Cross-reactive serum and memory B-cell responses to spike protein in SARS-CoV-2 and endemic coronavirus infection

Abstract

Pre-existing immunity to seasonal endemic coronaviruses could have profound consequences for antibody responses to SARS-CoV-2, induced from natural infection or vaccination. A first step to establish whether pre-existing responses can impact SARS-CoV-2 infection is to understand the nature and extent of cross-reactivity in humans to coronaviruses. Here we compare serum antibody and memory B cell responses to coronavirus spike proteins from pre-pandemic and SARS-CoV-2 convalescent donors using binding and functional assays. We show weak evidence of pre-existing SARS-CoV-2 cross-reactive serum antibodies in pre-pandemic donors. However, we find evidence of pre-existing cross-reactive memory B cells that are activated during SARS-CoV-2 infection. Monoclonal antibodies show varying degrees of cross-reactivity with betacoronaviruses, including SARS-CoV-1 and endemic coronaviruses. We identify one cross-reactive neutralizing antibody specific to the S2 subunit of the S protein. Our results suggest that pre-existing immunity to endemic coronaviruses should be considered in evaluating antibody responses to SARS-CoV-2.

Fig. 1. Reactivity of COVID and pre-pandemic human sera with cell surface-expressed human coronaviruses spikes and their soluble S-protein versions.

a Heatmap showing cell-based flow cytometry binding (CELISA) of COVID and pre-pandemic donor sera with 293 T-cell surface-expressed full-length spike proteins from β-(SARS-CoV-2, SARS-CoV-1, MERS-CoV, HCoV-HKU1, HCoV-OC43) and α-(HCoV-NL63 and HCoV-229E) human coronaviruses (HCoVs). Sera were titrated (six dilutions–starting at 1:30 dilution) and the extent of binding to cell surface-expressed HCoVs was recorded by % positive cells, as detected by PE-conjugated anti-human-Fc secondary Ab using flow cytometry. Area-under-the-curve (AUC) was calculated for each binding titration curve and the antibody titer levels are color-coded as indicated in the key. The binding of sera to vector-only plasmid (non-spike) transfected 293 T cells served as a control for non-specific binding. b ELISA binding of COVID and pre-pandemic donor sera to soluble S-proteins from β-(SARS-CoV-2, SARS-CoV-1, MERS-CoV, HCoV-HKU1, HCoV-OC43) and α-(HCoV-NL63 and HCoV-229E) HCoVs. Serum dilutions (eight dilutions–starting at 1:30 dilution) were titrated against the S-proteins and the binding was detected as OD405 absorbance. AUC representing the extent of binding was calculated from binding curves of COVID (left) and pre-pandemic (right) sera with S-proteins and comparisons of antibody binding titers are shown. The binding of sera with each protein is shown as scatter dot plots with a line at median. Binding to BSA served as a control for non-specific binding by the sera. The serum-binding experiments were carried out in duplicate and repeated independently at least once for reproducibility. Statistical comparisons between two groups were performed using a Mann–Whitney two-tailed test, (****p < 0.0001; ns p > 0.05).

Fig. 2. BioLayer interferometry binding of COVID and pre-pandemic serum antibodies to SARS-CoV-2 and endemic HCoV S-proteins.

a Heatmap summarizing the apparent BLI binding off-rates (k_off (1/s)) of the COVID and pre-pandemic human serum antibodies to SARS-CoV-2 S and endemic β-HCoV, HCoV-HKU1 and α-HCoV, HCoV-NL63 S-proteins. Biotinylated HCoV S-proteins (100 nM) were captured on streptavidin biosensors to achieve binding of at least one response unit. The S-protein-immobilized biosensors were immersed in 1:40 serum dilution solution with serum antibodies as the analyte and the association (120 s; 180–300) and dissociation (240 s; 300–540) steps were conducted to detect the kinetics of antibody-protein interaction. k_off (1/s) dissociation rates for each antibody–antigen interaction are shown. b Off-rates for binding of serum antibodies from COVID donors and from pre-pandemic donors to SARS-CoV-2 S and endemic HCoV, HCoV-HKU1, and HCoV-NL63, S-proteins. Significantly lower dissociation off-rates are observed for COVID compared with pre-pandemic sera. Statistical comparisons between the two groups were performed using a Mann–Whitney two-tailed test, (****p < 0.0001).

Fig. 3. SARS-CoV-2 S and endemic HCoV S-protein-specific cross-reactive IgG+ memory B cells from COVID donors and isolation and characterization of mAbs.

a–b. Flow cytometry analysis showing the single B-cell sorting strategy for COVID representative donor CC9 and frequencies of SARS-CoV-2 S and endemic β-HCoV, HCoV-HKU1 and α-HCoV, HCoV-NL63 S-protein-specific memory B cells in eight select COVID donors. The B cells were gated as SSL, CD3−, CD4−, CD8−, CD14−, IgD−CD19+, IgM−, IgG+. The frequencies of HCoV S-protein-specific IgG memory B cells were as follows; SARS-CoV-2 S (up to ~8%—range = ~1.6–8%), HCoV-HKU1 S (up to ~4.3%—range = ~0.2–4.3%), HCoV-NL63 S (up to ~0.6%–range = ~0.04–0.6%) protein single positive and SARS-CoV-2/HCoV-HKU1 S (up to ~2.4%–range = ~0.02–2.4%) and SARS-CoV-2/HCoV-NL63 S-protein (up to ~0.09%–range = ~0–0.09%) double positives. SARS-CoV-2-infected donors showed the presence of SARS-CoV-2/HCoV-HKU1 S-protein cross-reactive IgG memory B cells. Scatter dot plots show frequencies of S protein-specific B cells with a line at mean with SD. All differences between means with p values for each comparison are indicated. **p < 0.01; ***p < 0.001. A Mann–Whitney two-tailed test was used to compare the data groups. c Pie plots showing immunoglobulin heavy chain distribution of mAbs isolated from four COVID donors, CC9, CC10, CC36, and CC40. The majority of the mAbs were encoded by the IgVH3 immunoglobulin gene family. d Plots showing % nucleotide mutations in heavy (VH) and light (VL) chains of isolated mAbs across different individuals. The VH and VL mutations ranged from 0 to 11.6% and 0–4.4%, respectively, and are shown as scatter dot plots with a line at median. e CELISA-binding curves of isolated mAbs from four COVID donors with SARS-CoV-2 and HCoV-HKU1 spikes expressed on 293 T cells. Binding to HCoV spikes is recorded as % positive cells using a flow cytometry method. Five mAbs, three from the CC9 donor, and two from the CC40 donor show cross-reactive binding to SARS-CoV-2 and HCoV-HKU1 spikes. f Neutralization of SARS-CoV-2 by mAbs isolated from COVID donors. Four mAbs, two each from donors, CC36 and CC40, show neutralization of SARS-CoV-2. The neutralization experiments were performed in duplicate and repeated independently 1–2 times for reproducibility.

Fig. 4. Binding and ADE of SARS-CoV-2/HCoV-HKU1 S-protein-specific cross-reactive mAbs.

a Heatmap showing CELISA binding of COVID mAbs to seven HCoV spikes. Binding represented as area-under-the-curve (AUC) is derived from CELISA-binding titrations of mAbs with cell surface-expressed HCoV spikes and the extent of binding is color-coded. Five mAbs show cross-reactive binding across β-HCoV spikes. b BLI of SARS-CoV-2 and HCoV-HKU1 S-protein-specific cross-reactive mAbs. BLI binding of both IgG and Fab versions of three cross-reactive mAbs (CC9.2, CC9.3, and CC40.8) to SARS-CoV-2 and HCoV-HKU1 S-proteins was tested and the binding curves show association (120 s; 180–300) and dissociation rates (240 s; 300–540). BLI binding of antibody-S-protein combinations shows more stable binding (higher binding constants (KDs)) of cross-reactive mAbs HCoV-HKU1 compared to the SARS-CoV-2 S protein. c Antibody-dependent enhancement (ADE) activities of cross-reactive mAbs, CC9.2, CC9.3, and CC40.8 binding to SARS-CoV-2 live virus using FcγRIIa (K562) and FcγRIIb (Daudi)-expressing target cells. A dengue antibody, DEN3, was used as a control. Each data point in the curve is derived from the ADE experiment of mAbs with SARS-CoV-2 virus and shows virus titer obtained from technical replicates (n = 2); data representative of two independent experiments.

Fig. 5. Epitope specificities of SARS-CoV-2 and endemic HCoV S-protein specific cross-reactive mAbs.

a–b Organization of SARS-CoV-2 S protein subunits, domains, and subdomains (a). Epitope mapping of the mAbs binding to domains and subdomains of SARS-CoV-2 S-protein, NTD, RBD, RBD-SD1, and RBD-SD1-2 and heatmap showing BLI responses for each protein. The extent of binding responses is color-coded (b). Five mAbs were specific for RBD, two for NTD and the remaining mAbs displayed binding only to the whole S protein. c–d Negative stain electron microscopy of HCoV-HKU1 S-protein + Fab CC40.8 complex and comparison with MERS-CoV S + Fab G4 complex. c Raw micrograph of HCoV-HKU1 S in complex with Fab CC40.8. The Fab-HCoV-HKU1 S protein complexing was performed twice, and the data is representative of the two experiments. d Select reference-free 2D class averages with Fabs colored in orange for Fab CC40.8 and blue for Fab G4, which in 2D appear to bind a proximal epitope at the base of the trimer. 2D projections for MERS-CoV S-protein in complex with Fab G4 were generated in EMAN2 from PDB 5W9J.