown in (b–h). Source data of panels d–h are provided as a Source Data file.