Fig 1.
Location of amino acid substitutions in the spike protein of the original and SARS-Cov-2 VOIs and VOCs.
Wild-type (WT)- PDB: 6VXX; Alpha- PDB: 7R14; Mu-PDB; Gamma- PDB: 7VX1; Beta-PDB: 7VX1; Delta- PDB: 7W92; Omicron–PDB: 7T9K; other variants, Lambda, Kappa, Epsilon and Mu structure were modeled by using Swiss-model server and WT (PDB: 6VXX) as a template. The structures were shown as surfaces, with RBD colored in gray and mutations highlighted in pink. Group 1 includes the RBDs from the Alpha, Beta, Gamma, and Mu SARS-CoV-2 variants, Group 2 includes the RBDs from Epsilon, Delta, Kappa, and Lambda variants and Group 3 includes the RBDs from Omicron BA.1 and Omicron BA.5.
Fig 2.
Two major clonotypes demonstrate affinity for groups of SARS-CoV-2 variants.
The top sequence logos for CDR1, CDR2, and CDR3 are outputs of screening against the RBD from the Alpha and Beta variants Alpha/Beta (A, B, C). The bottom sequence logos for CDR1, CDR2, and CDR3 are the final sequences obtained after 4 rounds of screening against RBDs from SARS-CoV-2 variants Delta, Epsilon, Lambda, and Kappa (D, E, F). The sequence logo of the CDRs was aligned using the MUSCLE program and visualized using WebLogo, version 3. Bits represent the relative frequency of amino acids.
Fig 3.
Summary of the binding affinity KD (nM) of selected HCAbs against SARS-CoV -2 variant RBDs.
A HCAb is considered non-binding (n.b) for variant RBDs when the respective KD is > 125 nM. The data are the mean of triplicate experiments. To calculate the dissociation constants (KD) for each HCAb/variant combination, HCAb concentration–binding response data at Abs450nm through ELISA were fit to the titration curve equation with a background term using a maximum likelihood estimation non-linear regression through curve-fitting program in Prism 9 (GraphPad Software, Inc.).
Fig 4.
Summary of the neutralizing efficacy EC50 (nM) of selected HCAbs against VSV-SARS-CoV-2-GFP expressing variant Spikes.
A HCAb is considered non-neutralizing (n.n) for a particular VSV pseudotyped virus variant when the respective EC50 is > 625 nM (50 ug/ml). The EC50 values were processed and fitted into nonlinear regression curve using GraphPad Prism 9.0. The data are the mean of triplicate experiments.
Fig 5.
The relationship between predicted and observed HCAb dissociation constants for progressive stages of the iterative model building cycle as new variants are added.
(A) A model trained for all pre-Omicron variants with five-fold cross-validation. (B) Using pre-Omicron variants as a training set and Omicron BA.1 as a test set. (C) Improved model including additional Omicron BA.1 data in training set and cross-validation. (D) Non-Omicron BA.5 used as the training set, and Omicron BA.5 as the test set.
Fig 6.
Characterization of HCAbs identified by ML model against Omicron BA.1 and Omicron BA.5.
A, B: The binding affinity KD (nM) of HCAbs against Omicron BA.1 (A) and Omicron BA.5 RBDs (B). C, D: The neutralizing efficacy EC50 (nM) of HCAbs against Omicron BA.1 (C) and Omicron BA.5 (D) pseudotyped variants. All experiments were performed in triplicate. E: Table summarizing binding affinity and neutralizing efficacy. To calculate the dissociation constants (KD) for each HCAb/variant combination, HCAb concentration–binding response data at Abs450nm through ELISA were fit to the titration curve equation with a background term using a maximum likelihood estimation nonlinear regression through curve-fitting program in Prism 9 (GraphPad Software, Inc.). The EC50 values were processed and fitted into nonlinear regression curve using GraphPad Prism 9.0. All experiments were performed in triplicate.
Fig 7.
HCAbs identified by ML model compete with ACE2-rbFc for binding to RBDs from Omicron BA.1 (A), and Omicron BA.5 (B). Percentage ACE2_rbFc signal was calculated by dividing the maximum response in binding with SARS-CoV-2 RBD of the premixed ACE2_rbFc and HCAbs by the maximum response of ACE2_rbFc in solo binding of the SARS-CoV-2 RBD variant, multiplied by 100. The IC50 values were processed and fitted into nonlinear regression curve using GraphPad Prism 9.0. The experiments were performed in triplicate and the error is the standard deviation from the mean.
Fig 8.
Competition assay was conducted by using Bio-layer interferometry (BLI) to study binding against two epitopes on SARS-CoV-2 RBD variants, specifically Delta (A), Omicron BA.1 (B) and Omicron BA.5 (C). Epitope binning was performed by injecting a first HCAb for 600 s, followed by injection of a second HCAb for 600 s, with a final dissociation step for 600 s.
Fig 9.
Structural overlap of hACE2 with AP1C3: Delta RBD complex (A) and intermolecular interaction between RBD and AP1C3 at distinct epitopes (B). Since these epitopes are among the least conserved regions on the spike protein, a critical point mutation (L452R) results in improved binding of AP1C3 with ultrahigh-affinity, resulting in potent neutralization by AP1C3.
Fig 10.
Structural overlap of hACE2 with BP2A6: Omicron BA.1 RBD complex (A) and intermolecular interaction between RBD and BP2A6 (B). Additional BP2A6: Omicron RBD complex interactions: side chains of Asp102, T103 (BP2A6) form hydrogen bonds with the main chain carbonyl groups of L492 and Y449 and the side chain of T470 (RBD), respectively. The main-chain carbonyl group of Asp102 (BP2A6) also forms a hydrogen bond with Q493R (RBD). Besides these polar interactions, A105, P106, and L107 of BP2A6 and Y483 of RBD form a cluster of hydrophobic interactions, providing ultrahigh-affinity and selectivity for RBD binding.