|Home | About | Journals | Submit | Contact Us | Français|
Bicyclic nitroimidazoles, including PA-824, are exciting candidates for the treatment of tuberculosis. These prodrugs require intracellular activation for their biological function. We found that Rv3547 is a deazaflavin-dependent nitroreductase (Ddn) that converts PA-824 into three primary metabolites; the major one is the corresponding des-nitroimidazole (des-nitro). When derivatives of PA-824 were used, the amount of des-nitro metabolite formed was highly correlated with anaerobic killing of Mycobacterium tuberculosis (Mtb). Des-nitro metabolite formation generated reactive nitrogen species, including nitric oxide (NO), which are the major effectors of the anaerobic activity of these compounds. Furthermore, NO scavengers protected the bacilli from the lethal effects of the drug. Thus, these compounds may act as intracellular NO donors and could augment a killing mechanism intrinsic to the innate immune system.
Drug-resistant tuberculosis (TB) has emerged as a major threat to global health (1). Promising new candidate drugs are the bicyclic nitroimidazoles, including PA-824 and OPC-67683, which are currently in human clinical trials (2, 3). These molecules are active not only against actively replicating bacteria, but also against bacteria that are nonreplicating by virtue of hypoxia (4). These nonreplicating cells are thought to be particularly difficult to eradicate and may be a major determinant of the extended treatment periods (6 to 8 months) necessary to cure a patient and to avoid disease relapse (5). In addition, such nonreplicating bacteria have been proposed to be responsible for latent tuberculosis, a condition that affects one-third of the entire human population (6).
The mechanism of cell killing by these prodrugs is complex. Treatment with PA-824 or OPC-67683 disrupts the formation of mycolic acids, major constituents of the cell envelope of Mtb (4, 7). However, this effect seemed unlikely to be responsible for cell killing under nonreplicating conditions, because the bacilli do not extensively remodel mycolic acids under anaerobic conditions (8). The deazaflavin cofactor F420 and the nonessential F420-dependent glucose-6-phosphate dehydrogenase Fgd1 are both essential for the activation of bicyclic nitroimidazoles, and loss of either confers high-level resistance to PA-824 (4, 9, 10). We recently identified a rare third class of mutants that had lesions in Rv3547, a member of a large family of proteins in Actinobacteria of unknown function (fig. S1) (11). We speculated that Rv3547 was a deazaflavin (F420)-dependent nitroreductase and renamed it Ddn (12).
To test this, purified F420 was reduced by the use of recombinant Mtb Fgd1 and glucose-6-phosphate (12). Recombinant Ddn was capable of reoxidizing this cofactor in a time- and enzyme-dependent fashion in the presence of PA-824 (Fig. 1, A and B). To test the specificity of this reaction, the opposite enantiomer of PA-824 (which has the R configuration at the 6-position and is about 1/100th as active against whole cells of Mtb) was tested and found to be a poor substrate for Ddn (Fig. 1B).
We next studied the chemical structures of the products of PA-824 reduction by Ddn. By titrating the enzyme with F420H2 and analyzing an aliquot by liquid chromatography—mass spectrometry (LC-MS), we characterized three reduction products (Fig. 1C). The major product had a molecular mass of 315 and was labeled metabolite 1. The formation of this metabolite was mostly complete after addition of a single equivalent of F420H2, as was formation of a minor peak showing a mass/charge ratio (m/z) of 331 (metabolite 2). Additional F420H2 equivalents increased mainly metabolite 3 (mass 291), and an unstable intermediate with a molecular mass of 346 (triangles in Fig. 1C). This intermediate (compound C in Figs. 1C and and2)2) was not stable enough to isolate, and it decomposed quickly and cleanly to metabolite 3. Comparing the metabolites produced from radioactive PA-824 in vitro by using purified Ddn with those produced in whole cells of Mtb revealed only slight differences in the proportion of products formed (Fig. 1D).
On the basis of these results and the previous report that the major metabolite formed from OPC-67683 by whole cells of Mtb was the des-nitro form of this compound (7), we proposed the structures for these metabolites shown in Fig. 2. The structures and retention times on high-performance liquid chromatography (HPLC) of all three compounds were confirmed by chemical synthesis and complete analytical characterization (fig. S2). Using synthetic material, we showed that neither the des-nitro nor either of the other two stable metabolites formed by Ddn-mediated reduction of PA-824 had any detectable activity against Mtb, which suggested that the bactericidal event occurred during the process of reduction.
Unlike flavoproteins that perform nitroreduction by single-electron chemistry, F420 is an obligate two-electron, or hydride, donor resembling the nicotinamide cofactors (13). To explore the chemical mechanism for this unusual reduction, we performed the analogous chemical reduction of PA-824 by sodium borohydride (fig. S2). Desnitro PA-824, although not the major product, was also produced by borohydride-mediated reduction of PA-824. We unambiguously assigned H-2 (d6.68 ppm) and H-3 (d6.48 ppm) of desnitro PA-824 by observing a nuclear Overhauser effect (NOE) between the two protons at C-5 on the oxazine ring (δ4.01 ppm) and the imidazole proton at C-3 (δ6.48 ppm) (fig. S3). By performing the reduction with sodium borohydride in deuterated methanol and sodium borodeuteride in normal methanol, the proton at the C-3 position of des-nitro PA-824 was shown to be derived from reductant, whereas the proton at the C-2 position was derived from solvent (fig. S4).
Thus, we hypothesized that des-nitro PA-824 was formed by hydride transfer from F420H2 to C-3 of the imidazole ring followed by protonation of the resulting nitronic acid to give A, which can eliminate nitrous acid (Fig. 2). This is also observed in bacterial enzymes involved in bioremediation of 2,4,6-trinitrotoluene (TNT), at least one of which has been shown to be F420-dependent (14, 15). An electrochemical study of the sequential single-electron reduction of PA-824 by a radical mechanism supports the novel reduction of the imidazole ring preferentially to the nitro group (16). The nitrous acid released would be unstable and quickly disproportionate into nitric oxide (NO) and other reactive nitrogen intermediates (17). This same nitronic acid intermediate (or its tautomer) could instead be quenched by water giving B, followed by elimination of hyponitrous acid (HNO) to give product 2, the biochemical equivalent of the Nef reaction that typically requires strongly acidic conditions (18). Biological Nef reactions have previously been limited to nitroalkane dioxygenases that utilize flavins and molecular oxygen to convert an alkyl nitro group to a carbonyl (19). Although this is a minor product with PA-824, certain bicyclic nitroimidazoles (e.g., compound 7 in Table 1) show this as the major product. Further reduction of the nitronic acid intermediate produces the aromatic hydroxylamine metabolite C, which, upon further reduction and fragmentation, produces 3 (fig. S5) (20).
To test the hypothesis that reduction of PA-824 by Ddn resulted in the release of reactive nitrogen species, we used the Griess reagent to directly detect nitrous acid. A PA-824—dependent production of HNO2 by Ddn was consistently observed in the presence of F420H2 (Fig. 3A). NO could also be detected in PA-824—treated whole cells, under both aerobic (Fig. 3B) and hypoxic nonreplicating conditions (fig. S6), by preloading cells with diaminofluorescein (DAF-FM) diacetate, a probe that fluoresces specifically in the presence of NO (21, 22).
To correlate NO release with anaerobic killing, we selected a series of eight derivatives that had a broad range (from 2 to 125 μM, Table 1) of minimum anaerobicidal concentration (MAC) values (the aerobic potency of these compounds ranges from 0.04 to 6 μM). These derivatives revealed a strong correlation between the amount of the corresponding des-nitro product formed and anaerobic killing activity (Fig. 3C). Des-nitro formation did not correlate with aerobic activity, nor did any other metabolite correlate with either aerobic or anaerobic activity (fig. S7). The rate of NO release [an apparent Vmax/Km as RFU per hour divided by substrate affinity (the Michaelis constant) measured intracellularly] also correlated with anaerobic killing activity (Table 1). Finally, the effect of PA-824 on cells could be abrogated by a scavenger of NO, C-PTIO, which showed a dose-dependent decrease in NO production (Fig. 3D) and partially rescued hypoxic cells from killing by PA-824 (Fig. 3E). This effect was muted by both toxicity and poor cell penetrating ability of C-PTIO. Mtb mutants hypersensitive to NO, such as proteasome mutants, also display increased anaerobic sensitivity to PA-824 (fig. S8). The simplest interpretation of these data is that reactive nitrogen species are dominant in anaerobic activity but under aerobic conditions this is augmented by other effects.
Reactive nitrogen species play a major role in mammalian defense against mycobacterial infections, and mice deficient in the ability to produce NO are hypersusceptible to infection with Mtb (23). PA-824—resistant mutants in F420 biosynthesis have also been shown to be hypersensitive to killing by NO (24). Nitrite and NO both react with cytochromes and cytochrome c oxidase to interfere with electron flow and to cripple the coupling of respiration to reduction of oxygen. Thus, bicyclic nitroimidazoles may function as highly specific NO donors requiring activation by a powerful deazaflavin cofactor (F420) that is unique to a small group of microorganisms and extremophiles, which makes cross-activation by mammalian enzymes highly unlikely. This unusual mechanism for nitroimidazole reduction avoids the single-electron—reduced nitrogen species typical of oxygen-sensitive, flavin-dependent enzymes and defines the chemistry of a class of F420-dependent enzymatic nitroreductases.
NO-donating drugs have been shown to have potential for a wide variety of human diseases including cardiovascular disease, asthma, hypoxicischemic brain injury, glaucoma, and Alzheimer's disease (25). Indeed, it may be possible to more generally co-opt bacterial nitroreductases to reduce other nitro-containing heterocycles as an avenue for developing anti-infectives. It is noteworthy that the two candidate molecules in phase 2 trials (PA-824 and OPC-67683) were both optimized only for whole-cell aerobic activity and, as a result, have fairly modest anaerobic activity. Thus, these results have the potential to enable structure-based design of a drug specific for the treatment of latent tuberculosis.
We thank C. Nathan (Cornell University) and T. P. Begley (Cornell University) for insightful comments, S. Ehrt (Cornell University) for proteosome mutants, L. Daniels (Texas A&M University) for F420 and M. Goodwin (Tuberculosis Research Section) for analytical support. This work was funded (in part) by the intramural research program of National Institute of Allergy and Infectious Diseases, NIH, and (in part) by a grant from the Bill and Melinda Gates Foundation and the Wellcome Trust through the Grand Challenges in Global Health Initiative. This work was also supported by the Korea Foundation for International Cooperation of Science and Technology (KICOS) through a grant provided by the Korean Ministry of Education, Science and Technology (MEST) (No. K20501000001).
Supporting Online Material www.sciencemag.org/cgi/content/full/322/5906/1392/DC1