Bioinformatic and Empirical Analysis of Novel Hypoxia-Inducible Targets of the Human Antituberculosis T Cell Response

Abstract

We analyzed whole genome-based transcriptional profiles of Mycobacterium tuberculosis subjected to prolonged hypoxia to guide the discovery of novel potential Ags, by a combined bioinformatic and empirical approach. We analyzed the fold induction of the 100 most highly induced genes at 7 d of hypoxia, as well as transcript abundance, peptide-binding prediction (ProPred) adjusted for population-specific MHC class II allele frequency, and by literature search. Twenty-six candidate genes were selected by this bioinformatic approach and evaluated empirically using IFN-γ and IL-2 ELISPOT using immunodominant Ags (Acr-1, CFP-10, ESAT-6) as references. Twenty-three of twenty-six proteins induced an IFN-γ response in PBMCs of persons with active or latent tuberculosis. Five novel immunodominant proteins-Rv1957, Rv1954c, Rv1955, Rv2022c, and Rv1471-were identified that induced responses similar to CFP-10 and ESAT-6 in both magnitude and frequency. IL-2 responses were of lower magnitude than were those of IFN-γ. Only moderate evidence of infection stage-specific recognition of Ags was observed. Reconciliation of bioinformatic and empirical hierarchies of immunodominance revealed that Ags could be predicted, providing transcriptomic data were combined with peptide-binding prediction adjusted by population-specific MHC class II allele frequency.