Tuberculosis (TB) remains one of the most death-causing microorganism worldwide (World Health Organization, 2013). The increasing numbers of drug-resistant TB cases evidence that there is an urgent need for effective diagnosis, drugs and vaccines (Mwaba et al., 2011; Abubakar et al., 2013). The control of latent TB, a stage in which a person is infected with Mycobacterium tuberculosis (Mtb) but does not currently have active disease, plays an important role for disease control, since dormant bacilli are an enormous reservoir of potential TB cases (Rustad et al., 2009).

M.tb can live in a latent stage without causing any clinical symptom and has a potential of reactivation during all the infected individual lifetime. In fact, about one third of the world population is considered to be latently infected (Corbett et al., 2003). The diagnosis of latent tuberculosis infection (LTBI) through the classic tuberculin skin test (TST) has a lack of specificity, and its sensitivity is low in high-risk groups of progression to active TB. The new interferon (IFN)-γ release assays (IGRAs) are immunodiagnostic methods based on the in vitro quantification of the cellular immune response. The detection of IFN-γ released by sensitized T cells stimulated with specific M.tb antigens enables the identification of infected individuals. The main antigens used in IGRAs, the 6-kDa M.tb early-secreted antigenic target (ESAT)-6 protein, 10-kDa culture filtrate protein (CFP-10), coded in the region of difference (RD) 1, and TB7.7, coded in RD11, are present in M.tb but not in any Mycobacterium bovis bacillus Calmette-Guérin (BCG) vaccine strain nor in the majority of non-tuberculous mycobacteria (Andersen et al., 2000). Although their specificity is better than in TST, IGRAs do not discriminate between active disease and LTBI (Latorre et al., 2009) and do not clearly distinguish between a recently acquired infection and remote LTBI (Esmail et al., 2012; Pollock et al., 2013). Moreover, their sensitivity barely exceed 80%, and the response level against the antigens used does not seem to indicate high risk of progression to active TB. There is a need of new TB antigens as biomarkers for LTBI immunodiagnosis.

During LTBI, M.tb is contained within granulomas, which are formed by activated macrophages and other host components that isolate the infected cells in an organized structure and create an environment that suppresses M.tb replication (Esmail et al., 2012). Bacilli must adapt to a variety of environment stresses including reduced oxygen tension, iron limitation, nutrient deprivation, low pH and production of host factors such as nitric oxide and carbon monoxide. Some in vitro models demonstrated that M.tb is capable of an extensive repertoire of metabolic realignments to enter a defined non-replicating state. The initial response of M.tb is encoded by the dormancy survival regulon (DosR, also called DevR, Rv3133c), which leads to induction of a set of ~50 genes, many of unknown function. DosR controls the expression of genes that allow the bacteria to use alternative energy sources, especially lipids, and genes encoding factors that are selectively recognized by T cells from humans with LTBI (Ernst, 2012). This initial response is followed by a more extensible and more stable response called Enduring Hypoxic Response (EHR), which is comprised of 230 genes involved in the control of the regulatory factors and enzymatic machines of the long-term bacteriostasis program of non-replicating M.tb (Rustad et al., 2008). The antigens expressed by M.tb vary during the continued pressure mounted by host immune response in the course of the infection (Honer zu Bentrup and Russell, 2001; Demissie et al., 2006). Using in vitro models which mimic the conditions that the tubercle bacillus encounter within the host as infection progresses from latency to active disease, some infection phase-dependent genes have been identified and believed to be candidates for immunodiagnostics or for future vaccines (Mukamolova et al., 1998; Zvi et al., 2008; Ottenhoff and Kaufmann, 2012).

DosR regulon is crucial for rapid resumption of growth by involving resuscitation-promoting factors (Rpf) once M.tb exits the hypoxic non-respiring state. M.tb contains five Rpf -like proteins that are implicated in resuscitation of this microorganism from dormancy to reactivation via a mechanism involving hydrolysis of the peptidoglycan by Rpfs and partnering proteins (Ernst, 2012).

The in vitro models mentioned above supposed to recapitulate relevant environmental stress conditions that M.tb encounters upon host infection, which allow to identify differentially regulated M.tb genes. However, they present some limitations: many of these environmental stress factors may not be well known yet; there may be additive or synergistic effects between multiple stress factors in vivo that may easily be missed when studied in isolation in vitro; and certain key features of host response-induced stress cannot readily be recapitulated in vitro, including granuloma formation and TB necrosis (Commandeur et al., 2013).

For these reasons, different approaches have been developed to analyze the gene expression profiles of intracellular M.tb using infected human or murine macrophages, infected murine tissue or artificial granuloma mouse models (Schnappinger et al., 2003; Karakousis et al., 2004; Talaat et al., 2004; Cappelli et al., 2006). Specifically, several M.tb genes have been found to be differentially expressed in the lungs of mice strains with high susceptibility to TB during in vivo infection, the so-called in vivo-expressed M.tb (IVE-TB) genes. Interestingly, some of these IVE-TB genes had been also described as induced for nutrient deprivation in in vitro models (Commandeur et al., 2013).

Some M.tb infection phase-dependent antigens have already been tested in whole-blood assays or in peripheral blood mononuclear cells and may be differentially recognized in individuals with different TB status, that is, subjects with no risk of M.tb infection, LTBI individuals and active TB patients (Leyten et al., 2006; Lin et al., 2007, 2009; Black et al., 2009; Schuck et al., 2009; Goletti et al., 2010; Commandeur et al., 2013). The response is commonly measured through the release of IFN-γ. Additionally, IFN-γ plays a central role in the protection against M.tb (Cooper et al., 1993; Flynn et al., 1993; Kaufmann, 2001).

However the studies conducted in vivo, mice models of latency, seem to merely recapitulate primary disease and are closer to human HIV-TB co-infection. The infectious forms of TB arise after an adequate immune response, which itself may contribute to tissue destruction and cavitation. Hence, there has been a move away from considering IFN-γ as protective, except in the first encounter with the tubercle bacillus, with a renewed emphasis on polyfunctional T cells to contain TB infection (Ernst, 2012; Kaufmann, 2012). However, the ease of measuring IFN-γ gives it applicability to diagnostic tests.

The study of the immune response to the potential immunogenic M.tb antigens described above will enlarge our knowledge and will get us closer to the validation of a diagnostic LTBI candidate antigen. We hypothetized that these antigens expressed in latency conditions and involved in reactivation of the dormant bacterial will mainly induce IFN-γ response in LTBI infected patients, and not in no LTBI individuals.

This prospective study aims to evaluate the whole blood IFN-γ response to 60 M.tb recombinant antigens, the immunogenicity of some of them has not yet been assessed: 6 DosR regulon-encoded antigens, 12 TB reactivation-associated antigens, 1 Rpf antigen, 1 starvation antigen, 6 other stress response-associated TB antigens and 34 IVE-TB antigens (2 of those were EHR and 3 were EHR/starvation), in order to identify potential candidates for new LTBI diagnostic methods. We enrolled subjects with LTBI, active TB patients and controls not M.tb infected.

From the 578 participants, 60 were not tested for the determination of IFN-γ because of insufficient samples, eight subjects did not fulfill the inclusion criteria and five subjects were excluded from the study because the amount of IFN-γ in the negative control was too high. In the patients PHA induced high responses, which ratifies the validity of our methodology.

In this study we evaluated whether latency antigens induced a response which varied according to the group of individuals. Each category of antigens, that is, DosR regulon-encoded, TB reactivation-associated, Rpf, starvation, IVE-TB antigens and other stress response-associated TB antigens, contained at least one antigen whose statistical analysis was significative.

Regarding the control antigens we studied, in general, the response they induced was higher after 7 days of incubation. PPD and ESAT-6/CFP-10 were the antigens which induced a highest response, followed by the TB10.4 and finally Ag85A, which is in concordance with what Kassa et al. (2012) found. According to our finding, Chegou et al. (2012) observed that the majority of response to M.tb infection is largely driven by ESAT6/CFP-10, not by the other antigens that they used as controls (TB7.7, Ag85A/B, and HSP65), where the recognition was poor. Generally speaking, it was observed that infected individuals provided higher responses than those with the disease. In a study performed by Sutherland et al. (2013) it was found, instead, that PPD and ESAT-6/CFP-10 generated dominant responses but very few differences between active TB and LTBI subjects. In the present study, some antigens induced a lower response in active TB patients when compared with non-TB infected subjects. This fact could be explained by the criteria selection followed for including individuals in the non-infected group. In our study, patients from LTBI screening with negative QFN, but TST results under 10 mm (non BCG-vaccinated) or 15 mm (BCG-vaccinated) were considered non-infected, but in some cases some cross-reactivity with the antigens used as a control (including ESAT-6/CFP-10) and the latency antigens could not be rejected. Indeed, in general, in the group of non-infected individuals the IFN-γ responses against ESAT-6/CFP-10 are lower in patients with TST under 5 mm, than in patients with TST over 5 mm (data not shown). On the other hand, results in the literature regarding IFN-γ responses to the these antigens in active TB patients are inconsistent. Possible differences may reside in variations in host genetic makeup, M.tb strains, study methodologies or the extent of TB progression, with diminished IFN-γ production during advanced disease (Weir et al., 2003; Jabado and Gros, 2005; Tsenova et al., 2007; Winek et al., 2008; Day et al., 2011).

Among all M.tb recombinant antigens we studied, there is one which stands up significantly: Rv1733, which is a probable conserved transmembrane protein and is part of the DosR regulon. As it can be observed in Figure 1, this antigen induces a differentiated response between non-infected subjects, infected individuals and TB patients. Specifically, infected individuals are the ones that generate a highest response, followed by the non-infected, and by the patients with the active disease at the end. Rv1733 immunogenity has been previously analyzed, and a genomic study from Zvi et al. (2008) describes it as an immunodominant T cell antigen. Moreover, many authors agree that there is significantly higher T cell response in LTBI as compared to TB patients (Vordemeier et al., 1991; Leyten et al., 2006; Black et al., 2009; Schuck et al., 2009; Commandeur et al., 2011; Sutherland et al., 2013), and Rv1733 is one of the DosR antigens that induces a highest response. Interestingly enough, this response pattern against Rv1733 is produced after the stimulation during 24 h, which makes it very appealing to be used for diagnostic purposes. In contrast with our finding, Riaño et al. (2012) observed that LTBI and TB patients did not react to Rv1733. In another study with TB patients, a high response to Rv1733 was also obtained (Kassa et al., 2012).

Even though Rv1733 turned out to be a very immunogenic antigen, Rv2389, which belongs to the Rpf family, induced a high response in most of the individuals as well. Kassa et al. (2012) described that the Rv2389 was able to induce a high IFN-γ response in active TB patients. In fact, the immunoresponse to Rpf may play a protective role against bacilli reactivation (Riaño et al., 2012). Rv2389 was able to differentiate between non-infected individuals, individuals with LTBI and active TB patients when incubated with whole blood for 7 days, even though some overlapping is present. As observed by Chegou et al. (2012), active TB patients response was much lower than in non-TB individuals. Commandeur et al. (2011), demonstrated Rv2389 specific T cell response in long-term M.tb nonprogressors to active TB patients. Riaño et al., (2012), observed higher levels of IFN-γ in the supernatant of stimulated cells from LTBI compared to active TB patients. Huang et al. (2013) demonstrated that LTBI infected through household contacts possessed higher IFN-γ production to Rv2389c than did the community exposed individuals. In addition, QFN-positive individuals responded in a higher level to Rv2389 when compared to QFN-negative individuals. The Rv2389 ability of discriminate between these two groups of patients have been confirmed by receiver operating characteristics curve (ROC) analysis (area under curve = 0.877). Altogether, these findings indicate that Rv2389 would be a good biomarker of LTBI.

Concerning the recently identified IVE-TB antigens, the only research group that has studied their immunogenicity in vitro found that some of them induced high levels of IFN-γ (Commandeur et al., 2013). In our study, some antigens showed a certain difference when comparing the three groups of individuals. While the median value was quite low in all groups, a great number of antigens such as Rv0967, Rv1363, Rv1957, Rv2034, Rv3420, or Rv2642 among others were able to induce outstanding IFN-γ responses in some individuals. In concordance with the study performed by Commandeur and coworkers, it was the active TB patients the group who showed a lowest response. They observed that the individuals who generated a highest response were those that did it with ESAT-6/CFP-10 as well. They also observed that controls not exposed to M.tb and individuals with positive TST and negative response to ESAT-6/CFP-10 did not respond to IVE-TB antigens, which shows that there is a specificity linked to M.tb exposure. In our study, although some non-TB infected individuals produced IFN-γ after being stimulated with IVE-TB antigens, individuals with LTBI also responded as well. BCG vaccine was ruled out as the response trigger in subjects without LTBI, as T cells from non-BCG vaccinated individuals generated a response as well. However, due the difficulty of totally rule out the infection in the no LTBI individuals, we cannot reject that, alternatively, those responder non-BCG vaccinated individuals were really M.tb infected; and that the responder BCG-vaccinated individuals were, in fact, responding to the shared BCG antigens.

Interestingly, the response to Rv2380M and Rv2660 was higher in individuals with a negative QFN than subjects with a positive QFN, indicating that both antigens could be possible biomarkers for remote infection. In the opposite way, Rv2435n induced a higher response in subjects with positive QFN. It may be, therefore, a possible biomarker for recent infection.

Activated lymphocytes and effector T cells that produce IFN-γ from M.tb antigens sensitized individuals, persist for a limited time in the circulation once the antigen is cleared (Pathan et al., 2001). It is thought that central memory T cells, but not effector ones, may take several days (rather than hours) to produce effector cytokines (Kaech et al., 2002; Dheda et al., 2007). This is because, the commercial IGRAs are thought to reflect more recent, rather than remote infections. Therefore, contrary to the findings of the TST, in cases of remote infection, the IFN-γ level did not increase during the short period of exposure to the antigen in the ex vivo IFN-γ assay at baseline. For these reasons we chose to stimulate short and long term the blood samples with the different latency-related antigens. Interestingly, the higher IFN-γ responses have been obtained after long-term stimulation instead of short-term stimulation: Rv1733, Rv3862, Rv2662, Rv0244, Rv2389, Rv1909, Rv2435n, Rv0847, Rv0967, Rv1806, and Rv2642.

As far as we know, only Goletti et al. (2010) assessed the comparison between individuals recently and remotely infected to five latency mycobacterial antigens. They found that Rv2628 was able to differentiate recent from remote infection, being the individuals with remote infection the group that showed significantly higher IFN-γ whole blood responses. In a very preliminary results, using well TB status characterized individuals, we have observed that responses to some antigens (Rv2380M, Rv0967, Rv2435n, and Rv2913) could differentiate between recent and remote infection (data not shown).

The fact of not obtaining response to an antigen that other studies identify as immunogenic, can be due to different host immune responses, M.tb strains and variations in the methodology used (Ottenhoff et al., 1998; Caws et al., 2008; Homolka et al., 2010), and also some factors such as ethnicity (host genetics), nutritional status, and microbial environment (Sutherland et al., 2013). The discordance in results between studies could be also attributed to the lack of gold standard for defining LTBI, and the consequent heterogenicity in the study population included in the different studies. The difficulty of establishing a group of LTBI is demonstrated by the criteria followed by the different authors: Leyten et al. (2006) included both patients from contact-tracing studies and also from screening studies; Chegou et al. (2012) included contact-tracing studies individuals, where neither TST nor QFN results were available; Commandeur et al. (2013) included TST positive patients, with exposure to M.tb and or with history of traveling to high TB incidence countries; and Sutherland et al. (2013) included household contacts of TB patients or by random community selection or from HIV care clinics with TST higher than 10 mm in HIV negative, and higher than 5 in HIV positive (independently of the BCG status). According to a recent study (Sutherland et al., 2013), which includes individuals from different sub-Saharan African countries, despite possible differences in the criteria of study subjects, there were variations between sites in regards to antigen reactivity, suggesting that need to be considered.

In order to ensure the validity of the promising antigens, we decided to study if some antigens induced a different response depending on whether the individual had been vaccinated with BCG or not. Among the antigens that distinguished between non-infected individuals, infected individuals and patients with TB disease, three of them were also identified when we analyze the effect of the BCG: Rv2627, Rv1471, and Rv1957. In order to measure the magnitude of the BCG influence, the response of non-vaccinated individuals was assessed. Being p > 0.05 and the charts showing an overlapping of the response between the three groups it seems that BCG has a considerable influence in the results (data not shown).

The effect of the BCG on the immune response against latency M.tb antigens has been studied by other authors. Lin et al. (2007) found that, although the homology between the DosR regulon from the BCG strain and from M.tb was very high, BCG-vaccinated individuals did not present immune response against DosR. Instead, individuals exposed to M.tb did respond to DosR. Thus, it seems that the response to antigens linked to the control of LTBI is only generated when there is an exposure to M.tb, and it does not depend on whether the individual has been immunized by the BCG, probably because BCG fails to establish long-lived latent infections, and therefore it may not express (or under express) these antigens in vivo following vaccination (Honaker et al., 2008). However, this issue warrants further investigation.

The current study presents certain limitations which are worth mentioning. In the first place, it is worth highlighting the difficulty found in the classification of the individuals according to their TB status, specially among BCG-vaccinated individuals, since there is no gold standard assay for LTBI diagnosis. We therefore cannot rule out in some cases a misclassification. Secondly, it seems that some of the studied antigens could present certain lack of specificity; they could be shared in BCG strain and also in other mycobacteria (Lin et al., 2009), since some non-infected and non BCG-vaccinated individuals responded. Anyway, it is not clear whether the cross-reactivity to latency antigens in M.tb naive people contributes to the natural protection developed in 90% of the individuals who are infected but do not progress to active TB (Fine, 1995; Brandt et al., 2002). Thirdly, the sample size we could include was certainly limited for some antigens, including some antigens found as promising. Another limitation of our work lies in the fact that we only evaluated the immunoresponse in terms of IFN-γ production by T cells. Combination of other cytokines with IFN-γ can strengthen the diagnostic potential of M.tb antigen (Goldsack and Kirman, 2007). However, despite these limitations, this work obtained strong conclusions identifying potential antigens as candidates for further validation studies.

In conclusion, after screening the potential antigenicity in subjects across the spectrum of TB, we could identify promising antigens in all groups of antigens studied. Rv1733, which is encoded in DosR regulon, turned out to be very immunogenic and able to discriminate between the three defined TB status, thus considered a candidate biomarker. Rv2389 and Rv2435n, belonging to Rpf family and IVE-TB group of antigens, respectively, also stood out as LTBI biomarkers. Further work needs to be done in order to support our hypothesis and to have a pattern of host responses available so that by testing the response to a set of M.tb antigens we can define the TB status and make a clinical decision.

All authors listed contributed substantially in the conception or design of the work (Tom H. M. Ottenhoff, José Domínguez) or the adquisition of data (Irene Latorre, Kees L. C. M. Franken, Jéssica Díaz, Maria Luiza de Souza-Galvão, Irma Casas, José Maldonado, Cèlia Milà, Jordi Solsona, M. Ángeles Jimenez-Fuentes, Neus Altet, Alícia Lacoma, Juan Ruiz-Manzano, Cristina Prat), analysis (Vicente Ausina, M^adel Mar Serra-Vidal, José Domínguez, Tom H. M. Ottenhoff), or interpretation of data (M^adel Mar Serra-Vidal, José Domínguez, Tom H. M. Ottenhoff); in drafting the work or revising it critically for important intellectual content; in doing final approval of the version to be published; and in agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.