Abstract
AbstractIdentifying T cell epitopes is essential for studying and potentially tuning immune responses to pathogens. The polymorphic nature of major histocompatibility complex of class II (MHCII)-genes, and the complexity of the antigen processing mechanisms hinders the effective prediction of immunodominant patterns in humans, specially at the population level. Here, we combined the output of a reconstituted antigen processing system and ofin silicoprediction tools for SARS-CoV-2 antigens considering a broad-population coverage DRB1* panel to gain insights on immunodominance patterns. The two methods complement each other, and the resulting model improves upon single positive predictive values (PPV) from each of them to explain known epitopes. This model was used to design a minimalistic peptide pool (59 peptides) matching the performance reported for large overlapping peptide pools (> 500 peptides). Furthermore, almost 70 % of the candidates (23 peptides) selected for a frequent HLA background (DRB1*03:01/*07:01) feature immunodominant responsesex vivo, validating our platform for accessing T cell epitopes at the population level. The analysis of the impact of processing constraints reveals distinct impact of proteolysis and solvent accessible surface area on epitope selection depending on the antigen. Thus, considering these properties for antigens in question should improve available epitope prediction tools.
Publisher
Cold Spring Harbor Laboratory