Mapping and definition of HLA class I and II serologic epitopes using an unbiased reverse engineering strategy

Current models describing HLA epitopes are both theoretical and empirical. Each has limitations yielding discordant results and increasingly complex modeling. The models make a priori assumptions that epitopes must be present only on the mature protein, solvent accessible, on the 'top' (peptide binding surface) of the molecule, restricted to the same class as the antibody, and in the same position on the target allele if reactive to more than one locus. Results obtained counter to these assumptions are routinely discounted. For the 17th International Histocompatibility and Immunogenetics Workshop, we developed a reverse engineering algorithm to define epitopes without these assumptions on a cohort of 332 primary transplant pairs. Complete NGS typing of the transcribed (including leader) genomic DNA for 11 HLA loci of donor and recipient and DSA assignment by single antigen beads was performed. Our results show that, when grouped by 16 class I and II allele specific DSA, uniform clusters and 172 specific amino acid target epitopes are recognized by recipients despite originating from disparate HLA pairs. Data also show that these targets can be in the leader, alpha 3, transmembrane and cytoplasmic domains, thus calling into question current assumptions regarding immunogenic epitopes. Comparisons of amino acid epitopes defined by the Terasaki and Duquesnoy groups (TerEp and EpRegistry) are given.