In Vitro Inhibition of the Mycobacterium tuberculosis β-Ketoacyl-Acyl Carrier Protein Reductase MabA by Isoniazid

Abstract

The first-line specific antituberculous drug isoniazid inhibits the fatty acid elongation system (FAS) FAS-II involved in the biosynthesis of mycolic acids, which are major lipids of the mycobacterial envelope. The MabA protein that catalyzes the second step of the FAS-II elongation cycle is structurally and functionally related to the in vivo target of isoniazid, InhA, an NADH-dependent enoyl-acyl carrier protein reductase. The present work shows that the NADPH-dependent β-ketoacyl reduction activity of MabA is efficiently inhibited by isoniazid in vitro by a mechanism similar to that by which isoniazid inhibits InhA activity. It involves the formation of a covalent adduct between Mn^III-activated isoniazid and the MabA cofactor. Liquid chromatography-mass spectrometry analyses revealed that the isonicotinoyl-NADP adduct has multiple chemical forms in dynamic equilibrium. Both kinetic experiments with isolated forms and purification of the enzyme-ligand complex strongly suggested that the molecules active against MabA activity are the oxidized derivative and a major cyclic form. Spectrofluorimetry showed that the adduct binds to the MabA active site. Modeling of the MabA-adduct complex predicted an interaction between the isonicotinoyl moiety of the inhibitor and Tyr185. This hypothesis was supported by the fact that a higher 50% inhibitory concentration of the adduct was measured for MabA Y185L than for the wild-type enzyme, while both proteins presented similar affinities for NADP⁺. The crystal structure of MabA Y185L that was solved showed that the substitution of Tyr185 induced no significant conformational change. The description of the first inhibitor of the β-ketoacyl reduction step of fatty acid biosynthesis should help in the design of new antituberculous drugs efficient against multidrug-resistant tubercle bacilli.

Due to the presence of a thick and lipid-rich envelope, the tubercle bacillus (Mycobacterium tuberculosis) is intrinsically resistant to most of the broad-spectrum antibiotics. Moreover, like most pathogenic bacteria, the bacillus has developed resistance to the antituberculous agents used clinically. The outbreak of multidrug-resistant M. tuberculosis strains is one of the major causes of the resurgence of tuberculosis observed throughout the world since the mid-1980s (9, 12). Accordingly, the design of new efficient drugs has become one of the main objectives of tuberculosis control programs.

A high prevalence of resistance to isoniazid (INH), a first-line antituberculous antibiotic, has been observed around the world. The INH resistance phenotype correlates with mutations in the katG and inhA genes in more than half of the M. tuberculosis clinical isolates (4, 14, 15, 22, 23, 29, 30, 34). INH inhibits the growth of the tubercle bacillus after oxidation by the catalase-peroxidase KatG (35, 37). The resulting active species inhibits the activity of the InhA enzyme, a primary target of INH in M. tuberculosis (2, 18), by covalently binding to its cofactor, NADH, forming an isonicotinoyl-NAD adduct (19, 25, 31).

InhA belongs to a type II fatty acid elongation system (FAS), called FAS-II, which produces unusually long fatty acids (C₁₈ to C₃₀) (20). This complex is composed of several proteins (5), including KasA/KasB, MabA (for mycolic acid biosynthesis A), and InhA, which catalyze the first, second, and fourth steps of the elongation rounds, respectively (17, 21, 26, 32). Recent reports strongly suggest that KasA, a β-ketoacyl-acyl carrier protein (ACP) synthase, is not an INH target (16, 18). The mabA (fabG1) gene is localized directly upstream of the inhA gene on the M. tuberculosis chromosome (2), and both genes are believed to form an operon. InhA, an NADH-dependent 2-trans-enoyl-ACP reductase (26), and MabA, an NADPH-dependent β-ketoacyl-ACP reductase (KAR) (21), are both functionally and structurally related. The catalytic mechanisms used by these two types of enzymes are very similar (27), and so are their cofactors. Furthermore, MabA and InhA display the same specificity for long-chain substrates (21, 26). Structural studies of these proteins have shown that they both belong to the short-chain dehydrogenases reductases (SDR) or the reductases, epimerases, dehydrogenases (RED) structural superfamily (8, 10). Finally, they contain hydrophobic substrate binding pockets of similar sizes and hydrophobicities. Therefore, we investigated a possible interaction between MabA and a specific inhibitor of InhA, namely, INH. Our data showed that MabA activity is also inhibited by isoniazid in vitro. Hence, the mechanism of action of the antituberculous drug on this FAS-II enzyme was further investigated.

MATERIALS AND METHODS

Chemical synthesis of INH adducts.

The INH-NADP adduct was prepared by mixing 2 mM INH, 2 mM NADP⁺, and 4 mM Mn^III-PyrPh in 100 mM sodium phosphate buffer (pH 7.0), and the mixture was incubated for 20 min at room temperature. This reaction mixture was called the “inhibition mix.” The INH-NAD and INH-nicotinic acid adenine dinucleotide phosphate (NAADP) adducts were prepared as described above by replacing NADP⁺ by NAD⁺ and NAADP⁺, respectively. The spectrum of absorption between 200 and 500 nm of each reaction mixture was determined with a thermostated Uvikon 923 spectrophotometer (Bio-Tek Kontron Instruments). The total concentration of adducts in the reaction mixture was estimated from the optical density (OD) measurement at 326 nm by using a molar extinction coefficient of 6,900 M⁻¹ cm⁻¹ (19).

LC, HPLC, and MS analyses.

High-performance liquid chromatography (HPLC) analyses were performed by using a C₁₈ column (125 by 4.6 mm; particle size, 3 μm; Spherisorb) by elution with a linear gradient of 0 to 12% acetonitrile in an aqueous solution of 75 mM ammonium acetate and a flow rate of 0.5 ml/min (total elution time, 100 min); monitoring was performed with a Gilson HPLC apparatus coupled to a UV-visible wavelength detector set up at 260 nm. For liquid chromatography (LC)-electrospray ionization (ESI)-mass spectrometry (MS) analyses, performed in the positive mode, 50% of the flow eluted from the column was introduced into the electrospray turbosource (heated at 480°C) of a Perkin-Elmer SCIEX API 365 mass spectrophotometer. Commercial compounds (INH, isonicotinic acid [INA], NADP⁺) were used as standards.

The pool of INH-NADP adducts was separated from the contaminants (residual reactive compounds, degradation products, or INH metabolites) by chromatography on a Sep-Pak reverse-phase C₁₈ cartridge (360 mg or 5 g of resin). After an extensive wash of the resin with acetonitrile, the cartridge was equilibrated with an aqueous 75 mM ammonium acetate solution. After the reaction mixture was loaded, elution of the contaminants by the ammonium acetate solution was monitored by measurement of the OD at 260 nm. The subsequent elution of adducts by water was monitored by measurement of the OD at 326 nm. The pool of adducts was then frozen at −80°C overnight, lyophilized, and stored at −20°C either as a powder or as a suspension in water.

MabA overproduction and purification.

The mabA gene from M. tuberculosis H37Rv (Rv1483) was cloned into plasmid pET-15b (Novagen) as described previously (21). The Y185 mutation of the MabA protein was generated as described previously (8) by using the following oligonucleotide primers: 5′-CGC GAA TGT GGT GGC GCC GGG CTT AAT CGA CAC CGA TAT G-3′ and 5′-CAT ATC GGT GTC GAT TAA GCC CGG CGC CAC CAC ATT CGC G-3′. MabA codon 185, TAC (tyrosine), was changed to TTA (leucine); and a silent mutation (GCC to GCG) at position 182 was introduced to create a new restriction site (SmaI). The absence of a secondary mutation was verified by sequencing. The plasmids bearing either the wild-type (wt) mabA gene (pET-15b::mabA-wt) or the mabA Y185L gene (pET15b::mabA Y185L) were then used to transform strain Escherichia coli BL21(DE3) (Novagen). Both MabA wt and Y185L proteins were produced from the recombinant strains as N-terminal His-Tag fusion proteins and purified by chromatography on a nickel-nitrilotriacetic acid (Ni-NTA) agarose (Qiagen) column by previously described procedures (8, 21), except that morpholineethanesulfonic acid buffer and LiSO₄ were replaced by piperazine-N,N′-bis(2-ethanesulfonic acid) (PIPES) buffer and KCl, respectively, for crystallogenesis.

MabA activity assays in the presence of INH.

MabA activity was assayed as described previously (21) by adding the enzyme to the following standard reaction medium, which was thermostated at 25°C, in a final volume of 1 ml: 100 μM acetoacetyl coenzyme A (acetoacetyl-CoA) and 100 μM NADPH in 100 mM sodium phosphate buffer (pH 7.0). Oxidation of NADPH was followed by kinetic measurement of the decrease in absorbance at 340 nm. The activity was defined as the initial velocity. For each datum point, it was verified that the reaction medium displayed no activity prior to the addition of the enzyme.

The effect of INH on MabA activity was tested as follows: 3 μM MabA was premixed with 100 μM INH, 100 μM NADP⁺, and 200 μM Mn^III-pyrophosphate (Mn^III-PyrPh) (or 1 mM INH, 1 mM NADP⁺, and 2 mM Mn^III-PyrPh); the enzyme activity was assayed by adding a 40-μl aliquot to the standard reaction medium described above after various incubation times. For each experimental condition, control assays lacking one or several reactive compounds (INH, NADP⁺ or Mn^III-PyrPh) were performed. Mn^III-PyrPh was prepared as described previously (24).

Adduct forms were purified by HPLC, and their concentrations were determined spectrophotometrically at 260 nm (ɛ, 27,000 M⁻¹ cm⁻¹ [19]). Their effects on MabA activity were tested by adding the compound (final concentration, 2.7 μM) to the standard reaction medium described above before starting the reaction by adding the enzyme. Initial velocities were compared to those obtained in control experiments, in which the corresponding HPLC eluent alone was added in place of the adduct form. Note that the concentrations of acetonitrile and ammonium acetate in the eluent had little effect, if any, on MabA activity.

The 50% inhibitory concentrations (IC₅₀s) for the different inhibition mixes were calculated with the GraphPad Prism program after the following experiments were performed. The adducts were formed as described above (see “Chemical synthesis of INH adducts”). MabA (wt or Y185L) activity (initial velocity) was measured by using 400 nM enzyme in the presence of 100 μM acetoacetyl-CoA, 100 μM NADPH, and various adduct concentrations: for INH-NADP, 0.2 to 57 μM (MabA wt) or 0.2 to 100 μM (MabA Y185L); for INH-NAD, 0.2 to 57 μM (MabA wt); for INH-NAADP, 1.5 to 49 μM (MabA wt). The reactions were started by adding the enzyme to the reaction medium. Control experiments that measured the effect of the addition of INH, Mn^III, NADP⁺, NAADP⁺, or NAD⁺ alone were performed under similar conditions.

Fluorescence spectroscopy.

The MabA wt was dialyzed at 4°C against 1,000 volumes of 100 mM sodium phosphate buffer (pH 7.0) complemented with 10 or 20% (wt/vol) glycerol. Measurements were performed at room temperature by using a final protein concentration of 3 or 6 μM in the same buffer used for dialysis, in the presence or absence of a ligand, and in a total volume of 200 μl. In the case of the inhibition mix, a mixture composed of 2 mM INH, 2 mM NADP⁺, and 4 mM Mn^III-PyrPh was preincubated for 20 min; and then a 10-μl aliquot was added to the protein solution in order to have final theoretical concentrations of each reactive compound of 100, 100, and 200 μM, respectively. Control experiments with an incomplete inhibition mix lacking one or two of the reactive compounds were also performed in the same way. In other series of experiments, the ligands were used at 1.43 mM for acetoacetyl-CoA and at 125 μM for NADP⁺. Steady-state emission spectra were recorded with a spectrofluorimeter (SAFAS; Monaco) through a 10-nm-bandwidth monochromator in a quartz cuvette at room temperature. The excitation wavelength was set at 295 nm, and fluorescence emission was measured between 315 and 400 nm. The recorded spectra were corrected by use of the spectrum for the same medium without MabA. K_D (dissociation constant) values were measured as described previously (8) with various coenzyme concentrations, from K_D/2 to 10 × K_D.

Analysis of MabA-adduct complexes.

One volume of a 108 μM solution of MabA was added to 1 volume of the inhibition mix (see above) for 10 min at room temperature under gentle agitation. A 0.5-ml Ni-NTA agarose column, which had previously been equilibrated with 50 mM potassium phosphate buffer (pH 7.8) plus 5 mM imidazole, was loaded with the mixture. Unbound molecules were eliminated by extensive washing with the same buffer, and the process was monitored by measuring the optical density at 326 nm. Protein-ligand complexes were eluted from the column with 250 mM imidazole in buffer, with monitoring at 280 nm. The protein fraction was concentrated in an ultrafiltration unit (cutoff, 10 kDa; Centricon; Amicon) by centrifugation (3,000 × g) and then frozen at −20°C overnight. After the fraction was quickly thawed at room temperature, the insoluble material (corresponding to denatured protein) was precipitated by centrifugation. The supernatant was analyzed by HPLC as described above.

Molecular modeling.

Protein sequence database searches, alignment refinement, molecular modeling, and model evaluation were performed as described previously (21). Three-dimensional models of the cofactor-bound MabA and adduct-bound MabA were built by using the quaternary structure deduced from the crystal structure of the apo form of MabA (8) and either NADP⁺-bound FabG from Brassica napus (Protein Data Bank [PDB] accession number 1EDO) (11) or InhA-INH-NAD ternary complex (PDB1ZID) (31).

Crystallogenesis and crystallography.

Crystals of MabA Y185L were grown as described above for the wt protein (8). Crystals were obtained at 281 K by the hanging-drop vapor diffusion technique by mixing 1 μl of protein solution (10 mg/ml) with 1 μl of well solution containing about 10% polyethylene glycol 3350, 50 mM PIPES buffer at pH 6.2, 300 mM CsCl, and 10% glycerol. The data were collected with a Rigaku RU-200 rotating anode X-ray generator (graphite monochromated CuKα; λ = 1.5418 Å) and a MarResearch image plate detector (18 cm). Image processing and data scaling were performed with the DENZO and SCALEPACK software packages. Structure factor amplitudes were converted with the TRUNCATE program of the CCP4 suite of programs. All crystallographic refinement was performed by using the CNS program (version 1.0) (7). The wt MabA structure at a 2.03-Å resolution (8) was used as the starting model and for initial phasing. Manual rebuilding between refinement cycles was carried out with the TURBO-FRODO graphics package (Silicon Graphics, Mountain View, Calif.).

RESULTS

Inhibition of MabA activity by activated INH.

The His-tagged M. tuberculosis MabA protein was purified to homogeneity from an E. coli strain that overproduced the protein, and the action of INH on enzyme activity was studied. It has been shown that manganese ions in their Mn^III state are able to activate INH, thus mimicking the in vitro action of the catalase-peroxidase KatG, but at a much higher rate (24, 25). The unstable Mn^III ions can be stabilized by complexation with PyrPh (1). Thus, the effect of the addition of INH (100 μM [13.7 μg/ml]), NADP⁺ (100 μM), and Mn^III-PyrPh (200 μM) to MabA on the enzyme's ability to reduce acetoacetyl-CoA to β-hydroxybutyryl-CoA by using NADPH as a cofactor was tested. Strong inhibition (43%) of MabA activity was observed after just 1 min of incubation of the enzyme with INH, NADP⁺, and Mn^III-PyrPh (Fig. 1). The inhibition rate increased slightly with the incubation time (57% at 18 min). For comparison, InhA activity was inhibited by 77% under similar conditions (in which NADH replaced NADP⁺) (25). When MabA was exposed to higher concentrations of INH (1 mM), NADP⁺ (1 mM), and Mn^III-PyrPh (2 mM), its activity was much more impaired (e.g., about 86% inhibition after 30 s of incubation). No inhibition was observed in control experiments with one of the three molecules alone or when only Mn^III-PyrPh or INH was lacking. In contrast, incubation of the enzyme in the presence of only INH and Mn^III-PyrPh induced a decrease in enzymatic activity compared to those achieved in the control assays (Fig. 1). Similar observations were made when INH and Mn^III-PyrPh were preincubated for 20 min prior to incubation in the presence of MabA (data not shown). This phenomenon is most likely due to the presence of NADPH in the assay for enzyme activity, since NADPH could replace NADP⁺ in the formation of an inhibitory component. Indeed, spectrofluorimetry data (see below) suggested that the mixture of INH and Mn^III-PyrPh does not produce any molecular species able to bind to MabA. The highest levels of inhibition of MabA activity were obtained with the complete mixture (INH, NADP⁺, and Mn^III) (Fig. 1).

FIG. 1.

Inhibition of MabA activity by INH in the presence of Mn^III and NADP⁺. MabA (3 μM) was premixed with INH (100 μM), Mn^III-PyrPh (200 μM), and NADP⁺ (100 μM) prior to determination of its activity at different incubation times in the presence of acetoacetyl-CoA and NADPH (×). Control experiments in which the inhibition mix lacked one or several of the reactive compounds (INH, Mn^III-PyrPh, or NADP⁺) were performed. For more clarity, only data from experiments without NADP⁺ (▪), Mn^III-PyrPh (○), or INH (•) are displayed. The other controls gave results similar to those obtained without Mn^III-PyrPh or without INH. The inhibition rates reported in the text were calculated for a given time by using the two controls without INH and without Mn^III-PyrPh as references.

These results led to the conclusion that INH becomes an active inhibitor of MabA activity in the presence of the oxidized coenzyme NADP⁺ and Mn^III ions.

Effect of a preformed mix on MabA activity.

In order to determine whether the inhibitor of MabA activity could be formed in the absence of MabA or if the enzyme is required for the synthesis of its own inhibitor, NADP⁺, INH, and Mn^III-PyrPh were premixed (inhibition mix) and preincubated for 20 min. The effect of this mixture on MabA activity was measured in the presence of acetoacetyl-CoA and NADPH. The IC₅₀ of the inhibition mix was determined to be 2.2 μM (Table 1). Thus, an inhibitory molecule that is able to impair the activity of MabA can be produced outside the MabA active site by reaction between INH, Mn^III, and NADP⁺.

TABLE 1.

IC₅₀s of the mixture of INH-NADP adducts or adduct derivatives for MabA activity

Adduct	IC50 (μM)a
	MabA wt	MabA Y185L
INH-NADP	2.2 ± 0.4	6.8 ± 1.2
INH-NAD	12.9 ± 0.3	ND
INH-NAADP	5.3 ± 0.0	ND

^aInhibition experiments were performed in the presence of MabA (400 nM), acetoacetyl-CoA (100 μM), and NADPH (100 μM) and various concentrations of the adduct mixture. Data were fitted by nonlinear least-squares regression to determine IC₅₀; standard deviations are provided. ND, not determined. Control experiments showed that NAD⁺, INH, or Mn^III alone (concentrations, >5 mM) had no effect on MabA wt activity and that NAADP⁺ or NADP⁺ alone had little effect (IC₅₀s, >1 mM).

Binding of an inhibitory molecule to MabA.

Interaction of MabA with a potential inhibitor formed in the inhibition mixture was investigated by spectrofluorimetry. Indeed, the unique tryptophan residue of MabA was shown to be located in its substrate binding pocket. The presence of the intrinsic fluorophore allows the binding of a ligand to the active site to be detected (21). Thus, the fluorescence emission of MabA was quenched in the presence of NADP⁺ or acetoacetyl-CoA, and stronger quenching was observed in the presence of both substrates (Fig. 2A). The effects of different mixtures of ligands, which were preincubated for 20 min prior to the addition of MabA, were then tested. It appeared that INH had no significant impact on the fluorescence of the protein alone or in complex with NADP⁺, suggesting that the antibiotic in its native form does not bind to MabA (Fig. 2B). Furthermore, the fluorescence spectra were equivalent in the presence of Mn^III-PyrPh plus INH or Mn^III-PyrPh alone, which does not support the occurrence of a potential interaction between MabA and a metabolite of INH oxidation by Mn^III (described by Nguyen et al. [24]). In contrast, when INH was preincubated with both NADP⁺ and Mn^III-PyrPh (complete inhibition mixture), a quenching of fluorescence emission much stronger than that recorded for the control with NADP⁺ plus Mn^III-PyrPh was observed.

FIG. 2.

Fluorescence spectra of MabA in the presence of ligands. (A) Fluorescence emission of MabA alone (•) or in the presence of acetoacetyl-CoA (▪), NADP⁺ (▵), or both ligands (□). (B) Fluorescence emission of MabA in the presence of the complete inhibition mix (INH plus NADP⁺ plus Mn^III) (+); in the presence of NADP⁺ (▵), Mn^III (♦), INH (○), NADP⁺ plus INH (▴), NADP⁺ plus Mn^III (⋄), or INH plus Mn^III (□); or in the absence of ligand (•). The fluorescence of the protein was slightly affected by the presence of Mn^III-PyrPh, virtually because of the charge-induced quenching of the solvent-exposed tryptophan.

Altogether these experiments showed that at least one molecule formed during preincubation of the inhibition mixture is able to specifically bind to MabA.

Formation of an isonicotinoyl-NADP adduct.

In order to determine the nature of the MabA inhibitor, the inhibition mixture was first analyzed by spectrophotometry. The spectrum displayed an absorption maximum at 260 nm with a shoulder at 326 nm, while the latter was not observed for control samples lacking one of the three reactive compounds (INH, NADP⁺, or Mn^III-PyrPh) (data not shown). Kinetic experiments under the same conditions revealed that absorption at 326 nm started a plateau after about 10 min and reached the maximum (>99%) after 37 min of incubation. These data show that the molecule(s) formed in the total inhibition mix has a typical absorption at 326 nm, like the INH-NAD adduct which binds to the InhA protein (19), strongly suggesting that this molecule(s) corresponds to a structural analog(s) INH-NADP. The nature of the compound(s) was further investigated by LC-MS in the ESI mode. The mixture separated into 10 main peaks on the HPLC profile (Fig. 3). Peaks 2, 3, and 5 were identified by mass spectrometry as INA, as well as residual NADP⁺ and INH (Fig. 3). INA has previously been reported to be the main INH metabolite in oxidative conditions (33). Peaks 6 to 11 displayed the mass value expected for an INH-NADP adduct ([M + H]⁺ m/z 851). The mass spectra of four of these six peaks (peaks 6, 8, 10, and 11) also presented a pseudo-molecular ion with a loss of one molecule of water ([M + 18 − H]⁺ m/z 833). Analysis of INH-NAD adducts (6, 24) has shown that this phenomenon is linked to cyclization of the molecule (Fig. 4A, compound III), which facilitates dehydration (Fig. 4A, compound IV), as observed during mass analyses. As the cyclic derivative holds a new asymmetric center, the four peaks most likely correspond to the four diastereoisomers (6, 24). Peaks 7 and 9, which did not undergo dehydration, would represent the two epimers of the intact reduced form of the INH-NADP adduct (Fig. 4A, compound II). Peak 4 corresponded to a unique molecular ion (M⁺ of m/z 849) attributed to the oxidized form of the adduct (Fig. 4A, compound I).

FIG. 3.

HPLC profile of the inhibition mix. After a 30-min incubation, the reaction mixture containing NADP⁺, INH, and Mn^III-PyrPh in sodium phosphate buffer (pH 7.0) was loaded onto a C₁₈ reverse-phase HPLC column; and the compounds were eluted with a gradient of ammonium acetate and acetonitrile. The peak constituents were analyzed by ESI-MS in the positive mode and were attributed as follows: peak 1, solvent; peak 2, INA ([M + H]⁺ m/z 124); peak 3, NADP⁺ ([M + H]⁺ m/z 744); peak 4, oxidized form of INH-NADP adduct (M⁺ m/z 849); peak 5, INH ([M + H]⁺ m/z 138); peaks 7 and 9, reduced open forms of INH-NADP adduct ([M + H]⁺ m/z 851); peaks 6, 8, 10, and 11, reduced cyclic forms of INH-NADP adduct ([M + H]⁺ m/z 851) that undergo dehydration ([M + 18 − H]⁺ m/z 833) during mass spectrometry analysis. OD, optical density.

FIG. 4.

Proposed structures of different forms of adduct. (A) Isonicotinoyl-NADP adduct. I, oxidized form; II, reduced forms; III, cyclic reduced forms; IV, cyclic dehydrated forms. The HPLC peak numbers correspond to those in Fig. 3, and the proposed configurations of the corresponding molecules, as well as data from MS analysis, are presented. (B) Isonicotinoyl-NAADP adduct. *, asymmetric centers.

By comparison with the INH-NAD adducts (6, 24, 25), peaks 10 and 11 would correspond to the diastereoisomers 4S,7R, and 4R,7S of the cyclic forms, respectively; peaks 6 and 8 would correspond to the cyclic diastereoisomers 4S,7S and 4R,7R, respectively; and peaks 7 and 9 would correspond to the 4R and 4S open epimers, respectively (Fig. 4A). The small areas of peaks 6 and 8 compared to those of peaks 10 and 11 may be explained by a steric hindrance between the two six-carbon rings (Fig. 4A) (25). These data demonstrate the formation of an INH-NADP adduct with many different chemical structures.

The INH-NAD adduct that inhibits the InhA protein was also tested for its activity against MabA under the same conditions used for the INH-NADP pool. The IC₅₀ of the INH-NAD adduct was much greater than that of INH-NADP (Table 1). This observation is in agreement with the specificity of MabA for NADPH versus NADH as a cofactor (21) and shows that the presence of the isonicotinoyl moiety in the ligand does not totally compensate for the loss of the phosphate group. It also suggests that NADP binds to MabA in a similar manner whether it is a cofactor or part of the INH-NADP adduct.

Another derivative of the INH-NADP adduct was synthesized by incubating INH and Mn^III-PyrPh in the presence of NAADP⁺ (Fig. 4B). The reaction mixture exhibited the characteristic absorption at 326 nm, but it had a much simpler HPLC profile, due to the inability of the resulting adduct to cyclize (data not shown). The lower inhibitory potency of the INH-NAADP adduct on MabA activity compared to that of INH-NADP (Table 1) may be explained by the importance of either the amide group of nicotinamide or the cyclization process of the adduct in the interaction with the protein.

Identification of the active form of the INH-NADP adduct.

Since open, cyclic, and oxidized forms of the INH-NADP adduct are present in the inhibition mixture, it was essential to identify the species that inhibit MabA activity. While the mixture of adducts was stable for several hours at 0°C or for several weeks at −20°C, isolated adduct forms collected by HPLC were rapidly converted into the other adduct forms. A procedure was set up in order to concentrate the single molecules while preserving their integrity by charging the inhibition mixture on a Sep-Pak C₁₈ column to separate the pool of INH-NADP adducts from the residual reactive compounds and degraded products. The adduct mixture was eluted in water and then strongly concentrated (80 times) by lyophilization, and the different forms were separated by HPLC. Each form of the INH-NADP adduct was collected from the HPLC column and their effect on MabA activity was immediately tested to minimize the transformation into isomers. It was verified by HPLC analysis that the isolated molecules were not significantly metabolized at the temperature used for the assays (25°C) during the time course of one experiment. The effects of the different forms on MabA activity were compared by use of a fixed and identical concentration of each form. The molecules displaying the strongest inhibitory potency corresponded to the oxidized form in peak 4 and one of the major cyclic diastereoisomers in peak 10 (Fig. 5; Fig. 4A). The compound contained in peak 11 also induced a significant inhibition rate; however, since it was not perfectly separated from peak 10 by HPLC at the high sample concentrations used in this experiment, its activity may be linked to contamination by the molecule in peak 10. Interestingly, other isomeric adduct forms, including one of the open epimers (Fig. 4A, peak 9), poorly affected MabA activity (Fig. 5). In order to verify the results presented above by another technique, MabA was incubated in the presence of the inhibition mixture, and the enzyme-ligand complexes were then purified from free molecules by Ni-NTA affinity chromatography. The protein was then denatured by freezing-thawing, and the ligands released were analyzed by HPLC. Although some degradation of the ligands occurred during the experiments, peaks 4 and 10 were observed as the main constituents (data not shown), in agreement with the data presented above.

FIG. 5.

Effects of the isolated INH-NADP adduct forms on MabA activity. Each form of the INH-NADP adduct was collected from the HPLC column. The pool of adducts before HPLC separation (mix) or one of the isolated adduct forms was added at a final concentration of 2.7 μM to a reaction medium containing acetoacetyl-CoA (100 μM) and NADPH (100 μM) in sodium phosphate buffer (pH 7.0). After addition of MabA (final concentration, 400 nM), the activity of MabA was monitored spectrophotometrically. Inhibition rates correspond to the ratios of the activity assayed to the control activity, expressed as percentages. In control experiments, the corresponding HPLC eluent was added in place of the adduct form. Peak numbers correspond to those in Fig. 3 and 4.

Implication of a tyrosine residue in the stabilization of the INH-NADP adduct in the MabA active site.

Although MabA shares only 20% sequence identity (over 200 residues) with InhA, these two proteins are structurally closely related. The superimposition of their crystal structures showed a root mean square deviation of 1.4 Å over a common core of 112 C^α atoms (8) and, in particular, a perfect conservation of the topology of the cofactor binding site. The three-dimensional structures of apo-MabA (8) and the InhA-INH-NAD complex (PDB1ZID) (31) were used to create a molecular model of the binary complex MabA-INH-NADP adduct. Superimposition of the modeled or crystal structures of the protein-adduct complexes of both MabA and InhA allowed the position of the INH-NADP adduct inside the MabA active site to be predicted to be in the vicinity of the unique tryptophan (Trp145) (Fig. 6). The Phe149 residue, which has been shown to be the key factor for stabilization (by aromatic stacking) of the adduct in InhA (31), is replaced by a serine residue (Ser140) in MabA. This polar amino acid does not permit any hydrophobic interaction with the pyridine cycle of the isonicotinoyl moiety. However, a tyrosine residue (Tyr185) that faces Ser140 could readily make a π-π stacking with the adduct (Fig. 6). Alignment of the MabA sequence with that of KARs of 16 other organisms (bacteria or plants) revealed that the position equivalent to Tyr185 most often corresponds to a phenylalanine residue (data not shown). In one case only over 16 KARs, it corresponds to a leucine. In mycobacteria, e.g., M. tuberculosis, M. bovis, M. avium, and M. smegmatis, the tyrosine is conserved at this position; however, it is not in M. leprae (in which a phenylalanine is found). To investigate the potential involvement of Tyr185 in the binding of the INH-NADP adduct to MabA, the residue was changed to Leu by directed mutagenesis, and the mutated protein was purified.

FIG. 6.

Structural model of isonicotinoyl-NADP adduct in the MabA active site. Amino acid residues surrounding the adduct in InhA (PDB1ZID) are shown as white sticks (carbon atoms), while equivalent residues in MabA are in green; Trp145, however, is in yellow. The oxygen, nitrogen, and sulfur atoms of the residues are in red, blue, and orange, respectively. The isonicotinoyl moiety of the adduct is in pink, while the NAD(P) moiety is in blue (only the ribosyl-nicotinamide part is displayed). The figure was produced with the PyMol program (DeLano Scientific, South San Francisco, Calif.; http://pymol.sourceforge.net/).

The mutant protein was crystallized under conditions similar to those used for the wt protein (8). The three-dimensional structure of MabA Y185L was solved by X-ray diffraction. A crystal diffracting to a resolution of 2.5 Å was used to record the data to a level of completeness of 86.6% (R_merge = 11.5%). The crystal packing was isomorphous with that of the wt enzyme, with similar unit cell dimensions (space group C-2; a = 85.53 Å, b = 116.99 Å, c = 54.75 Å, γ = 124.7°) (8). Refinement of the structure to a final R_cryst factor of 25.0% (R_free = 31.9%) was achieved by a standard protocol. The crystal structure of the mutant enzyme revealed that the Y185L mutation did not induce any significant changes in the overall structure of the protein (root mean square deviation, 0.35 Å). The only clear difference in the electronic density appeared for residue 185, due to the side chain change. The leucine side chain points to the substrate binding site in a similar way as the tyrosine side chain (data not shown).

The similar K_D values for NADP⁺ for both the wt and the Y185L enzymes (8.7 ± 1.1 and 6.9 ± 0.9 μM, respectively), as measured by spectrofluorimetry, suggested that the cofactor binding site was little affected by the point mutation, in agreement with the crystallographic data. As expected, however, the activity of MabA Y185L tested in the presence of acetoacetyl-CoA and NADPH displayed a decrease of 70% ± 2% compared to that of the wt protein under identical experimental conditions. Indeed, the strong conservation of an aromatic residue at position 185 among the known KARs suggested that it is important for enzymatic activity. The observed loss of activity upon replacement of Tyr by another hydrophobic residue (leucine) is in agreement with its location within the active site. Moreover, the IC₅₀ of the INH-NADP adducts measured for MabA Y185L appeared to be three times greater than that for the wt protein (Table 1).

Taken together these results show that the interaction of MabA with a substrate or a cofactor derivative such as the INH-NADP adduct, but not with the cofactor, was significantly affected by the Y185L mutation. This supports the model of the MabA-adduct complex (Fig. 6) that suggests that the stabilization of isonicotinoyl-NADP in the MabA active site is partly based on the interaction between the isonicotinoyl moiety and the aromatic ring of residue Tyr185.

DISCUSSION

The present work provides convincing arguments for the in vitro action of INH on the MabA protein of M. tuberculosis by a mechanism similar to that previously described for InhA. Mn^III-oxidized INH covalently binds to the oxidized form of the MabA cofactor, NADP⁺, forming an isonicotinoyl-NADP adduct that binds to the MabA active site. The IC₅₀ of the INH-NADP adduct for MabA (2.2 μM) has the same order of magnitude as the IC₅₀ of the isonicotinoyl-NAD adduct for InhA, estimated to be 1 to 2 μM from previously published data (25). As in the case of the INH-NAD adduct, the MabA inhibitor is composed of a mixture of seven different isomers or derivatives (for the oxidized form) that are in dynamic equilibrium. Although isolated molecules are very unstable, we succeeded in separating them by LC and HPLC. Data obtained by distinct techniques converged and directed us toward the same conclusions. They revealed that mainly two forms of the adduct, the oxidized derivative and one of the cyclic diastereoisomers (whose proposed configuration is 4S,7R), are effective against the activity of MabA. Interestingly, it has been shown that the equivalent cyclic isonicotinoyl-NAD adduct (with the proposed configuration of 4S,7R) strongly inhibits the activity of InhA (25). Furthermore, it has recently been reported that the oxidized adduct would also be cyclic (6). The inhibitory activities of cyclic forms of INH-NADP might explain the higher IC₅₀ obtained for the INH-NAADP adduct that cannot cyclize, a hypothesis consistent with what has been observed for InhA in the presence of INH-NAD and INH-NAAD adducts (25).

Superimposition of a molecular model of the MabA-isonicotinoyl-NADP binary complex with the InhA-adduct crystal structure allowed us to propose the potential involvement of Tyr185, located in the substrate binding pocket, in the binding of the inhibitor to MabA. Kinetic experiments with the mutated MabA Y185L gave a significantly higher IC₅₀ than that obtained for the wt isoenzyme, in agreement with the modeling data. The crystal structure of the mutant protein positively demonstrated that the mutation effect was not caused by a global conformational change. Rather, the similar affinities of wt and mutant Y185L MabA enzymes for NADP⁺ suggested that the involvement of Tyr185 would be based on its interaction with the isonicotinoyl moiety of the adduct, as implied by the model of the MabA-adduct complex. This situation is reminiscent of that observed for InhA in complex with the INH-NAD adduct, in which the ligand is stabilized through π-π stacking between the isonicotinoyl moiety and the Phe149 residue of InhA (Fig. 6) (31).

The present work allowed to identify the first inhibitor of a KAR protein. Interestingly, this compound is a metabolite of the first-line antituberculous drug INH. Until now, inhibitors of most steps of the fatty acid elongation pathway monitored by type II systems were known, except for malonyl-CoA-ACP transacylation and ketoacyl-ACP reduction (13). To date, only a single isoform of KAR (FabG) has been described, and it is ubiquitous in the bacterial genomes that have been sequenced (13). Furthermore, the fabG gene is predicted to be essential in E. coli (36) and M. smegmatis (3). Accordingly, KARs could be used as targets for broad-spectrum antimicrobials, but we have previously shown that the mycobacterial KAR MabA has specific functional and structural properties compared to those of the homologous bacterial proteins, such as a large hydrophobic substrate binding pocket, which correlates with its preference for long-chain substrates (8, 21). Thus, MabA could be used as a target for the design of drugs that act specifically against mycobacteria. Furthermore, the mechanism of action of an inhibitor that has proved to be the most efficient antituberculous drug could serve as a model for rational drug design.

Point mutations were detected in the mabA gene of INH-resistant M. tuberculosis clinical isolates (28), as well as in the potential common promoter region of the mabA-inhA locus (14, 15, 22, 23, 30, 34), most likely generating INH resistance by inducing overexpression of the genes downstream (18). Although cloning of mabA in an expression vector did not confer INH resistance to M. smegmatis (3), one cannot rule out the possibility that the naturally mutated promoter or mutation of the gene itself could be responsible for a resistance phenotype in M. tuberculosis. We are investigating this question.