Abstract
Several new glutathione adducts (M3–M7) of trazodone were tentatively identified in human liver microsomal incubations using liquid chromatography-tandem mass spectrometry (LC/MS/MS). Following incubations with trazodone in the presence of glutathione, 1-(3′-chlorophenyl)piperazine (m-CPP), a major circulating and pharmacologically active metabolite of several antidepressants including trazodone, nefazodone, and etoperidone, was trapped with glutathione to afford the corresponding quinone imine-sulfydryl adducts M4 and M5. Two novel glutathione adducts of deschloro-m-CPP and deschloro-trazodone, M3 and M6, were also detected by tandem mass spectrometry. The identities of these m-CPP-derived glutathione adducts were further confirmed by LC/MS/MS analyses of microsomal incubations of m-CPP. To investigate the bioactivation mechanism, a regioisomer of m-CPP, 1-(4′-chlorophenyl)piperazine, was incubated in human liver microsomes. Blockage of bioactivation by 4′-chloro-substitution at least partially suggested that formation of m-CPP-derived glutathione adducts M3, M4, and M5 is mediated by a common quinone imine intermediate. A tentative pathway states that upon formation of the trazodone- and m-CPP-1′,4′-quinone imine intermediates through initial 4′-hydroxylation, glutathione attacks at the chlorine position by an ipso substitution, resulting in 4′-hydroxy-3′-glutathion-deschloro-trazodone (M6) and 4′-hydroxy-3′-glutathion-deschloro-m-CPP (M3), respectively. In contrast to CYP3A4-dependent bioactivation of trazodone itself, formation of M4 was mediated specifically by CYP2D6, as evidenced by cDNA-expressed CYP2D6-catalyzing formation of M4 from m-CPP, strong inhibition of formation of M4 by quinidine, a specific CYP2D6 inhibitor, in both incubations of trazodone and m-CPP with human liver microsomes, and concentration-dependent inhibition of M4 formation by quinidine.
Trazodone is a second-generation triazolopyridinone antidepressant drug (Scheme 1), which is structurally distinct from selective serotonin reuptake inhibitors, tri- and tetracyclics, and monoamine oxidase inhibitors. It is thought to act through combined 5-HT2 antagonism and 5-HT reuptake blockage (Haria et al., 1994). Trazodone is often coprescribed with other antidepressants as a sleep-inducing agent because of its more sedating and less anticholinergic side effects. Despite its therapeutic benefits, treatment with trazodone has been associated with rare but severe incidence of hepatic injury (Chu et al., 1983; Longstreth and Hershman, 1985; Beck et al., 1993; Hull et al., 1994), which is often described as idiosyncratic toxicity. Although the exact mechanism of trazodone hepatotoxicity is not clearly understood, a probable causal link between trazodone use and the onset of hepatic injury has been established (Fernandes et al., 2000; Rettman and McClintock, 2001).
As shown in Scheme 1, trazodone contains a triazolopyridinone moiety and a 3-chlorophenylpiperazine ring system. In humans, trazodone undergoes extensive hepatic metabolism mainly by hydroxylation, N-dealkylation, and N-oxidation (Baiocchi et al., 1974; Yamato et al., 1974; Jauch et al., 1976). Of particular interest in the biotransformation pathways of trazodone in humans is the detection and characterization of a dihydrodiol metabolite and 4′-hydroxytrazodone as major metabolites in urine (Baiocchi et al., 1974; Jauch et al., 1976). Formation of the dihydrodiol metabolite can presumably occur by nucleophilic addition of water to an electrophilic epoxide intermediate. On the other hand, 4′-hydroxytrazodone can undergo a two-electron oxidation leading to formation of an electrophilic quinone imine intermediate, which is capable of reacting with cellular proteins and other nucleophiles such as glutathione (Scheme 1). In human liver microsomal incubations, two glutathione adducts have been previously identified (Kalgutkar et al., 2005a), arising from quinone imine and epoxide intermediates (M1 and M2 depicted in Scheme 1). The epoxidation of triazolopyridinone, para-hydroxylation of 3-chlorophenylpiperazine, and subsequent oxidation to quinone imine were shown to be mediated by cytochrome P450 (P450) 3A4 (Kalgutkar et al., 2005a). These findings are significant as the first line of evidence to suggest that cytochrome P450-mediated reactive metabolites may play an important role in toxicity of the drug.
It is noteworthy that 1-(3′-chlorophenyl)piperazine (m-CPP, Scheme 1), resulting from N-dealkylation of trazodone, is a major circulating metabolite in humans common to several antidepressants including trazodone, nefazodone, and etoperidone (Melzacka et al., 1980; Fong et al., 1982; Otani et al., 1997; von Moltke et al., 1999). The metabolite m-CPP is of significant interest because it has 5-HT2C agonistic and 5-HT2A antagonistic properties (Conn and Sanders-Bush, 1987; Fiorella et al., 1995). It has also been suggested that m-CPP may contribute to the antidepressant efficacy of trazodone (Maes et al., 1997). Although oxidative metabolism of trazodone, nefazodone, and etoperidone are controlled by CYP3A4 (Rotzinger et al., 1998a; von Moltke et al., 1999; Zalma et al., 2000; Yan et al., 2002), 4′-hydroxylation of m-CPP is specifically mediated by CYP2D6 (Rotzinger et al., 1998b). Consistently, a clinical study showed that haloperidol, an inhibitor of CYP2D6, significantly increased plasma concentration of m-CPP, but not of trazodone (Mihara et al., 1997). Because 4′-hydroxy-m-CPP may undergo further two-electron oxidations to form an electrophilic quinone imine intermediate, characterization of reactive metabolites and responsible P450 enzymes would be very helpful for complete understanding of the biochemical mechanisms of idiosyncratic toxicity associated with m-CPP-containing drugs, such as trazodone. In this study, we report a total of five products of reactive metabolite formation in human liver microsomal incubations of trazodone, three of which were generated from bioactivation of the active metabolite m-CPP. Two novel deschloro glutathione adducts of m-CPP and trazodone presumably derived from an ipso substitution of chlorine by glutathione (GSH) were tentatively identified. In addition, it was found that formation of reactive metabolites of m-CPP was specifically mediated by CYP2D6. These data are important for further understanding the relationship between metabolic activation and hepatotoxicity of m-CPP-containing antidepressant drugs.
Materials and Methods
Materials. Reagents and solvents used in the current study were of the highest grade commercially available. The following chemicals were purchased from Sigma-Aldrich (St. Louis, MO): acetaminophen, dextromethorphan, dextrorphan, GSH, ketoconazole, α-naphthoflavone, trazodone, nefazodone, phenacetin, testosterone, 6β-hydroxy-testosterone, tolbutamide, sulfaphenazole, quinidine, trichloroacetic acid, NADPH, m-CPP hydrochloride, and 1-(4′-chlorophenyl)piperazine (p-CPP). Tranylcypromine was obtained from BIOMOL Research Laboratories (Plymouth Meeting, PA). (S)-Mephenytoin, 4′-hydroxy-(S)-mephenytoin, and 4-hydroxy-tolbutamide were purchased from SAFC Corp. (St. Louis, MO). Pooled human liver microsomes and supersomes containing cDNA-baculovirus-insect cell-expressed P450s (CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP2E1, CYP3A4, and CYP3A5) were obtained from BD Gentest (Woburn, MA). Formic acid, methanol, and acetonitrile were purchased from EM Scientific (Gibbstown, NJ).
Instrumentation. LC/MS/MS analyses were performed on an API 4000 Q-Trap hybrid triple quadrupole linear ion trap mass spectrometer (Applied Biosystems, Foster City, CA) interfaced online with a Shimadzu HPLC system (Shimadzu, Kyoto, Japan). For complete profiling of reactive metabolites, the precursor ion (PI) scan of m/z 272 was run in the negative mode with 0.2-Da step size, 5-ms pause between mass ranges, and 2-s scan rate or 50-ms dwell. The TurboIonSpray ion source conditions were optimized and set as follows: curtain gas = 35, collision gas = medium, ion spray voltage =-4500, and temperature = 500. Nitrogen was used as the nebulizer and auxiliary gas. Information dependent acquisition was used to trigger acquisition of enhanced product ion (EPI) spectra. The EPI scans were run in the positive mode at a scan range for daughter ions from m/z 100 to 1000. For NL-EPI analysis, the ion source conditions were set as follows: curtain gas = 35, collision gas = 6, ion spray voltage = 4500, and temperature = 500. Data were processed using Analyst 4.1 software (Applied Biosystems).
A Shimadzu HPLC system was coupled with an Agilent Eclipse XDB-Phenyl C18 column (3.0 × 150 mm, 3.5 μm; Agilent Technologies, Palo Alto, CA). The HPLC mobile phase A was 10 mM ammonium acetate in water with 0.1% formic acid, and the mobile phase B was acetonitrile with 0.1% formic acid. A Shimadzu LC-20AD solvent delivery module (Shimadzu) was used to produce the following gradient elution profile: 5% solvent B for 2 min, followed by 5 to 70% B in 20 min and 70 to 90% B in 2 min. The HPLC flow rate was 0.3 ml/min.
Microsomal Incubations. All incubations were performed at 37°C in a water bath. Stock solutions of the test compounds were prepared in methanol. The final concentration of methanol in the incubation was 0.2% (v/v). Pooled HLMs and the human cDNA-expressed P450 isozymes were carefully thawed on ice prior to the experiment. Trazodone, nefazodone, m-CPP, or p-CPP (10 μM) was individually mixed with HLM proteins (1 mg/ml) in 100 mM potassium phosphate buffer, pH 7.4, supplemented with 1 mM GSH. The total incubation volume was 1 ml. After 3 min of preincubation at 37°C, the incubation reactions were initiated by the addition of 1 mM NADPH. Reactions were terminated by the addition of 150 μl of trichloroacetic acid (10%) after 60 min of incubation. Incubations with the recombinant cDNA-expressed P450 isozymes were performed similarly except that liver microsomes were substituted by supersomes (100 pmol/ml). Control samples containing no NADPH or substrates were included. Samples were centrifuged at 10,000g for 15 min at 4°C to pellet the precipitated proteins, and supernatants were subjected to LC/MS/MS analysis of GSH adducts. Each incubation was performed in triplicate. For the negative precursor ion scanning of GSH adducts, supernatants were concentrated by solid-phase extraction as described below, prior to LC/MS/MS analyses.
P450 Inhibition by Chemical Inhibitors. The effect of specific inhibitors of individual P450 enzymes on the formation of reactive metabolites was examined using pooled human liver microsomes. Incubation mixtures consisted of trazodone (10 μM) or m-CPP (10 μM), individual chemical inhibitors, GSH (1 mM), and HLMs (1 mg/ml). The P450-specific inhibitors α-naphthoflavone (1 μM), sulfaphenazole (5 μM), tranylcypromine (15 μM), quinidine (2 μM), and ketoconazole (1 μM) were used to investigate the involvement of CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4, respectively. Incubations containing trazodone or m-CPP were started with the addition of 1 mM NADPH, and reactions were terminated by trichloroacetic acid. Controls containing no chemical inhibitors were included. Each incubation was performed in duplicate. The effectiveness of individual P450 inhibitors was also evaluated using P450 marker substrates 50 μM phenacetin (CYP1A2), 150 μM tolbutamide (CYP2C9), 100 μM (S)-mephenytoin (CYP2C19), 10 μM dextromethorphan (CYP2D6), and 100 μM testosterone (CYP3A4) in HLMs. Individual marker substrates were preincubated for 5 min at 37°C in the presence and absence of P450-specific inhibitors. Reactions were started with the addition of 1 mM NADPH, and terminated after 20 min. Formation of metabolites from individual P450 marker substrate was analyzed by LC/MS/MS as previously described (Walsky and Obach, 2004) with minor modifications. A comparison was made relative to the controls without inhibitors, and P450 activity was expressed as the percentage of control activity.
For concentration-dependent inhibition, the specific CYP2D6 inhibitor quinidine was used to further access the role of CYP2D6 for the reactive metabolite formation from incubations of m-CPP in human liver microsomes. Under similar incubation conditions described above, quinidine at various concentrations (0, 0.1, 0.3, 0.7, 1.2, 2, 4, or 6 μM) was added to the incubation mixture containing m-CPP (50 μM).
Solid-Phase Extraction. Samples resulting from incubations were desalted and concentrated by solid-phase extraction, prior to the negative precursor ion scan MS/MS analyses. Solid-phase extraction was performed using Oasis solid-phase extraction cartridges packed with 60 mg of sorbent C18 (Waters, Milford, MA). Cartridges were first washed with 2 ml of methanol and then conditioned with 2 ml of water. Supernatants resulting from centrifugation were loaded onto the cartridges, and cartridges were washed with 2 ml of water and then eluted with 2 ml of methanol. The methanol fractions were dried by nitrogen gas and reconstituted with 100 μl of a water-methanol (70:30) mixture. Aliquots (20 μl) of the reconstituted solutions were subjected to LC/MS/MS analysis.
LC/MS/MS Analysis. For complete profiling of reactive metabolites, samples were first subjected to chromatographic separations with a Shimadzu HPLC system coupled with an Agilent Eclipse XDB-Phenyl C18 column (3.0 × 150 mm, 3.5 μm; Agilent Technologies). The HPLC mobile phase A was 10 mM ammonium acetate in water with 0.1% formic acid, and mobile phase B was acetonitrile with 0.1% formic acid. A Shimadzu LC-20AD solvent delivery module (Shimadzu) was used to produce the following gradient elution profile: 5% solvent B for 2 min, followed by 5 to 70% B in 20 min and 70 to 90% B in 2 min. The HPLC flow rate was 0.3 ml/min. At 24 min, the column was flushed with 90% acetonitrile for 3 min before re-equilibration at initial conditions. LC/MS/MS analyses were performed on 20-μl aliquots of cleaned samples. A positive peak detected in the negative precursor ion scan over the range m/z 270 to 1000 was used to trigger the acquisition of a collision-induced dissociation (CID) MS/MS spectrum to further elucidate the structure of the GSH adduct.
For relative comparison of GSH adduct levels, the mass spectrometer was operated in the multiple reaction monitoring mode. Multiple reaction monitoring transitions were simultaneously monitored for detecting M1, m/z 693→564 and 693→420; for M2, m/z 695→566 and 695→422; and for M4, m/z 518→389 and 518→243. Data were analyzed using Analyst 4.1 version software (Applied Biosystems).
Molecular Modeling. The structure of m-CPP was drawn, and the geometry was optimized using the neglect of diatomic differential overlap semiempirical method of PM3 by steepest decent (Pople and Segal, 1965; Stewart, 2004) with ArgusLab 4.0 (Planaria Software, Seattle, WA). This was aligned and edited into the active site of the X-ray crystal structure of CYP2D6 (Protein Data Bank code 2F9Q) (Rowland et al., 2006). The structure was manipulated using WebLab Viewer Lite 4.0 (Accelrys, San Diego, CA) and ArgusLab 4.0, so that it was orientated in a geometrically reasonable position with no significant Van der Waals overlap with the protein, providing us a working model of the m-CPP-bound CYP2D6.
Results
Characterization of GSH Adducts of Trazodone. For the LC/MS/MS analysis of GSH adducts, samples generated from incubations with human liver microsomes were desalted and concentrated by solid-phase extractions, and resulting samples were subjected to the PI-EPI experiments. MS detection was carried out using the negative precursor ion scanning of m/z 272, corresponding to deprotonated γ-glutamyl-dehydroalanyl-glycine originating from the glutathionyl moiety (Dieckhaus et al., 2005). MS/MS spectra were acquired in positive ion mode using information-dependent data acquisition (Hopfgartner et al., 2003). As shown in Fig. 1A, a total of seven major components were detected by the negative precursor ion scanning of m/z 272, and they were arbitrarily designated as M1 (10.9 min), M2 (12.7 min), M3 (6.0 min), M4 (8.0 min), M5 (8.5 min), M6 (9.3 min), and M7 (9.8 min), respectively. None of these peaks was detected when either trazodone or NADPH was absent from the incubations. These data suggested that GSH adducts were formed from reactive metabolites of trazodone via oxidative metabolism.
Structures of these detected components were simultaneously identified based on MS/MS spectra in the positive ion mode. The PI-directed positive MS/MS spectrum of the most abundant adduct, M1, showed an [M + H]+ ion at m/z 693, suggesting that this component was one of the two GSH adducts previously identified (Kalgutkar et al., 2005a). This was supported by the MS/MS spectrum of [M + H]+ ion at m/z 693 that showed product ions at m/z 564, 547, 489, 444, 420, 283, 176, and 148. This component was subsequently identified as the corresponding quinone imine-sulfydryl adduct (M1, Scheme 1). The MS/MS spectrum of M2 showed an [M + H]+ ion of m/z 695, with product ions at m/z 566, 548, 422, 404, 372, 352, 237, and 208. This component was subsequently detected as the corresponding epoxide-sulfydryl adduct previously identified (Kalgutkar et al., 2005a).
There are five other components, M3 to M7, detected in the microsomal incubations of trazodone (Fig. 1A). Among these five components, M4 was the most abundant peak. The deprotonated molecular ion of component M4 was m/z 516 in the negative ion mode (Fig. 2A). A chlorine isotope peak was observed at m/z 518 (∼35% of the [M - H]- ion), which indicated that formation of a new GSH adduct that contained a chlorine atom. Under the positive ion mode, the MS/MS spectrum of [M + H]+ ion at m/z 518 provided characteristic product ions at m/z 443 and 389, resulting from neutral losses of glycine (75 Da) and pyroglutamate (129 Da), respectively (Fig. 2B). This confirmed that M4 was a new GSH adduct formed in the incubation of trazodone. The molecular ion [M + H]+ at m/z 518 was consistent with the addition of one molecule of glutathione to a monohydroxylated m-CPP metabolite of trazodone after P450-mediated N-dealkylation. Double cleavage at the piperazine ring with neutral losses of glycine and pyroglutamate formed the product ion at m/z 269. A loss of NH3 from the product ion at m/z 389 resulted in the fragment ion at m/z 372. The occurrence of the product ion at m/z 243 was consistent with the presence of an aromatic thioether motif in this GSH adduct (Baillie and Davis, 1993). A proposed structure for M4, which is consistent with the chlorine isotope cluster and the CID cleavage, is shown in Fig. 2B. The parent ion of component M5 was also m/z 516 in negative ion mode with a chlorine isotope pattern. The MS/MS analysis showed that components M4 and M5 had essentially identical spectra (Fig. 2B), suggesting that they are likely positional isomers.
The parent ion of component M3 was m/z 482 in negative ion mode with no chlorine isotope peak (Fig. 3A). Under the positive ion mode, the MS/MS spectrum of [M + H]+ ion at m/z 484 provided characteristic product ions at m/z 409 and 355, resulting from neutral losses of glycine and pyroglutamate, respectively (Fig. 3B). The molecular ion [M + H]+ of m/z 484 had a mass difference of 34 Da from that of M4 or M5, suggesting loss of a chlorine atom from m-CPP. Similar to the CID patterns in the MS/MS spectra of M4 and M5, the MS/MS spectrum of M3 provided the product ion at m/z 235, presumably resulting from double cleavage at the piperazine ring together with neutral losses of glycine and pyroglutamate. The occurrence of the product ion at m/z 209 was also consistent with the presence of an aromatic thioether motif in this GSH adduct (Baillie and Davis, 1993). Taken together, these data suggest that M3 is a GSH adduct of a deschloro-monohydroxylated m-CPP. A proposed structure for M3, which is consistent with the CID cleavage and loss of chlorine from m-CPP, is shown in Fig. 3B. Components M3, M4, and M5 were also identified in the incubations of nefazodone, which contains the same m-CPP moiety (data not shown).
Another component displaying no chlorine isotope cluster in MS detection was M6 (Fig. 4A). The [M + H]+ ion of m/z 659 was 34 Da less than that of M1, suggesting loss of a chlorine atom from trazodone. The MS/MS spectrum of [M + H]+ ion at m/z 659 afforded the diagnostic product ion at m/z 530, resulting from neutral loss of pyroglutamate (Fig. 4B). The presence of product ions at m/z 148 and 176 suggested that the N-propyl-triazolopyridinone moiety is unaltered. The product ions at m/z 249, 386, 410, 455, and 530 all had a mass shift of 34 Da from the corresponding product ions of M1 at m/z 283, 420, 444, 489, and 564, respectively. These data clearly suggested that the aromatic chlorine atom is lost in this GSH adduct, which is consistent with the absence of a chlorine isotope cluster. The occurrence of the product ion at m/z 386 also suggested that M6 contains an aromatic thioether motif. A proposed structure of the deschloro-monohydroxylated trazodone for M6 is shown in Fig. 4A.
Component M7 had a molecular ion [M + H]+ at m/z 689, which was 4 Da less than that of M1 (Fig. 5B). The deprotonated molecular ion of component M7 was m/z 687 in the negative ion mode (Fig. 5A). A chlorine isotope peak was observed at m/z 689 (∼35% of the [M - H]- ion), indicating retention of the chlorine atom. The MS/MS spectrum of M7 afforded the characteristic product ion at m/z 560, resulting from neutral loss of the pyroglutamate moiety (Fig. 5B). The presence of product ions at m/z 148 and 176 suggested that the N-propyl-triazolopyridinone moiety was unchanged. The product ions at m/z 279, 416, 440, 485, and 560 all had a mass decrease of 4 Da from the corresponding product ions of M1 at m/z 283, 420, 444, 489, and 564, respectively (Fig. 5B). These data suggested that the piperazine ring of trazodone may have undergone a P450-mediated dehydrogenation reaction. Such dehydrogenation reactions are often seen in drug metabolism (Ortiz de Montellano, 1989; Obach, 2001). The occurrence of the product ion at m/z 416 also suggested that M7 contains an aromatic thioether motif (Baillie and Davis, 1993). A proposed structure of M7, which is consistent with the CID cleavage, is shown in Fig. 5B.
Bioactivation of m-CPP. Characterization of GSH adducts formed from incubations of trazodone showed that three components, namely M3, M4, and M5, were generated from the metabolite m-CPP, which was presumably released from trazodone after N-dealkylation. To investigate the mechanisms of m-CPP bioactivation and further confirm the identities of m-CPP-derived GSH adducts generated from incubations of trazodone, m-CPP and a regioisomer, p-CPP, were incubated with human liver microsomes or recombinant CYP2D6. The regioisomer p-CPP was used to investigate whether a quinone imine was involved in the m-CPP bioactivation by blocking 4′-hydroxylation of m-CPP. As shown in Fig. 1B, three components, M3, M4, and M5, were detected in the incubation of m-CPP, and those components had the same HPLC retention times as the corresponding components from the incubation of trazodone (Fig. 1A). Also, the MS/MS spectra of these components were essentially identical as those of the corresponding GSH adducts from the trazodone incubation with HLMs or recombinant CYP2D6 (Figs. 1C, 2, and 3). These data confirmed the structural identities of m-CPP-derived GSH adducts formed in the incubations of trazodone. Under the same incubation conditions, the regioisomer p-CPP showed no bioactivation in human liver microsomes (Fig. 1D). The blockage of bioactivation of p-CPP was also confirmed by NL scanning of m/z 129 and 75, respectively (data not shown).
GSH Adduct Formation with Recombinant P450s. To investigate the roles of individual human P450 isozymes in the bioactivation of m-CPP and trazodone, the formation of GSH adducts M1, M2, and M4 was examined in incubations of m-CPP and trazodone with insect cell-expressed recombinant P450s. As shown in Fig. 6A, at the same enzyme concentration (100 pmol/ml), CYP2D6 was the predominant enzyme for the formation of M4 in the incubations of m-CPP. CYP3A4 also catalyzed M4 formation, but the level of M4 was less than 2% of that formed by CYP2D6. Only trace amounts or no M4 were detected in incubations with other P450 enzymes including CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2E1, and CYP3A5.
In the incubations of trazodone, formation of M1, M2, and M4 was examined. Consistent with observations by others (Kalgutkar et al., 2005a), it was found that formation of M1 and M2 was primarily mediated by CYP3A4 and to a lesser extent by CYP3A5 (Fig. 6B). CYP2D6 also catalyzed formation of M1 and M2, but both levels were less than 10% of those formed by CYP3A4. No single P450 isozyme was capable of catalyzing formation of M4 from the incubations of trazodone (Fig. 6B). This result agrees well with a previous report that the N-dealkylation of trazodone to form m-CPP was primarily mediated by CYP3A4 (Rotzinger et al., 1998a). In marked contrast, the formation of M4 was dramatically increased in the presence of both CYP2D6 and CYP3A4 (Fig. 6B). These data clearly suggested that although bioactivation and N-dealkylation of trazodone are mediated by CYP3A4/CYP3A5, bioactivation of the metabolite m-CPP is specifically mediated by CYP2D6.
Chemical Inhibition of Trazodone and m-CPP Bioactivation. The inhibitory effects of P450 isozyme-specific inhibitors on the formation of M1, M2, and M4 were examined using pooled human liver microsomes. Inhibitory activity was confirmed using P450 marker substrates (Walsky and Obach, 2004). In the incubations of trazodone, formation of M1 and M2 was greatly inhibited by the CYP3A4/3A5-selective inhibitor ketoconazole. Ketoconazole also strongly inhibited the formation of M4 (>85%) compared with the control value. This can be explained by the inhibition of CYP3A4-dependent N-dealkylation of trazodone (Rotzinger et al., 1998a). Quinidine, a specific CYP2D6 inhibitor, strongly inhibited M4 formation by 83% but did not inhibit the formation of M1 and M2 (Table 1).
In the incubations of m-CPP, formation of M4 was only inhibited by quinidine, which is consistent with the predominant role of CYP2D6 for M4 formation in the incubations with recombinant P450s. In both incubations of trazodone and m-CPP, the inhibitory effects on the formation of M1, M2, and M4 were minimal (<10%) for other P450-specific inhibitors including α-naphthoflavone (CYP1A2), sulfaphenazole (CYP2C9), and tranylcypromine (CYP2C19) (Table 1). It is noteworthy to point out that ketoconazole did not inhibit the formation of M4 from m-CPP (<10% inhibition).
Concentration-Dependent Inhibition with Quinidine. The formation of M4 was further examined using the specific CYP2D6 inhibitor quinidine in human liver microsomal incubations of m-CPP. As shown in Fig. 7, the CYP2D6 inhibitor quinidine resulted in a concentration-dependent inhibition of M4 formation from incubations of m-CPP. These results further support that CYP2D6 is the P450 enzyme involved in the formation of M4 from m-CPP.
Structural Modeling of the CYP2D6-m-CPP Complex. Although a crystal structure of the CYP2D6-substrate complex has not been reported yet, the recently published CYP2D6 structure (Rowland et al., 2006) provides an excellent template for modeling CYP2D6-m-CPP complex. It is well known that substrates of CYP2D6 typically contain a basic nitrogen atom and a planar aromatic ring, which is the structural feature of m-CPP.
Figure 8 depicts the binding mode of m-CPP in the CYP2D6 active site. The terminal nitrogen atom of the piperazine ring of m-CPP formed an ionic hydrogen bond with the negatively charged carboxylate group of Glu216, which lies on the underside of the F-helix. This is in agreement with mutagenesis studies in which Glu216 was identified as a key determinant in the binding of basic substrates (Paine et al., 2003). The chlorophenyl ring is proximate to the heme group, consistent with the fact that the oxidation of m-CPP by CYP2D6 occurs in this part of the molecule. Most noticeably, the binding orientation of m-CPP is precisely controlled by the hydrophobic π-π interaction between the chlorophenyl ring and Phe120 positioned on the B′-C loop and concomitantly by the hydrogen bond with Glu216 on the top (Fig. 8). The distance between C-4′ and the heme iron atom is 4.6 Å, suggesting that the active high-valent iron-oxo attacks the C-4′ during the oxidation reaction. From this CYP2D6-m-CPP binding mode, it was expected that docking of trazodone would encounter significant steric hindrance from residues on the F-G helices loop, situated on the top of the active site cavity. This result agrees well with the observation that even sharing a common structural feature of CYP2D6 substrates, trazodone is a poor substrate of this P450 isoform.
Discussion
In the present study, several new GSH adducts (M3–M7) were detected in the incubations of trazodone with human liver microsomes using LC/MS/MS. The results constitute the first report on bioactivation of m-CPP, a major circulating metabolite for several antidepressant drugs including trazodone, nefazodone, and etoperidone. It was found that formation of GSH adducts M3, M4, and M5 from m-CPP was mediated specifically by CYP2D6, in contrast to the CYP3A4-catalyzed bioactivation of trazodone and nefazodone (Kalgutkar et al., 2005a,b). In addition, two novel deschloro-GSH adducts, M3 and M6, derived from m-CPP and trazodone were tentatively identified by tandem mass spectrometry. These findings are important to fully understand the bioactivation pathways of trazodone and potential links to the mechanism of toxicity.
Direct evidence of bioactivation of m-CPP comes from incubations of m-CPP in human liver microsomes and recombinant P450 enzymes. Bioactivation of m-CPP is of particular interest because not only is it a major circulating metabolite of several antidepressants, but also it is a pharmacologically active serotonin receptor 5-HT2C agonist. Very recently, m-CPP was found used as an ecstasy-like substance, becoming a potential target for drug abuse (Bossong et al., 2005; Lecompte et al., 2006). The same set of m-CPP-derived GSH adducts was generated in incubations of trazodone and m-CPP, respectively (Fig. 1), suggesting that M3, M4, and M5 were formed via bioactivation of m-CPP. A two-step oxidation mechanism has previously been proposed for the bioactivation of the 3-chlorophenylpiperazine ring of trazodone; direct oxidation at the C-4′ position resulting in 4′-hydroxytrazodone, followed by further oxidation to form a quinone imine (Kalgutkar et al., 2005a). Sharing the identical 3-chlorophenylpiperazine ring structure, M4 and M5 are likely formed by the same bioactivation pathways (Scheme 2). Upon generation of m-CPP after CYP3A4-mediated N-dealkylation of trazodone followed by a two-step oxidation pathway, an m-CPP quinone imine was trapped by GSH to form M4 and M5.
Unlike M4 and M5, M3 was identified as a GSH adduct of m-CPP containing no chlorine atom. Recently, a similar deschloro GSH adduct of diclofenac was identified by LC/MS/MS and NMR and proposed to be derived from an ipso substitution of chlorine by GSH from a quinone imine intermediate (Yu et al., 2005). Because the proposed ipso substitution pathway is a two-step oxidation pathway and requires a initial oxidation on the C-4′ position, we further investigated the metabolic mechanisms using a regioisomer of m-CPP, p-CPP, to determine whether 4′-hydroxylation is required for formation of M3. There was no M3 detected in the incubations of p-CPP with human liver microsomes and recombinant P450 enzymes, suggesting formation of M3 requires the oxidation at the C-4′ position. Moreover, a total blockage of all three m-CPP-derived GSH adducts M3 to M5 in incubations of p-CPP suggested that they are likely formed via a common reactive quinone imine intermediate by two-electron oxidations, after the initial 4′-hydroxylation on the chlorophenyl ring (Scheme 2). Other evidence to support this bioactivation pathway is that 4′-hydroxy-m-CPP is the major metabolite in incubations of m-CPP (data not shown). Given these observations, we speculated that M3 detected in this study is 4′-hydroxy-3′-(glutathione-S-yl)-deschloro-m-CPP formed via a common quinone imine that is shared by M4 and M5, followed by the ipso substitution by GSH (Scheme 2). Similarly, the structure of M6 is proposed to be 4′-hydroxy-3′-(glutathione-S-yl)-deschloro-trazodone. Both proposed structures would have expected fragmentation patterns consistent with the CID MS/MS spectra (Figs. 3B and 4B, respectively). An alternate pathway for deschloro adduct formation is P450-mediated epoxidation between the C-3′ and C-4′ or the C-3′ and C-2′ positions on the chlorophenyl ring. Loss of chlorine would result from an attack of GSH on the epoxide. This mechanism has been proposed for the formation of the deschloro GSH adduct of diclofenac in human liver microsomal incubations (Yan et al., 2005). Although the current study cannot rule out this bioactivation pathway, one would expect that such potential epoxides could be formed in incubations of the regioisomer of m-CPP, p-CPP, for example, an epoxide between the C-4′ and C-3′ positions. The formed epoxides can readily be trapped by GSH to afford corresponding conjugates. However, both PI and NL scans failed to detect any GSH adducts in the incubations of p-CPP.
Our structural modeling of the CYP2D6-m-CPP complex revealed a well defined binding mode of m-CPP, with its orientation precisely controlled by a hydrophobic π-π interaction with Phe120 at the bottom and an ionic hydrogen bond with Glu216 from the top (Fig. 8). The binding orientation of m-CPP in the CYP2D6 active site clearly suggested that the C-4′ is the most susceptible site for an attack from the active high-valent iron-oxo, which is consistent with the fact that 4′-hydroxy-m-CPP is the major metabolite in incubations of m-CPP with both human liver microsomes and recombinant P450s. The abundance of 4′-hydroxy-m-CPP makes the ipso substitution pathway a more favorable mechanism for formation of M3 and M6. It is highly desirable to further confirm the absolute structures of these newly detected GSH adducts using synthetic standards. However, efforts to synthesize M3 and M6 have currently been hampered by the lack of a feasible chemistry strategy. Attempts were made to isolate M3 and M6 directly from incubations of trazodone with HLMs using preparative LC but failed to obtain enough of the adducts in the desired purity for NMR analysis, due to the low abundance of GSH adducts formed in the incubations. Regardless, formation of M3 and M6 represents unique reactive metabolites formed by trazodone.
Apart from CYP3A-mediated M1 and M2 formation, the formation of M4 from m-CPP was found to be mediated specifically by CYP2D6. This conclusion is supported by the following observations: 1) recombinant CYP2D6 catalyzed-formation of M4 in incubations of m-CPP; 2) formation of M4 was strongly inhibited by quinidine, a specific CYP2D6 inhibitor, in both incubations of m-CPP and trazodone; and 3) concentration-dependent inhibition of formation of M4 by quinidine. CYP3A4 also catalyzed M4 formation, but the conversion rate was much lower than that of CYP2D6. Therefore, CYP3A4 is unlikely to play a significant role in formation of M4 from m-CPP. This is supported by the observation that no single P450 isozyme was able to catalyze formation of M4 from the incubations of trazodone. Lack of M4 formation by CYP3A4 in incubations of trazodone can be explained by lower binding affinity of the formed m-CPP compared with trazodone, whereas the reason that CYP2D6 did not form M4 is that the N-dealkylation of trazodone to form m-CPP was primarily mediated by CYP3A4. This conclusion is further supported by the formation of M4 when trazodone was incubated with a mixture of CYP2D6 and CYP3A4 enzymes.
In conclusion, several novel reactive intermediates were detected in incubations of trazodone and m-CPP with human liver microsomes. In contrast to CYP3A-mediated metabolism and bioactivation of trazodone, formation of GSH adducts of m-CPP was found to be specifically mediated by CYP2D6, suggesting a possible relevance of CYP2D6 polymorphism and/or drug interactions to m-CPP toxicokinetics (Staack et al., 2007). Further studies are currently underway to study the balance of reactive metabolite formation from m-CPP and CYP2D6 phenotypes. It is our hypothesis that formation of GSH adducts from the 3-chlorophenylpiperazine ring moiety is mediated by a common quinone imine species. These findings are of significance in understanding biochemical mechanisms of idiosyncratic toxicity of several m-CPP-containing antidepressant drugs.
Acknowledgments
We thank Sid Nelson and Griff Humphreys for insightful suggestions and comments during preparation of the manuscript.
Footnotes
-
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
-
doi:10.1124/dmd.107.019471.
-
ABBREVIATIONS: 5-HT, 5-hydroxytryptamine; P450, cytochrome P450; m-CPP, 1-(3′-chlorophenyl)piperazine; GSH, glutathione; p-CPP, 1-(4′-chlorophenyl)piperazine; LC/MS/MS, liquid chromatography-tandem mass spectrometry; HPLC, high-performance liquid chromatography; PI, precursor ion; EPI, enhanced product ion; NL, neutral loss; HLM, human liver microsome; CID, collision-induced dissociation.
-
↵1 Current affiliation: Department of Drug Metabolism and Pharmacokinetics, Roche Palo Alto, Palo Alto, CA.
- Received October 28, 2007.
- Accepted January 28, 2008.
- The American Society for Pharmacology and Experimental Therapeutics