Convergence of a common solution for broad ebolavirus neutralization by glycan cap-directed human antibodies

Abstract

Antibodies that target the glycan cap epitope on the ebolavirus glycoprotein (GP) are common in the adaptive response of survivors. A subset is known to be broadly neutralizing, but the details of their epitopes and basis for neutralization are not well understood. Here, we present cryoelectron microscopy (cryo-EM) structures of diverse glycan cap antibodies that variably synergize with GP base-binding antibodies. These structures describe a conserved site of vulnerability that anchors the mucin-like domains (MLDs) to the glycan cap, which we call the MLD anchor and cradle. Antibodies that bind to the MLD cradle share common features, including use of IGHV1-69 and IGHJ6 germline genes, which exploit hydrophobic residues and form β-hairpin structures to mimic the MLD anchor, disrupt MLD attachment, destabilize GP quaternary structure, and block cleavage events required for receptor binding. Our results provide a molecular basis for ebolavirus neutralization by broadly reactive glycan cap antibodies.

Keywords: Ebola virus; antibody; antibody therapeutics; broadly neutralizing; ebolaviruses; filoviruses; glycan cap; mAb.

Figure 1.. Glycan cap antibody synergy and GP destabilization

(A) Jurkat cell surface-displayed EBOV GP binding was assessed using fluorescently labeled EBOV-515 or EBOV-520 after prior incubation of cells with individual unlabeled glycan cap antibodies. The blue dotted line represents basal binding of base antibodies without glycan cap antibodies. The orange dotted line represents maximal binding of base antibodies to GP_CL. Data are shown as mean ± SD of technical triplicates. (B) Negative-stain 2D-class averages of GP complexes bound to glycan cap antibodies and EBOV-515, demonstrating examples of intact trimeric complexes (left) and monomeric complexes (right). (C) Correlation analysis of antibody synergy and GP destabilization by glycan cap antibodies. Curve fitting was performed using simple linear regression analysis. The relationship between the two variables was determined using Pierson correlation analysis. r² quantifies goodness of fit, and the p value indicates whether the slope is significantly non-zero. See also Tables S1 and S2.

Figure 2.. Neutralizing and synergistic glycan cap antibodies bind GP across a wide range of orientations

(A) Low-pass-filtered glycan cap Fabs from cryo-EM structures solved in this study as well as elsewhere, bound to EBOV GPΔMuc, are overlaid to compare binding epitope and angle of approach. (B) Surface representations of cryo-EM structures solved in this study with a fitted ribbon model protomer. Shown are side (left) or top (right) views with respect tothe viral membrane. Fab HC is colored in dark tones and LC in light tones. Co-binding antibodies were removed from reconstructions for clarity. (C) Relationship between antibody angle of approach and Fab spacing. An angle of approach of 0° is considered parallel and 90° is considered perpendicular to the viral surface. An angle of approach greater than 90° indicates antibodies that bind inward toward the head domain, whereas less than 0° indicates antibodies that bind upward from the viral membrane. Fab spacing is determined by averaging the distance from the same point on modeled Fab hinge terminal residues in the HC and LC. Antibodies are labeled as in (A). Curve fitting was performed using simple linear regression analysis. The relationship between the two variables was determined using Pierson correlation analysis. r² quantifies goodness of fit, and the p value indicates whether the slope is significantly non-zero. See also Figures S2 and S4.

Figure 3.. Structural details of glycan cap antibody binding to the GP

Single protomers from structural models are shown with close-up views of interacting regions. HCs are rendered in darker colors and LCs in lighter colors, with GP1 colored white. Important residues that coordinate interaction and binding are highlighted. (A) Key residues in the EBOV-293 CDRH2 hydrogen bond along the length of β17 with an additional potential salt bridge between E65_H2 and K276_GP1. (B) EBOV-293 CDRH2 and CDRH3 make additional contacts, including at W275, and the LC forms potential hydrogen bonds between α2 and β17. (C) Similar to EBOV-293, the BDBV-43 CDRH2 loop binds along β17. (D) BDBV-43 CDRH2 makes additional contacts at W275 and also contacts the loop between α2 and β17 via its LC. (E) The EBOV-43 CDRH3 loop displaces the loop between α2 and β17, shifting N268 by ~8 Å (apo-GP in white and BDBV-43 bound GP in gray). (F) EBOV-437 makes contact with GP exclusively with its HC, hydrogen bonding along β17 with its CDRH3 and contacting the head domain in several places. (G) BDBV-289 makes extensive hydrogen bonds with its CDRH3 along β17. (H) BDBV-289 CDRH3 contacts W275 via methionine-aromatic and pi-pi interactions. Additional contacts are made with the head domain of GP via hydrophobic interactions with CDRH2. (I) BDBV-442 makes contact with GP exclusively with its HC. CDRH3 makes hydrogen bonds along β17, with W275 with hydrophobic interactions and along the loop between α2 and β17. (J) EBOV-296 binds to the GP along β17, contacting W275 via methionine-aromatic and pi-pi interactions. The LC makes further contact with the head domain of the GP with several potential salt bridges. (K) The EBOV-296 LC also makes contact with the loop between α2 and β17. See also Figures S3 and S4 and Tables S3 and S4.

Figure 4.. Glycan cap antibody paratopes feature CDR loops with β-hairpin structures that mimic and displace the β18-β18′ region in the glycan cap

(A) Ribbon models of the glycan cap antibody Fv domains with CDR loops highlighted. The HC is in dark gray (right) and the LC is in light gray (left). (B) Structures highlighting the interaction of each of the glycan cap antibodies with the β17 strand, which forms the basis of an extended b sheet in the glycan cap with the β17-β18 loop and β18-β18′ hairpin motif (shown on the left). (C) Crystal structure of the BDBV-289 Fab. Shown on the right is a comparison of the apo- and GP-bound forms of BDBV-289. *, from a previous study; †, shown as an initial homology model. EBOV-548 (PDB: 6UYE) and 13C6 (PDB: 5KEL) are included for comparison. See also Figure S5 and Tables S3 and S5.

Figure 5.. Glycan cap antibodies target a conserved hydrophobic cradle that anchors the MLDs to GP1

(A) Hydrophobicity surface rendering of the apo-EBOV GP protomer (PDB: 5JQ3), with the MLD anchor (β18-β18′) highlighted in red. Using the Kyte and Doolittle scale (Kyte and Doolittle, 1982), hydrophobic residues are colored orange and hydrophilic ones in blue. (B) Upon glycan cap mAb binding, the MLD anchor is displaced, exposing a hydrophobic pocket we call the MLD cradle. The cradle lies within a groove formed by α2 and β17, directly above the 3¹⁰ pocket. Key residues of the cradle are indicated. The MLD cradle is composed of residues from α2 and β17 as well as some additional residues that lie deeper in the core of GP1, including I218, F248, F252, L253, L256, I260, G264, L273, I274, W275, V277, and L244. The cradle is segmented in the middle by W275, which may explain this residue’s pivotal role in binding of many glycan cap antibodies to GP. (C) Interaction of glycan cap mAb HC loops with the MLD cradle (from the rectangle in B). Key hydrophobic residues from antibody paratopes are indicated. (D) Sequence alignment of the MLD anchor and cradle epitope for the five main ebolaviruses (EBOV; GenBank: QQN67572.1; BDBV, GenBank: AYI50382.1;SUDV, GenBank: AGB56678.1; Tai Forest virus [TAFV], GenBank: AWK96625.1; and Reston virus [RESTV], GenBank: QNF60335.1), with topology indicated below. Residues highlighted in orange are key hydrophobic residues that form the cradle; those in green form the base of the β17-β18 loop that interact with the base of the fusion loop, and those in red are key residues from β18 that interact with the cradle in apo-GP. Those marked with an asterisk are common escape mutants for this epitope. (E) Glycan cap antibody footprints highlighted on the structure of apo-GP, colored to reflect conservation, with dark purple indicating complete lack of conservation and white indicating complete conservation. (F) Sequence alignment of the CDRH3 region from each of the glycan cap antibodies analyzed in this study, with darker pink indicating complete conservation andlight pink indicating complete lack of conservation. The beta-turn-beta structure common to these paratopes is indicated above. Key sequences that are similar among these antibodies are boxed in black, with “Y” stretches from IGHJ6 gene use underlined in purple. See also Figure S5.

Figure 6.. Antibody reactivity to the human HeLa S3, 293F, and Jurkat cell lines

Intact or fixed and permeabilized cells were stained with 5 μg/mL of individual mAbs, followed by incubation with the detection phycoerythrin (PE)-conjugated anti-human immunoglobulin G (IgG) antibody and flow cytometry analysis. (A) Representative flow cytometry histograms showing binding of broadly neutralizing mAbs EBOV-442 (green) and EBOV-515 (blue) and the monoreactive mAb 2G4 (gray) to the indicated human cell lines. Binding of EBOV-442 to Jurkat-EBOV GP cells served as a control for antigen-specific binding (magenta). The mAb BDBV223 with known autoreactivity served as a control for autoreactivity (red). Cells stained with the detection PE-conjugated anti-human IgG antibody served as a control for the assay background. Cells were identified based on forward and side scatter analysis. (B) Median fluorescence intensity (MFI) quantifying binding to intact (extracellular staining) or fixed and permeabilized (extracellular and intracellular staining) cells of each antibody tested. The data are shown as a scatterplot of individual values from triplicate measurements for each mAb, with bar indicating the mean. Dotted line indicates the background level from the detection of antibody binding only as described in (A).

Figure 7.. Cleavage inhibition by glycan cap antibodies

(A) The Jurkat-EBOV GP was incubated with various concentrations of antibodies, treated with thermolysin, and then assayed using flow cytometry for exposure of the receptor binding site (RBS), as measured by binding of a fluorescently labeled MR78 antibody that recognizes the RBS. 50% inhibitory concentration (IC₅₀) and 90% inhibitory concentration (IC₉₀) values (left) and dose-dependent inhibition curves (right) are shown. Dotted line indicates percentage of RBS exposure in the presence of the 2D22 control. BDBV-329 was excluded because it does not bind to the EBOV GP, and BDBV-43 was not tested because of poor recombinant expression. Mean ± SD of technical triplicates from one experiment is shown. (B) Proposed model of GP inhibition by glycan cap antibodies (I) Enzyme cleavage of a loop draped over the outside of the GP (magenta) is thought to release the glycan cap and attached MLD. (II) Glycan cap antibodies that bind to the MLD cradle displace the MLD anchor and, thus, the MLDs themselves, potentially shifting their position and sterically blocking access to the cleavage loop by enzymes, especially on a GP-dense viral surface. See also Figure S6.