690
Chem. Res. Toxicol. 2009, 22, 690–698
In Vitro Screening of 50 Highly Prescribed Drugs for Thiol Adduct FormationsComparison of Potential for Drug-Induced Toxicity and Extent of Adduct Formation Jinping Gan,* Qian Ruan, Bing He, Mingshe Zhu, Wen C. Shyu, and W. Griffith Humphreys Pharmaceutical Candidate Optimization, Bristol-Myers Squibb Research and DeVelopment, Princeton, New Jersey 08543 ReceiVed October 6, 2008
Reactive metabolite formation has been associated with drug-induced liver, skin, and hematopoietic toxicity of many drugs that has resulted in serious clinical toxicity, leading to clinical development failure, black box warnings, or, in some cases, withdrawal from the market. In vitro and in vivo screening for reactive metabolite formation has been proposed and widely adopted in the pharmaceutical industry with the aim of minimizing the property and thus the risk of drug-induced toxicity (DIT). One of the most common screening methods is in vitro thiol trapping of reactive metabolites. Although it is well-documented that many hepatotoxins form thiol adducts, there is no literature describing the adduct formation potential of safer drugs that are widely used. The objective of this study was to quantitatively assess the thiol adduct formation potential of 50 drugs (10 associated with DIT and 40 not associated) and document apparent differences in adduct formation between toxic and safer drugs. Dansyl glutathione was used as a trapping agent to aid the quantitation of adducts following in vitro incubation of drugs with human liver microsomes in the presence and absence of NADPH. Metabolic turnover of these drugs was also monitored by LC/UV. Overall, 15 out of the 50 drugs screened formed detectable levels of thiol adducts. There were general trends toward more positive findings in the DIT group vs the non-DIT group. These trends became more marked when the relative amount of thiol adducts was taken into account and improved further when dose and total daily reactive metabolite burdens were considered. In conclusion, there appears to be a general trend between the extent of thiol adduct formation and the potential for DIT, which would support the preclinical measurement and minimization of the property through screening of thiol adduct formation as part of an overall discovery optimization paradigm. Introduction It is now generally accepted that the predominant pathway for the generation of certain classes of drug-induced toxicity (DIT)1 is through the generation of reactive metabolites (1-5). The toxicities usually ascribed to reactive intermediate formation are drug-induced liver, skin, and hematopoietic toxicities. Although there is much active research and debate on the exact mechanisms by which reactive intermediates provoke a toxic response, the common initial step in the process is widely accepted to be the formation of an intermediate with sufficient chemical reactivity to form a covalent bond with an intracellular molecule. Key to this understanding has been extensive studies both in vitro and in animals, demonstrating that drugs that cause clinical toxicity also generate reactive species. Notable examples where this positive correlation has been shown are acetaminophen, troglitazone, nefazodone, etc. There has also been a great deal of literature on methods of detecting the formation of reactive intermediates (6-9). These methods generally measure small molecule adduct formation or the formation of adducts with large molecules with radio* To whom correspondence should be addressed. Tel: 609-252-7885. E-mail:
[email protected]. 1 DIT, drug-induced toxicity; dansyl, 5-(dimethylamino)-1-naphthalenesulfonate; dGSH, dansyl glutathione; DTT, dithiothreitol; MRM, multiplereaction monitoring; NL, neutral loss; EPI, enhanced product ion; HLM, human liver microsome.
detection. A fundamental problem in putting these techniques into practice in a prospective manner is that there has been little if any examination of the reactive intermediate generation of compounds that are not associated with toxicity. As the correlation between extent of reactive intermediate formation and toxicity signals is far from precise, this lack of definition of what a “safe profile” looks like makes minimizing the property a more challenging exercise. Many druglike molecules will form some level of reactive intermediate when measured with sensitive analytical techniques so questions such as “can we tolerate any level of adduct formation?” and “how low do adduct levels need to be?” plague the development of the structure-activity relationships for many chemotypes. To better address the question of what the “safe profile” looks like, this study examined the adduct formation found when 40 drugs not associated with liver or other toxicities, including many of the most prescribed drugs in the Physician’s Desk Reference (PDR) (10), were incubated with human liver microsomes (HLMs) and dansylated glutathione.
Experimental Procedures Materials. The 10 drugs associated with DIT screened in this study (DIT group) were acetaminophen, carbamazepine, clozapine, diclofenac, flutamide, nimesulide, phenytoin, terbinafine, troglitazone, and zileuton. A known hepatotoxin, pulegone, was used as a positive control. The 40 drugs not considered to be associated with DIT screened in this study (non-DIT group) were amlodipine,
10.1021/tx800368n CCC: $40.75 2009 American Chemical Society Published on Web 03/02/2009
In Vitro Screening for Thiol Adduct Formation from 50 Drugs amoxicillin, atorvastatin, azithromycin, celecoxib, ciprofloxacin, citalopram, clarithromycin, duloxetine, enalapril, famotidine, fexofenadine, fluorexetine, furosemide, gabapentin, ibuprofen, isotretinoin, lansoprazole, levofloxacin, lisinopril, loratidine, losartan, montelukast, olanzapine, omeprazole, oxycodone, paroxetine, pioglitazone, raloxifene, ranitidine, risperidone, rofexicob, rosiglitazone, sertraline, sildenafil, simvastatin, sumatriptan, tomoxetine, tramadol, and venlafaxine. Amlodipine, pioglitazone, sildenafil, and zileuton were obtained from Chempacific (Baltimore, MD); celecoxib, ciprofloxacin, citalopram, fexofenadine, lisinopril, loratidine, paroxetine, risperidone, sertraline, venlafexine, rofecoxib, simvastain, and sumatriptan were obtained from Toronto Research Chemicals (North York, ON, Canada); clarithromycin was obtained from LKT (St. Paul, MN); nilutamide, losartan, and ibuprofen were obtained from Cayman Chemical (Ann Arbor, MI); terbinafine was obtained from Spectrum Chemicals (Gardena, CA); oxycodone was obtained from Cambridge Isotope Laboratories (Andover, MA); tomoxetine was obtained from Tocris (Ellisville, MO); and tramadol was obtained from Alltech (Nicholasville, KY). Atorvastatin, olanzapine, and duloxetine were extracted and purified from their respective brand name drugs. All other drugs were obtained from Sigma-Aldrich and its affiliates (St. Louis, MO). Pooled HLMs were obtained from BD Gentest (Woburn, MA), and dansylated reduced glutathione (dansyl glutathione, dGSH) was synthesized and purified in house according to published procedures (11). The dGSH was more than 99.9% pure as assessed by HPLC with fluorescence detection. All other reagents and solvents were of analytical reagent grade or better. In Vitro Incubations with HLM. All incubations were conducted in 96 well format. Test compounds (50 µM) were preincubated in 0.1 M potassium phosphate buffer (pH 7.4) with 1 mg/ mL HLM and 1 mM dGSH for 5 min at 37 °C. NADPH (1 mM final concentration) was then added to initiate the reaction. The final incubation volume was 0.2 mL. After 30 min, the reaction was terminated by the addition of 2 volumes of ice-cold methanol with 5 mM dithiothreitol (DTT). The plates were vortexed and centrifuged, and the resulting supernatants were stored at -80 °C before analysis. The screening was conducted in three steps. In the first step, all compounds were evaluated for additional fluorescence peaks in the LC/fluorescence chromatograms as compared to the blank samples (incubation without drugs) and the compound controls (incubations without the addition of NADPH and dGSH). In the next step, all compounds with additional fluorescence peaks were incubated with more control experiments including incubations in the absence of NADPH, or dGSH, or both. In the final step, all samples with confirmed dGSH adduct formation were analyzed by LC/MS to determine the mass of the adducts and potentially the structures of adducts by further MS fragmentation. HPLC/UV/Fluorescence Method. All incubation samples were injected onto a Shimadzu LC-10Avp HPLC system comprised of binary pumps, a well-plate autoinjector (chilled), a fluorescence detector, and a diode array detector. Analyte separation was accomplished using a reverse phase HPLC column (Phenomenex Prodigy ODS2, 4.6 mm × 150 mm). A shallow mobile phase gradient (3 min at 20% acetonitrile in 0.1% formic acid, ramps up to 50% acetonitrile in 20 min followed by another ramp to 90% in 10 min) was used to ensure adequate separation of adduct peaks (11). For the determination of metabolic turnover of substrate, UV peak areas of the respective substrate at the λmax were determined, and % turnover was calculated as the percentage of the peak area after incubation in the presence of NADPH and dGSH over that of control in the absence of NADPH and dGSH. For the determination of dGSH adduct concentration, dansyl-related fluorescence was monitored at λex of 340 nm and λem of 525 nm. The adduct peaks were identified by visual comparison of chromatograms from the incubated samples and their relevant controls. An external standard curve of dGSH was used for the quantitative determination of dGSH adduct concentrations.
Chem. Res. Toxicol., Vol. 22, No. 4, 2009 691 LC/MS for dGSH Adduct Mass Determination. LC/MS data were acquired using a Shimadzu LC-10Avp HPLC system equipped with a RFL-10A fluorescence detector (Shimadzu, Columbia, MD), a LEAP autosampler (Leap Technology, Carrboro, NC), and an API-4000 Qtrap mass spectrometer (Applied Biosystems, Foster City, CA). For adduct separation, a Synergy HydroRP column (250 mm × 4.6 mm, 4 µm, Phenomonex, Torrance, CA) using a 5 mM ammonium acetate buffer containing 0.02% formic acid as mobile phase A and acetonitrile as mobile phase B at a flow rate of 1.0 mL/min. The gradient started from 10 to 60% B in 30 min, followed by 60 to 100% B in 3 min, and maintained at 100% B for 2 min prior to column re-equilibration. The HPLC flow after column was diverted to the fluorescence detector (75%) and the mass spectrometer (25%) by an LC Packings Acurate diverter (Dionex, Sunnyvale, CA). The parameters for fluorescence detection were set as described above. A 50 µL aliquot of each processed sample was subjected to LC/ MS analysis in both the positive and the negative electrospray ionization modes. The deprotonated dansyl moiety (5-dimethylamino-1-naphthalenesulfinic acid, m/z ) 234) was used as the product ion for negative multiple-reaction monitoring (MRM), and the product ions formed by a neutral loss (NL) of dansyl-glutamic acid (362 Da) of the precursor ions were used for positive MRM. The precursor ions were set at (50 Da around the nominal mass of covalent conjugation of dGSH to either the parent molecules (P + dGSH-2H) or the potential metabolites with more than 50 Da mass difference from the parent molecules (N- or O-dealkylations and dissociation of the sulfinylbenzoimidazole bond). Data-dependent product ion scans (EPI) were employed following the MRM scans using IDA criterion of 500 counts. Daily Burden of Reactive Metabolites Calculation. To factor in the dose, extent of absorption, and metabolic clearance pathways, the daily burden of thiol-reactive metabolites (Drm) was estimated based on the following formula:
Drm ) D × fa × fm × frm in which D is the total daily dose (mg/day), fa is fraction absorbed, fm is fractional clearance via oxidative metabolism, and frm is the fraction of oxidative metabolism leading to thiol adduct formation. For the total daily dose values, the maximum doses recommended by the U.S. label are used. The values for fa and fm were extracted from the PDR (10), United States Prescribing Information, and, if available, publications of human ADME data for individual drugs (individual references are provided in Table 2). The value for frm was calculated as the ratio of the % dGSH adduct formation over the % metabolic turnover in the same incubation.
Results In the initial screening, 16 compounds were found to generate additional fluorescence peaks in the LC/fluorescence chromatograms in comparison with blank samples and respective controls. Of the 16 compounds, six (including pulegone) are from the DIT group, and 10 are from the non-DIT group (Figure 1). The structures of these 16 compounds are shown in Scheme 1. The drugs with positive signals were subsequently evaluated with control experiments without the addition of either NADPH or dGSH. The fluorescence peaks from all 16 drugs were not present in incubations without the addition of dGSH, indicating the absence of interference from fluorescent parent/metabolites. In contrast, the fluorescence peaks from incubations with omeprazole, lansoprazole, and montelukast were also present in their respective controls without the addition of NADPH, indicating that the adduct formation from these three drugs did not require oxidative metabolic activation. To characterize the dGSH adducts formed, MRM experiments followed by data-dependent product ion scans (MRM-EPI) were conducted on a hybrid Q-trap mass spectrometer (12). The samples were analyzed by LC/fluorescence/MS in both positive
692
Chem. Res. Toxicol., Vol. 22, No. 4, 2009
Gan et al.
Scheme 1. Chemical Structures of All Drugs with Detectable Adduct Formation but Are Not Associated with DIT and All Drugs That Are Associated with DITa
a
Not shown are the structures of the 30 drugs that neither are associated with DIT nor form detectable adducts.
and negative ionization modes. The negative MRM monitored the collision-induced dissociation (CID) of a characteristic product ion of dGSH (m/z ) 234), while the positive MRM monitored the CID of a characteristic NL of dGSH (NL ) 362 Da). The masses of the major adducts from each drug are listed in Table 1. The masses of drug adducts from acetaminophen, diclofenac, pulegone, clozapine, and troglitazone are consistent with previously reported values. Because of assay sensitivity issues and adduct stability in the case of paroxetine, adduct masses of duloxetine, rosiglitazone, and paroxetine could not be obtained. As an example of adduct mass determination, the LC/MS chromatogram and MS2 spectra from the omeprazole incubation
are shown in Figure 2. Both negative and positive MRM detected a peak of dGSH adduct at the same retention time (Figure 2A,B). The m/z values of the molecular ions in negative and positive modes were found as 685 and 687, respectively. The MS2 spectra of this adduct clearly exhibited the characteristic product ions and NLs of dGSH (Figure 2C,D). From the m/z of molecular ions and fragmentation patterns, the structure of this dGSH adduct was proposed as substitution by dGSH on the benzimidazole with replacement of the sulfinyl moiety. Similarly, LC/MS spectra of the structurally related lansoprazole also indicated that the dGSH adduct found by fluorescence detection was a dGSH conjugation to the benzimidazole moiety (Table 1). Therefore, a non-P450-mediated conjugation pathway of these two compounds was proposed in Scheme 2. Similar to
In Vitro Screening for Thiol Adduct Formation from 50 Drugs
Chem. Res. Toxicol., Vol. 22, No. 4, 2009 693
Table 1. Extent of dGSH Adduct Formation and Substrate Turnover (as a % of Initial Parent Concentrations) and the Mass Assignment of Major Adducts for All Drugs with Detectable dGSH Adduct Formationa drug
% adduct formation
mass of major adduct (Da)
risperidone acetaminophen duloxetine simvastatin rosiglitazone nimesulide paroxetine diclofenac losartan
0.05 0.5 0.10 0.11 0.12 0.35 0.51 0.52 0.82
montelukast pulegone raloxifene omeprazole clozapine troglitazone lansoprazole
1.47 1.82 2.03 2.50 3.15 3.93 6.48
864 (multiple) 689 NC 968 NC 817 NC 815 960 930 (two peaks) 1125 (two peaks) 688 1011 (three peaks) 686 864 (multiple) 979 656
proposed adduct composition M+ M+ NC M+ NC M+ NC M+ M+ M+ M+ M+ M+ M+ M+ M+ M+
dGSH-2H dGSH-2H (39) dGSH-10 dGSH + O-NO2-2H dGSH-Cl-2H (11, 20, 21) dGSH-2H dGSH-28 dGSH dGSH-4H (13, 40) dGSH-2H (18) dGSH-C9H13NO2S (15) dGSH-2H (23, 24) dGSH-2H (25, 27) dGSH-C9H10F3NO2S
% substrate turnover 10.2 ND 1.8 100 18.2 4.3 ND 68.2 12 ND 55.2 16.3 23 13.9 18.5 0.7
a NC, not characterized due to mass spectrometry sensitivity or adduct stability (paroxetine); ND, not determined due to either insufficient UV chromatographic separation of parent drug from interference.
Table 2. Estimated Daily Reactive Metabolite Formation from Drugs with Detectable dGSH Adductsa
drugs risperidone duloxetine simvastatin rosiglitazone paroxetine losartan raloxifene
maximum daily dose (mg) 8 60 80 8 60 100 60
fa
fm
not associated 0.70 0.95b ∼0.96 1.00 0.73 ∼0.95c 0.99 1.00 1.00 0.97 1.00 0.14d 0.60 0.01
frm
Drm (mg)
refs
0.005 0.05 0.001 0.01 NC 0.07 0.12
0.03 3.07 0.06 0.05 NC 0.98 0.04
41 42 43, 44 45 46, 47 48 49
NC 0.01 0.23 0.21 0.08
380e 1.29 166 12.7 5.3-14.4
50, 51 52, 53 54 55 56
associated acetaminophen diclofenac clozapine troglitazone nimesulide
4000 200 900 600 200
e
e
1.00 >0.8 0.50 ∼1
0.8-0.9 ∼1.00 0.20 0.33-0.9f
a fa, fraction absorbed; fm, fraction of clearance through oxidative metabolism; frm, fraction of oxidative metabolism leading to reactive metabolites; and Drm, total daily burden of reactive metabolites. b The fraction of oxidative metabolism in extensive metabolizers is used. c Detailed metabolite profiles in excreta were not reported for simvastatin. fm is estimated based on published drug interaction data of simvastatin with itraconazole (44). Coadministration of itraconazole, a CYP3A inhibitor, resulted in a simvastatin AUC ratio of 19, thus (1 fm) ) AUCc/AUCi ) 1/19, and fm ) 0.95. d fm of losartan is estimated based on the report that 14% of losartan dose was converted to its acid metabolite, E3174 (48). e Drm of acetaminophen is deduced based on a total of 9.5% of dose excreted as mercapturic acid and cysteine adduct in the urine (7.9%) and bile (1.6%) in man after oral dosing of acetaminophen (50, 51). f Nimesulide radioactivity was recovered in urine and feces as various oxidative and nitro reductive metabolites, mainly as M1 (hydroxylated) and M5 (nitro-reduced and hydroxylated). Only 33% of the dose was recovered as M1, thus definitively assigned to oxidative metabolism; around 10% was recovered as nitro-reduced metabolites, and the rest were metabolites with both oxidation and nitro reduction (56).
omeprazole and lansoprazole, montelukast also reacted with dGSH in the absence of NADPH to form two adducts with the same mass. The adduct mass was 1125 Da, consistent with direct conjugation of dGSH to montelukast (Table 1). An attempt was made to take into consideration dose, fraction absorbed, fraction of oxidative metabolism, and the partition ratio leading to adduct formation of all drugs that generated detectable adducts. Table 2 summarizes the calculated Drm values (daily burden) for 12 drugs with detectable adduct formation.
Figure 1. Extent of dGSH adduct formation from incubations of 50 drugs with HLM in the presence of NADPH and dGSH. The filled bars represent drugs associated with DIT, and the open bars represent drugs that are not associated with DIT. Asterisks point to the three drugs (omeprazole, lansoprazole, and montelukast) that form dGSH adducts without oxidative activation.
Most of the Drm values of DIT group drugs are higher than those of the not-DIT group drugs. Omeprazole, lansoprazole, and
694
Chem. Res. Toxicol., Vol. 22, No. 4, 2009
Gan et al.
Figure 2. LC/MS analysis of omeprazole HLM incubation. (A) TIC of the negative MRM chromatogram of the omeprazole incubation mixture. (B) TIC of the positive MRM of the omeprazole incubation mixture. (C) Negative product ion spectrum of the dGSH adduct peak (m/z ) 685) eluted at 16.7 min. (D) Positive product ion spectrum of the dGSH adduct peak (m/z ) 687) eluted at 16.7 min.
Scheme 2. Proposed Chemical Mechanism of NADPH-Independent dGSH Adduct Formation from Omeprazole and Lansoprazole
montelukast were not included in this calculation since their adduct formation is not thought to be a result of metabolic activation.
Discussion Bioactivation is believed to be the initial step toward druginduced host injury by many drugs, although the sequence of events following the bioactivation is still unclear. Bioactivation of drugs
can be mediated by various host cells (hepatocytes, neutrophils) and enzymatic systems (cytochrome P450s, FMO, esterases, peroxidases, UDP-glucuronosyl transferases) and can lead to different reactive species including epoxides, Michael acceptors, free radicals, acyl glucuronides, and hard electrophiles such as iminium ions. Thiol-containing reagents such as glutathione can only react with a subset of these reactive species such as epoxides and Michael acceptors; thus, this study is limited in its scope in
In Vitro Screening for Thiol Adduct Formation from 50 Drugs
Chem. Res. Toxicol., Vol. 22, No. 4, 2009 695
Figure 3. Scatter plot of % dGSH adduct formation (a) and estimated total daily burden (b) in the DIT and non-DIT groups. The open circles and triangles represent drugs not associated and associated with DIT, respectively. For illustrational purposes, a horizontal dotted line is plotted at 0.2% adduct level in panel a, and another is plotted at the 1 mg level in panel b. Adduct levels of omeprazole, lansoprazole, and montelukast are not shown in this figure.
evaluation of the potential for drugs to generate thiol-reactive species upon oxidative metabolic activation in incubations with liver microsomes. Furthermore, the formation of reactive species from some drugs is through multiple sequential steps with some of these steps not catalyzed by liver microsomal enzymes. One need also be cognizant that the reactivity of each reactive metabolite can be different toward thiols; thus, thiol trapping can be incomplete when a less reactive metabolite is generated. Although this is a limitation of the experimental design, many of the drugs that are associated with DIT form thiol-reactive metabolites, and trapping with a form of thiol-containing reagents is commonly used in the pharmaceutical industry as the “first line” screening tool to minimize reactive metabolite formation. The detection of a fluorescence signal with dGSH adducts enables semiquantitative measurement of thiol-reactive metabolite formation and makes it possible for the broad screening of highly prescribed drugs described in this study without the use of radiolabeled materials. Another feature of dGSH is that it does not serve as cofactor for glutathione S-transferases, so the trapping by dGSH measures the extent of chemical reactivity between reactive metabolites and dGSH (11). The dGSH trapping assay is used at BMS in selected discovery programs at the stage of SAR optimization after high levels of GSH adduct have been detected for a lead compound from a chemotype of interest. The semiquantitative nature of this assay enables a rapid test of hypotheses on bioactivation pathways by substitutions at suspected sites without the costly and timeconsuming synthesis of radiolabeled analogues. No absolute threshold is set for compound rejection, although for compounds projected to be administered at relatively high doses, less than 1% of adduct formation relative to starting drug concentrations are generally deemed acceptable as long as there is reasonable overall turnover of the compound. It is important to note that this assay is frequently followed up with in vivo studies to assess the property of reactive metabolite formation in a fully integrated biological system. Finally, the risk of reactive metabolite formation is only part of multifaceted, integrated benefit/risk assessment of a drug candidate and rarely serves as a stand-alone criterion to block the progression of a candidate molecule. This study was aimed at providing quantitative information on the potential to form reactive metabolites of commonly used drugs that are not associated with DIT and benchmark it against a panel of drugs that are associated with DIT. A total of 51 compounds were screened in the dGSH adduct assay, which included one positive control (natural product pulegone, which is a hepatotoxin) (13), 10 drugs in the DIT group, and 40 drugs in the non-DIT group. As shown in Table 1, while a majority of the drugs in the non-DIT group did not generate detectable adducts by the sensitive fluorescence assay (30 out of 40), half of the DIT group
drugs formed detectable adducts (5 out of 10 or 6 out of 11 if pulegone is included). As shown in Figure 3a, when the three drugs that formed adducts through nonoxidative conjugation are excluded (discussed below), the extent of adduct formation appears to be more in the DIT group. Nine drugs (three from the not-associated group and six from the DIT group) formed adducts at greater than 0.2%. Thus, there are 6 out 11, 55% (5 out of 10, 50% excluding pulegone) in the DIT group that formed adducts more than 0.2%, while there are 34 out of 40, 85% (34 out of 37, 92% excluding three drugs with nonoxidative conjugation) in the non-DIT group that formed adducts less than 0.2%. To put the percentage of adduct formation into perspective, 0.2% of adduct formation is equivalent to a rate of formation at 200 pmol thiol adduct/h/mg protein assuming linear kinetics. This number is reasonably consistent with the 50 pmol protein covalent binding/h/mg protein, which was proposed by Merck scientists as thresholds for a compound to be considered as potentially problematic (14). These values are generated with different assays, substrate concentrations (50 vs 10 µM), and readouts (thiol trapping vs protein covalent binding) than what was used by the Merck group. Because of the screening nature of this study, linearity of adduct formation was not investigated. As shown in Table 1, although the extents of dGSH adduct formation were all below 10% of initial parent concentration, the extents of parent drug turnover varied from minimal to 100% in this experiment; thus, the fraction of reactive metabolite to total oxidative metabolism (frm) calculation may be misleading for some drugs. The 0.2% adduct level is slightly lower than the aforementioned current BMS threshold of 1%; however, for a typical chemotype for which this assay was employed, the adduct levels of initial lead compounds were significantly above 1%, and the SAR optimization process through this assay generally greatly diminished the adduct levels. It is worth pointing out that three of the drugs that are positive for dGSH adducts, that is, omeprazole, lansoprazole, and montelukast formed dGSH adducts in the absence of NADPH. The formation of omeprazole GSH adducts has been previously reported with detection in the urine of rats as benzoimidazole N-acetylcysteine conjugate (15). The adduct masses observed from both omeprazole and lansoprazole in this study are consistent with the reported adduct structure; however, the lack of involvement of oxidative metabolism indicated a chemical substitution of the sulfinyl moiety rather than a sulfur oxidation pathway as previously suggested (15). Interestingly, proton pump inhibitors interact with cysteinyl moieties in the active site of H+/K+ ATPase (16), but the chemistry of their covalent binding to H+/K+ ATPase is different from what was observed in this study. The montelukast adduct mass is consistent with a direct addition of dGSH; however, the exact nature of this adduct
696
Chem. Res. Toxicol., Vol. 22, No. 4, 2009
was not characterized. A structurally related analogue verlukast (MK-0679) was reported to form GSH adducts in rat liver and kidney cytosols by 1,4-Michael addition to the bridge ethylene of the styryl quinoline (17). It is possible that dGSH reacted with montelukast in a similar manner. Of the 12 drugs and one positive control that formed dGSH adducts after oxidative bioactivation, eight of them were previously reported to form thiol adducts, namely, pulegone, rosiglitazone, raloxifene, diclofenac, clozapine, troglitazone, paroxetine, and acetaminophen (13, 18-29). In addition, reported here for the first time is the adduct formation from five drugs including risperidone, duloxetine, simvastatin, losartan, and nimesulide. The molecular weights of most adducts were identified by LC/MS; however, the exact structures of these adducts would require additional investigation. Five hepatotoxic drugs including carbamazepine, flutamide, phenytoin, terbinafine, and zileuton were not found to form detectable adducts in this study. Of them, flutamide and carbamazepine were recently reported to form thiol adducts detectable by LC/MS techniques (12, 30). In addition, a reversible glutathione adduct formation was also reported for terbinafine (31). Incubation of radiolabeled carbamazepine with HLM in the presence of NADPH also resulted in covalent protein binding; however, the level of covalent binding was very low (
