Bioactivation of Lumiracoxib by Peroxidases and Human Liver

106

Chem. Res. Toxicol. 2009, 22, 106–117

Bioactivation of Lumiracoxib by Peroxidases and Human Liver Microsomes: Identification of Multiple Quinone Imine Intermediates and GSH Adducts Ping Kang,* Deepak Dalvie, Evan Smith, and Matt Renner Pharmacokinetics Dynamics and Metabolism, Research Analytical, Pfizer Global Research and DeVelopment, 10724 Science Center DriVe, San Diego, California 92121 ReceiVed June 27, 2008

Lumiracoxib (Prexige; 2-[(2-fluoro-6-chlorophenyl)amino]-5-methyl-benzeneacetic acid) is a cyclooxygenase-2 selective inhibitor for the symptomatic treatment of osteoarthritis. Recently, the drug has been withdrawn in several countries due to serious liver side effects. Li et al. recently have shown that lumiracoxib is bioactivated to a proposed quinone imine that is trapped by N-acetylcysteine (NAC) to form two NAC adducts in human and rat liver microsomal incubations. The current study demonstrated that the lumiracoxib metabolite 4′-hydroxylumiracoxib (M5) can also be bioactivated by peroxidases such as horseradish peroxidase, myeloperoxidase, and prostaglandin H synthases. Efforts were also made to identify GSH adducts formed by P450s in human liver microsomal incubations of lumiracoxib. We herein report the detection and characterization of mono-, di-, tri-, and tetra-GSH adducts in these oxidizing systems. Most of the conjugates were generated as a result of bioactivation of M5 by both peroxidases and P450s. Quinone imine (M15) and two GSH-conjugated quinone imines (M17 and M18) were identified as intermediates in the formation of these conjugates. The latter two were formed through sequential elimination of the fluorine and chlorine groups of GSH-conjugated M15. An additional GSH adduct, which appeared to be formed directly from parent, was only observed in human liver microsomal incubations. A mechanism was proposed for the bioactivation of lumiracoxib and the formation of the observed GSH adducts. These results suggest that bioactivation of lumiracoxib and M5 may result in GSH depletion, covalent binding to proteins, and oxidative stress and may potentially lead to hepatotoxicity. Introduction Lumiracoxib (Prexige; 2-[(2-fluoro-6-chlorophenyl)amino]5-methyl-benzeneacetic acid) is a novel cyclooxygenase-2 (COX-2) selective inhibitor. It has been approved in more than 40 countries for the symptomatic treatment of osteoarthritis and/ or acute pain related to primary dysmenorrheal and dental or orthopedic surgery (2). Recently, several cases of serious liver toxicity in patients have been associated with the use of lumiracoxib. The drug has since been withdrawn in Australia (6), New Zealand (7), Canada (8), and all European Union markets (9). Lumiracoxib has not been approved in the United States. Lumiracoxib is a structural analogue of diclofenac. Despite diclofenac’s popularity as the most frequently used NSAID worldwide, it is hepatotoxic in about 4% of patients. Diclofenac and its metabolites (4′-hydoxydiclofenac and 5-hydroxydiclofenac) have been shown to be bioactivated to form various reactive metabolites such as acyl glucuronide, quinone imines, and arene oxides. These reactive metabolites are implicated as hazards in diclofenac-induced hepatotoxicity (10-13). Whether lumiracoxib undergoes any analogous bioactivations that could lead to the clinically observed hepatotoxicity needs to be evaluated. Lumiracoxib is extensively metabolized to 4′-hydroxy, 5-carboxy, and 4′-hydroxy-5-carboxy lumiracoxib derivatives in humans (14). Li et al. recently have shown that lumiracoxib is * To whom correspondence should be addressed. Tel: 858-622-7630. E-mail: [email protected].

bioactivated to a proposed quinone imine that is trapped by N-acetylcysteine (NAC) to form two NAC adducts in human and rat liver microsomal incubations (1). Both NAC adducts are derived from the metabolite 4′-hydroxylumiracoxib (M5), which contains a 4-hydroxyaniline moiety. Previous studies have shown that the 4-hydroxyaniline group is oxidatively metabolized to a quinone imine by peroxdases and P450s as observed in acetaminophen and 5-hydoxydiclofenac (12, 15-17). In the current study, we evaluated the potential metabolic activation of M5 by peroxidases such as horseradish peroxidase (HRP), myeloperoxidase (MPO), and prostaglandin H synthases (PGHS).1 Efforts were also made to identify GSH adducts of lumiracoxib in GSH- and NADPH-supplemented human liver microsomal incubations. We herein report the detection and characterization of multiple GSH adducts and multiple quinone imine intermediates of lumiracoxib in these oxidizing systems. A detailed bioactivation mechanism was proposed to account for the formation of various GSH adducts of lumiracoxib.

Materials and Methods Materials. Lumiracoxib was purchased from 3B Scientific Corp. (Libertyville, IL) and was dissolved in DMSO to a final concentration of 50 mM. HRP, NADPH, H2O2 (30 wt % in H2O), L-ascorbic acid, and glutathione (reduced form, GSH) were purchased from Sigma-Aldrich (St. Louis, MO). Arachidonic acid (sodium salt) was 1 Abbreviations: NSAIDs, nonsteroidal anti-inflammatory drugs; NAC, N-acetylcysteine; MPO, myeloperoxidase; HRP, horseradish peroxidase; PGHS, prostaglandin H synthases; COSY, homonuclear correlation spectroscopy; MS/MS, tandem mass spectrometry; PDA, photodiode array.

10.1021/tx8002356 CCC: $40.75  2009 American Chemical Society Published on Web 12/08/2008

BioactiVation of Lumiracoxib purchased from Cayman Chemical Co. (Ann Arbor, MI). MPO was purchased from EMD Chemicals Inc. (Gibbstown, NJ). Ram seminal vesicle microsomes were purchased from Oxford Biomedical Research (Oxford, MI). Human liver microsomes were prepared from human livers (BD Gentest, Woburn, MA) using standard protocols and characterized using P450-specific marker substrate activities. Aliquots from the individual preparations from 56 individual human livers were pooled on the basis of equivalent protein concentrations to yield a representative microsomal pool with a protein concentration of 20.4 mg/mL. All other commercially available reagents and solvents were of either analytical or HPLC grade. Human Liver Microsomal Metabolism. Lumiracoxib (10 and 100 µM) was incubated for 1 h at 37 °C in an incubation system consisting of 100 mM potassium phosphate buffer (pH 7.4), 2 mg of human liver microsomes, 5 mM GSH, and 1 mM NADPH in a final volume of 1 mL. Reactions were terminated by the addition of 5 mL of acetonitrile. Samples were mixed on a vortex mixer and centrifuged for 5 min. The supernatants were transferred into conical glass tubes and evaporated to dryness under N2 at 30 °C. The residues were reconstituted in 200 µL of acetonitrile:water (30: 70 with 0.1% formic acid, v/v), and aliquots (40 µL) were analyzed using an HPLC-MS system. Alternatively, the incubation mixture was quenched by adding ascorbic acid (2 mM final), and an aliquot (100 µL) was directly injected for LC/MS analysis. Control experiments were conducted in the absence of either GSH or NADPH. Metabolite profiling was performed on an Agilent 1100 HPLC system (Agilent Technologies, Palo Alto, CA) coupled with a Finnigan LCQ-Deca ion-trap mass spectrometer (Thermo Fisher Scientific, Waltham, MA). Separation was achieved using a Cosmosil 5PYE column (150 mm × 2.0 mm, Phenomenex, Torrance, CA) at a flow rate of 0.2 mL/min. A gradient of (A) water with 0.1% formic acid and (B) acetonitrile with 0.1% formic acid was as follows: initiated with 0% B for 5 min and then increased in a linear manner to 30% at 20 min and to 50% at 25 min, held at 50% until 28 min, changed linearly to 90% at 40 min, maintained at 90% for up to 43 min, and then decreased to 0% at 45 min. The column was allowed to equilibrate at 0% solvent B for 5 min prior to the next injection. The HPLC effluent going to the mass spectrometer was directed to waste through a divert valve for the initial 5 min after sample injection. Operating parameters for the ion-trap ESI-MS method were as follows (positive mode): capillary temperature, 270 °C; spray voltage, 4.5 kV; capillary voltage, 4.5 V; sheath gas flow rate, 90 (arbitrary value); and auxiliary gas flow rate, 30 (arbitrary value). The mass spectrometer was operated in data-dependent scanning mode. The ions were monitored over a full mass range of m/z 125-2000. For a full scan, the automatic gain control was set at 5.0 × 108, the maximum ion time was 100 ms, and the number of microscans was set at 3. For MSn scanning, the automatic gain control was set at 1.0 × 108, the maximum ion time was 400 ms, and the number of microscans was set at 2. For data-dependent scanning, the default charge state was 1, the default isolation width was 4.0, and the normalized collision energy was 40. Generation of Lumiracoxib Metabolites. Reaction mixtures consisting of lumiracoxib (100 µM), phosphate buffer (100 mM, pH 7.4), human liver microsomes (2 mg), and NADPH (1 mM) in final volume of 1 mL were incubated for 1 h at 37 °C. Ethyl acetate (3 mL) was added to quench the reaction. The reaction mixtures were vortexed for 1 min and then centrifuged. The ethyl acetate supernatants were transferred into conical glass tubes and evaporated to dryness under N2 at 30 °C. The crude mixture of lumiracoxib and its metabolites was used without further purification in subsequent peroxidase incubations. The yield of 4′-hydroxylumiracoxib (M5) was estimated to be ∼40% based on peak area of M5 in the UV chromatogram [photodiode array (PDA) total scan] of lumiracoxib and its metabolites. Incubations with Peroxidases. The crude mixture of lumiracoxib and its metabolites (total concentration, 100 µM final) was reconstituted in an incubation mixture containing phosphate buffer

Chem. Res. Toxicol., Vol. 22, No. 1, 2009 107 (100 mM, pH 7.4), reduced GSH (5 mM), and HRP (3 units) in a final volume of 1 mL. The reaction was initiated by adding H2O2 (500 µM final). After 15 min of incubation at 37 °C, the reaction was terminated by adding ascorbic acid (2 mM final), and an aliquot (100 µL) was injected for LC/MS analysis. Experiments were also carried out at lower concentrations of lumiracoxib and its metabolites (total concentrations at 10, 20, 40, and 60 µM, respectively). Similar incubations were performed for MPO except that KCl (150 mM final) and MPO (5 units/mL final) were added to the incubation mixture. Similarly, incubations were conducted with ram seminal vesicle microsomes (1 mg/mL final) where the reactions were initiated by adding H2O2 (500 µM final) or arachidonic acid (300 µM final). Control experiments without enzymes or cofactors were also performed. Identification of Multiple Quinone Imine Intermediates in the Oxidation of M5 by HRP. The crude mixture of lumiracoxib and its metabolites (total concentration, 100 µM) was reconstituted in an incubation mixture containing phosphate buffer (100 mM, pH 7.4) and HRP (3 units) in a final volume of 1 mL. The reaction was initiated by adding H2O2 (500 µM final). After a 2 min incubation at 37 °C, aliquots (100 µL) of the reaction mixture were transferred to glass tubes containing GSH of various concentrations (0-1000 µM) and were injected immediately for LC/MS analysis. Effect of GSH Concentration on the Formation of GSH Adducts of M5 by HRP. The HRP reaction mixtures containing phosphate buffer (100 mM, pH 7.4), HRP (3 units), H2O2 (500 µM), lumiracoxib, and its metabolites (total concentration, 100 µM) in a final volume of 100 µL were preincubated for 2 min, and then, GSH of various concentrations (25-1000 µM) was added. The reactions proceeded for 15 min at 37 °C. The reactions were terminated by adding ascorbic acid (2 mM final), and aliquots (100 µL) were injected for LC/MS analysis. The relative amounts of GSH adducts were represented by the percent ratio of UV peak areas (PDA total scan) of GSH adducts to that of lumiracoxib. Isolation of GSH Adduct M16 and NMR Characterization. A human liver microsomal incubation with lumiracoxib (100 µM) was performed as described above on a 10 mL scale. Ethyl acetate (30 mL) was added to stop the reaction. The reaction mixture was vortexed for 1 min and then centrifuged. The ethyl acetate supernatant was separated and evaporated to dryness under N2 at 30 °C. The crude mixture of lumiracoxib and its metabolites was reconstituted in an incubation mixture containing phosphate buffer (100 mM, pH 7.4), reduced GSH (5 mM), and HRP (3 units/mL) in a final volume of 3 mL. The reaction was initiated by adding H2O2 (500 µM final). After a 15 min incubation at 37 °C, the reaction was terminated by adding ascorbic acid (2 mM final). The major metabolite M8 was isolated but rapidly decomposed to M16 at room temperature. M16 was therefore isolated from this decomposed mixture. Separations were achieved using a COSMOSIL 5PYE column (150 mm × 4.6 mm, Phenomenex, Torrance, CA) at a flow rate of 1.0 mL/min with an Agilent 1100 HPLC system (Wilmington, DE). A gradient of (A) water with 0.1% formic acid and (B) acetonitrile with 0.1% formic acid was as follows: initiated with 100% A for 5 min, changed to 80% A from 5 to 10 min, changed to 50% A from 10 to 60 min, changed to 10% A from 60 to 70 min, held at 10% A from 70 to 75 min, changed to 100% A from 75 to 76 min, and held at 100% A from 76 to 80 min for the column to be equilibrated. All NMR spectra were acquired on a Bruker DPX-500 spectrometer equipped with a 5 mm TXI z-gradient cryoprobe (Bruker, Rheinstetten, Germany). M16 was dissolved in a mixture of D2O and acetonitrile-d3 (40:60, v/v). A 1H spectrum was acquired without water suppression using gradients pulses. Homonuclear correlation spectroscopy (COSY) spectrum was acquired without solvent suppression using gradient pulses for coherence selection. Chemical shifts are reported in ppm relative to tetramethylsilane (0 ppm).

Results Metabolites of Lumiracoxib in Human Liver Microsomal Incubations. Incubation of lumiracoxib in human liver

108

Chem. Res. Toxicol., Vol. 22, No. 1, 2009

Figure 1. UV chromatograms (PDA total scan) of metabolites in human liver microsomal incubations of lumiracoxib in the presence of GSH and NADPH. (A) UV chromatogram of lumiracoxib metabolites after sample concentration and (B) expended region of UV chromatogram showing multiple GSH adducts of lumiracoxib without sample concentration. See the experimental details in the Materials and Methods.

microsomes with GSH and NADPH generated several products. The major in vitro metabolite was M5 based on the UV chromatogram (Figure 1A). The MS spectrum of M5 showed an ion at m/z 310 ([M + H]). Upon further fragmentation, the molecular ion lost water (-18), CO (-28), and Cl (-35) sequentially to yield ions at m/z 292, 264, and 229, respectively (Table 1; see the Supporting Information for MS spectra). M5 was assigned to be 4′-hydroxylumiracoxib based on the similarity of MS fragmentation pattern observed to that reported in literature (14). The MS spectrum of M4 revealed an ion at m/z 292 that lost water easily to give a predominant fragment ion at m/z 274. Further fragmentation of the ion at m/z 274 generated ions at m/z 246 (-CO) and 211 (-Cl), respectively (Table 1). This metabolite was isolated for further spectroscopic analysis. 1H NMR of M4 (Supporting Information) showed the disappearance of the lumiracoxib methyl signal at 2.25 ppm and appearance of a new peak at 4.50 ppm (2H), which suggested that the methyl group was oxidized in M4. All six aromatic protons were retained in M4. M4 was proposed to be a indolin-2-one that resulted from initial hydroxylation of the methyl group followed by a facile cyclization of the carboxyl group with the amine group. The intermediate 5-hydroxymethyl lumiracoxib was not detectable. M4 has not been reported in previous studies; however, it could be the precursor to 5-carboxy lactam metabolites observed in human in vivo (14). Similar to that of M4, the tandem mass spectrometry (MS/MS) spectrum of M1 (molecular ion at m/z 308) showed a predominant ion at m/z 290 via the loss of water. Further fragmentation of the ion generated two ions at m/z 262 and 227, respectively (Table 1), both 16 Da greater than the corresponding fragment ions of M4. Thus, M1 was tentatively assigned to be a 4′-hydroxyl metabolite of M4. Formation of Lumiracoxib GSH Adducts in Human Liver Microsomes. The metabolic profile of lumiracoxib in human liver microsomes showed two major GSH adducts (Figure 1A; see the Supporting Information for MS spectra). The MS spectrum of M3 showed a molecular ion at m/z 599 ([M + H]+), 305 Da greater than that of lumiracoxib (Table 1). Fragmentation of the molecular ion generated ions at m/z 522 and 468 via the characteristic loss of glycine and pyroglutamate moieties, respectively, from the glutathionyl moiety.

Kang et al.

The predominant fragment ion of M3 at m/z 292 was generated from the elimination of the whole glutathionyl group. The adduct appeared to be derived from the parent drug only, as incubations of the isolated oxidative metabolites (M1, M4, and M5) in human liver microsomes failed to generate any detectable M3. M3 was proposed to be a benzylic GSH adduct of lumiracoxib. The facile cleavage of the glutathionyl moiety in the MS/MS spectrum of M3 was consistent with the formation of a resonance-stabilized bezylic carbon cation. Incubation of lumiracoxib in human liver microsomes also produced a GSH adduct M2 with a molecular ion at m/z 597 (Table 1). Characteristic neutral losses of the glycine and pyroglutamate moieties were observed in the MS/MS spectrum of M2 (ions at m/z 522 and 468, respectively). Loss of water from ion at m/z 468 generated the major ion at m/z 450. Further fragmentation of the ion at m/z 450 produced an ion at m/z 433 via elimination of the amino group and two ions at m/z 306 and 304 from cleavage of the C-S bond of glutathionyl group. The MS data of M2 suggested that the fluorine group was replaced by GSH. A similar NAC adduct of M5 has been reported recently where the fluorine group is substituted by a NAC molecule (1). Several di- and tri-GSH adducts of lumiracoxib (M6a, M6b, M7, and M8) at lower levels were also identified in human liver microsomal incubations of lumiracoxib (Figure 1B). Interestingly, when the supernatants of liver microsomal incubation mixture were evaporated to dryness and reconstituted for LC/ MS analysis, these GSH adducts were not observed. M6a showed a molecular ion at m/z 902, an addition of 305 Da to that of M2 (Table 1; see the Supporting Information for MS spectra). Sequential loss of two pyroglutamates (-129) from the molecular ion produced ions at m/z 773 and 644, respectively, suggesting the presence of two glutathionyl groups in M6. The structure of M6a was proposed to be a di-GSH adduct in which the fluorine was substituted by a GSH molecule (Table 1). Another adduct (M6b) with the same molecular ion at m/z 902 was detected at a trace level. It was a regioisomer of M6a (Scheme 1) as its MS/MS spectrum was identical to that of M6a. The definitive structure of M6a/M6b could not be determined based on MS data. The chlorine isotope pattern was lost in the molecular ion of M7 at m/z 868 (Table 1), consistent with substitution of the chlorine group of M2 by a GSH molecule. Sequential loss of two pyroglutamates was observed in the MS/MS spectrum of M7, resulting in the ions at m/z 739 and 610, respectively. A tri-GSH adduct (M8) showed a molecular ion at m/z 1173 (Table 1). Further fragmentation of this ion generated ions at m/z 1044, 915, and 786, respectively, through sequential cleavages of three glutamate bonds. The chlorine isotope pattern was absent for the molecular ion of M8, suggesting substitution of the chlorine by a GSH moiety. M8 was proposed to be a tri-GSH conjugate where both fluorine and chlorine substituents were replaced by GSH molecules. In addition to the GSH conjugates described above, several novel GSH conjugates (M9-12) were detected in incubations of lumiracoxib in human liver microsomes (Figure 1B). All of the metabolites showed molecular ions that were 28 Da less than those of M2, M6, M7, and M8, respectively, suggesting that M9-M12 were decarboxylated metabolites. The MS/MS spectra of these adducts all showed a predominant fragment ion via the loss of H2O, consistent with the structure of a benzylic alcohol. The molecular ion of M9 at m/z 569 was 28 Da less than that of M2 (Table 1; see the Supporting Information for MS spectra) and still retained the chlorine isotope pattern.

BioactiVation of Lumiracoxib

Chem. Res. Toxicol., Vol. 22, No. 1, 2009 109

Table 1. Molecular Ions, Proposed Structures, and MS Fragment Ions of Lumiracoxib Metabolites Formed in the Bioactivation of Lumiracoxib by Peroxidases and Human Liver Microsomes

110


Kang et al. Table 1. Continued


Chem. Res. Toxicol., Vol. 22, No. 1, 2009 111 Table 1. Continued

112


Kang et al. Table 1. Continued

Scheme 1. Proposed Mechanism for the Formation of Lumiracoxib GSH Adducts by Peroxidases and P450s

Elimination of a water molecule generated an ion at m/z 551. Further fragmentation of this ion generated ions at m/z 422 from loss of a pyroglutamate and m/z 278 from cleavage of the C-S bond of a glutathionyl moiety. M10 showed a molecular ion at m/z 874 with the chlorine isotope, 28 Da less than that of M6

(Table 1). Loss of water led to a predominant ion at m/z 856. This ion further fragmented to generate ions at m/z 727, 598, 583, and 454, respectively, through successive cleavage of glutathionyl C-S bond (-273) and elimination of pyroglutamate (-129). A molecular ion at m/z 840 was observed for M11



Figure 2. UV chromatograms (PDA total scan) of multiple GSH adducts of lumiracoxib from oxidation of a mixture of lumiracoxib and its metabolites by peroxidases in the presence of 5 mM GSH. (A) HRP, (B) Ram seminal vesicle microsome, and (C) MPO.

Figure 3. 1H NMR spectrum of M16 in a mixture of D2O and acetonitrile-d3 (40:60, v/v). The insert is the expended region showing two sets of GSH proton signals.

(Table 1). The MS/MS spectrum of M11 showed a predominant ion at m/z 822 as a result of dehydration. Subsequent fragment ions were at m/z 693 and 564 from successive loss of two pyroglutamates and at m/z 420 from the cleavage of a glutathionyl C-S bond of the ion at m/z 693. M12 (molecular ion at m/z 1145, Table 1) was identified as a decarboxy derivative of M8. Fragmentation of the dehydrated ion of M12 at m/z 1127 proceeded through successive cleavage of glutathionyl C-S bonds and elimination of pyroglutamates, resulting in ions at m/z 998, 869, 854, 725, and 596, respectively. Formation of Lumiracoxib GSH Adducts by Peroxidases. Because a synthetic standard of M5 was not available, this metabolite was generated using human liver microsomal incubation of lumiracoxib. Without separation, the crude ethyl acetate extract (see experimental section) of lumiracoxib and M5 was resuspended and subjected to oxidation by HRP/H2O2 in the presence of 5 mM GSH. This resulted in the formation of mono, di-, and tri-GSH adducts (M2, M6a, M6b, M7, and M8) that were previously observed in human liver microsomal incubations. However, the amount of these adducts was much greater in the peroxidase-mediated incubations (Figure 2A). The decarboxylated GSH adducts (M9, M10, M11, and M12) were formed at trace levels. Interestingly, direct injection of the reaction mixture to the LC/MS allowed detection of all of the GSH conjugates, whereas LC/MS analysis of the reconstituted supernatant showed only trace amounts of the multiple GSH adducts, presumably due to the difficulty of extracting different GSH adducts into the reconstitution buffer. Control experiments without HRP or H2O2 did not show turnover of M5 or formation of GSH adducts. Lumiracoxib and other oxidative metabolites (M1 and M4) were not oxidized by HRP. Incubations of lumiracoxib and its metabolites at lower concentrations (total concentration from 10 to 60 µM) with HRP/H2O2 and GSH generated similar metabolite profiles to that of 100 µM. A new, tetra-GSH adduct M13 was identified in HRPmediated oxidation. M13 exhibited a molecular ion at m/z 1478, 305 Da more than that of M8, suggesting addition of a glutathionyl group to M8 (Table 1). The MS/MS spectrum of M13 showed fragment ions at m/z 1349, 1220, 1091, and 962 from sequential loss of four pyroglutamate groups, consistent with four glutathionyl moieties in M13. HRP oxidation also afforded a cleavage product M14 (Table 1). A molecular ion at m/z 166 was obtained for M14, which upon collision-induced dissociation fragmented to ions at m/z 148 and 120 from dehydration and decarbonylation successively.

The tri-GSH adduct M8 was the most abundant metabolite in the HRP-mediated oxidation based on UV chromatogram (Figure 2A). To obtain a definitive structure of this adduct, an attempt was made to isolate M8 from the incubation mixture for NMR analysis. However, upon standing at room temperature, the isolated M8 quickly decomposed to a metabolite (M16). M16 showed a visible absorption maxima at 424 and 574 nm in the UV/vis spectrum (M8 absorbed at 342 nm), suggesting a structure of extended conjugation for M16. The molecular ion of M16 (m/z 896) was 277 Da less than that of M8, implying the cleavage of a glutathionyl C-S bond from M8 (Table 1). The MS/MS spectrum of M16 showed the loss of a pyroglutamate to form an ion at m/z 767 and further loss of a glutathionyl group to form an ion at m/z 460. M16 was purified and subjected to NMR analysis. Three aromatic proton signals at 7.50 (G), 7.34 (H), and 6.85 (I) ppm, respectively, were present in 1H NMR spectrum of M16 (Figure 3). Two sets of glutathionyl proton signals appeared in the region from 2.0 to 4.8 ppm. The benzyl methyl protons showed a chemical shift at 2.45 ppm. The COSY spectrum of M16 revealed a weak cross-peak between the two aromatic protons G and H (Figure 4A). It also exhibited couplings of the benzyl methyl to proton G and proton H, respectively (Figure 4B). Full assignment of the glutathionyl protons (labeled protons in Figure 3 insert) was achieved with the aid of the COSY experiment. NMR data suggested that there was only one aromatic proton (I) on ring R. However, available data were not sufficient to definitively assign the position of proton I. On the basis of the spectroscopic data of MS, UV-vis, and NMR, M16 was proposed to be a tricyclic di-GSH conjugate. We also studied the bioactivation of M5 by other peroxidases such as PGHS and MPO. Incubation of a mixture of lumiracoxib and M5 with ram seminal vesicle microsomes, H2O2/arachidonic acid, and GSH generated a metabolite profile similar to that of HRP (Figure 2C). Decarboxylated GSH adducts (M9, M10, M11, and M12) were detected at trace levels. A similar profile of GSH adducts was obtained for the incubation of lumiracoxib and M5 with MPO/H2O2/ Cl- in the presence of GSH (Figure 2B). Lumiracoxib and other oxidative metabolites (M1 and M4) were not oxidized by these peroxidases. The cleavage metabolite M14 was also formed in these incubations. Identification of Multiple Quinone Imine Intermediates in the Oxidation of M5 by HRP. LC/MS analysis of the incubation mixture of HRP without ascorbic acid quenching

114


Kang et al.

Figure 4. COSY spectrum of M16. (A) Expended region showing the coupling between aromatic protons G and H and (B) expended region showing the couplings of aromatic protons G and H to the methyl group.

showed that M5 was turned over to a metabolite (M15 in Table 1) with a molecular ion at m/z 308, 2 Da lower than that of M5. The MS/MS spectrum of M15 showed a major ion at m/z 262 from elimination of the carboxylic acid. Further fragmentation of this ion proceeded via dechlorination and decarbonylation (-CO) to yield ions at 234, 226, and 199, respectively. Extrusion of CO from the decarboxylated ion at m/z 262, which was not observed in the fragmentation of the corresponding decarboxylated ion (m/z 248) of M5, indicated the presence of a quinone moiety instead of a hydroxyl group in M15. Thus, the structure of M15 was proposed to be a quinone imine. M15 absorbed at 518 nm in the UV/vis spectrum (Figure 5), similar to that of a quinone imine of diclofenac (12). Ascorbic acid reduced M15 completely back to M5 and at the same time changed the color from pink to colorless.

Two other quinone imine intermediates were formed upon addition of GSH to the reaction mixture of HRP after the initial formation of M15. The first intermediate (M17) absorbed at 488 nm in the UV/vis spectrum (Figure 5). The molecular ion of M17 at m/z 595 was 2 Da less than that of M2, suggesting that M17 was a desaturated product of M2 (Table 1). Fragmentation of M17 generated ions at m/z 576 (-H2O), 551 (-CO2), 520 (-glycine), 466 (-pyroglutamate), 422 (-CO2 and -pyroglutamate), and 278 (-CO2 and the cleavage of a glutathionyl C-S bond). The second quinone imine intermediate (M18) absorbed at 364 and 482 nm in the UV/vis spectrum (Figure 5). The molecular ion of M18 at m/z 866 was 2 Da lower than that of M7, which suggested that M18 was a quinone imine of M7. The molecular ion at m/z 866 underwent fragmentation through dehydration,



Discussion

Figure 5. UV-vis spectra of quinone imine intermediates M15, M17, and M18.

Figure 6. Formation of GSH adducts with different concentrations of GSH in the oxidation of lumiracoxib M5 by HRP. The relative abundance of GSH adducts was represented by the percent ratio of UV peak areas (PDA total scan) of GSH adducts to that of lumiracoxib.

decarboxylation, deglutamation, and cleavage of the glutathionyl C-S bond to generate ions shown in Table 1. Effect of GSH on the Formation of Multiple Lumiracoxib GSH Adducts. The formation of GSH adducts was dependent on the concentration of GSH in HRP oxidation of M5 (Figure 6). At the lowest concentration of GSH (25 µM), the monoadduct M2 was formed in highest amount among the GSH adducts. The peak concentration of M2 could have been achieved at a GSH concentration less than 25 µM. As M2 decreased with increasing GSH, the diadduct M7 was formed early and peaked at round 100 µM GSH. The other diadduct M6 was generated in lower amount that did not change with GSH. The triadduct M8 increased with decreasing M2 and M7. The highest concentration of M8 was reached at round 500 µM GSH and remained constant thereafter. The formation of M17 and M18 was also dependent on the concentrations of GSH. M17 was most abundant with GSH at 25 µM, whereas M18 reached the highest concentration at 100 µM GSH (data not shown), which correlated with the optimal formation of the corresponding reduced GSH adducts M2 and M7.

A previous report has characterized two NAC adducts in liver microsomal incubations of lumiracoxib. The formation of these conjugates was proposed to proceed through a quinone imine intermediate (M15) that resulted from P450-catalyzed bioactivation of M5 (1). The current study demonstrated that M5 was also bioactivated by peroxidases, and the resulting M15 was characterized by UV/vis, MS, and chemical reduction. Additional GSH adducts and quinone imine intermediates were also identified. A proposed mechanism for the formation of multiple GSH adducts (M2, M6, M7, M8, and M13) from M5 is depicted in Scheme 1. The first step involves an overall two-electron oxidation of M5 by peroxidase or P450 enzymes to yield the quinone imine M15. The formation of the second quinone imine M17 results from a nucleophilic addition of GSH to M15 (formation of intermediate A) followed by displacement of the fluorine moiety (as HF). Reduction of M17 by ascorbic acid or GSH generates M2. Li et al. recently reported a fluorinesubstituted mono-NAC adduct in bioactivation studies of lumiracoixb (1). A similar chlorine-substituted mono-GSH adduct has also been reported for diclofenac (13). Nucleophilic attack of another GSH at the 3- or 5-position of ring R of M17 results in the formation M6a and M6b. As described in the Results, one isomer (M6a) was formed in greater amounts than M6b in the current study. Previous studies (1) have also demonstrated the formation of a mono NAC conjugate of lumiracoxib that results from addition of NAC at the carbon adjacent to the Cl moiety. M17 is converted to the third quinone imine M18 via nucleophilic displacement of Cl group of M17 by another GSH molecule. While reduction of M18 by GSH or ascorbic acid results in the formation of the diadduct M7, nucleophilic addition of GSH to M18 generates the triadduct M8. The proposed mechanism suggests that intermediate quinone imines M17 and M18 are generated from M15 and GSH by successive elimination of HF and HCl without involving peroxidase/P450s oxidation. The tetra-GSH adduct M13 is proposed to be derived from a quinone imine intermediate B (Scheme 1), which apparently requires an additional oxidation step. Recently, Madsen et al. (18) reported multiple deschloroGSH adducts from electrochemical as well as P450 oxidation of diclofenac, most likely through a similar mechanism. Indeed, we observed quinone imine, deschloro-mono-GSH quinone imine, deschloro-di-GSH quinone imine, and the corresponding multiple deschloro-GSH adducts of diclofenac in the oxidation of 4′-hydroxydiclofenac by HRP (unpublished results). Human liver microsomal incubation of lumiracoxib also produced decarboxylated GSH adducts (M9, M10, M11, and M12) in comparable abundance to those without decarboxylation (M2, M6, M7, and M8) (Figure 1B). The decarboxylated GSH adducts are proposed to be derived from a decarboxylated metabolite of M5. Decarboxylation of carbonyl-containing compounds by P450s has been documented in the literature (19, 22, 23). A similar bioactivation mechanism involving quinone imines of the decarboxylated M5 would account for the observed decarboxylated GSH adducts. GSH served as both a nucleophile and a reducing agent for lumiracoxib quinone imines, resulting in the formation of GSH adducts and presumably GSSG, respectively. Nucleophilic addition of GSH to M15 was much faster than the reduction of M15 by GSH, as M15 was completely turned over to GSH adducts in HRP incubations. In the presence of both ascorbic acid and GSH, M15 was reduced to M5 without generating detectable GSH adducts, which indicated a much higher

116


Kang et al.

Scheme 2. Proposed Mechanism for the Formation of M16 from M8

Scheme 3. Proposed Mechanism for the formation of M2 in Human Liver Microsomal Incubation of Lumiracoxib

reducing capability of ascorbic acid toward quinone imines than that of GSH. The relative amounts of the GSH adducts were dependent on the amount of GSH present in the reaction (Figure 6). A lower concentration of GSH favored the formation of mono-GSH adduct, whereas a higher concentration of GSH favored the di- and triadducts, which was consistent with the proposed sequential addition of GSH to quinone imines (Scheme 1). The major GSH adduct was M8 in peroxidase-mediated oxidation of M5. At room temperature, M8 decomposed to M16. A proposed mechanism for the formation of M16 is shown in Scheme 2. One-electron oxidation generates a N-centered radical C that undergoes C-S bond cleavage to form an intermediate D (pathway a). Another one-electron oxidation of D produces an oxygen-centered radical E that is in resonance with a thiol radical F. Coupling of the thiol radical with ring R′ results in the formation of M16a. Pathway b leads to a different tricyclic structure (M16b). However, only one product was observed. Available data were not sufficient to differentiate the two structures. Although the proposed mechanism for the formation of M16 would also be expected for the diadducts (M6, M7) or the tetra-adduct (M13), no such tricyclic products were detected. M3 was a direct GSH adduct of lumiracoxib observed only in human liver microsomal incubations. We propose that oxidation of lumiracoxib by P450s produces an imino methide, which is attacked by GSH at the benzylic position to form M3 (Scheme 3). When incubation of lumiracoxib was conducted in human liver microsomes supplemented with UDPGA, an acyl glucuronide of lumiracoxib was formed (data not shown), which nevertheless was a minor metabolite in humans in vivo (14). The formation of multiple GSH adducts in the bioactivation of lumiracoxib M5 indicated that one molecule of M5 could conjugate up to four GSH molecules, which could greatly facilitate GSH depletion. In addition to GSH adduct formation, quinone imines have been shown to covalently modify proteins

(20). In light of its high reactivity toward the thiol group of GSH, lumiracoxib quinone imine is expected to form covalent binding to proteins. Future study with radiolabeled lumiracoxib will provide more information on covalent binding to proteins. Redox cycling of quinone imines has also been shown to produce oxidative stress (10, 21). Altogether, these risk factors potentially contribute to the clinically observed hepatotoxicity. In summary, P450s and peroxidases (HRP, MPO, and PGHS) bioactivated M5 to a quinone imine M15 and several GSHconjugated quinone imine intermediates (M17 and M18), which were attacked by GSH to form mono-, di-, tri-, and tetra-GSH adducts. The current study implicates that bioactivation of lumiracoxib through quinone imines may result in GSH depletion, covalent binding to proteins, oxidative stress, and eventually lead to hepatotoxicity. These toxicological events have been proposed to be associated with quinone imine intermediates of other drugs such as diclofenac and acetaminophen (10, 21). Acknowledgment. We thank Bill Pool and Ellen Wu (Pfizer Global Research and Development) for reviewing the manuscript. Supporting Information Available: 1H NMR spectra of lumiracoxib and metabolite M4. MS spectra of metabolites M1-M18. This material is available free of charge via the Internet at http://pubs.acs.org.

References (1) Li, Y., Slatter, J. G., Zhang, Z., Li, Y., Doss, G. A., Braun, M. P., Stearns, R. A., Dean, D. C., Baillie, T. A., and Tang, W. (2008) In vitro metabolic activation of lumiracoxib in rat and human liver preparations. Drug Metab. Dispos. 36, 469–473. (2) Bannwarth, B., and Be´renbaum, F. (2007) Lumiracoxib in the management of osteoarthritis and acute pain. Exp. Opin. Pharmacother. 8, 1551–1564. (3) Farkouh, M. E., Kirshner, H., Harrington, R. A., Ruland, S., Verheugt, F. W., Schnitzer, T. J., Burmester, G. R., Mysler, E., Hochberg, M. C., Doherty, M., Ehrsam, E., Gitton, X., Krammer, G., Mellein, B.,


(4)

(5) (6) (7) (8) (9) (10)

(11) (12)

(13)

Gimona, A., Matchaba, P., Hawkey, C. J., and Chesebro, J. H. (2004) Comparison of lumiracoxib with naproxen and ibuprofen in the Therapeutic Arthritis Research and Gastrointestinal Event Trial (TARGET), cardiovascular outcomes: randomised controlled trial. Lancet 364, 675–684. Schnitzer, T. J., Burmester, G. R., Mysler, E., Hochberg, M. C., Doherty, M., Ehrsam, E., Gitton, X., Krammer, G., Mellein, B., Matchaba, P., Gimona, A., and Hawkey, C. J. (2004) Comparison of lumiracoxib with naproxen and ibuprofen in the Therapeutic Arthritis Research and Gastrointestinal Event Trial (TARGET), reduction in ulcer complications: Randomised controlled trial. Lancet 364, 665– 674. Topol, E. J., and Falk, G. W. (2004) A coxib a day won’t keep the doctor away. Lancet 364, 639–640. http://www.tga.gov.au/recalls/2007/lumiracoxib.htm. http://www.moh.govt.nz/moh.nsf/0/793489DCEF587AA2CC25733D0077C44B. http://www.novartis.ca/downloads/en/letters/prexige_fact_20071003_e. pdf. http://www.emea.europa.eu/pdfs/human/press/pr/PR_Lumiracoxib_57930107en.pdf. Boelsterli, U. A. (2003) Diclofenac-induced liver injury: A paradigm of idiosyncratic drug toxicity. Toxicol. Appl. Pharmacol. 192, 307– 322. Tang, W. (2003) The metabolism of diclofenacsEnzymology and toxicology perspectives. Curr. Drug Metab. 4, 319–329. Miyamoto, G., Zahid, N., and Uetrecht, J. P. (1997) Oxidation of diclofenac to reactive intermediates by neutrophils, myeloperoxidase, and hypochlorous acid. Chem. Res. Toxicol. 10, 414–419. Yu, L. J., Chen, Y., Deninno, M. P., O’Connell, T. N., and Hop, C. E. (2005) Identification of a novel glutathione adduct of diclofenac, 4′hydroxy-2′-glutathion-deschloro-diclofenac, upon incubation with human liver microsomes. Drug Metab. Dispos. 33, 484–488.

Chem. Res. Toxicol., Vol. 22, No. 1, 2009 117 (14) Mangold, J. B., Gu, H., Rodriguez, L. C., Bonner, J., Dickson, J., and Rordorf, C. (2004) Pharmacokinetics and metabolism of lumiracoxib in healthy male subjects. Drug Metab. Dispos. 32, 566–571. (15) Potter, D. W., Miller, D. W., and Hinson, J. A. (1985) Identification of acetaminophen polymerization products catalyzed by horseradish peroxidase. J. Biol. Chem. 260, 12174–80. (16) Potter, D. W., Miller, D. W., and Hinson, J. A. (1986) Horseradish peroxidase-catalyzed oxidation of acetaminophen to intermediates that form polymers or conjugate with glutathione. Mol. Pharmacol. 29, 155–62. (17) Potter, D. W., and Hinson, J. A. (1989) Acetaminophen peroxidation reactions. Drug Metab. ReV. 20, 341–358. (18) Madsen, K. G., Olsen, J., Skonberg, C., Jurva, U., Cornett, C., Hansen, S. H., Johansen, T. N., and Olsen, J. (2008) Bioactivation of diclofenac in vitro and in vivo: Correlation to electrochemical studies. Chem. Res. Toxicol. 21, 1107–1119. (19) Ortiz de Montellano, P. R. (2005) Substrate oxidation by cytochrome P450 enzymes. In Cytochrome P450: Structure, Mechanism and Biochemistry (Ortiz de Montellano, P. R., Ed.) pp 226-228, Plenum Press, New York. (20) Ju, C., and Uetrecht, J. P. (2002) Mechanism of idiosyncratic drug reactions: reactive metabolite formation, protein binding and the regulation of the immune system. Curr. Drug Metab. 3, 367–377. (21) Hinson, J. A., Reid, A. B., McCullough, S. S., and James, L. P. (2004) Acetaminophen-induced hepatotoxicity: Role of metabolic activation, reactive oxygen/nitrogen species, and mitochondrial permeability transition. Drug Metab. ReV. 36, 805–822. (22) Komuro, M., Higuchi, T., and Hirobe, M. (1995) Application of chemical cytochrome P-450 model systems to studies on drug metabolismsVIII. Novel metabolism of carboxylic acids via oxidative decarboxylation. Bioorg. Med. Chem. 3, 55–65. (23) Grillo, M. P., Ma, J., Teffera, Y., and Waldon, D. J. (2008) A novel bioactivation pathway for diclofenac initiated by P450 mediated oxidative decarboxylation. Drug Metab. Dispos. 36, 1740–1744.

TX8002356

Bioactivation of Lumiracoxib by Peroxidases and Human Liver

Recommend Documents