Epitope mapping of SARS-CoV-2 Spike protein using naturally-acquired immune responses to develop monoclonal antibodies

Abstract

COVID-19 vaccination strategies are already available almost worldwide. However, it is also crucial to develop new therapeutic approaches, especially for vulnerable populations that may not fully respond to vaccination, such as the immunocompromised. In this project, we predicted 25 B-cell epitopes in silico in the SARS-CoV-2 Spike (S) protein and screened these against serum and plasma samples from 509 COVID-19 convalescent patients. The aim was to identify those epitopes with the highest IgG reactivity to produce monoclonal antibodies against them for COVID-19 treatment. We implemented Brewpitopes, a computational pipeline based on B-cell epitope prediction tools, such as BepiPred v2.0 and Discotope v2.0, and a series of antibody-epitope accessibility filters. We mapped the SARS-CoV-2 S protein epitopes most likely to be recognised by human neutralizing antibodies. Linear and structural epitope predictions were included and were further refined considering accessibility factors influencing their binding to antibodies like glycosylation status, localization in the viral membrane and accessibility on the 3D-surface of S. Blood samples were collected from 509 COVID-19 patients prospectively recruited 35 days after symptoms initiation, positive RT-qPCR or hospital/ICU discharge. Presence of IgG against SARS-CoV-2 was confirmed by lateral flow immunoassays. Epitopes immunogenicity was tested through the analysis of IgG levels and seropositivity in the convalescent serum and plasma samples and 126 pre-pandemic negative controls by Luminex to identify those with the highest reactivity. The seropositivity cut-offs for each epitope were calculated using a set of 126 pre-pandemic samples as negative controls (NC). Twenty-five SARS-CoV-2 S epitopes were predicted in silico as potentially the most immunogenic. These were synthesized and tested in a multiplex immunoassay against sera/plasmas from convalescent COVID-19 patients (5.7% asymptomatic, 35.6% mild, 13.8% moderate, 23% severe and 22% unknown because of anonymous donation). Among the 25 epitopes tested, 3 exhibited significantly higher IgG reactivity compared to the rest. The proportion of seropositive patients towards these 3 epitopes, based on median fluorescence intensity (MFI or Log₁₀ MFI) above that from NC, ranged between 11 and 48%. Two out of the three most immunogenic epitopes were scaled up, resulting in the generation of two monoclonal antibodies (mAbs). These two mAbs exhibited comparable levels of S protein affinity to commercialized mAbs. Our data shows that the candidate S epitopes predicted in silico are recognised by IgG present in convalescent serum and plasma. This evidence suggests that our computational and experimental pipeline is able to yield immunogenic epitopes against SARS-CoV-2 S. These epitopes are suitable for the development of novel antibodies for preventive or therapeutic approaches against COVID-19.

Keywords: Antibody; Epitope mapping; IgG; Immunoassay; Luminex; Multiplex; Predictive models; Recovered patients of COVID-19; SARS-CoV-2; Serology; Spike protein.

Fig. 1

Study samples flow-chart.

Fig. 2

Accessibility of candidate epitopes in the 3D conformation of the Spike protein. In silico analysis identified (A) 12 epitopes in the RBD domain, (B) 8 epitopes between both RBD and NTD domains, and (C) 5 epitopes in the NTD domain. Accessible epitopes are shown as sphered regions. Each chain of the Spike (S) protein trimeric structure is depicted in red, blue, and green backbones.

Fig. 3

IgG reactivity to the 25 synthetized peptides. IgG reactivity is depicted in violin plots; each plot represents one of the twenty-five peptides. Red lines indicate the seropositivity cut-off value for each peptide (mean + 3SD), and individual outliers are shown as dots.

Fig. 4

IgG reactivity to the 8 selected peptides. Violin plots show IgG reactivity distribution for each of the eight peptides. Red lines indicate the seropositivity cut-off value for each peptide (mean + 3SD), and individual outliers are represented as dots.

Fig. 5

Surface Plasmon Resonance (SPR) binding analysis of monoclonal antibodies to different ligands. (A) E1 mAb11; (B) G12 mAb11; (C) F9 mAb25; (D) H6 mAab25; and (E) anti-RBD mAb (commercial neutralizing antibody targeting the RBD region). Peptides 11 and 25, as well as the S1 protein from SARS-CoV-2 (employed for comparison purposes), were covalently immobilized on the sensor chip surfaces as described in the Methods section. Overlayed sensorgrams (solid lines) and fitted curves (dotted lines) show the association and dissociation kinetics at different antibody concentrations. Association and dissociation rate constants, along with equilibrium dissociation constants, were extracted from the fitted equations in Table 2. Model fitting was evaluated using the Akaike Information Criterion (AIC), with values exceeding 99.99% in all cases, confirming the reliability of the kinetic model.