Bacterial strains, growth media and culture conditions

Inoculum was initially grown from frozen stock stored at −70°C. Mtb strain H37Rv was grown for 8 days (up to OD600 = 2.0) in Middlebrook 7H9 liquid medium (HiMedia, India) supplemented with 0.05% Tween 80 and 10% ADC (albumin, glucose, catalase) growth supplement (HiMedia, India). 200 μl of this culture was added to 50 ml unmodified Sauton medium, containing (per litre): KH₂PO₄, 0.5 g; MgSO₄·7H₂O, 1.4 g; L-asparagine, 4 g; glycerol, 60 ml; ferric ammonium citrate, 0.05 g; sodium citrate, 2 g; 1% ZnSO₄·7H₂O, 0.1 ml; H₂O, to l L; pH 7.0 (adjusted with 1 M NaOH) and supplemented with 0.05% Tween 80 and 10% ADC and this culture was grown for 10 days at 37°C with agitation (200 rpm). The culture, grown in this medium, served as an inoculum that was added to modified medium to reach the concentration ~ 10⁵–10⁶ cells per ml (ca. 2 ml / 200 ml). The composition of the modified Sauton medium is following: KH₂PO₄, 0.5 g; MgSO₄·7H₂O, 1.4 g; L-asparagine, 4 g; glycerol, 2 mL; ferric ammonium citrate, 0.05 g; citric acid, 2 g; 1% ZnSO₄·7H₂O, 0.1 mL; pH 6.0–6.2 (adjusted with 1 M NaOH). Moreover this modification includes a supplement: 0.5% BSA (Cohn Analog, Sigma), 0.025% tyloxapol and 5% glucose. Cultures were incubated in 500 mL flasks containing 200 mL of modified Sauton medium at 37°C with shaking at 200 rpm (Innova, New Brunswick) for 30–50 days, and pH values were periodically measured. In log phase, the pH of the culture reached 7.5–8.0 and then decreased in the stationary phase. When the medium in post-stationary phase Mtb cultures reached pH 6.0–6.2 (after 30–45 days of incubation), cultures (50 mL) were transferred to 50 mL plastic tightly capped tubes and kept under static conditions, without agitation, at room temperature for up to 5 months post-inoculation. At the time of transfer, 2-(N-morpholino)-ethane sulphonic acid (MES) was added at a final concentration of 100 mM to dormant cell cultures to prevent fast acidification of the spent medium during long-term storage.

Evaluation of cell viability by CFU and MPN assays

Bacterial samples (age of active mycobacteria—10 days, dormant—5 months) were serially diluted in fresh Sauton medium supplemented with 0.05% tyloxapol; three replicates of each sample from each dilution were spotted on agar plates (1.5%) with Middlebrook 7H9 medium (HiMedia, India), supplemented with 10% (v/v) ADC (HiMedia, India). Plates were incubated for 30 days at 37°C, after which the number of CFU was counted.

The most probable number (MPN) assay is a statistical approach based on diluting a bacterial suspension until reaching the point when no cell is transited into any well. The growth pattern in three tubes is compared to the published statistical tables in order to estimate the total number of bacteria (either viable or dormant) present in the sample [19].

The MPN assay was carried out in 48-well plastic plates filled with 0.9 mL of special medium for the most effective reactivation of dormant Mtb cells. This medium contains 3.25 g of nutrient broth (HiMedia, India) dissolved in 1 L of a mixture of Sauton medium (0.5 g KH₂PO₄; 1.4 g MgSO₄·7H₂O; 4 g L-asparagine; 0.05 g ferric ammonium citrate; 2 g sodium citrate; 0.01% (w/v) ZnSO₄·7H₂O per litre, pH 7.0), Middlebrook 7H9 liquid medium (HiMedia, India) and RPMI (Thermo Fisher Scientific, USA) (1 : 1 : 1) supplemented with 0.5% v/v glycerol, 0.05% v/v Tween 80 and 10% ADC (HiMedia, India) [18]. Appropriate serial dilutions of the cells (100 μL) were added to each well. Plates were sealed and incubated at 37°C statically for 30 days. The MPN values, based on the well pattern originating from the visible bacterial growth at the corresponding dilution point, were calculated [19].

Measurement of L-alanine and L-glycine dehydrogenase activities

For the reductive amination of pyruvate (L-alanine dehydrogenase activity), the photometric measurement of the rate of NADH decrease was carried out. The reaction mixture consisted of 20 mM pyruvate and 0.5 mM NADH in 50 mM NH₄OH/(NH₄)₂SO₄ (pH 7.6). L-glycine dehydrogenase activity was measured similarly, however, except for pyruvate, 20 mM glyoxylate was used. The activity was defined as turnover of 1 μM of NADH/NAD+ in per s, per mg of whole cell extract.

Preparation of samples for proteome analysis

Active and dormant cells were prepared in three biological replicates. Bacteria were harvested by centrifugation at 8000 g for 15 min and washed 10 times with a buffer containing (per litre) 8 g NaCl, 0.2 g KCl and 0.24 g Na₂HPO₄ (pH 7.4). The bacterial pellet was re-suspended in ice-cold 100 mM HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulphonic acid) buffer (pH 8.0) containing complete protease inhibitor cocktail (Sigma, USA) and PMSF (phenylmethanesulphonyl fluoride) then disrupted with zirconium beads on a bead beater homogenizer (MP Biomedicals FastPrep-24) for 1 min, 5 times for active cells and 10 times for dormant cells. The bacterial lysate was centrifuged at 25,000 g for 15 min at 4°C. To maximize protein isolation, SDS (2% w/v) extraction was carried out. The extracts were precipitated using a ReadyPrep 2D Cleanup kit (Bio-Rad, USA) to remove ionic contaminants such as detergents, lipids and phenolic compounds from protein samples.

In-solution digestion

The protein extracts were dissolved in 20 μL of 6 M urea in 0.1 M ammonium bicarbonate. The protein concentration was measured using a BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA). An equal amount of the total proteins was used for further trypsinolysis and profiling.

Proteins were reduced using TCEP (Tris-(2-carboxyethyl)-phosphin) at 25°C for 45 min, following alkylation with iodoacetamide (IAA) in the dark for 30 min, having final concentrations of 5 mM and 55 mM, respectively. Excess IAA was quenched with 1,4-dithiothreitol, then 50 mM ammonium bicarbonate was added to a total volume of 200 μL. Trypsin was added in a protein-to-trypsin ratio of 50 : 1, and tryptic digestion was done at 37°C overnight. The reaction was stopped by adding formic acid (FA) to a final concentration of 5%. The peptide mixtures obtained were pre-fractionated and purified using a C18 Empore™ disk (3M, St. Paul, USA) as described in [20].

LC-MS/MS analyses

Prior to analysis, peptides were dissolved in 20 μL of 3% ACN/0.1% FA. The injection volume was 1 μL with a flow rate of 5 μL/min. For the mobile phases, 0.1% FA in MS-safe water (A) and 0.1% FA in 100% ACN (B) were used.

Analysis of peptides was done on a Synapt G2-Si High Definition Mass Spectrometer employing T-wave ion mobility powered mass spectrometry coupled to an ACQUITY UPLC M-Class System (Waters) using data-independent acquisition.

The UPLC system was equipped with a trapping column (nanoEase MZ Symmetry C18 Trap Column, 180 μm × 20 mm, 5.0 μm particle diameter, 100 Å pore size, Waters) and an analytical column (nanoEase MZ HSS T3 Column, 75 μm × 100 mm, C18, 1.8 μm particle diameter, 100 Å pore size, Waters). The sample was loaded onto the trapping column for 2 min and transferred to the analytical column. The peptides were eluted from the analytical column with a flow rate of 0.4 μL/min with a linear gradient of increasing concentration of mobile phase B from 5% to 35% for 70 min. Within the nanoFlow^TM ESI-source, a capillary voltage of 2.5 kV was used, the sampling cone voltage was 40 V and the source offset was 80 V. The source temperature was set to 80°C. In the trapping, ion mobility and transfer chambers, the wave velocity of the TriWave was set to 1000, 650 and 175 m/s, respectively, and ramping wave heights of 40, 40 and 4.0 V were used, respectively. In the HDMS^E mode, the collision energy was ramped to 22–45 eV. Data acquisition was carried out using MassLynx software (Waters). HDMS^E mass data were acquired in low and high energy modes. Raw data were noise-reduced using the Noise Compression Tool (Waters).

The acquired data were submitted for processing and database searching using Progenesis software against the database prepared in-house containing protein sequences from Mycobacterium tuberculosis strain ATCC 25618/H37Rv downloaded from the UniProt database (version 20190304) supplemented with sequences of common contaminants (Max Planck Institute of Biochemistry, Martinsried, Germany). The low-energy and high-energy threshold values were set to 150 and 50 counts, respectively. The parameters used for the database search were: enzyme specificity: trypsin, allowed missed cleavages: 2, fixed modification: carbamidomethylation (Cys), variable modifications: N-terminal protein acetylation, and oxidation (Met); the false discovery rate was 1%; minimum fragment ion matches per peptide, minimum fragment ion matches per protein and minimum peptides per protein were set to 1, 4 and 2, respectively. Proteins were considered as significant if they were identified in at least two independent biological replicates.

Statistical analysis

Data analysis and visualization were carried out in R. The generated scripts are available from the corresponding author on reasonable request.

Data normalization

Before the analysis, the total protein concentration was quantified and an equal amount of protein extracts was used for further trypsinolysis and profiling–pre-instrumental normalization on total protein concentration was carried out [21]. To make the profiles comparable for the subsequent statistical analysis, post-instrumental pre-treatment methods were carried out: to deal with heteroscedasticity and to make skewed distributions symmetric, log₂ transformation was performed. To force the samples to have the same distributions for the protein intensities, quantile normalization was carried out [22].

Principal component analysis (PCA) and hierarchical clustering

For illustration of intragroup and between-group variances, principal component analysis (PCA) was carried out, applying the preliminarily normalized, log₂-transformed and Pareto-scaled datasets [23–25]. Similarity/dissimilarity between the groups was determined by a method of hierarchical clustering using the method of Ward [26].

Z-score calculations and heatmap visualization

A pre-treated protein matrix, consisting of proteins arranged in rows and groups of samples arranged into columns, was centred and scaled [27]. The centring represents the difference between the value of a protein’s abundancy and the total mean of the protein. The scaling is the result of division of the scaled metabolite by its σ. The resulting Z-scores, being in a similar range for the palette of proteins, make feasible comparison of the variances between the variables (proteins).

Statistical significance analysis

The corresponding p-values of comparison of the two groups of cells (dormant vs active) were calculated from a two-sided t-test. Since the data matrix was initially log₂-transformed, the resulting log₂(FC_dorm/mean) is the difference between the log₂(mean_dorm) and log₂(mean_act).

Subtractive proteomic analysis of mycobacterial proteins prevailing in dormancy with antigenic properties

The basic principles of subtractive analysis developed previously were used in the current work [28,29]. The significantly changed proteins revealing in dormancy a positive fold change (FC) were subjected to subtractive analysis. In the first step, the paralogue proteins were identified using a CD-HIT online platform [30,31]. The proteins with 60% identity were considered to be paralogous and discarded from the further analytical steps. In the next step, the BLASTp procedure was applied to select non-homologues to Homo sapiens proteins in the query set, using a threshold expectation value of 0.001 and similarity below 35%. The resulting list of proteins was further categorized according to their subcellular localization using the CELLO online platform [32].

The shortened list of proteins was further compared with previously published datasets with annotated Mtb proteins with antigenic properties: i) a list of 181 proteins detected exclusively in the sera of TB-positive patients from a study on the dynamic antibody response to the proteome [33]; ii) a list of 62 proteins resulting from serodiagnosis of latent Mtb infection [34]; iii) a list of 22 proteins resulting from the analysis of TB-specific IgG antibody profiles, detectable on two different platforms (Luminex and MBio) [35].

Annotation of the list of enzymes–prodrug activators

KEGG API was used to prescribe EC numbers to the initially prescribed UniProt KO code [36].

Dormant mycobacterial cells subjected to the proteomic analysis and the results of proteomic profiling

Dormant Mtb cells subjected to the further proteomic profiling bore the typical traits of ‘non-culturability’ [13,18], particularly incapability to grow on rich solid medium: the experimentally estimated viability by the CFU method was 1.32 ± 0.48 × 10⁴ cells/mL (± SE) for dormant cells and 1.17 ± 0.44 × 10⁸ cells/mL (± SE) for active cells. Analysis of the 16S RNA sequences of cultures revealed 100% identity to M. tuberculosis. At the same time, estimation of the total number of viable but ‘non-culturable’ cells after resuscitation in fresh liquid medium by the MPN assay resulted in 1.21 ± 0.65 × 10⁹ cells/mL (± SE) for dormant cells and 0.97 ± 0.44 × 10⁹ cells/mL (± SE) for actively growing cells (Fig 1). Before the analysis, the total protein concentration was quantified as 3.56 ± 0.24 mg/g wet weight (± SE) for dormant cells and 6.01 ± 0.14 mg/g wet weight (± SE) for active cells. An equal amount of the total proteins was used for further trypsinolysis and LC-MS/MS profiling [21]. LC-MS profiling resulted in the detection of 1379 proteins (whose encoding genes Rv were further used as identifiers), covering 35% of the coding M. tuberculosis H37Rv genome.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. General scheme of dormant cells production and viability assessment of M. tuberculosis H37Rv cells subjected to further proteomic analysis.

A) The outline of the model of gradual acidification of the bacterial culture in the post-stationary phase used in the current study [13]. See details in the Material and Methods section; B) The results of viability assays CFU and MPN of Active cells (akt) were probed in the late log phase–the 10^th day of cultivation in unmodified Sauton medium. Viability of the dormant cells (dorm) was analysed after 5 months of storage after gradual acidification of the bacterial culture in the post-stationary phase. Estimation based on probing of 6–9 independent samples of each type of cell. Statistically significant decrease (***—p 0.001) in the CFU value at dormant state, while maintaining the unchanged MPN value (ns–non-significant), indicates the transition of bacteria to the state of viability, but “non-culturability”. Hereinafter, unless otherwise stated, the group of dormant cells is colored blue, the group of active cells is colored orange.

https://doi.org/10.1371/journal.pone.0269847.g001

However, the raw instrumental data showed skewed distributions of protein intensity, tending to lower values (Fig 2A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Normalization of proteomic profiles of dormant (5 month of storage) and active M. tuberculosis cells.

A) Distribution of raw protein intensities demonstrates skewed distributions of protein intensity, tending to lower values; B) average median intensity of nine samples (biological and technical replicates) for each physiological condition after log₂ transformation and quantile normalization of the protein profiles results in comparable normally distributed profiles with equal medians, making the protein profiles eligible for further statistical analyses.

https://doi.org/10.1371/journal.pone.0269847.g002

For the subsequent statistical analysis, post-instrumental pre-treatment methods were carried out: to deal with heteroscedasticity and to make skewed distributions symmetric, log₂-transformation was performed. To force the samples to have the same distributions for the protein intensities, quantile normalization was carried out [22], resulting in comparable normally distributed profiles with equal medians, making the protein profiles eligible for further statistical analyses (Fig 2B).

PCA and hierarchical clustering

The results of PCA [25] depict ca. 80% of all variance in the protein profiles. To perform PCA, the data matrix was preliminary Pareto-scaled [24]. The results demonstrate that samples are clearly distinguished by their physiological state (‘dormant’ vs ‘active’) (Fig 3A). Hierarchical cluster analysis, based on calculations of the amount of within-cluster dissimilarity analogously results in two distinct clusters, separating dormant cells from active ones. Ward’s minimum variance method was used for the calculations [26] (Fig 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3.

Results of the principal component analysis (A) and hierarchical clustering (B) of M. tuberculosis cells used for proteomic profiling. PC1 with the largest variance of each protein level separates the samples into two distinct clusters. The elliptical boundary is analogous to the 95% CI for both bivariate distributions. Ward’s method [26] of hierarchical clustering similarly demonstrates segregation of the two types of physiological group.

https://doi.org/10.1371/journal.pone.0269847.g003

Differential analysis

The results of differential analysis are listed in the (S1 Table “Results of differential analysis”) and graphically depicted as a volcano plot (Fig 4A). Generally, the abundancy of 894 proteins (out of 1379) was found to change significantly (p < 0.05) in the transition to the dormant state–among them 468 with a positive FC (‘stored in dormancy proteins’) and 426 with a negative FC (‘decreased in dormancy proteins’). The rest of the proteins (485) lay beyond the scope of differential analysis–there was greater variance in their fluctuation in transition to dormancy and consequently their changes were insignificant. Analysis of the first 30 proteins stored up significantly differently in dormancy, bearing the maximum log₂FC, discloses a cohort of regulatory proteins and virulence factors (and not enzymatic proteins), implying biochemical adaptation to the stress conditions (Fig 4B). According to the main stress factor–pH decrease–we observed upregulation of the transcriptional regulator PhoP (Rv0757, p < 0.05, log₂FC = 1.43) of the two-component PhoP–PhoR system, conferring the fitness of mycobacteria to the lowered pH values [18,37,38]. In addition, PhoPR controls secretion of the antigens Ag85A (Rv3804c, p < 0.05, log₂FC = 1.28) and Ag85B (Rv0129c, p > 0.05, log₂FC = 1.11) as well as ESAT-6 EsxA (Rv3875, p < 0.05, log₂FC = 4.14)–a secreted strong human T-cell antigen protein that plays a number of roles in modulating the host’s immune response to infection [39] (Fig 4C). Tightly associated with the ESX-1 and ESX-5 secretion systems are proteins of the PE/PPE family [40]. Out of 95 KEGG-annotated PE/PPE proteins we detected six: PPE18 (Rv1196, p < 0.05, log₂FC = 2.44), PPE19 (Rv1361c, p < 0.05, log₂FC = 2.11), PPE32 (Rv1808, p < 0.05, log₂FC = 3.15), PPE60 (Rv3478, p < 0.05, log₂FC = 1.05), PE15 (Rv1386, p < 0.05, log₂FC = 3.64) and PE31 (Rv3477, p > 0.05, log₂FC = 0.39). The low discovery rate may be explained by the paucity of trypsin-cleavage sites in these proteins and their structural homology [41]. For the proteins of this class, a variety of functions is assumed: many PE and PPE genes have been shown to be upregulated in starvation and during stress conditions [42]. A chaperone function of PE proteins has been reported, particularly in transporting of MPT64 antigen protein (Rv1980c, p < 0.05, log₂FC = 0.69) [43]. Additionally, the PhoP transcriptional regulator has been shown to downregulate DevR (Rv3133c, p > 0.05, log₂FC = 0.01), a regulator of the DosR ‘dormancy’ regulon [37,44]. Correspondingly, out of ca. 50 postulated genes of the DosR regulon, only 19 proteins were detected in the current study, 14 of which were found significantly stored in the dormant state (Fig 4D). Similarly, we observed a down-shift of a number of other transcriptional regulatory proteins, e.g., Rv0275c, Rv1556, Rv1019, Rv0818, Rv3058c, Rv3249c, etc. Analogously, in the absence of translational processes, the abundancy of translation initiation factors (Rv3462c, Rv2839c) as well as 30s and 50s ribosomal proteins (Rv0717, Rv2875c, Rv0641) was similarly decreased [45].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. General statistical results of comparative proteomic analysis of dormant vs active M. tuberculosis cells.

A) Results of differential comparison of dormant cells vs active cells reveal 894 statistically significantly changed proteins, among them 468 proteins prevailing in dormancy (with a positive FC value) and 426 proteins whose abundancy decreased in the transition to dormancy (with a negative FC value); B) visualization of the 30 proteins most abundant in dormancy with the maximum FC; C) STRING-based interaction network of PhoP and DevR regulators. Upregulation of pH-dependent regulator PhoP suppresses DevR, a regulator of the DosR system, however activating the expression of mycobacterial virulence factors: ESAT and PE/PPE proteins; D) heat map of the detected proteins of DosR regulon (14 upregulated out of 50 postulated for Wayne’s model of dormancy) [46].

https://doi.org/10.1371/journal.pone.0269847.g004

There was no consistency in behaviour of proteins of the ABC trehalose transporter complex LpqY–SugA–SugB–SugC (Rv1235, Rv1236, Rv1237, Rv1238): out of four annotated in the Mtb genome only Rv1235 and Rv1238 were detected in the current study, albeit with a different sign of FC in dormancy, rather indicating a decrease in their abundancy and efficacy in the current stress conditions and confirming the dependency of the dormant cells on trehalose to replenish intracellular pathways–particularly glycolysis and pentose phosphate pathway [47,48]. On the contrary, the components of the phosphate ABC transporter PstS1-3 (Rv0928, Rv0934, Rv0932c) were found concordantly to be stored in dormancy, indicating a possible dependency of dormant cells on extracellular sources of inorganic phosphate. Part of the ABC efflux pump Rv1218c, conferring the cells the resistance to a wide range of drugs, was found to be similarly stored.

For the analysis of biochemical adaptation to the transition to dormancy, the 666 detected enzymes (both with positive and negative log₂FC, based on KEGG orthology annotation) were placed on the metabolic map and the corresponding processes which they catalyse were highlighted: in orange for those enzymes with a positive FC (enzymes stored in dormancy) and in blue for those demonstrating a negative FC (a relative decrease in dormancy). The width of the highlighted lines corresponds to the p-value (the thicker lines are for p < 0.01, the thinnest lines are for insignificant results (p > 0.05) (Fig 5A). A glance at the general metabolic visualization (Fig 5A) gives an impression that, in dormancy, a scattered ‘mosaic’ intactness of enzymes can be seen: the flawless functionality of the many pathways is hindered by enzymes significantly decreased in dormancy, e.g., the functionality of the urea cycle may be broken because of the decreased level of argininosuccinate synthase (Rv1658, p < 0.05, log₂FC = −0.69). To annotate precisely the metabolic pathways differentially preserved in dormancy, pathway enrichment analysis was carried out on the enzymatic proteins with FC > 0 and p < 0.05 [49,50] (Fig 5B). Detailed pathway analysis revealed nine upregulated proteins of TCA; however, isocitrate dehydrogenase (Rv0066c, p < 0.05, log₂FC = −0.41) and subunits of α-ketoglutarate dehydrogenase (Rv2454 and Rv2455) demonstrated negative values of log₂FC, impeding the normal flow of the TCA cycle (Fig 5C). The subunits of membrane-bound fumarate reductase were not technically detected; however, subunits of the succinate dehydrogenase complex (Rv3318 and Rv0247c) were detected, potentially enabling the reverse flow of the processes in the ‘reductive branch’ of the TCA cycle (part of the Arnon–Buchanan cycle) from pyruvate to succinate, providing the cell with reoxidized NAD⁺ [51,52]. To the NADH regeneration may contribute L-alanine dehydrogenase (catalysing the reductive amination of pyruvate to L-alanine) and glycine dehydrogenase (catalysing the reductive amination of glyoxylate to glycine), whose activity was shown to increase in microaerophilic conditions[53,54] and was confirmed experimentally in the currents study: the activity of the former was 27.87±5.58 μM(NADH)·s^-1·mg_extract^-1 (± SE) for active cells and 63.82±1.28 μM(NADH)·s^-1·mg_extract^-1 (± SE) for dormant cells, and the activity of the latter was 15.59±2.83 μM(NADH)·s^-1·mg_extract^-1 (± SE) for active cells and 15.08±1.46 μM(NADH)·s^-1·mg_extract^-1 (± SE) for dormant cells.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Proteomic composition of enzymatic proteins in dormant M. tuberculosis.

A) Pathway visualization, based on KO annotation: the detected enzymes (either stored in the dormant state, or those whose abundancy decreased) were placed on a metabolic map and the corresponding processes highlighted–orange for enzymes stored in dormancy (FC > 0), dark blue for degraded enzymes (FC < 0); line thickness corresponds to statistical significance–the thickest lines depict statistically valuable processes; B) GO enrichment analysis for the pathways assembled by the enzymes stored in dormancy. Gene ratio–the ratio of the annotated involved genes of the concrete pathway to the total annotated genes of these pathways; C) reconstruction of central carbon metabolism pathways of dormant M. tuberculosis cells–Z-scores are mapped on the corresponding Rv coding enzymes; D) reconstruction of ‘de novo’ UMP pathway. UMP accumulation in the dormant state was similarly recently observed in the same model conditions in M. smegmatis cells [55]; E) mycothiol salvage pathway.

https://doi.org/10.1371/journal.pone.0269847.g005

The intactness of enzymes acting in the β-oxidation of fatty acids may provide the cell with a surplus of acetyl-CoA (in the case of oxidation of even-chained fatty acids) and propionyl-CoA (in the case of oxidation of odd-chained fatty acids). Dormant Mtb cells demonstrated the presence of enzymes of the methylcitrate cycle, which may allow the conversion of propionyl-CoA and oxaloacetate to succinate and pyruvate. The latter can be converted to fumarate by the detected succinate dehydrogenase (Rv3318, Rv0247c). If this pathway is functioning, the accumulation of organic acids (contributing to sustainment of the energized level of transmembrane potential in dormancy [56,57]) in the cells and medium should be expected; that, however, was found experimentally for dormant cells of M. smegmatis, obtained in the same model of self-acidification of growth medium [55]. Contributing to the accumulation of these acids and to detoxification of surplus acetyl-CoA is the glyoxylate pathway, whose enzymes isocitrate lyase (Rv0467) and malate synthase (Rv1837c) were found to be accumulated in dormant cells.

The classical flow of the glycolytic pathway is hampered by deficient glyceraldehyde-3-phosphate dehydrogenase (Rv1436, p < 0.05, log₂FC = −0.43) and phosphoglycerate kinase (Rv1437, p < 0.05, log₂FC = −0.61)–what is supported by metabolomic studies [58]; however, there is a bypass, consisting of the formation of glycerone phosphate, which can further spontaneously convert into methylglyoxal, which in further detoxification pathways leads to pyruvate through lactate formation. Similarly, all enzymes of glycerol metabolism were detected–the catabolism of glycerol or glycerol-3-phosphate is conjugated with the toxic carbonyl compounds methylglyoxal and lactaldehyde (the latter was detected experimentally for M. smegmatis [55]). However, functioning aldehyde dehydrogenases (Rv0147, Rv2858c) or glyoxylase (Rv0577) detoxify these intermediates, ultimately to lactate. The further transformation of forming lactate by a quinone-dependent lactate dehydrogenase finally supplies the cells with pyruvate capable of further catabolism in the TCA, glyoxylate and methylcitrate pathways. Mtb used to have annotated two lactate dehydrogenases, LldD2 (Rv1872c, p < 0.05, log₂FC = 1.04) and LldD1 (Rv0694, p < 0.05, log₂FC = 2.67); however, lactate dehydrogenase activity was confirmed for LldD2 only [59]. On the other hand, upregulation of Rv0694 was observed in the current conditions of dormancy in glycerol-glucose medium. Rv0694 is a member of the Rv0691–Rv0694 gene cluster, involved in biosynthesis of a recently discovered peptide-derived alternative electron carrier–mycofactocin [60,61].

The decreased detection of particular enzymes may be due to the preservation of proteolytic peptidases abundant in the dormant proteome: Rv0319 (p < 0.05, log₂FC = 1.1), Rv0457c (p < 0.05, log₂FC = 0.82), Rv3883c (p < 0.05, log₂FC = 2.53) and Rv0983 (p < 0.05, log₂FC = 0.64).

The enzymes of glycerol catabolism that are stored unprocessed in dormancy contribute to lipid metabolism and may serve as a source of reoxidation of reductive equivalents (Fig 5C). The gapped glycolytic pathway fails to serve substrate-level ATP generation (if any): however, accumulated polyphosphate (the transport of inorganic phosphate remains intact) may serve a phosphagen–a reservoir of phosphoryl groups, that can be used to generate ATP, thus we detected a reversible polyphosphate kinase Ppk1 (Rv2984, p < 0.05, log₂FC = 0.57).

Additionally, we were able to reconstruct a de novo unimpaired uridine monophosphate (UMP) pathway, serving the precursor of all pyrimidine nucleotides (Fig 5D).

Similarly to the known involvement of protective redox systems as a response to ROS attacks in macrophages, upregulation of superoxide dismutase was observed in our dormancy conditions SodA (Rv3846, p < 0.05, log₂FC = 1.04). Since mycobacteria lack glutathione, the ubiquitous low-molecular weight thiol in other organisms, Mtb is dependent on mycothiol for antioxidant activity; the enzymes encoding its biosynthesis were found to be stored in the dormant state (Fig 5E).

To the oxidative stress protective system belong F₄₂₀-dependent glucose-6-phosphate dehydrogenase Fgd1 (Rv0407, p > 0.05, log₂FC = 0.71) and F₄₂₀H₂-dependent menaquinone reductase Rv1261c (p > 0.05, log₂FC = 0.52), contributing to the protection of bacteria from the formation of toxic semiquinones [62]. In shaping the response to ROS and RNS, the accumulation of the regulator SigH was observed (Rv3223c, p < 0.05, log₂FC = 0.48).

Annotation of immunogenic proteins accumulated in dormant cells

It was interesting to search for immunogenic proteins among the detected proteins with a positive FC in dormant Mtb, as they are potentially applicable for latent TB diagnosis. In the current study we enriched the possible candidates, based on the subtractive proteomic concept [28]. The essence of this approach presumes the identification of non-paralogue proteins followed by the identification of non-homologues to H. sapiens proteins and the analysis of subcellular localization (Fig 6A). Thus, the initially applied 468 proteins significantly represented in dormancy with a positive FC resulted in 273 non-homologue proteins after the first two filters, further localization analysis of which allowed the identification of 53 membrane proteins, 175 cytoplasmic proteins and 45 extracellular proteins.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Determination of potential antigenic proteins in dormant M. tuberculosis proteome.

A) General workflow for establishment of potential antigenic proteins. The complete list of antigenic proteins is available in the (S1 Table “Antigenic proteins”); bold-annotated proteins are those with immunogenic properties selected by the sera of latently infected patients [34]; B) Venn analysis of the reference antigenic proteins annotated from previously published works [33–35], with the list of proteins resulting from the subtractive proteomic analysis of the current data set.

https://doi.org/10.1371/journal.pone.0269847.g006

The short-listed data set was compared with the reference lists from published works on studies of the sera of human TB patients [33–35]; (S1 Table “Antigenic proteins”).

The comparative analysis reference lists comprised the most reactive proteins (in total 265 antigens) as described in the Material and Methods.

Comparative analysis of the proteins in these reference sets with the selected proteins in the current study with a positive FC in dormancy revealed 40 potentially immunogenic proteins specific for dormancy (Fig 6A). Remarkably, these studies (including our data set) demonstrate only one shared protein, Rv1284 –a carbonic anhydrase (p < 0.05, log₂FC = 0.38), which was previously annotated among the latency-related antigens of Mtb [63] (Fig 6B).

Annotation of proteins accumulated in dormancy–potential prodrug activators

The accumulation of a significant number of enzymatic proteins in the dormant state may imply the potency of ‘non-culturable’ Mtb cells to perform biotransformation reactions.

Analysis of the distribution of enzyme classes in dormancy revealed the preservation of a bulk bundle of oxidoreductases and hydrolases (Fig 7A). The list of 73 oxidoreductases detected in dormancy with an FC > 0 is summarized in the (S1 Table “Enzymatic proteins”). The majority of the annotated enzymes of this class are NAD- and FAD-dependent, bearing relatively low standard reduction potentials (−0.32 and −0.22 V correspondingly). There are, however, two F₄₂₀-dependent enzymes, Rv0407 (F₄₂₀-dependent glucose-6-phosphate dehydrogenase Fgd1) and deazaflavin (F₄₂₀)-dependent nitroreductase (Ddn), to detect. F₄₂₀-dependent enzymes are known to bear a lower reduction potential (−0.32 V) and have been shown to participate in bicyclic nitroimidazole compounds reduction [62,64]. Similarly, the FAD-dependent decaprenylphosphoryl-beta-D-ribose-2’-epimerase (catalysing the conversion of decaprenyl-phosphoryl-D-ribose to decaprenyl-phosphoryl-D-arabinose–the sole arabinose precursor in the pathway of arabinogalactan biosynthesis) is capable of nitroreducing recently developed benzothiazinones (BTZ) [65].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Enzymes of dormancy as potential intracellular prodrugs activators.

A) Distribution of the enzyme classes within the enzymes stored in dormancy compared to whole-genome enzyme classes; B) overlap of the detected 48 hydrolytic enzymes with the annotated group of serine hydrolases and esterases [66].

https://doi.org/10.1371/journal.pone.0269847.g007

The component of the non-proton pumping NADH:quinone oxidoreductase (Rv0392c, p < 0.05, log₂FC = 3.81) accumulated in dormancy is hypothesized to reduce clofazimine, an antibiotic whose mode of action is mediated through ROS formation [67].

Hydrolytic enzymes (esterases and lipases) are considered to play a crucial role in mycobacterial persistence [68,69]. Among the enzymes detected with a positive FC in the dormant state, 48 could be classified as hydrolytic enzymes (S1 Table “Enzymatic proteins”). This list was compared with a list of annotated serine hydrolases and esterases [66] (Fig 7B). In the query data set of enzymes with a positive FC, three enzymes potentially keep their activity in dormancy (Rv2970c, Rv2224c and Rv1683) [66]. The list of enzymes of other EC classes can be found in the (Table “Enzymatic proteins”).