Hydrophobicity is a hallmark of immunogenic MHC class I T cell epitopes (APP2P.111)

Abstract

Abstract The molecular mechanisms underlying MHC class I-peptide immunogenicity are not well defined. Here, we investigated the biochemical properties that are associated with immunogenicity from immunogenic (n=5035) and non-immunogenic (n=4853) MHC-I peptides curated from the Immune Epitope Database (IEDB). Using a bioinformatic comparative analysis, we evaluated three amino acid properties (hydrophobicity, side chain mass, and polarity) for an association with immunogenicity within the pMHC dataset. We found a strong bias towards hydrophobic amino acid usage within immunogenic peptides (Spearman ρ = 0.71, P = 4.24 x 10^-4); this difference in hydrophobicity is due to preferential usage of non-polar hydrophobic amino acids in TCR-contact residues for peptides restricted by HLA-A2, H-2Db, H-2Kb (P-values < 0.01). We used amino acid hydrophobicity to design an artificial neural network (ANN) model of immunogenicity for MHC-I peptides. We trained this model to predict mouse H2Db-restricted immunogenic epitopes for 4 different antigens from LCMV, Adenovirus, and Influenza A virus. Our findings suggest that hydrophobicity of TCR contact residues may be a fundamental biochemical hallmark that favors immunogenicity of MHC/peptide complexes. Incorporating hydrophobicity into epitope prediction algorithms may improve prediction of immunogenic CTL epitopes.