Recognition of mycobacterial epitopes by T cells across mammalian species and use of a program that predicts human HLA-DR binding peptides to predict bovine epitopes.

Bioinformatics tools have the potential to accelerate research into the design of vaccines and diagnostic tests by exploiting genome sequences. The aim of this study was to assess whether in silico analysis could be combined with in vitro screening methods to rapidly identify peptides that are immunogenic during Mycobacterium bovis infection of cattle. In the first instance the M. bovis-derived protein ESAT-6 was used as a model antigen to describe peptides containing T-cell epitopes that were frequently recognized across mammalian species, including natural hosts for tuberculosis (humans and cattle) and small-animal models of tuberculosis (mice and guinea pigs). Having demonstrated that some peptides could be recognized by T cells from a number of M. bovis-infected hosts, we tested whether a virtual-matrix-based human prediction program (ProPred) could identify peptides that were recognized by T cells from M. bovis-infected cattle. In this study, 73% of the experimentally defined peptides from 10 M. bovis antigens that were recognized by bovine T cells contained motifs predicted by ProPred. Finally, in validating this observation, we showed that three of five peptides from the mycobacterial antigen Rv3019c that were predicted to contain HLA-DR-restricted epitopes were recognized by T cells from M. bovis-infected cattle. The results obtained in this study support the approach of using bioinformatics to increase the efficiency of epitope screening and selection.