Mutational Analysis of Tradeoffs between Receptor Affinity and Antibody Escape for SARS-CoV-2 Variants of Concern

Emerging SARS-CoV-2 variants are of broad interest because they may be more resistant to current vaccines and associated immune responses. Toward the long-term goal of understanding how receptor-binding domain (RBD) mutations impact transmissibility, it is critical to elucidate the impacts of RBD mutations on ACE2 affinity and antibody escape. This information is important because RBD mutations can strongly modulate ACE2 affinity, which is linked to changes in viral infectivity, and antibody escape, which is linked to changes in antibody neutralization potency. Moreover, this information is also important because of the inherent tradeoffs between ACE2 affinity and antibody escape, as many RBD mutations that strongly increase one property also strongly decrease the other property, suggesting that evaluating either property in isolation is unlikely to explain how RBD mutations impact SARS-CoV-2 transmissibility. Therefore, the Tessier lab has developed machine learning models to describe the impact of single and multisite RBD mutations on ACE2 affinity and antibody escape. This approach uses large but sparsely sampled experimental datasets that measure the impact of single and multisite RBD mutations on ACE2 affinity and antibody escape to train machine learning models. Next, the models are used to predict the impact of vast numbers of additional RBD mutations that are absent in the experimental datasets. The goal of this proposal is to use machine learning models and multiple experimental techniques to predict and experimentally evaluate the impacts of additional RBD mutations in Variants of Concern, such as the Delta variant, on ACE2 affinity and antibody escape. The hypothesis is that the models will be able to identify additional single and multisite mutations in the RBDs of key Variants of Concern that strongly modulate ACE2 affinity and/or antibody escape. To test this hypothesis, in Aim 1, predictions of the impact of additional single and multisite mutations in the RBDs of Variants of Concern on ACE2 affinity and infectivity will be tested. This Aim will involve testing these predictions using i) yeast surface display of RBDs and flow cytometry to measure ACE2 affinity, and ii) pseudovirus assays to measure infectivity. No live viruses will be generated or tested in this work. Next, in Aim 2, predictions of the impact of additional single and multisite mutations in the RBDs of Variants of Concern on antibody escape and neutralization will be tested. The human serum samples that will be used are from donors that were either infected, vaccinated, or infected and subsequently vaccinated. This Aim will involve testing the model predictions using i) yeast surface display of RBDs and flow cytometry to measure antibody binding, and ii) pseudovirus assays to measure antibody neutralization. A key expected outcome will be the optimization and validation of models that can be used to aid in the rapid identification of the most threatening emerging SARS-CoV-2 variants. This integrated experimental and computational approach holds great potential for use in improving vaccine and therapeutic antibody development to address current and future pandemics.