NADPH-Dependent Covalent Binding of - American Chemical Society

Groton, Connecticut, and La Jolla, California, and Research Technology Center, Pfizer Global Research and. DeVelopment, Cambridge, Massachusetts...
0 downloads 0 Views 225KB Size
Download PDF
Chem. Res. Toxicol. 2007, 20, 1649–1657

1649

NADPH-Dependent Covalent Binding of [3H]Paroxetine to Human Liver Microsomes and S-9 Fractions: Identification of an Electrophilic Quinone Metabolite of Paroxetine Sabrina X. Zhao,† Deepak K. Dalvie,‡ Joan M. Kelly,§ John R. Soglia,† Kosea S. Frederick,† Evan B. Smith,‡ R. Scott Obach,† and Amit S. Kalgutkar*,† Pharmacokinetics, Dynamics and Metabolism Department, Pfizer Global Research and DeVelopment, Groton, Connecticut, and La Jolla, California, and Research Technology Center, Pfizer Global Research and DeVelopment, Cambridge, Massachusetts ReceiVed April 25, 2007

The primary pathway of clearance of the methylenedioxyphenyl-containing compound and selective serotonin reuptake inhibitor paroxetine in humans involves P450 2D6-mediated demethylenation to a catechol intermediate. The process of demethylenation also results in the mechanism-based inactivation of the P450 isozyme. While the link between P450 2D6 inactivation and pharmacokinetic interactions of paroxetine with P450 2D6 substrates has been firmly established, there is a disconnect in terms of paroxetine’s excellent safety record despite the potential for bioactivation. In the present study, we have systematically assessed the NADPH-dependent covalent binding of [3H]paroxetine to human liver microsomes and S-9 preparations in the absence and presence of cofactors of the various phase II drugmetabolizing enzymes involved in the downstream metabolism/detoxification of the putative paroxetine–catechol intermediate. Incubation of [3H]paroxetine with human liver microsomes and S-9 preparations resulted in irreversible binding of radioactive material to macromolecules by a process that was NADPH-dependent. The addition of reduced glutathione (GSH) to the microsomal and S-9 incubations resulted in a dramatic reduction of covalent binding. Following incubations with NADPH- and GSHsupplemented human liver microsomes and S-9, three sulfydryl conjugates with MH+ ions at 623 Da (GS1), 779 Da (GS2), and 928 Da (GS3), respectively, were detected by LC-MS/MS. The collisioninduced dissociation spectra allowed an insight into the structure of the GSH conjugates, based on which, bioactivation pathways were proposed. The formation of GS1 was consistent with Michael addition of GSH to the quinone derived from two-electron oxidation of paroxetine–catechol. GS3 was formed by the addition of a second molecule of GSH to the quinone species obtained via the two-electron oxidation of GS1. The mechanism of formation of GS2 can be rationalized via (i) further two-electron oxidation of the catechol motif in GS3 to the ortho-quinone, (ii) loss of a glutamic acid residue from one of the adducted GSH molecules, and (iii) condensation of a cysteine–NH2 with an adjacent carbonyl of the ortho-quinone to yield an ortho-benzoquinoneimine structure. Inclusion of the catechol-O-methyltransferase cofactor S-adenosylmethionine (SAM) in S-9 incubations also dramatically reduced the covalent binding of [3H]paroxetine, a finding that was consistent with O-methylation of the paroxetine–catechol metabolite to the corresponding guaiacol regioisomers in S-9 incubations. While the NADPH-dependent covalent binding was attenuated by GSH and SAM, these reagents did not alter paroxetine’s ability to inactivate P450 2D6, suggesting that the reactive intermediate responsible for P450 inactivation did not leave the active site to react with other proteins. The results of our studies indicate that in addition to the low once-a-day dosing regimen (20 mg) of paroxetine, efficient scavenging of the catechol and quinone metabolites by SAM and GSH, respectively, serves as an explanation for the excellent safety record of paroxetine despite the fact that it undergoes bioactivation. Introduction The propensity of a xenobiotic to undergo bioactivation to reactive intermediates, usually electrophiles, is a function of the chemistry that ensues following metabolism. Information to qualify certain organic functional groups as “structural alerts” or “toxicophores” has been inferred from myriad examples of * To whom correspondence should be addressed. Tel: 860-715-2433. Fax: 860-686-1059. E-mail: [email protected]. † Pharmacokinetics, Dynamics and Metabolism Department, Groton, Connecticut. ‡ Pharmacokinetics, Dynamics and Metabolism Department, La Jolla, California. § Research Technology Center, Cambridge, Massachusetts.

protoxins containing these motifs, which upon bioactivation generate reactive metabolite(s) (1, 2). The concept can be further expanded to include numerous drugs that contain putative structural alerts and are associated with type B idiosyncratic toxicity (1–4). Indeed, for many such drugs, reactive metabolite formation has been demonstrated using in vitro systems derived from human hepatic tissues. Furthermore, examples wherein elimination of the bioactivation liability in prototype drugs associated with toxicity translates into a markedly improved safety profile with the new agent have served to strengthen the link between bioactivation and toxicity. Overall, on the basis of such retrospective analysis of structure–toxicity relationships, there often exists a concern amongst medicinal chemists that

10.1021/tx700132x CCC: $37.00  2007 American Chemical Society Published on Web 10/02/2007

1650 Chem. Res. Toxicol., Vol. 20, No. 11, 2007

Zhao et al.

Scheme 1. Metabolic Pathways of Paroxetine in Preclinical Species and Humans

functional groups that have been associated with bioactivation in the past should be generally avoided in drug design efforts. However, not all drugs that contain structural alerts prone to bioactivation are associated with toxicity. For instance, paroxetine, the selective serotonin reuptake inhibitor, which contains the 1,3-benzdioxole (methylenedioxyphenyl) structural alert, has enjoyed worldwide commercial success as an antidepressant, and despite over a decade of clinical use, life-threatening adverse drug reactions with this drug are extremely rare (5–7). The major route of paroxetine metabolism in animals and humans involves cytochrome P450 2D6-mediated demethylenation of the methylenedioxyphenyl group, yielding a catechol metabolite (8, 9) that can potentially undergo further oxidation to a reactive orthoquinone intermediate. As is the case with numerous methylenedioxyphenyl-containing compounds (10), the process of metabolism on the methylenedioxyphenyl group in paroxetine by P450 2D6 also results in the mechanism-based inactivation of the P450 isozyme (11), a phenomenon that is consistent with the observed nonstationary pharmacokinetics of paroxetine in P450 2D6 extensive metabolizers (8) and clinical drug–drug interactions (DDIs1) between paroxetine and coadministered drugs whose clearance is largely dependent on the activity of P450 2D6. Examples of characterized clinical DDIs with paroxetine include its effects on the kinetics of desipramine (12, 13), metoprolol (14), risperidone (15), and atomoxetine (16), where the clearance of the affected drugs is impaired by 5–8fold. The fold increases in the victim drug(s) area under the curve are in reasonable agreement with the values obtained upon scaling of the in vitro kinetic constants for mechanism-based inactivation of P450 2D6 by paroxetine (17, 18). While the link between the in vitro P450 2D6 inactivation and the observed clinical DDIs has been firmly established, there still exists a disconnect in terms of paroxetine’s excellent safety record despite the bioactivation liability. A plausible explanation for this disparity may be the rapid detoxication of the catechol metabolite by phase II enzymes such that bioactivation to the 1 Abbreviations: DDIs, drug–drug interactions; COMT, catechol-Omethyltransferase; GSH, reduced glutathione; UDPGA, uridine-5′-diphosphoglucuronic acid; PAPS, phosphoadenosine 5′-phosphosulfate; SAM, S-adenosyl methionine; LC-MS/MS, liquid chromatography tandem mass spectrometry; CID, collision-induced dissociation; Rt, retention time; UGT, uridine 5′-diphosphoglucuronosyl transferase; ST, sulfotransferase; GS1, monoglutathionyl conjugate of paroxetine catechol; GS2, 5-(3-(3-((carboxymethyl)carbamoyl)-7-((4S)-4-(4-fluorophenyl)piperidin-3-yl)methoxy)5-oxo-3,5-dihydro-2H-benzo[b][1,4]thiazin-8-ylthio)-1-(carboxymethylamino)1-oxopropan-2-ylamino)-2-amino-5-oxopentanoic acid; GS3, diglutathionyl conjugate of paroxetine catechol; MI complex, metabolite–inhibitor complex.

ortho-quinone does not occur. Certainly, the identification of the O-methylated metabolite of paroxetine mediated by catecholO-methyltransferase (COMT) and the further metabolism of this guaiacol derivative to corresponding glucuronide and sulfate metabolites in humans and animals supports this notion (8, 9) (Scheme 1). In an attempt to gain an insight into this hypothesis, we have systematically assessed the NADPH-dependent covalent binding of [3H]paroxetine to human liver microsomes and S-9 preparations in the absence and presence of cofactors of the various phase II drug-metabolizing enzymes involved in the downstream metabolism/detoxification of the putative paroxetine–catechol intermediate. The occurrence of the bioactivation sequence paroxetine f paroxetine catechol f paroxetine quinone has also been examined in human liver microsomes and S-9 fractions using glutathione (GSH) as a trapping agent in the presence or absence of various cofactors of phase 2 enzymes. The results of this overall analysis are summarized herein.

Experimental Procedures Materials. Paroxetine hydrochloride was synthesized (chemical and isomeric purity >99% by mass spectrometry and NMR) and purified at Pfizer Global Research and Development (Groton, CT). [Phenyl-6-3H]paroxetine (see Scheme 1) was purchased from Perkin Elmer (Waltham, MA). Unless specified, all other chemicals and solvents were obtained from Sigma-Aldrich (Milwaukee, WI) and were reagent grade or better. NADPH, uridine-5′-diphosphoglucuronic acid (UDPGA), 3′-phosphoadenosine-5′-phosphosulfate (PAPS), S-adenosyl methionine (SAM), and GSH were purchased from Sigma-Aldrich Chemical Co. Human liver microsomal fractions pooled from 53 individual donors were purchased from BD Gentest (Woburn, MA). A pooled human liver S-9 (15 human livers) fraction was obtained from In Vitro Technologies, Inc. (Baltimore, MD). In Vitro Covalent Binding of [3H]Paroxetine to Human Liver Microsomes and S-9 Fractions. [3H]Paroxetine stock solution (24.4 Ci/mmol) was prepared in ethanol. Unlabeled paroxetine stock solution (30 mM) was prepared in methanol. The determination of the extent of radioactivity covalent binding over time was assessed by incubating paroxetine (25 µM paroxetine/ [3H]paroxetine, 0.151 µCi per tube) at 37 °C in human liver microsomes (2 mg protein/mL, P450 concentration ) 0.7 µM) or human liver S-9 (5 mg protein/mL, P450 concentration ) 0.07 µM) containing 100 mM potassium phosphate buffer (pH 7.4) and 3.3 mM MgCl2. NADPH (1.3 mM final) was added to initiate the reaction, and a sample without NADPH was used as a control. Aliquots (1 mL) were taken at T ) 0, 2, 5, 10, 20, and 30 min, and

Paroxetine BioactiVation to an Electrophilic Quinone Intermediate the reaction was terminated by adding 5 mL of acetonitrile. Samples were centrifuged (3000g, 5 min, 4 °C), and the supernatant was decanted. The pellet was subsequently washed and centrifuged, and the supernatant was decanted a total of four more times: one using acetonitrile and the final three washes using HPLC grade water. The final water wash supernatant was checked in the liquid scintillation counter (Wallac 1409 DSA, Perkin Elmer Life Sciences, IL) to ensure that counts were below background (approximately 40 dpm). The protein pellet was resuspended in 0.5 mL of 2 N NaOH at room temperature for 24 h. Water (0.5 mL) followed by 18 mL of scintillation fluid (Ultima Gold, Perkin Elmer, Boston, MA) was then added, and the radioactivity in each sample was counted. The effect of phase II metabolic pathways such as GSH conjugation, methylation, glucuronidation, and sulfation on covalent binding of [3H]paroxetine to liver proteins was also assessed in liver microsomes and S-9. Thus, microsomal (2 mg protein/mL) and/or S-9 (5 mg protein/mL) mixtures containing paroxetine (2.5 µM paroxetine/[3H]paroxetine, 0.10 µCi per tube) and NADPH (1.3 mM) were incubated at 37 °C for 20–30 min in the presence of GSH (5 mM), SAM (0.5 mM), UDPGA (5 mM):alamethicin (5 µg/mL):saccharolactone (1 mM), or PAPS (0.25 mM):alamethicin (5 µg/mL):saccharolactone (1 mM). Samples without NADPH were used as no reaction controls, and samples without phase II cofactors (GSH, SAM, UDPGA, and PAPS) were used to assess covalent binding. The following equation was used to calculate radiolabeled covalent binding (pmol/mg protein):

[

× substrate concentration (nmoL ⁄ mL) (radioactivity (dpm) ) radioactivity of substrate (dpm) microsomal protein concentration (mg ⁄ mL)

]

× 1000

Paroxetine Metabolism Studies in Human Liver Microsomes and S-9 Fractions. Stock solutions of paroxetine were prepared in methanol. The final concentration of methanol in the incubation media was 0.2% (v/v). Incubations were carried out at 37 °C for 60 min in a shaking water bath. The incubation volume was 1 mL and consisted of the following: 0.1 M potassium phosphate buffer (pH 7.4) containing MgCl2 (10 mM), human liver microsomes or S-9 preparations (final protein concentration ) 2 mg/mL), NADPH (1.3 mM), paroxetine (20 µM), and GSH (5 mM). Incubations that lacked either NADPH or GSH served as negative controls, and reactions were terminated by the addition of ice-cold acetonitrile (1 mL). Secondary phase II conjugation reactions (glucuronidation, sulfation, and methylation) of paroxetine were also assessed in human liver S-9 incubations via the addition of UDPGA (3 mM), PAPS (0.25 mM), and SAM (0.5 mM), respectively. The solutions were centrifuged (3000g, 15 min), and the supernatants were dried under a steady nitrogen stream. The residue was reconstituted with mobile phase and analyzed for metabolite formation by liquid chromatography tandem mass spectrometry (LC-MS/MS). Bioanalytical Methodology for Metabolite Identification. The separation of metabolites was achieved at ambient temperature on a Kromasil C4 100A column (3.5 µm, 150 mm × 2.0 mm; Phenomenex, Torrance, CA) by reverse phase chromatography. The mobile phase consisted of 0.1% formic acid (solvent A) and acetonitrile (solvent B) and was delivered at 0.200 mL/ min. A gradient was used to separate paroxetine and its metabolites as well as conjugates derived from trapping of reactive intermediates with GSH. The initial composition of solvent B was maintained at 1% for 10 min and then increased in a linear manner as follows: 30% at 28 min, 50% at 30 min, and 90% at 35 min. It was then maintained at 90% for up to 37 min and then decreased to 1% in the next 3 min. The column was allowed to equilibrate at 1% solvent B for 5 min before 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. Mass spectrometric analyses were performed on a ThermoFinnigan Deca XP ion trap mass spectrometer, which was interfaced to an Agilent HP-1100 HPLC system (Agilent Technologies, Palo Alto, CA) and equipped with an electrospray ionization source. The values for electrospray ionization were as follows: capillary temperature, 270 °C; spray

Chem. Res. Toxicol., Vol. 20, No. 11, 2007 1651 voltage, 4.0 kV; capillary voltage, 4.0 V; sheath gas flow rate, 90; and auxiliary gas flow rate, 30. The mass spectrometer was operated in a positive ion mode with data-dependent scanning. The ions were monitored over a full mass range of m/z 100–1000. 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 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 3.0, and the normalized collision energy was 45.0. Metabolites were identified by comparing t ) 0 samples to t ) 60 min samples (with or without cofactors), and structural information was generated from collision-induced dissociation (CID) spectra of the corresponding protonated molecular ions. P450 2D6 Inhibition Studies in Human Liver Microsomes and S-9 Fractions. For nonpreincubation (time-independent) P450 2D6 inhibition, a pooled human liver microsomal or S-9 preparation (0.25 mg/mL) was incubated with dextromethorphan (25 µM) and paroxetine at concentrations ranging from 0.5 to 25 µM at 37 °C for 20 min in the presence of NADPH (1.3 mM). For preincubationdependent (time-dependent) inhibition, microsomes or S-9 (protein concentration ) 0.25 mg/mL), NADPH (1.3 mM), and paroxetine at concentrations ranging from 0.5 to 25 µM were incubated for 30 min at 37 °C in the absence or presence of GSH (5 mM), after which incubation with dextromethorphan (25 µM) and a second aliquot of NADPH was carried out for 20 min. With human liver S-9 preincubation-dependent inhibition studies, the effect of addition of SAM (1 mM) on the time-dependent inhibition of P450 2D6 activity was also assessed. Incubations were quenched by the addition of 2 volumes of acetonitrile containing levallorphan (100 ng/mL) as internal standard. The remaining P450 2D6 activity as measured by the formation of dextrorphan was determined by LCMS/MS. Inhibition studies were conducted in triplicate. Following centrifugation at 3000g for 10 min, the supernatant was removed and placed in a 96 well 1 mL polystyrene block, dried under nitrogen, and then reconstituted in 100 µL of 30/70% acetonitrile/ 10 mM ammonium acetate with 1% isopropanol and 0.1% formic acid. The sample was injected (10 µL) onto a Waters AtlantisdC18 reverse phase column (2.1 mm × 30 mm, 3 µm) (Waters, Milford, MA). The analyte and internal standard were eluted using a linear gradient of 95% 10 mM ammonium acetate, 1% isopropronol and 0.1% formic acid/5% acetonitrile to 95% acetonitrile/ 5% 10 mM ammonium acetate, 1% isopropronol, and 0.1% formic acid at 0.45 mL/min between 2 and 3 min. Detection was accomplished by multiple reaction monitoring at m/z ) 285 f 157 for dextrorphan and at m/z ) 284 f 157 for internal standard in positive ionization mode on a API 4000 linear Ion trap coupled to a Shimadzu LC20 controller and LC20 vp liquid pumps. Data analysis was conducted as described previously (11). For all HPLC analyses, peak areas of the metabolite (dextrorphan) were expressed as a ratio to the internal standard (levallorphan) peak area for each concentration of the inhibitor. These peak area ratios represent the remaining dextromethorphan demethylase activity in the microsomes and S-9 and were expressed as a percentage of the timematched control samples without inhibitor.

Results Irreversible Binding of [3H]Paroxetine to Human Liver Microsomal Proteins. As shown in Figure 1, a high level of irreversible binding of radioactivity to microsomal proteins was observed when the [3H]paroxetine (final paroxetine concentration ) 25 µM) was incubated with human liver microsomes. The process was NADPH- and time-dependent with maximal incorporation of radioactivity (∼200 pmol bound/mg protein based on an average of two independent determinations) observed after a 30 min incubation with liver microsomes. The effect of GSH and UDPGA on covalent binding was also assessed in NADPH-supplemented human liver microsomal

1652 Chem. Res. Toxicol., Vol. 20, No. 11, 2007

Figure 1. Time course of [3H]paroxetine covalent binding to human liver microsomes. The determination of the extent of radioactive covalent binding over time was assessed by incubating paroxetine (25 µM paroxetine/[3H]paroxetine, 0.151 µCi per tube) at 37 °C in human liver microsomes (2 mg protein/mL) containing 100 mM potassium phosphate buffer (pH 7.4) and 3.3 mM MgCl2. NADPH (1.3 mM final) was added to initiate the reaction, and samples without NADPH were used as controls.

Figure 2. Effect of GSH and UDPGA on the covalent binding of [3H]paroxetine to human liver microsomal proteins. Microsomal incubations contained paroxetine (2.5 µM paroxetine/[3H]paroxetine, 0.10 µCi per tube), NADPH (1.3 mM), 5 mM GSH or 5 mM UDPGA, 5 µg/mL alamethicin, and 1 mM saccharolactone. Incubations were conducted at 37 °C for 30 min. Data are expressed as means ( SD of three independent determinations. Samples without NADPH were used as no reaction controls, and samples without GSH or UDPGA/ alamethicin/saccharolactone were used as covalent binding controls.

incubations containing [3H]paroxetine at a pharmacologically relevant concentration of 2.5 µM. As is shown in Figure 2, the addition of GSH to the incubation medium significantly reduced (by >90%) the level of irreversible binding of radioactivity, whereas the addition of the uridine 5′-diphosphoglucuronosyl transferase (UGT) cofactor UDPGA resulted in a net reduction of covalent binding by ∼30%. Irreversible Binding of [3H]Paroxetine to Human Liver S-9 Proteins. Covalent incorporation of radioactivity in human liver S-9 was also observed upon incubation of [3H]paroxetine (final paroxetine concentration ) 25 µM) in a time- and NADPH-dependent fashion with maximal incorporation of radioactivity (∼70 pmol bound/mg protein based on the average of two independent experiments) observed after a 20 min incubation with S-9 (Figure 3). The effect of various cofactors on covalent binding was also assessed in NADPH-supplemented human liver S-9 incubations containing [3H]paroxetine at a

Zhao et al.

Figure 3. Time course of [3H]paroxetine covalent binding to human liver S-9. The determination of the extent of radioactive covalent binding over time was assessed by incubating paroxetine (25 µM paroxetine/[3H]paroxetine, 0.151 µCi per tube) at 37 °C in human liver S-9 (5 mg protein/mL) containing 100 mM potassium phosphate buffer (pH 7.4) and 3.3 mM MgCl2. NADPH (1.3 mM final) was added to initiate the reaction, and samples without NADPH were used as controls.

Figure 4. Effect of phase 2 enzyme cofactors on the covalent binding of [3H]paroxetine to human liver S-9 proteins. Microsomal incubations contained paroxetine (2.5 µM paroxetine/[3H]paroxetine, 0.10 µCi per tube), NADPH (1.3 mM), 5 mM GSH or 5 mM UDPGA, 5 µg/mL alamethicin, and 1 mM saccharolactone or PAPS (0.25 mM) or SAM (0.5 mM). Incubations were conducted at 37 °C for 30 min. Data are expressed as means ( SD of three independent determinations. Samples without NADPH were used as no reaction controls, and samples without phase 2 enzyme cofactors were used as covalent binding controls.

physiologically relevant concentration of 2.5 µM. As is shown in Figure 4, the addition of GSH and SAM to the incubation medium significantly reduced covalent binding by ∼70 and 56%, respectively. The effect of UDPGA and PAPS, cofactors of UGT and sulfotransferase (ST) enzymes, on the reduction of covalent binding was less pronounced. Effect of Exogenously Added GSH on the Preincubation-Dependent Inhibition of P450 2D6 Activity in Human Liver Microsomes and S-9. Consistent with published findings (11), a time- and concentration-dependent inhibitory component was evident for paroxetine, as the IC50 value for P450 2D6-mediated dextromethorphan-O-demethylase activity was significantly reduced from 3.95 ( 0.19 to 0.19 ( 0.03 µM (i.e., more potent inhibition) when the agent was preincubated with NADPH-supplemented human liver microsomes. Likewise, the IC50 value for P450 2D6-mediated

Paroxetine BioactiVation to an Electrophilic Quinone Intermediate

Chem. Res. Toxicol., Vol. 20, No. 11, 2007 1653

Figure 5. Product ion spectrum obtained by CID of the MH+ ion (m/z 623) of GS1 (Rt ) 18.7 min) from NADPH-supplemented human liver microsomal incubations containing paroxetine (10 µM) and GSH (3 mM). The origins of the characteristic ions are as indicated.

dextromethorphan-O-demethylase activity in NADPH-supplemented human liver S-9 preparations was reduced from 1.29 ( 0.13 to 0.11 ( 0.02 µM. The addition of SAM (1 mM) to liver S-9 incubations or the addition of GSH (5 mM) to liver microsomal or S-9 incubations did not alter the IC50 value for time-dependent inhibition of P450 2D6 activity by paroxetine [IC50 (+SAM), 0.11 ( 0.01 µM; IC50 (+GSH), 0.10 ( 0.02 µM]. Overall, these findings suggest that the paroxetine reactive intermediate responsible for P450 2D6 inactivation did not leave the active site to react with other proteins. Metabolism/Bioactivation of Paroxetine in Human Liver Microsomes and S-9 Fractions. 1. Microsomes. LCMS/MS analysis of extracts of human liver microsomal incubations containing paroxetine (10 µM), NADPH, and GSH revealed the presence of two GSH conjugates [monoglutathionyl conjugate of paroxetine catechol (GS1): MH+ ) 623, retention time (Rt) ) 18.7 min] and 5-(3-(3-((carboxymethyl)carbamoyl)7-((4S)-4-(4-fluorophenyl)piperidin-3-yl)methoxy)-5-oxo-3,5-dihydro-2H-benzo[b][1,4]thiazin-8-ylthio)-1-(carboxymethylamino)1-oxopropan-2-ylamino)-2-amino-5-oxopentanoic acid (GS2): MH+ ) 779, Rt ) 24.4 min), none of which were evident in incubations that lacked either of the cofactors. The molecular weight of GS1 was consistent with the addition of one molecule of GSH to paroxetine catechol. The product ion spectrum of GS1 obtained by CID of the MH+ at m/z 623 (Figure 5) produced fragment ions at m/z 494, 350, and 192. The fragment ion at m/z 192 was also present in the product ion spectrum of paroxetine and was assigned to the 4-fluorophenylpiperidine motif. The fragment ion at m/z 494 was consistent with the elimination of the pyroglutamate residue of GSH (19), whereas the fragment ion at m/z 350 was assigned as a cleavage adjacent to the cysteinyl thioether moiety with charge retention on the paroxetine–catechol residue. The occurrence of the fragment ion at m/z 350 is consistent with the presence of an aromatic thioether motif in GS1 (19). A proposed structure for GS1, which is consistent with the product ion spectrum, is shown in Figure 5. The molecular weight of GS2 suggested the addition of 449 Da to paroxetine. The product ion spectrum of GS2 obtained by CID of the MH+ at m/z 779 (Figure 6) produced fragment ions at m/z 650, 506, 474, and 313. The fragment ion at m/z 650 was consistent with the elimination of the pyroglutamate residue of GSH, suggesting the presence of at least one GSH

moiety. Thus, the addition of 449 Da to paroxetine can be rationalized as 305 Da, plus the remainder 144 Da. On the basis of these observations, a general structure, consistent with the observed fragmentation pattern, is shown in Figure 6 for GS2. The observed UV spectrum of GS2 (Figure 6, inset) depicts a λmax at 366 nm that is consistent with the nonaromatic keto form of GS2 as opposed to the aromatic enol form that could be present as well (20). The structure depicted in Figure 6 and Scheme 2, which represents one of the two possible regioisomers, is shown for illustrative purposes only. The mass spectrum did not provide additional insight into the regiochemistry of GS2. 2. S-9. In addition to GS1, LC-MS/MS analysis of extracts of human liver S-9 incubations containing paroxetine (10 µM), NADPH, and GSH revealed the presence of two isomeric GSH conjugates diglutathionyl conjugate of paroxetine catechol (GS3a) (Rt ) 15.9 min) and GS3b (Rt ) 16.5 min) with molecular ion 928 (MH+). The cyclized GSH conjugate GS2 observed in human liver microsomes was not observed in NADPH- and GSH-supplemented human S-9 incubations of paroxetine. The molecular weight of GS3a/GS3b and the product ion spectrum obtained by CID of the MH+ at m/z 928 were consistent with the addition of two molecules of GSH to paroxetine–catechol as shown with a general structure in Figure 7. Diagnostic fragment ions at m/z 797 and m/z 670 were derived with the sequential elimination of the pyroglutamate residue within the two GSH residues. Incidentally, the addition of SAM, UDPGA, and PAPS to the NADPH- and GSH-supplemented human liver S-9 incubations with paroxetine led to complete disappearance of both GS1 and GS3a/GS3b. In the place of the GSH conjugates, there appeared the previously reported metabolites of paroxetine (8, 9) including the regioisomeric guaiacol metabolites (Rt ) 21.5 and 22.4 min) derived from O-methylation of paroxetine–catechol and the downstream glucuronide (Rt ) 18.9 and 20.5 min) and sulfate conjugates (Rt ) 21.3 min) obtained via conjugation of the guaiacol derivatives. The corresponding CID spectra of these metabolites are shown in Figure 8. The CID spectra of the glucuronide and sulfate conjugates of the guaiacol metabolite(s) were identical.

Discussion In this report, we have described findings demonstrating that paroxetine can covalently bind to human hepatic tissue in vitro

1654 Chem. Res. Toxicol., Vol. 20, No. 11, 2007

Zhao et al.

Figure 6. Product ion spectrum obtained by CID of the MH+ ion (m/z 779) of GS2 (Rt ) 24.4 min) from NADPH-supplemented human liver microsomal incubations containing paroxetine (10 µM) and GSH (3 mM) (A). Panel B depicts the MS3 spectrum of fragment ion m/z 650 observed in the product ion spectrum of GS2. The corresponding UV spectrum of GS2 is also shown (inset). The origins of the characteristic ions are as indicated. The structure depicted represents one of the two possible regioisomers. The exact regiochemistry is not known, and the drawn structure is for illustration only.

Scheme 2. Proposed Mechanism of Paroxetine Bioactivation in Human Liver Microsomes and S-9

and that inclusion of GSH and/or cofactors for phase 2 enzymes, involved in conjugation of phenols/catechols, can dramatically diminish covalent binding. Paroxetine exhibits nonlinear pharmacokinetics in the clinic (8, 21), a phenomenon that is consistent with the agent being both a substrate (22) and an inhibitor of P450 2D6 (11, 17, 18). Paroxetine possesses a methylenedioxyphenyl group, a structural alert, known to exhibit mechanism-based inactivation of P450 enzymes. P450-catalyzed metabolism of the methylenedioxyphenyl substituent results in initial hydroxylation at the methylene carbon. This unstable intermediate can partition between demethylenation yielding a catechol intermediate and formaldehyde/formate and dehydration to a carbene (10, 23). Carbenes can form quasi-irreversible metabolite–inhibitor (MI) complexes with the P450Fe(II) form of the enzyme in a time-dependent fashion, thereby resulting in enzyme inactivation (10, 23). Such MI complexes exhibit an

absorbance maximum in the Soret region between 448 and 456 nm when the heme iron is in the Fe(II) state. The observation that time-dependent inactivation of P450 2D6 by paroxetine is also accompanied by an increase in absorbance at 456 nm (11) suggests that paroxetine inactivation of P450 2D6 indeed occurs via the formation of a MI complex. Besides the formation of reactive carbene intermediates that inactivate P450, demethylenation of the methylenedioxyphenyl group will lead to the formation of catechol metabolites. Catechols are precursors to ortho-quinones, which can cause toxicity via (i) covalent interactions with biomacromolecules, (ii) GSH depletion, and/ or (iii) oxidative stress due to redox cycling with the corresponding semiquinones/catechol forms (24–27). Several catechols are also good substrates for COMT, UGT, and ST enzymes, which can potentially circumvent the oxidation pathway to electrophilic quinones (28, 29).

Paroxetine BioactiVation to an Electrophilic Quinone Intermediate

Figure 7. Product ion spectrum obtained by CID of the MH+ ion (m/z 928) of GS3a (Rt ) 15.9 min) from NADPH-supplemented human liver S-9 incubations containing paroxetine (10 µM) and GSH (3 mM) (A). Panel B depicts the MS3 spectrum of fragment ion m/z 670 observed in the product ion spectrum of GS3a. The product ion spectrum of GS3b (Rt ) 16.5 min) was identical to that of GS3a. The origins of the characteristic ions are as indicated.

Our findings on the NADPH-dependent covalent binding of [3H]paroxetine to proteins in liver microsomal and S9 preparations represent a classical assessment of bioactivation potential of xenobiotics and drugs (2). In the absence of any additional metabolism data, such an isolated finding can be interpreted as being a harbinger of a potential toxicological response in the clinic. However, in vitro incubation of xenobiotics with cofactors

Chem. Res. Toxicol., Vol. 20, No. 11, 2007 1655

that only support P450 activity can be misleading in covalent binding assessments since the full complement of drugmetabolizing enzymes is not active. There are a few examples where the precursor to a reactive intermediate is also known to undergo competing biotransformation by phase 2 enzyme(s) to “nonreactive” metabolites (3). This is what was observed in the present analysis; while paroxetine could be bioactivated by P450 to a reactive intermediate that covalently bound to microsomal or S-9 protein, when a more complete incubation was done, this covalent binding was significantly reduced. For instance, inclusion of SAM in the S-9 incubations resulted in diminished covalent binding due to competing O-methylation (vs oxidation to ortho-quinone) of the intermediate catechol to the corresponding guaiacol isomers. The formation of the guaiacol regioisomers and their further metabolism to glucuronide/sulfate conjugates as observed in the course of our metabolism studies is consistent with previous reports on paroxetine metabolism in humans (8, 9). Adduction of reactive intermediates with GSH represents one of the most important pathways leading to detoxication of xenobiotics including drugs. When GSH was included in liver microsome and S-9 incubations at physiologically relevant concentrations, covalent binding by [3H]paroxetine was greatly diminished. Consistent with this finding, the formation of several GSH conjugates was discernible in microsomes and S-9 incubations indicating that GSH had reacted with the orthoquinone. Whether the conjugation of GSH with the orthoquinone is facilitated by GSH-S-transferase remains unclear at the present time. However, it is interesting to point out that no sulfydryl conjugates of paroxetine are discernible in microsomal incubations supplemented with dansylated GSH (30). Three types of GSH adducts were observed, namely, GS1, GS2, and GS3 (in the form of two isomers, GS3a and GS3b). The

Figure 8. Product ion spectra obtained by CID of the MH+ ions of paroxetine metabolites following incubation of paroxetine (10 µM) with NADPH-, GSH-, SAM-, UDPGA-, and PAPS-supplemented human liver S-9 protein. The origins of the characteristic ions obtained via CID of the MH+ of the various paroxetine metabolites are also depicted. The exact regiochemistry of methylation of the catechol and subsequent glucuronidation/ sulfation conjugation is not known, and the drawn structures are for illustration only. Furthermore, the CID spectra of the two observed glucuronide isomers (Rt ) 18.9 min and 20.5 min) and the sulfate conjugate (Rt ) 21.3 min) were identical.

1656 Chem. Res. Toxicol., Vol. 20, No. 11, 2007

molecular ion at m/z 623 (MH+) of GS1 is consistent with simple Michael addition of one molecule of GSH to the orthoquinone (Scheme 2). The formation of GS2 (MH+ ) 779) can be envisioned to arise from rearrangement of a di-GSH adduct, in which cyclization and loss of a glutamic acid residue from one of the GSH molecules has occurred (see Scheme 2). Precise mechanistic details for the formation of GS2 remain unclear, but it is not unprecedented since structurally analogous GSH conjugates have been observed with catechol derivatives including dopamine and its analogues (25, 31, 32). The molecular ion at m/z 928 (MH+) of GS3a/GS3b is consistent with Michael addition of a second molecule of GSH to GS1 and is indicative of further two-electron oxidation of the catechol motif in GS1 to the ortho-quinone (see Scheme 2), similar to previously published reports on analogous catechols (26, 33, 34). The regiochemistry of GSH attachment to the phenyl ring in GS1, GS2, or GS3a/GS3b could not be determined by NMR spectroscopy since these conjugates were unstable under the conditions of the preparative chromatography for their purification. Our findings on the covalent binding of paroxetine to liver proteins via bioactivation contrast with the excellent safety record of this drug, especially when considering the link between the bioactivation and the occurrence of idiosyncratic hepatotoxicity with drugs. The answer to this may lie in one or more of several plausible explanations. First, relative to other drugs believed to cause idiosyncratic hepatotoxicity via covalent modification of “critical” proteins and/or oxidative stress mechanisms, paroxetine is dosed at a relatively low amount per day. A standard daily regimen of paroxetine is 20 mg/day, while other drugs known to be associated with idiosyncratic adverse drug reactions are dosed at considerably greater levels—acetaminophen is dosed at 4 g/day, felbamate at 2.4 g/day, and several NSAIDs known to be hepatotoxic are (or were, if already removed from clinical use) dosed at several hundred milligrams per day. Thus, the total body burden to reactive metabolite exposure is likely to be considerably less for paroxetine as compared with other hepatotoxic drugs and may not exceed a threshold needed for eliciting a toxicological response. Second, the present data showed that covalent binding was markedly reduced when the incubations were conducted in the presence of other cofactors needed for the activities of enzymes commonly associated with the detoxication of reactive intermediates. GSH reduced covalent binding the most. This is analogous to the case of acetaminophen, where hepatotoxicity is observed at doses believed to exceed the capacity of the liver to form GSH adducts of the reactive quinone–imine intermediate of acetaminophen. The amount of ortho-quinone formed in vivo after administration of paroxetine at 20 mg/day may be readily handled by the liver’s pool of GSH, as well as the alternate metabolism of the paroxetine catechol via O-methylation. While the formation of sulfydryl conjugates of paroxetine catechol has not been demonstrated in vivo, mass balance studies do reveal a significant proportion of unidentified polar metabolite(s) in urine and feces from human volunteers (8). It is likely that some of the unidentified metabolites may be the GSH and/or mercapturic acid derivatives of paroxetine catechol. Finally, while the P450 2D6-mediated bioactivation of the methylenedioxyphenyl group, which ultimately can give rise to the catechol and ortho-quinone, appears to be responsible for the metabolism of paroxetine after a single dose, it is known that the clearance of paroxetine in P450 2D6 extensive metabolizers drops with repeated dosing, presumably due to sustained inactivation of P450 2D6 (35). Thus, upon chronic dosing of paroxetine, other clearance mechanisms that are ordinarily not observed with

Zhao et al.

single dosing may be the important pathways of paroxetine clearance. With P450 2D6 inactivated, the amount of paroxetine undergoing metabolism at the methylenedioxyphenyl group to the catechol and ortho-quinone would be lower than anticipated. Each of these hypotheses remains speculative without further experimentation, but an overall conclusion that can be made from these findings is that covalent binding observed in vitro cannot necessarily be assumed to be predictive of toxicity. Other factors are also likely involved such as the total dose of drug, the amount that proceeds through a metabolic pathway with a chemically reactive intermediate, and knowledge of the contribution of metabolism that would detoxicate the reactive intermediate. Efforts to better understand these relationships are currently ongoing in our laboratories and will be reported in due course. Acknowledgment. We dedicate this article to Prof. Lawrence J. Marnett on the occasion of his 60th birthday.

References (1) Baillie, T. A. (2006) Future of toxicology—Metabolic activation and drug design: Challenges and opportunities in chemical toxicology. Chem. Res. Toxicol. 19, 889–893. (2) Evans, D. C., Watt, A. P., Nicoll-Griffith, D. A., and Baillie, T. A. (2004) Drug-protein adducts: an industry perspective on minimizing the potential for drug bioactivation in drug discovery and development. Chem. Res. Toxicol. 17, 3–16. (3) Kalgutkar, A. S., and Soglia, J. R. (2005) Minimising the potential for metabolic activation in drug discovery. Exp. Opin. Drug Metab. Toxicol. 1, 91–142. (4) Kalgutkar, A. S., Gardner, I., Obach, R. S., Shaffer, C. L., Callegari, E., Henne, K. R., Mutlib, A. E., Dalvie, D. K., Lee, J. S., Nakai, Y., O’Donnell, J. P., Boer, J., and Harriman, S. P. (2005) A comprehensive listing of bioactivation pathways of organic functional groups. Curr. Drug Metab. 6, 161–225. (5) Colakoglu, O., Tankurt, E., Unsal, B., Ugur, F., Kuperlioglu, A., Buyrac, Z., and Akpinar, Z. (2005) Toxic hepatitis associated with paroxetine. Int. J. Clin. Pract. 59, 861–862. (6) Azaz-Livshits, T., Hershko, A., and Ben-Chetrit, E. (2002) Paroxetine associated hepatotoxicity: A report of 3 cases and a review of the literature. Pharmacopsychiatry 35, 112–115. (7) Benbow, S. J., and Gill, G. (1997) Paroxetine and hepatotoxicity. Br. Med. J. 314, 1387. (8) Kaye, C. M., Haddock, R. E., Langley, P. F., Mellows, G., Tasker, T. C., Zussman, B. D., and Greb, W. H. (1989) A review of the metabolism and pharmacokinetics of paroxetine in man. Acta Psychiatr. Scand. Suppl. 350, 60–75. (9) Haddock, R. E., Johnson, A. M., Langley, P. F., Nelson, D. R., Pope, J. A., Thomas, D. R., and Woods, F. R. (1989) Metabolic pathways of paroxetine in animals and man and the comparative pharmacological properties of the its metabolites. Acta Psychiatr. Scand. Suppl. 350, 24–26. (10) Murray, M. (2000) Mechanisms of inhibitory and regulatory effects of methylenedioxyphenyl compounds on cytochrome P450-dependent drug oxidation. Curr. Drug Metab. 1, 67–84. (11) Bertelsen, K. M., Venkatakrishnan, K., Von Moltke, L. L., Obach, R. S., and Greenblatt, D. J. (2003) Apparent mechanism-based inhibition of human CYP2D6 in vitro by paroxetine: Comparison with fluoxetine and quinidine. Drug Metab. Dispos. 31, 289–293. (12) Brøsen, K., Hansen, J. G., Nielsen, K. K., Sindrup, S. H., and Gram, L. F. (1993) Inhibition by paroxetine of desipramine metabolism in extensive but not in poor metabolizers of sparteine. Eur. J. Clin. Pharmacol. 44, 349–355. (13) Alderman, J., Preskorn, S. H., Greenblatt, D. J., Harrison, W., Penenberg, D., Allison, J., and Chung, M. (1997) Desipramine pharmacokinetics when coadministered with paroxetine or sertraline in extensive metabolizers. J. Clin. Phychopharmacol. 17, 284–291. (14) Hemeryck, A., Lefebvre, R. A., De Vriendt, C., and Belpaire, F. M. (2000) Paroxetine affects metoprolol pharmacokinetics and pharmacodynamics in healthy volunteers. Clin. Pharmacol. Ther. 67, 283– 291. (15) Spina, E., Avenoso, A., Facciola, G., Scordo, M. G., Ancione, M., and Madia, A. (2001) Plasma concentrations of risperidone and 9-hydroxyrisperidone during combined treatment with paroxetine. Ther. Drug Monit. 23, 223–227. (16) Belle, D. J., Ernest, C. S., Sauer, J. M., Smith, B. P., Thomasson, H. R., and Witcher, J. W. (2002) Effect of potent CYP2D6 inhibition

Paroxetine BioactiVation to an Electrophilic Quinone Intermediate

(17)

(18)

(19) (20) (21)

(22)

(23)

(24) (25) (26)

by paroxetine on atomoxetine pharmacokinetics. J. Clin. Pharmacol. 42, 1219–1227. Venkatakrishnan, K., and Obach, R. S. (2005) In vitro-in vivo extrapolation of CYP2D6 inactivation by paroxetine: Prediction of nonstationary pharmacokinetics and drug interaction magnitude. Drug Metab. Dispos. 33, 845–852. Obach, R. S., Walsky, R. L., and Venkatakrishnan, K. (2007) Mechanism-based inactivation of human cytochrome P450 enzymes and the prediction of drug-drug interactions. Drug Metab. Dispos. 35, 246–255. Baillie, T. A., and Davis, M. R. (1993) Mass spectrometry in the analysis of glutathione conjugates. Biol. Mass Spectrom. 22, 319– 325. Napolitano, A., Memoli, S., and Prota, G. (1999) A new insight in the biosynthesis of pheomelanins: characterization of a labile 1,4benzothiazine intermediate. J. Org. Chem. 64, 3009–3011. Sindrup, S. H., Brøsen, K., and Gram, L. F. (1992) Pharmacokinetics of the selective serotonin reuptake inhibitor paroxetine: Nonlinearity and relation to the sparteine oxidation polymorphism. Clin. Pharmacol. Ther. 51, 288–295. Bloomer, J. C., Woods, F. R., Haddock, R. E., Lennard, M. S., and Tucker, G. T. (1992) The role of cytochrome P4502D6 in the metabolism of paroxetine by human liver microsomes. Br. J. Clin. Pharmacol. 33, 521–523. Fukuto, J. M., Kumagai, Y., and Cho, A. K. (1991) Determination of the mechanism of demethylenation of (methylenedioxy)phenyl compounds by cytochrome P450 using deuterium isotope effects. J. Med. Chem. 34, 2871–2876. Monks, T. J., and Jones, D. C. (2002) The metabolism and toxicity of quinones, quinonimines, quinone methides, and quinone-thioethers. Curr. Drug Metab. 3, 425–438. Bolton, J. L., Trush, M. A., Penning, T. M., Dryhurst, G., and Monks, T. J. (2000) Roles of quinones in toxicity. Chem. Res. Toxicol. 13, 135–160. Monks, T. J., Hanzlik, R. P., Cohen, G. M., Ross, D., and Graham, D. G. (1992) Quinone chemistry and toxicity. Toxicol. Appl. Pharmacol. 112, 2–16.

Chem. Res. Toxicol., Vol. 20, No. 11, 2007 1657 (27) Bolton, J. L. (2002) Quinoids, quinoid radicals, and phenoxyl radicals formed from estrogens and antiestrogens. Toxicology 177, 55–65. (28) Butterworth, M., Lau, S. S., and Monks, T. J. (1996) 17 beta-Estradiol metabolism by hamster hepatic microsomes. Implications for the catechol-O-methyl transferase-mediated detoxication of catechol estrogens. Drug Metab. Dispos. 24, 588–594. (29) Taskinen, J., Ethell, B. T., Pihlavisto, P., Hood, A. M., Burchell, B., and Coughtrie, M. W. (2003) Conjugation of catechols by recombinant human sulfotransferases, UDP-glucuronosyltransferases, and soluble catechol O-methyltransferase: Structure-conjugation relationships and predictive models. Drug Metab. Dispos. 31, 1187–1197. (30) Gan, J., Harper, T. W., Hsueh, M. M., Qu, Q., and Humphreys, W. G. (2005) Dansyl glutathione as a trapping agent for the quantitative estimation and identification of reactive metabolites. Chem. Res. Toxicol. 18, 896–903. (31) Samuel, K., Yin, W., Stearns, R. A., Tang, Y. S., Chaudhary, A. G., Jewell, J. P., Lanza, T., Lin, L. S., Hagmann, W. K., Evans, D. C., and Kumar, S. (2003) Addressing the metabolic activation potential of new leads in drug discovery: A case study using ion trap mass spectrometry and tritium labeling techniques. J. Mass Spectrom. 38, 211–221. (32) Wong, K. S., Goyal, R. N., Wrona, M. Z., Blank, C. L., and Dryhurst, G. (1993) 7-S-glutathionyl-tryptamine-4,5-dione: A possible aberrant metabolite of serotonin. Biochem. Pharmacol. 46, 1637–1652. (33) Erve, J. C., Svensson, M. A., von Euler-Chelpin, H., and KlassonWehler, E. (2004) Characterization of glutathione conjugates of the remoxipride hydroquinone metabolite NCQ-344 formed in vitro and detection following oxidation by human neutrophils. Chem. Res. Toxicol. 17, 564–571. (34) Miller, R. T., Lau, S. S., and Monks, T. J. (1997) 2,5-Bis-(glutathionS-yl)-alpha-methyldopamine, a putative metabolite of (+/-)-3,4methylenedioxyamphetamine, decreases brain serotonin concentrations. Eur. J. Pharmacol. 323, 173–180. (35) Ozdemir, V., Tyndale, R. F., Reed, K., Herrmann, N., Sellers, E. M., Kalow, W., and Naranjo, C. A. (1999) Paroxetine steady-state plasma concentration in relation to CYP2D6 genotype in extensive metabolizers. J. Clin. Psychopharmacol. 19, 472–475.

TX700132X