Anal. Chem. 2007, 79, 6831-6839
Generation and Identification of Reactive Metabolites by Electrochemistry and Immobilized Enzymes Coupled On-Line to Liquid Chromatography/Mass Spectrometry Wiebke Lohmann and Uwe Karst*
Westfa¨lische Wilhelms-Universita¨t Mu¨nster, Institut fu¨r Anorganische und Analytische Chemie, Corrensstrasse 30, 48149 Mu¨nster, Germany
The detection of reactive metabolites using conventional in vivo and in vitro techniques is hampered because the intermediately formed reactive species are prone to covalent binding to cellular macromolecules. Therefore, the application of improved methods is required. The online coupling of an electrochemical reactor and horseradish peroxidase immobilized on magnetic microparticles with liquid chromatography/mass spectrometry (EC/LC/ MS or HRP/LC/MS) allows the direct detection of reactive metabolites of the model compounds amodiaquine, amsacrine, and mitoxantrone, which are all known for readily binding to cellular macromolecules after metabolization by cytochrome P450. EC/LC/MS and HRP/LC/MS experiments were compared to rat liver microsome incubations and proved to be valuable complementary methods since reactive quinone, quinone imine, and quinone diimine species could be detected directly and not only after trapping with glutathione. Furthermore, N-dealkylation and N-oxidation of amodiaquine were successfully simulated by electrochemical oxidation reactions, as well as the formation of an aldehyde. Therefore, EC/LC/MS and HRP/LC/MS are promising tools for the identification of both reactive and stable metabolites in drug development. Metabolism is generally regarded as a process contributing to the detoxification of xenobiotics. During this process, xenobiotics such as drugs are converted into more polar compounds that are easily eliminated from the body. In many cases, the resulting metabolites are less toxic than the corresponding parent drug. However, some drugs undergo metabolic reactions that lead to the formation of reactive species.1,2 These reactive intermediates may modify cell proteins and DNA by covalent binding to the macromolecules and thus cause drug-induced toxicity and cell damage.3 Reactive metabolites can roughly be divided into electrophiles and free radicals.4 Most reactive metabolites are electrophiles that react with nucleophiles such as nitrogen- or * To whom correspondence should be addressed. E-mail:
[email protected]. Fax: +49-251-83 36013. (1) Kalgutkar, A. S.; Dalvie, D. K.; O’Donnell, J. P.; Taylor, T. J.; Sahakian, D. C. Curr. Drug Metab. 2002, 3, 379-424. (2) Evans, D. C.; Watt, A. P.; Nicoll-Griffith, D. A.; Baillie, T. A. Chem. Res. Toxicol. 2004, 17, 3-16. 10.1021/ac071100r CCC: $37.00 Published on Web 08/08/2007
© 2007 American Chemical Society
sulfur-containing compounds. Typical soft electrophiles are reactive intermediates with quinone, quinone imine, or quinone diimine structures that emerge from the dehydrogenation of drugs containing dihydroxybenzene, aminophenol, or phenylenediamine moieties by enzymes involved in the oxidative metabolism such as cytochrome P450s (CYP, Figure 1). In a healthy human liver, there are large amounts of glutathione (GSH) present. GSH is a natural trapping agent that may bind to soft electrophiles via the thiol group of the cysteine residue and thus detoxify reactive intermediates.5 However, in case of depletion of GSH stores, toxic effects may be observed. Drugs that extensively form reactive intermediates are, for example, amodiaquine (AQ), amsacrine (AMSA), and mitoxantrone (MX) (Figure 2) which all are known for toxic side effects. AQ is a 4-aminoquinoline antimalarial agent that is used alternatively to chloroquin because of spreading chloroquin-resistant parasites.6 Besides N-deethylation and oxidative deamination,7 AQ is mainly excreted in rat bile as a 5′-thioether conjugate with GSH or cysteine,8 an observation that indicates the intermediate formation of a reactive amodiaquine quinone imine (AQQI) or semiquinone imine (AQSQI).9 The formation of AQQI and associated agranulocytosis and hepatotoxicity10 finally led to its withdrawal from prophylactic use. AMSA and MX are both intercalating antitumor agents that are used in the treatment of leukemia and other cancer types. AMSA is a 9-anilinoacridine agent that is metabolized by CYP to a reactive quinone diimine (AQDI).11 AQDI is either hydrolyzed to a quinone imine or detoxified by glutathione conjugation via the 5′- or 6′-position of (3) Zhou, S.; Chan, E.; Duan, W.; Huang, M.; Chen, Y. Z. Drug Metab. Rev. 2005, 27, 41-213. (4) Uetrecht, J. In Drug Metabolizing Enzymes: Cytochrome P450 and Other Enzymes in Drug Discovery and Development; Lee, J. S., Obach, R. S., Fisher, M. B., Eds.; Marcel Dekker: New York, 2002; pp 87-146. (5) Rinaldi, R.; Eliasson, E.; Swedmark, S.; Morgenstern, R. Drug Metab. Dispos. 2002, 30, 1053-1058. (6) Peters, W. Chemotherapy and Drugs Resistance in Malaria; Academic Press: London, 1987. (7) Jewell, H.; Maggs, J. L.; Harrison, A. C.; O’Neill, P. M.; Ruscoe, J. E.; Park, B. K. Xenobiotica 1995, 25, 199-217. (8) Harrison, A. C.; Kitteringham, N. R.; Clarke, J. B.; Park, B. K. Biochem. Pharmacol. 1992, 43, 1421-1430. (9) Maggs, J. L.; Tingle, M. D.; Kitteringham, N. R.; Park, B. K. Biochem. Pharmacol. 1988, 37, 303-311. (10) Neftel, K. A.; Woodtly, W.; Schmid, M. Br. Med. J. 1986, 292, 721-723. (11) Shoemaker, D. D.; Csysyk, R. L.; Gormley, P. E.; DeSouza, J. J. V.; Malspeis, L. Cancer Res. 1984, 44, 1939-1945.
Analytical Chemistry, Vol. 79, No. 17, September 1, 2007 6831
Figure 1. Oxidative metabolism of dihydroxybenzenes (X, Y ) O), aminophenols (X ) NR, Y ) O), or phenylenediamines (X ) NR, Y ) NR′): Dehydrogenation by CYP to quinones, quinone imines, or quinone diimines and subsequent binding to glutathione or cellular macromolecules.
Figure 2. Structures of the selected model compounds.
the aniline ring.12 MX is an anthraquinone species that is cytotoxic due to its CYP-mediated metabolism to a reactive quinone (MXQ).13 MXQ is interconnected by a tautomeric equilibrium with a second electrophilic species being a quinone diimine (MXQDI).14 MXQ is detoxified by glutathione conjugation, while intramolecular reaction of the reactive MXQDI with the secondary amino group of the side chain yields a cyclization product. Furthermore, oxidation of the side chain leads to carboxylic acid metabolites. The detection and investigation of reactive metabolites is an urgent issue in drug development. Since reactive species are generally short-lived and unstable, the detection is not easy. One approach to obtain information about the nature and abundance of reactive metabolites is the use of trapping agents that form stable adducts with the reactive intermediates.15 Trapping agents include GSH and other thiols for soft electrophiles; semicarbazide, methoxylamine, and potassium cyanide for hard electrophiles and 5,5-dimethylpyrrolidine-N-oxide for free radicals. Especially the use of isotopically labeled GSH and analysis of the adducts by LC/MS/MS experiments such as neutral loss scans have proven to be a valuable tool in high-throughput screening for reactive metabolites.16 However, these techniques lack the direct detection of reactive intermediates. Once a metabolite is trapped, only the adduct is characterized and the nature of the reactive species has to be derived therefrom. Thus, the development of techniques that allow the direct detection of reactive species immediately after their generation is required to avoid binding or degradation of the analytes. Ideally, such a technique includes the on-line generation of metabolites to meet these criteria. The on-line coupling of an electrochemical reactor with liquid chromatography and mass spectrometry (EC/LC/MS) has already been used widely to mimic CYP-mediated reactions and might have the (12) Robertson, I. G. C.; Palmer, B. D.; Shaw, G. J. Biol. Mass Spectrom. 1993, 22, 661-665. (13) Duthie, S. J.; Grant, M. H. Br. J. Cancer 1989, 60, 566. (14) Mewes, K.; Blanz, J.; Ehninger, G.; Gebhardt, R.; Zeller, K. P. Cancer Res. 1993, 53, 5135-5142. (15) Ma, S.; Subramanian, R. J. Mass Spectrosc. 2006, 41, 1121-1139. (16) Yan, Z.; Caldwell, G. Anal. Chem. 2004, 76, 6835-6847.
6832
Analytical Chemistry, Vol. 79, No. 17, September 1, 2007
potential to become an interesting alternative in the detection of reactive metabolites as well. First results in using electrochemistry for modeling cyctochrome P450 reactions were obtained in the late 1980s when N-dealkylation reactions were successfully mimicked.17 At the same time, the on-line coupling of an electrochemical reactor and mass spectrometry was presented and the oxidative metabolism of paracetamol including trapping of an electrochemically generated reactive metabolite with bioavailable thiols was simulated.18 The redox behavior of dopamine was investigated using off-line19,20 and on-line21 EC/MS setups. More recently, electrochemical modeling reactions were studied more systematically to investigate which metabolic phase I reactions can be simulated electrochemically and which not, with the intention to obtain insight into the potential of EC/MS for the simulation of drug metabolism.22-24 Research on this field has been summarized in various reviews.25-27 The simulation of phase II reactions, including the coupling of glutathione or other thiols to electrochemically generated phase I metabolites was achieved in recent years. For example, the metabolism of clozapine including phase I and phase II reactions was successfully mimicked28 as well as the detoxification of paracetamol29 using EC/LC/MS. All (17) Hall, L. R.; Iwamoto, R. T.; Hanzlik, R. P. J. Org. Chem. 1989, 54, 24462451. (18) Getek, T. A.; Korfmacher, W. A.; McRae, T. A.; Hinson, J. A. J. Chromatogr. 1989, 474, 245-256. (19) Zhang, F.; Dryhurst, G. J. Electroanal. Chem. 1995, 398, 117-128. (20) Xu, R.; Huang, X.; Kramer, K. J.; Hawley, M. D. Bioorg. Chem. 1996, 24, 110-126. (21) Deng, H.; Van Berkel, G. J. Electroanalysis 1999, 11, 857-865. (22) Jurva, U.; Wikstro¨m, H. V.; Bruins, A. P. Rapid Commun. Mass Spectrom. 2000, 14, 529-533. (23) Jurva, U.; Wikstro ¨m, H. V.; Weidolf, L.; Bruins, A. P.; Rapid Commun. Mass Spectrom. 2003, 17, 800-810. (24) Gamache, P.; Smith, R.; McCarthy, R.; Waraska, J.; Acworth, I. Spectroscopy 2003, 18, 14-41. (25) Volk, K. J.; Jost, R. A.; Brajther-Toth, A. Anal. Chem. 1993, 64, 21A-33A. (26) Hayen, H.; Karst, U. J. Chromatogr. A 2003, 1000, 549-565. (27) Karst, U. Angew. Chem., Int. Ed. 2004, 43, 2476-2478. (28) Van Leeuwen, S. M.; Blankert, B.; Kauffmann, J.-M.; Karst, U. Anal. Bioanal. Chem. 2005, 382, 742-750. (29) Lohmann, W.; Karst, U. Anal. Bioanal. Chem. 2006, 386, 1701-1708.
these results indicate that the use of EC/LC/MS might also be a promising approach for the detection of reactive metabolites. Besides this purely instrumental technique, the enzymatic online generation of reactive metabolites might be an interesting approach as well. For the on-line coupling of enzymatic reactions to instrumental separation and detection techniques, immobilization of the involved enzymes is required. Since the immobilization of CYP itself is difficult due to the limited stability of the enzymes, the horseradish peroxidase (HRP) system, which is known to exhibit similar oxidation characteristics,30,31 was considered to be a good alternative. Even though the reaction mechanisms of HRPand CYP-catalyzed oxidations differ slightly, HRP has proven to simulate oxidative metabolism reactions well, as was demonstrated in several applications.32,33 As the HRP-catalyzed dehydrogenation for the selected model compounds AQ and MX has been investigated previously34,35 and diimine formation of AMSA has at least been examined by another peroxidase system,36 it was decided to use immobilized HRP on-line with LC/MS to study reactive metabolites. Both on-line techniques, EC/LC/MS and HRP/LC/MS, are compared and correlated with the incubation with rat liver microsomes as an established in vitro method in terms of its value for the investigation of reactive intermediates and modeling drug metabolism. EXPERIMENTAL SECTION Chemicals. Amodiaquine dihydrochloride dihydrate, amsacrine hydrochloride, mitoxantrone dihydrochloride, horseradish peroxidase (type II), magnesium chloride hexahydrate, glycine, PBS buffer, and ammonium formate were obtained from SigmaAldrich Chemie GmbH (Steinheim, Germany). 2-(4-Morpholino)ethanesulfonic acid (MES), 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC), potassium dihydrogenphosphate, dipotassium hydrogenphosphate, formic acid, and ammonium hydroxide solution (25%) were purchased from Fluka Chemie GmbH (Buchs, Switzerland). Hydrogen peroxide solution (35%) was ordered from Acros Organics (Geel, Belgium) and NADPH from AppliChem GmbH (Darmstadt, Germany). Methanol and acetonitrile for HPLC and N-hydroxysuccinimide (NHS) were obtained from Merck KGaA (Darmstadt, Germany). All chemicals were used in the highest quality available. Water used for HPLC was purified using a Milli-Q Gradient A 10 system and filtered through a 0.22-µm Millipak 40 (Millipore, Billerica, MA). Magnetic latex microparticles (micromer-M, 3 µm, COOH functionalized) were obtained from micromod Partikeltechnologie GmbH (Rostock, Germany). Instrumentation. For the LC/MS measurements, a HPLC system from Shimadzu (Duisburg, Germany), consisting of two LC-10ADVP pumps, a DGC-14A degasser, a SIL-HTA autosampler, a CTO-10AVP column oven, and a SPD-10AVVP UV detector was (30) Hollenberg, P. F. FASEB J. 1992, 6, 686-694. (31) De Visser, S. P.; Shaik, S.; Sharma, P. K.; Kumar, D.; Thiel, W. J. Am. Chem. Soc. 2003, 125, 15779-15788. (32) Anzenbacher, P.; Niwa, T.; Tolbert, L. M.; Sirimanne, S. R.; Guengerich, F. P. Biochemistry 1996, 35, 2512-2520. (33) Dalvie, D. K.; O’Connell, T. N. Drug Metab. Dispos. 2004, 32, 49-57. (34) Naisbitt, D. J.; Williams, D. P.; O’Neill, P. M.; Maggs, J. L.; Willock, D. J.; Pirmohamed, M.; Park, B. K. Chem. Res. Toxicol. 1998, 11, 1586-1595. (35) Blanz, J.; Mewes, K.; Ehninger, G.; Proksch, B.; Waidelich, D.; Greger, B.; Zeller, K. P. Drug Metab. Dispos. 1991, 19, 871-880. (36) Kettle, A. J.; Robertson, I. G. C.; Palmer, B. D.; Anderson, R. F.; Patel, K. B.; Winterbourn, C. C. Biochem. Pharmacol. 1992, 44, 1731-1738.
coupled to an API 2000 or QTRAP mass spectrometer from Applied Biosystems (Darmstadt, Germany) or a micrOTOF mass spectrometer from Bruker Daltonics (Bremen, Germany), all equipped with an electrospray ionization (ESI) source. The software used for controlling HPLC, API 2000, and QTRAP was Analyst 1.4.1 (Applied Biosystems). For controlling the micrOTOF and data handling, micrOTOFControl 1.1 and DataAnalysis 3.3 (Bruker Daltonics) software were used. The equipment for the electrochemical generation of metabolites was obtained from ESA Biosciences Inc. (Chelmsford, MA). It consisted of a model 5020 Guard Cell and a Coulochem II potentiostat. The Guard Cell contains a glassy carbon working electrode, a Pd counter electrode, and a Pd/H2 reference electrode. All potentials mentioned in this work are based on this reference electrode. A PEEK in-line filter was placed in front of the cell inlet to protect the working electrode. For the on-line coupling of the functionalized magnetic microparticles to the LC/MS system, the particles were filled in a capillary (Radel tubing, 1/16 in. × 0.03 in.) from Upchurch Scientific (Oak Harbor, WA). They were fixed in their position by two surrounding magnets mounted in a homemade magnet holder. A PEEK in-line filter was positioned downstream of the magnet holder to avoid particles that were flushed onto the HPLC column. Immobilization of HRP on Magnetic Microparticles. A 0.5 mL sample of micromer-M suspension (surface functionalization, COOH) was transferred into an Eppendorf tube and separated with a permanent magnet. The supernatant was removed, and the particles were resuspended in 1 mL of MES buffer (pH adjusted to 6.3 with 0.5 M Na2CO3). After separation of the particles, the supernatant was removed again and the particles were suspendend in 1 mL of 0.1 M MES buffer containing 4 mg of EDC and 8 mg of NHS. After vortexing for 1 min, the suspension was incubated with continuous mixing for 2 h at room temperature. The activated particles were washed twice with 1 mL of MES buffer at a permanent magnet and resuspended in 1 mL of PBS buffer (pH 7.4) containing 200 µg of HRP. The suspension was incubated with continuous mixing for 3 h at room temperature. After separating the particles at a permanent magnet and removing the supernatant, 1 mL of PBS buffer containing 25 mM glycine was added to cover the free reaction sites on the particle surface. The suspension was incubated with continuous mixing for 30 min at room temperature. The particles were washed twice with 1 mL of PBS buffer and stored at +4 C after resuspension in 1 mL of PBS buffer. On-Line Generation of Metabolites by Electrochemical and HRP Oxidation. The setup used for on-line generation of metabolites by either electrochemical or HRP oxidation and subsequent LC/MS analysis is shown in Figure 3. For electrochemical oxidation, 50 µM analyte or 50 µM analyte and 0.5 mM GSH in ammonium formate solution (20 mM, pH adjusted to 7.4 with ammonium hydroxide solution) and acetonitrile (50/50, v/v) were conducted through the EC cell using a syringe pump model 74900 (Cole Parmer, Vernon Hills, IL) at a flow rate of 50 µL/ min. The potential at the electrochemical flow-through cell was varied between 0 and 1000 mV versus Pd/H2. The eluate from the EC cell was collected in a 10-µL loop. After switching on the EC cell, the system was allowed to equilibrate under the selected conditions for ∼5 min. After the equilibration time, the 10-port Analytical Chemistry, Vol. 79, No. 17, September 1, 2007
6833
Figure 3. Schematic setup of the systems for the on-line generation, separation, and identification of metabolites. The on-line reactor is either an EC flow-through cell connected to a potentiostat (EC/LC/MS system) or functionalized magnetic particles that are fixed with two magnets in a capillary (HRP/LC/MS system). In loading position, the reaction mixture from the EC cell or the HRP particles is collected in the 10-µL loop. After switching the valve, the loop is emptied by the HPLC pumps and the content is injected on the column.
valve was switched and the eluate from the EC cell was injected either on the HPLC column or directly into the mass spectrometer (flow injection analysis). For oxidation of the drug standards by immobilized HRP, 50 µM analyte and 0.5 mM H2O2 or 50 µM analyte, 0.5 mM GSH, and 0.5 mM M H2O2 in water were conducted through a capillary (0.03-in. i.d.) filled with the functionalized particles (path length of the reactor, ∼4 mm) using a syringe pump at a flow rate of 20 µL/min. After an equilibration time of ∼10 min, the reaction mixture was collected in a 10-µL loop and analyzed by LC/MS or flow injection analysis according to the EC conditions described above. Rat Liver Microsomal Incubations. Rat liver microsomes were kindly provided by R. Maul from the University of Hamburg (Hamburg, Germany). For incubations of the drug standards with rat liver microsomes, the incubation mixture (1 mL) contained 1.3 mg mL-1 microsomal protein (determined by the method of Bradford),37 0.45 nmol mL-1 CYP (measured according to the method of Omura and Sato),38 50 µM substrate, 1.2 mM NADPH, and 0.5 mM MgCl2 in 50 mM phosphate buffer, pH 7.4. In case of reaction in the presence of GSH, the reaction mixture additionally contained 5 mM GSH. After 5-min preincubation of microsomes, substrate, and GSH at 37 °C, the reaction was initiated by addition of MgCl2 and NADPH. Incubations were carried out for 90 min at 37 °C. The reaction was stopped by adding 600 µL of the reaction mixture to 600 µL of ice-cold acetonitrile to precipitate proteins. After 5 min, the samples were centrifuged and the supernatant was directly used for injection into the LC/ MS system. For each incubation, a negative control was carried out without NADPH. HPLC Conditions. Separation of the drug standards and their metabolites was carried out using a Zorbax Eclipse XDB-C8 (37) Bradford, M. M. Anal. Biochem. 1976, 72, 248-254. (38) Omura, T.; Sato, R. J. Biol. Chem. 1964, 239, 2370-2378.
6834 Analytical Chemistry, Vol. 79, No. 17, September 1, 2007
Table 1. Gradient Profiles Used for the Separation of the Drug Standards and Their Metabolites amodiaquine
amsacrine
mitoxantrone
time (min)
CH3CN (%)
time (min)
CH3CN (%)
time (min)
CH3CN (%)
0.01 2.00 23.00 25.00 27.00 28.00 30.00
5 5 30 80 80 5 stop
0.01 2.00 10.00 12.00 13.00 15.00
5 5 90 90 5 stop
0.01 2.00 10.00 12.00 14.00 15.00 17.00
5 5 23 90 90 5 stop
column with the following dimensions: 150 mm × 4.6 mm i.d., 5-µm particle size (Agilent, Waldbronn, Germany). The column had a temperature of 40 °C. The flow rate of the mobile phase was 1 mL/min. For all separations, eluent A of the mobile phase was a solution of 10 mmol of ammonium formate and 200 µL of formic acid in 1 L of deionized water (pH ∼4). Eluent B was acetonitrile. The injection volume was 10 µL. The gradient profiles are shown in Table 1. MS Conditions. On the API 2000 and QTRAP instruments, ESI(+)-MS was performed either in the full scan mode (m/z 2001000, integration time 1 s) or in the selected ion monitoring mode. The MS conditions are presented in Table 2. On the micrOTOF instrument, full scan spectra (m/z 100-1200) were recorded using ESI(+)-MS under the following conditions: end plate offset, -200 V; capillary, 4500 V; nebulizer gas (N2), 1.6 bar; dry gas (N2), 8.0 L min-1; dry temperature, 200 °C; capillary exit, 90.0 V; skimmer 1, 30.0 V; skimmer 2, 23.0 V; hexapole 1, 20.0 V; hexapole 2, 20.6 V; hexapole rf, 150 V; transfer time, 49.0 µs; prepulse storage, 5.0 µs; detector, -1000 V.
Table 2. MS Parameters Used for the Detection of the Drug Standards and Their Metabolites
declustering potential entrance potential focusing potential curtain gas (N2) ion spray voltage temperature nebulizer gas (N2) dry gas/heating gas (N2)
amodiaquine API 2000
amsacrine QTRAP
mitoxantrone QTRAP
20 V 10 V 400 V 30 psi 5000 V 400 °C 50 psi 70 psi
60 V 10 V 25 psi 5500 V 400 °C 50 psi 70 psi
60 V 10 V 25 psi 5500 V 400 °C 50 psi 70 psi
RESULTS AND DISCUSSION On-Line Electrochemical Generation of Metabolites. To obtain first insight into the electrochemical behavior of AQ, AMSA, and MX, an EC/MS setup was employed without chromatographic separation of the electrochemically generated metabolites. A solution containing the analyte of interest was pumped through an electrochemical flow-through cell containing a porous glassy carbon working electrode. The large surface area of the working electrode allows up to 100% turnover of the analytes, depending on flow rate, pH, nature of the analyte, and other conditions. The pH of the analyte solution was adjusted to 7.4 to simulate the physiological conditions as far as possible. However, a certain amount of acetonitrile was needed to prevent the analytes from adsorbing at the surface of the working electrode. The reaction mixture eluting from the electrochemical cell was directly injected into the mass spectrometer using an injection loop mounted on a 10-port switching valve (Figure 3). This injection technique is set up easily and allows optimizing the conditions for both electrochemical oxidation and HPLC separation independently. This is a significant advantage compared with the previously employed EC/LC/MS systems for the simulation of metabolic reactions, which used the EC cell in-line with the HPLC.28,29 Using that former setup, the electrochemical oxidation has to take place under the same conditions as the LC separation with respect to the parameters flow rate, pH, and composition of the mobile phase. In the system presented here, electrochemical oxidation and separation of the oxidation products can be carried out under completely different conditions, thus allowing us to apply the optimum conditions for both. In Figure 4, the mass spectra of AQ, AMSA, and MX without and with electrochemical oxidation at different potentials are compared. Without electrochemical oxidation, [M + H]+ peaks for the three analytes at m/z 356 (AQ, Figure 4a), 394 (AMSA, Figure 4d), and 445 (MX, Figure 4g) were observed. Additional signals in the mass spectra were a result of either fragmentation or oxidation in the electrospray source. When potentials between 300 and 400 mV versus Pd/H2 were applied at the working electrode, all drug standards underwent almost quantitative oxidation by dehydrogenation to the respective quinone imine (for AQ, m/z 354, AQQI, Figure 4b), quinone diimine (for AMSA, m/z 392, AQDI, Figure 4e), or quinone (for MX, m/z 443, MQ, Figure 4h). In the case of MX, formation of a quinone diimine (MXQDI) may be considered as well, since the formation of two distinct dehydrogenated species, a quinone and a quinone diimine, interconnected by a tautomeric equilibrium has been reported in the literature.14 Additionally, a signal was observed at m/z 441,
which indicates the generation of both quinone and quinone diimine (MXQQDI) simultaneously. The increase of the potential at the EC cell up to 600 or 700 mV resulted in further oxidation of AQ (Figure 4c) and MX (Figure 4i), whereas the quinone diimine of AMSA (Figure 4f) remained unchanged. N-Dealkylation is a process that is easily induced electrochemically.23 Hence, N-deethylation was observed upon electrochemical oxidation of AQ at a potential of 700 mV versus Pd/H2. Due to the fact that quinone imine formation already takes place at lower potentials, the monodeethylated product was detected as the respective quinone imine at m/z 326 (DESAQQI). The presence of the original deethylated aromatic structure without quinone imine formation (DESAQ) cannot be excluded, but the respective signal at m/z 328 overlaps with the 37Cl isotopic peak from m/z 326. Later investigations including HPLC separation of the electrochemically generated metabolites and high-resolution mass spectrometry showed that a small amount of DESAQ was formed besides DESAQQI. Further oxidation resulted mainly in a signal at m/z 299. The expected bis-deethylation at the tertiary amine function would result in m/z 300 or 298, depending on whether quinone imine formation occurs or not. However, instead of these m/z ratios, a signal at m/z 299 was observed. To elucidate the nature of this compound, the measurements were repeated using time-of-flight mass spectrometry for determining exact masses (data not shown). The determined exact mass of m/z 299.0578 indicates the formation of an ion with the sum formula C16H12ClN2O (calculated: m/z 299.0582, difference 1.34 ppm). These results suggest the generation of an aldehyde, since the formation of aldehyde structures is a wellknown process during electrochemical oxidation of aliphatic amines.23 Increase of the electrochemical oxidation potential up to 1000 mV resulted in a signal with m/z 315 (data not shown), tentatively identified as the respective carboxylic acid as a product of further oxidation of the aldehyde. The formation of a carboxylic acid metabolite of AQ as a product of an oxidative deamination process has been reported in the literature for the use of rat liver microsomes.7 Oxidation of MX at 600 mV versus Pd/H2 mainly led to the formation of both quinone and quinone diimine simultaneously, resulting in a signal at m/z 441 (MXQQDI, Figure 4i). Furthermore, signals at m/z 457 and 473 were observed. This suggests the introduction of one and two oxygen atoms into the molecule, shown by a mass increase of 16 and 32 Da compared with m/z 441. A further assignment of the site of oxidation, either the anthracene moiety or the side chain of MX, cannot be derived from these data and requires additional investigations that were not focus in this work. For on-line HPLC separation of the electrochemically generated metabolites, a potential of 700 mV versus Pd/H2 was selected for AQ since the variety of formed metabolites at this potential simulated the known AQ metabolism best. AMSA and MX were oxidized at 300 mV. All separations were carried out using a C8 column and a binary gradient of acetonitrile and aqueous ammonium formate solution (pH 4). Chromatograms were recorded in both total ion current (TIC) and selected ion monitoring (SIM) mode using m/z ratios that were observed in EC/MS experiments. The following figures display the combined SIM traces. Analytical Chemistry, Vol. 79, No. 17, September 1, 2007
6835
Figure 4. MS spectra of the electrochemical oxidation of the drug standards in acetonitrile/aqueous ammonium formate solution (20 mM, pH adjusted to 7.4 with ammonium hydroxide solution).
Figure 5. ESI(+)-TIC chromatograms (combined from SIM measurements) of (a) AQ and metabolized AQ using (b) on-line electrochemistry at 700 mV vs Pd/H2, (c) on-line HRP, (d) off-line microsomes, and (e) off-line microsomes in the presence of GSH. Peak assignment: 1, AQ (m/z 356); 2, DESAQ (m/z 328); 3, AQQI (m/z 354); 4, DESAQQI (m/z 326); 5, AQ aldehyde* (m/z 299); 6, AQQI N-oxide* (m/z 370); 7, AQ N-oxide* (m/z 372); 8, bis-DESAQ (m/z 300); 9, GS-AQ (m/z 661); 10, GS-DESAQ (m/z 633). Structures with * are identified tentatively.
Scans a and b in Figure 5 show the chromatograms of AQ before and after electrochemical metabolization. All generated metabolites could be well resolved and identified. The elution order is in accordance with the polarity of the suggested structures of the metabolites from the EC/MS experiments. Generally, dehydrogenated species elute later than aromatic analytes, and the deethylated compounds DESAQ (m/z 328) and DESAQQI (m/z 326) elute before the respective non-ethylated species AQ (m/z 356) and AQQI (m/z 354). The retention time of 17.9 min for the compound with m/z 299 confirms the assumption that an aldehyde might have been formed from the amino group. Although the comparably high retention time could indicate the formation of a compound with a quinone imine structure, this was excluded for two reasons: First, in contrast to all other quinone 6836 Analytical Chemistry, Vol. 79, No. 17, September 1, 2007
imine species, a respective aromatic structure with m/z 301 was absent, and second, the species with m/z 299 remained unchanged upon addition of ascorbic acid to an off-line EC mixture, whereas all other quinone imine species (AQQI and DESAQQI) underwent reduction to the respective aromatic compounds (data not shown). Hence, additional evidence for the formation of an aldehyde was obtained. Furthermore, the formation of a species with m/z 370 and a retention time of 21.7 min was observed. The assumption that one oxygen atom was introduced into the AQQI moiety was confirmed by time-of-flight mass spectrometric measurements (data not shown). However, hydroxylation could be excluded, because the compound with m/z 370 had a higher retention time than AQQI. Hence, N-oxidation of AQ is a more likely alternative. Since the focus of this work is directed to the identification of reactive metabolites, further investigations on the nature of the N-oxide were not carried out. Figure 6 resumes the electrochemical metabolism of AQ as described in more detail above. LC/ESI(+)-MS chromatograms of AMSA and its metabolites that were generated by different means of oxidation are presented in Figure 7. The electrochemical oxidation of AMSA resulted in a less complex metabolite pattern than AQ (Figure 7b): Only two species were observed, AMSA itself (m/z 394) eluting at 7.5 min and the quinone diimine AQDI (m/z 392) eluting at 8.6 min. The electrochemical metabolism pathway is summarized in Figure 8. The electrochemical oxidation of MX is presented in Figure 9. As already shown in the EC/MS experiments, MX was oxidized to a species with m/z 443 being either a quinone (MXQ) or a quinone diimine (MXQDI) and to a compound with m/z 441, which corresponds to both quinone and quinone diimine formation (MXQQDI). Chromatograms of MX before and after electrochemical oxidation are shown in Figure 9a and b. MX (m/z 445) elutes at 10.0 min, but a second peak is observed with the same m/z ratio at 9.6 min. Apparently, electrochemical oxidation results in the formation of a slightly more polar isomer of MX. However, it was not possible to identify the nature of this isomer, and further investigations, e.g., by combination with NMR spectroscopy,
Figure 6. Electrochemical metabolism pathway of AQ at 700 mV vs Pd/H2. The aromatic nitrogen was chosen as site of N-oxidation, but oxidation of one of the other nitrogen atoms is also possible.
Figure 7. ESI(+)-TIC chromatograms (combined from SIM measurements) of (a) AMSA and metabolized AMSA using (b) on-line electrochemistry at 300 mV vs Pd/H2, (c) on-line HRP, (d) off-line microsomes, and (e) off-line microsomes in the presence of GSH. Peak assignment: 1, AMSA (m/z 394); 2, AQDI (m/z 392); 3, AMSA + O (m/z 410); 4, GS-AMSA (m/z 699).
Figure 8. Electrochemical metabolism pathway of AMSA at 300 mV vs Pd/H2.
appear to be promising. For m/z 443, only one peak at 12.8 min is observed, being rather MXQ than MXQDI for two reasons: On the one hand, oxidation of hydroquinones occurs more readily compared to oxidation of p-phenylenediamines.39 On the other hand, oxidation of the phenylenediamine moiety of MX may cause cyclization by reaction of the side chain with the quinone diimine.14 (39) Weinberg, N. L.; Weinberg, H. R. Chem. Rev. 1968, 68, 449-523.
Figure 9. ESI(+)-TIC chromatograms (combined from SIM measurements) of (a) MX and metabolized MX using (b) on-line electrochemistry at 300 mV vs Pd/H2, (c) on-line HRP, (d) off-line microsomes, and (e) off-line microsomes in the presence of GSH. Peak assignment: 1, MX (m/z 445); 2, MX isomer (m/z 445); 3, MXQ* (m/z 443); 4, MXQQDI (m/z 441); 5, GS-MX (m/z 750). Structures with * are identified tentatively.
Since only one chromatographic peak was observed for m/z 443, formation of MXQDI is less likely than formation of MXQ. The twofold dehydrogenated MXQQDI with m/z 441 elutes at 12.5 min. The electrochemical metabolism pathway of MX at a potential of 300 mV versus Pd/H2 is presented in Figure 10. To obtain information about the reproducibility of the EC/ LC/MS system, the dehydrogenation of AQ resulting in AQQI at 300 mV versus Pd/H2 and further N-dealkylation resulting in DESAQQI at 700 mV were exemplarily selected for multiple experiments. Injections were repeated threefold, and average peak areas and standard deviations were calculated. For AQQI at 300 mV, the average peak area was (6.05 ( 0.28) × 107 arbitrary units, representing a relative standard deviation of 4.6%. For DESAQQI at 700 mV, the average peak area was (9.67 ( 0.65) × 106 arbitrary units, representing a relative standard deviation of 6.7%. These exemplary figures illustrate the good reproducibility of the Analytical Chemistry, Vol. 79, No. 17, September 1, 2007
6837
Figure 10. Electrochemical metabolism pathway of MX at 300 mV vs Pd/H2.
metabolite generation with the EC/LC/MS setup. Generally, relative standard deviations did not exceed 10% as long as the experimental parameters were kept constant and a regular cleaning of the EC cell by rinsing with acetonitrile was performed. In summary, EC experiments reproduce previously reported CYP-catalyzed metabolic reactions of AQ, AMSA, and MX quite well, especially dehydrogenation and N-dealkylation reactions, although a systematical comparison with microsome experiments is required for further conclusions as is shown below. On-Line HRP-Catalyzed Generation of Metabolites. For HRP-catalyzed on-line generation of metabolites, HRP was immobilized by carbodiimide coupling using N-hydroxysuccinimide and 1-ethyl-3-(3-dimethylaminopropyl)carbodiimde on magnetic microparticles (MMP) coated with carboxylic acid groups. Remaining reactive sites of the activated MMP after HRP immobilization were covered with glycine. The HRP-functionalized MMP were filled in a capillary and fixed with two magnets mounted in a magnet holder. The enzyme reactor had a path length of ∼4 mm, long enough to allow a sufficient turnover of the analytes to a detectable amount of reactive metabolites. As with the EC/LC/MS experiments, a solution containing the analyte of interest and H2O2 was pumped over the immobilized HRP and collected in a loop for injection onto the HPLC column (Figure 3). The general setup of this coupling has already been described above. LC/ESI(+)-MS chromatograms of the on-line generation of metabolites by immobilized HRP are shown in Figures 5c (AQ), 7c (AMSA), and 9c (MX). AQ was oxidized by HRP to AQQI (m/z 354) and a species with m/z 370. The latter compound was already observed upon electrochemical oxidation, being presumably a product of N-oxidation of AQ. In contrast to the electrochemical generation of metabolites, N-dealkylation was not observed. HRPcatalyzed oxidation of AMSA resulted in quinone diimine formation (AQDI, m/z 392) as already observed in EC experiments. Corresponding to the electrochemical generation of metabolites, no additional oxidation products were found in the HRP experiments. The HRP-catalyzed oxidation of MX yields analogous results to the electrochemical oxidation as well. Besides residual MX (m/z 445), MXQ (m/z 443) was observed at 12.8 min and MXQQDI (m/z 441) at 12.6 min. For this model compound, the turnover to dehydrogenated species was almost quantitative under the reaction conditions described above. In general, the statements about the reproducibility for the EC/LC/MS system apply accordingly to the HRP experiments. Standard deviations were minimized by keeping the experimental parameters constant and by regularly replacing the functionalized microparticles. Normally, the relative standard deviations were slightly higher than for the EC/LC/MS system but did not exceed 15% in multiple experiments. 6838 Analytical Chemistry, Vol. 79, No. 17, September 1, 2007
Table 3. Overview of Metabolites of Amodiaquine, Amsacrine and Mitoxantrone Generated by Electrochemistry, HRP, and CYPa
amodiaquine
amsacrine mitoxantrone
metabolite
EC
HRP
CYP
AQ AQQI DESAQ DESAQQI bisDESAQ AQ aldehyde AQ N-oxide AQQI N-oxide AMSA AQDI AMSAOH MX MXQ MXQQDI MX isomer
x x x x × x × x x x × x x x x
x x × × × × × x x x × x x x ×
x (x) x (x) x x x × x × x x x × ×
a x, metabolite was generated by the respective technique; ×, metabolite was not generated by the respective technique; (x), metabolite was generated by the respective technique, but detection was only possible after trapping with GSH.
In summary, HRP-catalyzed generation of metabolites was comparable to EC oxidation except for the fact that N-dealkylations, which are readily obtained in electrochemical experiments, were not observed under HRP catalysis. Consequently, HRP is less suited for mimicking the complete pattern of CYP-catalyzed reactions than the presented electrochemical reactor. Comparison with Rat Liver Microsome Experiments. To verify whether the on-line generated metabolites are comparable to those generated by the pharmacologically relevant CYP, AQ, AMSA, and MX were metabolized using rat liver microsomes. The incubation mixtures were analyzed by off-line LC/ESI(+)-MS. Negative controls were carried out without adding NADPH to the reaction mixture, thus preventing CYP catalysis. Table 3 compares the metabolic reactions that were observed for AQ, AMSA, and MX using the presented on-line techniques EC/LC/MS and HRP/ LC/MS as well as the conventionally employed in vitro technique using rat liver microsomes. For AQ, N-dealkylation to the already described DESAQ and further oxidation to AQ aldehyde was observed (Figure 5d). The retention times are in good accordance with the EC experiments. In contrast to the EC experiments, a twofold deethylated metabolite (bis-DESAQ, m/z 300) was observed at 8.5 min. This suggests that bis-DESAQ is indeed a metabolite of AQ, but reacts readily to AQ aldehyde under EC conditions despite the short reaction time and was thus not found in EC experiments. Furthermore, a peak with m/z 372 and a retention time of 14.8 min indicates the introduction of one oxygen atom into the AQ moiety. As already discussed for the EC
experiments, this species presumably is an N-oxide due to the higher retention time compared with AQ itself. For AMSA, metabolization by rat liver microsomes (Figure 7d) led only to the formation of a species with m/z 410 and a retention time of 7.2 min that was not observed during EC and HRP metabolism. This compound probably originates from the introduction of one oxygen atom into the AMSA moiety (AMSA + O). In contrast to AQ, where only N-oxidation was observed, this metabolite of AMSA may be a hydroxylated species since the retention time is lower than for AMSA itself. Only one metabolite was found after metabolization of MX, presumably the dehydrogenated MXQ (m/z 443) at a retention time of 12.8 min (Figure 9d). This species is the same as was already described for EC and HRP experiments. Interestingly, quinone imine and quinone diimine species for AQ and AMSA were only observed as oxidation products generated in the electrospray source. None of the reactive compounds was present in the incubation mixture that was injected onto the HPLC column. Only for MX, a small amount of the dehydrogenated MXQ was observed. The absence of dehydrogenated products might give rise to the conclusion that dehydrogenation of AQ and AMSA did not occur in the microsome experiments and dehydrogenation of MX happened only to a small extent. However, this is not true as analogous microsome experiments with addition of GSH showed. For all three compounds, the formation of one or more GSH adducts was observed. This observation indicates that dehydrogenation of the three standard compounds inevitably had happened in the course of the incubation with microsomes, since otherwise, no or at least a much smaller amount of GSH adduct would have been found as was shown in negative controls without using NADPH (data not shown). In the case of AQ (Figure 5e), reaction with GSH yielded two adducts, GS-AQ (m/z 661) resulting from trapping of AQQI (coeluting with DESAQ) and GS-DESAQ (m/z 633) from trapping of DESAQQI. AMSA (Figure 7e) and MX (Figure 9e) reacted with GSH to one adduct each, GS-AMSA (m/z 699) formed by trapping AQDI and GS-MX (m/z 750) as a result of trapping MXQ. From these observations, it was concluded that reactive species had covalently bound to cellular macromolecules in microsome experiments without GSH (according to Figure 1) and therefore could not be extracted from the reaction mixture anymore. Hence, in contrast to on-line EC and HRP oxidation, the detection of dehydrogenated reactive species of AQ, AMSA, and partly MX was only possible using an indirect approach by trapping the reactive intermediates with GSH. Additionally, trapping experiments were carried out with online EC and immobilized HRP as well. The drug standards were oxidized accordingly to the already described procedures in the presence of GSH (data not shown). For AQ and AMSA, the same GSH adducts as in the microsome experiments were observed except for GS-DESAQ in the HRP experiment, since deethylation was not catalyzed by HRP. Respective EC and HRP experiments with MX did not yield a GSH adduct. This is due to the fact that MXQ seems to be comparably stable, leading on the one hand to incomplete binding to cellular macromolecules and thus the appearance of MXQ in microsome experiments and on the other hand to no coupling with GSH in on-line metabolic experiments,
where the reaction time is rather short before the coupling partners MXQ and GSH reach the HPLC column. The previous observations show that the generation of metabolites by either on-line EC or immobilized HRP is favorable for the detection of reactive species over microsomal incubations for two reasons: First, the generation of metabolites occurs directly before LC separation and identification, meaning that semistable metabolites with a lifetime of only a few minutes or even less still can be detected. Even if the metabolite degrades during HPLC separation, its presence can be concluded from the resulting broad peak shape. Second, reactive metabolites have no possibility to bind to macromolecules when they are generated electrochemically. This is valid as well for the generation of reactive metabolites by immobilized HRP, because the time the analytes are in direct contact with the functionalized magnetic particles is comparably short (∼5 s in our experiments) so that binding of the reactive metabolites to HRP is not favored. These facts show that on-line EC and immobilized HRP are powerful supplementary tools in metabolism studies that complement the conventionally used in vitro and in vivo techniques for the detection of both reactive and stable metabolites. CONCLUSIONS In this work, the combinations of EC/LC/MS and HRP/LC/ MS were presented as new tools for the on-line generation and detection of reactive metabolites. The results from EC- and HRPinitiated metabolism were systematically compared with conventional off-line in vitro experiments using rat liver microsomes. Furthermore, the simulation of metabolic reactions by EC or immobilized HRP resulting in stable metabolites as for example dealkylated products, an aldehyde (both only in EC experiments), or N-oxides was possible. In summary, the on-line generation of metabolites in an electrochemical reactor or by immobilized HRP with subsequent LC/MS analysis has the advantage over conventionally used in vivo and in vitro techniques that not only stable metabolites but also reactive species like quinones, quinone imines, or quinone diimines can be detected and characterized directly and not only after trapping with glutathione or other agents. Although differences in the amount and the nature of the formed metabolites occur between the presented on-line techniques and the processes in the living cell, the presented on-line techniques are valuable tools that complement the established techniques in metabolism research and may be additionally employed in search for novel pharmaceuticals. ACKNOWLEDGMENT Rat liver microsomes were kindly provided by R. Maul from the University of Hamburg (Hamburg, Germany). We thank U. Jurva from AstraZeneca R&D Mo¨lndal (Mo¨lndal, Sweden) and J.-M. Kauffmann from the Free University of Brussels (Brussels, Belgium) for helpful discussions. Financial support by the Deutsche Forschungsgemeinschaft (DFG, Bonn, Germany) and the Fonds der Chemischen Industrie (Frankfurt/Main, Germany) is gratefully acknowledged.
Received for July 10, 2007.
review
May
25,
2007.
Accepted
AC071100R Analytical Chemistry, Vol. 79, No. 17, September 1, 2007
6839
