Substrate specificity of the deazaflavin-dependent nitroreductase from Mycobacterium tuberculosis responsible for the bioreductive activation of bicyclic nitroimidazoles


  • M. G. and T. M. contributed equally to this work

U. H. Manjunatha, Novartis Institute for Tropical Diseases, 10 Biopolis Road, Singapore 138670
Fax: +65 6722 2917
Tel: +65 6722 2976
C. E. Barry III, Tuberculosis Research Section, NIAID, NIH, Bethesda, MD 20892, USA
Fax: +1 301 480 5705
Tel: +1 301 435 7509


The bicyclic 4-nitroimidazoles PA-824 and OPC-67683 represent a promising novel class of therapeutics for tuberculosis and are currently in phase II clinical development. Both compounds are pro-drugs that are reductively activated by a deazaflavin (F420) dependent nitroreductase (Ddn). Herein we describe the biochemical properties of Ddn including the optimal enzymatic turnover conditions and substrate specificity. The preference of the enzyme for the (S) isomer of PA-824 over the (R) isomer is directed by the presence of a long hydrophobic tail. Nitroimidazo-oxazoles bearing only short alkyl substituents at the C-7 position of the oxazole were reduced by Ddn without any stereochemical preference. However, with bulkier substitutions on the tail of the oxazole, Ddn displayed stereospecificity. Ddn mediated metabolism of PA-824 results in the release of reactive nitrogen species. We have employed a direct chemiluminescence based nitric oxide (NO) detection assay to measure the kinetics of NO production by Ddn. Binding affinity of PA-824 to Ddn was monitored through intrinsic fluorescence quenching of the protein facilitating a turnover-independent assessment of affinity. Our results indicate that (R)-PA-824, despite not being turned over by Ddn, binds to the enzyme with the same affinity as the active (S) isomer. This result, in combination with docking studies in the active site, suggests that the (R) isomer probably has a different binding mode than the (S) with the C-3 of the imidazole ring orienting in a non-productive position with respect to the incoming hydride from F420. The results presented provide insight into the biochemical mechanism of reduction and elucidate structural features important for understanding substrate binding.


deazaflavin-dependent nitroreductase


F420-dependent glucose-6-phosphate dehydrogenase


minimal inhibitory concentration


Mycobacterium tuberculosis




photomultiplier tube






One-third of the human population is infected with tuberculosis (TB) and nearly 2 million die every year from active disease [1]. Mycobacterium tuberculosis (Mtb), the causative agent of TB, invades human alveolar macrophages where it has evolved to evade or survive the hostile conditions within these professional phagocytes [2]. Partial containment of infection is accomplished by formation of a highly organized, multicellular granuloma within which nutrient deprivation and hypoxia contain, but do not kill, the invading microbe [3]. Perhaps because of the complexity of the intracellular environment and the variable replication state of the organism, current treatment for TB requires a combination of isoniazid, rifampicin, pyrazinamide and ethambutol for 2 months followed by isoniazid and rifampicin for 4 months [4]. Inadequate treatment and failure to comply with drug regimens have resulted in the emergence of multi-drug resistant and extensively drug resistant Mtb strains [5,6]. New agents with the potential to shorten treatment, for example by killing anaerobically non-replicating Mtb, are urgently needed to reverse current trends.

Nitroimidazoles are commonly used antibiotics for anaerobic microbial infections in humans. They show better activity against obligate anaerobes than aerobic organisms because their bactericidal activity requires an oxygen-sensitive bioreduction of the aromatic nitro group [7]. Metronidazole (Mtz) is a widely used 5-nitroimidazole for the treatment of infections caused by anaerobic bacteria [8,9]. Mtb cultures become sensitive to Mtz under low oxygen conditions. However, the compound has no activity against aerobically growing bacteria [10]. Aerobic antitubercular activity of bicyclic nitroimidazoles was first reported from a series of 4- and 5-nitroimidazole[2,1-b]oxazoles [11]. A highly active lead compound, CGI-17341, showed activity in a murine infection model but was subsequently dropped [12,13] because of in vitro mutagenic results in the Ames test. Subsequent elaboration of the bicyclic nitroimidazole series resulted in the identification of Ames negative analogs PA-824 [14] and OPC-67683 [15], two compounds that are currently in phase II clinical development (Fig. 1). Both compounds are pro-drugs that require bioreductive activation and show no cross-resistance to other front-line TB drugs. Activated PA-824 inhibits lipid biosynthesis in a dose-dependent manner [14]. Transcriptional profiling analyses of PA-824 treated cells suggest that inhibition of both cell wall biosynthesis and respiration contribute to the cidal activity of this compound [16].

Figure 1.

 Chemical structures of F420 and nitroimidazole substrates used in this study. The C-5 carbon of F420 carrying hydride and the proposed site of initial hydride addition on C-3 of PA-824 are each shown by an asterisk.

Several studies have characterized spontaneously generated PA-824 resistant mutants as a means to identify the cellular machinery involved in its activation. The original observation that the F420-dependent glucose-6-phosphate dehydrogenase (FGD1) was required for sensitivity [14] was followed by a series of papers describing resistance in isolates that had lost the ability to biosynthesize the deazaflavin cofactor F420 [17,18] and culminated in identification of a conserved hypothetical protein encoded by Rv3547 that was essential for susceptibility to the compound [19]. FGD1 catalyzes the oxidation of glucose-6-phosphate to phosphogluconolactone and in turn reduces F420 to F420H2. F420, an obligate two electron transfer agent, is a low redox potential, soluble 7,8-didemethyl-8-hydroxy-5-deazariboflavin with a ribosyl-phospholactyl moiety and polyglutamate chain (Fig. 1) [20]. F420H2 is the active form of the cofactor that is utilized by a protein encoded by Rv3547, an enzyme henceforth referred to as a deazaflavin-dependent nitroreductase (Ddn). The physiological role of Ddn is unknown. The F420H2-dependent reduction of PA-824 by Ddn produces three stable metabolites resulting from reduction of the imidazole ring at C-3 [21]. One of the major products formed is the des-nitro metabolite with subsequent release of nitrous acid which degrades to NO. We have evaluated several PA-824 analogs and observed that the amount of des-nitro metabolite formed (and the corresponding amount of NO generated) correlates well with their anaerobicidal activity [21]. However, the aerobicidal and anaerobicidal activities of PA-824 correlate poorly [22,23] and our mechanistic understanding of the reaction course leading to aerobic or anaerobic activity remains incomplete.

In order to further enhance our understanding of the mechanism of action of PA-824, we have studied the F420H2-dependent nitroreductase activity of Ddn using PA-824 and a selected collection of chemically distinct nitroimidazole analogs by investigating reoxidation of F420H2, production of NO and by determining binding constants of the analogs to the protein. These results suggest that the tail portion of the nitroimidazole determines the binding orientation of the head group, conferring stereospecificity in orientation of the molecule towards reduction.


Optimization of Ddn catalyzed PA-824 reduction

Recombinant Ddn was expressed and purified as an MBP-His6 tagged protein in Escherichia coli (see Experimental procedures). F420 was isolated from Mycobacterium smegmatis and reduced to F420H2 as described in Experimental procedures. Enzymatic activity of Ddn was determined spectrophotometrically by monitoring F420H2 oxidation at A400 nm. The enzyme activity of Ddn was determined between pH 6.0 and pH 8.0; the optimum pH for Ddn mediated F420H2 oxidation was 7.0–8.0 (Fig. S1D). At pH 8.0, the enzyme was active at ambient temperature whereas its activity was compromised at 48 °C (Fig. S1E). The inclusion of various monovalent (Na+, K+) and divalent ions (Mg2+, Mn2+) had no effect on enzyme activity (data not shown). However, the presence of a detergent (0.01% Triton X-100) improved the enzyme activity significantly; thus the final buffer contained 200 mm Tris/HCl (pH 8.0) with 0.01% Triton X-100 (Fig. S1F). Addition of 10% dimethylsulfoxide (DMSO), 2 mm dithiothreitol (DTT) and 20 mm EDTA to the Ddn buffer had no effect on enzyme activity (Fig. S1G).

The F420H2 specificity of Ddn was investigated by monitoring Ddn catalyzed PA-824 reduction in the presence of various other redox active cofactors. The enzyme was not able to utilize NADH and NADPH in the reduction of PA-824 and was highly specific for F420H2 (data not shown). F420 is present within cells as a series of polyglutamylated forms containing two to six glutamates in the tail in a species-specific pattern. Both M. smegmatis and Mtb produce predominantly F420 species with five to six glutamate residues (Fig. 1) [24]. The ability of Ddn to utilize the F420 isolated from Methanobacterium thermoautotrophicum, which produces F420 with two glutamate residues in its side chain (F420-2) [25], was therefore evaluated. Ddn enzyme was able to utilize reduced F420-2 (Km,app 26 μm) as efficiently as F420-5 from M. smegmatis (Km,app 29 μm).

The lipophilic tail of nitroimidazole substrates determines reduction selectivity and efficiency

To understand the effect of substrate structure on the enzyme kinetic parameters, we first determined the apparent steady state kinetic constants Km,app and kcat/Km,app for the Ddn reaction with PA-824 in the presence of 100 μm F420H2 and 1 μm Ddn (Fig. 2A,B). The Km,app for PA-824 was 29 μm and kcat/Km,app was 0.16 μm−1·min−1 (Fig. 2B, Table 1). To explore reduction selectivity, we determined the kinetic parameters of Ddn with a limited set of structurally diverse PA-824 analogs including (R)-PA-824, des-nitro PA-824, Mtz, (R) and (S) stereoisomers of CGI-17341 and (R) and (S) stereoisomers of a simple phenyloxazole series (Fig. 1). Ddn was not able to oxidize F420H2 when using des-nitro PA-824 or (R)-PA-824 as substrates (Fig. 2B), supporting the importance of the nitro group and the previously noted enantiomeric specificity for reduction of these nitroimidazo-oxazines [21]. Ddn showed no biochemical activity using Mtz, a monocyclic 5-nitroimidazole, as a substrate (Fig. 2B) reinforcing the notion that Mtz killing of anaerobic Mtb occurs via a mechanism that is independent of F420 and Ddn. Aza-PA-824 was equally good as a substrate as PA-824 for Ddn with a kcat/Km,app of 0.12 μm−1·min−1 (Fig. 2B and Table 1).

Figure 2.

 Ddn kinetics monitored via F420H2 oxidation. (A) Oxidation of F420H2 at varying PA-824 concentrations monitored by the increase in absorbance at 400 nm. (B) Michaelis–Menten plots for the Ddn catalyzed reduction of PA-824, (R)-PA-824, des-nitro PA-824, metronidazole, Aza- PA-824, (R)- and (S)-CGI-17341 and (R)- and (S)-phenyloxazole derivatives at 100 μm F420H2. Each value represents the mean ± standard deviation of the initial velocity of formation of oxidized F420 from a linear regression analysis. (C), (D) Two-substrate profile analysis for Ddn. Michaelis–Menten plot of (C) initial velocity versus [PA-824] and (D) initial velocity versus [F420H2] in the presence of 1 μm Ddn and varying concentration of the other substrate. Each value represents the mean ± standard deviation of the initial velocity of formation of oxidized F420 from a linear regression analysis. (E) Schematic representation of Ddn reaction mechanism: progression of reaction via a ternary complex Ddn : F420H2 : PA-824 to form F420 and PA-824 metabolite products.

Table 1.   Characterization of various PA-824 analogs. MIC, minimal inhibitory concentration; ns, not a substrate.
 F420H2 oxidation assayMIC (μm)
Km,appm)kcat,app (min−1)kcat/Km,appm−1·min−1)Mtb WTMtba: F420Mtbb: Ddn
  1. H37Rv-5A1-ΔfbiC. H37Rv-14A1-ΔDdn. Not determined because of poor aqueous solubility of OPC-67683 (< 1 μm).

PA-82428.6 ± 3.64.7 ± 0.30.160 ± 0.0200.8> 100> 100
(R)-PA-824nsnsns> 100> 100> 100
(S)-CGI-17341207.2 ± 70.82.4 ± 0.60.011 ± 0.0020.5> 1000.4
(R)-CGI-17341123.2 ± 20.12.2 ± 0.20.020 ± 0.0050.25> 1000.2–0.4
(S)-Phenyloxazole67.8 ± 37.51.7 ± 0.50.029 ± 0.01112.5> 10012.5
(R)-Phenyloxazole83.6 ± 15.74.8 ± 0.50.061 ± 0.0030.8> 1000.8
OPC-67683ccc0.01> 20> 20
Aza-PA-82412.9 ± 3.41.6 ± 0.10.117 ± 0.0150.8> 100> 100

Previously, we have shown that Mtb Ddn mutants were equally as sensitive as wild-type (WT) Mtb to the nitroimidazo-oxazole analog CGI-17341 [19] suggesting involvement of other Ddn homologs in its activation. CGI-17341 is a racemic mixture that proved to be a modest substrate for Ddn (data not shown). In order to understand the enantiomeric specificity of nitroimidazo-oxazoles, we synthesized the (R) and (S) stereoisomers of CGI-17341 (Data S1). Both (R) and (S) forms had highly potent Mtb cellular activity against WT H37Rv as well as against H37Rv-14A1-ΔDdn, suggesting a Ddn independent mode of activation. Nevertheless, activation was still F420-dependent as the H37Rv-5A1-ΔfbiC strain was resistant to both the stereoisomers (Table 1). These genetic studies suggest that while PA-824 is activated exclusively by Ddn in Mtb, activation of CGI-17341 is carried out by Ddn or other Ddn homologs (Rv1261c, Rv1558 and Rv3178). Although Ddn is not likely to be the only relevant reductase for activation of CGI-17341, these compounds nonetheless were Ddn substrates in vitro. Substrate-dependent F420H2 oxidation was observed with both the (R) and (S) isomers of CGI-17341 at about half the maximal velocity achieved for PA-824 and there was no significant rate difference between the two enantiomers (Fig. 2B, Table 1).

Guided by our earlier observations showing that PA-824 resistant Ddn mutants were partially resistant to nitroimidazo-oxazoles with longer hydrophobic substituents on the oxazole ring [19], we suspected that substrates with a longer lipophilic tail would perhaps stereoselectively interact with Ddn. To probe this hypothesis further, we synthesized the (R) and (S) stereoisomers of a nitroimidazo-oxazole with a bulkier hydrophobic phenyl side-chain (Data S2). As expected, the (R)-phenyloxazole was 15-fold more active in whole cell activity against WT H37Rv than its corresponding (S) enantiomer but similar activities of both isomers were observed when minimal inhibitory concentrations (MICs) were determined with a Ddn mutant (Table 1). Once again, this implicates the role of other homologs of Ddn in Mtb such as Rv1261c, Rv1558 and Rv3178 in the cellular activation of simple nitroimidazo-oxazole analogs. When tested against purified Ddn, both (R)- and (S)-phenyloxazole analogs were poor substrates for Ddn with kcat/Km,app values of 0.06 μm−1·min−1 and 0.03 μm−1·min−1 respectively (Table 1). Interestingly, the low specificity constants (kcat/Km,app) for all four tested nitroimidazo-oxazole analogs are largely driven by their high Km,app values. This is possibly because of the absence of the trifluoromethoxy benzyl group which seems to be important for Ddn enzyme activity [22].

Notably the maximal velocity of the (R) isomer was nearly equal to that of PA-824 although the affinity was significantly lower (Km,app 84 μm). Thus it seems that Ddn displays stereospecificity with oxazole substrates with larger lipophilic tails but the stereo preference is the reverse of that observed with the oxazine series.

Resistance to OPC-67683, the (R) isomer of a bicyclic 4-nitroimidazo-oxazole with a hydrophobic tail on the oxazole ring, has also been mapped to Ddn [15]. We found OPC-67683 to be highly active against Mtb H37Rv (MIC90 0.01 μm) whereas the Ddn mutant, as expected, was resistant to this compound, confirming that Ddn is the primary biological activator for the compound (Table 1). OPC-67683 showed substrate-dependent F420H2 oxidation by Ddn (with an initial velocity of 63 nmin−1 with 5 μm OPC-67683). However, owing to solubility limitations with the compound (< 1 μm in aqueous buffer), in vitro enzyme kinetics of Ddn with OPC-67683 could not be evaluated. Enzymatic formation of the des-nitro form of OPC-67683 with Ddn was detected by LC-MS analysis (Fig. S2, m/z 490 compared with 535 for the parent) consistent with the whole cell metabolite analysis reported previously [15].

Kinetic mechanism of Ddn catalysis

The kinetic mechanism of Ddn was studied using preliminary two-substrate profile analyses (Fig. 2C,D). The increase in the Vmax,app values with increasing concentrations of both PA-824 and F420H2 suggest that reaction chemistry must occur through a ternary complex of Ddn : F420H2 : PA-824 (Fig. 2E). However, further analysis such as product inhibition studies are required to determine precise binding constants and to distinguish between a random versus an ordered kinetic mechanism. Detailed understanding of Ddn : F420H2 : PA-824 catalysis would aid nitroimidazole optimization.

The kinetics of NO generation by Ddn

Previously, we used the Griess assay to study the in vitro release of reactive nitrogen intermediates produced from the Ddn catalyzed PA-824 reaction [21]. In this assay, a two-step diazotization reaction under acidic conditions results in the formation of an azo-complex which can be monitored spectrophotometrically. Although a reliable endpoint assay for Ddn as reported, its low sensitivity (∼ 1 μm) is of concern in the study of enzyme kinetics. Here, we employed an NO analyzer based on chemiluminescence detection with a sensitivity of ∼ 1 nm (1000-fold better) to study in vitro Ddn kinetics. Steady state kinetics of Ddn with PA-824 and F420H2 in the presence of saturating concentrations of either substrate were investigated by monitoring NO generation (Fig. 3A,B). The Km,app for PA-824 was 8 μm and for F420H2 was 22 μm with a kcat/Km,app value of 0.01 μm−1·min−1. As expected, des-nitro PA-824 and (R)-PA-824 were not substrates for Ddn [21] and therefore did not show NO release (Fig. 3A).

Figure 3.

 Monitoring Ddn kinetics using an NO analyzer. (A) Michaelis–Menten plots of the reaction of PA-824, des-nitro PA-824 and (R)-PA-824 with 1 μm Ddn at 100 μm F420H2 (inset shows the time course of NO release profiles at varying PA-824 concentrations). (B) Michaelis–Menten plot of the reaction of F420H2 with 1 μm Ddn at 50 μm PA-824. Each value represents the mean ± standard deviation of the initial velocity of formation of oxidized F420 from a linear regression analysis.

Ddn binding with (R)- and (S)-PA-824 and F420 by fluorescence quenching

The N-terminal portion of Ddn contains a critical tryptophan (Trp20) involved in binding the hydrophobic tail of PA-824. Fluorescence emission spectra of 1 μm Ddn in the presence of increasing concentrations of PA-824 revealed that PA-824 efficiently quenched a significant fraction of the intrinsic fluorescence of this protein upon binding (Fig. 4A). The extent of ligand binding to Ddn was determined by monitoring changes in maximal fluorescence emission at 334 nm. At saturating concentrations of PA-824, 69.6% of the intrinsic Ddn fluorescence was quenched with no significant change in the maximum emission wavelength. To further explore the role of stereochemistry in determining substrate binding, we also explored the inactive (R)-PA-824 in this assay, which was incapable of turnover by Ddn. Surprisingly (R)-PA-824 was able to bind to the enzyme causing a reduction in intrinsic fluorescence by 78% at saturating concentrations (Fig. 4B). However, under similar conditions, des-nitro PA-824 failed to bind Ddn (Fig. 4C). The relative change in intrinsic fluorescence 1 − F/F0 was plotted against the ligand concentration to generate saturation isotherms (Fig. 4D). In the absence of F420H2, Ddn was able to bind PA-824 and (R)-PA-824 with Kd values of 13.95 μm and 19.24 μm, respectively. Binding to the oxidized or reduced F420 was evaluated in a similar manner (Fig. 4E,F). Ddn displayed extremely tight binding to oxidized F420 with a Kd of 0.30 μm in contrast to the reduced form that showed a Kd of 3 μm (Fig. 4G). Moreover, binding of F420-2 to Ddn, monitored by the same method, was comparable to that of F420-5 from M. smegmatis suggesting that the extended polyglutamate tail did not contribute significantly to binding affinity (Fig. S3).

Figure 4.

 Binding of nitroimidazole and F420 substrates to Ddn evaluated by intrinsic tryptophan fluorescence quenching studies. Emission spectra of 1 μm of Ddn protein in the presence of increasing concentrations of (A) PA-824, (B) (R)-PA-824, (C) des-nitro PA-824, (E) F420 and (F) F420H2. Arrows indicate quenching of the fluorescence signal at the wavelength of maximum emission with increasing concentrations of ligand. (D), (G) Saturation isotherms for Ddn binding with various ligands. Kd values are indicated in the graphs. The experiments were carried out at least three independent times and one such representative profile is shown.

In order to obtain further understanding of how the inactive (R) form of PA-824 still bound to Ddn despite lack of turnover, we examined potential binding modes of both forms to the co-crystal structure of Ddn with F420 that we recently reported [21a]. The full-length Ddn protein structure was not solved; however, the high resolution crystal structure of a full-length related enzyme from Nocardia (nfa33440; see Fig. S4 for sequence alignment) as well as extensive mutagenesis studies support a structural model of the active site. We used this model to manually dock both isomers of PA-824 to arrive at putative binding modes. In both cases, a water molecule H-bonded to Ser72 was used as a docking point for the nitro group. Figure 5A shows that the C-3 of the nitroimidazole of the (S) isomer is directly positioned above the C-5 of F420H2 (shown by an asterisk in Fig. 1), and this allows a favorable hydride transfer from the C-5 of F420H2. However, the C-3 of the (R) isomer docked in a flipped orientation (Fig. 5B) and is skewed away from the C-5 of F420H2 and this makes the hydride transfer very unlikely. In both cases, the aromatic tail portion is aligned well with Trp16 for aromatic–aromatic interactions and this is consistent with the fluorescence quenching experiments. While not conclusive, this docking provides a plausible explanation for the observed stereoselectivity of the (S) isomer of PA-824 over the (R) isomer by Ddn as well as their comparable binding affinity to Ddn.

Figure 5.

 Proposed docking modes of (S) and (R) isomers of PA-824. Binding modes of (A) catalytically active (S)-PA-824 and (B) catalytically inactive (R)-PA-824 to F420 bound Nocardia Ddn homolog, nfa33440 (top panels). The respective bottom panels show the orientation of the geometry optimized enantiomers (in yellow) docked into the active site of a Nocardia Ddn homolog complexed with F420 with a few important residues in the binding pocket labeled. (A), Bottom panel: One of the oxygen atoms of NO2 of (S)-PA-824 was placed onto the crystal water molecule to H-bond with Ser72 of the Nocardia homolog. In this orientation, the C-3 of the imidazole is 4.1 Å away from the C-5 of F420, and also properly positioned for the incoming hydride transfer in terms of the angle of C-5–(H)–C-3; note that the hydride (H) of C-5 is not seen in the X-ray structure. Atoms represented by colors are as follows: green, carbon; blue, nitrogen; red, oxygen. (B), Bottom panel: The enantiomer (R)-PA-824 was docked in a flipped orientation resulting in the C-3 of the imidazole at a distance of 4.3 Å away from the C-5 of F420 and an improper positioning for the incoming hydride transfer in terms of the angle of C-5–(H)–C-3.


The possibility of emergence of drug resistance warrants further study and optimization of second generation nitroimidazoles as anti-tuberculars [22,23,26,27]. The kinetic properties of analogs of this class as substrates for Ddn are important determinants of their activity against the whole organism. Therefore a more detailed biochemical characterization of the reaction is important to support a deeper understanding of the structure–activity relationship in this series and to guide optimization efforts. Ddn enzyme activity was unaffected by the inclusion of monovalent or divalent cations. Also, the lack of any effect upon addition of a metal chelating reagent EDTA suggests that the enzyme activity is metal ion independent. Ddn is a 151 amino acid protein with only one cysteine residue at position 149 (Fig. S4), implying that only intermolecular disulfide bonds may exist, if at all. DTT, up to 2 mm, had no effect on the activity of Ddn suggesting the lack of any structurally or functionally important intermolecular disulfide bonds. Addition of Triton X-100 improved the enzyme activity significantly; the presence of a surfactant possibly stabilizes Ddn, therefore contributing to improvement in activity.

Consistent with recently reported F420 bound crystal structure of the core enzyme lacking the N-terminus [21a], Ddn was highly specific to F420H2 and was not able to utilize several other cofactors which exhibit similar reduction potentials (nicotinamide based cofactors) or resemble F420 structurally (flavins) [25]. Ddn could utilize reduced F420H2 as efficiently as the reduced F420-5 from M. smegmatis. Therefore, although the polyglutamate chain of F420 seems to be important for Ddn activity, the number of glutamate residues (ranging from 2 to 6) did not seem to contribute to nitroimidazole conversion.

The kinetic constant kcat/Km,app for the Ddn enzymatic reaction with PA-824 was 0.16 μm−1·min−1. Ddn was not able to metabolize either the (R) stereoisomer or the des-nitro derivative of PA-824. Tryptophan quenching studies indicated that Ddn was able to bind PA-824 and (R)-PA-824 with similar affinity. These results demonstrate that the physical binding of Ddn to PA-824 is not affected by the stereochemistry of the compound; however, the complete loss of activity of the enzyme with the (R) stereoisomer suggest that (R)-PA-824 may bind to Ddn non-productively and interfere with the catalysis of the enzyme probably due to the incorrect positioning of C-3 of the nitroimidazole ring to C-5 of the F420H2 from which the hydride transfer takes place. Molecular docking studies of (R)-PA-824 with Ddn strengthen this hypothesis (Fig. 5). Although we have some preliminary evidence that (R)-PA-824 is a competitive inhibitor of Ddn mediated reduction of (S)-PA-824 (data not shown), the very poor solubility of this compound precluded full characterization of the inhibition. These results, however, suggest that clinical development of a racemic mixture would not be advisable. The des-nitro derivative showed negligible binding to Ddn supporting the lack of enzyme activity with des-nitro PA-824 as a substrate, suggesting a critical role for the nitro group in binding. Binding studies of Ddn with the F420 indicated higher affinity for F420 (Kd 0.3 μm) than F420H2 (Kd 3 μm). This binding of oxidized F420 in the active site could prevent the entry of a reduced F420 into the active site possibly explaining the low turnover of the enzyme. However, the low binding constant for the product might change in the presence of the nitroimidazole substrate, which requires further investigation.

The F420H2-dependent reduction of PA-824 by Ddn produces three stable metabolites. As proposed earlier [21], the first hydride transfer from the C-5 of F420H2 to the C-3 of PA-824 results in a nitronic acid intermediate of PA-824 from which all of the three metabolites arise. The des-nitro derivative of PA-824 (a major metabolite) is formed by elimination reaction with the release of nitrous acid, which in turn disproportionates to release NO and other reactive nitrogen species. Further chemical transformations and F420H2 mediated reduction result in the formation of exocyclic ketone and exocyclic imine. The chemistry involving reduction of the C=C bond by F420H2 through the formation of a Meisenheimer complex is well precedented in enzymes of the bacterial old yellow enzyme (OYE) family involved in bioremediation of 2,4,6-trinitrotoluene and at least one of these enzymes has been shown to be F420-dependent [28,29]. The deazaflavin cofactor F420 is unique in that under normal conditions it is an obligate two electron or hydride ion donor [20]. However, the precise role of the low redox potential of the 5-deaza-cofactor (F420H2) in the mechanism of bicyclic nitroimidazole reduction is yet to be understood.

Here we also describe a highly sensitive and robust method for examining in vitro Ddn enzyme kinetics based on NO detection using an NO analyzer. The anaerobicidal activity of bicyclic 4-nitroimidazoles correlated with the formation of des-nitro species and release of reactive nitrogen intermediates [21]. Although the two nitroimidazole clinical candidates for TB in phase II trials, PA-824 and OPC-67683, have modest anaerobic activity, they have been optimized only for aerobic whole cell activity. An assay that determines the release of NO will therefore facilitate target based selection of PA-824 analogs that have improved anaerobic activity. The kcat/Km,app of PA-824 determined from the NO assay was 0.013 μm−1·min−1 which is 10 times lower than what is observed in the F420 oxidation assay. The efficiency of NO generation or detection seems to be reduced possibly due to quenching of NO in solution under the conditions tested or the non-enzymatic reaction between NO and F420H2 [30]. Consequently the utility of this assay needs to be reconfirmed after careful analysis of detection of various reactive nitrogen intermediates and reaction conditions. However, in the presence of saturating concentrations of F420H2 (100 μm) and with 4 μm PA-824, 1.12 μm NO was produced in 2 h along with 7.2 μm of oxidized F420 (implying 155 nm of NO produced per micromole of F420 oxidized). Thus in Mtb cells release of micromolar concentrations of intracellular NO by Ddn would have a significant effect on cell viability.

The physiological role of Ddn is currently unknown. Sequence homologs primarily exist in actinobacteria; however, there is little information on their potential cellular functions. Ddn was first experimentally detected in a study that identified membrane proteins in Mtb suggesting that Ddn might be a membrane associated protein [31]. Rv1261c, Rv1558 and Rv3178 are the three homologs of Ddn in Mtb that have other orthologs in other mycobacterial species. Ddn and its homologs form a class of previously uncharacterized F420H2-dependent nitroreductases with no identified physiological substrate. A recent bioinformatics study that carried out phylogenetic profiling of 1451 bacterial and archaeal genomes based on F420 biosynthesis nominated three dominant families, one of which was the Ddn family [32]. The study indicates that the Ddn family is restricted to F420 producing bacteria. Another study that identified F420H2-dependent reductases in the degradation of aflatoxins showed that a Ddn homolog in M. smegmatis, Msmeg_5998, utilized F420H2 in the reduction of the unsaturated ester moiety of aflatoxins, thereby activating them for spontaneous hydrolysis [33,34]. Both the degradation of aflatoxin and reduction of PA-824 by Ddn seem to share a common mechanism in which hydride is transferred from the low redox potential deazaflavin to the electron deficient ring systems of the substrates.

The biochemical characterization of Ddn described here with PA-824 and other analogs, and the methods developed, will prove useful in future attempts to characterize additional members of the Ddn enzyme family. In addition, since Ddn appears to be capable of rigorous structure-based specificity, characterizing this protein for its binding to the various nitroimidazole analogs will prove to be beneficial to optimize a backup clinical candidate for PA-824.

Experimental procedures

Bacteria and culture conditions

E. coli DH5α strain used for cloning and E. coli BL21 (DE3) Tuner strain used for protein expression were grown in Luria broth as per standard protocol. MIC90 determination and culture conditions for Mtb H37Rv WT, H37Rv-5A1-ΔfbiC and H37Rv-14A1-ΔDdn have been described earlier [19].

Drugs and chemicals

Mtz and isopropyl thio-β-d-galactoside (IPTG) were obtained from Sigma-Aldrich (Singapore, Singapore). (S)-PA-824, (R)-PA-824 and Aza-PA-824 [23] and OPC-67683 [35] were synthesized as described. Synthesis of (R) and (S) isomers of both CGI-17341 and a phenyloxazole derivative are described in Data S1 and Data S2. All compounds were dissolved in 90% DMSO as 50 mm stocks.

F420 purification from M. smegmatis

F420 was purified from M. smegmatis mc2155 cells using a modified version of a protocol described earlier [36]. Briefly, M. smegmatis was grown in large batches in modified strullu-romand (MSR) medium in a fermentor; cells were harvested, resuspended in water and lysed by three rounds of autoclaving (10 min at 121 °C). The lysed cell extract was filtered and passed through an HiQ (Biorad) ion exchange 250 mL column followed by Florisil (Sigma) 250 mL column chromatography. F420 was further purified using reverse phase (Source RPC 30; GE Healthcare) chromatography (acetonitrile gradient, 400 nm). A final yield of 1.18 μmol·L−1 culture was obtained which was comparable with the 1.43 μmol·L−1 reported in the literature [36]. Purified F420 was characterized by LC-ESI MS and fluorescence emission scanning. As expected, F420 showed an emission maximum at 470 nm when excited at 400 nm. Each of the peaks in the LC profile corresponded to one of the F420 analogs (Fig. S1A,B). As reported earlier [24], F420 with four to six glutamate residues accounted for the majority of the F420 analogs in the purified sample.

Cloning, expression and purification of recombinant Ddn protein

The coding sequence for Ddn (Rv3547) was amplified by PCR (Pfx polymerase; Invitrogen) from H37Rv genomic DNA using forward (5′-caggatcccgaaatcaccgccgcggtt-3′) and reverse (5′-tagcggccgctcagggttcgaaaccacga-3′) primers. PCR amplified fragments were cloned into BamHI and NotI sites of an entry vector of a Gateway expression system [37]; further the gene was sub-cloned into various destination vectors. A construct (pMBPKan5-Ddn) coding for N-terminal MBP-His6-tagged Ddn was transformed into BL21 (DE3) Tuner cells and protein expression was induced with 0.1 mm IPTG for 20 h at 18 °C. Soluble recombinant MBP-His6-Ddn was purified on a nickel affinity binding column and the fusion protein was cleaved by PreScission protease (GE Healthcare). The digested protein was subjected to ion exchange chromatography (Resource S; GE Healthcare) and the elutant was passed through a second nickel affinity binding column to remove any remaining tagged protein and non-removed tags. The purified protein was stored at −80 °C in a buffer containing 20 mm Tris/HCl pH 7.5, 100 mm NaCl, 10% glycerol, 1 mm DTT. The purified protein was a soluble aggregate and had a calculated size of ∼ 17.5 kDa on SDS/PAGE (Fig. S1C) which corresponded to the predicted molecular weight. Protein concentration of 3.03 μg·μL−1 was obtained as determined by the standard Bradford assay [38] and the final yield of the purified protein was about 30 mg·L−1 of culture.

Evaluation of Ddn enzyme activity

Enzymatic activity was determined spectrophotometrically by monitoring F420H2 oxidation at A400 nm (isosbestic point for oxidized F420) [21]. F420H2 was prepared as described previously from F420 by FGD1 enzyme in the presence of glucose-6-phosphate as the substrate [21]. F420 concentration was measured at A400 (e = 25.7 mm−1·cm−1). Typically, the assay mixture to determine enzyme activity contained 50 μm PA-824, 100 μm F420H2 and 1 μm Ddn in a final volume of 100 μL unless specified otherwise. Control reactions without the enzyme and without PA-824 were included for each set of experiments. To determine the pH optimum for Ddn activity, F420H2 oxidation was monitored at room temperature at pH ranging from 6.0 to 8.0 (Mes/OH pH 6.0–6.5, Tris/HCl pH 7.0–7.5, HEPES/NaOH pH 8.0–8.5). For temperature optima, the reaction was carried out in an optimal buffer (referred to as Ddn buffer: 200 mm Tris/HCl, pH 8.0, with 0.01% Triton X-100) at varying temperatures between 20 and 50 °C. The effect of detergent (Triton X-100), monovalent cations [Na+, K+ (0–300 mm)], divalent cations [Mg2+, Mn2+ (0–20 mm)], DMSO (0–10%), EDTA (0–20 mm) and reducing agent DTT (0–2 mm) on enzyme activity was evaluated by including them in the reaction mixture. To determine specificity of Ddn for F420H2, this component in the assay reaction was replaced with NADH and NADPH. For Ddn kinetics experiments, the initial velocities of F420H2 oxidation were plotted against the substrate concentration (PA-824 and its analogs) and analyzed using nonlinear regression to the Michaelis–Menten equation using graphpad prism 5 (GraphPad Software Inc., La Jolla, CA, USA). When nanomolar concentrations of Ddn were used in order to achieve steady state kinetics at lower concentrations of nitroimidazole substrates, detection of F420H2 oxidation was not feasible since the rate of Ddn catalyzed F420H2 oxidation was comparable to the spontaneous re-oxidation of F420H2 in the absence of the enzyme. Therefore 1 μm of Ddn was used in all reactions and apparent kinetic constants Vmax,app, Km,app and kcat/Km,app for any given reaction were determined from the plotted data.

Ddn kinetics studies via NO release assay

NO release was directly monitored using the Sievers NO analyzer 280i. The NO analyzer consists of an ozone generator, a chemiluminescent reaction chamber and a red-sensitive photomultiplier tube (PMT). In an aqueous reaction buffer, reactive nitrogen species react with dissolved O2 to form nitrite, which is reduced to NO by sodium iodide (1%) in acetic acid. The chemiluminescence assay is based on a gas phase reaction between NO and ozone to generate electronically excited nitrogen dioxide (NO2*) which emits light in the red and near-infrared and is detected by a thermoelectrically cooled, red-sensitive PMT. The reaction chamber containing glacial acetic acid–NaI with 100 μL of 1× antifoaming agent was purged with nitrogen continuously. The reaction chamber was connected to the PMT; standard 50 μL solutions of sodium nitrite (1 nm to 1 μm) were injected and a calibration curve was generated with nitrite concentration as a function of the PMT output (mV). Varying concentrations of nitroimidazole substrates or F420H2 were used with 1 μm of Ddn in a Ddn buffer. NO release profiles were monitored over a 10-min period injecting 50 μL of the reaction mixture at each time point. Apparent kinetic constants Vmax,app, Km,app and kcat/Km,app for each substrate were determined by fitting the initial rate of NO release as a function of substrate concentration using nonlinear regression fitted to the Michaelis–Menten equation.

Ddn binding studies to PA-824, F420 and F420H2

Fluorescence emission spectra of Ddn (excitation at 280 nm) in a 100 μL volume with or without ligand were recorded between 310 and 420 nm using a Tecan Infinite M1000 plate reader. Ddn, 1 μm, was incubated with increasing ligand concentrations for 30 min before fluorescence measurements. The extent of ligand binding to Ddn was determined by monitoring changes in maximal fluorescence emission at 334 nm. The dissociation constant Kd value ([ligand] at which 50% of the protein’s intrinsic tryptophan fluorescence is quenched) was calculated by fitting the relative change in intrinsic fluorescence at 334 nm (1 − F/F0) of Ddn versus ligand concentration to a one-site binding (hyperbolic model) using nonlinear regression, where F0 is the intrinsic intensity of fluorescence of 1 μm Ddn alone and F is the fluorescence at a given concentration of ligand; corrections were made to subtract ligand-associated fluorescence.

Modeling of PA-824 at the binding pocket

Putative binding modes of both stereoisomers of PA-824 in the binding pocket of Ddn were investigated using high resolution co-crystal structures of F420 with Ddn and with a homolog from Nocardia (nfa33440) [21a]. The pseudo-equatorial conformer of (S)-PA-824 was constructed by replacing the methyl of 7R-methyl-PA-824 [39] with a hydrogen, and then energy minimized in the gaseous phase with the density functional theory at the level of B3LYP/6-31G* [40]. The geometry of (R)-PA-824 was obtained by inversion of (S)-PA-824 coordinates. Docking of both (S)-PA-824 and (R)-PA-824 to the binding pocket was done manually with Quanta 2008 (Accelrys).


This work was funded, in part, by a grant from the Bill and Melinda Gates Foundation and the Wellcome Trust through the Grand Challenges in Global Health Program 11 to NITD through Imperial College London; by NITD and also in part by the intramural research program of NIAID (to CEB) and through the Center for Information Technology (to YSL). We thank Lacy Daniels for F420-2. The quantum chemical study utilized PC/LINUX clusters at the Center for Molecular Modeling of the NIH (