Diagnosis of tuberculosis often relies on the ex vivo IFN-γ release assays QuantiFERON-TB Gold In-Tube and T-SPOT.TB. However, understanding of the immunological mechanisms underlying their diagnostic use is still incomplete. Accordingly, we investigated T cell responses for the TB Ags included in the these assays and other commonly studied Ags: early secreted antigenic target 6 kDa, culture filtrate protein 10 kDa, Rv2031c, Rv2654c, and Rv1038c. PBMC from latently infected individuals were tested in ex vivo ELISPOT assays with overlapping peptides spanning the entirety of these Ags. We found striking variations in prevalence and magnitude of ex vivo reactivity, with culture filtrate protein 10 kDa being most dominant, followed by early secreted antigenic target 6 kDa and Rv2654c being virtually inactive. Rv2031c and Rv1038c were associated with intermediate patterns of reactivity. Further studies showed that low reactivity was not due to lack of HLA binding peptides, and high reactivity was associated with recognition of a few discrete dominant antigenic regions. Different donors recognized the same core sequence in a given epitope. In some cases, the identified epitopes were restricted by a single specific common HLA molecule (selective restriction), whereas in other cases, promiscuous restriction of the same epitope by multiple HLA molecules was apparent. Definition of the specific restricting HLA allowed to produce tetrameric reagents and showed that epitope-specific T cells recognizing either selectively or promiscuously restricted epitopes were predominantly T effector memory. In conclusion, these results highlight the feasibility of more clearly defined TB diagnostic reagent.