Combining computational modeling and experimental library screening to affinity-mature VEEV-neutralizing antibody F5
Engineered monoclonal antibodies (mAbs) have proven to be highly effective therapeutics in recent viral outbreaks due to their specificity and ability to provide immediate protection, regardless of immune status. However, despite technical advancements in the field, an ability to rapidly adapt or increase antibody affinity and by extension, therapeutic efficacy, has yet to be fully realized. We endeavored to stand-up such a pipeline using molecular modeling combined with experimental library screening to increase the affinity of a given antibody, F5, to recombinant E1E2 antigen from Venezuelan Equine Encephalitis Virus (VEEV) subtype IAB (TC-83). F5 is a monoclonal antibody with potent neutralizing activity against VEEV that was isolated from human bone marrow donors. F5 is known to bind to spikes on the surface of VEEV made up of a trimer of heterodimers of the glycoproteins E1 and E2. In this work we modeled the interaction of F5 with the E1E2 trimer of VEEV (TC-83) and generated predictions for mutations to improve binding using a Rosetta-based approach and dTERMen, an informatics approach. Modeling the structure of the complex was complicated by the fact that a high-resolution structure of F5 is not available and the H3 loop of F5 exceeds the length for which current modeling approaches can determine a unique structure. To overcome these challenges nine F5 structures with varying H3 loop conformations were generated using RosettaAntibody, PIGS (Prediction of ImmunoGlobulin Structure), and SWISS-Model and these base antibody structures were evaluated in docking trials to recombinant VEEV E1E2 based on relative binding affinity for several subtypes. The structure that gave the best agreement with the experimental trend in relative binding affinity was used for mutation analysis. A subset of the predicted mutations from both methods were incorporated into a phage display library of scFvs (single-chain variable fragments) and screened for binding affinity to the recombinant E1E2 antigen. Results from this screen were used to identify favorable mutations which were then incorporated into twelve human-IgG1 variants. All twelve variants showed increased binding relative to the parental F5 human-IgG1. The best case showed > 60x increased binding to recombinant E1E2 relative to the parental antibody, notably showing a drastic improvement of the Kd or "off rate" compared to the parental F5 IgG. These results demonstrate the ability of our methods to rapidly increase affinity and could be leveraged for increasing Ab binding breadth to additional viral variants.