In the last few decades, the hallmarks of cell-mediated protection against Mycobacterium tuberculosis (Mtb), the causative agent of tuberculosis (TB), have been a subject of intense investigation. The production of the T helper cell type 1 cytokine IFNγ is considered key in Mtb immunity, since it is a central factor in activating macrophages to disarm intracellular mycobacteria (1, 2). A wide landscape of Mtb antigens targeted by human T cells is being uncovered, including proteins (3–6), lipoglycans (7–9), and lipoproteins (10–12) that are processed and exposed by antigen-presenting cells in the context of various presentation platforms. These can be either polymorphic classical MHC class I (HLA-A, -B, and -C) or MHC class II (HLA-DR, -DQ, and -DP) molecules (3–6, 10, 12), oligomorphic MHC class Ib molecules (HLA-E) (13–16) or CD1 isoforms (7–9, 11, 17–19). Relevant to the development of immunodiagnostic tests and vaccine candidates, strong human IFNγ responses consistently pointed at a range of immunodominant protein antigens, including members of the so-called PE/PPE and ESX protein families (5, 20–25). Whether these responses are for the greater part beneficial to the host by providing protection against Mtb or might actually help the pathogen to spread after damaging lung tissue is, for most of them, currently unanswered. Hyperconservation of human Mtb T cell peptide epitopes has been described, perhaps arguing for a beneficial effect of recognition by the host for the pathogen (26, 27), yet epitope sequence variability has also been reported (3, 28, 29).

Several genome-wide screens and bioinformatics-guided approaches further added to the identification of novel protein antigens and immunodominant epitopes for a number of antigen presentation platforms (5, 13, 24, 29–33). Altogether, the picture emerging from these studies is consistent with a multi-epitopic, multi-antigenic IFNγ response during Mtb infection. To investigate whether different protein classes have the same or diverse functional characteristics, Lindestam Arlehamn et al. combined genome-wide HLA class II binding predictions with high-throughput cellular screens of peptides to interrogate CD4+ T cell responses from latently infected individuals. A significant clustering was seen of the majority of targeted proteins, representing 42% of the total response to three broadly immunodominant antigenic islands, to only 0.55% of the total open reading frames (ORFs) (5). However, no quantitative, functional, or phenotypical distinction was observed between T cells elicited by the various protein classes involved, such as those assigned to be secreted or others belonging to secretion systems themselves, or to cell wall or cellular processes. Hence, because of equal functionality, no antigen class could be implied in a more protective (or non-protective) profile over others.

Even though greatly informative, preselecting epitope candidates from the full Mtb proteome of approximately 4,000 ORFs based on bioinformatics has limitations. Binding algorithms may not be 100% effective and certain protective Mtb epitopes with weaker binding properties could perhaps rank too low in the assignment to be selected.

Moreover, the assumption that the immunoproteome is merely a direct translation of the coding genome is an oversimplification. As an additional level of proteome complexity, primary protein structures can be modified after translation. Multiple post-translational modifications (PTMs) occur in higher and lower organisms, involving proteolytic events or transfer of modifying groups to one or more amino acids of the proteins. These PTMs may influence the protein’s active state, compartmentalization, turnover, and/or interactions with other proteins. The rich nature of PTMs of prokaryotic proteomes has started to become unraveled only recently (34), essentially through advances in mass spectrometry (MS) (35). However, their presence in the Mtb proteome and their role in virulence and immunity have not received sufficient attention yet. Here, we review PTMs currently known to occur in the Mtb proteome and discuss whether they modify the Mtb immunoproteome indirectly, by engaging eukaryotic innate receptor signaling or antigen-processing pathways, or directly by persisting as structural moieties in the immunogenic epitopes. In addition, we highlight technologies enabling the unbiased detection and identification of the Mtb T cell epitope repertoire, modified or unmodified.

Current advances in MS-based proteomics have revealed that, like in eukaryotes, PTMs can create an enormous diversity and complexity of gene products in prokaryotes, as was reviewed recently elsewhere (34). PTMs are covalent-processing events chemically changing protein structure, often catalyzed by substrate-specific enzymes. Hundreds of types of PTMs are known, some of which can occur in parallel to create even more heterogeneity in the protein arsenal (36, 37). There are several technical obstacles still to overcome in PTM analysis. In proteome measurements, each protein can be identified based on combined mass and fragmentation patterns from various cleaved peptides. In PTM measurements, each modification site is only represented by a single peptide species. Modified peptides can be of low abundance and furthermore may have chemical properties requiring optimization of liquid chromatography (LC) separation techniques or fragmentation modules, used in MS identification. As a solution, robust MS-based proteomic workflows have been designed, including affinity-based enrichment strategies that can assist in the identification of, e.g., the phosphoproteome, the glycoproteome, or the acetylated proteome (35).

Over the last two decades, multiple proteomic studies were performed on Mtb. In one recent study, using dedicated subcellular fractionation combined with affinity enrichment and liquid chromatography mass spectrometry (LC-MS) based proteomics, Bell et al. were able to bona fide identify 1,051 protein groups present in the Mtb H37Rv proteome, including lipoproteins, glycoproteins, and glycolipoproteins (38). While data are accumulating, our insight into Mtb PTMs is still far from complete (see Table 1 for summary and structure examples of PTMs discussed).

The rich variety of PTMs to a large proportion of the Mtb proteome is likely to play a major role in the successful intracellular lifestyle of Mtb during chronic and sometimes lifelong infections. In the quest of novel vaccines, urgently needed to improve the limited protective capacity of BCG, it may be useful to understand the role of these PTMs in the host response to Mtb infection. Over thousands of years, a balance has been reached in which Mtb avoids excessive immunity allowing it to survive in the host, and in which a certain level of immunity allows the host not to succumb to the infection.

While the primary Mtb proteome shows features of hyperconservation, suggesting an evolutionary advantage to ensure stable epitope recognition by CD4+ T cells (26), PTMs superimpose a high level of complexity. This may complicate the identification of protective protein antigens based on in silico analyses and recombinant DNA technologies. Once protective protein antigens have been identified, the exact structural features need to be known for optimization and process development of the antigen. Furthermore, it will be important to know whether a particular PTM acts as an immune modulator, or/and whether it is part of the structural antigen moiety targeted by the adaptive immune system. Illustrative in this respect are three examples of Mtb protein antigens with PTMs, currently considered as vaccine candidates because of their immunodominance in humans and/or protective effect in animal models.

The 45–47 kDa secretory and cell-surface adhesin Apa is a major mycobacterial antigen with different O-mannosylation patterns in pathogenic versus non-pathogenic mycobacterial species that are critical for its T cell antigenicity in vivo and in vitro (70, 91). T cells from BCG-vaccinated PPD-responsive individuals recognize either both native mannosylated Apa (nApa) and recombinant non-mannosylated Apa (rApa), or nApa only. These latter T cells did, in contrast to the former, not recognize synthetic peptides corresponding to the Apa protein sequence. Together with the finding that recognition of nApa required active antigen processing, these data suggest that mannosylation does not induce alternate processing of nApa but rather that the carbohydrate moiety is an intrinsic part of the T cell epitope(s) (92). Protection by Apa was shown in guinea pig and mouse models in the context of various vaccine platforms (protein, DNA, and poxvirus boost) and routes (intanasal and subcutaneous), as a subunit or as a BCG-booster vaccine (70, 92–94). In a mouse model, adjuvanted nApa was found to induce higher frequencies of CD4+ T cells, producing more cytokines, compared to adjuvanted rApa. However, both antigens were equally protective against virulent Mtb infection when used as a subunit vaccine or as a BCG-booster vaccine (92). This indicates that O-mannosylation is not required for the protective effect in this model. However, understanding of the impact of the different immune responses evoked by nApa and rApa, as well as the nature of the putative naturally processed glycopeptide(s), need further investigation.

In contrast to Apa, the natural PTM of HBHA, methylation, is essential for providing high levels of protection against Mtb challenge in mice, in addition to its antigenicity in Mtb-infected human individuals (95, 96). However, immunization of mice with purified non-methylated HBHA induces antibodies and Th1 cytokines at levels similar to those induced by immunization with methylated HBHA. Also, the antibody isotype profiles are similar in both instances. Interestingly, however, only splenocytes isolated from mice immunized with methylated HBHA, and not with non-methylated HBHA, induce IFNγ secretion upon incubation with Mtb-pulsed macrophages. Methylated HBHA-specific T cell responses are likely to participate in protection against disease in humans, since T cells from patients with active TB secrete significantly lower amounts of IFNγ after stimulation with methylated HBHA than subjects with latent Mtb infection (Figure 1B) (23, 97, 98). HBHA is being considered as a BCG-booster vaccine (99), as responses to methylated HBHA were found to be primed in BCG-vaccinated infants (100). It is not yet known whether the PTM affects the presentation of non-modified protective T cell epitopes via modulation of antigen uptake or processing, or whether methylation is part of the protective T cell epitope(s) involved.

The N-terminal-Thr acetylated antigen ESAT-6 is known as an immunological hotspot in humans (6). During natural infection or after subunit vaccination in mice, vigorous Th1 type CD4⁺ T cell responses are directed to the N-terminal immunodominant epitope ESAT-6_1–15, whereas other epitopes are masked (101). These can be revealed by redesigning ESAT-6 analogs in which the dominant epitope is removed, resulting in the engagement of protective CD4⁺ T cell responses that resist infection-driven terminal differentiation (102). To our knowledge, the role of the N-acetylation at Thr2 in generating the ESAT-6 peptide repertoire has not been interrogated, yet in view of ESAT-6’s current status as a vaccine candidate in clinical testing (99, 103), such assessment may be important.

In order to fully characterize these candidate vaccine antigens, it will be important to elucidate the exact roles of the added glyco-, methyl-, or N-acetyl moieties, respectively. Does their presence modulate effective antigen processing, perhaps by steering proteolysis and immunodominance through masking certain enzyme cleavage sites as was shown for O-linked glycans (104), or are they part of the protective immunoproteome itself ? Clearly more studies are needed, including epitope identification approaches to unravel, in these and other targeted vaccine candidates, the role of PTMs in the Mtb immunoproteome. Knowledge on the precise role of the PTM of Mtb vaccine candidates may be of great help to optimize vaccine candidates and potentially to simplify vaccine design and process development.

Protein antigens, modified or not, are translated for T cell surveillance into immunogens in antigen-processing pathways of antigen-presenting cells. This translation consists of enzymatic cleavage and rescue of protein fragments onto the molecules of a relevant antigen-presenting platform, such as classical class I or II MHC molecules (105), non-classical MHC molecules, including class Ib MHC molecules (16), or CD1 isoforms (17). The identification of the exact nature of the naturally processed and presented Mtb immunoproteome would require dedicated technologies such as LC-MS, first pioneered MHC class I ligands by Hunt et al. more than two decades ago (106, 107). Typically, cell lines would be grown at large scale (>1 × 10⁹ cells) and, after detergent solubilization and immunoaffinity purification of MHC-ligand complexes, bound peptide epitopes would be eluted. The purified endogenous MHC class I ligands were characterized by dedicated LC-MS and MS/MS sequencing.

Nowadays, ever evolving LC-MS/MS systems have greatly added to our understanding of the endogenous peptide repertoire and binding motifs of many MHC class I and II molecules (108–111), as well as of class1b MHC molecules (112). For the classical MHC pathways, the notion has emerged that antigen-presenting cells express approximately 100,000 MHC class I and II molecules at their surface, presenting thousands of different endogenous peptides, at widely divergent abundances (113). LC-MS/MS sequencing can unambiguously identify the epitopes as they are eluted from their antigen-presenting molecules in a qualitative and quantitative manner, revealing both primary epitope sequences, as well as any modifications to them (114). LC-MS/MS analyses have shown that processing inside antigen-presenting cells can generate modified or unpredictable MHC epitopes, such as deamidated (115), citullinated (116), or cysteinylated (117) ligands, as well as ligands arising from protein splicing (118–120) or from alternative reading frames or read-throughs of protein-encoding genes (121–123).

Pathogen-encoded immunoproteomes, including PTMs, generated from the proteome inside infected or antigen endocytosing antigen-presenting cells, should be detectable through LC-MS/MS sequencing approaches as well, although pathogen-derived ligands will be needles in the haystack of eluted self epitopes. To facilitate the identification of these non-self pathogen-derived antigens, targeted LC-MS/MS approaches have been developed (124–127). Foreign epitopes that originate from proteins synthesized during infection inside antigen-presenting cells, such as viral MHC class I epitopes during infection, can be traced using algorithms detecting isotopic patterns in the mass chromatograms of MHC immunoproteomes from carefully mixed infected and non-infected cell cultures that were metabolically labeled during growth (128). Alternatively, epitopes that arise from exogenous proteins endocytosed by antigen-presenting cells during infection, such as bacterial MHC class II epitopes, can be traced back in the MHC-bound peptide repertoire after metabolic labeling of antigen during the prokaryotic cell growth (126, 129). However, if PTMs are suspected in the foreign MHC immunoproteome, chromatography, ion fragmentation strategy, and even affinity enrichment strategies will have to be considered accordingly. Until now, only a single study has reported the identification of several Mtb epitopes presented by MHC class I via LC-MS (130). More approaches are underway to extend our knowledge on the naturally processed and MHC-presented Mtb epitopes, including those derived from methylated HBHA, using dedicated LC-MS. These studies include large-scale human monocyte or dendritic cell cultures and either in vitro Mtb infection or targeted antigen pulsing. Inhibition of MHC class II presentation upon incubation with live Mtb, mycobacterial lysates, or purified antigens may frustrate these attempts (131, 132). Dedicated isolation and analytical discovery procedures should then help to identify the Mtb epitope “needles” in the self “haystack,” and increase our knowledge on the role of PTM in the Mtb immunoproteome.

Fast developments in LC-MS/MS-based proteomics have enabled the detection of many types of PTMs in proteomes of prokaryotes, including Mtb. Elucidating the role of PTMs in the immunoproteome of protective Mtb protein antigens is important for the molecular optimization of vaccine candidates, and will also greatly benefit from technical advancements in LC-MS/MS.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.