Rational Design of CD8+ T Cell Vaccines for SARS-CoV-2

Abstract

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2, of the sarbecovirus subgenus), the etiologic agent for CoronaVirus Disease 2019 (COVID-19) disease, has infected over 700 million people and caused over seven million deaths in the last four years. Soon after the pandemic’s onset, the urgent need for effective vaccines was clear, with multiple primarily antibody-inducing vaccines available within a year. However, waning antibody titers and variant escape from convalescent and vaccine-induced antibody neutralization underscored the importance of engagement of the cellular arm of the adaptive immune response to mitigate against severe COVID-19 and provide durable cross-protection of emerging variants. To design immune-focusing CD8+ T cell vaccines for SARS-CoV-2, we applied structure-based network analysis and assessments of human leukocyte antigen (HLA) class I-peptide stability to identify highly networked, mutation-constrained epitopes across the viral proteome. These epitopes impair pseudotyped lentivirus infectivity when mutated, elicit CD8+ T cell reactivity in convalescent individuals, and have resisted ongoing viral evolution over the course of the pandemic, highlighting their potential utility in second-generation vaccines. Upon emergence of the highly infectious Omicron variant, numerous studies demonstrated escape from naturally occurring and vaccine-induced neutralizing antibodies and preservation of T cell reactivity in convalescent and vaccinated individuals. However, we observed that a subset of individuals had >50% reduction in effector and memory CD8+ T cell responses to the Omicron spike protein. The ability of SARS-CoV-2 variants to escape T cell responses emphasized the need for a mutation-constrained T cell epitope vaccine that includes epitopes from proteins beyond the highly mutable spike protein. We therefore designed an immunogen comprised of highly networked, HLA-stabilizing epitopes from SARS-CoV-2 structural, accessory, and RNA-dependent RNA polymerase proteins as well as immunogenic epitopes across the viral proteome that enhance HLA coverage of the vaccine cassette. Prime-boost intramuscular vaccinations of our immunogen delivered via lipid nanoparticle-encapsulated mRNA to K18-hACE2 and B cell-deficient mice were immunogenic, with induction of robust proliferative and reactive CD8+ T cell responses. Co-immunization of our T cell cassette with a covalently linked spike N-terminal domain (NTD) and receptor binding domain (RBD) immunogen revealed augmentation of CD8+ and CD4+ T cell responses to non-spike epitopes within the T cell cassette and NTD-RBD epitopes within the domain-based NTD-RBD immunogen. This demonstrates the potential synergism between the T cell- and antibody-inducing mRNA vaccines, providing a novel basis for generating enhanced vaccine-induced immunity against SARS-CoV-2. Our experience with SARS-CoV-2 revealed the difficulty of rapidly identifying putative T cell epitopes and the potential for computational tools to accelerate T cell vaccine design. This led us to leverage deep learning approaches to develop MUNIS, a novel HLA class I epitope prediction tool. We show that MUNIS outperforms current state-of-the-art prediction algorithms in identifying peptides presented by HLA class I alleles and predicting immunodominance hierarchies for viral pathogens. In addition, application of MUNIS to protein sequences from the Epstein-Barr virus and subsequent experimental validation led to the identification of new CD8+ T cell epitopes, illustrating the capacity for deep learning models to rapidly identify immunogenic CD8+ T cell epitopes in never-before-seen viral proteomes. In summary, this thesis demonstrates the rational design of a CD8+ T cell vaccine for SARS-CoV-2 to mitigate against the emergence of T cell escape in more recent SARS-CoV-2 variants, the potential synergy afforded by dual immunization with T cell- and antibody-inducing vaccines to provide broad protection against SARS-CoV-2 variants, and the potential for deep learning approaches to rapidly accelerate the design of T cell vaccines for future pandemic preparedness.