Bioactivation of a Novel 2-Methylindole-Containing DualChemoattractant Receptor-Homologous Molecule Expressed on

T-Helper Type-2 Cells/D-Prostanoid Receptor Antagonist Leads toMechanism-Based CYP3A Inactivation: Glutathione Adduct

Characterization and Prediction of In Vivo Drug-Drug Interaction

Simon G. Wong, Peter W. Fan, Raju Subramanian, George R. Tonn,1 Kirk R. Henne,Michael G. Johnson, Michelle Tadano Lohr, and Bradley K. Wong

Departments of Pharmacokinetics and Drug Metabolism (S.G.W., P.W.F., R.S., G.R.T., K.R.H., M.T.L., B.K.W.) and MedicinalChemistry (M.G.J.), Amgen Inc., South San Francisco, California

Received November 23, 2009; accepted January 25, 2010

ABSTRACT:

The 2-methyl substituted indole, 2MI [2-(4-(4-(2,4-dichlorophenyl-sulfonamido)-2-methyl-1H-indol-5-yloxy)-3-methoxyphenyl)aceticacid] is a potent dual inhibitor of 1) chemoattractant receptor-homologous molecule expressed on T-helper type-2 cells and2) D-prostanoid receptor. During evaluation as a potential treat-ment for asthma and allergic rhinitis, 2MI was identified as amechanism-based inactivator of CYP3A4 in vitro. The inactivationwas shown to be irreversible by dialysis and accompanied by anNADPH-dependent increase in 2MI covalent binding to a 55- to60-kDa microsomal protein, consistent with irreversible binding toCYP3A4. Two glutathione (GSH) adducts, G1 and G2, were identi-fied in vitro, and the more abundant adduct (G1) was unambigu-ously determined via NMR to be GSH adducted to the 3-position ofthe 2-methylindole moiety. The potential for a clinical drug-drug

interaction arising from mechanism-based inactivation of CYP3A4by 2MI was predicted using a steady-state model, and a 4.3- to7.5-fold increase in the exposure of midazolam was predicted atanticipated therapeutic concentrations. To better assess the po-tential for in vivo drug-drug interactions, the Sprague-Dawley ratwas used as an in vivo model. An excellent in vitro-in vivo corre-lation was observed for the reduction in enzyme steady-state con-centration (E�ss/Ess) as well as the change in the exposure of aprototypical CYP3A substrate, indinavir (area under the curve(AUC) for indinavir/AUC). In summary, 2MI was identified as apotent mechanism-based inactivator of CYP3A and was predictedto elicit a clinically relevant drug-drug interaction in humans at ananticipated therapeutic concentration.

Prostaglandin D2, the major cyclooxygenase product formed inactivated mast cells, is an important mediator of the inflammatoryresponse associated with asthma. PGD2 activity is mediatedthrough activation of chemoattractant receptor-homologous mole-cule expressed on T-helper type-2 cells (CRTH2) and D-prostanoid(DP) receptor, and antagonism of CRTH2 and DP has been pro-posed as a potential therapeutic target for the treatment of asthmaand allergic rhinitis (Pettipher, 2008). The 2-substituted methylindole, 2MI [2-(4-(4-(2,4-dichlorophenylsulfonamido)-2-methyl-

1H-indol-5-yloxy)-3-methoxyphenyl)acetic acid] (Fig. 1) is a po-tent dual DP and CRTH2 antagonist and a potential treatment for asthmaand allergic rhinitis. Preliminary absorption, distribution, metabolism, andexcretion studies revealed that it was also a potent mechanism-based inhibitorof CYP3A4.

Mechanism-based inactivation (MBI) of CYP3A4 has been asso-ciated with clinically relevant drug-drug interactions (DDIs) (Zhou etal., 2005; Kalgutkar et al., 2007; Riley et al., 2007). Unlike compet-itive inhibition, metabolic activity after MBI is restored only after denovo enzyme synthesis (Silverman, 1995; Hollenberg, 2002), result-ing in a prolonged reduction of enzyme activity. This irreversibleinactivation occurs as the result of P450-mediated bioactivation of theparent compound to reactive species that are unable to leave the activesite, leading to modifications of the enzyme either through tight

1 Current affiliation: Department of Drug Metabolism and Pharmacokinetics, ElanBiopharmaceuticals, South San Francisco, California.

Article, publication date, and citation information can be found athttp://dmd.aspetjournals.org.

doi:10.1124/dmd.109.031344.

ABBREVIATIONS: CRTH2, chemoattractant receptor-homologous molecule expressed on T-helper type-2 cells; DP, D-prostanoid; 2MI, (2-(4-(4-(2,4-dichlorophenylsulfonamido)-2-methyl-1H-indol-5-yloxy)-3-methoxyphenyl)acetic acid); MBI, mechanism-based inactivation; DDI, drug-drug interactions; P450, cytochrome P450; HLM, human liver microsomes; RLM, rat liver microsomes; OH, hydroxy; ACN, acetonitrile; LC, liquidchromatography; MS/MS, tandem mass spectrometry; TAO, troleandomycin; ABT, 1-aminobenzotriazole; GSH, glutathione; PAGE, polyacryl-amide gel electrophoresis; MRM, multiple reaction monitoring; HRP, horseradish peroxidase; SPE, solid-phase extraction; 1D, one dimensional;2D, two dimensional; HMBC, heteronuclear multiple quantum correlation; AUC, area under the curve; AUCi, AUC in the presence of the inhibitor;MS, mass spectrometry; 3MI, 3-methylindole.

0090-9556/10/3805-841–850$20.00DRUG METABOLISM AND DISPOSITION Vol. 38, No. 5Copyright © 2010 by The American Society for Pharmacology and Experimental Therapeutics 31344/3574564DMD 38:841–850, 2010 Printed in U.S.A.

841

at ASPE

T Journals on M

ay 24, 2018dm

d.aspetjournals.orgD

ownloaded from

coordination with the prosthetic heme or covalent modification ofP450 apoprotein and/or heme.

Recognizing the potential for 2MI to elicit an adverse DDI precip-itating from MBI, the present studies were undertaken to 1) elucidatethe putative reactive intermediate(s) and the mechanism of enzymeinactivation, 2) to predict the magnitude of in vivo DDIs based on invitro inactivation kinetics using a steady-state model, and 3) to inves-tigate the propensity for 2MI to elicit MBI and DDI using theSprague-Dawley rat as an in vivo model.

Materials and Methods

Chemicals and Enzymes. 2MI was synthesized by the Amgen MedicinalChemistry Department (Amgen, South San Francisco, CA). [35S]2MI (50mCi/mmol, �99% purity) was synthesized by Moravek Biochemicals (Brea,CA). [1-14C]Naphthalene and horseradish peroxidase type XII were purchasedfrom Sigma-Aldrich (St. Louis, MO). Pooled HLM, RLM, and recombinantCYP3A4-expressed microsomes (Supersomes, containing 960 pmol of cyto-chrome b5/mg of protein) were purchased from BD Biosciences (San Jose,CA). Deuterated 6-�-OH-testosterone (6�-hydroxytestosterone-16,16,17-D3,�99% purity) was purchased from Cerilliant Corporation (Round Rock, TX).All other chemicals were obtained from commercial sources and were of thehighest purity available.

Handling of and Treatment of Animals. Studies with rats were performedin accordance with the Institutional Animal Care and Use Committee ofAmgen. Male Sprague-Dawley rats (250–300 g) were obtained from CharlesRiver Laboratories (Wilmington, MA).

Microsomal Studies. To evaluate the potential for time-dependent P450inactivation, incubations containing final concentrations of 2MI (10 �M), HLM, orRLM (1.0 mg/ml) in potassium phosphate buffer (100 mM, pH 7.4) were initiatedwith NADPH (1 mM). After initiation, aliquots of this primary incubation weretransferred at 0, 5, 10, 15, and 30 min to a secondary incubation (at 1:10 dilution)containing a saturating concentration of the CYP3A substrate testosterone (200�M), NADPH (1 mM), and phosphate buffer and incubated for an additional 20min. Formation of 6-�-OH-testosterone under these conditions was confirmed tobe linear with protein concentration and incubation time (data not shown). Thesecondary reaction was terminated by addition of 1 volume of acetonitrile (ACN)containing deuterated 6-�-OH-testosterone as the internal standard. The sampleswere centrifuged and CYP3A activity was determined by measuring 6-�-OH-testosterone formation by an LC-MS/MS method as described previously (Walskyand Obach, 2004). To ascertain the contribution of CYP3A4, the preincubationwas also performed in the presence of a competitive CYP3A4 inhibitor, ketocon-azole (1 �M). For assessment of MBI of other human P450 isoforms, incubationconditions were identical to those described previously and used the followingselective substrates: 200 �M phenacetin (CYP1A2), 20 �M diclofenac(CYP2C9), and 50 �M bufuralol (CYP2D6). The respective reaction productswere monitored (acetaminophen, 4�-OH-diclofenac, and 1-OH-bufuralol) by LC-MS/MS assay as described previously (Walsky and Obach, 2004).

Kinetic Studies. The kinetics of CYP3A inactivation were determinedusing a range of 2MI concentrations (0.1–100 �M). Nonspecific binding of2MI to microsomal protein was determined to be �10%, and therefore nominal

concentrations were used for kinetic calculations. Incubations were conductedas described above, with a 0- to 30-min preincubation followed by a secondaryincubation. Data were plotted as the natural log (LN)-linear plot of thepercentage of remaining CYP3A activity versus preincubation time, andobserved initial inactivation rate constants (kobs) were calculated from thepseudo-first-order kinetics slope. The maximum rate constant for enzymeinactivation (kinact) and the inactivation constant (KI) were calculated (Graph-Pad Prism; GraphPad Software Inc., San Diego, CA) according to the follow-ing hyperbolic equation:

kobs �kinact � Io

KI � Io

where kobs is the initial rate constant for CYP3A inactivation and Io is the initialinhibitor concentration (Waley, 1985; Silverman, 1995).

Partition Ratio. The partition ratio, the number of moles of product formedper mole of enzyme inactivated, was determined using recombinant CYP3A4and the titration method (Silverman, 1995). In brief, 2MI (0–10 �M) waspreincubated with a fixed amount of recombinant CYP3A4 (200 pmol/reac-tion) in the presence of NADPH (1 mM) in phosphate buffer for 60 min toensure complete inactivation. The percentage of remaining CYP3A activity ateach substrate concentration was plotted against the molar ratio of 2MI toCYP3A4 enzyme. The intercept between the linear regression line obtained atmolar ratios less than 5 and the horizontal line obtained from saturatingconditions were extrapolated to the x-axis to yield the turnover number. Thepartition ratio was calculated by subtracting a value of 1 from the turnovernumber.

Assessment of Reversibility. The reversibility of time-dependent CYP3Ainhibition was assessed by dialysis. In brief, 2MI (20 �M), troleandomycin(TAO) (10 �M), or raloxifene (10 �M) was incubated with HLM (1 mg/ml)in the presence of NADPH (1 mM) in phosphate buffer for 30 min. After theincubation, an aliquot was removed for determination of the predialysisCYP3A activity. An aliquot (0.5 ml) of the remaining reaction mixture wastransferred to a Slide-a-Lyzer (Pierce Biotechnology, Rockford, IL) dialysiscassette (10,000 mol. wt. cutoff) and dialyzed against 2 liters of phosphatebuffer (0.1 M, pH 7.4) for approximately 22 h at 4°C. The buffer was replacedonce during the dialysis at approximately 6 h of dialysis. Pre- and postdialysisCYP3A activity, after a 10-fold dilution, was determined by monitoring6-�-OH-testosterone formation as described above. CYP3A activities werethen normalized to protein content, which was determined using a BCA assaykit (Pierce Biotechnology).

CO Binding and Spectrophotometric Studies. The effect of 2MI oncarbon monoxide binding to reduced P450 was determined in HLM and RLM.2MI (100 �M) was incubated with 1 mg/ml HLM or RLM in phosphate bufferin the presence and absence of NADPH. After 30 min, an aliquot was removed,reduced with sodium dithionite (2–5 mg), and equally divided between two1-ml (volume), 1-cm path length cuvettes. After recording baseline absor-bance, carbon monoxide was bubbled into the sample cuvette for 1 min, andthe absorbance was scanned from 400 to 600 nm using a Cary UV spectro-photometer. Incubations containing 1-aminobenzotriazole (ABT) (5 mM) wereused as a positive control.

Microsomal Covalent Binding Assay. Covalent binding to microsomalprotein was assessed for [35S]2MI (20 �M) and the positive control[1-14C]naphthalene (20 �M). Incubations (0.2 ml) with HLM or RLM (1mg/ml protein) were conducted in phosphate buffer for 1 h at 37°C and wereperformed in the absence of NADPH (used as a negative control) or in thepresence of NADPH (1 mM) with or without GSH (10 mM). Covalently boundradioactive equivalents to protein were measured with a semiautomated proteinwashing procedure using a cell harvester (Semi-Automated Cell HarvesterSystem; Brandel, Gaithersburg, MD) according to the method described byDay et al. (2005). After 1 h of incubation, aliquots (200 �l) of the incubationswere taken and transferred to a second 96-well plate (2-ml well volume)containing 1 ml of acetone/well. The samples were vortexed (30 s) and washedusing the Semi-Automated Cell Harvester System (48-Sample; Brandel). Theprecipitated protein was trapped onto a Whatman glass fiber filter mat, washedwith 80% acetone/water (2 ml/well, v/v), and treated further by 10 washes with80% methanol-water (2 ml/well, v/v). The captured washed proteins on theglass fiber filter mat sections were transferred to another 96-well plate (2-ml

FIG. 1. Chemical structure of 2MI [2-(4-(4-(2,4-dichlorophenylsulfonamido)-2-methyl-1H-indol-5-yloxy)-3-methoxyphenyl)acetic acid].

842 WONG ET AL.

at ASPE

T Journals on M

ay 24, 2018dm

d.aspetjournals.orgD

ownloaded from

well volume) and treated with KOH (1 ml, 0.1 N) for 1 h in a heated water bath(95°C) with moderate horizontal shaking. The hydrolyzed samples were al-lowed to cool on ice (15 min), and aliquots (25 �l) were taken for proteinconcentration determination using the BCA protein assay (Pierce Biotechnol-ogy). Finally, aliquots (600 �L) of the hydrolyzed solutions were added toscintillation vials (6 ml) containing Pico-Fluor 40 liquid scintillant (5 ml;PerkinElmer Life and Analytical Sciences, Waltham, MA). Radioactivity wascounted on a Tri-Carb 2800TR liquid scintillation analyzer (PerkinElmer Lifeand Analytical Sciences). Covalent binding to protein is expressed as picomoleequivalents per milligram of protein.

SDS-PAGE. For SDS-PAGE analysis, aliquots (200 �l) from covalentbinding incubations similar to those described above were quenched with100% ACN (2 ml) and microcentrifuged (14,000g for 5 min). The supernatantswere removed, and the pellets were washed by microcentrifugation three moretimes with ACN (2 ml) to remove all unbound radioactivity. The final super-natants were shown not to contain radioactivity when 0.5-ml aliquots wereanalyzed by scintillation counting as described above (data not shown). SDS-PAGE was performed using the Mini-PROTEAN 3 Cell electrophoresis sys-tem, in which approximately 50 �g of protein was loaded onto 10% Tris-HClpolyacrylamide gels (height 7.0 cm, width 8.3 cm, and thickness 1.0 mm) perwell, and electrophoresis was conducted for 1 h at 120 V. The gels werestained for protein with Coomassie Blue R-250 for 2 min and destained witha solution containing distilled water, 10% glacial acetic acid, and 20% meth-anol for 30 min, followed by treatment with distilled water (30 min). The gelswere dried using the model 3 gel dryer (1 h; Bio-Rad Laboratories, Hercules,CA) and then exposed to Fuji Film Imaging Plates (Fuji Film, Tokyo, Japan)for 1 month. Captured phosphorimages of radiolabeled proteins were analyzedusing an FLA-5100 Bio-Imaging Analyzer and MultiGauge 3.0 software (FujiFilm).

Metabolite Profiling in Human Liver Microsomes. 2MI (1 �M) wasincubated with HLM (1 mg/ml) in phosphate buffer (0.1 M, pH 7.4) in thepresence of NADPH (1 mM) for 30 min. After incubation, the reaction mixturewas quenched with 2 volumes of ACN followed by microcentrifugation(4000g for 5 min). The supernatants were removed and evaporated to drynessunder a stream of nitrogen gas at room temperature, reconstituted with water-ACN (1:1), and then analyzed on an ABI 4000 QTrap LC-MS/MS system(Applied Biosystems, Foster City, CA). The mobile phase consisted of 0.1%formic acid in water (solvent A) and 0.1% formic acid in ACN (solvent B), anda C18 column (ACE 250 � 4.6 mm, 5 �m; MAC-MOD Analytical, Inc.,Chadds Ford, PA) was used at a 0.5 ml/min flow rate under the followinggradient conditions: 0 to 5 min 95% solvent A with a linear increase to 75%A over 65 min. Because of the low microsomal metabolic turnover, a multiplereaction monitoring (MRM)-information dependent acquisition scan was per-formed (collision energy � 40 with � 10 spread) to monitor the formation ofpotential metabolites (i.e., mono-oxidative, o-demethylated, and dihydrodiolproducts).

GSH Adduct Generation and Scale-Up. GSH-trapping experiments with2MI were conducted with HLM or RLM (1 mg/ml) and NADPH (1 mM) orhorseradish peroxidase (HRP) (1 mg/ml) and H2O2 (0.1%) in the presence ofGSH (5 mM) for 60 min at 37°C. All incubations were quenched with 1volume of ACN before centrifugation and LC-MS/MS analysis. Because of thenear complete conversion of 2MI to G1 observed in HRP, G1 was isolatedfrom large-scale (50 ml) incubations with HRP and 2MI (50 �M). After a60-min incubation, the sample was centrifuged (3000g, 10 min) and thesupernatant was transferred to a preconditioned SPE cartridge (500-mg OasisHLB; Waters, Milford, MA). The contents of the SPE cartridge were elutedserially three times, each with 25, 50, 75, and 95% ACN-water containing0.1% formic acid. SPE fractions containing G1 were evaporated to dryness andreconstituted in 1 ml of ACN-water (1:1; v/v). Purification of the GSH adductwas achieved using a semipreparative HPLC system (Alliance 2695; Waters)in line with a diode array detector (996; Waters) and fraction collector (FoxyII; Teledyne-Isco, Lincoln, NE). Separation was achieved on a C18 [LunaC18(2), 5 �M, 100Å; Phenomenex, Torrance, CA] column maintained at40°C. The mobile phase consisted of 0.1% formic acid in water (solvent A) ormethanol (solvent B) with the following linear gradient conditions: 0 to 3 min,95% A; 3 to 33 min, 95 to 5% A; 33 to 38 min, 5% A; and 40 min, 95%. Theflow rate was 4 ml/min. 2MI and pooled fractions containing G1 (retentiontime of 17.5 min) were combined and evaporated to dryness. The G1 isolate,

approximately 600 �g, was reconstituted in 165 �l of deuterated d4-methanoland transferred to a 3-mm tube for NMR analysis.

NMR Analysis. NMR spectra of 2MI and G1 were acquired on a 600-MHzspectrometer equipped with a 5-mm cryoprobe (Bruker, Newark, DE). Com-plete structural assignment was performed by analyzing the 1D 1H, 2D 1H/1Hproton-proton total correlation spectroscopy, and 2D 1H/13C heteronuclearsingle quantum coherence and HMBC data sets. In addition, the isolated GSHadduct was analyzed by high-resolution mass spectrometry using an Orbitrapmass spectrometer (Thermo Fisher Scientific, Waltham, MA).

Prediction of In Vivo Mechanism-Based Inactivation and the Magni-tude of DDIs. The potential for 2MI to cause an increase in exposure(AUCi/AUC) of a coadministered “victim” drug was determined using asteady-state model (eq. 1) (Wang et al., 2004):

AUCi

AUC�

F�G

FG�

1

fmCYP3A4

1 �kinact � Iu

kdeg � �KI � Iu�

� �1 � fmCYP3A4�

(1)

where Iu is the unbound inhibitor concentration and FG and F�G are theintestinal wall availabilities in the absence and presence of the inhibitor,respectively. For human predictions, a value of 0.000321 min1 was used asthe P450 degradation rate constant, kdeg (Fromm et al., 1996), which has beendemonstrated to yield good in vitro-in vivo correlation for known clinical DDIs(Einolf, 2007). For the rat, a value of 0.000825 min1 was used for kdeg

(Watkins et al., 1986). For clinical DDI predictions with midazolam, a value of0.89 was used for fm CYP3A4 (Chen et al., 2006a) and a value of 0.57 was usedfor FG (Thummel et al., 1996). For the rat, the interaction with the prototypicalCYP3A substrate indinavir was predicted using an fm CYP3A2 value of 1 (Chibaet al., 1995) and a value of 0.94 for FG (Lin et al., 1999). The steady-statemodel was also used to predict the decrease in CYP3A2 steady-state concen-tration (E�ss/Ess) after 2MI administration using eq. 2 (Mayhew et al., 2000):

E�ss

Ess�

kdeg

kdeg �kinact � Iu

KI � Iu(2)

where E�ss and Ess are the enzyme steady-state concentrations in the presenceand absence of the inhibitor, respectively, and Iu is the unbound inhibitorconcentration.

Bioanalysis and Pharmacokinetics of 2MI in the Rat. Pharmacokineticparameters were determined in the rat after 2MI administration (2 and 20mg/kg p.o. administered in 1% Tween-1% methylcellulose suspension).Plasma samples were quenched 3:1 with acetonitrile containing internal stan-dard, and after centrifugation the supernatant was transferred to a 96-well plateand analyzed by LC-MS/MS using electrospray ionization with MRM inpositive ionization mode on an API 4000 triple quadrupole spectrometer(Applied Biosystems). The LC system consisted of Shimadzu LC-10adVPpumps (Shimadzu Scientific Instruments, Columbia, MD) and a CTC PALautosampler. Separation of the analytes was achieved using a linear gradientwith a mobile phase flow rate of 0.5 ml/min (mobile phase A, aqueous 0.1%formic acid; and mobile phase B, ACN containing 0.1% formic acid) and aLuna C18 column (5 �M, dimensions 30 � 3 mm; Phenomenex). The initialmobile phase composition (95% A and 5% B) was held for 0.5 min, increasedlinearly over 1.5 min to 2% A and 98% B, held for an additional 1.2 min, andfinally reequilibrated at the initial conditions for 1.2 min. The mass transitionmonitored was m/z 535 3 326. Pharmacokinetic parameters were calculatedusing standard noncompartmental methods (Watson LIMS; Thermo FisherScientific).

Measurement of Unbound Fraction. The unbound fraction of 2MI inplasma and microsomes was determined by an ultracentrifugation method. Inbrief, 2MI (200–1000 ng/ml) was added to plasma or hepatic microsomes (1mg/ml) at 1000 ng/ml. After incubation for 15 min at 37°C, samples werecentrifuged for 16 h at 355,000g, and unbound concentrations of 2MI weredetermined by an LC-MS/MS assay as described above.

In Vivo Mechanism-Based Inactivation of CYP3A and Interaction withIndinavir in Male Sprague-Dawley Rats. In vivo enzyme inactivation wasdetermined by measuring hepatic microsomal CYP3A activity ex vivo in rats

8432MI DERIVATIVE CAUSES MECHANISM-BASED INACTIVATION OF CYP3A

at ASPE

T Journals on M

ay 24, 2018dm

d.aspetjournals.orgD

ownloaded from

after oral gavage administration of 2MI (0, 2, or 20 mg/kg q.d. in 1%Tween-1% methylcellulose suspension) for 3 days. ABT (20 mg/kg q.d. for 3days), a nonselective P450 mechanism-based inactivator, was used as a posi-tive control (Mugford et al., 1992; Strelevitz et al., 2006). Four hours after thefinal dose, animals were sacrificed, and hepatic microsomes were preparedusing standard methods (Fleischer and Kervina, 1974). The change in enzymesteady-state concentration (E�ss/Ess) was calculated by dividing the microsomalCYP3A activity from 2MI-treated animals by the corresponding activity mea-sured in the vehicle control group.

In a separate study, the propensity for 2MI to increase the exposure of aCYP3A2 probe substrate in vivo was determined in rats. 2MI was administeredorally (2 or 20 mg/kg q.d. in 1% Tween-1% methylcellulose suspension) for 3days. Indinavir (10 mg/kg) was selected as a prototypical CYP3A2 substrate(Chiba et al., 1997) and was coadministered orally (10 mg/kg in 0.05 M citricacid) with the final dose of 2MI or vehicle control on the last day of the study.Plasma indinavir concentrations were determined by LC-MS/MS assay usingelectrospray ionization with MRM in positive ionization mode on an API 4000triple quadrupole spectrometer. The mass transition monitored was m/z 6143421, and LC conditions were identical to those described above for 2MI.Indinavir pharmacokinetic parameters were calculated using standard noncom-partmental analysis.

Results

In Vitro Mechanism-Based Inactivation. The effect of 2MI on theactivities of several P450 isoforms evaluated in pooled HLM is summa-rized in Table 1. 2MI (10 �M, initial preincubation concentration) causeda time- and NADPH-dependent decrease in CYP3A4 activity and fol-lowed pseudo-first-order kinetics. In contrast, CYP1A2, CYP2C9, andCYP2D6 activities were not affected by 2MI, and inhibition of

CYP3A4 activity was almost completely prevented by ketoconazole.In addition, CYP3A4 inactivation was not appreciably attenuated bythe addition of GSH. In separate experiments, 2MI also did not causecompetitive inhibition of CYP3A4, CYP2C9, or CYP2D6 with IC50

values �10 �M (data not shown). Figure 2 illustrates the effect ofincreasing concentrations of 2MI (0.1–100 �M) on the time-depen-dent decrease in CYP3A activity in HLM (Fig. 2A) and RLM (Fig.2C). Kinetic parameters were determined by nonlinear regression ofthe kobs versus 2MI concentration plot shown in Fig. 2, B (HLM) andD (RLM). In HLM, 2MI was a potent time-dependent inactivator witha calculated KI of 18 �M and kinact of 0.16 min1. In RLM, 2MI wasless potent (KI � 32 �M), but its maximal inactivation rate (kinact �0.83 min1) was 5-fold higher than that measured in HLM. With theuse of recombinant human CYP3A4, the partition ratio (molar ratio of2MI metabolized per CYP3A4 inactivated) was estimated to be ap-proximately 5.4 (Fig. 3). The effect of dialysis on the inhibition ofCYP3A4 activity in HLM after a 30-min preincubation with 2MI,raloxifene, or TAO is shown in Fig. 4. Dialysis was unable to restoreCYP3A4 activity in incubations containing 2MI, and approximately60% inhibition was observed both before and after dialysis. Ralox-ifene, which inactivates CYP3A4 through irreversible covalent bind-ing (Chen et al., 2002), elicited CYP3A4 inhibition that was also notrecovered after dialysis. In contrast, �85% activity was recovered inincubations containing TAO, which forms a reversible metabolite-intermediate complex with CYP3A4 (Pessayre et al., 1981). In sepa-rate experiments, 2MI was not able to decrease carbon monoxidebinding to reduced P450 (recombinant CYP3A4) nor was there anyevidence for metabolite-intermediate complex formation (i.e., charac-teristic absorbance at 455 nm) (data not shown).

Covalent Binding of [35S]2MI to Microsomal Proteins and SDS-PAGE Analysis. Table 2 summarizes the total radioactive covalentbinding of [35S]2MI or [14C]naphthalene to microsomal protein. InRLM, 2MI exhibited NADPH-dependent covalent binding to protein(156 � 18 pmol/mg protein) that was minimally attenuated by GSH(15% reduction). The total NADPH-dependent covalent binding toprotein in incubations with HLM was 79.4 � 2.2 pmol/mg protein(50% of that observed in the rat) and was modestly attenuated byGSH (33% reduction) (Table 2).

The protein specificity of the irreversible covalent modification wasinvestigated by SDS-PAGE autoradiography analysis after incubation

FIG. 2. Time- and concentration-dependent inactivation of CYP3Aactivity by 2MI in HLM (A and B) and RLM (C and D). KI and kinact

were determined from nonlinear regression of the kobs versus timeplot (B and D). In human, KI � 18 �M, kinact � 0.16 min1; in rat,KI � 32 �M, kinact � 0.83 min1. LN, natural logarithm.

TABLE 1

Effect of 2MI on P450 activities in HLM

Percentage of activity remaining relative to vehicle control after a 15-min preincubationwith 2MI (10 �M) expressed as the mean of two determinations

Isozyme (Treatment) Substrate Remaining Activity

%

CYP3A4 (NADPH) Testosterone 94.6CYP3A4 (�NADPH) Testosterone 41.7CYP3A4 (�NADPH, � 1 �M ketoconazole) Testosterone 87.9CYP3A4 (�NADPH, � 5 mM GSH) Testosterone 52.2CYP2D6 (�NADPH) Bufuralol 91.9CYP2C9 (�NADPH) Diclofenac 77.4CYP1A2 (�NADPH) Phenacetin 92.3

844 WONG ET AL.

at ASPE

T Journals on M

ay 24, 2018dm

d.aspetjournals.orgD

ownloaded from

of [35S]2MI with RLM and HLM (Fig. 5A). In the presence ofNADPH, a band of radioactivity representing a 55- to 60-kDa proteinwas observed, which is consistent with irreversible binding to one ormore P450 isoforms (He et al., 1999). Incubations of [35S]2MI withrecombinant CYP3A4 yielded a similar band of radioactivity repre-senting a 55- to 60-kDa protein (Fig. 5B). A weak radioactive bandrepresenting binding to low molecular weight proteins (�21 kDa) was

also observed in the presence and absence of NADPH, and no furthereffort was directed to characterizing this band.

In Vitro Metabolism of 2MI and GSH Adduct Formation. Invitro turnover of 2MI in HLM in the presence of NADPH was low,with �90% parent compound remaining after 30 min. Two hydroxy-lated metabolites were detected via MRM-information dependentacquisition scan (data not shown). MS/MS fragmentation patternssuggested that the indole aromatic ring and the 3-methoxyphenyl ringwere the sites of oxidation. No evidence for dihydrodiol formationwas observed. When 2MI was incubated in RLM or HLM in thepresence of NADPH and GSH, two GSH adducts (MH� 840) desig-nated as G1 and G2, were formed, and the collisionally inducedproduct ion spectra are shown in Fig. 6. Incubations with HRP (1mg/ml) in the presence of H2O2 and GSH also yielded two GSHadducts with fragmentation patterns and retention times identical tothose observed in HLM and RLM. Therefore, it was concluded thatG1 and G2 were formed in RLM, HLM, and HRP, and because of thenear-complete conversion of 2MI to G1 by HRP, this system was usedfor scale-up and structural characterization.

Structural Characterization of G1 and G2 Adducts. G1. Afull-scan MS spectrum obtained on the Q-trap instrument revealed aprotonated ion at m/z 840, indicating a net addition of 305 Da(�GSH 2H) to 2MI. MS/MS analysis (Fig. 6A) of the m/z 840 iongave rise to ions characteristic of GSH adduction: m/z 711, neutralloss of pyroglutamic acid (129 Da); and m/z 765, neutral loss ofglycine (75 Da) (Baillie and Davis, 1993). High-resolution MS andMSn data were also obtained on an Orbitrap mass spectrometer for G1and were identical to the nominal-resolution data. High-resolutiondata for an ion at m/z 177 [observed m/z 177.04780 (C9H9N2

32S),calculated m/z 177.0481, 1.79 ppm error] provided specific evi-dence for the site of adduction of the glutathionyl sulfur to a carbonon the indole ring. Additional high-resolution MS/MS fragments andtheir assignments are detailed in Table 3.

1D proton and 2D 1H/13C HMBC NMR spectra of purified G1 areshown in Fig. 7, A and B, respectively. NMR data confirmed thatGSH adduction had occurred at the 3-position (labeled C7 in Fig. 7A)of the indole moiety. The G1 NMR spectrum, compared with that ofthe parent 2MI, revealed that all protons from the parent molecule,with the exception of H7 at 6.41 ppm, were intact. In addition, GSHresonances from each of its nonexchangeable protons were also ob-served (Fig. 7A). The G1 rotating-frame Overhauser effect spectros-copy spectrum (data not shown) revealed a weak cross-peak from themethyl (H10 at 2.53 ppm) to one of the methylene protons (Hc� at3.29 ppm) on the GSH cysteine (data not shown). The HMBC spec-trum from G1 (Fig. 7B) displayed correlations from Hc� (3.09 and3.29 ppm) and H10 (2.53 ppm) to C7 (at 98.6 ppm). Relevant 1H and13C NMR data for 2MI and G1 are listed in Table 4.

G2. A full-scan MS spectrum obtained on the Q-trap instrumentalso revealed a protonated ion at m/z 840, indicating a net addition of305 Da (�GSH 2H) to 2MI. LC-MS/MS analysis (Fig. 6B) gave

FIG. 3. Partition ratio for 2MI-mediated inactivation of CYP3A4 showing loss ofCYP3A4 activity as a function of the ratio of 2MI to CYP3A4. Each valuerepresents the mean � S.D. from four determinations. The partition ratio wascalculated to be 5.4.

FIG. 4. CYP3A4 activity after dialysis (f) compared with activity before dialysis(u) after incubation with 2MI, raloxifene, or troleandomycin. Each bar representsthe mean percentage of control CYP3A4 activity � S.D. from three determinations.

TABLE 2

Total radioactive covalent binding to microsomal protein after incubation with [35S]2MI (10 �M) or [14C]naphthalene

Data are means � S.D. from three determinations.

TreatmentCovalent Binding to Protein

Species NADPH �NADPHGSH �NADPH�GSH

pmol/mg protein

�35S 2MI Human 20.2 � 3.1 79.4 � 2.2 53.1 � 3.1�35S 2MI Rat 18.6 � 3.6 156 � 18 132 � 7.5�14C Naphthalene Human 9.7 � 0.82 157 � 12 13.8 � 2.5�14C Naphthalene Rat 9.9 � 0.84 298 � 39 20.5 � 1.7

8452MI DERIVATIVE CAUSES MECHANISM-BASED INACTIVATION OF CYP3A

at ASPE

T Journals on M

ay 24, 2018dm

d.aspetjournals.orgD

ownloaded from

rise to fragments identical to those obtained from the G1 adduct. Thepresence of the m/z 177 ion, as observed in the G1 high-resolutionMS/MS spectrum, indicated the site of GSH adduction was on theindole ring of G2. In addition, the similarity of G1 and G2 MSfragments favors the site of glutathione adduction to be on the aro-matic ring instead of the 2-methylindole methyl carbon. However, nofurther attempt was made to isolate and characterize G2 because of itslow abundance in all conditions tested.

Assessment of Drug-Drug Interaction Potential. Effect of 2MI onCYP3A2 steady-state concentration in rats. Figure 8 illustrates theeffect of 2MI or ABT on ex vivo microsomal CYP3A2 activity. 2MI(20 mg/kg for 3 days) elicited a 62.4% decrease in CYP3A2 activity,which corresponds to a decrease in enzyme steady-state concentration(E�ss/Ess) of 0.376.

At the lower dose, a minimal decrease in CYP3A activity (E�ss/Ess � 0.8) was observed. The nonselective P450 inhibitor ABT wasused as a positive control and elicited an approximately 80% decreasein CYP3A2 activity (E�ss/Ess � 0.18).

Predictions using the steady-state model and comparison withobserved values. Table 5 summarizes the predicted changes in enzyme

activity (E�ss/Ess) and the exposure (AUCi/AUC) of a prototypicalCYP3A2 substrate, indinavir, using a steady-state model and themeasured kinetic parameters (KI and kinact) for 2MI in the rat. Theexperimentally observed values for E�ss/Ess (determined from ex vivomicrosomal CYP3A2 activity) and AUCi/AUC (determined by mea-suring indinavir AUC in the absence and presence of 2MI) are alsosummarized in Table 5. The in vitro-in vivo correlation was comparedusing three different values for the inhibitor concentration (Iu) for theprediction: the maximal systemic concentration (Cmax), the 24-h av-erage concentration (Cave), and the concentration at the time of liverharvesting (4 h postdose, Cterminal). These values were determinedafter a single 2MI dose of 2 or 20 mg/kg and were corrected for themeasured unbound fraction in plasma (fu) in rats (0.005). Because ofthe high gut availability of indinavir (FG � 0.94) (Lin et al., 1999), theeffect of intestinal CYP3A inhibition mediated by 2MI (F�G) wasassumed to be minimal (i.e., F�G/FG � 1) when the change in AUCwas predicted. At the higher dose group, a 65% decrease in CYP3A2activity (E�ss/Ess � 0.35) and a 2.65-fold increase in the AUC ofindinavir was predicted when the unbound Cmax (60.8 nM) was usedas the value for Iu. These predicted values (using unbound Cmax)

FIG. 5. A, SDS-PAGE analysis of covalent binding of [35S]2MI inRLM and HLM (1 mg/ml). B, SDS-PAGE analysis of covalentbinding of [35S]2MI to recombinant human CYP3A4 (33 and 330�g/ml).

FIG. 6. A, LC-MS/MS spectrum of the GSHconjugate G1 (MH� ion at m/z 840.11615) withthe assignments of characteristic fragment ions.B, LC-MS/MS spectrum of GSH conjugate G2(MH� ion at m/z 840). amu, atomic mass units.

846 WONG ET AL.

at ASPE

T Journals on M

ay 24, 2018dm

d.aspetjournals.orgD

ownloaded from

correlated well with the experimentally observed in vivo values of0.376 for E�ss/Ess and 3.62 for AUCi/AUC in the rat. In contrast, bothunbound Cterminal (1.77 nM) and Cave (2.71 nM) underpredicted thechange in CYP3A2 level (E�ss/Ess � �0.92) and increase in indinavirexposure (AUCi/AUC � 1.0). At the lower dose group (2 mg/kg),unbound Cmax, Cave, and Cterminal values were less than 2 nM, and aminimal change in enzyme steady-state concentration or indinavirexposure was predicted, in agreement with the observed in vivovalues.

Predicted Human DDI. The potential for 2MI to cause a DDIwith the CYP3A4 substrate midazolam in humans was estimatedusing the in vitro kinetic parameters for KI and kinact and thesteady-state model (eq. 2). Compared with indinavir, midazolamhas lower gut availability (FG � 0.57) (Thummel et al., 1996), andthus inhibition of intestinal CYP3A4 could have a significantaffect on the predicted change in AUC. Therefore, the predictedAUCi/AUC was calculated with and without an intestinal inhibi-tion component. 2MI was predicted to cause a 4.3-fold increase inthe AUC of midazolam when inhibition of intestinal CYP3A4 (F�G

and FG � 0.57, F�G/FG � 1) was ignored and a 7.5-fold increase

FIG. 7. G1 NMR. A, 1D proton spectrum. B, 2D 1H/13C HMBCspectrum. solv., solvent.

TABLE 3

High-resolution LC-MS/MS fragmentation pattern of G1

N

NO

COOH

S

Cl

H

H

S

NH

H2N

COOH

O

HN

HOOC O

H

OMeCl

OO

765

711502

631

358177

358 314631 587-CO2

-CO2

�M � H � m/z 840.11615 (observed); 840.11735 (calculated); C34 H36O12N535Cl2

32S2; difference 1.2 mDa (1.43 ppm)MS2 on m/z 840.1: 765.08344 (C32H31O10N4

35Cl232S2), 711.07318

(C29H29O9N435Cl2

32S2), 631.19287 (C28H33O10N532S), 587.20321

(C27H33O8N532S), 502.15032 (C23H26O7N4

32S)MS3 on m/z 631.1: 587.20272 (C27H33O8N5

32S), 358.09721 (C18H18N2O432S),

314.10761 (C17H18N2O232S)

MS4 on m/z 358.0: 177.04783 (C9H9N232S)

8472MI DERIVATIVE CAUSES MECHANISM-BASED INACTIVATION OF CYP3A

at ASPE

T Journals on M

ay 24, 2018dm

d.aspetjournals.orgD

ownloaded from

when complete intestinal CYP3A4 inhibition (F�G/FG � 1/0.57)was assumed.

Discussion

2MI was a selective, potent, and efficient mechanism-based inac-tivator of CYP3A4. Inhibition of CYP3A4 activity was time- andNADPH-dependent, followed pseudo-first order kinetics, and wasirreversible after dialysis. GSH was unable to markedly attenuateCYP3A4 inhibition, and the low partition ratio (5.4) suggested thatthe reactive species involved in MBI may not appreciably leave theP450 active site. No spectrophotometric evidence for metabolite-intermediate formation or decreases in carbon monoxide binding toreduced P450 were observed, indicating that binding or interactionwith the P450 active site did not involve perturbations of the hememoiety. The irreversible enzyme inactivation was accompanied byNADPH-dependent covalent binding of [35S]2MI to microsomalprotein, and SDS-PAGE analysis revealed that this binding wasassociated with a 55- to 60-kDa protein, consistent with irrevers-ible adduction to CYP3A4 (He et al., 1999). Taken together, thesedata indicate that 2MI is bioactivated to a reactive intermediate thatcovalently binds to the CYP3A active site, leading to irreversiblemechanism-based inactivation.

Recognizing that P450-mediated bioactivation and subsequent MBIcan lead to potentially adverse DDIs, subsequent efforts were focusedon characterizing the reactive intermediate(s) formed during oxidativemetabolism of 2MI. Two GSH adducts were detected, with identicalMS/MS fragmentation patterns. The major adduct, G1, was deter-mined unambiguously by NMR to comprise GSH linked, via athioether bond, on the 3-position of the 2-methylindole ring. G1 wasproposed to be formed via two possible mechanisms: A) through a2,3-epoxy-2-methylindole intermediate (I1) or B) through a 2-meth-ylindole diimine intermediate (I5), which is susceptible to Michaeladdition of GSH. Through pathway A, GSH preferentially attacks theless sterically hindered 3-position of the epoxide (I1) and forms G1predominantly after loss of a water molecule (Scheme 1, pathway A).Through pathway B, two successive 1-electron abstractions from 2MIand a loss of H� yield a conjugated diimine (I5), which is alsosusceptible to GSH adduction at the 3-position of the indole moiety(Scheme 1, pathway B). After rearrangement of the epoxide (I1), bothproposed pathways A and B may proceed through the diimine (I5).

A literature precedent for the proposed pathway A has been dem-onstrated for zomepirac and tolmetin. These 2-alkylpyrrole deriva-tives are bioactivated via an analogous mechanism involving anepoxide intermediate, and a GSH adduct has been identified with theglutathionyl moiety on the 3-position of the pyrrole ring (Chen et al.,2006b). With respect to pathway B, a mechanism involving twosequential 1-electron abstractions is consistent with the previouslyreported HRP-mediated bioactivation of zafirlukast (Kassahun et al.,2005). The efficient conversion of 2MI to G1 by HRP in the presentstudy implies that this mechanism (B), proceeding through thediimine, may predominate for the peroxidase enzyme system. Incontrast, for the P450-mediated bioactivation of 2MI, the relativecontribution of the proposed pathway A or B is not clear and is furthercomplicated by the potential interconversion of the diimine and ep-oxide. It is interesting to note that a diimine intermediate has beenpreviously reported for the peroxidative metabolism of benzidine,followed by subsequent GSH and DNA adduct formation (Yamazoeet al., 1988; Lakshmi et al., 1994). In addition, unlike 2MI, bioacti-vation of 3-methylindole (3MI) proceeds through P450-mediated de-hydrogenation to a highly reactive 3-methyleneindolenine intermedi-ate capable of mechanism-based P450 inactivation and proteinalkylation (Skiles and Yost, 1996; Skordos et al., 1998; Yan et al.,

FIG. 8. Ex vivo CYP3A2 activity in hepatic microsomes prepared from rats aftertreatment with 2MI (2 or 20 mg/kg), ABT (20 mg/kg), or vehicle for 3 days q.d.Each bar represents the mean � S.D. from three determinations.

TABLE 4

Summary of proton and carbon chemical shifts from affected portions for 2MI and G1

The labeling scheme is provided in Fig. 7A. NMR data were acquired in CD3OD; the solvent signal appeared at �H 3.32 ppm/�C 47.9 ppm. Numbers in parenthesis are J couplings in Hertz.

Label2MI G1

1H 13C 1H 13C

ppm ppm

1 7.07 (d, 8.7) 110.1 7.13 (d, 8.7) 110.62 6.42 (d, 8.7) 111.2 6.48 (d, 8.7) 112.13 144.7 146.34 133.8 133.15 129.0 127.06 116.1 117.17 6.41 (s) 98.8 98.68 136.9 144.0

10 2.44 (s) 12.1 2.52 (s) 10.6a� 3.72, 3.77 (AB multiplet) 41.5b� 4.46 (dd, 3.87, 9.6) 53.5c� 3.09 (dd, 9.6, 13.5) 3.29 (buried peak) 38.3d� 3.63 (dd, 6.3) 54.2e� 2.12 (m) 26.3f� 2.52, 2.60 (m) 31.9

J coupling notations: s, singlet; d, doublet; dd, doublet of a doublet; m, multiplet.

848 WONG ET AL.

at ASPE

T Journals on M

ay 24, 2018dm

d.aspetjournals.orgD

ownloaded from

2007; Kartha and Yost, 2008). Both dehydrogenation and oxidativepathways participate in 3MI bioactivation, but the predominant mech-anism is P450-isoform dependent (Yost, 2001). In the present study,formation of an analogous reactive species for 2MI [i.e., a 2-methyl-eneindolenine, or imine methide (I7)] (Scheme 1, pathway C) was notimplied, because there was no evidence for GSH adduction on the2-methyl carbon of the 2-methylindole ring. In contrast to the bioac-tivation of 3MI, the lack of evidence for the formation of a 2-meth-ylene-indolenine suggests that dehydrogenation does not play animportant role in the bioactivation of 2MI. To the best of our knowl-edge, the present study is the first report of 2-methylindole bioacti-vation leading to MBI. Although more mechanistic studies are required,preliminary evidence suggests that a 2,3-epoxy-2-methylindole and/or adiimine is the potential reactive intermediate responsible for covalent adduc-tion to the CYP3A active site.

Risk assessment for the likelihood of an in vivo DDI has becomeroutinely implemented in the pharmaceutical industry when furtheradvancement of a compound with an identified MBI liability is

considered. One commonly implemented strategy uses a steady-stateapproach (Mayhew et al., 2000; Wang et al., 2004), which requires thefollowing four key input parameters: 1) innate enzyme degradationrate constant (kdeg), 2) inhibitor potency (KI), 3) maximal inactivationrate (kinact), and 4) unbound concentration (Iu) of the inhibitor. Thismodel has been shown retrospectively to be an effective tool for theprediction of in vivo DDI in humans (Galetin et al., 2006; Obach etal., 2007). It has also been successfully implemented preclinically topredict a 3-fold increase in the exposure of indinavir caused by MBIof CYP3A2 in the rat elicited by a novel melanocortin-4 receptoragonist (Tang et al., 2008). Although acceptable correlations havebeen established between the predicted and observed AUCi/AUC fora victim drug using the steady-state model, in vitro-in vivo correlationfor the prediction of the change in enzyme steady-state concentration(E�ss/Ess) is not routinely evaluated, due, in part, to the complexity ofassessing in vivo enzyme levels.

In the present study, the steady-state model was used to predict bothAUCi/AUC and E�ss/Ess in the Sprague-Dawley rat. E�ss/Ess was

SCHEME 1. Proposed mechanism of CYP3A4-mediated bioactivation of 2MI.

TABLE 5

Comparison of the predicted change in CYP3A2 enzyme steady-state concentration (E�ss/Ess) and indinavir exposure (AUCi/AUC) with that observed in vivo in ratsusing different parameters for the inhibitor concentration (Iu) of 2MI

Dose Parameter Used for Iu Unbound ConcentrationE�ss/Ess AUCi/AUC

Predicted Observed Predicted Observed

nM

2 mg/kg Cmax 2.32 � 1.5 0.93 � 0.04 1.07 � 0.043Cave 0.15 � 0.06 1.00 � 0.002 0.795 � 0.04 1.00 � 0.0017 1.19 � 0.90Cterminal 0.24 � 0.12 0.99 � 0.004 1.01 � 0.0034

20 mg/kg Cmax 60.8 � 8.1 0.35 � 0.029 2.65 � 0.20Cave 2.71 � 0.36 0.92 � 0.010 0.376 � 0.026 1.08 � 0.011 3.62 � 0.84Cterminal 1.77 � 0.39 0.95 � 0.011 1.12 � 0.011

8492MI DERIVATIVE CAUSES MECHANISM-BASED INACTIVATION OF CYP3A

at ASPE

T Journals on M

ay 24, 2018dm

d.aspetjournals.orgD

ownloaded from

determined by comparing ex vivo CYP3A2 activity in hepatic micro-somes prepared from rats pretreated with 2MI with that in ratsreceiving vehicle control. An excellent correlation was observed(within 10%) for the change in E�ss/Ess when unbound systemic Cmax

was used as the input Iu concentration, which is consistent with reportsindicating that the maximal systemic unbound concentration yieldsgood prediction of AUCi/AUC (Obach et al., 2007). The model alsoaccurately predicted a 3-fold increase in indinavir exposure, furtheremphasizing its relevance to in vivo MBI and DDIs mediated by 2MI.The excellent in vitro-in vivo correlation increased the confidence forusing the steady-state model for risk assessment of the likelihood of aclinically relevant DDI. Subsequently, at a therapeutically relevant2MI concentration (225 nM), a 4.3- to 7.5-fold increase in the AUCof a midazolam was predicted. Because of the unfavorable risk for thepotential of an in vivo DDI, strengthened by the excellent in vitro-invivo correlation observed in the rat, further progression of 2-methyl-indole-containing compounds was de-prioritized, prompting investi-gation of alternative chemical series as dual CRTH2/DP antagonists.

In summary, the 2-methylindole-containing compound 2MI wasidentified as a potent mechanism-based inactivator of CYP3A thatbinds covalently to the CYP3A apoprotein. A GSH adduct linked tothe 3-position of the 2-methylindole moiety was fully characterizedvia MS and NMR. The putative reactive intermediate(s) were inferredto be a 2,3-epoxide and/or a conjugated diimine. The steady-statemodel was shown to accurately predict the interaction between 2MIand indinavir in vivo in the rat, and the molecule was predicted tocause a DDI at clinically relevant doses.

Acknowledgments. We acknowledge Craig Uyeda, Jade Koide,and Hoa Le for assistance with in vivo experiments, bioanalysis, anddetermination of pharmacokinetics. In addition, we thank Dr. MarkGrillo for assistance in GSH adduct characterization and manuscriptreview.

References

Baillie TA and Davis MR (1993) Mass spectrometry in the analysis of glutathione conjugates.Biol Mass Spectrom 22:319–325.

Chen M, Nafziger AN, and Bertino JS Jr (2006a) Drug-metabolizing enzyme inhibition byketoconazole does not reduce interindividual variability of CYP3A activity as measured byoral midazolam. Drug Metab Dispos 34:2079–2082.

Chen Q, Doss GA, Tung EC, Liu W, Tang YS, Braun MP, Didolkar V, Strauss JR, Wang RW,Stearns RA, et al. (2006b) Evidence for the bioactivation of zomepirac and tolmetin by anoxidative pathway: identification of glutathione adducts in vitro in human liver microsomesand in vivo in rats. Drug Metab Dispos 34:145–151.

Chen Q, Ngui JS, Doss GA, Wang RW, Cai X, DiNinno FP, Blizzard TA, Hammond ML,Stearns RA, Evans DC, et al. (2002) Cytochrome P450 3A4-mediated bioactivation ofraloxifene: irreversible enzyme inhibition and thiol adduct formation. Chem Res Toxicol15:907–914.

Chiba M, Hensleigh M, and Lin JH (1997) Hepatic and intestinal metabolism of indinavir, anHIV protease inhibitor, in rat and human microsomes. Major role of CYP3A. BiochemPharmacol 53:1187–1195.

Chiba M, Nishime JA, and Lin JH (1995) Potent and selective inactivation of human livermicrosomal cytochrome P-450 isoforms by L-754,394, an investigational human immunedeficiency virus protease inhibitor. J Pharmacol Exp Ther 275:1527–1534.

Day SH, Mao A, White R, Schulz-Utermoehl T, Miller R, and Beconi MG (2005) A semi-automated method for measuring the potential for protein covalent binding in drug discovery.J Pharmacol Toxicol Methods 52:278–285.

Einolf HJ (2007) Comparison of different approaches to predict metabolic drug-drug interactions.Xenobiotica 37:1257–1294.

Fleischer S and Kervina M (1974) Subcellular fractionation of rat liver. Methods Enzymol31:6–41.

Fromm MF, Busse D, Kroemer HK, and Eichelbaum M (1996) Differential induction ofprehepatic and hepatic metabolism of verapamil by rifampin. Hepatology 24:796–801.

Galetin A, Burt H, Gibbons L, and Houston JB (2006) Prediction of time-dependent CYP3A4

drug-drug interactions: impact of enzyme degradation, parallel elimination pathways, andintestinal inhibition. Drug Metab Dispos 34:166–175.

He K, Woolf TF, and Hollenberg PF (1999) Mechanism-based inactivation of cytochromeP-450–3A4 by mifepristone (RU486). J Pharmacol Exp Ther 288:791–797.

Hollenberg PF (2002) Characteristics and common properties of inhibitors, inducers, andactivators of CYP enzymes. Drug Metab Rev 34:17–35.

Kalgutkar AS, Obach RS, and Maurer TS (2007) Mechanism-based inactivation of cytochromeP450 enzymes: chemical mechanisms, structure-activity relationships and relationship toclinical drug-drug interactions and idiosyncratic adverse drug reactions. Curr Drug Metab8:407–447.

Kartha JS and Yost GS (2008) Mechanism-based inactivation of lung-selective cytochrome P450CYP2F enzymes. Drug Metab Dispos 36:155–162.

Kassahun K, Skordos K, McIntosh I, Slaughter D, Doss GA, Baillie TA, and Yost GS (2005)Zafirlukast metabolism by cytochrome P450 3A4 produces an electrophilic �,�-unsaturatediminium species that results in the selective mechanism-based inactivation of the enzyme.Chem Res Toxicol 18:1427–1437.

Lakshmi VM, Zenser TV, and Davis BB (1994) Mechanism of 3-(glutathion-S-yl)-benzidineformation. Toxicol Appl Pharmacol 125:256–263.

Lin JH, Chiba M, Chen IW, Nishime JA, deLuna FA, Yamazaki M, and Lin YJ (1999) Effect ofdexamethasone on the intestinal first-pass metabolism of indinavir in rats: evidence ofcytochrome P-450 3A [correction of P-450 A] and p-glycoprotein induction. Drug MetabDispos 27:1187–1193.

Mayhew BS, Jones DR, and Hall SD (2000) An in vitro model for predicting in vivo inhibitionof cytochrome P450 3A4 by metabolic intermediate complex formation. Drug Metab Dispos28:1031–1037.

Mugford CA, Mortillo M, Mico BA, and Tarloff JB (1992) 1-Aminobenzotriazole-induceddestruction of hepatic and renal cytochromes P450 in male Sprague-Dawley rats. Fundam ApplToxicol 19:43–49.

Obach RS, Walsky RL, and Venkatakrishnan K (2007) Mechanism-based inactivation of humancytochrome P450 enzymes and the prediction of drug-drug interactions. Drug Metab Dispos35:246–255.

Pessayre D, Descatoire V, Konstantinova-Mitcheva M, Wandscheer JC, Cobert B, Level R,Benhamou PJ, Jaouen M, and Mansuy D (1981) Self-induction by triacetyloleandomycin of itsown transformation into a metabolite forming a stable 456 nm-absorbing complex withcytochrome P-450. Biochem Pharmacol 30:553–558.

Pettipher R (2008) The roles of the prostaglandin D2 receptors DP1 and CRTH2 in promotingallergic responses. Br J Pharmacol 153 (Suppl 1):S191–S199.

Riley RJ, Grime K, and Weaver R (2007) Time-dependent CYP inhibition. Expert Opin DrugMetab Toxicol 3:51–66.

Silverman RB (1995) Mechanism-based enzyme inactivators. Methods Enzymol 249:240–283.Skiles GL and Yost GS (1996) Mechanistic studies on the cytochrome P450-catalyzed dehydro-

genation of 3-methylindole. Chem Res Toxicol 9:291–297.Skordos KW, Skiles GL, Laycock JD, Lanza DL, and Yost GS (1998) Evidence supporting the

formation of 2,3-epoxy-3-methylindoline: a reactive intermediate of the pneumotoxin 3-meth-ylindole. Chem Res Toxicol 11:741–749.

Strelevitz TJ, Foti RS, and Fisher MB (2006) In vivo use of the P450 inactivator 1-aminoben-zotriazole in the rat: varied dosing route to elucidate gut and liver contributions to first-passand systemic clearance. J Pharm Sci 95:1334–1341.

Tang W, Stearns RA, Wang RW, Miller RR, Chen Q, Ngui J, Bakshi RK, Nargund RP, Dean DC,and Baillie TA (2008) Assessing and minimizing time-dependent inhibition of cytochromeP450 3A in drug discovery: a case study with melanocortin-4 receptor agonists. Xenobiotica38:1437–1451.

Thummel KE, O’Shea D, Paine MF, Shen DD, Kunze KL, Perkins JD, and Wilkinson GR (1996)Oral first-pass elimination of midazolam involves both gastrointestinal and hepatic CYP3A-mediated metabolism. Clin Pharmacol Ther 59:491–502.

Waley SG (1985) Kinetics of suicide substrates. Practical procedures for determining parameters.Biochem J 227:843–849.

Walsky RL and Obach RS (2004) Validated assays for human cytochrome P450 activities. DrugMetab Dispos 32:647–660.

Wang YH, Jones DR, and Hall SD (2004) Prediction of cytochrome P450 3A inhibition byverapamil enantiomers and their metabolites. Drug Metab Dispos 32:259–266.

Watkins PB, Wrighton SA, Schuetz EG, Maurel P, and Guzelian PS (1986) Macrolide antibioticsinhibit the degradation of the glucocorticoid-responsive cytochrome P-450p in rat hepatocytesin vivo and in primary monolayer culture. J Biol Chem 261:6264–6271.

Yamazoe Y, Zenser TV, Miller DW, and Kadlubar FF (1988) Mechanism of formation andstructural characterization of DNA adducts derived from peroxidative activation of benzidine.Carcinogenesis 9:1635–1641.

Yan Z, Easterwood LM, Maher N, Torres R, Huebert N, and Yost GS (2007) Metabolism andbioactivation of 3-methylindole by human liver microsomes. Chem Res Toxicol 20:140–148.

Yost GS (2001) Bioactivation of toxicants by cytochrome p450-mediated dehydrogenationmechanisms. Adv Exp Med Biol 500:53–62.

Zhou S, Yung Chan S, Cher Goh B, Chan E, Duan W, Huang M, and McLeod HL (2005)Mechanism-based inhibition of cytochrome P450 3A4 by therapeutic drugs. Clin Pharmaco-kinet 44:279–304.

Address correspondence to: Dr. Simon G. Wong, 1120 Veterans Blvd., SouthSan Francisco, CA 94080. E-mail: simon.wong@amgen.com

850 WONG ET AL.

at ASPE

T Journals on M

ay 24, 2018dm

d.aspetjournals.orgD

ownloaded from