Mycobacterium tuberculosis is one of the worlds' most successful and sophisticated pathogens. It is estimated that over 2 billion people today harbour latent M. tuberculosis infection without any clinical symptoms. As most new cases of active tuberculosis (TB) arise from this (growing) number of latently infected individuals, urgent measures to control TB reactivation are required, including post-exposure/therapeutic vaccines. The current bacille Calmette-Guérin (BCG) vaccine and all new generation TB vaccines being developed and tested are essentially designed as prophylactic vaccines. Unfortunately, these vaccines are unlikely to be effective in individuals already latently infected with M. tuberculosis. Here, we argue that detailed analysis of M. tuberculosis genes that are switched on predominantly during latent stage infection may lead to the identification of new antigenic targets for anti-TB strategies. We will describe essential host-pathogen interactions in TB with particular emphasis on TB latency and persistent infection. Subsequently, we will focus on novel groups of late-stage specific genes, encoded amongst others by the M. tuberculosis dormancy (dosR) regulon, and summarise recent studies describing human T-cell recognition of these dormancy antigens in relation to (latent) M. tuberculosis infection. We will discuss the possible relevance of these new classes of antigens for vaccine development against TB.