Mycobacterium tuberculosis (Mtb), the causative agent of tuberculosis (TB), remains a major health threat. Each year, 8 million new TB cases occur and 2 million individuals die of TB 1. Moreover, it is estimated that one third of the population is latently infected with Mtb, of which ~10% will develop active disease during lifetime. The development of active TB occurs when the balance between natural immunity and the pathogen changes (e.g. upon waning of protective immune response during adolescence and in HIV patients, 2). In addition, at present ~50 million individuals are probably infected with multi drug-resistance (MDR) strains of Mtb (WHO, 2006), rendering antibiotic treatment difficult.

The current vaccine, introduced over 80 years ago, is the live attenuated bacterium Mycobacterium bovis Bacillus Calmette-Geurin (BCG), designed as a prophylactic vaccine for pre-infection administration. BCG is known to protect young children against severe forms of TB however it does not efficiently and consistently protect adults against the most prevalent form of the disease, namely, pulmonary TB (variable protective efficacy ranges from 0% to 80%), nor does BCG offer protection from establishment of latent TB and subsequent reactivation 3456789.

In principle, current putative vaccination strategies against TB can be divided into two groups: (a) prophylactic vaccines (aimed at disease prevention), based on BCG with or without antigens secreted by replicating bacteria recognized during the early stage of infection (e.g. ESAT-6 or Ag85A/B), or protective mutants (live attenuated BCG substitutes) and (b) as yet undefined post-exposure vaccines (boosting BCG) aimed at elimination/containment of latent TB and prevention of reactivation. Ideally, a prime-boost approach comprising of a prophylactic vaccine with subunit, viral vectored or DNA-based vaccines, based on late-stage antigens induced in the dormant stage (transition from replicating to non-replicating stage, latency antigens) or resuscitation/reactivation stage, should have maximum impact on all stages of Mtb infection 26101112131415.

In the past few years there have been important breakthroughs in the development of improved prophylactic TB vaccines. Novel vaccine candidates, mostly selected as single gene products, include: rBCG vaccines (e.g. rBCG30 – expressing Ag85B or ΔureCHly+rBCG – an urease deficient strain expressing listeriolysin), virus-based recombinant vaccines (e.g. MVA85A – Rv3804c expressed in replication-deficient vaccinia virus), live attenuated Mtb strains and subunit vaccines comprising of dominant secreted antigens which could boost the immune response after priming with BCG (e.g. Mtb72F, Ag85B-ESAT-6 fusion), (reviewed by 145671016).

Post-exposure vaccine development requires identification of gene products participating in adaptation of Mtb to the intracellular habitat as Mtb changes from replication to dormancy or from dormancy to resuscitation. Moreover, in the absence of effective prediction models and animal models assessing protective immunity, evaluation of a large number of individual antigens remains laborious 12. In vitro gene expression studies under conditions which mimic dormancy and/or reactivation constitute the major source of information. Availability of whole genome sequences of diverse mycobacterial strains in conjunction with data from global analyses such as whole-genome DNA microarray and proteomic technologies have been the subject of intensive research in the recent years, both at site of pulmonary TB and in ex vivo macrophages 1117. To identify genes expressed specifically during latency, different in vitro conditions have been suggested to model the harsh environment within macrophages or granulomas. These include oxygen deprivation, nutrient starvation and iron limitation. Together with expression studies in lungs/granulomas, in phagocytized bacteria inside activated macrophages or in murine/guinea pig infection models, these datasets provide better understanding of the transition of bacteria through stages of active multiplication, dormancy and resuscitation. Based on bioinformatic studies and analysis of multiple dormancy-related datasets, novel drug targets against the dormant phase of Mtb infection have been recently identified 18.

Several attempts have been made to modify the immunogenicity or antigenicity of BCG by generating recombinant strains expressing cytokines, pore-forming listeriolysin/perfringolysin, immunodominant antigens or additional antigens missing from the genome of the avirulent M. bovis BCG genome. Over 200 genes located in 14 gene segments assigned to defined regions of difference (RDs) are missing in the vaccine strain. In principle, potency of BCG vaccines could be improved by supplementing with missing immunodominant RD genes; however critical evaluation is required since RD regions are associated with virulence. Indeed, when BCG was supplemented by the classical RD-1 antigen ESAT-6 (with or without Ag85B 19), enhanced protection was observed in mice however the recombinant strain was more virulent as compared to the wild-type. In a more recent study, Kalra & Grover 20 demonstrated the enhanced prophylactic potential of selected combinations of RD antigens supplementing BCG, improving protection in aerosol exposed mice.

Mtb is an intracellular pathogen and as such cell mediated immunity rather than antibody-mediated immunity is essential for the control of bacterial replication and subsequent protection against TB 7. It is thought that a coordinated response of the cellular immune response is fundamental to the protective immunity 212223, including both CD4 and CD8 T-cells and several cytokines such as IFNγ and TNFα. It has been suggested that CD4 T-cells are mainly crucial during the early phases of infection, while the CD8 T-cells play an important role in the chronic phase of the disease 212223, however little is known about antigen-specific human T-cell responses against persisting mycobacteria, particularly in the context of latent infection and protection against disease, with the exception of some products of the DosR regulon 2425. Therefore, rational selection of sequences that may function as T-cell epitopes in vaccine formulations is crucial. Relatively few human CD8 T-cell epitopes have been found by conventional methods. The availability of the genomic sequence 26 provides new horizons in analyzing the potential immunome of the bacilli, using in silico identification of CTL binders. Several prediction screens of Mtb T-cell epitopes were reported to date 25272829303132, nevertheless these analyses were mostly restricted either to few MHC alleles or to a limited number of preselected subset of genes.

In this study, we have combined multiple published datasets from: (a) gene expression and DNA microarray experiments mimicking conditions leading to dormancy or at the dormant state; (b) genome-wide insertional mutagenesis examining gene essentiality under different conditions; (c) genes expressed in macrophages under conditions mimicking persistence; (d) genes expressed in lungs/granulomas; (e) expression under iron limiting conditions; (f) identification of genes products reported to elicit humoral and cellular response or potential vaccine candidates; (g) in silico identification of T-cell epitopes by dedicated epitope-mapping algorithms and databases compiling relevant experimental data. We designed a whole-genome scoring, ranking and prioritization algorithm and used it to analyze the combined dataset. Accordingly, we selected a set of 189 putative vaccine candidate proteins covering all disease phases, from which we further derived a shorter list of 45 top-ranking antigens for vaccine studies.

Genome sequences and annotations of Mycobacterium tuberculosis H37Rv [GenBank: AL123456] were downloaded from the NCBI 33. Sequence similarity searches were conducted using the Blast algorithm, vs. the non-redundant (nr, NCBI) database. Comparisons against secondary databases of domains and motifs were performed as follows: CDD 34, SMART 35, Pfam 36, Interpro 37, Prosite 38. Secretion signals were analyzed by SignalP 3940 for the secretion via the SecA pathway, and TATfind for searching of the Twin-Arginine Translocation (Tat) motif of the TAT pathway 41. Transmembrane helical segments were predicted by Tmpred 42. The database of bacterial lipoproteins, Dolop 43 was searched for putative lipoproteins. Predictions of CTL epitopes were carried out using the NetCTL integrative approach 44, which combines predictions of MHC class I binding, TAP transport efficiency, and proteasomal cleavage. The analysis was conducted for each of 12 HLA supertypes. A threshold of 1.25 was set on the combined prediction score. The results were parsed by in-house perl scripts. Immune epitope information was complemented by querying the IEDB database 45 for documented B- and T-cell epitopes. The following Mtb-related servers and databases were screened for additive relevant data: TB-sgc – The TB Structural Genomics Consortium 46, Tuberculist-the database on Mycobacterium tuberculosis genetics 47, TBDB – an integrated platform for TB drug discovery 48, MTBreg – The Database of Conditionally Regulated Proteins in Mycobacterium tuberculosis 49 and BioHealthBase – the Biodefense and Public Health Database 50. Functional categories provided for the selected ORF products were as implemented in 2651, and their assignment to each of the antigen was derived from the Tuberculist database (see above). Classes/phases of Mtb infection were assigned as well to the selected antigens: the well established classical antigens, DosR-regulated antigens (as reported in 5253), reactivation antigens (antigens listed by Talaat and his coworkers 54, as well as documented putative implications of individual antigens), and the 5 known resuscitation antigens 55; antigens not belonging to any of the above classes, were denoted as "Others". p-values were calculated using the binomical distribution of frequencies.

The strategy for whole genome-based selection of candidate genes to be included in a vaccine platform, consisted of: (1) compilation of documented data originating from global analyses pertaining to criteria relevant to vaccine and to M. tuberculosis (Mtb) pathogenesis (2) selection of a subset of genes for further evaluation, based on the accumulated data, by cross matching the data in the different criteria; (3) bioinformatic analyses aimed at both further characterization of the candidate genes, in terms of annotation, gene context and cellular localization and; immunoinformatic analysis conducted for the prediction of T-cell epitopes; (4) development and application of a ranking scheme, based on qualitative and quantitative measures, as a tool for prioritizing the selected candidates. A schematic presentation of the various steps, as well as the reductive scheme generating a list of best-hit antigens, is given in Figure 1.