Tool Development for Studying and Manipulating Peptide-MHC Interactions in a Globally-Representative Manner

Abstract

Major histocompatibility complex (MHC) proteins play a critical role in the adaptive immune system, presenting peptide fragments on the surface of cells for surveillance by T cells. In this way, T cells are able to sense cellular dysfunction associated with disease, such as the presence of pathogen-derived peptides. The ability to assess and predict peptide-MHC binding is, therefore, an important component of understanding and engineering immune responses. Peptide-MHC binding is complex, in part due to the immense diversity on both sides of the interaction: MHCs are encoded by the most polymorphic genes in the body and can bind to a subset of trillions of potential peptides. Further, MHC alleles have not been uniformly studied, which presents challenges when designing therapies for diverse patient populations. In this work, we develop tools to study and manipulate peptide-MHC interactions in a more globally-representative manner. First, we study highly polymorphic class II MHC alleles, utilizing data from high-throughput yeast display screens to train algorithms for antigen prediction. Next, we adapt the yeast display platform to screen user-defined libraries of peptides and apply the approach for optimizing peptides and profiling whole viral pathogens for MHC binding. To further increase the MHC throughput of these approaches, we develop a second-generation platform that opens the pipeline for MHC alleles. Finally, we take an orthogonal approach to studying peptide-MHC binding in a representative manner, studying the highly conserved, class Ib MHC HLA-E. We characterize the HLA-E peptide repertoire and train prediction algorithms to identify novel proteome-derived binders. Taken together, these works advance our toolset for studying peptide-MHC interactions across patient populations, with applications in infectious disease, cancer, and autoimmunity.