A bovine antibody possessing an ultralong complementarity-determining region CDRH3 targets a highly conserved epitope in sarbecovirus spike proteins

Abstract

Broadly neutralizing antibodies have huge potential as novel antiviral therapeutics due to their ability to recognize highly conserved epitopes that are seldom mutated in viral variants. A subset of bovine antibodies possess an ultralong complementarity-determining region (CDR)H3 that is highly adept at recognizing such conserved epitopes, but their reactivity against Sarbecovirus Spike proteins has not been explored previously. Here, we use a SARS-naïve library to isolate a broadly reactive bovine CDRH3 that binds the receptor-binding domain of SARS-CoV, SARS-CoV-2, and all SARS-CoV-2 variants. We show further that it neutralizes viruses pseudo-typed with SARS-CoV Spike, but this is not by competition with angiotensin-converting enzyme 2 (ACE2) binding. Instead, using differential hydrogen-deuterium exchange mass spectrometry, we demonstrate that it recognizes the major site of vulnerability of Sarbecoviruses. This glycan-shielded cryptic epitope becomes available only transiently via interdomain movements of the Spike protein such that antibody binding triggers destruction of the prefusion complex. This proof of principle study demonstrates the power of in vitro expressed bovine antibodies with ultralong CDRH3s for the isolation of novel, broadly reactive tools to combat emerging pathogens and to identify key epitopes for vaccine development.

Keywords: antibody; antibody engineering; antiviral agent; epitope mapping; hydrogen-deuterium exchange.

Figure 1

Cell display and binding of ultralong scFvs to SARS-CoV-2 Spike protein.A, structure of an ultralong CDRH3 stalk and knob domain (PDB: 4K3D) color coded to show the parts encoded by the V_H1-7, D_H8-2, and J_H2-4 gene segments. B, construct used to express membrane-bound scFvs with an ultralong V_H under the control of the cytomegalovirus promoter (CMV). The C-terminal Myc tag and platelet-derived growth factor receptor transmembrane domain (PDGFR; for cell surface expression) are shown. C, 293T cells were transfected with round 0 scFv library. Upper: FACS plots of these cells without (left) and with (right) incubation with α-Myc-FITC. Lower: FACS plots showing α-His-PE and α-Myc-FITC staining in the absence (left) or presence (right) of 40 nM Spike. The red box indicates cells expressing scFvs that bind Spike. D, enrichment of Spike-binding scFvs after two rounds of plasmid-based selection. 293T cells were transfected with round 2 plasmid library incubated without (upper) and with (lower) 40 nM Spike. The red box shows cells expressing scFvs that bind Spike. E, the amino acid sequence of the Spike-binding B9-scFv. The regions encoded by V_1-7 (blue), the VD junction (orange), D_8-2 (dark gray), and J_2-4 (green) are shown. Cysteine residues are highlighted in yellow. FACS, fluorescence-activated cell sorting.

Figure 2

Isolation of an ultralong scFv that binds to WT SARS-CoV-2 Spike receptor-binding domain.A, cartoon depicting the SARS-CoV-2 Spike protein subdomains across aa 2 to 1211. NTD: N-terminal domain, RBD: receptor-binding domain, SD1: sub-domain 1, SD2: sub-domain 2, FCS: furin cleavage site, FP: fusion peptide, HR1: heptad repeat 1, CH: central helix, CD: connector domain, HR2: heptad repeat 2. B, FACS plots of 293T cells transfected with plasmids encoding B9-scFv (top panels) or CR3022-scFv (bottom panels) and incubated with or without 2 μM of the 8xHis-tagged SARS-CoV-2 subdomains shown. SSC refers to side scatter. C, relative binding of cell surface–expressed scFvs to Spike subdomains summated from FACS experiments shown in (B). Relative binding was calculated as the fold enrichment in percent positive cells from scFv transfected cells incubated with the indicated subdomain, compared to percent positive, nontransfected cells incubated with the same subdomain. Data are presented as mean ± SD (n = 3). Lower binding to the S1 domain compared to the RBD may be due to occlusion of the epitope in S1. FACS, fluorescence-activated cell sorting.

Figure 3

Purified B9-scFv is broadly reactive and binds strongly to SARS-CoV RBD.A, cartoon depicting the positions of the mutations in the SARS-CoV-2 VOC Spike RBD (aa 319–591; upper) and Omicron VOC RBD (lower). All mutations are on the Wuhan-Hu-1 + D614G background. B, FACS plots showing B9-scFv (5 μM) binding to the indicated mutant SARS-CoV-2 Spike proteins expressed on the 293T cell surface. B9-scFv binding is indicated by the red box. C, FACS plots showing RBD variants (2 μM) binding to nontransfected 293T cells (NTC) and 293T cells transfected with B9-scFv. RBD binding is indicated by the red box. SSC refers to side scatter. D, FACS-based quantification of binding of B9-scFv or CR3022-scFv to MERS-CoV, SARS-CoV-2, or SARS-CoV RBD (Urbani variant). Mean fluorescence intensity (MFI) of PE staining was used to quantify binding (n = 3). Data are presented as mean ± SD. E, FACS-based half-maximal binding estimates of B9-scFv or CR3022-scFv binding to SARS-CoV RBD (Urbani variant). Data points are plotted as a percentage of the MFI obtained at the highest RBD concentration tested in each case (n = 2). Data are presented as mean ± SD. Estimated K_D and R² values are indicated. F, FACS plots showing binding of purified B9-scFv (200 nM) to 293T cells that were nontransfected (upper) or transfected with full-length SARS-CoV Spike (Urbani variant; lower). FACS, fluorescence-activated cell sorting, RBD, receptor-binding domain; VOC, variants of concern.

Figure 4

B9-scFv neutralizes SARS-CoV but does not compete with ACE2.A, neutralization of viruses pseudotyped with SARS-CoV Spike protein (left; n = 9), SARS-CoV-2 (Wu-1/D614G; middle; n = 3), or VSV-G (right; n = 3) by the indicated concentrations CR3022-scFv or B9-scFv. Data are presented as mean ± SD. B, estimation of the IC₅₀ of CR3022- (n = 3) or B9-scFv (n = 4) for the neutralization of SARS-CoV pseudotyped lentiviruses. Data are presented as mean ± SD. Estimated IC₅₀ values are shown. C, FACS-based quantification of SARS-CoV RBD binding to hACE2 in the presence of the indicated concentrations of ACE2-Fc, B9-scFv, or CR3022-scFv (n = 4). Data are presented as mean ± SD, p = 0.0004 (unpaired Student’s t test). FACS, fluorescence-activated cell sorting, RBD, receptor-binding domain.

Figure 5

B9-scFv binds to a conserved, cryptic site on the RBD.A, Wood’s plots showing the summed differences in deuterium uptake in SARS-CoV RBD at 2 min of exposure to deuterium, comparing RBD alone to RBD in the presence of B9-scFv. Wood’s plots were generated using Deuteros (46). Peptides colored in blue are protected from exchange in the presence of B9-scFv. Peptides with no significant difference between conditions, determined using a 99% confidence interval (dotted line), are shown in gray (n = 3). B, amino acid sequence alignments of SARS-CoV (Urbani) and SARS-CoV-2 (Wu-1) RBDs (aa 319–591). Orange boxes around the sequence indicate a protected region on the SARS-CoV RBD when incubated with B9-scFv as identified by HDX. A cyan asterisk above the sequence indicates the residues in SARS-CoV-2 that were mutated to their SARS-CoV equivalent in the binding studies shown in (C). C and D, binding of B9-scFv to SARS-CoV-2 (Wu-1) Spike or the SARS-CoV-2 Spike carrying the indicated mutations. Quantification of binding is shown in (C; n = 3); selected FACS plots are shown in (D). For (C), data presented is mean ± SD (n = 3). E, modeling the proposed epitope of B9-scFv onto a crystal structure of SARS-CoV-2 RBD (blue) bound to ACE2 (red) (PDB: 6M0J). HDX protected region 1 (RBD residues 449–467) is in yellow, while residues within 5 Å of this are in green or cyan. Those in cyan indicate those mutations tested in (C and D), arrows indicate the mutations that increase binding. The alignments of the relevant regions of the SARS-CoV and SARS-CoV-2 RBDs are presented beneath the structural models. Mutations that improve binding in (C and D) are highlighted in red and labeled above the sequence. FACS, fluorescence-activated cell sorting, HDX, hydrogen-deuterium exchange; PDB, Protein Data Bank; RBD, receptor-binding domain.

Figure 6

B9-scFv binding to the cryptic epitope interferes with a stabilizing glycan interaction.A, molecular surface map of the full-length SARS-CoV spike (PDB: 5X5B), colored in magenta, was generated using UCSF Chimera. The yellow surface indicates the proposed epitope of B9-scFv; this region is relatively occluded in all contexts. In an RBD-down conformation, this epitope is pressed against the N-terminal domain of the neighboring protomer, while in the RBD-up conformation it is likely to be inaccessible without extensive clashes. B, ribbon diagram showing the epitope for B9-scFv in yellow on Spike protomer A (magenta). This lies adjacent to the NTD from protomer B (blue) where a glycan on N165 potentially occludes conserved epitope. PDB file: 6VXX. C, Estimation of the half-maximal binding of B9-scFv to cell surface SARS-CoV Spike (n = 2). NTD, N-terminal domain; PDB, Protein Data Bank; RBD, receptor-binding domain.