Converting non-neutralizing SARS-CoV-2 antibodies into broad-spectrum inhibitors

Abstract

Omicron and its subvariants have rendered most authorized monoclonal antibody-based treatments for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) ineffective, highlighting the need for biologics capable of overcoming SARS-CoV-2 evolution. These mostly ineffective antibodies target variable epitopes. Here we describe broad-spectrum SARS-CoV-2 inhibitors developed by tethering the SARS-CoV-2 receptor, angiotensin-converting enzyme 2 (ACE2), to known non-neutralizing antibodies that target highly conserved epitopes in the viral spike protein. These inhibitors, called receptor-blocking conserved non-neutralizing antibodies (ReconnAbs), potently neutralize all SARS-CoV-2 variants of concern (VOCs), including Omicron. Neutralization potency is lost when the linker joining the binding and inhibitory ReconnAb components is severed. In addition, a bi-functional ReconnAb, made by linking ACE2 to a bi-specific antibody targeting two non-overlapping conserved epitopes, defined here, shows sub-nanomolar neutralizing activity against all VOCs, including Omicron and BA.2. Given their conserved targets and modular nature, ReconnAbs have the potential to act as broad-spectrum therapeutics against SARS-CoV-2 and other emerging pandemic diseases.

Fig. 1. Conservation of the SARS-CoV-2 spike protein.

a, Sequence conservation from 44 related spike proteins overlaid on the SARS-CoV-2 spike protein structure (left) and the SARS-CoV-2 RBD (right; residues 319–541) (PDB ID: 6VXX) identifies a highly conserved patch in S2. Color gradient is a step gradient of conservation containing nine total steps of sequence conservation identified from the ConSurf database; gradient is shown on the bottom. b, Sequence identity for all residues in the SARS-CoV-2 spike protein compared to a set of 44 related coronavirus spike proteins shows higher conservation in the S2 relative to the S1. A value of 1.0 means perfect identity across all compared coronavirus proteins. RBD and NTD domains of SARS-CoV-2 spike are labeled on the top; S1 and S2 domains are labeled on the bottom.

Fig. 2. Non-RBD antibodies, selected to prioritize diversity, were used to identify non-RBD SARS-CoV-2 antibodies that bind SARS-CoV-1, a surrogate for epitope conservation.

a, A phylogenetic tree of 422 HC sequences, from our curated library of 696 anti-SARS-COV-2 spike antibodies, generated using Geneious Prime. Germline alleles are not shown. Forty-eight selected clones are shown as stars. Histograms of the HC and LC (b) CDR3 lengths and (c) V-gene usage from the 48 selected non-RBD clones indicated in a. d, A binding profile of the scFv-yeast library produced from the sequences identified in a and Extended Data Fig. 1. e, BLI binding of identified cross-reactive clones expressed as IgGs at 100 nM to SARS-CoV-2 spike (left) or SARS-CoV-1 spike (right). f, BLI competition binding assay of the seven cross-reactive antibodies binding to SARS-CoV-2 (left) and SARS-CoV-1 (right). White indicates no binding of the tested antibody, indicating that the antibodies compete for binding. Antibodies that compete are surrounded by dotted lines; unique competition groups are surrounded by solid lines. The five unique competition groups are labeled on the SARS-CoV-2 binding competition map. Sites A.1–A.4 are indicated as an overlapping supersite. Loading antibodies are indicated in columns, and competing antibodies are indicated in rows. g, Binding of antibody Fab fragments at 200 nM against SARS-CoV-2 spike. Hashed lines show K_D fit determined using Prism.

Fig. 3. scFv-based ReconnAbs tether ACE2 to cross-reactive, non-neutralizing antibodies.

a, A schematic of an scFv-based ReconnAb protein before and after TEV cleavage. Estimated molecular weights of cleavage products are shown beneath both. b, Schematic of ReconnAb activity and the dependence on the tether. The scFV binds to a conserved site, and then ACE2 interacts with the RBD. Upon TEV cleavage, the ACE2 has lower apparent affinity and does not bind the RBD (right). Conserved sites are shown as teal; the remainder of the spike monomers are shown as tints of brown and ACE2 as dark brown. c, SDS-PAGE demonstrates that ReconnAbs are readily cleaved by TEV. 1. Full-length ReconnAb; 2. ACE2; 3. scFv. d, BLI binding of uncleaved and TEV-cleaved ReconnAbs to either SARS-CoV-2 spike (left) or SARS-CoV-1 spike (right) shows a reduction in binding upon TEV cleavage. e, BLI binding of hFc-ACE2 to SARS-CoV-2 spike, which has been pre-associated with ReconnAbs either uncleaved (solid lines) or cleaved (hashed lines), shows that TEV-cleaved ReconnAbs do not compete with hFc-ACE2 binding. Binding of hFc-ACE2 without competitor is shown on the right (dotted line).

Fig. 4. ReconnAbs show broad-spectrum neutralization of SARS-CoV-2 VOCs.

a, Intact scFv-based ReconnAbs (orange) show potent neutralization of SARS-CoV-2 VOCs. Inhibition is markedly reduced upon TEV cleavage (teal). Pseudoviral IC₅₀ for ReconnAbs against a range of SARS-CoV-2 VOCs with and without TEV cleavage. IC₅₀ values shown are the average of two independent experiments. b–d, A bi-functional IgG ReconnAb shows potent neutralization of SARS-CoV-2 VOCs. b, A schematic representation of the CV10-2449–ACE2 CrossMAb indicates linkage of ACE2 and bi-functional Fab arms. c, Pseudoviral IC₅₀ for CV10-2449–ACE2 CrossMAb against a range of SARS-CoV-2 VOCs with and without TEV cleavage. IC₅₀ values shown are the average of two independent experiments. The red dotted line indicates the average neutralization of Fc-ACE2 as in Supplementary Fig. 12. d, Intact IgG CrossMAb ReconnAbs (orange) show neutralization of SARS-CoV-2 BA.2. Inhibition is markedly reduced upon TEV cleavage (teal). The CrossMAb IgG ReconnAb neutralizes the BA.2 Omicron variant at ~50 pM NT₅₀. Error bars denote standard deviation. A representative plot is shown.

Extended Data Fig. 1. The non-RBD library was selected to prioritize diversity.

A phylogenetic tree generated using Geneious Prime of 436 light chain sequences from a curated library of 696 anti-SARS-COV-2 spike antibody sequences. Labels same as Fig. 1. Germline alleles are not shown. Antibodies denoted with names were cross-reactive for SARS-CoV-1.

Extended Data Fig. 2. hCoV proteins expression is optimized with a GCN4 tag.

(A) A schematic representation of constructs tested for expression. 6 hCoV spike proteins, either FL or truncated, tested with 4 different C-terminal trimerization domains and tags. ∆C truncations shown in the middle, SARS-CoV-2 spike numbering. (B) linkers tested. For those without Avi tags, the underlined portions were removed. (C) (top) BLI binding responses from isolated Expi supernatants binding to His 1 K octet tips. Higher response corresponds to higher protein expression. (bottom) An anti-his tag dot blot assay of the supernatants tested in the BLI binding. Dot blot shows good correspondence with the BLI binding. Influenza hemagglutinin used as a positive control, mock transfection as a negative control. (D) Yield determined for full length 2 P hCoV constructs containing the GCN4-Avi-His tag from a duplicate experiment of 100 mL transfections, error bars denote standard deviation.

Extended Data Fig. 3. The non-RBD yeast library was sorted for SARS-CoV-1 binding antibodies.

The gating scheme utilized to collect cross-reactive SARS-CoV-2 and SARS-CoV-1 binding yeast. Both the high and low gates were combined following the sort.

Extended Data Fig. 4. Antibodies isolated from yeast sorts are cross-reactive.

(A) SDS-page analysis of of IgG proteins made from clones identified in the SARS-CoV-1 FACS sorts. MW ladders are in the left-most lanes of the two gels. (B) ELISA binding of IgG proteins made from scFv clones identified by FACS sorting for SARS-CoV-1 binding. Biotinylated hCoV antigens were plated and dilutions of IgGs were tested for binding. Normalized A₄₅₀ calculated by adjusting for pathlength. Except for COV2-2490, CV21, COVA2-18 all IgGs bind to SARS-CoV-2 and SARS-CoV-1. (C) BLI binding for the 7 identified SARS-CoV-1 cross-reactive clones against MERS (left) or OC43 (middle) spike proteins. Only COV2-2449 shows any binding affinity for MERS or weakly to OC43 spike proteins. No Fabs tested bind to the SARS-CoV-2 NTD, consistent with its low sequence conservation between SARS-CoV-2 and SARS-CoV-1.

Extended Data Fig. 5. Antibodies isolated from yeast sorts are non-neutralizing.

Single dilution (100 nM) neutralization against SARS-CoV-2 Wuhan-Hu-1 for the 7 identified cross-reactive antibodies. All antibodies show no neutralization at a high concentration of 100 nM. Error bars denote standard deviation.

Extended Data Fig. 6. A longer linker length does not impact neutralization potency of ReconnAbs.

(A) The linker used to tether ACE2 to either the C-terminus of the scFv or C-terminus of the COV2-2449 LC. The linker was designed to contain a Hexa-His tag for purification and a TEV site to facilitate proteolysis. (B) A longer linker length did not alter ReconnAb activity. One scFv (targeting either epitope A or B – as in Fig. 2f) was mutated to contain a linker that was 7aa longer than the previous study (B). An SDS-Page gel demonstrates that these proteins proteolyzed identically to those with the shorter linkers (C). The neutralization potency was not impacted by the increased linker length (D). Data shown is the average of 2 neutralization experiments.

Extended Data Fig. 7. The bifunctional IgG-based ReconnAb expresses and binds to SARS-CoV-2 spike as expected.

(A) The CV10-2449-ACE2-CrossMAb ReconnAb contains all designed components by SDS-PAGE, assayed before or after TEV cleavage and with or without 2-mercaptoethanol (BME). 1. Full-length ReconnAb, 2. Cleaved CrossMAb CV10-2449 IgG, 3. COV2-2449-LC-ACE2 fusion, 4. ACE2 (reduced ACE2 shows double banding), 5. HC, 6. Cleaved 2449-LC with linker, bottom band CV10 LC. (B) BLI binding of CV10-2449-ACE2-CrossMAb (brown) or the TEV cleaved form (teal) to SARS-CoV-2 spike. Binding is reduced upon TEV cleavage.

Extended Data Fig. 8. The bifunctional IgG-based ReconnAb blocks ACE2 binding and interacts with FcγRI.

(A) Relative hFc-ACE2 binding to SARS-CoV-2 spike protein which has been pre-associated with 200 nM CV10-2449-XMAb (left) or the TEV cleaved form (right). No competitor was set to a value of 1.0. (B) The CrossMAb IgG ReconnAb binds to FcγRI, CrossMAb IgG was loaded on the octet using FAB2G tips and then associated with human FcγRI at 200 nM. The results demonstrate specific binding at an appropriate affinity.

Extended Data Fig. 9. The CrossMAb IgG ReconnAb neutralizes better than bivalent ACE2 (Fc-ACE2).

(A) Schematic depicting the protein design for Fc-tagged ACE2. Fc fusion was put C-terminal of the ACE2 ectodomain. (B) Neutralization of Fc-ACE2 against the same VOCs tested as in Fig.6 depicts weaker neutralization compared to the IgG ReconnAb. (C) a table of neutralizing potency of the IgG ReconnAb compared to Fc-ACE2 and the fold improvement for the IgG ReconnAb compared to Fc-ACE2.