Isotopic labeling experiments that elucidate the mechanism of DNA strand cleavage by the hypoxia-selective antitumor agent 1,2,4-benzotriazine 1,4-di-N-oxide

Abstract

The 1,2,4-benzotriazine 1,4-dioxides are an important class of potential anticancer drugs that selectively kill the low-oxygen (hypoxic) cells found in solid tumors. These compounds undergo intracellular one-electron enzymatic reduction to yield an oxygen-sensitive drug radical intermediate that partitions forward, under hypoxic conditions, to generate a highly reactive secondary radical that causes cell killing DNA damage. Here we characterized bioreductively-activated, hypoxia-selective DNA-strand cleavage by 1,2,4-benzotriazine 1,4-dioxide. We found that one-electron enzymatic activation of 1,2,4-benzotriazine 1,4-dioxide under hypoxic conditions in the presence of the deuterium atom donor methanol-d₄ produced non-deuterated mono-N-oxide metabolites. This and the results of other isotopic labeling studies provided evidence against the generation of atom-abstracting drug radical intermediates and are consistent with a DNA-damage mechanism involving release of hydroxyl radical from enzymatically-activated 1,2,4-benzotriazine 1,4-dioxides.

Introduction

In cancer chemotherapy, there is a continuing need for the development of drugs that selectively target malignant tissue.¹ Toward this end, the poorly developed vascular system in tumors creates a low oxygen environment that may be exploited by some small molecules for the selective killing of cancer cells.² The 1,2,4-benzotriazine 1,4-di-N-oxides are an important class of hypoxia-selective cytotoxins.^2-7 Tirapazamine (1a) is the lead compound in this class of drugs, but many analogs in this structural family display similar bioactivities,^3-6 and the relatively new agent SN30000/CEN-209 (4) is advancing toward clinical development.^8,9

The medicinal action of tirapazamine and other 1,2,4-benzotriazine 1,4-dioxides involves intracellular one-electron enzymatic reduction to yield an oxygen-sensitive drug radical intermediate (2).^10-13 In normally-oxygenated tissue, 2 is simply oxidized by O₂ back to the parent compound 1.^13,14 However, under hypoxic conditions, the radical intermediate 2 partitions forward to generate a highly reactive secondary radical that kills cells via abstraction of hydrogen atoms from the 2′-deoxyribose phosphate backbone of DNA.^15-22 The mono-N-oxide 3 is the major drug metabolite generated in this process.^15,16,23 The exact nature of the ultimate DNA-damaging radical generated following bioreductive activation of the 1,2,4-benzotriazine 1,4-dioxides remains a topic of ongoing study, with two different possibilities considered in the recent literature. First, evidence has been presented that is consistent with the release of hydroxyl radical (HO•) from the protonated drug radical 2 (Scheme 1).^21,24-28 This mechanism adequately accounts for the DNA-damaging properties and the hypoxia-selective cytotoxicity of many different heterocyclic N-oxides including 1,2,4-benzotriazine 1,4-dioxides,^21,24-26,29 quinoxaline 1,4-dioxides,^30-33 1,2-dihydro-4-phenylimidazo[1,2-a]pyrido[3,2-e]pyrazine 5-oxides,³⁴ and phenazine 5,10-dioxides.³⁵ A second mechanistic proposal suggests that the tirapazamine radical 2 can undergo dehydration to generate a highly reactive benzotriazinyl radical 5 that abstracts hydrogen atoms from DNA.^36-40 To account for the bioactivities of other 1,2,4-benzotriazine 1,4-dioxide analogues such as 1b, the dehydration mechanism was expanded to include a family of hypothetical processes that generate aryl radical intermediates such as 6.³⁸

Scheme 1.

Isotopic labeling experiments have the potential to distinguish which of these two mechanistic pathways is operative. For example, the proposed generation of the aryl radical 6 from 2b in the presence of a deuterium atom donor will yield the corresponding deuterated drug metabolite 3c,^41,42 whereas release of hydroxyl radical from 2b will produce the non-deuterated metabolite 3b. Here we report the results of such isotopic labeling studies carried out in the context of compound 1b. We reasoned that 1b was good subject for mechanistic studies because this agent displays hypoxia-selective cytotoxicity comparable to tirapazamine (1a)³ and has been proposed to generate the DNA-damaging aryl radical 6.³⁸ However, the ability of 1b to carry out in vitro reductively-activated, hypoxia-selective DNA strand cleavage in a manner similar to tirapazamine had not been previously demonstrated. Therefore, before undertaking isotopic labeling studies, we characterized the DNA-damaging properties of 1b.^13,21,25,43 We found that 1b displays bioreductively-activated, hypoxia-selective DNA-cleaving properties analogous to the lead compound tirapazamine (1a). The results of several different isotopic labeling studies provided evidence against the generation of atom-abstracting drug radicals such as 6 following one-electron enzymatic activation of 1b under hypoxic conditions.

Materials and Methods

Materials

Materials were of the highest purity available and were obtained from the following sources: Cytochrome P450 reductase, NADPH, sodium phosphate, mannitol, DMSO, desferal, catalase, superoxide dismutase (SOD), tert-butyl nitrite, silica gel (0.04-0.063 mm pore size) for column chromatography, and silica gel plates for thin layer chromatography from Sigma Chemical Co. (St. Louis, MO); agarose from Seakem; HPLC grade solvents (acetonitrile, methanol, ethanol, tert-butyl alcohol, ethyl acetate, hexane, and acetic acid) from Fischer (Pittsburgh, PA); ethidium bromide from Roche Molecular Biochemicals (Indianapolis, IN); deuterated NMR solvents (deuterium oxide, D 99.96%, methanol-d₄, D 99.8%, N,N-dimethylformamide-d₇, D 99.5%; acetone-d₆, D 99.8%) were from Cambridge Isotope Laboratories (Andover, MA). Compounds 1a and 1b were prepared by routes described previously in the literature.^3,23,44 Compound 3b was prepared via the same general procedure used for 1b.³ and the spectral data of this product matched that published previously.³⁸

DNA Cleavage Assays

DNA strand cleavage reactions were carried out as described previously.^32,35 Briefly, in a hypoxic DNA-cleavage assay, agents such as 1a, 1b, or 1d (25-150 μM) were incubated with supercoiled plasmid DNA (33 μg/mL, pGL-2 Basic), NADPH (500 μM), cytochrome P450 reductase (33 mU/mL, where one unit is defined as the amount of enzyme required to cause the reduction of 1.0 μmole of cytochrome c by NADPH per minute at pH 7.4 at 37 °C), catalase (100 μg/mL), superoxide dismutase (10 μg/mL), sodium phosphate buffer (50 mM, pH 7.0), acetonitrile (0.5-3.3% v/v), and desferal (deferoxamine mesylate salt, 10 mM) under anaerobic conditions at 25 °C for 4 h. All components of the reactions except enzymes, NADPH, and DNA were degassed by three freeze-pump-thaw cycles. Enzymes, NADPH, and DNA were diluted with degassed water in an argon-filled glove bag to prepare stock solutions. Reactions were initiated by the addition of cytochrome P450 reductase, wrapped in aluminum foil to prevent exposure to light, and incubated in an argon-filled glove bag. Following incubation, the reactions were quenched by addition of 4 μL of loading buffer containing bromophenol blue dye and were loaded onto a 0.9% agarose gel. The gel was electrophoresed for approximately 2 h at 83 V in 1×TAE buffer. The gel was then stained in a solution of aqueous ethidium bromide (0.5 μg/mL) for 15 min, destained in fresh distilled water for 15 min, and the DNA in the gel visualized by UV-transillumination. The amount of DNA in each band was quantified using an Alpha Innotech IS-1000 digital imaging system. Buffers for the reactions performed in D₂O were prepared using the same mixture of phosphate buffer salts used in the H₂O experiments, as described by others.^45,46

Synthesis of 4-deutero-1,2,4-benzotriazine 1,4-dioxide (1c)

In a 25 mL round bottom flask, tert-butyl nitrite (0.25 mL, 2.10 mmol) was added to DMF-d₇ (2 mL) and heated to 65 °C. To this mixture, a solution of 1a (100 mg) in DMF-d₇ (2 mL) and was added dropwise over the course of 15 min. The reaction was monitored by thin layer chromatography every 10 min until all starting material was consumed. The reaction mixture was then cooled to room temperature and the DMF-d₇ was removed by high vacuum. Column chromatography on silica gel eluted with 1:1 ethyl acetate-hexane gave 1c as a yellow solid (23 mg, 25%). R_f 0.60 (100% ethyl acetate): ¹H NMR (DMSO, 500 MHz,): δ 8.39 (d, J = 8.5 Hz, 1H), 8.36 (d, J = 8.5 Hz, 1H), 8.14 (ddd, J = 8.5, 7.5, 1.0 Hz, 1H), 8.10 (ddd, J = 8.5, 7.5, 1.0 Hz, 1H). ¹³C-NMR (DMSO, 125.8 MHz): δ 142.2 (t, J = 33.3 Hz), 141.0, 136.0, 135.6, 133.3, 121.5, 119.4. HRMS (ES⁺, [M+H]) m/z calcd C₇H₅DN₃O₂ calcd mass 165.0523; actual mass 165.0515.

Synthesis of 1,2,4-benzotriazine (7)

The benzotriazine 7 was prepared via a modification of the procedures described by Fuchs et al. and Mason and Tennant.^23,47 To a solution of 3b (300 mg, 2.04 mmol) in 70% ethanol-water (40 mL) was added sodium dithionite (710 mg, 4.08 mmol). The resulting suspension was refluxed for 2 h, an additional aliquot of sodium dithionite (305 mg, 2.04 mmol) added, and the suspension refluxed for another 30 min, at which time all starting material was consumed (as judged by TLC). The solvent was removed by rotary evaporation and the resulting pale yellow residue purified by column chromatography on silica gel eluted with 30% ethyl acetate-hexane to yield 7 as pale yellow solid (154 mg, 58%). R_f = 0.57 (30% ethyl acetate/hexane): ¹H-NMR (DMSO, 500 MHz,): δ 10.13 (s, 1H), 8.61 (d, J = 8.5 Hz, 1H), 8.18-8.24 (m, 2H), 8.10 (ddd, J = 8.5, 6.5, 1.5 Hz, 1H). ¹³C-NMR (DMSO, 125.8 MHz): δ 154.1, 148.1, 140.1, 136.8, 132.2, 129.5, 129.0. Spectral data matched that reported by Boyd et al.⁴⁸

LC-MS/MS Analysis of Mixtures Generated by In Vitro Hypoxic Metabolism of 1b and 1c

In the LC-MS isotope incorporation assay, all stock solutions were made in D₂O except for 1b. Stock solutions used in the assays were prepared as follows: a solution of 1b (15 mM) was prepared by dissolving 10 mg of 1b in 4 mL acetonitrile, a solution of sodium phosphate (500 mM) was prepared from 409 mg Na₂HPO₄ and 254 mg NaH₂PO₄ dissolved in 10 mL D₂O), a solution of desferal (10 mM) was prepared by dissolving 3 mg desferal in 455 μL D₂O), a solution of NADPH (30 mM) was prepared by dissolving 1 mg of NADPH sodium salt in 40 μL D₂O, the stock of cytochrome P450 reductase (120 U/mL) was prepared by dissolving 3 μL of a 260.4 U/mL stock solution of enzyme dissolved in 3.5 μL D₂O, and a solution of CD₃OD (15 M) was prepared from 610 μL CD₃OD in 390 μL D₂O. All stock solutions except cytochrome P450 reductase were degassed by three cycles of freeze-pump-thaw on a vacuum line. For preparation of the assays, the di-oxide 1b (10 μL) was mixed with desferal (30 μL), sodium phosphate buffer (30 μL), NADPH (10 μL), and CD₃OD (10 μL of the 15 M CD₃OD stock in D₂O) and 209 μL of D₂O. Finally, the reactions were inititated by the addition of cytochrome P450 reductase stock solution (1 μL). The final volume of the assay mixture was 300 μL. These assays contained small amounts of residual exchangeable protium. The final mixtures contained approximately 99.5% D content in the water and, thus, small amounts (∼0.5%) of CD₃OH will be present alongside much larger amounts of CD₃OD. Nonetheless, these traces of protium in the deuterium atom donor are not expected to represent a confounding factor in the interpretation isotope incorporation assays. Bond enthalpies significantly favor atom abstraction from the methyl group of methanol rather than the hydroxyl group (see Supporting Information for further discussion of this topic). Experimental evidence presented below confirmed that atom abstraction in this system occurs at the methyl group of methanol. Specifically, the formation of significant amounts of metabolites 8 clearly indicated that •CD₂OH/D and •CH₂OH were generated. In addition, assays in which additional steps were taken to remove the residual exchangeable protium sources gave identical results (Figure S8).

After 4 h incubation under argon at 25 °C, the proteins in the assay mixtures were removed by centrifugation through Amicon Microcon (YM3) filters. The filtrate was analyzed by Beckman Coulter HPLC employing a C18 reverse phase Betabasic column (5 μm particle size, 150 Å pore size, 25 cm length, 4.6 mm i.d.) eluted with gradient starting from 5% B (0.1% trifluoacetic acid in acetonitrile) and 95% A (0.1% trifluoacetic acid in water) for 5 min followed by linear increase to 11% B in another 24 min, then the gradient was increased linearly to 80% of B within in 6 min, and finally 80% of B was used to wash the column for 10 min. A flow rate of 1.0 mL/min was used and the products were monitored by their UV-absorbance at 240 nm. LC/ESI-MS experiments were carried out using an ion trap mass analyzer, on a LCQ FLEET instrument (Thermo Fisher Scientific). Positive ion electrospray was used as the means of ionization. The heated inlet capillary temperature was 375 °C and electrospray needle voltage was 5 kV. Nitrogen sheath gas was supplied at 45 psi and the LC/ESI-MS analysis was done in the positive ion mode. Relative collision energies of 35% were used when the ion trap mass spectrometry was operated in the MS/MS mode. Parent ions were selected manually (164, 165, 148, 149, 132, 133, 178, 180) and subsequent tandem mass spectrometry was performed automatically by XCalibur software (Thermo Fisher Scientific).

Results and Discussion

Bioreductively-Activated, Hypoxia-Selective DNA Strand Cleavage by 1b

We synthesized 1b by a literature route involving treatment of 1a with t-butyl nitrite in DMF.^3,44 We prepared the expected metabolite 3b via the analogous reaction on 3a.³⁸ In addition, we prepared a second expected^15,16,23 metabolite 7 by treatment of 3b with sodium dithionite in ethanol-water. We employed a plasmid-based assay to measure the ability of 1b to cause bioreductively-activated DNA strand cleavage under hypoxic conditions. In this assay, oxidative DNA strand cleavage causes conversion of a supercoiled plasmid substrate (form I) to the open circular form II.^49-52 These two forms of plasmid DNA are then separated using agarose gel electrophoresis and visualized by staining with ethidium bromide.^49-52 This assay is especially useful for monitoring the direct DNA strand breaks generated by radical intermediates.^53,54 Direct strand breaks induced by radicals typically arise via abstraction of hydrogen atoms from the 2-deoxyribose phosphate backbone of DNA.^53,54 We used recombinant human NADPH:cytochrome P450 reductase to carry out one-electron reductive activation of 1b because this or a related enzyme⁵⁵ is thought to be responsible for the intracellular activation of 1,2,4-benzotriazine 1,4-dioxides.^43,56,57 For reactions carried out under hypoxic conditions, molecular oxygen was removed from stock solutions by freeze-pump-thaw degassing and the assay mixtures assembled and incubated in an inert atmosphere glove bag. Catalase, superoxide dismutase, and desferal were included as described previously^21,32 to prevent potential background strand cleavage stemming from the conversion of adventitious molecular oxygen to reactive oxygen species.

We found that 1b caused DNA strand cleavage when incubated with the plasmid substrate and the NADPH/NADPH:cytochrome P450 reductase enzyme system under hypoxic conditions (Figure 1). Across a range of concentrations (25-150 μM), the yields of DNA strand breaks generated by 1b were comparable to those produced by 1a and 1d, two biologically-active compounds whose DNA-strand cleaving properties have been characterized previously (Figure 1).^{3,13,21,25,43} Control experiments showed that neither 1b alone nor the NADPH/NADPH:cytochrome P450 reductase enzyme system (without 1b) caused significant amounts of DNA strand cleavage (Figure 1). Likewise, when 1b was incubated with the enzyme alone (no NADPH), NADPH alone (no enzyme) or with the complete enzymatic reducing system under aerobic conditions no significant amounts DNA strand cleavage were observed (Figure S1). Further control experiments showed that compounds 3b and 7, two major metabolites generated in the hypoxic metabolism of 1b (see below), did not cause DNA strand cleavage either alone or in the presence of the NADPH:cytochrome P450 reductase enzyme system under hypoxic conditions. We found that DNA strand cleavage by 1b was substantially inhibited (60-90%) by the classical radical-scavenging agents⁵⁸ ethanol, methanol, t-butanol, DMSO, and mannitol (500 mM, Figure 2), analogous to the results observed previously for 1a and 1d. Of special importance to the studies described below, we found that methanol-d₄ (500 mM) efficiently quenched the DNA-cleaving intermediate generated upon enzymatic activation of 1b (Figure S2). Overall, the reductively-activated, hypoxia-selective DNA-cleaving properties of 1b closely resemble those of tirapazamine and other 1,2,4-benzotriazine 1,4-dioxides.^13,21,25,43

Figure 1.

Bioreductively-activated DNA strand cleavage by 1b under hypoxic conditions. Panel A. Concentration-dependent DNA strand cleavage by 1b. Supercoiled plasmid DNA (pGL2-basic, 33 μg/mL) was incubated with 1b (25-150 μM), NADPH:cytochrome P450 reductase (33 mU/mL), NADPH (500 μM), catalase (100 μg/mL), superoxide dismutase (10 μg/mL), sodium phosphate buffer (50 mM, pH 7.0) and desferal (1 mM) under anaerobic conditions at room temperature for 4 h, followed by agarose gel electrophoretic analysis. Strand breaks per plasmid DNA molecule (S) were calculated using the equation S = –ln f₁ where f₁ is the fraction of plasmid present as form I. Lane 1, DNA alone (S = 0.35 ± 0.14); lane 2, enzyme system alone (S = 0.32 ± 0.11); lane 3, 1b alone (150 μM, S = 0.38 ± 0.17); Lanes 4-9, enzyme system + 1b at 25 μM (S = 0.49 ± 0.13), 50 μM (S = 0.76 ± 0.05), 75 μM (S = 1.07 ± 0.08), 100 μM (S = 1.30 ± 0.09), 125 μM (S = 1.42 ± 0.07), 150 μM (S = 1.62 ± 0.14). Panel B. Comparison of concentration-dependent DNA strand breakage by 1a, 1b, and 1d under hypoxic conditions. Strand cleavage by various concentrations (25-150 μM) of the three different compounds was carried out as described above. Background cleavage present in the untreated plasmid was subtracted to allow direct comparison of DNA cleavage yields between different experiments.

Figure 2.

Radical-scavenging agents inhibit bioreductively-activated DNA strand cleavage by 1b. Column 1 shows the strand cleavage yield generated by 1b (50 μM) under the standard hypoxic reaction conditions described in the Experimental Section and the Legend of Figure 1. Columns 2-6 illustrate the cleavage yields under those same standard conditions except in the presence of the indicated radical scavenging agent (500 mM). Background cleavage present in the untreated plasmid was subtracted to allow direct comparison of DNA cleavage yields between different experiments. The number of strand breaks per plasmid DNA molecule (S) was calculated using the equation S = –ln f₁ where f₁ is the fraction of plasmid present as form I.

In Vitro Bioreductive Metabolism of 1b

In the case of other 1,2,4-benzotriazine 1,4-dioxides that have been characterized, DNA strand cleavage is accompanied by generation of the corresponding 1,2,4-benzotriazine 1-oxide 3 as the major metabolite (Scheme 1).^{15,23,25,27,28} Here, we used LC-MS/MS analysis to characterize the products generated by in vitro, cell-free, hypoxic metabolism of 1b by NADPH:cytochrome P450 reductase in the presence of methanol as a hydrogen atom source. UV-vis detection of the resulting products (240 nm) revealed a mixture consisting of the starting material 1b alongside several new metabolites (Figure 3). MS/MS analysis of the products, along with comparison to authentic synthetic standards, showed that the 1,4-dioxide 1b eluted at 6.7 min, the expected mono-N-oxide 3b eluted at 18.8 min and, the expected “no-oxide” metabolite 7 eluted at 17.7 min (Table 1). The minor metabolite eluting at 14.7 min was tentatively assigned the structure 8a based upon its m/z of 178 in the LC-MS. Such a product could arise by a number of different mechanisms including addition of the methanol-derived radical, •CH₂OH, to 1b, followed by either dehydration or loss of HO•.⁵⁹ An analogous mono-N-oxide product was previously observed to arise via the reaction of methyl radical with 1a.⁵⁹

Figure 3.

Reverse-phase HPLC chromatogram (UV 254 nm) of the products generated by in vitro hypoxic metabolism of 1b in the presence of methanol.

Table 1.

LC-MS/MS properties of the mixture generated by in vitro metabolism of 1b under hypoxic conditions in the presence of CH₃OH.

Metabolites	Retention time	[M+H]+	MS/MS fragments
1b	6.5	164	147, 137, 119, 93, 92, 65
7	17.7	132	105, 77
3b	18.8	148	120, 93, 92, 65
8a	14.6	178	160, 132, 120, 104, 93, 92, 77, 65

In Vitro Bioreductive Metabolism of 1b in Phosphate-Buffered D₂O-methanol-d₄

The results described above set the stage for isotopic labeling experiments designed to probe whether bioreductive activation of 1b generates an atom-abstracting drug radical intermediate such as 6. As noted above, generation of an aryl radical such as 6 from 2b in the presence of a deuterium atom donor will yield the corresponding deuterated drug metabolite 3c,^41,42 whereas release of hydroxyl radical from 2b will generate the non-deuterated metabolite 3b. It may be important to reiterate that the experiments described above demonstrated that methanol-d₄ efficiently quenches bioreductively-activated DNA strand cleavage by 1b, thus providing a clear indication that the DNA-cleaving intermediate generated following bioreductive activation of 1b does indeed react with methanol-d₄.

Thus, 1b was incubated with the NADPH:cytochrome P450 reductase enzyme system under hypoxic conditions in phosphate-buffered D₂O containing methanol-d₄ (500 mM) and the metabolites generated in this reaction were analyzed by LC-MS/MS. We observed that the spectrum of metabolites generated in the presence of phosphate-buffered D₂O/methanol-d₄ was identical to that observed in the H₂O/methanol experiment shown in Figure 3 (Figure S3). Importantly, LC-MS analysis revealed no significant deuterium incorporation into the major metabolite 3 (Figure 4). The relative abundances of ions in the isotope cluster for the [M+H]⁺ ion of 3 observed in these experiments matches that expected to arise from the natural abundances of ²H, and ¹³C in a molecule of this molecular formula. Similarly, the no-oxide metabolite showed no deuterium incorporation. Our LC-MS analysis revealed that the mass-to-charge ratio of the metabolite eluting at 14.7 min generated in the D₂O/methanol-d₄ experiment was two Daltons greater than the analogous product generated in the presence of non-deuterated methanol (Figure S4). This observation is consistent with the notion that the 14.7 min metabolite contains a CH₂OH or CD₂OH fragment derived from methanol or methanol-d₄ (the deuterium on the hydroxyl group of the CD₂OD fragment presumably was exchanged with protons from trifluoroacetic acid in the HPLC eluent during the LC-MS analysis). As noted above, 8a and 8b could reasonably arise via multiple reaction pathways; therefore, we cannot consider the formation of these products diagnostic for any particular mechanism.

Figure 4.

Hypoxic metabolism of 1b in phosphate-buffered D₂O/CD₃OD does not lead to incorporation of deuterium into the metabolite 3. LC-MS of the mono-N-oxide metabolite 3 generated by hypoxic metabolism of 1b. Panel A, isotope cluster for the [M+H]⁺ ion of 3 generated in phosphate-buffered H₂O/CH₃OH (relative peak intensities for m/z 148:149:150 = 100:7.9:0.5). Panel B, isotope cluster for the [M+H]⁺ ion of 3 generated in phosphate-buffered D₂O/CD₃OD (relative peak intensities for m/z 148:149:150 = 100:8.2:0.6).

Bioreductive Metabolism of the Deuterium-Containing Analog 1c

Finally, we synthesized the deuterated analogue 1c via treatment of 1a with t-butyl nitrite in DMF-d₇.^44,60 Availability of this labeled compound enabled us to undertake isotopetracking experiments complementary to the solvent-labeling experiments described above. Specifically, the dehydration mechanism proposed³⁸ to generate the radical 6 would result in loss of deuterium content during the conversion of 1c to 3 when the reaction is carried out in the presence of protic solvents and hydrogen atom sources. In contrast, the hydroxyl radical mechanism will proceed with complete retention of label in the conversion of 1c to 3. LC-MS analysis of the mixture generated by NADPH:cytochrome P450 reductase-catalyzed metabolism of the deuterated analog 1c revealed no significant loss of deuterium content in the enzymatic transformation of 1c to the corresponding mono-N-oxide metabolite 3 (Figure S5). The observation that the deuterium content of the starting material 1c was retained in 3 had the side benefit of demonstrating that a deuterium label does not “wash out” of the metabolite 3 under our reaction conditions. This serves as a control for the methanol-d₄ experiments described in the previous section, ruling out the possibility, albeit remote, that, in those experiments, deuterium was incorporated into the metabolite 3, but subsequently was completely washed out prior to the mass spectrometric analysis.

Conclusions

The results presented here show that 1,2,4-benzotriazine 1,4-dioxide (1b) is able to carry out enzyme-activated, hypoxia-selective DNA strand cleavage analogous to the other benzotriazine di-N-oxides that have been characterized.^21,25,43 Two complementary isotopic labeling experiments provided evidence against the generation of atom-abstracting drug radicals such as 6 following one-electron reductive activation of 1,2,4-benzotriazine 1,4-dioxide under hypoxic conditions. Accordingly, the data is consistent with a mechanism involving release of the well-known DNA-damaging agent, hydroxyl radical from the reductively-activated drug intermediate 2b.

It may be useful to consider how the results described here mesh with the earlier suggestion that a spin-trapped adducts of 2 and 6 were detected following microsomal reduction of 1b in the presence of the nitrone spin-trap PBN (N-tert-butylphenyl-α-nitrone).³⁸ Importantly, recent computational analyses suggested that spin-trapping agents may not serve as “neutral observers” that report on the normal reactions of the bioreductively-activated radical intermediate 2.²⁸ Rather, spin-trapping agents have the potential to react directly with 2, in effect, short-circuiting its normal reactivity. Thus, various reactions of spin-trapping agents with 2b – rather than with 6 – may yield the observed ESR spectra.^28,37,38 It is difficult to rigorously define the structure of a spin-trapped radical using ESR and further work is required to determine the exact nature of the radicals captured by spin-trapping reagents during bioreduction of 1b and other analogs.

Overall, the results are consistent with the hypothesis that bioreductive activation of heterocyclic N-oxides under hypoxic conditions leads to the release of the highly cytotoxic reactive species, hydroxyl radical. This molecular mechanism may be relevant to the biological properties of many, structurally varied heterocyclic N-oxides under investigation as potential therapeutic agents for the treatment of cancer,^7,61,62 tuberculosis,^63,64 malaria,^65,66 leishmaniasis,⁶⁶ Chagas disease,⁶⁷ and bacterial infections.⁶⁸

Structures.