Introduction

Isoniazid (isonicotinic acid hydrazide) is one of the most potent first-line drugs used in the treatment of TB, the others being rifampicin, ethambutol and pyrazinamide. Its clinical activity against MTB was discovered in the early 1950s [1]. INH is bactericidal against MTB in vitro and in vivo including in humans [2]. MTB is most susceptible to INH during its logarithmic phase of growth; the minimum inhibitory concentration (MIC) is 0.2 µg/ml [3]. Interestingly, INH does not have bactericidal activity on non-replicating MTB (one of the reasons why tuberculosis requires prolonged treatment as well as multidrug therapy). INH is a nicotinamide analog, structurally related to ethionamide and pyrazinamide. INH is converted to the active form by MTB catalase-peroxidase (KatG) [4]. The active drug is an isonicotinoyl radical that reacts non-enzymatically with pyridine nucleotide coenzymes, such as NAD, to form an adduct with INH (INH-NAD+) [5, 6]. The INH-NAD+ adduct binds to and inhibits enoyl–acyl carrier protein (ACP) reductase (InhA), resulting in the inhibition of mycolic acid biosynthesis [7,8,9]. Mycolic acids are high molecular weight, α-alkyl, β-hydroxy fatty acids, which are the major outer cell wall components of mycobacteria [10]. The INH-NAD+ adduct formation has been studied using cell-free assays, and the establishment of the molecular interactions and kinetics of InhA adduct generation have added greater insight into the structure–activity relationship of the drug [11, 12]. Subsequently, binding of INH-NAD+ adduct to recombinant InhA isolated from an E. coli strain co-expressing inhA and katG of MTB was also demonstrated [13].

Molecular mechanism of resistance of MTB to INH is attributed to mutations in the katG gene [4, 14]. Mutations reduce the ability of KatG to activate the pro-drug INH, thus leading to drug resistance [15,16,17,18]. The most characterized one is the missense mutation S315T (serine 315 to threonine) [19, 20]. The mutated katG produces a functional catalase-peroxidase with enzymatic activity, but impaired in its ability to form an INH-NAD+ adduct, reducing the INH bactericidal activity [21]. INH resistance also arises due to substitutions (T→A; C→A) in inhA at the 3′ end of the presumed ribosome binding site located in the region upstream [22]. A single point mutation in inhA (Ser 94 Ala) confers a five-fold increase in resistance to INH, as well as inhibition of mycolic acid biosynthesis [23].

Even though INH is effective in killing actively growing bacilli, it possesses little or no activity against MTB under conditions of nutrient starvation [24] or progressive oxygen depletion in vitro [25]. This resistance mechanism is exhibited by MTB inactivated by a non-replicating persistent state (NRP), which does not involve active cell wall synthesis. To date, it was thought that dormant MTB may not convert INH to the bioactive form, which contributes to resistance [26]. In this study, we report that dormant MTB does actively metabolize INH to its active form. Although none of the proposed mechanisms independently can explain a singular mechanism of INH-resistance in dormant MTB, our study essentially rules out the possibility of non-formation of active INH.

Materials and methods

Bacterial strains and growth conditions

Virulent laboratory strain MTB H37Rv was cultured on Löwenstein–Jensen slants and incubated at 37 °C for 4 to 6 weeks. Broth cultures were prepared by inoculation of one loopful of bacterial colony into Middlebrook 7H9 (BD Difco) medium supplemented with 10% albumin–dextrose–catalase (ADC) and incubated on an orbitary shaker at 130 rpm at 37 °C. All steps involving handling of MTB were carried out in a biosafety level three (BSL3) facility. Validated INH-resistant (INH mono-resistant (Ref no: 567) and multidrug-resistant (MDR) (Ref no: 1565)) strains were a kind gift from the District TB Center, Trivandrum and were cultured by the same protocol. Establishment of dormancy using a modified Wayne’s model was performed as described by Gopinath et al. [27].

Drug treatment

To examine the effect of INH on actively growing MTB, a 15-day-old culture was treated with INH (0.2 µg/ml) [28] and was incubated further for 4 days at 130 rpm at 37 °C. Plate count assay was performed on Middlebrook 7H10 agar medium supplemented with 0.5% glycerol, 0.05% Tween 80, and 10% oleate–albumin–dextrose–catalase supplement (Becton Dickinson, Franklin Lakes, NJ, USA). For testing the effect of INH on dormant MTB, INH (0.2 µg/ml) was added to the hypoxic MTB cultures by means of a 25-gauge needle; methylene blue dye did not change color during this process in comparison to the aerobically grown control, indicating that the reduced oxygen tension had not been compromised [26]. Aliquots were withdrawn from each tube on days 1, 2, 3, 7 and 14 after INH treatment and plated onto 7H10 agar plates.

RNA isolation, cDNA synthesis, and quantitative PCR

RNA isolation from dormant/normoxic MTB was carried out using Trizol (Sigma-Aldrich) reagent following the protocol described by Gopinath et al. [27]. Total RNA was resuspended in nuclease-free water and treated with DNase I (Sigma-Aldrich) at 37 °C for 30 min. Complementary DNA (cDNA) was prepared using Reverse Transcriptase Core kit (Eurogentec, Seraing, Belgium) according to the manufacturer’s protocol. qPCR reaction was performed using SYBR Green Supermix kit (Bio-Rad). The cycling conditions were set as follows: an initial denaturation at 94 °C for 5 min; 35 cycles of denaturation at 94 °C for 30 s, primer annealing at 59 °C for 30 s, and extension at 72 °C for 40 s, followed by a melt curve analysis. Relative gene expression was performed using −ΔΔCt method, with sigma factor A (sigA) as the housekeeping gene control.

Electron microscopy

Twenty milliliters of MTB culture from each condition was centrifuged at 5000 × g for 10 min at 25 °C. The pellets were washed thrice in 10 ml ice-cold PBS and centrifuged at 5000 × g for 5 min at 25 °C. Pellets were dissolved in 1 ml Dubos Difco broth and vortexed vigorously to disperse the bacterial clumps. Suspensions with an OD of 0.15 were prepared in fresh Dubos broth. Three hundred microliters from each sample was then loaded onto polylysine-coated coverslips that were prepared according to the manufacturer’s guidelines (Sigma-Aldrich, St. Louis, MO, USA). Bacteria were allowed to settle for 30 min before gently decanting the medium. One milliliter of 2.5% glutaraldehyde in PBS (pH 7.4) was added to the bacterial film on the coverslip and incubated overnight at 4 °C. A series of dehydration steps was performed for 10 min each in 50, 70, 95 and 100% ethanol and the coverslips were dried at room temperature. Samples were gold-sputtered using JEOL-1200 (Peabody, MA, USA) and imaged using a scanning electron microscope (JEOL-JSM-5600 LV).

Total metabolite isolation from the bacteria

MTB in the states of normoxia and dormancy was centrifuged (5000 × g, 10 min, 25 °C). Cell pellets were washed twice in sterile water (10 ml), and 400 mg of the pellet from each condition was then resuspended in 1 ml of sterile PBS. The suspensions were transferred to 2 ml screw cap microcentrifuge tubes containing glass beads (0.5 mm) and incubated on ice for 5 min. The tubes were then subjected to three, one-minute pulses at 4200 rpm with 1 min interval in a Mini Bead beater (BioSpec Products Inc., Bartlesville, UK). The suspensions were centrifuged for 10 min at 13,000 × g (Eppendorf, Hauppauge, NY), and the proteins in the supernatants were precipitated by adding 200 µl of methanol and further incubating at 4 °C for 1 h. The precipitated proteins were removed by centrifuging for 10 min at 4000 × g at 25 °C. Supernatant collected was dried using speed vacuum concentrator and resuspended in 50 µl of sterile water.

LC–MS analysis

Liquid chromatography

Analysis was performed using a Waters ACQUITY UPLCTM system (Waters, Milford, MA, USA) coupled to a quadrupole–time-of-flight (Q–TOF) mass spectrometer (SYNAPT-G2, Waters). Both the systems were operated and controlled by MassLynx4.1 SCN781 software (Waters). Metabolite separation was achieved using reverse-phase liquid chromatography employing a C18 (high-strength silica, 2.1 × 100 mm, 1.8 µm; Waters) column. Briefly, a 7.5 µl aliquot of each sample was injected into the column maintained at 40 °C. For each sample, the run time was 20 min with a flow rate of 400 µl/min. The mobile phase consisted of aqueous (A) and organic (B) solvent components, where A was 0.1% formic acid in ultrapure water and B was 0.1% formic acid in acetonitrile. The gradient was 0 min, 1% B; 2 min, 10% B; 6 min, 30% B; 8 min, 50% B; 12 min, 75% B; 15 min, 99% B; and 20 min, 1% B. Each sample was injected in triplicate with blank injections between each sample.

Mass spectrometry

The SYNAPT® G2 High Definition MS™ System mass spectrometer (Waters) was operated in positive and negative resolution mode with electrospray ionization (ES+) source over a mass range of 50–1500 Da. Each spectrum was acquired for 0.25 s with an interscan delay of 0.024 s. Ion source and desolvation gas (nitrogen) temperatures were kept at 130 °C and 450 °C, respectively. The cone and desolvation gas flow rates were 80 L/h and 600 L/h, respectively. Capillary voltage was set at 3.0 kV, and sampling cone voltage was set at 30 kV. Lock mass acquisition was performed every 30 s by leucine–enkephalin (556.2771 [M+H]+) for accurate on-line mass calibration. Data acquisition was carried out in centroid mode and processed using MassLynx software.

Data processing

MassLynx4.1 SCN781 (Waters) was used for data acquisition and collection. Markerlynx XS software (Waters) was employed for peak/feature picking and raw data deconvolution, noise filtering, peak detection, isotope peak removal, alignment of retention time and mass, and optional peak/feature normalization. Method parameters for Markerlynx processing were as follows: peak width at 5% height - 1 S; marker intensity threshold-10 counts; mass window - 0.05 Da, and retention time window - 0.2 min. Chemical formula substitution and naming of compounds were performed using MassLynx4.1 SCN781 within 5 ppm range and further verified using ChemSpider chemical structure database. The masses were further validated using MassTRIX available at http://masstrix3.helmholtz-muenchen.de/masstrix3/start.html.

Results

Dormant MTB is resistant to INH

To test INH resistance in dormant MTB, we performed a series of plate count assays after treating dormant MTB and actively growing bacilli (control) with lethal concentration of INH (0.2 µg/ml). Upon INH treatment, actively growing MTB showed significant reduction in bacterial count with 100% killing 16 days after drug treatment (Fig. 1a). Viable count of dormant MTB was not affected by treatment with INH even after 35 days and the cell numbers remained relatively constant throughout the treatment time points, indicating that bacteria are indeed dormant and resistant to INH (Fig. 1b). Effect of INH on the bacterial morphology was analyzed using scanning electron microscopy. Log-phase MTB treated with INH showed characteristic morphological changes (Fig. 1c, d). Interestingly, no such morphological changes were observed in dormant MTB treated with the same concentration of INH (Fig. 1e, f). However, dormant MTB appeared fragile due to cavitation and thinning of the cell wall (Fig. 1c, e).

Fig. 1
figure 1

Effect of INH (0.20 μg/ml) on the viability of log-phase (Control) and dormant MTB. a 15-day-old MTB grown in Middlebrook 7H9 medium was treated with 0.20 µg/ml of INH, and viable cells were counted (CFU/ml) on days 16, 17, 18 and 19. b 21-day-old dormant MTB was treated with the same concentration of INH, grown under hypoxic condition, and viable cells were counted on days 22, 23, 24, 28 and 35. In both experiments, untreated MTB at same time points was used as control. Results from a representative sample of two biological replicates are shown. The error bars indicate standard deviations of the mean colony counts. Morphology of log-phase and dormant MTB upon treatment with INH. Scanning electron microscopic (SEM) images showing c log-phase MTB, d INH-treated log-phase cells, e dormant MTB, and f INH-treated dormant MTB. Magnification-5000× for (c) and (d), and 10,000× for (e) and (f), respectively

Full size image

Expression of katG and inhA is comparable in INH-treated and untreated dormant MTB

To test if katG expression played a role in INH resistance in dormant MTB, we compared the expression of katG during normoxia and dormancy by qPCR. Interestingly, katG expression was found to be similar in dormant MTB and log-phase bacteria (Fig. 2a). Reanalysis of KatG protein levels in NRP-1 and NRP-2 stages of dormancy from our previous proteomics study also showed similar results (Fig. 2b) [27]. katG expression remained comparable in dormant and actively growing MTB treated with INH, compared with untreated bacteria (Fig. 2c). Next we examined the expression levels of inhA under the same conditions. Similar to katG, expression of inhA also remained similar in both dormant and actively growing MTB (Fig. 2d). From the analysis of the proteomics data from our previous studies, we found that InhA levels were comparable in dormant and actively growing MTB (Fig. 2e) [27]. inhA expression remained unchanged in dormant and actively growing MTB treated with INH, compared with untreated controls (Fig. 2f).

Fig. 2
figure 2

katG and inhA expression in log-phase and dormant MTB. a katG expression in log-phase and dormant MTB. b quantitative proteomic analysis of KatG during non-replicating persistence stage-1 (NRP-1) and non-replicating persistence stage-2 (NRP-2) [27]. c katG expression during log-phase and dormancy in INH-treated and untreated MTB. d inhA expression in log-phase and dormant MTB. e quantitative proteomic analysis of InhA during non-replicating persistence stage-1 (NRP-1) and non-replicating persistence stage-2 (NRP-2) [27]. f inhA expression during log-phase and dormancy in INH-treated and untreated MTB. The results are expressed as relative fold expression after normalizing with sigA, the endogenous control. Results are represented as mean of two biological replicates ± standard deviation

Full size image

Drug efflux pumps in MTB are downregulated during dormancy, but do not respond to INH treatment

Since katG and inhA expression remained unaltered during dormancy, we wondered if drug efflux pumps were used by the dormant bacteria to expel INH from the cell. We selected five of the best-known MTB efflux pump proteins for analysis – IniA, PstB, BacA, MmR, and EfpA [29,30,31,32,33]. Surprisingly, the mRNA levels of all these efflux pump proteins were significantly down in dormant MTB when compared to actively growing bacteria (Fig. 3a). Interestingly, when dormant MTB was treated with INH, it did not display an upregulation of any of the genes coding for these efflux pump proteins (Fig. 3c), whereas iniA, pstB, and efpA were found to be upregulated significantly in normoxially grown INH-treated MTB (Fig. 3b).

Fig. 3
figure 3

a Expression profile of efflux pump protein genes - iniA, pstB, bacA, mmR, and efpA during normoxia (control) and dormancy. b Expression profile of the same genes in log-phase MTB and in MTB treated with INH. c Expression of the same genes in dormant MTB and INH-treated dormant MTB. The results are expressed as relative fold expression after normalizing with sigA, the endogenous control. Results are represented as the mean of two biological replicates ± standard deviation (Student’s t-test). ***P ≤ 0.001, **P ≤ 0.01 and *P ≤ 0.1

Full size image

Meta-analysis of INH-treated MTB

Identification of INH and INH-NAD+ adduct in bacterial metabolome

To test if INH is metabolized to active INH-NAD+ adduct by dormant MTB, we isolated total metabolites from dormant and actively dividing MTB treated with INH for 24 and 48 h. INH (Sigma-Aldrich) was used as mass standard for LC–MS analysis. In a 20 min run using water:acetonitrile gradient, we observed a peak at 0.84 s which corresponded to INH with a molecular mass of 138.067 Da (Fig. 4a, e), and this mass was absent in metabolites isolated from untreated MTB (Fig. 4d, h). Similarly, we were able to identify small amounts of the pro-drug (INH) from actively dividing (Fig. 4b, c, f, g) and dormant (Fig. 5b, c, f, g) MTB (24 and 48 h post treatment) at the same retention time. We were able to identify INH (Fig. S1A, C, B & D) even in the clinical isolates (INH mono-resistant and MDR) after 48 h of INH treatment. Interestingly, there was a significant reduction in the quantities of INH at 24-48 h in actively growing MTB.

Fig. 4
figure 4

Meta-analysis to identify INH in log-phase MTB. a Extracted ion chromatograms (EIC) of INH (500 ng/µl) dissolved in water (standard). EIC of INH from total metabolites identified from log-phase MTB after (b) 24 and (c) 48 h of INH treatment. d EIC of metabolites isolated from untreated MTB culture. MS spectra (showing m/z) of the EIC. e INH standard. MTB after 24 (f) and 48 h (g) of INH treatment. h Untreated MTB culture. m/z of INH is marked by arrows in all panels. In all the panels, intensities of expected mass are given at the top right of the mass spectra

Full size image
Fig. 5
figure 5

Meta-analysis to identify INH from dormant MTB. a Extracted ion chromatograms (EIC) of INH (500 ng/µl) dissolved in water (standard). EIC of INH from total metabolites identified from dormant MTB after (b) 24 and (c) 48 h of INH treatment. d EIC of metabolites isolated from untreated MTB culture. MS spectra (showing m/z) of the EIC. e INH standard; dormant MTB after 24 (f) and 48 h (g) of INH treatment. h Untreated dormant MTB metabolites. m/z of INH is marked by arrows in all panels. In all the panels, intensities of expected mass is given at the top right of the mass spectra

Full size image

Active form of INH (INH-NAD+ adduct) has a molecular mass of 769.34 Da [34]. From our meta-analysis, we were able to identify this mass from actively dividing MTB treated with INH (24 and 48 h). In a 20 min metabolite run, a peak which corresponded to 769.34 Da was observed at 10.03 min from actively dividing MTB (Fig. 6a, b, e, f). This mass was absent in metabolites from untreated MTB. Quite interestingly, INH-NAD+ adduct was absent among the metabolites isolated from INH-resistant MTB strains treated with INH (Fig. S2A, B, C &D). This was expected because both the strains (INH mono-resistant and MDR) have S315T mutation in their KatG, as a result of which they could not convert the pro-drug into the active form. Remarkably, we observed that dormant MTB could convert the pro-drug to INH-NAD+ adduct at 24 and 48 h after INH treatment (Fig. 7a, b, d, e). The efficiency of conversion was similar to that by actively growing the bacteria as evidenced by comparable peak heights (Figs. 6e, f and 7d, e).

Fig. 6
figure 6

Meta-analysis to identify INH-NAD+ adduct from log-phase MTB. Extracted ion chromatograms (EIC) of INH-NAD+ adduct (769.34 Da), with a retention time of 10 min identified from metabolites of log-phase MTB treated with INH after 24 (a) and 48 h (b) of INH treatment. However, this mass is absent in standard (c), and untreated MTB metabolites (d) at same retention time. MS spectra (showing m/z) of the EIC of the peak eluted at the 10th min of the run from log-phase MTB after 24 (e) and 48 h (f) of INH treatment. Active INH-NAD+ adduct is absent in standard (g), and untreated MTB metabolites (h). m/z of INH-NAD+ adduct is marked by arrows in all panels. In all the panels, the intensities of expected mass are given at the top right of the mass spectra

Full size image
Fig. 7
figure 7

Meta-analysis to identify INH-NAD+ adduct from dormant MTB. Extracted ion chromatograms (EIC) of INH-NAD+ adduct (769.34 Da), with a retention time of 10 min identified from metabolites of dormant MTB treated with INH after 24 (a) and 48 h (b) of INH treatment. However, this mass is absent in untreated dormant MTB metabolites at the same retention time (c). MS spectra (showing m/z) of the EIC of the peak eluted at the 10th min of the run (e) from dormant MTB after 24 (d) and 48 h (e) of INH treatment. Active INH-NAD+ adduct is absent in untreated dormant MTB metabolites (f). m/z of INH-NAD+ adduct is marked by arrows in all panels. In all the panels, the intensities of expected mass are given at the top right of the mass spectra

Full size image

Discussion

Two distinct mechanisms contribute to resistance against INH in MTB, the first and most studied being chromosomal mutations in INH response genes. The second mechanism is the acquisition of growth phase-dependent phenotypic resistance to INH. Alternatively called INH tolerance, this occurs during the stationary phase of growth, hypoxia, or dormancy and is one of the reasons for the need for prolonged chemotherapy due to emergence of a subpopulation of bacteria called persistors. Conditions of progressive hypoxia or nutrient starvation can induce INH resistance in MTB [26]. Transcriptional profiling of MTB during these conditions revealed that lack of INH response genes correlated with INH resistance [26]. A similar study that used Wayne’s model of dormancy to probe the sensitivity of bacilli to INH by transcriptional profiling reports that INH treatment failed to elicit a transcriptional response from non-replicating bacteria (NRP-2) [35]. Our analysis of katG and inhA transcripts also confirmed that the expression levels of these genes are unaltered during dormancy and INH treatment. Another hypothesis is that the cell wall of dormant MTB prevents entry of the pro-drug into the intracellular compartment. There is compelling evidence that there is a metabolic shift from lipid biosynthesis to lipid degradation in the cell wall during dormancy [27]. It is evident from the photomicrographs that dormant MTB cells showed significant cavitation and thinning of the cell wall that lent them a fragile appearance. The cells also showed a decrease in size (1-2 µm) when compared to the log-phase MTB (8-10 µm). However, treatment of dormant MTB with INH did not cause any alterations in cell morphology compared to the untreated dormant bacteria. INH penetrates host cells readily; its antimycobacterial activity was found to be equal against intracellular and extracellular MTB in vitro [36]. Detection of INH and its metabolic byproducts from INH-treated MTB has been recognized as a challenging task requiring high sensitivity and precision metabolomics approaches [37]. Most of the chemical modifications that INH undergoes inside the bacilli including formation of INH-NAD+ adduct have been demonstrated only through cell-free assays [38]. The first attempt in this direction used a LC/MS-based metabolomics approach to detect INH or any of its byproducts in the urine collected from TB patients undergoing drug treatment. Mahapatra et al. (2015) were unable to detect INH-NAD+ adducts from the urine samples of INH-treated TB patients, although they detected 4-isonicotinoyl nicotinamide (4-INN), which they propose to result from the hydrolysis of the INH-NAD+ adduct. Host enzymes like lactoperoxidases are also capable of metabolizing INH, and the same authors found that mice which were not infected with MTB, but treated with INH, also secreted 4-INN. 4-INN was found in the urine of culture-negative TB patients also [34]. As this conversion is possibly mediated by host enzymes, the utility of 4-INN as a marker for INH activation in TB infection is limited [39]. In another study which analyzed the plasma metabolome of adult pulmonary TB patients treated with first-line drugs detected the m/z for INH at low intensity, but did not detect any INH metabolites other than acetylhydrazine (m/z 97.0389) and isoniazid pyruvate (m/z 208.0682) [40]. In our in vitro model of dormancy, which is based on the Wayne’s model of progressive hypoxia, we show that INH penetrates dormant bacilli and can be readily detected in the metabolome. Also for the first time, using targeted metabolomics approach, we demonstrate that dormant MTB can actively metabolize INH to active INH-NAD+ adduct. Our observations contradict the current hypothesis that INH resistance in dormant MTB is due to the inability of dormant MTB to convert INH to its active metabolite. We propose that the phenotypic resistance of dormant bacteria could be attributed to hitherto unknown factors that target biochemical steps downstream of INH adduct formation. Further investigation is needed to tease out the mechanism of growth phase-induced INH tolerance in MTB.