Single- and Two-Electron Reduction of Nitroaromatic Compounds by Flavoenzymes: Mechanisms and Implications for Cytotoxicity

Nitroaromatic compounds (ArNO2) maintain their importance in relation to industrial processes, environmental pollution, and pharmaceutical application. The manifestation of toxicity/therapeutic action of nitroaromatics may involve their single- or two-electron reduction performed by various flavoenzymes and/or their physiological redox partners, metalloproteins. The pivotal and still incompletely resolved questions in this area are the identification and characterization of the specific enzymes that are involved in the bioreduction of ArNO2 and the establishment of their contribution to cytotoxic/therapeutic action of nitroaromatics. This review addresses the following topics: (i) the intrinsic redox properties of ArNO2, in particular, the energetics of their single- and two-electron reduction in aqueous medium; (ii) the mechanisms and structure-activity relationships of reduction in ArNO2 by flavoenzymes of different groups, dehydrogenases-electrontransferases (NADPH:cytochrome P-450 reductase, ferredoxin:NADP(H) oxidoreductase and their analogs), mammalian NAD(P)H:quinone oxidoreductase, bacterial nitroreductases, and disulfide reductases of different origin (glutathione, trypanothione, and thioredoxin reductases, lipoamide dehydrogenase), and (iii) the relationships between the enzymatic reactivity of compounds and their activity in mammalian cells, bacteria, and parasites.


Introduction
Over the decades, nitroaromatic compounds (ArNO 2 ) maintain their importance in relation to industrial processes, environmental pollution, and pharmaceutical application. Current estimates have their production, that is, the synthesis of pigments, polymers, pesticides, explosives, or pharmaceuticals, up to 10 8 tons per year ( [1][2][3][4][5][6], and references therein). Because of contamination of groundwater and soil at military and industrial sites by ArNO 2 that exhibit toxic, mutagenic, and cancerogenic activities, there has been a significant increase in research to understand and apply biological processes for their degradation.
On the other hand, the electron-attracting ability and redox activity make the nitro group a versatile and unique group in medicinal chemistry. Nitroaromatic compounds have a long history of use as antibacterial and antiparasitic drugs and their application as radiosensitizers and hypoxia-selective anticancer agents ( [6], and references therein) (Figures 1 and 2). The resurgence of interest in their use is caused by the reevaluation of the problems with their mutagenicity and the new potential fields of their application, e.g., the treatment of oxic tumors, including the development of antibody-or gene-directed therapies employing bacterial nitroreductases [7,8].
Alternatively, the values of E 1 of the number of nitroaromatics were determined by cyclic voltammetry. Typically, the electrochemical reduction in ArNO 2 in aqueous media proceeds irreversibly with the formation of ArNHOH. However, this process takes place in two steps, with the pH-independent transfer of the first electron and pH-dependent transfer of three electrons: At pH 11-12, the redox potential of a second step may become more negative than the potential of ArNO 2 /ArNO 2 − · couple. In this case, a separate reversible process of singleelectron transfer at E m = E 1 7 is observed in cyclic voltammetry ( [31], and references therein). There is some interest in the prediction of E 1 7 of ArNO 2 from quantum mechanical calculations or the use of substitute descriptors such as the electron affinities of ArNO 2 or the heats of formation (∆Hf) of ArNO 2 − . However, the calculations in vacuo frequently do not provide reliable predicted E 1 7 values due to the large data scattering or may be confined only to a series of homologous compounds [32][33][34][35]. Some improvement may be expected upon the introduction of solvent matrix effects into calculations [36]. On the other hand, the E 1 7 values of ArNO 2 may be predicted from linear log (rate constant) vs. E 1 7 relationships in single-electron reduction in nitroaromatics by flavoenzymes dehydrogenases-electrontransferases or their redox partners, FeS proteins [11][12][13]. The use of the geometric average of rate constants obtained in several enzymatic systems improves the prediction accuracy. The calculated reduction potentials (E 1 7(calc.) , Table A1, Appendix A) deviate from the experimental ones by no more than 40 mV (standard devia-tion, ± 18 mV) and thus should be considered as realistic. Importantly, this approach may be applied for groups of structurally diverse ArNO 2 .
The redox properties of ArNO 2 multielectron reduction products are insufficiently characterized in quantitative terms. In aqueous medium, ArNO 2 are electrochemically reduced into ArNHOH directly, bypassing ArNO (Equations (1) and (2)). On the other hand, the use of mixed ethanol-aqueous solution with pH 1.0-4.0 enabled the detection of reduction intermediate dihydroxylamine (ArN(OH) 2 ), which further undergoes dehydration [38]: However, the voltammetric characteristics of this reaction could not be extrapolated into aqueous medium with pH 7.0. Following this approach, it was assumed that the ratelimiting step of enzymatic two-electron reduction in ArNO 2 is a net hydride transfer with the formation of ArN(OH)O − [39]. The calculated heats of formation (∆Hf(ArN(OH)O − ) increase with the electron-accepting potency of substituents and roughly correlate with ∆Hf(ArNO 2 − ) and E 1 7 of nitroaromatics. In aqueous medium, nitrosobenzene is reversibly electrochemically reduced into phenylhydroxylamine at E 0 7 = 0.184 V [40]. It is suggested that the intermediate free radical ArNOH· is unstable, and that the redox potentials of first and second electron transfer are separated by −0.5 V. Because ArNO are more powerful oxidants than ArNO 2 , they can be directly reduced by NAD(P)H, GSH, ascorbate, and other reductants. For example, nitrosobenzene is reduced by NADPH and ascorbate with k = 124 M −1 s −1 and 2.8 × 10 3 M −1 s −1 , respectively [41,42]. The reactivity of nitrosobenzenes increases with their electron-accepting properties. The reactions of ArNO with GSH proceed with The energetics of formation of ArNH2 from ArNHOH in aqueous medium is also incompletely characterized. At metal or carbonaceous electrodes, nitrobenzene is reduced into phenylhydroxylamine with half-wave potentials (E1/2) of −0.30-−0.45 V, whereas the latter is reduced into aniline with E1/2 = −0.55-−0.70 V at pH 7.0 [52]. However, the overvoltage of this reaction depends on the electrode material [53]. On the other hand, the second electron transfer in this reaction, reduction in aniline radical into aniline, is characterized by E 1 = 1.03 V at pH 6.9 [10]. Thus, in spite of the uncertain value of E 0 7 of phenylhydroxylamine/aniline redox couple, it is clear that the reduction in phenylhydroxylamine into aniline radical should proceed at very negative potential. This may impose certain barriers toward the enzymatic formation of ArNH2 from ArNHOH, in particular, single-electron transfer steps.

Mechanisms of Reduction in Nitroaromatic Compounds by Flavoenzymes
An early study of nonenzymatic reduction in nitroaromatics by reduced FMN under anaerobic conditions demonstrated a linear dependence of log k on E 1 7 of ArNO2 [54]. Its extrapolation to ∆E 1 7 = 0 gives k ~ 10 7 M −1 s −1 , which agrees with an "outer-sphere" electron transfer model (Appendix B). The products of the reduction in nitroaromatics were Scheme 1. Fragmentation of reduction products of nitrobenzyl carbamates (a), nitrobenzyl sulfonates (b), and nitrobenzyl phosphoramide mustards (c). Adapted from the work of [47][48][49][50][51].
The energetics of formation of ArNH 2 from ArNHOH in aqueous medium is also incompletely characterized. At metal or carbonaceous electrodes, nitrobenzene is reduced into phenylhydroxylamine with half-wave potentials (E 1/2 ) of −0.30-−0.45 V, whereas the latter is reduced into aniline with E 1/2 = −0.55-−0.70 V at pH 7.0 [52]. However, the overvoltage of this reaction depends on the electrode material [53]. On the other hand, the second electron transfer in this reaction, reduction in aniline radical into aniline, is characterized by E 1 = 1.03 V at pH 6.9 [10]. Thus, in spite of the uncertain value of E 0 7 of phenylhydroxylamine/aniline redox couple, it is clear that the reduction in phenylhydroxylamine into aniline radical should proceed at very negative potential. This may impose certain barriers toward the enzymatic formation of ArNH 2 from ArNHOH, in particular, single-electron transfer steps.

Mechanisms of Reduction in Nitroaromatic Compounds by Flavoenzymes
An early study of nonenzymatic reduction in nitroaromatics by reduced FMN under anaerobic conditions demonstrated a linear dependence of log k on E 1 7 of ArNO 2 [54]. Its extrapolation to ∆E 1 7 = 0 gives k~10 7 M −1 s −1 , which agrees with an "outer-sphere" electron transfer model (Appendix B). The products of the reduction in nitroaromatics were hydroxylamines. Since that time, a substantial amount of information accumulated in this area, evidencing the diversity of reaction mechanisms, which will be analyzed in subsequent subsections.

Single-and Mixed Single-and Two-Electron Reduction in Nitroaromatic Compounds by Flavoenzymes Dehydrogenases-Electrontransferases
Flavoenzymes dehydrogenases-electrontransferases transform two-electron (hydride) transfer into a single-electron one, and, most frequently, possess single-electron transferring redox partner, heme-or FeS-containing protein. Their action is characterized by the formation of neutral (blue) flavin semiquinone, FMNH or FADH·, as a reaction intermediate. In this section, the properties of flavohemoenzymes or heme-reducing flavoenzymes and flavoenzymes FeS reductases are discussed separately. This is related not to the different properties or action mechanisms of their flavin cofactors but to the different roles of the heme or FeS redox centers in the reduction in nitroaromatics.
NADPH: cytochrome P-450 reductase (P-450R) is a 78 kD enzyme associated with the endoplasmic reticulum of a variety of eukaryotic cells. It is responsible for electron transfer from NADPH to the cytochromes P-450 and to other microsomal enzyme systems ( [55], and references therein). Rat liver P-450R has a hydrophobic 6 kD N-terminal membranebinding domain, the FMN-binding domain next to it, the connecting domain, and the FAD-and NADPH-binding domains at the C-terminal side [56]. In catalysis, the transfer of redox equivalents follows the pathway NADPH → FAD → FMN → cytochrome P-450 (cytochrome c). The pyrimidine part of the isoalloxazine ring of FMN is accessible to solvent. The edge-to-edge distance between the isoalloxazine rings of FMN and FAD is 3.5-4.5 Å. The negatively charged FMN-binding domain and hydrophobic membranebinding domain are involved in the complex formation between P-450R and cytochromes P-450 or cytochrome c. The potentiometric and kinetic characteristics of P-450R are presented in Table 1. The large differences between the redox potential values for the first and second electron transfer point to the high stability of flavin semiquinones. A catalytic cycle involving one-, two-, and three-electron-reduced states of reductase with FMNH 2 acting as the principal donor of electrons to oxidants is proposed [57]. It is suggested that the rates of interflavin electron transfer are higher or similar to the rate of FAD reduction by NADPH [58]. Cytochromes P-450 reoxidize P-450R with low rates, 0.15-1.0 s −1 [59]. P-450R reduces nitroaromatic compounds in a single-electron way and is the enzyme used most frequently to demonstrate their redox cycling reactions. For this reaction, the linear dependence of logarithms of reaction rate or k cat /K m on E 1 7 of ArNO 2 was observed [68,69]. This is in line with an "outer-sphere" electron transfer model [28]. Moreover, ArNO 2 are systematically less reactive than quinones with the same E 1 7 values [69][70][71]. This additionally supports an "outer-sphere" reaction model because the electron selfexchange constants for ArNO 2 are by two orders of magnitude lower than those for quinones [25,26,29]. Subsequently, the linear log (k cat /K m ) vs. E 1 7 relationships were used for the estimation of unknown E 1 7 values of nitroaromatics [11][12][13]. The redox partners of P-450R, cytochromes P-450, catalyze the denitration of ArNO 2 with the formation of corresponding hydroxyl derivatives [72,73] and reverse the accumulation of the amine products of polinitrobenzene reduction, catalyzing the formation of their hydroxylamines [74]. On the other hand, cytochrome P-450 101A1 is able to reduce m-nitroacetophenone into a corresponding amine [75]. However, the impact of these reactions on the cytotoxicity of ArNO 2 has not been studied in detail.
Nitric oxide synthases (NOS) are dimeric flavohemoproteins that catalyze the conversion of L-arginine to citrulline and nitric oxide (NO·) at the expense of NADPH. Each monomer of NOS consists of a heme domain with tetrahydrobiopterin bound at its Nterminus and a FAD-and FMN-containing reductase domain at its C-terminus. The reductase and by domains are linked by a calmodulin (CAM)-binding sequence ( [76], and references therein). The reductase domain is highly similar to P-450R. In catalysis, the redox equivalents are transferred in the pathway NADPH → FAD → FMN → heme (tetrahydrobiopterin), involving mainly one-and three-electron reduced states of the reductase domain in the turnover ( [77,78], and references therein). The potentiometric and kinetic characteristics of neuronal NOS are presented in Table 1. Although these data may vary in various publications, it is accepted that the redox potential of cofactors decreases in the order FMN/FMNH·>> FMNH·/FMNH 2 ≥ FAD/FADH > FADH·/FADH 2 ≥ heme, and that the flavins are reduced much faster than heme. NOS reduces quinones and ArNO 2 in a single-electron way via FMNH 2 [79][80][81]. The reasons why the heme moiety is not involved in the reduction process are unclear. It may be partly explained by an increase in its redox potential after the binding of ligands. Under anaerobic conditions, nilutamide (15) and CB-1954 (14) are reduced into corresponding hydroxylamines [80,81]. TNT (4) and dinitrobenzenes inhibit the formation of NO· by NOS [82,83], most possibly by trapping of NO· with the product of their redox cycling, O 2 − , resulting in the formation of peroxynitrite. An alternative mechanism is the diversion of electron flux from FMNH 2 to heme. Like in P-450R-catalyzed reactions, the reactivity of ArNO 2 was characterized by the linear log k cat /K m vs. E 1 7 relationship with some possible discrimination against the negatively charged compounds [64]. ArNO 2 were less reactive than quinones with the same E 1 7 values. Flavohemoglobins (FHbs) have been found in a wide variety of bacteria and fungi and play a key role in their resistance to nitrosative stress. They consist of an N-terminal hemebinding domain and of C-terminal FAD-and NAD(H)-binding modules. During turnover, NADH reduces FAD, which further reduces the Fe 3+ form of hemoglobin (HbFe 3+ ); oxyhemoglobin (HbFe 2+ O 2 ) is finally formed under aerobic conditions. The reaction of the HbFe 2+ O 2 with NO· leads to NO· detoxification, i.e., the formation of nitrate instead of the toxic peroxynitrite (ONOO − ). The reactions proceed with a high turnover rate, ca. 100 s −1 [84]. The crystal structures of FHb from various sources show that the pyrimidine ring of the FAD isoalloxazine is partly accessible to solvent, whereas the access to heme may be partly hampered by a bound phospholipid molecule 66 [85]. The potentiometric and kinetic characteristics of FHb are given in Table 1. The steady-state reduction in quinones or ArNO 2 by S. aureus FHb follows a "ping-pong" mechanism with the oxidative half-reaction as a rate-limiting catalysis step [67]. During the turnover in the presence of an oxidant, the reduced FAD is reoxidized by >10 times more rapidly than HbFe 2+ O 2 moiety, i.e., it acts as a preferred electron donor. The reoxidation of heme may be hampered by a bound phospholipid molecule; moreover, the binding of O 2 may significantly increase the potential of Fe 3+ /Fe 2+ O 2 couple. The log k cat /K m of nitrobenzenes and nitrofurans displays a well-expressed parabolic dependence on their E 1 7 . In contrast to reactions of other electrontransferases, TNT and p-nitrobenzaldehyde were reduced with a 35-40% single-electron flux. The mixed character of reduction is possibly determined by the relatively low stability of the FAD semiquinone state of FHb, 15% at equilibrium [65].
Flavoenzymes FeS reductases either have the FeS proteins as redox partners or perform the intramolecular electron transfer to a FeS redox center. Some of their representatives may transfer electrons in both directions to and from FeS centers. Typically, both FeS reductases and their redox partners participate in the reduction in ArNO 2 and other groups of redox-active xenobiotics (Scheme 2): NADPH:adrenodoxin reductase (ADR) is a monomeric 51 kD FAD-containing enzyme, first isolated from bovine adrenal cortex mitochondria. It reduces 13.3 kD Fe2S2 protein adrenodoxin (ADX), which in turn reduces mitochondrial cytochromes P-450 that participate in the biosynthesis of steroid hormones. The ADR-ADX complex formation is determined by the electrostatic interaction between the positively charged amino acid residues of ADR and negatively charged residues of ADX ( [86], and references therein). The potentiometric and kinetic characteristics of ADR and ADX are presented in Table 2   Table 2 NADPH:adrenodoxin reductase (ADR) is a monomeric 51 kD FAD-containing enzyme, first isolated from bovine adrenal cortex mitochondria. It reduces 13.3 kD Fe 2 S 2 protein adrenodoxin (ADX), which in turn reduces mitochondrial cytochromes P-450 that participate in the biosynthesis of steroid hormones. The ADR-ADX complex formation is determined by the electrostatic interaction between the positively charged amino acid residues of ADR and negatively charged residues of ADX ( [86], and references therein). The potentiometric and kinetic characteristics of ADR and ADX are presented in Table 2.   [95] 3500 e − /s (NADH to FMN (ferricyanide), steady-state) [96,97]; 150-380 s −1 (NADH to ubiquinone, steady-state, proteoliposomes) [98] NADPH reduces ADR with the final formation of the FADH 2 -NADP + charge-transfer complex (K d~1 0 −8 M). NADP + also binds to the oxidized and semiquinone form of ADR but with a lower affinity. NADPH binds to the FAD semiquinone more efficiently than NADP + [99].
The steady-state reactions of quinone or ArNO 2 reduction by ADR proceed with k cat = 20-25 s −1 , which is close to the maximal rate of enzyme reduction by NADPH [100]. However, NADPH acts as a strong (K i = 5.0-6.0 µM) inhibitor with respect to oxidants. Calculated by extrapolation to [NADPH] = 0, k cat /K m values for the reduction in nitrofurans by ADR are low, 1.0 ÷ 4.0 × 10 3 M −1 s −1 [101]. It was concluded that quinones and ArNO 2 oxidize free enzyme form and its complexes with NADPH and NADP + , although with slower rates. Most probably, the rate-limiting step of this reaction is the oxidation of FAD semiquinone. On the other hand, ADX stimulates the reduction in ArNO 2 by ADR, eliminating the inhibition by NADPH and providing a more efficient alternative reduction pathway via reduced ADX [100,101]. The k cat /K m values of nitroaromatics are much higher than in the oxidation of ADR. Their logarithms increase with their E 1 7 ; moreover, ArNO 2 are less reactive than quinones with the same E 1 7 values [70,71]. Mycobacterium tuberculosis contains ADX-type 50 kD NADPH-feredoxin reductase FprA [102]. The E 0 7 of FAD is equal to −0.235 V [103]. The rate of FAD reduction by NADPH exceeds 160 s −1 ; however, the reaction analysis is complicated by the enzyme reoxidation by NADP + . Blue FAD semiquinone is formed during the reduction in the enzyme by NADPH or during its photoreduction in the presence of NADP + . The enzyme reduces M. smegmatis Fe 7 S 7 ferredoxin with k cat = 3.4 s −1 [102]. However, the data on the reactions of FprA and its redox partner(s) with ArNO 2 are absent.
Most ferredoxin:NADP(H) oxidoreductases (FNRs) are 35-36 kD monomeric FADcontaining enzymes that transfer electrons from reduced FeS-protein ferredoxin (Fd) to NADP + , thus providing NADPH for CO 2 assimilation in plants and cyanobacteria. However, some FNRs act in the opposite direction, reducing Fd at the expense of NADPH, thus supplying reduced Fd for nitrate assimilation (roots) or biosynthesis of isoprenoids (malaria parasite Plasmodium falciparum).
In this part, we discuss the mechanisms of ArNO 2 reduction by enzymes that function in opposite directions, FNR from cyanobacterium Anabaena PCC7119 and from P. falciparum (Pf FNR). Their potentiometric and kinetic characteristics are presented in Table 2. The isoalloxazine ring of FAD of both FNRs is partly exposed to solvent [104][105][106]. Electrostatic interaction between the negatively charged Fd and positively charged FNR and hydrophobic interactions are involved in the formation of the Anabaena FNR-Fd complex [89]. An analogous electrostatic complex is formed between Pf FNR and Pf Fd [105]. Both enzymes reduce quinones and ArNO 2 in a single-electron way [107][108][109]. Reactions follow the "ping-pong" mechanism with the oxidative half-reaction as a partly rate-limiting step, 100 s −1 for Anabaena FNR, and up to 63 s −1 for Pf FNR. They are characterized by parabolic dependences of log k cat /K m of quinones and nitroaromatics on their E 1 7 value. Again, ArNO 2 comprised a separate series of compounds with lower reactivity. In both enzymes, the rate-limiting step in FADH − reoxidation is the oxidation of FADH·. Like in reactions of ADR, Fd stimulated the reduction in nonphysiological oxidants by both FNRs, providing an alternative reduction pathway with lower k cat and higher k cat /K m of oxidants. In another study, several low-potential nitroimidazoles were found to oxidize reduced Anabaena Fd with k = 630 ÷ 3500 M −1 s −1 [110], which was much faster than the reduction in their analogs by FNR alone [102]. Pf Fd also stimulated the reduction in quinones and nitroaromatics by Pf FNR, providing the faster pathway of their reduction with k cat equal to the rate of Pf FNR-Pf Fd electron transfer. Like spinach FNR, both Anabaena FNR and Pf FNR catalyze the reductive denitration of tetryl (2) into N-methylpicramide (3) accompanied by redox cycling of free radical intermediate [111] The above data indicate that although PfFNR possesses low homology, 20%-30%, with plant FNRs and functions in the opposite direction, its nitroreductase reaction mechanism and absence of substrate specificity are the same as of Anabaena FNR. However, because Fds possess higher nitroreduction rates, an important but insufficiently studied problem is their possible nitroaromatic substrate specificity.
NADH:ubiquinone reductase (CoQR, complex I) is localized in the inner mitochondrial membrane. It is a large (1 MD) enzyme of 45 subunits, catalyzing NADH oxidation by ubiquinone and performing transmembrane proton translocation. The data of numerous studies of bovine mitochondrial complex I [94,98,[112][113][114] may be summarized as follows: (i) FMN and 8 FeS clusters N1-6, separated by 7.6-14 Å distances, are localized in the hydrophilic arm of L-shaped complex I that extends into the mitochondrial matrix. FMN is located in 51 kD subunit; (ii) the hydrophobic domain of the complex, localized in the inner mitochondrial membrane, pumps 4 protons from the matrix to intermembrane mitochondrial space per molecule of NADH oxidized; and (iii) the sequence of transfer of redox equivalents is NADH The reduction in ubiquinone is inhibited by tightly binding rotenone and piericidin. The potentiometric and kinetic characteristics of bovine complex I are presented in Table 3.
The mechanism of reduction in soluble quinones, nitroaromatics, and other artificial electron acceptors by CoQR is still a matter of debate. In this context, the reference reaction is its reduction in ferricyanide, where ferricyanide presumably directly oxidizes reduced FMN [96,97]. This reaction proceeds according to a "ping-pong" mechanism with double competitive substrate inhibition, which shows that both substrates compete for the same binding site in reduced and oxidized enzyme form. The use of 4-S− 2 H-NADH decreases kcat of reaction and kcat/Km of NADH by two times, which shows that the rate-limiting step of the process is the reduction in FMN by NADH. The reduction in soluble quinones and nitroaromatics by complex I is insensitive to rotenone and is characterized by a common parabolic dependence of log kcat/Km on E 1 7 of oxidants. For the most active oxidants, kcat of reaction reaches 100 s −1 . Importantly, ArNO2 are reduced in a mixed single-and twoelectron way with a single-electron flux of 45%-70%. The studies of the complex I inhibition by NADH, NAD + , redox inactive ADP-ribose, and slowly reacting quinones enabled us to conclude that quinones and nitroaromatics bind close to the ferricyanide binding site [115][116][117]. In contrast to ferricyanide, they may oxidize both free enzyme form and its complexes with NADH (Kd = 3.0 µM) and NAD + (Kd = 30-60 µM), although with slower rates. Because the above Kd differ from NADH and NAD + inhibition constants toward ferricyanide, it is possible that ferricyanide and quinones or ArNO2 oxidize different redox states of the enzyme. The possible involvement of FeS centers in nitroreduction warrants further studies.
Among the similar redox systems that may contribute to cytotoxic/therapeutic action of ArNO2, Trichomonas vaginalis contain a partly characterized Fd-dependent system. T. vaginalis ferredoxin (E 1 7 = −0.347 V) plays a central role in hydrogenosomal electron The above data indicate that although Pf FNR possesses low homology, 20-30%, with plant FNRs and functions in the opposite direction, its nitroreductase reaction mechanism and absence of substrate specificity are the same as of Anabaena FNR. However, because Fds possess higher nitroreduction rates, an important but insufficiently studied problem is their possible nitroaromatic substrate specificity.
NADH:ubiquinone reductase (CoQR, complex I) is localized in the inner mitochondrial membrane. It is a large (1 MD) enzyme of 45 subunits, catalyzing NADH oxidation by ubiquinone and performing transmembrane proton translocation. The data of numerous studies of bovine mitochondrial complex I [94,98,[112][113][114] may be summarized as follows: (i) FMN and 8 FeS clusters N1-6, separated by 7.6-14 Å distances, are localized in the hydrophilic arm of L-shaped complex I that extends into the mitochondrial matrix. FMN is located in 51 kD subunit; (ii) the hydrophobic domain of the complex, localized in the inner mitochondrial membrane, pumps 4 protons from the matrix to intermembrane mitochondrial space per molecule of NADH oxidized; and (iii) the sequence of transfer of redox equivalents is NADH The reduction in ubiquinone is inhibited by tightly binding rotenone and piericidin. The potentiometric and kinetic characteristics of bovine complex I are presented in Table 3.
The mechanism of reduction in soluble quinones, nitroaromatics, and other artificial electron acceptors by CoQR is still a matter of debate. In this context, the reference reaction is its reduction in ferricyanide, where ferricyanide presumably directly oxidizes reduced FMN [96,97]. This reaction proceeds according to a "ping-pong" mechanism with double competitive substrate inhibition, which shows that both substrates compete for the same binding site in reduced and oxidized enzyme form. The use of 4-S− 2 H-NADH decreases k cat of reaction and k cat /K m of NADH by two times, which shows that the rate-limiting step of the process is the reduction in FMN by NADH. The reduction in soluble quinones and nitroaromatics by complex I is insensitive to rotenone and is characterized by a common parabolic dependence of log k cat /K m on E 1 7 of oxidants. For the most active oxidants, k cat of reaction reaches 100 s −1 . Importantly, ArNO 2 are reduced in a mixed single-and twoelectron way with a single-electron flux of 45-70%. The studies of the complex I inhibition by NADH, NAD + , redox inactive ADP-ribose, and slowly reacting quinones enabled us to conclude that quinones and nitroaromatics bind close to the ferricyanide binding site [115][116][117]. In contrast to ferricyanide, they may oxidize both free enzyme form and its complexes with NADH (K d = 3.0 µM) and NAD + (K d = 30-60 µM), although with slower rates. Because the above K d differ from NADH and NAD + inhibition constants toward ferricyanide, it is possible that ferricyanide and quinones or ArNO 2 oxidize different redox states of the enzyme. The possible involvement of FeS centers in nitroreduction warrants further studies.
Among the similar redox systems that may contribute to cytotoxic/therapeutic action of ArNO 2 , Trichomonas vaginalis contain a partly characterized Fd-dependent system. T. vaginalis ferredoxin (E 1 7 = −0.347 V) plays a central role in hydrogenosomal electron transport, reversibly transferring electrons from pyruvate:ferredoxin oxidoreductase (PFOR) to hydrogenase or to the NADH dehydrogenase module that contains FMN in 51 kD subunit, and Fe 2 S 2 cluster in 24 kD subunit (FOR) [118][119][120]. Hypothetically, FOR can reduce nitroaromatics; however, the data on its nitroreductase reactions are absent. On the other hand, using the hydrogenosomal extracts of T. vaginalis, PFOR catalyzed pyruvatedependent reduction in a series of ArNO 2 (E 1 7 = −0.564 V-−0.243 V) under anaerobic conditions [121]. At fixed compound concentration, a linear log (reduction rate) vs. E 1 7 relationship is observed. T. vaginalis Fd stimulated the reduction in ArNO 2 ; however, the reaction rate almost did not depend on E 1 7 . Furthermore, it has been shown that T. vaginalis Fd reduces low-potential metronidazole (40) and other nitroimidazoles with an unexpectedly high rate, k = 4.2 × 10 5 ÷ 1.0 × 10 6 M −1 s −1 [110]. On the other hand, metronidazole and another low-potential compound, chloramphenicol (23), are also rapidly reduced by another NADH oxidizing 26 kD FMN and FeS-containing protein, with k cat = 56 s −1 and k cat /K m = 2.0 × 10 6 M −1 s −1 , and k cat = 130 s −1 and k cat /K m = 1.7 × 10 6 M −1 s −1 , respectively [122]. The functions of this protein are unknown.
Microaerophilic bacterium Helicobacter pylori contains a similar partly characterized system, consisting of PFOR and flavodoxin:quinone oxidoreductase (FqrB) [123]. The electrons between these flavoenzymes are reversibly transferred by a low-potential electron carrier flavoprotein flavodoxin. Importantly, the reduction in NADP + by FqrB was inhibited by nitrothiazole nitazoxanide (52) and a number of nitrochromanes, nitroben-zenes, and nitrobenzoxadiazoles, which were binding to flavodoxin [124]. The system consisting of PFOR, ferredoxin:NAD + reductase, and ferredoxin, the latter participating in ArNO 2 reduction, is also present in Giardia spp. [125]. However, the catalytic properties of its components are insufficiently characterized.
Mammalian xanthine oxidase (XOD) attracted some attention as a model system for the single-electron reduction in ArNO 2 . The reactions with nitroimidazoles [126] and nitroacridines [127] were characterized by the absence of structure specificity, i.e., an increase in log (reaction rate) with E 1 7 of oxidants. However, one may note that XOD is a product of proteolysis of native NAD + -reducing xanthine dehydrogenase (XDH) under a variety of pathophysiological conditions ( [128], and references therein). While XDH prevails intracellularly, XOD is prevalent in body fluids such as milk and plasma, where it may be secreted or released from dead cells. XDH is a 2 × 145 kD dimer, with each subunit containing molybdopterin cofactor, FAD, and two Fe 2 S 2 clusters. During the catalysis, electrons are transferred from the purine substrate to molybdopterin, then to FAD via FeS clusters, and ultimately to the final electron acceptor, NAD + (XDH) or O 2 (XOD). The rate-limiting catalysis step is the reductive half-reaction [129]. Partly purified XDH under aerobic conditions reduces nitrofurazone into several products, including its amino metabolite [130]. The fractions of XDH and XOD in the cytosol under anaerobiosis reduced 1-and 2-nitropyrenes and 4-nitrobiphenyl into their amino metabolites [131]. However, the studies of nitroreductase reactions of XDH did not receive further attention.
Summing up, the single-electron reduction in ArNO 2 by P-450, NOS, and FNR may be attributed to the high stability of their flavin semiquinone state. Evidently, the reaction follows an "outer-sphere" electron transfer mechanism. The distances of electron transfer (R p ) calculated according to this model (Appendix B, Equation (A3)), are equal to 4.2 (P-450R), 3.9 (nNOS), 4.4 (Anabaena FNR), 4.9-5.6 (Pf FNR) [109], and 2.1-3.7 Å (bovine ADX) [71,101]. These orientational values are consistent with the partial exposure of their redox centers to solvent. In all these cases, however, the principal factor determining the reactivity of ArNO 2 is their E 1 7 . This leaves relatively little space for the improvement of the enzymatic reactivity of compounds. The reasons for the mixed single-and two-electron way of ArNO 2 reduction by CoQR and FHb are unclear. Because of the limited amount of data, the factors determining nitroaromatic oxidant specificity for the single-electron transferring flavoenzymes of M. tuberculosis, T. vaginalis, H. pylori, and Giardia spp. are unclear. On the other hand, the application of Equation (A3) in the analysis of reactions of T. vaginalis Fd [110] gives R p ≤ 1.5 Å, which points to strong electronic coupling and deviation from an "outer-sphere" electron transfer model. This is in accordance with the possible binding of ArNO 2 at the unique cavity near the FeS cluster of T. vaginalis Fd ( [110], and references therein) and points to the possible substrate structure specificity.

Two-Electron Reduction in Nitroaromatic Compounds by NAD(P)H:Quinone Oxidoreductase (NQO1) and Bacterial Nitroreductases
Mammalian NAD(P)H:quinone oxidoreductase (NQO1, DT-diaphorase) is a dimeric 2 × 30 kD enzyme containing one molecule of FAD per subunit. It catalyzes two-electron reduction in quinones and ArNO 2 at the expense of NADH or NADPH. The physiological functions of NQO1 are incompletely understood. It is supposed that it maintains vitamins K1, K2, and K3 in a reduced state and participates in the stabilization of transcription factor p53 ( [132], and references therein). The activity of NQO1 is frequently elevated in various tumors ( [133], and references therein). For this reason, there exists a continuing interest in the development of NQO1-directed bioreductive drugs. The studies of rat, mice, and human NQO1 revealed that the bound nicotinamide ring of NADP(H) displaces Phe178 , interacts with Tyr126 and Tyr128 , and is stacked with FAD isoalloxazine ring at a distance of 3.4 Å. Duroquinone, other quinones, and dicoumarol, competitive to NAD(P)H, occupy the nicotinamide binding site [134][135][136][137]. In contrast, another inhibitor, cibacron blue, binds at the adenosine binding site and almost does not interact with the nicotinamide binding site. The potentiometric and kinetic characteristics of NQO1 are given in Table 3. Under redox equilibrium, NQO1 contains 8% red (anionic) semiquinone. During the reoxidation of reduced NQO1 by quinones and single-electron acceptor ferricyanide, the transient semiquinone formation was not observed. It means that the rate-limiting step of this process is the oxidation of two-electron reduced FAD [138].
The mechanism of reduction in quinones by NQO1 is understood better than the reduction in ArNO 2 . The reactions follow the "ping-pong" mechanism and are frequently characterized by quinone substrate inhibition due to the formation of dead-end complexes with the oxidized enzyme. The reactivity (log k cat /K m ) of rat NQO1 with the number of quinones of different structures increased with their E 1 7 and decreased with their van der Waals volume (VdWvol) [139]. Their optimal VdWvol was estimated to be ≤200 Å 3 . The entropies of activation (∆S = ) calculated according to k cat /K m of reduction in rapidly reacting quinones were equal to −84 ÷ −76 J mol −1 K −1 , whereas the reduction in "slow" quinones was characterized by ∆S = = −36 ÷ −11 J mol −1 K −1 . This demonstrates more efficient electronic coupling of "fast" oxidants in the transition state [140].
The reactivity of a series of nitrobenzimidazoles toward rat NQO1 increases with their E 1 7(calc.) [155]. However, for a larger set of compounds, the dependence of log k cat /K m on E 1 7 is almost absent. The reactivity of ArNO 2 increases with the stability of their complexes with the oxidized form of NQO1 [141,155]. However, the criteria for efficient complex formation are poorly understood. Another factor enhancing the reactivity of ArNO 2 is an increased torsion angle of the nitro group. This, in particular, explains why NQO1 reduces the 4-nitro group of CB-1954 (13). The combined studies of the effects of dicoumarol and cibacron blue and interdependent quinone and ArNO 2 inhibition enabled us to conclude that nitroaromatics and bulky quinones bind partly outside the nicotinamide/dicoumarol binding pocket and may partly occupy adenosine/cibacron blue binding pocket [139,141].
Bacterial type I oxygen-insensitive nitroreductases (NRs) catalyze NAD(P)H-dependent multistep reduction in ArNO 2 with the formation of nitroso, hydroxylamine and amino products. Among several groups of type I NRs, two major ones are distinguished according to their similarity with E. coli nitroreductases A or B [156]. The physiological functions and physiological electron acceptors for bacterial type I NRs are not clearly characterized. It is suggested that they participate in the antioxidant defense of microorganisms as part of the soxRS regulon, whose genes are upregulated in response to oxidative stress [157]. Oxygen-insensitive NRs play a considerable role in the bioreductive degradation of polynitroaromatic explosives such as 2,4,6-trinitrotoluene (TNT) (4) [3]. They also are potential candidates for gene-directed enzyme-produg cancer therapy (GDEPT) using nitroaromatic compounds. This approach is based on the much higher rates of two-electron reduction in ArNO 2 by bacterial NRs than by human NQO1 [7,8,15]. In this case, genes encoding NRs are inserted into the tumor cells using either a virus or a plasmid. Further, we will address the properties of NRs with well-characterized catalysis mechanisms and/or well-established biomedical importance.
Most type I reductases are 2 × 24 ÷ 27 kD dimers, containing one FMN per subunit. They may reduce a wide spectrum of oxidants, including nitroaromatics, quinones, riboflavin derivatives, and inorganic complexes [142,158,159]. Typically, they are inhibited by micromolar or lower concentrations of dicoumarol, which binds at the dihydronicotinamide binding site of NAD(P)H, and acts as a competitive inhibitor toward NAD(P)H. Their reactions proceed according to the "ping-pong" mechanism. The potentiometric and kinetic characteristics of NRs are given in Table 3.
The best-characterized members of group B NRs are E. coli nitroreductase B (NfsB) and Enterobacter cloacae nitroreductase (E. cloacae NR). The X-ray studies of NfsB show that the FMN isoalloxazine ring is localized in the intersubunit domain, its re-plane is solventaccessible [143,160,161]. The nicotinamide ring of bound NAD(P) is stacked between isoalloxazine and Phe124'. Nitroaromatic compounds bind close to the nicotinamide binding domain of NfsB; however, there exist several potential binding sites of CB-1954 (14), with the participation of Lys14, Lys74, Ser12, or Phe124, Asn71, and Gly166 [150]. On the other hand, nitrofurazone interacts with Glu165 and Phe70 and binds in nonproductive conformation because its nitrofuran group does not stack over the isoalloxazine ring of FMN [143]. This may point to the flexibility of the active center of NfsB during catalysis and to its ability to accommodate the oxidants of different sizes and shapes. This is particularly indicated by the ability of NfsB to reduce either the 2-or 4-nitro group of CB-1954 and only the 2-nitro group of SN-23682 (19) [162]. The kinetic parameters of NfsB (Table 3) show that it reduces CB-1954 much more rapidly than does NQO1. As the maximal rates of reduction in nitroaromatics are lower than the rate of reduction in FMN (Table 3), the oxidative half-reaction is the rate-limiting step of catalysis of NfsB. The early studies [158,159] point to the possible absence of substrate specificity and the increase in their reactivity with reduction potential. According to the studies of a series of derivatives of CB-1954 and SN-23682, their k cat /K m varied in the range of 4.8 × 10 2 -6.3 × 10 4 M −1 s −1 [162]. This data scattering points to a definite effect of the size and position of substituents on the reactivity of compounds, which, however, is difficult to characterize. On the other hand, the reactivity of compounds starts to decrease at their VdWvol ≥ 400 Å 3 .
E. cloacae NR possesses 88% homology with NfsB [163]. The potentiometric and kinetic characteristics of the enzyme are given in Table 4. The semiquinone state of FMN of E. cloacae NR is extremely unstable, ca. 0.01% at equilibrium [146]. This possibly determines the two-electron character of reduction in ArNO 2 . The oxidative half-reaction is the ratelimiting step of the catalytical cycle. Using 4-R-2 H-NADH, kinetic isotope effects are observed in both reductive and oxidative half-reactions [147]. It shows that the H atom, transferred from dihydronicotinamide to N-5 position of isoalloxazine, is subsequently transferred to oxidant during two-electron reduction if the exchange of proton at N-5 position with the solvent is sufficiently slow. In contrast with single-electron transferring flavoenzymes and NQO1, E. cloacae NR reduced ArNO 2 faster than quinones with the same E 1 7 value [39,164]. The reactivity of nitroaromatics increased with their E 1 7 or correlated with their ∆Hf(ArN(OH)O − ) or ∆Hf(ArNO 2 − ), thus showing little structure specificity. The k cat /K m of the most efficient oxidants, derivatives of tetryl (2), reached 10 7 M −1 s −1 . TNT (4) oxidized 4 NADH equivalents in two steps, apparently being reduced to dihydroxylamino derivative, whereas tetryl oxidized 6 NADH equivalents. Importantly, E. cloacae NR did not catalyze reductive N-denitration of tetryl (Scheme 3) and reduced it into currently unidentified products.
E. coli NfsA, an FMN-dependent NADPH-specific enzyme, is the best-characterized member of group A nitroreductases [166]. The kinetic and potentiometric characteristics of the enzyme are given in Table 3. NfsA follows the "ping-pong" mechanism with the rate-limiting oxidative half-reaction [149]. The reactivity of ArNO 2 is systematically higher than the reactivity of quinones possessing the same E 1 7 . NfsA reduced tetryl (2)  Like in NfsA, nitrofurantoin binds at the active center of NfsA in nonproductive orientation (Figure 4), its nitrofuran ring interacts with Arg15 and Lys167, and its Scheme 4. A general scheme of reduction in nitroaromatics by oxygen-insensitive nitroreductases involving an enzymatic (pathway (a)) and nonenzymatic (pathway (b)) reduction in nitroso intermediate.
Like in NfsA, nitrofurantoin binds at the active center of NfsA in nonproductive orientation (Figure 4), its nitrofuran ring interacts with Arg15 and Lys167, and its imidazole group binding over the isoalloxazine ring interacts with Arg225 [142]. This shows that the catalysis of NfsA, NfsB, and E. cloacae NR may share some common features. However, there may exist certain differences in their action. The computer modeling study suggests that the binding CB-1954 to NfsA involves Ser40 (Ser41 in NfsB) and Ile129 (Phe124 in NfsB); however, it may also involve Phe42 that is absent in NfsB [167]. In contrast to E. cloacae NR and NfsB, dicoumarol does not quench the fluorescence of FMN, which points to its relatively weak interaction with isoalloxazine [149].

Scheme 4.
A general scheme of reduction in nitroaromatics by oxygen-insensitive nitroreductases involving an enzymatic (pathway (a)) and nonenzymatic (pathway (b)) reduction in nitroso intermediate.
Like in NfsA, nitrofurantoin binds at the active center of NfsA in nonproductive orientation (Figure 4), its nitrofuran ring interacts with Arg15 and Lys167, and its imidazole group binding over the isoalloxazine ring interacts with Arg225 [142]. This shows that the catalysis of NfsA, NfsB, and E. cloacae NR may share some common features. However, there may exist certain differences in their action. The computer modeling study suggests that the binding CB-1954 to NfsA involves Ser40 (Ser41 in NfsB) and Ile129′ (Phe124′ in NfsB); however, it may also involve Phe42 that is absent in NfsB [167]. In contrast to E. cloacae NR and NfsB, dicoumarol does not quench the fluorescence of FMN, which points to its relatively weak interaction with isoalloxazine [149]. In NfsA-catalyzed reactions, log kcat/Km of a series of nitrobenzenes including CB-1954 and nitrofurans correlated well with their E 1 7 [149]. The kcat/Km of the most efficient oxidant, tetryl, reaches 7.9 × 10 6 M −1 s −1 . The reduction rate constants of several 2nitroimidazoles (E 1 7 ~ −0.390 V), dinitrobenzene PR-104 (13), and metronidazole (40) obtained in other studies [15,168,169] are also close to this correlation. Thus, the data available so far demonstrate that the reactivity is determined mainly by the reduction potential of ArNO2 and not by their structural peculiarities.
Mycobacterium smegmatis enzyme MsPnBA is classified as group A nitroreductase [150]. The kinetic and potentiometric characteristics of the enzyme are presented in Table  3. This nitroreductase reduces antitubercular benzothiazinones into their amines and In NfsA-catalyzed reactions, log k cat /K m of a series of nitrobenzenes including CB-1954 and nitrofurans correlated well with their E 1 7 [149]. The k cat /K m of the most efficient oxidant, tetryl, reaches 7.9 × 10 6 M −1 s −1 . The reduction rate constants of several 2nitroimidazoles (E 1 7~− 0.390 V), dinitrobenzene PR-104 (13), and metronidazole (40) obtained in other studies [15,168,169] are also close to this correlation. Thus, the data available so far demonstrate that the reactivity is determined mainly by the reduction potential of ArNO 2 and not by their structural peculiarities.
Mycobacterium smegmatis enzyme MsPnBA is classified as group A nitroreductase [150]. The kinetic and potentiometric characteristics of the enzyme are presented in Table 3. This nitroreductase reduces antitubercular benzothiazinones into their amines and confers M. smegmatis resistance to these drugs [170]. Like in NfsA-catalyzed reactions, the reactivity of a series of examined nitrobenzenes increases with the σ value of their substituents [150]. Other less-characterized NfsA-like nitroreductases from Neisseria meningitidis and Bartonella henselae reduce CB-1954 and metronidazole with similar rates to those of NfsA [169].
A relatively well-characterized oxygen-insensitive NR is the peroxiredoxin-nitroreductase (Prx-NR) hybrid enzyme of Thermotoga maritima, which consists of a Prx domain at the N-terminus, and an FMN-containing NR domain at the C-terminus. These domains function independently without the exchange of redox equivalents [151]. The NR domain of Prx-NR (residues 142-321) possesses a 20-24% homology with E. coli NfsB and E. cloacae NR, and 18% homology with NfsA [152], and does not contain the residues analogous to Phe124 , Phe70, Ser40, Lys14 and Lys74 of group B NRs. The rate-limiting step of Prx-NR catalysis is the oxidative half-reaction. Importantly, the substrate specificity of Prx-NR differs from that of E. cloacae NR and NfsA. Although the log k cat /K m of ArNO 2 increased with their E 1 7 , nitroaromatics were less reactive than quinones with the same E 1 7 values [152]. Because this phenomenon is characteristic of single-electron reduction (see Section 3.1), it is possible that ArNO 2 are reduced in a multistep (e − , H + , e − ) way with a rate-limiting first electron transfer.
Helicobacter pylori contains nitroreductase RdxA, an FMN-containing 2 × 26 kD dimer that shows no more than 29% sequence identity with other homologous structures of NRs [171,172]. RdxA exhibits high NADPH oxidase activity, 2.8 s −1 . The maximal reduction rate of nitrofurazone and CB-1954 was similar to RdxA oxidase activity. They were reduced with k at /K m of 1.4 × 10 6 M −1 s −1 and 3.0 × 10 5 M −1 s −1 , respectively. Metronidazole was reduced with a much lower rate, k cat = 0.22 ÷ 0.62 s −1 and k at /K m = 2.0 × 10 3 M −1 s −1 . Nitrothiazole nitazoxanide (52) was reduced by RdxA with a rate similar to that of metron-idazole [173]. Another H. pylori nitroreductase, FrxA, is even more scarcely characterized. In this case, the reactivity of ArNO 2 increases in the order metronidazole < nitrofurans < nitazoxanide [173]. The factors determining substrate specificity of both NRs are unclear.
Nitroreductase-type enzymes (NTR) were found in parasites Trypanosoma cruzi and T. brucei [174][175][176]. They represent 2 × 30 kD dimers containing one FMN per monomer. T. cruzi NTR is NADPH-specific, whereas the T. brucei enzyme oxidizes both NADH and NADPH. They are inhibited by dicoumarol with K i = 258 nM and K i = 14 nM, respectively. These enzymes reduce nitrofurans, nitroimidazoles, and nitrobenzenes with k cat = 0.2 ÷ 1.2 s −1 , and k cat /K m ranging from 7.3 × 10 2 to 2.5 × 10 5 M −1 s −1 . The final product of reduction in nifurtimox under aerobic conditions is unsaturated open-chain nitrile, whose antitrypanosomal activity was close to that of nifurtimox. Benznidazole is metabolized into hydroxylamine products, which further undergoes secondary reactions. The factors determining enzyme-substrate specificity are unclear. It is evident that the activity of compounds does not correlate with their E 1 7.
There exist limited data on the formation of amines as the final product of reduction in ArNO 2 by type I NRs, although this problem is relevant both to biomedicine and ecotoxicology. NfrA from Bacillus LMA exhibits 40% homology with NfsB and reduces 3,5dinitro-trifluoromethylbenzene to diamine product with k cat = 18 s −1 [180]. Nitrofurazone is reduced to its amine derivative at a lower rate. S. typhimurium NR quantitatively reduces nitrobenzene into aniline [165]. It is unclear whether the tendency for amine formation is determined by the properties of ArNO 2 or nitroreductase or by both factors. It has been suggested that the possibility of amine formation increases with the reduction potential of ArNO 2 and the size of their aromatic system [150]. A recent study shows that Haemophilus influenza NR-B reduces chloroamphenicol (23) into a corresponding amine with k cat = 10.2 s −1 and k at /K m = 2.0 × 10 4 M −1 s −1 [181]. This NR possesses unusual and undisclosed substrate specificity because it reduces more powerful oxidant metronidazole (40) ( Table 1) with a lower rate, k cat = 0.34 s −1 and k at /K m = 4.6 × 10 3 M −1 s −1 with the formation of its hydroxylamine metabolite.
Summing up, the two-electron reduction in ArNO 2 by NQO1 and bacterial oxygeninsensitive NRs may be attributed to the low stability of their flavin semiquinone state. However, the relative stability of FAD − of NQO1, 8% under equilibrium [138], may enable this enzyme to perform the reductive denitration of tetryl (2) (Scheme 3) in a mixed singleand two-electron way [143]. This reaction is not characteristic for E. cloacae NR-B and E. coli NfsA [39,149], evidently due to the much lower stability of their FMN semiquinone [146].
The crystallographic studies of NRs from E. coli [142,143,160,161] point to the flexibility of their active sites and to their ability to accommodate the substrates of various sizes. The kinetic studies of several A-and B-type NRs demonstrate that the reactivity of ArNO 2 is strongly influenced by their reduction potential [39,149,150]. However, this leaves some space for the improvement of the activity of compounds. Another unresolved problem is the factors determining substrate specificity of nitroreductases from H. pylori, H. influenza, Leishmania, and Trypanosoma spp.

Single-and Two-Electron Reduction in Nitroaromatic Compounds by Flavoenzymes Disulfide Reductases
Flavoenzymes disulfide reductases contain FAD and redox-active disulfide group, which participate in the transfer of redox equivalents in a sequence NAD(P)H → FAD → catalytic disulfide → low-M r or protein disulfide substrate. In most cases, they perform antioxidant functions. These reactions proceed via obligatory two-electron (hydride) transfer without the formation of free radical intermediates ( [184,185], and references therein). Although being slow, the nitroreductase reactions of disulfide reductases received significant attention because of the combined action of ArNO 2 , redox cycling, and inhibition of physiological reactions of disulfide reductases. It is important to note that these compounds are reduced by flavin but not by reduced disulfide cofactor due to unfavorable energetics of single-electron oxidation of dithiols [186].
Glutathione reductase (GR) and trypanothione reductase (TR), the 2 × 55 kD homodimers, contain one FAD and catalytic disulfide per subunit. GR catalyzes the reduction in glutathione disulfide (GSSG), and TR catalyzes the reduction in trypanothione (TS 2 ), a glutathione-spermidine conjugate. The structure and reaction mechanism of both enzymes are similar [187][188][189]. GR performs antioxidant functions in various organisms. TR is found exclusively in trypanosomes and leishmanias, the causative parasites of several tropical diseases, including African sleeping sickness and Chagas disease. These parasites do not contain GSSG/GSH, and their antioxidant defense relies mainly on TR-catalyzed regeneration of T(SH) 2 . The presence of amino acids with different charge in the disulfide substrate-binding domain of HGR and T. congolense TR [187,190] enable the discrimination between the negatively charged GSSG and positively charged TS 2 .
GR and TR are reduced by NADPH to two-electron reduced form (EH 2 ), which is the FAD-thiolate charge-transfer complex with the main electron density being localized on thiolate. Next, EH 2 is reoxidized by disulfide. These reactions follow a "ping-pong" mechanism with reductive half-reaction as a rate-limiting step. The k cat values of human erythrocyte GR (HGR), P. falciparum GR (Pf GR), and TRs span between 120 and 240 s −1 [185,[191][192][193][194][195]. More precisely, their mechanism should be classified as "hybrid ping-pong" because during turnover, GSSG reoxidizes not free EH 2 form, but its tight complex with repeatedly bound NADPH (K d = 2.1 µM, yeast GR [196]). In this case, GSSG oxidizes free EH 2 and its complexes with NADPH and NADP + with sufficiently close rates. The redox potentials of HGR, Pf GR and T. congolense TR are equal to −0.227 (pH 8.2 [193]), −0.206 (pH 6.9 [197]), and −0.275 V (pH 7.5 [198]), respectively. Under artificial conditions, GR may be further reduced into the four-electron reduced state (EH 4 ); however, this form is not formed during the enzyme turnover. This is attributed to the tight binding of NADPH, which stabilizes the EH 2 form. ArNO 2 are reduced by GR and TR in a single-electron way. The most efficient oxidant of HGR and Pf GR is tetryl (2) (k cat ≥ 5 s −1 , k cat /K m = 2.0 ÷ 7.5 × 10 3 M −1 s −1 [195]). The low reaction rates complicate the substrate specificity studies. However, the introduction of basic substituents into nitrofuran molecule enhances their reactivity toward TR (k cat = 2.5 ÷ 3.0 s −1 , k cat /K m = 3.3 ÷ 9.2 × 10 4 M −1 s −1 [199,200]). A specific feature of quinone and nitroreductase reactions of GR and TR is the activation by the reaction product NADP + [192,201]. Although the main electron density in the FAD-thiolate charge-transfer complex is localized on thiolate, its minor part remains on FAD. The binding of NADP + to EH 2 with a concomitant displacement of NADPH increases the electron density on FAD, which may accelerate the reduction in xenobiotics. The order of reactivity of various redox forms of GR and TR with quinones and presumably with ArNO 2 is EH 2 < EH 2 -NADPH < EH 2 -NADP + < EH 4 [192,202]. However, the site(s) of their reduction are not characterized.
An important aspect of the interaction of ArNO 2 with GR and TR is the inhibition of the reduction in their physiological disulfide oxidants [17,195,[199][200][201][202][203]. In these cases, compounds act as non-or uncompetitive inhibitors with respect to NADPH and disulfide substrate and bind at the intersubunit domain of GR or TR close to the binding sites of GSSG or TS 2 . The amino acid residues of this domain of HGR, Pf GR, and TR are strikingly different [189,204]. Currently, significant efforts are devoted to the studies of efficient and specific inhibitors targeting this domain, with an aim to develop new antimalarial or trypanocidal drugs [194,205]. The efficiency of ArNO 2 or other redox-active compounds is evaluated by the ratio (k cat /K m )/K i , which combines their ability to serve as redox cycling agents and inhibit GSSG/TS 2 reduction [194]. For most active compounds, this ratio is above 10 11 M −2 s −1 (naphthoquinones) and above 10 10 M −2 s −1 (nitrofurans) [194,199,200].
Lipoamide dehydrogenase (LipDH, NADH:lipoamide oxidoreductase) is structurally similar to GR and TR, although it does not perform antioxidant functions. Mammalian LipDH catalyzes the NAD + -dependent oxidation of covalently bound dihydrolipoate of the pyruvate dehydrogenase or α-ketoglutarate dehydrogenase complexes [184]. Pig heart enzyme catalyzes the rapid transfer of two redox equivalents in both directions, from dihydrolipoamide to NAD + and from NADH to lipoamide (LipS 2 NH 2 ). In the latter case, the rate-limiting step of the process is the oxidative half-reaction [184,206]. Like GR or TR, LipDH shuttles between E ox and EH 2 states.
Pig heart and T. cruzi LipDH reduce nitroaromatic compounds with mboxemphk cat = 0.1 ÷ 2.0 s −1 [203,207] that makes 0.02-0.5% of LipS 2 NH 2 reduction rate. Some information on the reaction mechanism may be obtained from the studies of quinone reduction, which takes place with 20% (pig heart LipDH [206]), or even with 80% of maximal disulfide oxidant reduction rate (M. tuberculosis LipDH [208]). It was concluded that the EH 4 fraction of LipDH, which is formed in the absence of NAD + , is mostly responsible for quinone reduction. Because M. tuberculosis LipDH efficiently reduces quinones, k cat =165-190 s −1 [208], one may also expect its significant nitroreductase activity, which, however, was not investigated. NAD(P)H-thioredoxin reductases (TrxRs) are important antioxidant enzymes in both prokaryotic and eukaryotic organisms [209]. There exist two groups of TrxRs: high-M r (2 × 54-58 kD) enzymes contain FAD, catalytic disulfide, and a third cofactor, selenylsulfide (mammalians), or additional disulfide (P. falciparum) in their active center. Low-M r enzymes (2 × 35 kD) from plants, bacteria, and archaea contain FAD and catalytic disulfide. The physiological substrates of TrxRs are 10-12 kD disulfide proteins thioredoxins (Trxs), and sometimes disulfide or monothiol-containing glutaredoxins (Grxs). Apart from the antioxidant action, they perform other numerous physiological functions [210,211]. It is important to note that human TrxR is overexpressed in numerous cancers [212].
First, let us discuss the properties of mammalian (rat, human) TrxR. The catalytic cycle of the enzyme is similar to GR, except for the participation of selenylsulfide, which is located at the C-end of the protein [213,214]. The rate-limiting catalysis step, 30 ÷ 40 s −1 , is the two-electron transfer between dithiolate and selenylsulfide, which is responsible for the reduction in Trx and other numerous substrates of mammalian TrxR. During catalysis, the enzyme cycles between EH 2 and EH 4 redox states (E 0 7 = −0.294 V [215]). Mammalian TrxR may directly reduce 5,5 -dithiobis(2-nitrobenzoic acid) (DTNB), with the maximal rate being equal to that of Trx reduction. Other less efficient oxidants are cystine, lipoate, alloxan, dehydroascorbate, vitamin E, coenzyme Q10, and vitamin K, lipid peroxides, and H 2 O 2 .
Rat TrxR reduces ArNO 2 either in a two-electron way (p-dinitrobenzene) or in a singleelectron way (tetryl) [216]. During the reaction, TrxR cycles between four-and two-electron reduced states. The reactivity of ArNO 2 , k cat = ≤ 0.1-2.8 s −1 , and k cat /K m = < 10 2 -1.4 × 10 4 M −1 s −1 [216,217], is comparable with that of quinones, which follow poorly expressed linear log k cat /K m vs. E 1 7 relationship [215]. The reduction in nitroaromatics that may alkylate -SH/-Se − groups (1-chloro−2,4-dinitrobenzene, tetryl (2), dinitrobenzo-furoxane (48), 2-phenylsulfonylnitropyridines) is accompanied by a rapid covalent modification of the enzyme with a loss of DTNB and Trx reductase activity and to an increase in NADPH oxidase activity [216,218,219]. The most important feature of TrxR is that in addition to reduced FAD, the reduced selenylsulfide may participate in the reduction in fully substituted quinones such as phenanthrene quinone or toxoflavin [215,220], and, presumably, nitroaromatics. This was demonstrated by the selective suppression of these reactions by Au-thioglucose, which specifically reacts with selenol. The participation of selenocysteine but not cysteine in these reactions may be explained by the relative ease of its single-electron oxidation since E 7 (Sec·/Sec − ) is equal to 0.43 V, whereas E 7 (Cys·/CysH) is equal to 0.92 V [186]. Dinitrobenzenes and nitrofurans acted as weak noncompetitive inhibitors with respect to DTNB (K i = 17-400 µM), whereas tetryl and dinitrobenzofuroxane (48) acted as competitive inhibitors with K i values of 12.5 and 5.0 µM, respectively [216]. This points to the existence of several binding sites of ArNO 2 .
Another representative of high-M r enzymes, P. falciparum TrxR, contains FAD and two catalytic disulfides, Cys88, Cys93, and Cys535, Cys540 in the active center [221]. The enzyme reduces its physiological oxidant, Pf Trx1, and quinones with relatively high and similar rates, 51.7 and 31-67 s −1 , respectively [222]. This is >10 times faster than the reduction in quinones by Pf GR or HGR. However, the nitroreductase reactions of Pf TrxR have not been examined.
Low-M r TrxRs from E.coli and other species contain one FAD and one redox-active disulfide per subunit. In contrast with GR and TR, their FAD and disulfide redox groups function separately. The E 0 7 values of FAD/FADH 2 and S 2 /(SH) 2 redox couples are equal to −0.243 ÷ −0.260 and −0.254 ÷ −0.271 V, respectively [223]. In E. coli TrxR, NADP(H) binds at the vicinity of catalytic disulfide and at 17 Å distance from the isoalloxazine ring of FAD. In order to reduce FAD, the disulfide/NADP(H)-binding domain should undergo substantial rotation, which in turn exposes the catalytic disulfide into solution and makes it accessible to Trx [224]. This makes the formation of charge-transfer complex formation impossible [225,226]. The catalysis rate-limiting step, 25 ÷ 30 s −1 , is either the conformational transition or Trx reduction. In catalysis, TrxR cycles between two-and four-electron reduced forms, FADH 2 /S 2 and FADH 2 /(SH) 2 . Interestingly, in spite of an obligatory two-electron character of physiological reactions of E. coli TrxR, it forms neutral (blue) FAD semiquinone during irradiation under anaerobic conditions [227].
The reactions of ArNO 2 and other prooxidant xenobiotics with E. coli TrxR were not studied. However, they were examined with structurally similar TrxRs from Arabidopsis thaliana (E 0 7 (FAD/FADH 2 ) = −0.244 V) [228][229][230] and Thermotoga maritima (E 0 7 (FAD/FADH 2 ) = −0.230 V) [151,[231][232][233]. The latter enzyme preferentially oxidizes NADH, and its physiological oxidant is a 25.2 kD Grx-1. Although its crystal structure is unavailable, this enzyme possesses two catalytic cysteines Cys147 and Cys150, a vicinal NAD(P)H bonding motif, FAD bonding motif, and interdomain motif Gly251, Pro254, which possibly participates in domain rotation [231]. Both enzymes reduced ArNO 2 with 70-90% single-electron flux [230,232] and catalyzed N-denitration of tetryl with the formation of N-methylpicramide (Scheme 3). This reaction was characterized by k cat = 1.5 s −1 and k cat /K m = 2.7 × 10 4 M −1 s −1 (A. thaliana TrxR), and k cat = 6.3 s −1 and k cat /K m = 6.6 × 10 4 M −1 s −1 (T. maritima TrxR). In both cases, the log k cat /K m of reduction in ArNO 2 increased with E 1 7 . The reactivity of nitroaromatics was by one order of magnitude lower than that of quinones with the same E 1 7 values. On the analogy with quinone reduction by T. maritima TrxR, the rate-limiting reaction step in nitroreductase reactions of low-M r TrxRs may be the oxidation of FAD semiquinone [233]. It is important to note that the single-electron character of ArNO 2 reduction and reactivity of A.thaliana and T. maritima TrxRs are similar to the properties of low-M r TrxRs from T. vaginalis and Giardia lamblia [234,235]. Thus, they may be considered as the model enzymes for the studies of this pathway of ArNO 2 activation in parasites.
The structure and functions of Salmonella typhimurium alkyl hydroperoxide reductase AhpF share some similarities both with high-M r and low-M r TrxRs ( [236], and references therein). Its 58 kD monomer contains FAD and two redox-active disulfide centers, Cys345-Cys348 and Cys129-Cys132. C-terminal FAD-binding domain of AhpF possesses 34% identity with the analogous domain of E. coli TrxR. Through putative large interdomain movements, electrons are transferred in a sequence NADH → FAD → Cys345-Cys348 → Cys129-Cys132. The latter center, located in N-terminus, transfers electrons to the catalytic disulfide of another hydroperoxide reductase, AhpC. In contrast to E. coli TrxR, reduced AhpF may form 540 nm absorbing charge-transfer complex under high ionic strength of the solution. AhpF from E. coli reduces nitrofurans in a single-electron way [237]. Although the kinetic parameters of these reactions are not reported, AhpH is partly responsible for the sensitivity of E. coli to nitrofurans.
Summing up, the low nitroreductase activity of disulfide reductases may be attributed to the low electron density on FAD in FAD-thiolate charge-transfer complexes. This diminishes their role in redox cycling or other modes of activation of ArNO 2 in the cell. However, low-M r TrxRs where FAD and catalytic disulfide function separately may more significantly contribute to the redox cycling of ArNO 2 . The frequently observed singleelectron reduction in nitroaromatics is not inconsistent with the redox properties of FAD of disulfide reductases. Apart from the formation of FADH· of low-M r TrxR of E. coli under the artificial conditions [227], the complex of reduced GR with NADP + also gradually converts into a product, characterized as the complex of NADP + with FAD − ·( [184], and references therein). Another relevant problem is the exploitation of the bioreductive potential of a reduced selenylsulfide moiety of mammalian TrxR in nitroreductase reactions.

Role of Bioreductive Processes in Mammalian Cell Cytotoxicity of Nitroaromatics
Redox cycling is an intrinsic property of ArNO 2 ; therefore, the oxidative stress will obligatorily take place during their action under aerobic conditions. In our opinion, the deviation from the limits predicted by the redox cycling activity could be instrumental in the characterization of additional mechanisms of cytotoxicity or therapeutic action of nitroaromatics. It is commonly accepted that if the main cytotoxicity factor is the rate of enzymatic formation of ArNO 2 − · and their redox cycling, their cytotoxicity may be described by a following quantitative structure-activity relationship (QSAR): where cL 50 is compound concentration causing 50% cell death or analogous quantitative parameter, log P is octanol/water partition coefficient, and log D is octanol/water distribution coefficient at pH 7.0 [13,155,[238][239][240][241] (Table 4). These dependences mirror the log (rate constant) vs. E 1 7 relationships in single-electron reduction in ArNO 2 by P-450R, NOS, or other single-electron transferring flavoenzymes (see Section 3.1). An additional diagnostic test is antioxidant protection against cytotoxicity. However, this approach may not be applied in the assessment of the mechanisms of action of ArNO 2 that possess alkylating, bioreductively alkylating, or leaving groups (see below), or may directly interact with DNA, e.g., the positively charged nitracrine (49) derivatives [242]. Nevertheless, this approach, in particular, the use of E 1 7(calc.) obtained from enzymatic single-electron reduction reactions (Table A1, Appendix A), enabled to demonstrate that the main cytotoxicity mechanism of explosives pentryl (1), tetryl (2), TNT (4), tetranitrobenzimidazolone (42), dinitrobenzofuroxane (48), ANTA (54), NTO (55), and nitroaromatic antiandrogens nilutamide (15) and flutamide (16) could be oxidative stress [13,71,155,241,[243][244][245][246]. On the other hand, although nifurtimox (27) induces oxidative stress in neuroblastoma cells and is synergistic with the GSH synthesis inhibitor buthionine sulfoximine, it is unclear whether this is the main mechanism of its anticancer action [247,248]. Table 4. Structure-activity relationships in mammalian cell cytotoxicity of nitroaromatics with possible relevance to bioreductive activation.

Compounds and Cells
Structure-Activity Relationships

Hypoxic Conditions
Nitroimidazoles, nitrofurans, V79-379A cells [254] ∆log cL 50 /∆E 1 7 = −10 V −1 4-(Alkylamino)-5-nitroquinolines, CHO-AA8 cells [249] Cytotoxicity and hypoxic selectivity roughly increase with the rate of drug-stimulated O 2 consumption Nitrofurans and nitrothiophenes with alkyl-N-aziridine or alkyloxirane side chains, V79-379A cells [251][252][253] Alkylating side chains do not increase hypoxic selectivity N,N-bis(2-chloroethyl)-N-methyl-N-(2-nitrobenzyl) ammonium chloride and its derivatives, CHO-AA8 cells [250] Alkylating leaving groups increase cytotoxicity, no relationship between E 1 7 and cytotoxicity or hypoxic selectivity Nitrochloromethylbenzindolines (nitroCBIs, 56), Skov3 and HT29 cells [23] Hypoxic toxicity does not correlate with E 1 7 , hypoxic selectivity roughly increases with E 1 7 In this context, one may note that because of the similarity of log (rate constant) vs. E 1 7 relationships in the reduction in ArNO 2 by flavoenzymes dehydrogenases-electrontransferases, it is difficult to assess their individual contribution. Some information may be obtained from the hypoxic cytotoxicity studies (see below), assuming that the extent of formation of reduced ArNO 2 metabolites is proportional to the rate of their singleelectron enzymatic reduction. In this case, the main enzyme responsible for redox cycling of ArNO 2 should be NADPH:cytochrome P-450 reductase (P-450R) [255,256]. The roles of other P-450R-type enzymes such as inducible NO synthase, human novel oxidoreductase NR1 [257], and methionine synthase reductase [258] depend on the cell type. Although the nitroreductase reactions of the latter two enzymes are uncharacterized, one may expect the similarity of their log (rate constant) vs. E 1 7 relationships with those of P-450R and NOS. On the other hand, the contribution of mitochondrial NADH:ubiquinone dehydrogenase and, possibly, other flavoenzymes of mitochondrial respiratory chain to the cytotoxicity of ArNO 2 was recently demonstrated [259]. The comparison of reactivities of P-450R and NOS with those of disulfide reductases (Tables 1 and 2 and Section 3.3) clearly rules out the latter enzymes as the important sources of free radicals of ArNO 2 . The inhibitors of cytochromes P-450 protect against the cytotoxicity of nitrobenzenes, nitrofurans, and nitrothiophenes [71,205] in several cell lines; however, the roles of cytochromes P-450 in ArNO 2 cytotoxicity remain unclear so far.
The presence of directly alkylating alkyl-N-aziridine or alkyloxirane groups in ArNO 2 molecules evidently increases their cytotoxicity and abrogates the log (cytotoxicity) vs. E 1 7 relationships (Table 4). It is important to distinguish the contribution of redox cycling and alkylation in their action, in particular, the role of putative alkylation targets. It is suggested that 2-phenylsulfonylnitropyridines, which possess micromolar cL 50 levels against the NCI cancer cell lines, exert their action through thioredoxin reductase inactivation [219]. However, because of the high E 1 7 value of nitropyridine moiety (Table A1), the redox cycling reactions with flavoenzymes electrontransferases may significantly contribute to their cytotoxicity. The influence of bioreductively activated alkylating groups such as aziridine of nitrogen mustard or bioreductively activated leaving tetraalkylammonium group in the aerobic cytotoxicity of ArNO 2 remains equivocal ( Table 4). In these cases, an increase in cytotoxicity of ArNO 2 may be attributed to the formation of a fraction of stable reduction products in cell compartments with reduced O 2 tension such as nucleus and mitochondria. In cell lines with 100-200 U/mg NQO1 (determined according to quinone reduction rate), dicoumarol partly protects against the cytotoxicity of ArNO 2 . However, the cytotoxicity of CB-1954, SN-23682, and their derivatives is close to the limits predicted by their E 1 7 [240], and dicoumarol similarly protects against the cytotoxicity of ArNO 2 that do not contain aziridine or nitrogen mustard groups [71,243].
ArNO 2 are considered to be the hypoxia-selective toxins or therapeutic agents because their hydroxylamine reduction products formed under hypoxia may alkylate DNA and other nucleophiles. Frequently, their -NHOH or -NH 2 metabolites demonstrate even higher cytotoxicity than the parent ArNO 2 ([23,260,261], and references therein). Additionally, the electron-donating character of these groups increases the reactivity of alkylating aziridine or nitrogen mustard groups and accelerates the dissociation of leaving groups. In certain cases, the hypoxic selectivity, i.e., the ratio of cL 50 under oxic and hypoxic conditions, may exceed 1000. It was established that NADPH:cytochrome P-450 reductase (P-450R) plays a major role in the formation of reduced metabolites of ArNO 2 under hypoxia [255,256]. One may expect the correlation of hypoxic toxicity with E 1 7 analogously with the rate of ArNO 2 reduction. However, this type of correlation is frequently absent (Table 4). Evidently, this reflects the superposition of two opposite factors, the ease of reduction increasing with E 1 7 of the parent compound and the alkylating reactivity, decreasing with electron-accepting potency of substituents. The studies of CB-1954, SN-23682, and their reduction product have shown that the cytotoxicity is determined mainly by the alkylating reactivity of reduction products, their stability, and interconversion rate [260].
Nitroreductase-transfected cancer cells are the objects of gene-directed enzyme-prodrug therapy (GDEPT). Most typically, NRs, initially E. coli NfsB, activated the alkylating compounds such as CB-1954 and SN-23682 or released a drug molecule bound to ArNO 2 after its reduction (Scheme 1) [7]. Further, we will address the reductive activation of alkylating ArNO 2 , because in the other case, the cytotoxic outcome mostly depends on the properties of the released drug. Early studies have shown that in the presence of exogenous NfsB and NADH in the cell growth medium, the cytotoxicity of CB-1954, SN-23682, and their derivatives roughly correlated with the rate of their enzymatic reduction [162]. The sensitivity to CB-1954 was shown to correlate closely with the level of nitroreductase expression in the cell ( [7], and references therein). CB-1954 demonstrated a substantial bystander effect, i.e., killing of non-transfected cells, shown to be mediated via the cell-permeable hydroxylamine metabolite. Among ArNO 2 that do not possess bioreductively activated alkylating groups, nitrofurazone, but not misonidazole (35) or nitracrine (50) showed significant activity in NfsB-transfected cell lines. This may be attributed to its better activity as an enzyme substrate.
However, the further use of NfsB in GDEPT was complicated by the relatively low tolerance of patients to large concentrations of CB-1954, required for its efficient NfsBcatalyzed conversion in the cell ( [8,15], and references therein). This has spurred a search for alternative nitroreductase enzymes that are more effective at converting CB-1954 at its low concentrations and for alternative nitroaromatic compounds. The large-scale screening of various NRs pointed to the versatility of NfsA-group nitroreductases, which in general possess higher k cat /K m values of CB-1954 than NfsBs. It is suggested that not a higher k cat but a much higher affinity (i.e., much lower K m ) for a given substrate is a defining characteristic of the efficiency of the process [262]. However, the nitroreductase gene expression was unpredictable in transfected HCT-116 cells, and there was little correlation between the CB-1954 sensitivity of those cell lines and the k cat /K m of the purified nitroreductases. Possibly, the long-term stable expression of some nitroreductases is not tolerated in human cell lines. Another potential drawback of the NfsA enzymes is their preference for NADPH, as NADPH is usually present at lower concentrations than NADH in mammalian cells. Currently, efforts are undertaken for the construction of selective mutants of NRs and the improvement of their expression [263,264]. Further studies are performed with new compounds such as PR-104A (13), TH-302 (36), and nitroCBI (56).
Nitroaromatic compounds induce both necrotic and apoptotic mammalian cell death. Thus, finally, one has to address the expression of cell signaling proteins during the apoptotic action of ArNO 2 . The current studies are confined to the changes of expression and phosphorylation of the 'classical' signaling proteins such as tumor suppressor p53, cell cycle inhibitor p21, proapoptotic protein Bax, antiapoptotic protein Bcl-2, protooncogene protein N-Myc, various caspases (cysteine-dependent aspartate-directed proteases, which are the executioners of the apoptotic cell death), mitogen-activated kinases JNK, p38 and ERK, and other kinases (Table 5). In certain cases, the upregulation of p21, p53, Bax/Bcl-2 ratio, and caspase expression was attenuated by antioxidants [265][266][267]. This shows that ArNO 2 − initiated oxidative stress contributes to these events.

Role of Bioreductive Processes in Antibacterial and Antiparasitic Action of Nitroaromatics
The current progress in the design and development of nitroaromatic antibacterial and antiparasitic agents was recently presented in several comprehensive reviews [6,268]. In this subsection, we will address the cases with evidence and, preferentially, quantitative demonstration of the contribution of reductive processes to the action of ArNO 2 .  Activation of phosphorylation 1,3-dinitrobenzene, rat Sertoli cells [272] AKT kinase Inhibition of phosphorylation 2-(2-(5-Nitrofuran-2-yl)-vinyl)-quinoline derivative, H1299 [275]; 5-nitroimidazole DYT-40, U251 and U87 [279]; nifurtimox, human neuroblastoma [280] Nitroreductase-deficient mutants of E. coli were resistant to all types of nitroaromatic compounds [281,282]. This points to the principal role of NRs in their bioactivation. Reflecting the correlation between the reactivity of NfsA and NfsB and E 1 7 of nitroaromatics, the rates of ArNO 2 aerobic reduction by E. coli increased with their electron-accepting potency [283]. However, the activity of ArNO 2 against E. coli and other aerobic bacteria poorly correlated with their electron affinity [284]. This may point to the role(s) of their other insufficiently studied targets such as AhpF [237] or to a parallel activity dependence on the alkylating potency of reduction products, which decreases with the electron-donating potency of substituents [285]. In particular, the activity of a large series of nitroaromatics against anaerobic Bacteroides fragilis decreased with their E 1 7 [284]. On the other hand, there was a reasonable correlation between the k cat /K m of 20 purified NRs with CB-1954 as substrate and their ability to induce CB-1954-mediated DNA damage in E. coli host cells [15].
Metronidazole and other 5-nitroimidazoles exert a high antitrichomonal activity under both aerobic and anaerobic conditions; in most experiments, however, inhibitory concentrations are higher in aerobiosis than in anaerobiosis [286]. Trichomonad flagellates and anaerobic bacteria become resistant to metronidazole after a loss of PFOR and NADH:ferredoxin reductase (FOR) functions [120,286]. The unexpectedly high rate of reaction of nitroimidazoles by ferredoxin that initiates redox cycling [110] possibly explains the high sensitivity of T. vaginalis to metronidazole. Because the rate of reduction ArNO 2 by ferredoxin is apparently insensitive to their E 1 7 [121], it points to the perspectives of application of low-potential compounds that are nontoxic to mammalian cells. However, the further steps are complicated by the uncharacterized nitroreductase reactions of FOR and a parallel involvement of 26 kD FMN and FeS-containing protein with uncharacterized functions [122], which, under anaerobiosis, reduces metronidazole to amino metabolite and contributes to its detoxification. Because of low nitroreductase activity, the role of TrxR in the activation of ArNO 2 in T. vaginalis remains a matter of debate [234,235].
The current studies of Helicobacter pylori most frequently address the mechanisms of its resistance to metronidazole [287]. The sensitivity to metronidazole correlated with NADH-and NADPH-dependent nitroreductase activity in the cytosolic fraction of clinical isolates [288]. In turn, resistance to metronidazole results from loss-of-function mutations in genes encoding nitroreductases RdxA and FrxA ( [268], and references therein). However, the individual roles of the above NRs in the action of other nitroaromatic compounds remain poorly understood so far. Moreover, another potential candidate for the efficient reduction in ArNO 2 is the complex of PFOR, FqrB, and flavodoxin [124]. Nitroaromatic compounds that bind to flavodoxin at micromolar concentrations exhibit activity against H. pylori at similar concentrations. Importantly, these compounds were active against metronidazole-resistant isolates, which may point to a different mechanism of their action [289]. Further studies in this direction warrant detailed examination of nitroreductase activity of FqrB and, especially, flavodoxin.
Giardiasis is treated primarily with metronidazole, but other ArNO 2 , including benznidazole, nitazoxanide, and nitrofurans, are also effective [125]. Giardia spp. possess numerous potential sources of their activation, such as PFOR, ferredoxin, thioredoxin reductase, and several nitroreductases ( [125,290,291], and references therein). However, their individual contributions to the action of particular drugs remain a matter of debate.
The recent review [6] did not reveal the relationships between the antitubercular activity of nitroaromatics and their redox properties. The activity of a series of derivatives of PA-824 (57) against M. tuberculosis did not depend on their E 1 7 value in the range of −0.570-−0.338 V [24] or their reduction potentials in acetonitrile [292]. On the other hand, 6,8-dinitro-1,3-benzothiazin-4-ones were more active than their 6-trifluoromethyl-8-nitro analogs [293]. Although the downregulation of F-420-dependent nitroreductase confers the resistance of M. tuberculosis to PA-824 and other nitroaromatics [294], the studies of several series of PA-824 derivatives did not demonstrate the relationship between their enzymatic reactivity and antibacterial activity [182,295]. In our opinion, the further development of antitubercular nitroaromatics could be extended by studies of their interaction with M. tuberculosis flavoenzymes, e.g., FprA [102] or LipDH [208].
Nifurtimox (27) and benznidazole (34) were the only nitroaromatic compounds to be recently considered for antitrypanosomal drug development ( [296], and references therein). Because of its genotoxicity, megazol (38) was not further developed. The mechanisms of antitrypanosomal action of nifurtimox and other nitrofurans are a matter of debate. The early studies suggested that the main mechanism of action is their single-electron reduction and redox cycling of free radicals, the potential candidates being trypanothione reductase, lipoamide dehydrogenase, and, possibly, P-450R-type enzymes [199,203,297]. However, redox cycling was only observed at high nifurtimox concentrations, which were by two orders of magnitude higher than its micromolar concentrations required for antiproliferative activity [297]. On the other hand, the two-electron transferring trypanosomal NTRs may play a leading role in the bioactivation of nifurtimox and, possibly, other nitrofurans and nitroimidazoles [174][175][176]. Their upregulation or downregulation increased the sensitivity or resistance of parasites, respectively, toward nifurtimox, benznidazole, and fexinidazole. However, because of limited kinetic data on NTR reactions, including the new groups of ArNO 2 such as 3-nitrotriazoles and nitroquinolin-2(1H)-ones [298][299][300][301][302], it is difficult to demonstrate the relationship between their trypanocidal activity and reactivity toward NTR. Moreover, in certain cases, the upregulation or downregulation of NTR causes limited effects on parasite sensitivity. In this context, parallel studies of other flavoenzyme targets of nitroaromatics may be suggested. For example, the nitroreductase activity of trypanothione reductase (TR) is similar to that of NTRs [199,200], although the reaction proceeds in a single-electron way. Moreover, nitrofurans act as TR inhibitors [199,200]. There did not exist a relationship between the trypanocidal activity and TR inhibition potency of a series of nitrofurans, which were weak enzyme inhibitors (K i = 61-1180 µM) [303]. However, the trypanocidal activity of chinifur (24) and its derivatives which were more efficient TR inhibitors (K i = 2.3-110 µM), increased with their TR inhibition potency [304].
NR1 of Leishmania spp. participates in the activation of the new antiparasitic agent fexinidazole (38) and nifurtimox (27), because its upregulation increases the parasite sensitivity to these drugs. On the other hand, the knockout of NR2 caused the parasite complete resistance R-PA-824 (57) and only marginal resistance to nifurtimox [179,296]. This is in line with their relative efficiency as NR2 substrates.
The action of ArNO 2 against the malaria parasite P. falciparum received relatively little attention, although a number of nitrofurans, nitrobenzenes, nitroimidazoles, and 4-nitrobenzothiadiazole were shown to possess in vitro antiplasmodial activity at micromolar or lower concentrations [17,195,305,306]. The antiplasmodial activity of a series of nitrobenzenes and nitrofurans increases with their E 1 7 , log D, and their efficiency of inhibition of P. falciparum glutathione reductase (Pf GR) [195]. The relationship between the activity of ArNO 2 and the inhibition of erythrocyte GR is uncertain. Given the data available, P. falciparum ferredoxin:NADP + oxidoreductase (Pf FNR), and, possibly, ferredoxin may act as the most efficient reductants of ArNO 2 in plasmodia. P. falciparum thioredoxin reductase may be next to it according to quinone reductase activity [221]. Because P. falciparum, during its intraerythrocyte stage, adopts microaerophilic metabolism and relies mainly on anaerobic processes [307], an increase in the activity of ArNO 2 with E 1 7 may equally well reflect the rates of formation of free radicals or DNA-damaging hydroxylamines under the action of Pf FNR. On the other hand, the series of tetrazole-containing nitroimidazooxazines and quinoline-based nitroimidazoles exhibited antiplasmodial activity at micromolar concentrations [306], which is beyond the limits expected from their redox potentials (Table A1). This may be attributed to additional modes of their action arising from the hybrid structure of these compounds.

Conclusions
This review is not meant to be exhaustive in addressing the numerous and complex ways of interaction of nitroaromatic compounds with flavoenzymes and their cytotoxic/therapeutic consequences. However, one may conclude that with some minor exceptions, the single-electron reduction potential of nitroaromatics is the principal factor determining the rate of their single-electron reduction by flavoenzymes dehydrogenaseselectrontransferases and their FeS redox partners. This leaves relatively little space for the improvement of the enzymatic reactivity of compounds. On the other hand, although the studies of several two-electron transferring bacterial nitroreductases point to an important role of redox potential in the reactivity of nitroaromatics, the factors determining substrate specificity of nitroreductases from H. pylori, H. influenza, Leishmania, and Trypanosoma spp. remain unclear. This leaves more space for the improvement of the activity of compounds.
Low nitroreductase activity of flavoenzymes disulfide reductases diminishes their role in redox cycling or other modes of activation of nitroaromatics in the cell. However, low-M r TrxRs where FAD and catalytic disulfide function separately may more significantly contribute to the redox cycling of nitroaromatics. Another relevant problem is the exploitation of the bioreductive potential of a reduced selenylsulfide moiety of mammalian TrxR. In our opinion, this review may highlight the potential relevance of a number of flavoenzymes and/or their metalloprotein redox partners whose nitroreductase reactions were studied less comprehensively. Because of the definite perspectives of nitroaromatic compounds in the treatment of oxic tumors, the potential but insufficiently characterized targets of their action are the mitochondrial respiratory chain, in particular, its complex I and cytochromes P-450. Concerning the flavoenzymes of bacteria and parasites, more thorough studies of nitroreductase reactions of FprA and LipDH of M. tuberculosis, flavodoxin of H. pylori, TrxR, and Fd of P. falciparum, and PFOR/Fd/FOR systems of Trichomonas and Giardia spp. may foster the development of new groups of nitroaromatic agents or the repurposing of existing ones. Appendix A Table A1. Single-electron reduction midpoint potentials of nitroaromatic compounds determined by pulse radiolysis (E 1 7 ) or calculated according to the data of enzymatic single-electron reduction (E 1 7(calc.) ).   [10] unless specified otherwise, b taken from Refs. [11][12][13], c taken from Ref. [14], d taken from Ref. [15], e taken from