Zafirlukast Metabolism by Cytochrome P450 3A4 Produces an

Chem. Res. Toxicol. 2005, 18, 1427-1437

1427

Zafirlukast Metabolism by Cytochrome P450 3A4 Produces an Electrophilic r,β-Unsaturated Iminium Species That Results in the Selective Mechanism-Based Inactivation of the Enzyme Kelem Kassahun,†,‡ Konstantine Skordos,†,§ Ian McIntosh,‡ Donald Slaughter,‡ George A. Doss,| Thomas A. Baillie,‡ and Garold S. Yost*,§ Department of Pharmacology and Toxicology, University of Utah, Salt Lake City, Utah 84112-5820, and Department of Drug Metabolism, Merck Research Laboratories, West Point, Pennsylvania 19486 and Rahway, New Jersey 07065 Received April 1, 2005

Zafirlukast is a leukotriene antagonist indicated for the treatment of mild to moderate asthma, but the drug has been associated with occasional idiosyncratic hepatotoxicity. Structurally, zafirlukast is similar to 3-methylindole because it contains an N-methylindole moiety that has a 3-alkyl substituent on the indole ring. The results presented here describe the metabolic activation of zafirlukast via a similar mechanism to that described for 3-methylindole. NADP(H)-dependent biotransformation of zafirlukast by hepatic microsomes from rats and humans afforded a reactive metabolite, which was detected as its GSH adduct. Mass spectrometry and NMR data indicated that the GSH adduct was formed by the addition of GSH to the methylene carbon between the indole- and methoxy-substituted phenyl rings of zafirlukast. The formation of this reactive metabolite in human liver microsomes was shown to be exclusively catalyzed by CYP3A enzymes. Evidence for in vivo metabolic activation of zafirlukast was obtained when the same GSH adduct was detected in bile of rats given an iv or oral dose of the drug. On the basis of results with model peroxidases and of the structures of product alcohols from incubations containing H218O, it appeared that zafirlukast underwent dehydrogenation by two sequential one-electron oxidations. In addition, zafirlukast proved to be a mechanism-based inhibitor of CYP3A4 activity in human liver microsomes and in microsomes containing cDNA-expressed CYP3A4. The enzyme inhibitory property of zafirlukast was selective for this enzyme among all of the P450 enzymes that were tested in human liver microsomes. The inactivation was characterized by a KI of 13.4 µM and kinact of 0.026 min-1. In summary, zafirlukast dehydrogenation to an electrophilic R,β-unsaturated iminium intermediate may be associated with idiosyncratic hepatotoxicity and/or cause drug-drug interactions through inactivation of CYP3A4.

Introduction Zafirlukast (Accolate) is a leukotriene receptor antagonist which is used clinically to treat mild to moderate asthma. Several isolated cases of idiosyncratic hepatotoxicity have been associated with the use of zafirlukast (1-3). Clinical reports describing the hepatotoxicity characterize it as an idiosyncratic drug reaction based on several factors including the following: other pathophysiological causes of hepatic disease were eliminated; liver biopsy results were consistent with toxic injury; systemic hypersensitivity was observed in at least one case; and rechallenge with zafirlukast produced recurrent hepatitis (1). Structurally, zafirlukast can be considered to be a derivative of 3-methylindole since it contains an N* To whom correspondence should be addressed. Tel: 801-581-7956. Fax: 801-585-3945. E-mail: [email protected]. † These two authors contributed equally to this research and should both be considered “first authors”. § University of Utah. ‡ Merck Research Laboratories, West Point. | Merck Research Laboratories, Rahway.

methylindole moiety that is substituted at position 3 of the indole ring. The potent pneumotoxic compound 3-methylindole is known to be metabolized to a reactive electrophilic intermediate (3-methyleneindolenine) by P4501-mediated dehydrogenation (4). The structure of the intermediate was inferred through the characterization of its GSH adduct (5). Additional reactive intermediates that are formed by oxygenation of the indole moiety may also contribute to the toxicities of 3-methylindole (6, 7). The dehydrogenation mechanism is believed to proceed by an initial hydrogen atom abstraction from the methyl group followed by a subsequent one-electron oxidation (8). Both protein and DNA adducts of 3-methyleneindolenine have been documented (9, 10), and these adducts may initiate lung-selective toxicity of 3-methylindole. In addition to reactive intermediate formation that may ultimately lead to toxicity, the dehydrogenation process also may alter drug metabolism. Reactive intermediates formed in the P450 active site may covalently modify the enzyme, producing mechanism-based inacti1 Abbreviations: CID, collision-induced dissociation; P450, cytochrome P450; psi, pounds per square inch.

10.1021/tx050092b CCC: $30.25 © 2005 American Chemical Society Published on Web 08/12/2005

1428

Chem. Res. Toxicol., Vol. 18, No. 9, 2005

vation. This process has been demonstrated for the compound furafylline with the enzyme CYP1A2 (11) and may occur during the dehydrogenation of 3-methylindole by CYP2F1 (12). Given the evidence that zafirlukast may elicit idiosyncratic hepatotoxicity and the demonstrated potential for P450 inhibition (13, 14), an investigation of the possibility of reactive metabolite formation was warranted. It was hypothesized that zafirlukast may be susceptible to a P450-mediated dehydrogenation reaction to form a reactive iminium intermediate, in a manner similar to that observed for 3-methylindole. This hypothesis was tested using both in vitro and in vivo systems. Thus, the objectives of the current study were (1) identification and structural characterization of reactive intermediates resulting from zafirlukast metabolism and elucidation of the mechanism of their formation, (2) identification of the specific P450 isoforms responsible for reactive metabolite formation, and (3) investigation of the potential for zafirlukast to serve as a mechanism-based inhibitor of the P450 involved.

Materials and Methods Chemicals. NADP(H), reduced GSH, formic acid, and potassium phosphate were obtained from the Sigma Chemical Company (St. Louis, MO). Zafirlukast was a generous gift from AstraZeneca Pharmaceuticals. All other chemicals were of general laboratory reagent grade or higher. Incubations with Human and Rat Liver Microsomes. The microsomal metabolism of zafirlukast was studied using liver microsomal preparations (Xenotech LLC, Kansas City, KS) from rats and humans. Incubations contained 0.1 M potassium phosphate buffer, zafirlukast (50 µM), microsomal protein (2 mg/ mL), and NADP(H) (1 mM) in a final volume of 1 mL. Incubations also were carried out in the presence of GSH (5 mM). Additional incubations (incubation volume of 0.5 mL) were performed in the presence of H218O [48%]). Incubations were carried out at 37 °C for 45 min in a shaking water bath. Incubations that lacked either NADP(H) or GSH served as negative controls. Reactions were terminated by the addition of 2 vol of acetonitrile. After centrifugation, the supernatant from each incubation was removed and evaporated to dryness. The residue was reconstituted in 30% acetonitrile in water (200 µL), vortex-mixed, and centrifuged. Aliquots (10 µL) of the final supernatant were analyzed by LC/MS and LC/MS/MS. The relative amounts of M4 and M7 formed in microsomal incubations with and without GSH were determined by comparing the LC/MS/MS peak area ratios of M4 and M7 (monitoring the transition m/z 592.3 f 353.1) to that of the internal standard (m/z 431.2 f 306.1). Incubations with Peroxidases. The peroxidase-mediated metabolism of zafirlukast (50 µM) was studied using horseradish peroxidase (20 units/mL) or myeloperoxidase (5 units/mL). Incubations were performed in the presence of phosphate buffer (50 mM containing 0.5 mM EDTA), H2O2 (50-100 µM), and GSH (5 mM) in a total volume of 1 mL. Following incubation (37 °C) for 60 min, the reaction was terminated, and the samples were processed as described above for liver microsomes. Incubations with Microsomes Containing cDNA-Expressed Enzymes. The individual P450s (3A4, 2B6, 2D6, 1A2, 2C9, and 2C19) were obtained from BD Biosciences (formerly Gentest, San Jose, CA) as insect cell microsomes containing both P450 reductase and cytochrome b5. Incubations were conducted with 50-100 pmol P450 and 50 µM zafirlukast in 0.1 M potassium phosphate buffer, pH 7.4, in the presence and absence of 2 mM GSH, and 2 mM NADP(H). All incubation components excluding NADP(H) were prepared and mixed on wet ice. The incubation mixtures were transferred to a 37 °C water bath and equilibrated for 5 min prior to the addition of NADP(H) to

Kassahun et al. initiate the reaction. The incubations were stopped after 20 min by the addition of an equal volume of ice-cold acetonitrile. The precipitated protein was removed by centrifugation in a Beckman microcentrifuge (14 000g). The supernatant was concentrated in a Savant Speed Vac to an approximate volume of 200 µL and stored at -20 °C until LC/MS analysis. Inhibition Studies. The effect of P450 isozyme-selective chemical inhibitors on the formation of the GSH conjugate of zafirlukast in human liver microsomes was studied using the following inhibitors at the indicated concentration: fluvoxamine (10 µM; CYP1A2), sulfaphenazole (5 µM; CYP2C9), quinidine (10 µM; CYP2D6), and ketoconazole (1 µM; CYP3A). The inhibitors, which were dissolved in 50% acetonitrile in water, were incubated individually with human liver microsomes as described above under Incubations with Human and Rat Liver Microsomes. The same protocol was used to determine the enzymes involved in the biotransformation of zafirlukast to its hydroxylated metabolites using a zafirlukast concentration of 5 µM. A low concentration of zafirlukast was used to simulate the low plasma concentrations that would be produced by normal dosages of this drug. The extent of inhibition was estimated by comparing full-scan MS peak height (extracted ion chromatograms) of the metabolite obtained from the control incubation (no inhibitor) with those obtained in the presence of an inhibitor. In Vivo Studies. Two groups of three male Sprague-Dawley rats equipped with a bile duct cannula received intravenous (5 mg/kg) or oral (20 mg/kg) doses of zafirlukast. For oral administration, the compound was suspended in 0.5% methylcellulose, while samples for intravenous dosing were dissolved in DMSO. The bile samples were filtered and diluted with an equal volume of water and introduced into the LC/MS system. Identification of Metabolites. Metabolites were identified by electrospray LC/MS/MS analysis using the Finnigan LCQ mass spectrometer. Extracts from in vitro incubations and bile samples were introduced into the mass spectrometer using a Zorbax eclipse XDB-C8 column (4.6 mm × 25 cm, 5 µm). The mobile phase consisted of 0.1% aqueous formic acid (solvent A) and 0.1% formic acid in acetonitrile (solvent B), and the gradient was as follows: 0-2 min:,10% B; 35 min, 50% B; 40 min, 80% B; and 45 min, 10% B. The flow rate was 1 mL/min, and the LC effluent was split such that 200 µL went to the mass spectrometer and the remaining to a UV detector. Full-scan mass spectra were obtained in positive ion mode, and product ion spectra were generated for components of interest by comparing profiles obtained from treatments with the appropriate control. The GSH adduct formed in vitro was isolated by HPLC (using the above conditions) from incubations containing horseradish peroxidase and analyzed by NMR (using a Varian 500 MHz instrument and CD3OD as the solvent). Another LC/MS/MS system (a Finnigan TSQ Quantum mass spectrometer) was used to analyze phase I metabolites of zafirlukast. The extract from the incubations was injected onto a Phenomenex Synergi HydroRP C18, 150 mm × 2 mm, 4 µm column at a flow rate of 200 µL/min. The mobile phase consisted of 0.1% aqueous formic acid (solvent A) and 0.1% formic acid in acetonitrile (solvent B), and the gradient was as follows: 0 min, 5% B; 2 min, 5% B; 30 min, 90% B; 40 min, 90% B; and 41 min, 5% B. Positive ions were monitored over the full-mass range of m/z 500-1000. The instrument was operated with a sheath gas pressure of 60 psi and an auxiliary gas pressure of 20 psi. The capillary temperature was 350 °C, and the ESI spray voltage was 3.0 kV. Time-Dependent Inactivation of CYP3A4 by Zafirlukast. Primary incubation mixtures were prepared containing 0.1 M potassium phosphate buffer, pH 7.4, 2 mM NADP(H), and zafirlukast (0-50 µM). Insect cell microsomes containing CYP3A4, cytochrome P450 reductase cytochrome b5, and the primary incubation mixture were prewarmed at 37 °C for 5 min. The preincubation reaction was initiated by the addition of the microsomal solution to the primary incubation mixture (final concentration of CYP3A4 was 0.5 µM). Control incubations were performed in the absence of NADP(H). At intervals of 0, 2, 5,

Zafirlukast Inactivates CYP3A4

Chem. Res. Toxicol., Vol. 18, No. 9, 2005 1429

Figure 1. LC/MS/MS spectrum obtained by CID of the MH+ ion (m/z 881) of the GSH conjugate of zafirlukast detected in human liver microsomal incubations. The putative assignments of characteristic fragment ions are shown. 10, and 20 min, aliquots of the primary incubation mixtures (10 pmol CYP3A4) were transferred to secondary incubation mixtures. Secondary incubation mixtures contained 0.1 M potassium phosphate, pH 7.4, 2 mM NADP(H), and 0.2 mM testosterone. Secondary incubations were conducted for 20 min at 37 °C and stopped by the addition of ice-cold acetonitrile. Testosterone and 6β-hydroxytestosterone were extracted from the incubation mixtures with 2 vol of dichloromethane. 6βHydroxytestosterone formation was quantified using a standard HPLC assay with UV detection and utilized as an index of CYP3A4 activity. Effect of Glutathione, Excess Substrate, and Dialysis on the Time-Dependent Inactivation of CYP3A4 by Zafirlukast. Primary incubation mixtures were prepared as described above, except that reduced GSH (4 mM) or testosterone (0.2 mM) were included in the primary incubations. Secondary incubations and quantification of 6β-hydroxytestosterone were performed exactly as described above. To determine the reversibility of the time-dependent inhibition by zafirlukast, a dialysis experiment was conducted as follows. Primary incubations using a zafirlukast concentration of 50 µM were performed as described above for 20 min at 37 °C. Control incubations were performed in the absence of zafirlukast. At the conclusion of the primary incubation period, the entire contents of the reaction were placed into a Slide-A-Lyzer dialysis cassette (Pierce Biotechnology, Inc., Rockford, IL). The primary incubation mixtures were dialyzed against 0.1 M potassium phosphate buffer, pH 7.4, at 4 °C overnight. Secondary incubations and quantification of 6β-hydroxytestosterone were performed exactly as described above. P450 Isozyme Selectivity of the Time-Dependent Inhibition by Zafirlukast. The isozyme selectivity of the timedependent inhibitory effect of zafirlukast was studied using human liver microsomal testosterone 6β-hydroxylase (CYP3A4), diclofenac 4′-hydroxylase (CYP2C9), (S)-mephenytoin 4′-hydroxylase (CYP2C19), bufuralol 1′-hydroxylase (CYP2D6), and phenacetin-O-dealkylase (CYP1A2) activities. The following time-dependent inhibitors were used as positive controls: L-754394 (CYP3A4 and CYP2D6), tienilic acid (CYP2C9), ticlopidine (CYP2C19), and furafylline (CYP1A2). NADP(H)-fortified human liver microsomes (2 mg/mL final concentration) were incubated in the presence of solvent alone, zafirlukast (10 µM), or positive control (1-10 µM) in a 37 °C shaking water bath (final volume of 25 µL). After a 30 min incubation period, preheated buffer (225 µL) containing NADP(H) (1 mM) and the marker substrate (concentrations were greater than 5-fold higher than the Km values of each substrate/enzyme pair) were added, and the reaction was allowed to proceed for a period of 10-30 min. Subsequently, the reaction was terminated with the addition of 2 vol of acetonitrile containing the appropriate internal standard, and the samples were processed and analyzed by LC/MS/MS using a previously published method (15).

Results Metabolic Activation of Zafirlukast by Human and Rat Liver Microsomes. NADP(H)-dependent biotransformation of zafirlukast by hepatic microsomes from rats and humans afforded a reactive metabolite, which was detected as its GSH adduct. The GSH adduct formed by microsomes from the two species, and by horseradish peroxidase, had a molecular mass of 880 Da (parent -H + GS). The conjugate was shown by MS/MS and NMR analysis to be formed by the addition of the elements of GSH to the methylene carbon between the indole and phenyl rings (Figures 1 and 2). The GSH adduct was produced in reasonably high amounts by horseradish peroxidase, so this source was used for the NMR experiments. All other spectral and chromatographic techniques confirmed that the adduct from horseradish peroxidase was identical to the adduct produced by human and rat hepatic microsomes and one of the adducts that was identified from the bile of rats dosed with zafirlukast. There was no evidence (selected ion monitoring at m/z 881) for the formation of a GSH adduct at the C-2 position of the indolenine electrophile, or for additional adducts with this molecular mass. In addition, there were five GSH adducts that had an m/z of 899 that presumably were formed by addition of GSH to epoxides on the aromatic moieties. The NMR assignments for the parent compound (lower trace, Figure 2) were made by a combination of a COSY and NOE difference spectra. The key observation in the identification of the structure of the GSH conjugate was the loss of the typical signal for the CH2 bridge (4.0 ppm in the parent) and the appearance of a new signal at 5.95 ppm (upper trace, Figure 2) which is consistent with the addition of GSH on this group. The spectrum of the GSH conjugate showed two closely related species, in agreement with the presence of a diastereomeric mixture (apparently not separable by the HPLC conditions used) that was produced from the creation of a new chiral center at the benzylic carbon. Figure 3 shows the effects of enzyme-selective P450 inhibitors on the formation of the GSH adduct of zafirlukast. The selective CYP3A4 inhibitor ketoconazole almost completely inhibited the formation of the adduct, while inhibitors selective for CYP1A2, CYP2C9, and CYP2D6 had no effect. Consistent with these results, we also measured the formation of the GSH adduct by the 3A4, 2E1, 2B6, 2D6, 1A2, 2C9, and 2C19 recombinant enzymes. Only CYP3A4 was able to catalyze the

1430


Kassahun et al.

Figure 2. 1H NMR spectra (500 MHz) of the GSH conjugate of zafirlukast (upper trace), isolated from in vitro incubations, and zafirlukast (lower trace) in CD3OD.

formation of the GSH adduct from zafirlukast (data not shown). Metabolic Activation of Zafirlukast by Peroxidases. Zafirlukast readily formed (∼10% conversion in 60 min) a GSH adduct with molecular mass of 880 Da upon incubation with horseradish peroxidase (Figure 4). This adduct had identical HPLC retention time and product ion spectrum to that obtained from the GSH conjugate formed in human liver microsomes. The same conjugate was generated when zafirlukast was incubated with human myeloperoxidase (data not shown). Detection of GSH Adducts in Vivo. A total of five GSH conjugates were detected in the bile of rats dosed with zafirlukast at 5 mg/kg iv or 20 mg/kg po (Figure 5). The metabolite with an MH+ ion at m/z 881 was identical to that detected in vitro (incubations with liver microsomes or peroxidases), based on retention time and MS/MS spectrum. The remaining four isomeric GSH conjugates had a molecular mass of 898 Da and thus may have been derived from arene oxide metabolites of zafirlukast. Detection of Hydroxylated Metabolites in Vitro. A total of seven oxidative metabolites (parent + 16 Da) were detected in incubations of zafirlukast with NADP(H)-fortified human liver microsomes (Figure 6). Structural assignments for these hydroxylated metabolites were based on information from prior investigation of zafirlukast metabolism (16, 17) and the MS/MS spectra of the metabolites obtained in the current investigation. Thus, the two major metabolites M1 and M4 (Figure 7) were identified as being products of P450-mediated oxidation reactions involving the cyclopentyl ring and the N-methyl moiety of the indole ring, respectively. The combined weight of evidence provided structural assignments that were somewhat tenuous, but still consistent with all available data. M2 and M3, like M1, were believed to be hydroxylated on the cylopentyl ring based on the fragment ion at m/z 464 (480 in the other four

metabolites). The site of hydroxylation of M5 and M6 is likely to be the toluyl ring, because an MS/MS diagnostic fragment, the ion at m/z 405 (421 in all other five metabolites) was present. The site of hydroxylation in M7 was determined to be the methylene carbon between the indole- and methoxy-substituted phenyl rings based on the structurally informative fragment at m/z 353. Because the ion at m/z 353 is present in both M4 and M7, the structural assignments of these two metabolites was made on the basis that (1) N-methylindole hydroxylation (formation of M4) is known to be the major pathway of zafirlukast biotransformation in vivo in humans and in vitro in human liver microsomes (similar to the current finding) and (2) formation of M4 is known to be catalyzed by CYP2C9 (similar to the current finding, see below), whereas M7 appears to be formed exclusively by CYP3A4 (see below). The stability of M4 may appear initially to be unususal for a carbinolamine, but demethylation of N-methylindoles is not a common metabolic pathway. The lone pair of electrons on the nitrogen participates in indole aromaticity and is not available to be protonated and facilitate the loss of formaldehyde from the carbinolamine. Therefore, we would predict that the carbinolamine M4 should be stable. If M7 was formed predominantely by hydration of the unsaturated iminium ion, addition of excess GSH should decrease the relative amount of this alcohol, and not change the relative amounts of M4. This hypothesis was evaluated by comparing the LC/MS/MS peak area ratios of M4 and M7 to that of the internal standard, obtained from extracts of incubations conducted with excess GSH (5 mM) or incubations without GSH. The results showed that the formation of M7 was reduced by 33% in the presence of GSH, while the levels of M4 were unaffected (data not shown). The effect of enzyme-selective P450 inhibitors on the formation of the hydroxylated metabolites of zafirlukast was determined by comparing LC/MS peak height ratios



Figure 3. LC/MS extracted ion (m/z 881) chromatograms showing the detection of the GSH adduct of zafirlukast in NADPHfortified human liver microsomes (labeled control) and the effects of form-selective P450 inhibitors on the formation of the adduct.

of analyte to internal standard obtained in the presence of various inhibitors with that generated from control (no inhibitor) incubations. The selective CYP3A4 inhibitor ketoconazole inhibited the formation of M7 by g94%. Fluvoxamine, sulfaphenazole, and quinidine inhibited the formation of M4 by e53%. In agreement with these results, among the cDNA-expressed P450 enzymes tested, M7 was formed only by CYP3A4, while M4 was formed predominantly by CYP2C9. The other enzymes that

showed some activity with respect to formation of M4 were CYP3A4, 2D6, 2C19, and 1A2 (data not shown). Similarly, CYP3A4 was also able to catalyze the formation of the other hydroxylated metabolites. Metabolism in the Presence of H218O. Incubations were carried out in the presence of H218O to determine the extent of 18O incorporation in hydroxylated metabolites of zafirlukast. LC/MS analysis of the extracts of incubations carried out in the presence of H218O revealed

1432


Kassahun et al.

Figure 4. LC/MS extracted ion (m/z 881) chromatogram showing the formation of the GSH adduct of zafirlukast in incubations containing horseradish peroxidase. Under the conditions of LC/MS/MS analysis using the mass range of 450-950 Da, this conjugate was the only significant compound identified.

Figure 5. LC/MS extracted ion chromatograms showing GSH adducts derived from zafirlukast metabolism in vivo in rats. The putative molecular ion of GSH added to the R,β-unsaturated imine would be m/z 881 (upper panel) and of GSH added to epoxides on the aromatic moieties would be m/z 899 (lower panel).

Figure 6. Reconstructed ion current (m/z 592, parent + 16 Da; m/z 576, parent) chromatogram showing hydroxylated metabolites formed upon incubation of zafirlukast in human liver microsomes in the presence of NADPH.

that, of the seven hydroxylated metabolites that were detected, only M7 had incorporated 18O. The percent 18O incorporation in M7 was 81%. Time-DependentInhibitionofCYP3A4byZafirlukast. Insect cell microsomes containing CYP3A4 were prein-

cubated with various concentrations of zafirlukast (050 µM). The testosterone 6β-hydroxylase activity of CYP3A4 was then determined from a subsequent incubation. Zafirlukast produced time-, concentration-, and NADP(H)-dependent inactivation of CYP3A4. The en-



Figure 7. LC/MS/MS product ion spectra obtained by CID of the MH+ ion (m/z 592) of the hydroxylated metabolites of zafirlukast. The spectra of M2 and M3 were similar to that of M1, and the spectrum of M6 was identical to that of M5. The putative assignments of characteristic fragment ions are shown.

zyme was maximally inactivated to 20% of control activity. These results are shown graphically in Figure

8, as a plot of the log of 3A4 activity remaining versus preincubation time. The slopes of these lines represent

1434


Kassahun et al. Table 1. Time-Dependent Inhibition of Human Liver Microsomal Enzymes by Zafirlukast, Compared to Prototypical Inhibitors of the Enzymes % inhibition enzyme

zafirlukast

positive controla

3A4 2D6 2C9 2C19 1A2

51 4 6 0 4

89 80 86 26 80

a The enzymes and their respective inhibitors were 3A4 and 2D6, L-754394; 2C9, tienilic acid; 2C19, ticlopidine; and 1A2, furafylline. The concentration of zafirlukast was 10 µM, while that of the positive controls ranged from 1 to 10 µM.

Figure 8. Time- and concentration-dependent inactivation of P450 3A4 by zafirlukast: (b) 0 µM zafirlukast, (9) 10 µM, ([) 20 µM, (2) 35 µM, and (1) 50 µM. Values shown are the means of 3-6 determinations, and in all cases, the standard deviation was less than 18%. The inset is a Kitz-Wilson double reciprocal linearization, from which KI and kinact were determined.

Figure 9. Effects of GSH, testosterone, and dialysis on the inactivation of CYP3A4 by zafirlukast.

the observed first-order rate constants (kobs) of the inactivation reaction at a given zafirlukast concentration. The reciprocal of kobs was plotted against the reciprocal of the zafirlukast concentration to obtain the inactivation constants KI (inactivator concentration at half-maximal inactivation rate) and kinact (inactivation rate constant). The KI was determined to be 13.4 µM, and the kinact was determined to be 0.026 min-1. The partition ratio, or ratio of noninactivating catalytic events to inactivation events, was also determined. Briefly, the percent of CYP3A4 activity remaining following preincubation at the longest time-point tested, was plotted against the ratio of the concentrations of zafirlukast to CYP3A4. A partition ratio of ∼44 was determined from the intercept of the regression line formed at lower zafirlukast concentrations and the straight line formed from higher concentrations. Substrate and Glutathione Protection. In classical mechanism-based enzyme inactivation, the presence of excess alternative substrate in the preincubation mixtures will protect against the inactivation reaction (18). Additionally, inactivation is believed to occur prior to release of the reactive species from the enzymes active site. Therefore, inclusion of GSH in the primary incubation reaction will not protect the enzyme (18). Testosterone was included in primary incubations of CYP3A4 with 100 µM zafirlukast. An increase in the remaining activity of the enzyme was observed, when compared to preincubations performed in the absence of testosterone (Figure 9). Conversely, the inclusion of 2 mM GSH in the preincubation mixtures did not protect the enzyme to an appreciable extent (Figure 9). These data are consistent

with the conclusion that zafirlukast is a mechanismbased inactivator of CYP3A4. Irreversible Nature of the Inactivation of CYP3A4 by Zafirlukast. Following the conclusion of primary incubation of CYP3A4 in the presence and absence of zafirlukast, the reaction mixtures were dialyzed against 0.1 mM potassium phosphate buffer overnight. Under these conditions, inhibiting species that are not covalently bound to the enzyme are free to diffuse away, restoring enzyme function. Incubation mixtures containing zafirlukast that were dialyzed contained 30% of control enzyme activity (Figure 9). As stated previously, CYP3A4 was maximally inactivated to 20-25% of control enzyme activity. Therefore, the CYP3A4 inactivation process by zafirlukast did not appear to be reversible under these conditions. Isozyme Specificity of the P450 Inactivation by Zafirlukast. The isozyme specificity of the time-dependent inhibitory effect of zafirlukast was studied using human liver microsomal P450 activities shown in Table 1. Zafirlukast had little or no effect on the enzymes that were tested, other than CYP3A4. As expected, the respective positive controls were effective time-dependent inhibitors.

Discussion The primary objective of the present study was to determine if zafirlukast is subject to metabolic activation in a fashion similar to that described for the pneumotoxin 3-methylindole, because zafirlukast contains an N-methylindole moiety that is substituted at position 3 of the indole ring and thus can be considered to be a derivative of 3-methylindole. NADP(H)-dependent biotransformation of zafirlukast by hepatic microsomes from rats and humans gave rise to a chemically reactive intermediate, which was detected as its GSH conjugate. The structure of the GSH adduct was elucidated using MS and NMR spectroscopy, and it was determined that the adduct was formed by the addition of the elements of GSH to the methylene carbon between the indole- and methoxysubstituted phenyl rings (Figure 2). Such a substitution in the zafirlukast molecule creates a chiral center, and the NMR spectrum (Figure 2) of the adduct showed two closely related molecules. The presence of these similar adducts is consistent with the formation of a diastereomeric mixture by horseradish peroxidase (apparently not separable by the HPLC conditions used). Several P450 enzymes were studied for their ability to activate zafirlukast, but only CYP3A4 was capable of catalyzing the formation of the GSH conjugate. Evidence for in vivo metabolic activation of zafirlukast was ob-



Figure 10. Proposed scheme for CYP3A4-mediated metabolism of zafirlukast to the reactive iminium intermediate identified in the present study.

tained when the GSH adduct was detected in the bile of rats given an iv or oral dose of the drug. In addition, both horseradish peroxidase and myeloperoxidase readily converted zafirlukast to the same reactive metabolite as that formed by CYP3A4. A proposed mechanism for CYP3A4-mediated formation of the reactive metabolite of zafirlukast is shown in Figure 10. Thus, CYP3A4mediated one-electron oxidation, either via hydrogen atom abstraction from the methylene carbon (pathway 2) or electron abstraction from the nitrogen (pathway 3), followed by a second one-electron oxidation, would produce the highly electrophilic R,β-unsaturated iminium species shown in Figure 10. This electrophilic intermediate would then spontaneously react with GSH or water to afford the GSH adduct and the hydroxylated metabolite, M7, respectively. Further evidence for the R,βunsaturated iminium species as a reactive metabolite of zafirlukast was obtained from the pattern of 18O incorporation in the hydroxylated metabolites of zafirlukast, as discussed below. A mechanism employing two sequential single-electron oxidations would also be consistent with the fact that the same reactive metabolite is generated by horseradish peroxidase and myeloperoxidase, enzymes that are known to oxidize substrates by this mechanism. A similar mechanism has been invoked

for the peroxidase-catalyzed metabolism of acetaminophen to N-acetyl-p-benzoquinone imine (19). To gain insights into the mechanism of formation of the reactive metabolite, microsomal incubations were carried out in the presence of H218O and the extent of 18 O incorporation into hydroxylated metabolites of zafirlukast was determined. Seven hydroxylated metabolites were detected in human liver microsomes (Figure 6), and the major metabolites (M1 and M4) were identified on the basis of MS/MS spectra and published data from previous investigations of zafirlukast metabolism (16, 17). These two alcohols were products of P450mediated oxidation reactions involving the cyclopentyl ring and the N-methyl moiety of the indole ring, respectively. N-Methylindole hydroxylation (formation of M4) is a major pathway of zafirlukast biotransformation in humans and is known to be catalyzed by CYP2C9 (17). The hydroxylated metabolite M7 was identified as the benzylic alcohol that is shown in Figure 10 on the basis of its MS/MS spectrum and the fact that it is a unique metabolite of CYP3A4. From the seven hydroxylated metabolites that were detected, only M7 incorporated 18O from water in the medium. Lack of 18O incorporation into M4, coupled with the absence of the corresponding GSH adduct, argues

1436


against the formation of the exomethylene iminium intermediate depicted in pathway 1 (Figure 10). Furthermore, while M4 is a product of CYP2C9 catalysis, the reactive metabolite was formed exclusively by CYP3A4 (Figure 3). The incorporation of 18O (81%) into the benzylic alcohol, M7, suggests that the R,β-unsaturated iminium species is indeed the reactive metabolite of zafirlukast that produces the GSH conjugate. Therefore, it can be concluded that, while M4 is formed solely by “oxygen-rebound” (pathway 4, Figure 10), M7 is derived both from “oxygen-rebound” (minor component; pathway 6, Figure 10) and hydration of the R,β-unsaturated iminium reactive intermediate (major component; pathway 7, Figure 10). The reverse reaction, that is, dehydration of M7, is not expected to contribute significantly to the formation of the iminium reactive intermediate because it is unlikely that this alcohol would eliminate water more readily than the GSH adduct, which appeared to be stable. This conclusion is in agreement with data from 3-methylindole in which it was shown that indole-3-carbinol did not significantly contribute to the formation of the reactive intermediate 3-methyleneindolenine at physiological pH (8). In addition, the finding that the amount of M7 was reduced by 33% in the presence of excess GSH in the microsomal incubation (while the level of M4 was not affected) is consistent with the view that the iminium intermediate is the source of both the GSH adduct and M7. In the present study, it was also shown that zafirlukast was an inhibitor of CYP3A4 via a process determined to be mechanism-based. The inactivation of CYP3A4 was time- and concentration-dependent (Figure 8) and required NADP(H). The inactivation showed pseudo-firstorder kinetics with a KI, kinact, and partition ratio of 13.4 µM, 0.026 min-1, and 44, respectively. The inactivation of CYP3A4 by zafirlukast was found to be irreversible because extensive dialysis could not restore enzyme activity (Figure 9). In addition, an exogenous nucleophile, GSH, did not attenuate enzyme inactivation, while an alternative substrate afforded partial protection. However, the studies presented herein did not demonstrate that zafirlukast was covalently attached to CYP3A4 protein during the inactivation process. The mechanism-based enzyme inhibition by zafirlukast was found to be highly selective for CYP3A4 (Table 1). On the basis of the finding that the formation of the R,βunsaturated iminium intermediate was shown to be catalyzed only by CYP3A4, it is reasonable to conclude that the same reactive metabolite is responsible for the suicide inhibition of CYP3A4. The finding that zafirlukast is biotransformed to a highly reactive electrophilic species that was found to alkylate GSH and may be responsible for mechanismbased inhibition of CYP3A4 has a number of implications with respect to the use of zafirlukast in humans. First, these results imply that zafirlukast may potentially have pharmacokinetic drug-drug interactions with CYP3A4 substrates. To our knowledge, the only clinical interaction study conducted with a CYP3A4 substrate was with terfenadine; however, zafirlukast had no effect on plasma levels of terfenadine (20). Conversely, zafirlukast was shown to inhibit CYP3A enzyme activity in vitro (nonpreincubation inhibition) with a KI of 2 µM in one report (17), but higher values have been published (13, 14), and our KI for the inactivation process was 13.4 µM. In addition, zafirlukast has high plasma protein binding

Kassahun et al.

(>99%), which would lower the free-drug concentration. Thus, free-drug plasma concentrations may not be high enough to produce clinically significant inhibition of CYP3A4. In addition, the relatively slow maximal inactivation rate and relatively high partition ratio of the mechanism-based inhibition suggest that zafirlukast is not an efficient inactivator of the enzyme. Therefore, although zafirlukast is a mechanism-based inactivator of CYP3A4 in vitro, it does not appear to cause significant clinical drug-drug interactions with substrates of this enzyme. The other major conclusion from our studies is that zafirlukast is metabolized to a highly reactive intermediate, and this finding may be related to zafirlukastinduced hepatotoxicity in individuals on zafirlukast therapy. In clinical trials, zafirlukast has caused infrequent asymptomatic elevations in serum liver enzymes (21) and more recently has been associated with cases of severe hepatotoxicity (1). The histopathologic changes observed in cases where liver biopsy was completed are consistent with drug-induced liver injury. Our study is the first to report the formation of a chemically reactive, potentially hepatotoxic intermediate through metabolism of zafirlukast by hepatic enzymes. The dehydrogenation pathway leading to the formation of the reactive metabolite of zafirlukast is considered to be an uncommon pathway (22), but several additional examples have been elucidated. Some important examples which elicit toxicity after P450-mediated dehydrogenation include acetaminophen (23), butylated hydroxytoluene (24, 25), and 4-hydroxytamoxifen (26). In summary, it may be concluded that zafirlukast undergoes two successive one-electron oxidations to afford a highly electrophilic R,β-unsaturated iminium species, which reacts spontaneously with GSH to yield the observed adduct. This chemically reactive intermediate is likely responsible for the mechanism-based inactivation of CYP3A4 and may play a role in zafirlukast-induced hepatotoxicity.

Acknowledgment. We thank Ping Lu for some of the time-dependent CYP3A inhibition experiments conducted in human liver microsomes. This study was supported by United States Public Health Service Grants HL13645 and HL60143 from the National Heart, Lung, and Blood Institute and GM074249 from the National Institute of General Medical Sciences (to G.S.Y.).

References (1) Reinus, J. F., Persky, S., Burkiewicz, J. S., Quan, D., Bass, N. M., and Davern, T. J (2000) Severe liver injury after treatment with the leukotriene receptor antagonist zafirlukast. Ann. Intern. Med. 133, 964-968. (2) Moles, J. R., Primo, J., Fernandez, J. M., and Hinojosa, J. E. (2001) Acute hepatocellular injury associated with zafirlukast. J. Hepatol. 35, 541-542. (3) Su, C. W., Wu, J. C., Huang, Y. H., Huang, Y. S., Chang, F. Y., and Lee, S. D. (2002) Zafirlukast-induced acute hepatitis. Zhonghua Yixue Zazhi (Taipei, Taiwan) 65, 553-556. (4) Nocerini, M. R., Carlson, J. R., and Yost, G. S. (1984) Electrophilic metabolites of 3-methylindole as toxic intermediates in pulmonary oedema. Xenobiotica 14, 561-564. (5) Nocerini, M. R., Yost, G. S., Carlson, J. R., Liberato, D. J., and Breeze, R. G. (1985) Structure of the glutathione adduct of activated 3-methylindole indicates that an imine methide is the electrophilic intermediate. Drug Metab. Dispos. 13, 690-694. (6) Skordos, K. W., Laycock, J. D., and Yost, G. S. (1998) Thioether adducts of a new imine reactive intermediate of the pneumotoxin 3-methylindole. Chem. Res. Toxicol. 11, 1326-1331.

Zafirlukast Inactivates CYP3A4 (7) Skordos, K. W., Skiles, G. L., Laycock, J. D., Lanza, D. L., and Yost, G. S. (1998) Evidence supporting the formation of 2,3-epoxy3-methylindoline: a reactive intermediate of the pneumotoxin 3-methylindole. Chem. Res. Toxicol. 11, 741-749. (8) Skiles, G. L., and Yost, G. S. (1996) Mechanistic studies on the cytochrome P450-catalyzed dehydrogenation of 3-methylindole. Chem. Res. Toxicol. 9, 291-297. (9) Thornton-Manning, J., Appleton, M. L., Gonzalez, F. J, and Yost, G. S. (1996) Metabolism of 3-methylindole by vaccinia-expressed P450 enzymes: correlation of 3-methyleneindolenine formation and protein-binding. J. Pharmacol. Exp. Ther. 276, 21-29. (10) Regal, K. A., Laws, G. M., Yuan, C., Yost, G. S., and Skiles, G. L. (2001) Detection and characterization of DNA adducts of 3-methylindole. Chem. Res. Toxicol. 14, 1014-1024. (11) Racha, J. K., Rettie, A. E., and Kunze, K. L. (1998) Mechanismbased inactivation of human cytochrome P450 1A2 by furafylline: detection of a 1:1 adduct to protein and evidence for the formation of a novel imidazomethide intermediate. Biochemistry 37, 7407-7419. (12) Lanza, D. L., Code, E., Crespi, C. L., Gonzalez, F. J., and Yost, G. S. (1999) Specific dehydrogenation of 3-methylindole and epoxidation of naphthalene by recombinant human CYP2F1 expressed in lymphoblastoid cells. Drug Metab. Dispos. 27, 798803. (13) Shader, R. I., Granda, B. W., von Moltke, L. L., Giancarlo, G. M., and Greenblatt, D. J. (1999) Inhibition of human cytochrome P450 isoforms in vitro by zafirlukast. Biopharm. Drug Dispos. 20, 385388. (14) Liu, K. H., Lee, Y. M., Shon, J. H., Kim, M. J., Lee, S. S., Yoon, Y. R., Cha, I. J., and Shin, J. G. (2004) Potential of pranlukast and zafirlukast in the inhibition of human liver cytochrome P450 enzymes. Xenobiotica 34, 429-438. (15) Lu, P., Schrag, M., Slaughter, D., Raab, C., Shou, M., and Rodrigues, D. (2003) Mechanism-based inhibition of human liver microsomal cytochrome P450 1A2 by zileuton, a 5-lipoxygenase inhibitor. Drug Metab. Dispos. 31, 1352-1360. (16) Savidge, R. D., Bui, K. H., Birmingham, B. K., Morse, J. L., and Spreen, R. C. (1998) Metabolism and excretion of zafirlukast in dogs, rats, and mice. Drug Metab. Dispos. 26, 1069-1076.

Chem. Res. Toxicol., Vol. 18, No. 9, 2005 1437 (17) Grimm, S. W., Stams, K. R., and Aaron, E. J. (1996) Zafirlukast is a substrate for CYP2C9 and an inhibitor of CYP2C9 and CYP34 in vitro [abstract]. 7th North American International Society for the Study of Xenobiotics Meeting, Oct 1996, San Diego, CA, p 392. (18) Kent, U. M., Juschyshyn, M. I., and Hollenberg, P. F. (2001) Mechanism-based inactivators as probes of cytochrome P450 structure and function. Curr. Drug Metab. 2, 215-243. (19) Bessems, J. G. M., de Groot, M. J., Baede, J. M., te Koppele, J. M., and Vermeulen, N. P. E. (1998) Hydrogen atom abstraction of 3,5-disubstituted analogues of paracetamol by horseradish peroxidase and cytochrome P450. Xenobiotica 28, 855-875. (20) Suttle, A. B., Birmingham, B. K., and Vargo, D. L. (1997) The pharmacokinetics of zafirlukast and terfenadine after administration to healthy men. Allergy 52 (Suppl. 37), 184. (21) Chung, K. F, and Barnes, P. J. (1998) Zafirlukast (ACCOLATETM). Drugs Today 34, 375-388. (22) Guengerich, F. P. (2001) Uncommon P450-catalyzed reactions. Curr. Drug Metab. 2, 93-115. (23) Raucy, J. L., Lasker, J. M., Lieber, C. S., and Black, M. (1989) Acetaminophen activation by human liver cytochromes CYPIIE1 and CYPIA2. Arch. Biochem. Biophys. 271, 270-283. (24) Bolton, J. L., and Thompson, J. A. (1991) Oxidation of butylated hydroxytoluene to toxic metabolites. Factors influencing hydroxylation and quinone methide formation by hepatic and pulmonary microsomes. Drug Metab. Dispos. 19, 467-472. (25) Lewis, M. A., Yoerg, D. G., Bolton, J. L., and Thompson, J. A. (1996) Alkylation of 2′-deoxynucleosides and DNA by quinone methides derived from 2,6-di-tert-butyl-4-methylphenol. Chem. Res. Toxicol. 9, 1368-1374. (26) Fan, P. W., Zhang, F., and Bolton, J. L. (2000) 4-Hydroxylated metabolites of the antiestrogens tamoxifen and toremifene are metabolized to unusually stable quinone methides. Chem. Res. Toxicol. 13, 45-52.

TX050092B

Zafirlukast Metabolism by Cytochrome P450 3A4 Produces an

Recommend Documents