Mechanism of thioamide drug action against tuberculosis and leprosy

To identify the active form of ETH, we and others have attempted to use purified EthA to activate ETH and inhibit InhA in vitro but have never been able to observe any InhA inhibition (unpublished data). Because in vitro activation of the drugs ETH and PTH has not been possible by either chemical or enzymatic approaches, we developed a cell-based activation method. In this system, recombinant M. tuberculosis EthA and InhA were co-overexpressed in the same Escherichia coli cell, and ETH or PTH was added to the culture to test whether the drugs would inhibit InhA upon activation. Although ETH and PTH are both potent drugs against M. tuberculosis (MIC = ∼0.5 μg/ml) (24), they do not affect E. coli growth, even at very high concentrations (100 μg/ml), which is primarily caused by the absence of an EthA homologue in E. coli.

InhA and EthA from M. tuberculosis were coexpressed in E. coli BL21 (DE3) in the presence of 100 μg/ml ETH. InhA was rapidly purified, and an in vitro enzyme assay was performed. InhA isolated from the experimental sample had <1% of the specific activity of InhA purified without the addition of ETH under the same assay condition. Mass analysis of denatured InhA from the experimental sample indicated the presence of a small molecule with a molecular weight of 798.2 (Fig. S1, A and B). This corresponds to the exact molecular weight of an ethyl- isonicotinic-acyl-NAD covalent adduct. Moreover, pure fractions of this small molecule showed strong inhibition to native InhA in vitro (Ki = 7 ± 5 nM), which is as potent as the INH-NAD adduct, the active form of INH (Ki = 5 nM) (8). When PTH was used in the same coexpression experiment, a compound with a molecular weight of 812.2 was identified that corresponds to the exact weight of a propyl-isonicotinic-acyl-NAD adduct (Fig. S1 C). This compound is also extremely potent against InhA in vitro (Ki = 2 ± 0.8 nM).

M. tuberculosis InhA in complex with the adducts was crystallized. X-ray diffraction data to 2.2 and 2.5 Å resolution were collected from single crystals of ETH and PTH complexes, respectively (Table S1). Unbiased electron density maps of each complex clearly indicated the presence of a modified NAD with an ethyl-isonicotinic-acyl or propyl-isonicotinic-acyl group covalently attached to the 4 position of the nicotinamide ring in a 4S configuration (Fig. 2 A). The chemical structures of both inhibitors are consistent with the molecular weights obtained by the mass analysis. Similar to the structure of InhA bound with the adduct INH-NAD (7), the ethyl-isonicotinic-acyl, or the propyl-isonicotinic-acyl, moiety is found in a hydrophobic pocket that was formed by the rearrangement of the side chain of Phe¹⁴⁹ (Fig. 2, B–D). The ethyl-isonicotinic acyl or the propyl- isonicotinic-acyl group also forces the side chain of Phe¹⁴⁹ to rotate ∼90°, forming an aromatic ring-stacking interaction with the pyridine ring (Fig. 3 A). The pocket is predominantly lined by hydrophobic groups from the side chains of Tyr¹⁵⁸, Phe¹⁴⁹, Met¹⁹⁹, Trp²²², Leu²¹⁸, Met¹⁵⁵, Met¹⁶¹, and Pro¹⁹³, and is adjacent and partly overlapped with the fatty acyl substrate-binding site. Indeed, the atoms common to ETH-NAD, PTH-NAD, and INH-NAD are in nearly identical positions. The only difference is the extra ethyl or propyl group at the 2 position of the pyridine ring of ETH or PTH. The ethyl group contributes to the binding of ETH-NAD adduct by forming π-stacking interactions with the aromatic side chain of Tyr¹⁵⁸ at a distance of ∼3.3 Å. It is also within van der Waal interaction distances with side chains of Leu²¹⁸ (3.3 Å) and Met¹⁵⁵ (3.2 Å). The hydrogen-bonding interactions between the phosphate group of the adduct and residues of the nucleotide-binding site are well conserved. Therefore, it is very likely that mutations, such as S94A, that decrease the binding of NAD(H) and the INH-NAD adduct would also weaken the binding of ETH-NAD and PTH-NAD (Fig. 4). This explains why the S94A mutant strain of M. tuberculosis is coresistant to both INH and ETH.

Other than ETH and PTH, thioamide drugs such as thiacetazone and isoxl have been shown to be activated by EthA (20). The same cell-based method was applied to test thiacetazone. The isolated InhA was not inhibited under the same assay condition. As expected, mass analysis did not show the existence of any tightly bound inhibitor. These results indicate that, unlike ETH or PTH, thiacetazone does not target InhA, even though all of these thioamides are activated by EthA in M. tuberculosis.

Unlike INH, which is not effective against M. leprae, most likely because of the dysfunction of M. leprae katG, ETH and PTH are used in the treatment of leprosy. Genome analysis indicated that both M. leprae and M. avium have ethA and inhA homologues with high sequence similarities to their M. tuberculosis counterparts. Therefore, the M. leprae ethA and inhA were coexpressed, in the presence of PTH using a cell-based activation method similar to M. tuberculosis, as described in the second paragraph in Results and discussion. A compound with the same molecular weight as PTH-NAD adduct was identified, which also showed strong inhibition to M. leprae InhA (Ki = 11 ± 6 nM) in vitro.

The crystal structure of M. leprae InhA in complex with PTH-NAD was solved to 2.1 Å resolution to compare the binding mode of the inhibitor with the enzyme with M. tuberculosis InhA. The overall structure of M. leprae InhA is similar to M. tuberculosis InhA (RMSD = 1.3 Å for C_αs), and the active site residues are conserved. The propyl-isonicotinic-acyl moiety of the adduct is observed inside a hydrophobic pocket. The pyridyl ring forms a π-stacking interaction with the aromatic side chain of Phe¹⁴⁹, and the propyl group is π stacking with the aromatic side chain of Tyr¹⁵⁸. Residues, including Ser⁹⁴ in the nucleotide-binding site, form hydrogen-bonding interactions with the phosphate group of the adduct in a similar manner as M. tuberculosis InhA (Fig. 3 B). These results supported our hypothesis that the active form of PTH inhibits M. leprae InhA in a similar way to M. tuberculosis InhA. Although no clinical or experimental mutant of M. leprae InhA has been reported thus far, based on the binding mode of the PTH-NAD adduct, it is very likely that the mutations of InhA found in ETH-resistant M. tuberculosis mutant strains, such as S94A, would also confer resistance of M. leprae to ETH and PTH.

Although ETH and INH NAD adducts are similar, their activation mechanisms are very different. INH, a hydrazid, is activated by the heme-using catalase-peroxidase KatG (16, 18). We proposed that INH was oxidized to generate an isonicotinic-acyl free radical, which subsequently attacked the NAD⁺ to form INH-NAD (7). ETH, a thioamide, has been shown to be metabolized by EthA, a FAD enzyme found in M. tuberculosis. It has been proposed that a free radical metabolite intermediate could be generated through EthA oxidation of ETH, similar to the activation of INH. However, no active species that inhibit InhA were isolated in vitro (20), which suggests that an unknown cell component, either a protein or cell membrane, is required for the formation of the adduct by the free radical intermediate. We believe that the inactive metabolites isolated in previous attempts could result from side reactions and quenching of the free radical intermediate in solution. It is still not clear how the thioamide is oxidized by EthA. Tokuyama et al. demonstrated that a thioamide could be used as a precursor of a synthon equivalent to an imidoyl radical in converting thioamides to corresponding indole derivatives. Bu₃SnH/Et₃B has been used as a free radical initiator in pure organic solvent (25). Similarly, we postulate that ETH is converted to an imidoyl radical, and this imidoyl radical subsequently attacks NAD⁺ to form an adduct, which is then converted to ethyl-isonicotinic-acyl-NAD adduct after hydrolysis to release the amine group. It is also possible that the imidoyl anion is the intermediate before forming the adduct with NAD (Fig. 5). However, based on the current evidence, we are not certain if this reaction is catalyzed by EthA alone or requires the involvement of additional enzymes.

The experiments presented in this paper describe the molecular mechanism of the drug action of ETH and PTH against M. tuberculosis and M. leprae. The identification of the ETH-NAD and PTH-NAD adducts, like the INH-NAD, represent a novel paradigm in the history of drug action. Efforts to detect the INH-NAD and ETH-NAD adduct in M. tuberculosis or other mycobacteria have been unsuccessful. However, four different mechanisms of coresistance to INH and ETH have been discovered: (a) overexpression of the InhA target (6, 12); (b) target modification to prevent adduct binding (8); (c) activator modification to prevent prodrug activation and adduct formation (20); and (d) mutations in ndh, the gene encoding the essential NADH dehydrogenase II that regulates intracellular NADH/NAD ratios (26). All of these observations are consistent with our working model that INH, ETH, and PTH are all prodrugs that form NAD adducts to inhibit InhA. The discovery of ETH-NAD and PTH-NAD adducts, which are generated through completely different routes from INH-NAD, further validates this model. This information is important for the optimization of drug activity and the understanding of drug resistance. Because the molecular target of ETH, PTH, and INH is the same, it validates InhA as an outstanding antituberculosis and antileprosy drug target. However, because most of the clinical strains resistant to ETH and PTH contain mutations in the ethA gene, it is advantageous to find agents that inhibit InhA, without the need for EthA activation, as effective chemotherapy against resistant bacteria.

The M. tuberculosis inhA has been previously cloned (14). M. tuberculosis ethA was cloned from genomic DNA (National Institutes of Health contract N01-AI-75320; Colorado State University). The amplified product was inserted into pET28b using the NdeI and NotI restriction sites. M. leprae ethA and inhA were cloned from genomic DNA. The amplified product of M. leprae ethA was inserted into pET15b using the NdeI and BamHI restriction sites. M. leprae inhA was inserted into pET30b using the NdeI and HindIII restriction sites.

The plasmids of M. tuberculosis inhA and ethA were singly and doubly transformed into E. coli BL21 (DE3; EMD Biosciences, Inc.). The strain containing plasmids of inhA and ethA was cultured in LB-Miller media containing 50 μg/ml kanamycin and 50 μg/ml carbanicillin at 37°C until OD₆₀₀ reached 0.8. Expression of both genes was induced for 20 h at 16°C by addition of 1 mM isopropyl β-D-thiogalactopyranoside. At the same time as induction, 100 μg/ml ETH or PTH was also added to the culture. The same protocol was used for the strain containing only the inhA plasmid.

Recombinant InhA was purified according to the previous method (14).

The coexpression and purification of M. leprae ethA and inhA were conducted using protocols similar to those used for the M. tuberculosis enzymes.

InhA purified from the experimental strain containing both inhA and ethA genes was concentrated and heated for 40 s at 100°C. After the heat treatment, ETH-NAD or PTH-NAD was separated from denatured enzymes by filtration, using a centricon device (cutoff size = 30 kD). The concentration of ETH-NAD and PTH-NAD was determined by its absorbance at 260 and 326 nM (16). The molecular weight of both adducts was determined by matrix-assisted laser desorption/ionization (MALDI) performed on an ABI Voyager-DE STR (AME Bioscience): ETH-NAD, calculated weight = 797.2 and found weight = 797.3 (negative mode), and calculated weight = 799.2 and found weight = 799.2 (positive mode); PTH-NAD, calculated weight = 811.2 and found weight = 811.3 (negative mode).

All assays were performed on a spectrophotometer (Cary 100 Bio Spectrophotometer; Varian, Inc.) at 25°C by monitoring oxidation of NADH at 340 nm. Reactions were initiated by adding 50 μM of substrate dodecenoyl-CoA to assay mixtures containing 1 nM InhA, 100 μM NADH, and 3–2,000 nM of adduct inhibitors.

The IC₅₀ was determined from the dose-response plot of enzyme fractional activity as a function of inhibitor concentration. K_i was obtained by dividing the IC₅₀ value by 1 + [S₁]/K_m1 + [S₂]/K_m2, where [S₁] and [S₂] are the concentrations of dodecenoyl-CoA and NADH, and K_m1 and K_m2 are their Michaelis constants.

Crystallization was accomplished by the hanging drop vapor diffusion method (27). M. tuberculosis InhA in complex with inhibitors was cocrystallized in hanging droplets containing 2 μl of protein solution at 10 mg/ml and 2 μl of buffer (12% MPD, 4% DMSO, 0.1 M Hepes, 0.025 M sodium citrate) at 16°C in Linbro plates against 1 ml of the same buffer. Diamond-shaped protein crystals formed ∼4 d later.

M. leprae InhA in complex with inhibitor was cocrystallized in a similar manner, and the crystal had a cubic shape.

Data were collected at 121 K using cryoprotection solution containing reservoir solution with an additional 30% MPD. Crystals of M. tuberculosis InhA:ETH-NAD and M. leprae InhA:PTH-NAD diffracted x rays to 2.2 and 1.8 Å using the beamline 23-ID at the Advanced Photon Source (Argonne National Laboratory, Argonne, IL). Diffraction data were collected from a single crystal with 1° oscillation widths for a range of 120°. Crystals of M. tuberculosis InhA:PTH-NAD were diffracted to 2.5 Å using a Raxis image plate detector coupled to a Rigaku x-ray generator using a copper rotating anode (CuK_α, λ = 1.54 Å). The data were integrated and reduced using HKL-2000 (HKL Research, Inc.; Table S1) (28).

Crystals produced from InhA in complex with ETH-NAD were isomorphous to those of the native enzyme. Initial phases were obtained by molecular replacement using the apo-InhA structure (1ENY) and refined with CNS software (Table S1) (29). F_o – F_c and 2F_o – F_c electron density maps were calculated, and an additional density resembling the inhibitor was found. The ligand was fit into the additional density, and the whole model was rebuilt using XtalView (30). During the final cycles of the refinement, water molecules were added into peaks above 3 σ of the F_o − F_c electron density maps that were within hydrogen-bonding distances from the appropriate protein atoms. The final refinement statistics are listed in Table S1.

Table S1 provides data collection, processing, and refinement statistics. Fig. S1 depicts MALDI mass spectra showing that two inhibitors bound to InhA are compounds with the apparent weights of 798 and 812, respectively. Fig. S2 shows the crystal structure of the adducts superimposed onto the simulated annealing omit electron density maps contoured at 1 and 1.5 σ.

We thank Shane Tichy for his excellent technical assistance in mass spectrometry.

This work was supported by National Institutes of Health grant PO1AIO68135, Einstein/MMC Center for AIDS Research grants AI51519 and AI43268, and by the Robert A. Welch Foundation. G.S. Besra acknowledges support from Mr. James Bardrick in the form of a Personal Research Chair and from the Medical Research Council.

The authors have no conflicting financial interests.