Detection and Structural Characterization of Glutathione-Trapped

Oct 6, 2007 - MS data sets were also utilized for the elimination of false positive peaks ... adducts.1r7 Neutral loss (NL) scanning analysis using a ...
0 downloads 0 Views 305KB Size
Anal. Chem. 2007, 79, 8333-8341

Detection and Structural Characterization of Glutathione-Trapped Reactive Metabolites Using Liquid Chromatography-High-Resolution Mass Spectrometry and Mass Defect Filtering Mingshe Zhu,* Li Ma, Haiying Zhang, and W. Griffith Humphreys

Department of Biotransformation, Pharmaceutical Candidate Optimization, Bristol-Myers Squibb Research and Development, Princeton, New Jersey 08543

The present study was designed to apply the mass defect filter (MDF) approach to the screening and identification of reactive metabolites using high-resolution mass spectrometry. Glutathione (GSH)-trapped reactive metabolites of acetaminophen, diclofenac, carbamazepine, clozapine, p-cresol, 4-ethylphenol, and 3-methylindole in human liver microsomes (HLM) were analyzed by HPLC coupled with Orbitrap or Fourier transform ion cyclotron resonance mass spectrometry. Through the selective removal of all ions that fall outside of the GSH adduct MDF template windows, the processed full scan MS chromatograms displayed GSH adducts as major components with no or a few interference peaks. The accurate mass LCMS data sets were also utilized for the elimination of false positive peaks, detection of stable oxidative metabolites with other MDF templates, and determination of metabolite molecular formulas. Compared to the neutral loss scan by a triple quadrupole instrument, the MDF approach was more sensitive and selective in screening for GSH-trapped reactive metabolites in HLM and rat bile and far more effective in detecting GSH adducts that do not afford the neutral loss of 129 Da as a significant fragmentation pathway. The GSH adduct screening capability of the MDF approach, together with the utility of accurate mass MS/MS information in structural elucidation, makes high-resolution LC-MS a useful tool for analyzing reactive metabolites. The detection of metabolic activation of lead compounds has become an integral part of drug metabolism research in the discovery process in order to lessen the risk posed by development of drugs that form large amounts of reactive intermediates.1-7 Various screens have been devised that allow multiple compounds to be evaluated for their propensity to form small or large molecule * To whom correspondence should be addressed. Dr. Mingshe Zhu, BristolMyers Squibb, Biotransformation, PCO, Princeton, New Jersey 08540-4000. Telephone: (609) 252-3324. E-mail: [email protected]. (1) Uetrecht, J. Drug Discovery Today 2003, 8, 832-837. (2) Ma, S.; Subramanian, R. J. Mass Spectrom. 2006, 41, 1121-1139. (3) Evans, D. C.; Watt, A. P.; Nicoll-Griffith, D. A.; Baillie, T. A. Chem. Res. Toxicol. 2004, 17, 3-16. (4) Gan, J.; Harper, T. W.; Hsueh, M. M.; Qu, Q.; Humphreys, W. G. Chem. Res. Toxicol. 2005, 18, 896-903. 10.1021/ac071119u CCC: $37.00 Published on Web 10/06/2007

© 2007 American Chemical Society

adducts.1-7 Neutral loss (NL) scanning analysis using a triple quadrupole mass spectrometer was the first LC-MS/MS method employed for screening for glutathione (GSH)-trapped in vitro reactive metabolites.8,9 In the NL analysis, GSH adducts formed in human liver microsome (HLM) incubations in the presence of GSH are monitored for a neutral loss of 129 Da, which is a common fragmentation pathway of GSH adducts. The NL experiment can be easily operated following a generic acquisition protocol and is especially useful for detecting certain uncommon GSH adducts. However, the NL scanning method has relatively poor sensitivity and selectivity, often resulting in the appearance of many intense false positive peaks in the NL ion chromatograms. Furthermore, the fragmentation patterns of GSH adducts upon collision-induced dissociation (CID) are dependent on the linkage structures between the drug and GSH moieties. Some classes of GSH adducts, such as benzylic thioether GSH adducts, do not afford a neutral loss of 129 Da as the primary fragmentation pathway.10 To improve the selectivity of the NL scanning, a mixture of GSH and stable-isotope labeled GSH (1:1 ratio) has been employed as a trapping agent.11-13 With this method, false positive peaks can be readily recognized based on the lack of a unique double isotope peak ratio in their NL MS/MS spectra. A similar approach using a peptide containing both cysteine and lysine and its isotopic analog as a trapping agent was developed for analyzing “soft” and “hard” reactive metabolites.14 In addition, (5) Nelson, S. D. In Biological Reactive Intermediates; Dansette, P. M., Delaforge, R. S. M., Gibson, G. G., Greim, H., Jollow, D. J., Monks, T. J., Sipes, I. G., Eds.; Kluwer Academic/Plenum Publishers: New York, 2001; Vol. VI. (6) Soglia, J. R.; Harriman, S. P.; Zhao, S.; Barberia, J.; Cole, M. J.; Boyd, J. G.; Contillo, L. G. J. Pharm. Biomed. Anal. 2004, 36, 105-116. (7) Day, S. H.; Mao, A.; White, R.; Schulz-Utermoehl, T.; Miller, R.; Beconi, M. G. J. Pharmacol. Toxicol. Methods 2005, 52, 278-285. (8) Baillie, T. A.; Davis, M. R. Biol. Mass Spectrom. 1993, 22, 319-325. (9) Chen, W. G.; Zhang, C.; Avery, M. J.; Fouda, H. G. In Biological Reactive Intermediates; Dansette, P. M., Delaforge, R. S., M., Gibson, G. G., Greim, H., Jollow, D. J., Monks, T. J., Sipes, I. G., Eds.; Kluwer Academic/Plenum Publishers: New York, 2001; Vol. VI, pp 521-524. (10) Dieckhaus, C. M.; Fernandez-Metzler, C. L.; King, R.; Krolikowski, P. H.; Baillie, T. A. Chem. Res. Toxicol. 2005, 18, 630-638. (11) Yan, Z.; Caldwell, G. W. Anal. Chem. 2004, 76, 6835-6847. (12) Yan, Z.; Maher, N.; Torres, R.; Caldwell, G. W.; Huebert, N. Rapid Commun. Mass Spectrom. 2005, 19, 3322-3330. (13) Mutlib, A.; Lam, W.; Atherton, J.; Chen, H.; Galatsis, P.; Stolle, W. Rapid Commun. Mass Spectrom. 2005, 19, 3482-3492. (14) Yan, Z.; Maher, N.; Torres, R.; Huebert, N. Anal. Chem. 2007, 79, 42064214.

Analytical Chemistry, Vol. 79, No. 21, November 1, 2007 8333

a precursor ion (PI) scanning method that monitors for a negative ion at m/z 272 in the negative electrospray ionization mode has been developed to improve the effectiveness of GSH adduct screening.10 Alternatively, a multiple reaction monitoring (MRM)based method using a hybrid triple quadrupole-linear ion trap instrument has been developed to enhance the sensitivity and selectivity of GSH adduct detection.15 The MRM experiment is carried out by following more than 100 transitions from protonated molecules of potential GSH adducts to their product ions derived from a neutral loss of either 129 or 307 Da. Recently, hybrid high resolution/accurate mass instruments, including Q-TOF, LTQ FTMS, and LTQ Orbitrap mass spectrometers, have been introduced and applied to metabolite identification.16-21 These instruments have high sensitivity, fast scan rate, and/or MSn scanning capability.22 However, unlike triple quadrupole LC-MS, these hybrid high-resolution mass spectrometers often do not have true NL, PI, and MRM scanning capabilities, which greatly limit their routine use in the detection of metabolites in complex biological matrixes, especially uncommon or unpredicted metabolites. To overcome the drawback, a mass defect filter (MDF) technique was developed to facilitate the detection of oxidative metabolites by high resolution/accurate mass LC-MS based on predicable, narrow ranges for changes in mass defects of metabolites.23,24 Compared to NL and PI scan techniques, the MDF approach has been shown to be more comprehensive in detecting both common and uncommon oxidative metabolites, including metabolites that have significantly smaller molecular weights than those of parent drugs.24 The effectiveness of MDF in analyzing oxidative metabolites in complex biological samples, including liver microsomes, bile, and plasma,18,21,23-25 has been demonstrated. Alternatively, an exact mass pseudo NL acquisition method has been developed on a Q-TOF instrument for screening GSH adducts.26 In the analysis, the neutral loss monitoring is achieved via alternating the data acquisition with low- and highcollision energy. Once a mass difference of 129.0426 Da (within a narrow mass tolerance window) between ions in the low- and high-energy full scan mass spectra is detected, product ion acquisition of the potential precursor ion is triggered. The main objective of the present study was to apply the highresolution LC-MS based MDF technique to the screening and identification of GSH-trapped reactive metabolites. The sensitivity (15) Zheng, J.; Ma, L.; Xin, B.; Olah, T.; Humphreys, W. G.; Zhu, M. Chem. Res. Toxicol. 2007, 20, 757-766. (16) Peterman, S. M.; Duczak, N.; Kalgutkar, A. S.; Lame, M. E.; Soglia, J. R., Jr. J. Am. Soc. Mass Spectrom. 2006, 17, 363-375. (17) Makarov, A.; Denisov, E.; Kholomeev, A.; Balschun, W.; Lange, O.; Strupat, K.; Horning, S. Anal. Chem. 2006, 78, 2113-2120. (18) Bateman, K. P.; Castro-Perez, J.; Wrona, M.; Shockcor, J. P.; Yu, K.; Oballa, R.; Nicoll-Griffith, D. A. Rapid Commun. Mass Spectrom. 2007, 21, 14851496. (19) Sanders, M.; Shipkova, P. A.; Zhang, H.; Warrack, B. M. Curr. Drug Metab. 2006, 7, 547-555. (20) Liu, D. Q.; Hop, C. E. J. Pharm. Biomed. Anal. 2005, 37, 1-18. (21) Mortishire-Smith, R. J.; O’Connor, D.; Castro-Perez, J. M.; Kirby, J. Rapid Commun. Mass Spectrom. 2005, 19, 2659-2670. (22) Ma, S.; Chowdhury, S. K.; Alton, K. B. Curr. Drug Metab. 2006, 7, 503523. (23) Zhang, H.; Zhang, D.; Ray, K. J. Mass Spectrom. 2003, 38, 1110-1112. (24) Zhu, M.; Ma, L.; Zhang, D.; Ray, K.; Zhao, W.; Humphreys, W. G.; Skiles, G.; Sanders, M.; Zhang, H. Drug Metab. Dispos. 2006, 34, 1722-1733. (25) Castro-Perez, J. M. Drug Discovery Today 2007, 12, 249-256. (26) Castro-Perez, J.; Plumb, R.; Liang, L.; Yang, E. Rapid Commun. Mass Spectrom. 2005, 19, 798-804.

8334 Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

Figure 1. Structures of test compounds.

and selectivity of the MDF approach was evaluated by analyzing reactive metabolites of seven model compounds in HLM incubations. Results were compared to those from neutral loss scanning on a triple quadrupole instrument. The accurate mass LC-MS data sets acquired for the selection of GSH adducts were also employed for the determination of GSH adduct formulas, elimination of false positive peaks, and identification of stable oxidative metabolites. Additionally, the matrix effect of rat bile on the detection of GSH adducts by the NL scanning and MDF approach was investigated. EXPERIMENTAL SECTION Materials. Pooled HLM were purchased from BD Biosciences (Woburn, MA). GSH, NADPH, acetaminophen, carbamazepine, diclofenac, clozapine, p-cresol, 4-ethylphenol, and 3-methyindole were purchased from Sigma-Aldrich (St. Louis, MO). Microsomal Incubation. Test compounds (10 µM) (Figure 1) were incubated separately with HLM (1.0 mg/mL), GSH (1 mM), and NADPH (1 mM) in potassium phosphate buffer (100 mM, pH 7.4) for 30 min. The total incubation volume was 2 mL. The incubation reactions were initiated by the addition of NADPH after a 3 min preincubation and were stopped by the addition of 300 µL of trichloroacetic acid (10%). After centrifugation (13 000 rpm for 10 min), the supernatants were loaded onto solid-phase extraction cartridges (Oasis extraction cartridges, Waters Corp., Milford, MA). The cartridges were washed with 1 mL of water and then eluted with 2 mL of methanol. The methanol fractions were dried and reconstituted with 200 µL of a water-acetonitrile mixture (v/v, 95/5). Aliquots (20 µL) of the reconstituted solutions were injected into the LC-MS system. Triple Quadrupole LC-MS. The HPLC system consisted of Shimadzu LC10ADvp bipumps (Columbia, MD), an HTC PAL autosampler (Leap Technologies, Cary, NC), and a Zorbax SB C-18 HPLC column (2.1 mm × 150 mm, 5 µm, Agilent Technologies, Palo Alto, CA). HPLC mobile phase A was formic acid in water (0.1%) and mobile phase B was acetonitrile (100%). For analysis of GSH-trapped reactive metabolites in HLM, the HPLC gradient started at 2% B for 3 min, increased linearly to 90% B over 27 min, held at 90% B over 5 min, then returned to the initial conditions over 2 min. The HPLC flow rate was 0.25 mL/min. For analysis of rat bile, the HPLC gradient started at 2% B for 3 min, increased linearly to 15% B over 17 min, increased linearly

Table 1. Mass Defect Shifts of GSH-Trapped Reactive Metabolites That Are Formed via Common P450-Mediated Bioactivation Reactionsa mass shift of the drug moiety

GSH adduct composition

mass defect shift of GSH adductb

functional group

-34

P + GSH - HCl

+0.0389

-32 -30 -18

P + GSH - S - 2 H P + GSH - 2 N - 4 H P + GSH + O - HCl

+0.0279 -0.0218 +0.0339

-14 -12 -2 0

P + GSH - CH2 - 2 H P + GSH - C - 2 H P + GSH + O - HF P + GSH - 2 H

-0.0157 0 +0.0043 0

P + GSH + O - NH - 2 H P + GSH

-0.016 +0.0157

+14

P + GSH + O - 4 H

-0.0208

aromatic ether methylenedioxy aromatic fluoride substituted phenol derivative aromatic amine R,β-unsaturated carbonyl benzylamine

+16 +18 +28 +32

P + GSH + O - 2 H P + GSH + O P + GSH + CO - 2 H P + GSH + 2 O - 2 H

-0.0051 +0.0106 -0.0051 -0.0102

phenol furan benzylamine benzene

+1 +2

β-Cl to a nitrogen or sulfur thiourea hydrazine aromatic chlorine

common metabolic activation reaction loss of Cl to form aziridinium or episulfonium R-NH-C(dS)-NH-R′ f R-NH-C(SG)dN-R′ R-NH-NH2 f R-SG formation of quinone imine or epoxide followed by GSH attack and loss of HCl demethylation followed by oxidation to quinone formation of quinone epoxidation followed by GSH attack and loss of HF formation of quinone methide formation of quinone via oxidative deamination CH2dCH-C(dO)-R f GS-CH2-CH-C(dO)-R formation of hydroxylamine followed by oxidation to nitrile oxide formation of quinone formation of epoxide followed by GSH attack formation of isocyanate formation of quinone

a Most listed common bioactivation reactions, examples of these reactions and associated references were adapted from a previously published document.15 b The mass defect shifts of GSH adducts are defined as the difference in the mass defects between a detected GSH adduct and the GSH-adduct MDF template (MH+ of the drug + GSH - 2 H).

to 90% B over 15 min, held at 90% B for 10 min, and then returned to the initial conditions over 0.1 min. The HPLC flow rate was 0.25 mL/min. The HPLC system was interfaced to a Finnigan TSQ quantum triple quadrupole mass spectrometer (Thermo Finnigan, San Jose, CA). For analysis of GSH adducts in HLM, NL scans for losses of 129 and 307 Da in a mass range of 350-700 Da and full scan MS analysis in a range from 350 to 800 Da were performed in the positive ion electrospray mode. For analysis of GSH adducts in bile, NL scans for loss of 129 Da in a mass range of 85-850 Da were conducted. The spray voltage was set at 4.0 kV, and the capillary temperature was set at 350 °C. The NL experiment was operated using argon as the collision gas with a collision energy at 26 eV. The product ion spectral scan of 4-ethyphenol GSH adducts over a mass range of 50-450 Da was conducted under similar MS conditions with a collision energy of 26 eV. LC-Orbitrap MS. The same HPLC system and method as described above were employed. An LTQ Orbitrap mass spectrometer with positive ion electrospray ionization (Thermo Fisher Scientific, Bremen, Germany) was operated using an accurate mass full scan MS acquisition with a mass range of 100-700 Da and a data-dependent MS2 and MS3 acquisition with the Orbitrap. All experimental data were acquired using external calibration 1 or 2 day(s) prior to the analysis. The resolving power used was 30 000 for the full scan event and 7500 for the MS2 and MS3 scan events. LC-FTMS. A Surveyor HPLC system was interfaced to an LTQ FT mass spectrometer (FTMS) (Thermo Finnigan, San Jose, CA). The same HPLC methods as described above were used. The mass spectrometer was operated in the positive ion electrospray mode at a capillary temperature of 320 °C. The accurate mass full scan MS data were acquired in a mass range of 85-850

Da with a resolving power of 50 000. The MS/MS data were acquired using nitrogen as the collision gas with a collision energy set at 25%. Mass Defect Filter Processing Methods. The high-resolution LC-MS data acquired with the LTQ Orbitrap instrument were processed using MetWorks 1.0.1 (Thermo Fisher Scientific, Bremen, Germany). The MDF processed data files were analyzed using QualBrowser (Thermo Fisher Scientific, Bremen, Germany). The high-resolution LC-MS data acquired with the LTQ FTMS were converted to the NetCDF format using the File Converter tool in Xcalibur then processed using a mass defect filter software developed at Bristol-Myers Squibb.23 After the processing, the output file was converted to its native data file format using the original file conversion utility to facilitate comparison to the original data. For the detection of GSH adducts, the mass defect filter window was set to (0.040 Da around the mass defect of a GSH adduct template (MH+ of a drug + GSH - 2 H) over a mass range of (50 Da around the mass of the filter template. For the detection of stable oxidative metabolites, the mass defect filter windows were set to (0.040 Da around the mass defect of a compound and its core substructures over a mass range of (50 Da around the masses of filter templates applied.24 Determination of Matrix Effects of Rat Bile. A mixture of rat bile (2 µL) and a reconstituted solution (20 µL) from the diclofenac HLM incubation sample was analyzed using the MDF approach (with LTQ FTMS) or neutral loss scanning. In addition, various amounts of rat bile (0, 1, 5, and 10 µL) were spiked into a reconstituted solution containing GSH adducts of acetaminophen and carbamazepine. GSH adducts in these samples were analyzed by neutral loss scanning and full-scan analysis on the triple quadrupole instrument. The HPLC method and mass spectrometAnalytical Chemistry, Vol. 79, No. 21, November 1, 2007

8335

ric data acquisition and processing methods were described above. RESULTS AND DISCUSSION GSH Adduct Screening Strategy. The strategy outlined for the screening and structural characterization of GSH adducts using MDF included three steps: (1) acquisition of a full scan LC-MS data set using high-resolution mass spectrometry, (2) selection of GSH adduct ions by processing the LC-MS data set with GSH adduct MDF templates, and (3) structural characterization of the GSH adducts based on m/z values and empirical formulas of GSH adducts and the interpretation of product ion spectra acquired by either data-dependent analysis or in a second LC-MS run. Table 1 lists common bioactivation reactions leading to GSH adduct formation predicted to occur in HLM. Although these GSH adducts represent a variety of changes in the structures of the drug moieties, the differences in the mass defects between the GSH adducts and the GSH adduct filter template (MH+ of the drug + GSH - 2 H) are no greater than 0.040 Da. Therefore, mass defect filters from -0.040 to +0.040 Da around the mass defect of the filter template are able to select for all the GSH adducts listed in Table 1. In some cases, reactive metabolites are formed by cleaving a parent drug into two significantly smaller molecules via N- and O-dealkylation or amide hydrolysis. GSH adducts of these reactive metabolites can be detected using core substructure-GSH adduct filter templates (a protonated core structure of the drug + GSH - 2 H). A similar process of using core-structure filter templates was shown to be effective in detecting small oxidative metabolites formed by dealkylation reactions of nefazodone.24 Method Validation. The effectiveness of the MDF approach for the detection of GSH-trapped reactive metabolites was examined by analyzing reactive metabolites of seven model compounds (10 µM) (Figure 1) in HLM incubations supplemented with GSH (1 mM). The total ion chromatogram (TIC) of the full scan MS data set for the clozapine HLM incubation sample displayed one GSH adduct (OM2) and several intense peaks corresponding to endogenous components and stable oxidative metabolites (Figure 2A). After MDF processing with the drug-GSH filter template, nondrug-related components were almost completely filtered out, resulting in the detection of an additional two minor GSH adducts, OM1 and OM3, at the TIC level (Figure 2B). All these GSH adducts of clozapine had similar mass defects to that of the drugGSH filter template (e0.0051 Da) (Table 2). The MDF approach not only enabled the detection of minor GSH adducts, such as OM1 and OM3, but also simplified the full scan spectra of these adducts by completely removing the interference ions that fall outside of the MDF windows (Figure 1 of the Supporting Information). On the basis of the accurate mass LC-MS data (Figure 2B and Table 2) and previously reported information of clozapine-GSH adducts in HLM,11,27,28 OM2 and OM1 were tentatively assigned as C6-glutathinylclozapine and one of its isomers, respectively. OM3 was tentatively identified as a GSH adduct of clozapine-N-oxide since the N-oxide metabolite was (27) Maggs, J. L.; Williams, D.; Pirmohamed, M.; Park, B. K. J. Pharmacol. Exp. Ther. 1995, 275, 1463-1475. (28) Williams, D. P.; Pirmohamed, M.; Naisbitt, D. J.; Maggs, J. L.; Park, B. K. J. Pharmacol. Exp. Ther. 1997, 283, 1375-1382.

8336 Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

Figure 2. MDF analysis of GSH adducts and oxidative metabolites of clozapine (10 µM) in an HLM incubation supplemented with GSH. (A) TIC of accurate mass full scan MS analysis of the HLM incubation. (B) MDF processed ion chromatogram obtained using the drug-GSH adduct filter template (MH+ of the drug + GSH - 2 H). (C) MDP processed ion chromatogram obtained using the drug filter template (MH+ of the drug). (D) Ion chromatogram of the NL scanning of 129 Da. The accurate mass LC-MS data were acquired by LC-FTMS, and the NL scanning data were acquired by triple quadrupole LC-MS.

found to be the only monooxidation product present in the HLM incubation (Figure 2C). The usefulness of the MDF approach for detecting GSH adducts was further demonstrated by analyzing reactive metabolites of other model compounds, including 3-methylindole,29-31 4-ethtylphenol,32-34 p-cresol,34,35 acetaminophen,32 diclofenac,36,37 and carbamazepine,38,39 that are known to undergo bioactivation in HLM. The MDF processed ion chromatograms of these (29) Yan, Z.; Easterwood, L. M.; Maher, N.; Torres, R.; Huebert, N.; Yost, G. S. Chem. Res. Toxicol. 2007, 20, 140-148. (30) Yost, G. S. Adv. Exp. Med. Biol. 2001, 500, 53-62. (31) Regal, K. A.; Laws, G. M.; Yuan, C.; Yost, G. S.; Skiles, G. L. Chem. Res. Toxicol. 2001, 14, 1014-1024. (32) Chen, W.; Koenigs, L. L.; Thompson, S. J.; Peter, R. M.; Rettie, A. E.; Trager, W. F.; Nelson, S. D. Chem. Res. Toxicol. 1998, 11, 295-301. (33) Thompson, D. C.; Perera, K.; London, R. Chem. Res. Toxicol. 1995, 8, 5560. (34) Yan, Z.; Zhong, H. M.; Maher, N.; Torres, R.; Leo, G. C.; Caldwell, G. W.; Huebert, N. Drug Metab. Dispos. 2005, 33, 1867-1876. (35) Thompson, D. C.; Perera, K.; Fisher, R.; Brendel, K. Toxicol. Appl. Pharmacol. 1994, 125, 51-58. (36) Tang, W.; Abbott, F. S. J. Mass Spectrom. 1996, 31, 926-936. (37) Tang, W.; Abbott, F. S. Chem. Res. Toxicol. 1996, 9, 517-526. (38) Pearce, R. E.; Uetrecht, J. P.; Leeder, J. S. Drug Metab. Dispos. 2005, 33, 1819-1826.

Table 2. GSH Adducts Detected in HLM Incubations Using the MDF Approach

test compound (MH+)

GSH adducta

acetaminophen (152.0711) carbamazepine (237.1028)

AM2 CM2 CM1 CM3 CM4 CM5 CM6 CM7 OM1 OM2 OM3 DM2 DM1 PM1 PM2 PM3 EM1 EM2 TM1 TM2 TM3 TM4 TM5 TM6 TM7 TM8 TM9

clozapine (327.1376) diclofenac (296.0245) p-cresol (109.0653) 4-ethylphenol (123.0810) 3-methylindole (132.0813)

theoretical MH+ of GSH adduct (molecular formula) 457.1393 (C18H25N4O8S) 576.1765 (C25H30N5O9S) 560.1815 (C25H30N5O8S) 558.1659 (C25H28N5O8S) 632.2058 (C28H35ClN7O6S) 648.2007 (C28H35ClN7O7S) 617.0876 (C24H27Cl2N4O9S) 583.1265 (C24H28ClN4O9S) 414.1335 (C17H24N3O7S) 430.1284 (C17H24N3O8S) 428.1491 (C18H26N3O7S) 444.1440 (C18H26N3O8S) 469.1393 (C19H25N4O8S) 453.1444 (C19H25N4O7S)

437.1495 (C19H25N4O6S)

MH+

determined of GSH adduct

error (ppm)

457.1378 576.1761 560.1815 560.1818 560.1816 558.1656 558.1655 558.1653 632.2065 632.2051 648.2010 617.0867 583.1251 414.1321 430.1271 430.1272 428.1478 444.1429 469.1384 469.1386 453.1434 453.1435 453.1434 453.1437 453.1436 437.1485 437.1484

-3.3 -0.7 0 +0.5 +0.2 -0.5 -0.7 -1.1 1.1 -1.3 0.5 -1.5 -2.6 -3.4 -3.0 -2.7 -3.4 -2.8 -1.9 -1.5 -2.3 -2.0 -2.2 -1.6 -1.7 -2.4 -2.6

composition of GSH adduct

∆ of GSH adduct mass defectb

P + GSH - 2 H P + GSH + 2 O P + GSH + O

0 +0.0055 +0.0106

P + GSH + O - 2 H

-0.0051

P + GSH - 2 H

0

P + GSH + O - 2 H P + GSH + O - 2 H P + GSH + O - HCl P + GSH - 2 H P + GSH + O - 2 H

-0.0051 -0.0051 +0.0339 0 -0.0051

P + GSH - 2 H P + GSH + O - 2 H P + GSH + 2 O - 2 H

0 -0.0051 -0.0102

P + GSH + O - 2 H

-0.0051

P + GSH - 2 H

0

a The GSH adducts of acetaminophen, carbamazepine, and diclofenac were previously structurally characterized in HLM incubations.15 b ∆ of GSH adduct mass defect is defined as the difference between the mass defect of a detected GSH adduct and that of the GSH-adduct MDF template (MH+ of the drug + GSH - 2 H).

incubation samples displayed GSH adducts as major components with a few false positives. The GSH adducts of acetaminophen (Supporting Information, Figure 2B), diclofenac (Supporting Information, Figure 2C), and carbamazepine (Supporting Information, Figures 3 and 4) were also detected in HLM incubations by MRM analysis in a previous study,15 in which the structures and formation pathways of these GSH adducts were discussed. Two GSH adducts of p-cresol, PM1 and PM2 (Table 2), were detected in the HLM incubation by LC-Orbitrap MS (Supporting Information, Figure 2A). PM1 and PM2 were tentatively identified as a GSH adduct of a quinone methide intermediate and a GSH adduct of 4-methyl-[1,2]-benzoquinone, respectively. Similar GSH adducts were previously observed in HLM incubations.34,35 Multiple GSH adducts of 3-methylindole, including two major adducts TM8 and TM9 and seven minor adducts, were detected by the LC-Orbitrap using the MDF approach (Figure 3 and Table 2). TM8 and TM9 had the same protonated molecules at m/z 437.1484 ( 0.0001, corresponding to the direct addition of GSH to 3-methylindole (drug + GSH - 2 H). TM3-TM7 had the same protonated molecules at m/z 453.1435 ( 0.0002 (Figure 3C), corresponding to the addition of an oxygen atom and a GSH molecule to the drug (drug + GSH + O - 2 H). These GSH adducts were previously observed in HLM.29 Minor GSH adducts TM1 and TM2 had protonated molecules approximately at m/z 469.196, corresponding to the incorporation of two oxygen atoms into the drug (39) Pearce, R. E.; Vakkalagadda, G. R.; Leeder, J. S. Drug Metab. Dispos. 2002, 30, 1170-1179.

moiety (drug + GSH + 2 O - 2 H) (Table 2). The formation of these two GSH adducts in HLM has not been reported in the literature. The Utility of Accurate Mass Spectra. In addition to facilitating GSH adduct screening with MDF, accurate mass LC-MS data were employed for empirical formula determination to aid in the confirmation of GSH adducts and elimination of false positives. For example, a minor peak (retention time 9.1 min, Figure 3A) in the MDF-processed profile of the 3-methylindole incubation sample was determined to be a false positive on the basis of its empirical molecular formula. Furthermore, accurate MS/MS and MS3 spectral data acquired using a data-dependent method with an LTQ Orbitrap instrument were utilized to facilitate spectral interpretation. This, in come cases, can be crucial for structural elucidation. For example, TM8 and TM9 had protonated molecules at m/z 437.1487 and 437.1483 (C19H25N4O6S), respectively (Table 2), corresponding to the direct addition of GSH to 3-methylindole (drug + GSH - 2 H). Accurate MS/MS spectra of TM8 and TM9 showed a single major product ion at m/z 308.09039 and 308.10584, respectively (Figure 4A and Figure 4C). If solely based on their nominal mass (m/z 308), one would think that both product ions were formed from the neutral loss of 129 Da. However, on the basis of the accurate mass MS/MS data, the elemental composition of the product ion of TM8 was determined to be C10H18N3O6S, corresponding to the protonated GSH molecule, while the element composition of the product ion of TM9 was determined to be C14H18N3O3S, corresponding to a product Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

8337

Figure 3. MDF analysis of GSH adducts of 3-methylindole (10 µM) in an HLM incubation supplemented with GSH. (A) MDF processed ion chromatogram obtained using the drug-GSH adduct template (MH+ of the drug + GSH - 2 H). (B) Extracted ion chromatograms (EIC) of the MDF processed profile for m/z 437. (C) EIC of the MDF processed profile for m/z 453. (D) EIC of the MDF processed profile for m/z 469. The accurate mass LC-MS data were acquired by LC-Orbitrap MS.

ion derived from a neutral loss of 129 Da. The accurate mass MS3 spectra of TM8 and TM9 further confirmed the structures of these two product ions (Figure 4B and Figure 4D). Interestingly, two product ions shown in the MS3 spectra of both TM8 and TM9 had the same nominal masses at m/z 162 and 233, while their structures assigned based on the accurate mass measurement were totally different (Figure 4B and Figure 4D). TM8 was tentatively identified as 3-(glutathion-S-yl-methyl)-indole with the GSH moiety on the C-3 methylene group (Figure 4A). Like most benzylic thioether GSH adducts,10 TM8 fragmented primarily via the cleavage of the thioether bond (the loss of 307 Da). TM9 was tentatively assigned as 3-methyl-2-(glutathione-S-yl)-indole with the GSH moiety on the C-2 position of the indole group. Detection of Stable Oxidative Metabolites. One of the advantages of the MDF approach is that accurate mass LC-MS data sets acquired for GSH adduct screening were also applicable to the detection of stable oxidative metabolites and other classes of metabolites. Thus, a full scan LC-MS data acquisition with a generic method in a single LC-MS run can provide all the data needed for detecting a variety of metabolites with different MDF templates. For example, in addition to three GSH adducts OM1, OM2, and OM3 detected in the clozapine incubation sample using the GSH adduct filter template (Figure 2B), two major oxidative metabolites were revealed by MDF processing with the drug filter 8338 Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

template (Figure 2C). These metabolites were tentatively assigned as desmethylclozapine (m/z 313.1212) and clozapine-N-oxide (m/z 342.1315), which is consistent with previous observations.27,40 The identification of stable oxidative metabolites in the reactive metabolite trapping experiment can facilitate the elucidation of chemical mechanisms of bioactivation since these oxidative metabolites are often mechanistically and structurally related to the reactive metabolite(s). Comparison with NL Scanning Method. Neutral loss scan with a triple quadrupole instrument has been frequently employed for GSH adduct screening. One of the disadvantages of the NL scanning method is its incapability to detect low levels of GSH adducts.10,15 In the current study the NL scanning of 129 and 307 Da failed to detect several minor GSH adducts of test compounds, which were detected by the MDF approach. For example, a minor clozapine-GSH adduct (OM1) (Figure 2D) and six low abundance 3-methylindole-GSH adducts (TM1-TM3 and TM4-TM6) (data not shown) were not detected by the NL scanning. These results indicate that the MDF approach was more sensitive in the detection of in vitro GSH adducts than the NL scanning analysis. The effectiveness of the NL scan for GSH adducts is dependent on MS/MS fragmentation patterns of GSH adducts that can be (40) Fang, J.; Coutts, R. T.; McKenna, K. F.; Baker, G. B. Naunyn-Schmiedeberg’s Arch. Pharmacol. 1998, 358, 592-599.

Figure 4. Accurate mass MS/MS and MS3 spectra and proposed structures of 3-methylindole adducts TM8 and TM9. (A) MS/MS spectrum of TM8 (m/z 437.1485). (B) MS3 spectrum of TM8, which was recorded by performing CID of the precursor ion at m/z 308.09039. (C) MS/MS spectrum of TM9 (m/z 437.1484). (D) MS3 spectrum of TM9, which was recorded by performing CID of the precursor ion at m/z 308.10584. The MS/MS and MS3 spectra were acquired by LC-Orbitrap with a data-dependent acquisition method.

Figure 5. Product ion spectra and proposed structures of 4-ethylphenol-GSH adducts (A) EM1 and (B) EM2. The spectra were acquired by a triple quadrupole instrument.

significantly different among different classes of GSH adducts.2 As shown in the 4-ethylphenol example, the fragmentation patterns of GSH adducts of 4-ethylphenol, EM1 and EM2, were completely different (Figure 5). Like other benzylic thioether-GSH adducts,10 EM1 showed an insignificant neutral loss of 129 Da (Figure 5A). As a result, the NL scanning of 129 Da was not able to find EM1 (Figure 6B). In contrast, the MDF approach detected EM1 (Figure

6A) since the detection mechanism of the MDF approach is associated with the similarity of the mass defects of GSH adducts to that of the GSH adduct filter template rather than predicted MS/MS fragmentation patterns of GSH adducts. The selectivity of the MDF approach for in vitro GSH adducts, in general, was significantly better than that of the NL scanning method, which was evident in the clozapine (Figure 2) and 4-ethylphenol (Figure Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

8339

Figure 6. MDF and NL scanning analyses of GSH adducts of 4-ethylphenol in an HLM incubation supplemented with GSH. (A) MDF processed ion chromatogram obtained using the drug-GSH adduct filter template. (B) Ion chromatogram of the NL scanning of 129 Da. (C) Ion chromatogram of NL scanning of 307 Da. The accurate mass LC-MS data were acquired by LC-Orbitrap, and the NL scanning data were acquired by a triple quadrupole LC-MS instrument.

5) examples. The use of stable isotope GSH as a trapping agent has greatly improved the selectivity of the NL scanning of 129 Da; however, this strategy would not be effective in the detection of GSH adducts that afford the NL of 307 Da instead of 129 Da as a primary fragmentation pathway, such as EM1. Since the MDF approach utilizes full scan MS data for the metabolite search, the stable isotope GSH trapping experiment can be readily adapted to further enhance its selectivity regardless of fragmentation patterns of GSH adducts. To compare the effectiveness of the MDF and NL scanning methods for the detection of GSH adducts in complex biological

matrixes, diclofenac adducts in rat bile were analyzed by both LC-MS methods (Figure 7). The TIC recorded from the highresolution LC-MS analysis showed many intense peaks and high background noise produced by endogenous components (Figure 7A). The MDF processing with the drug-GSH filter template almost completely removed these interference ions, resulting in a very clear ion chromatogram that displayed the diclofenac adducts, DM1 and DM2, as predominant peaks (Figure 7B). In contrast, the NL scanning analysis of the same bile sample failed to detect the diclofenac adducts (Figure 7C). The superior selectivity of MDF for drug-GSH adducts in a bile sample is also illustrated in the MDF-processed full scan mass spectrum of DM2 (Figure 7E) that only displayed the protonated molecule at m/z 617.0904 and its isotopic ion at m/z 619.0877. In contrast, the unprocessed mass spectrum of DM2 exhibited several significant interference ions that made the identification of the GSH adduct ions difficult (Figure 7D). Matrix Effect of Rat Bile. To further investigate the matrix effect of rat bile on GSH adduct detection by NL scans and fullscan MS analysis, a mixture of the acetaminophen and carbamazepine incubation samples with or without the addition of rat bile (up to 10 µL) were analyzed. The GSH adducts AM2 and CM2 were shown to be predominant peaks in the NL chromatogram in the absence of rat bile (Figure 8A). The background noise levels and intensities of interference peaks were elevated with an increase of the amounts of bile added to the injected samples (Figure 8). Some of these background noise and interference peaks may be related to endogenous GSH conjugates present in rat bile, which can be detected by the NL scanning for 129 Da. As a result, the signal-to-noise ratios of the AM2 and CM2 peaks decreased. CM2 and AM2 were not detected by neutral loss scanning when 10 µL of rat bile was added into an injection sample (data not shown). In contrast, the absolute intensities of the AM2 and CM2 peaks detected by the NL scanning of 129 Da and the full scan MS analysis remained unchanged with the addition of rat bile up to 10 µL, suggesting that the ionization of the GSH adducts was not affected by rat bile. These observations indicate that the increase in the background noise levels and intensities

Figure 7. Analysis of diclofenac-GSH adducts in rat bile by the MDF approach and the NL scanning method. (A) TIC of unprocessed accurate mass full scan LC-MS data. (B) MDF processed ion chromatogram obtained using the drug-GSH adduct filter template. (C) Ion chromatogram of the NL scanning of 129 Da. (D) Unprocessed accurate mass full scan mass spectrum of DM2. (E) MDF-processed accurate mass full scan mass spectrum of DM2. The accurate mass LC-MS data were acquired by LC-FTMS, and the NL scanning experiment was carried out using a triple quadrupole LC-MS instrument. 8340

Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

Figure 8. Analysis of GSH adducts of acetaminophen and carbamazepine with or without the addition of rat bile by the NL scanning of 129 Da: (A) no bile, (B) 1 µL rat bile, and (C) 5 µL rat bile. Rat bile was added to a mixture of the HLM incubation samples of acetaminophen and carbamazepine prior to injection.

of interference peaks, rather than ionization suppression, played a key role in the sensitivity reduction of the NL scanning analysis by rat bile. As demonstrated in the diclofenac example (Figure 7B), the matrix effect of rat bile on the MDF analysis was relatively insignificant because the accurate mass full scan MS analysisbased MDF technique has better selectivity and can discriminate drug-GSH adducts from endogenous GSH adducts present in rat bile. CONCLUSIONS This study describes the application of the mass defect filtering technique to the detection and structural characterization of GSHtrapped reactive metabolites by high-resolution LC-MS. Unlike triple quadrupole LC-MS-based NL scan, PI scan, and MRM methods, the detection of GSH adducts by the MDF approach is

dependent on the similarity of mass defects of GSH adducts to those of GSH adduct filter templates regardless of fragmentation patterns of the GSH adducts. Results from this study demonstrated that the MDF approach has several advantages over the neutral loss scanning method: (1) The MDF approach, in general, was more sensitive in detecting in vitro GSH adducts and far more effective in the analysis of the GSH adducts that do not afford a neutral loss of 129 Da as a significant fragmentation pathway. (2) The MDF approach was more selective in detecting drug-GSH adducts in rat bile, in which there are large amounts of endogenous GSH conjugates that undergo the neutral loss fragmentation pathways similar to those of drug-GSH adducts. (3) The accurate mass spectral data acquired for the GSH adduct detection were also useful in the determination of molecular formulas of GSH adducts and elimination of false positives. (4) In addition to GSHtrapped reactive metabolites, stable oxidative metabolites and other classes of metabolites can be detected via processing the same LC-MS data sets with different MDF templates. The exceptional metabolite detection capability of the MDF technique, together with the utility of accurate mass MS/MS information in the elucidation of metabolite structures, makes high-resolution mass spectrometry a useful LC-MS platform for the comprehensive screening and identification of GSH adducts in biological matrixes. ACKNOWLEDGMENT We thank Dr. Mark Sanders for his support during this research endeavor and Dr. Qian Ruan for technical assistant in the use of the Orbitrap LC-MS instrument. SUPPORTING INFORMATION AVAILABLE MDF processed ion chromatograms of the p-cresol, acetaminophen, and diclofenac incubation samples and MDF processed full scan MS spectra of selected GSH adducts and/or stable oxidative metabolites clozapine and carbamazepine. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review May 29, 2007. Accepted August 14, 2007. AC071119U

Analytical Chemistry, Vol. 79, No. 21, November 1, 2007

8341