Abstract
A strategy is proposed to profile compounds for mechanism-based inactivation of CYP3A4, CYP2C19, CYP2C9, CYP2D6, and CYP1A2 based on an apparent partition ratio screen. Potent positives from the screen are confirmed by time- and concentration-dependent inactivation assays. Quasi-irreversible inhibitions are then differentiated from irreversible inactivations by oxidation with potassium ferricyanide and/or dialysis. The three-step screening procedure has been validated with acceptable accuracy and precision for detection and confirmation of mechanism-based inactivators in drug discovery. We report here the apparent partition ratios for 19 mechanism-based inactivators and four quasi-irreversible inhibitors obtained under the same experimental conditions. The apparent partition ratio screen was automated to provide throughput for determining structure-mechanism-based inactivation relationships. Information about reversibility can be used to assess potential toxicity mediated by covalent adducts, as well as the potential for pharmacokinetic drug-drug interactions. Direct comparison of known mechanism-based inactivators and quasi-irreversible inhibitors, based on our screening of apparent partition ratios, has identified ritonavir, mibefradil, and azamulin as highly effective mechanism-based inactivators; e.g., 1 mol of CYP3A4 was inactivated on turnover of about 2 mol of compound. Other mechanism-based inactivators we identified include bergamottin (CYP1A2 besides previously reported CYP3A4), troglitazone (CYP3A4), rosiglitazone (CYP3A4), and pioglitazone (CYP3A4). Comparison of the apparent partition ratios and inactivation clearance data for the three glitazones suggests that the chromane moiety on troglitazone contributes to its greater potency for mechanism-based inactivation.
Reports of adverse, sometimes fatal, events in patients on multiple drug therapies are increasing. A significant number of those events are due to elevated concentrations of drugs in plasma, resulting from pharmacokinetic drug-drug interactions (DDI) that occur because cytochrome P450 (P450)-mediated drug metabolism is inhibited. Reports of fatal DDI were responsible for the recent withdrawal of terfenadine (Monahan et al., 1990), astemizole (Desager and Horsmans, 1995), cisapride (Wysowski and Bacsanyi, 1996), mibefradil (Krayenbuhl et al., 1999), and cerivastatin (Sica and Gehr, 2002) from the market. To avoid costly failures in drug development, pharmacokinetic DDI from P450 inhibition is now being investigated earlier in the drug discovery process across the pharmaceutical industry.
Most adverse pharmacokinetic DDIs are attributable to reversible inhibition of P450 isozymes. The potential for a discovery compound to be subjected to reversible inhibition is typically investigated by in vitro studies to determine IC50 values using recombinant human P450 isozymes (Crespi et al., 1997) or pooled human liver microsomes (HLMs) (Dierks et al., 2001). That approach weeds out new chemical entities that are potent inhibitors. The IC50 values of new chemical entities from the same chemical series can be used to establish structure-inhibition relationships. Those data, together with a computational model of a specific P450 isozyme, can be used to design out the inhibition liability.
Adverse pharmacokinetic DDI can also result from mechanism-based inactivation of P450 isozymes, illustrated by some fatal adverse events observed with mibefradil (Posicor) (Krayenbuhl et al., 1999), a potent mechanism-based inactivator (MBI) of CYP3A4 (Prueksaritanont et al., 1999). Irreversible inactivation generally involves metabolism of the inactivator to a reactive metabolite, which covalently modifies the P450 isozyme and results in loss of activity (Silverman, 1996). Enzymatic activity can only be restored through de novo protein synthesis; therefore, inhibitions are prolonged in vivo even after elimination of the inactivator. Hence, failure to take into consideration mechanism-based inactivation of P450 isozymes can lead to gross underestimation of the potential extent of pharmacokinetic DDI (Bjornsson et al., 2003). Mechanism-based inactivation of P450 has also been associated with idiosyncratic toxicity (Tucker et al., 2001), a consequence that is unpredictable because we cannot identify individuals at risk. Therefore, our strategy is to design out structural features responsible for mechanism-based inactivation in drug discovery using a method of screening described herein.
Traditionally, screening for mechanism-based inactivation of a P450 isozyme relies on detection of changes in enzymatic activity with and without preincubation of the test compound for a defined period of time. A shift of IC50 to a lower value with preincubation indicates mechanism-based inactivation (Favreau et al., 1999); conversely, a higher IC50 with preincubation indicates that the test compound is a substrate of the enzyme (Favreau et al., 1999). Measurement of IC50 shift is often conducted by incubating a recombinant human P450 isozyme with a substrate that generates a fluorescent metabolite for high throughput screening. However, fluorescent assays are susceptible to interference by fluorescent test compounds or their metabolites. Furthermore, the effect of metabolites generated by one isozyme on other isozymes cannot be tested in a recombinant single-enzyme system. This difficulty is best illustrated by amiodarone, which is an MBI of CYP3A4 but its N-deethylated metabolite inactivates CYP1A2, CYP2B6, and CYP2D6 as well (Ohyama et al., 2000). Therefore, a definitive assay must include pooled HLM that contain a complete array of P450 isozymes.
This report describes the development, validation, and automation of a method for initial profiling of new chemical entities for mechanism-based inactivation of CYP3A4, CYP2C9, CYP2C19, CYP2D6, and CYP1A2. This method is based on measuring apparent partition ratios incorporating pooled, mixed gender HLM, Food and Drug Administration-recommended marker substrates, and fast analysis by liquid chromatography/tandem mass spectrometry (LC/MS/MS). Potent positive MBIs from the apparent partition ratio (APR) screen are confirmed for time- and concentration-dependent P450 inactivations. Reversibility of inactivation is then evaluated by oxidation with potassium ferricyanide and/or dialysis. Application of this strategy has led to our identification of troglitazone, rosiglitazone, and pioglitazone as MBIs of CYP3A4.
Materials and Methods
Materials. 1′-Hydroxymidazolam, 4-hydroxymidazolam, 6β-hydroxytestosterone, (±)-bufuralol hydrochloride salt, (±)-1′-hydroxybufuralol maleate salt, 4′-hydroxydiclofenac, S-(+)-mephenytoin, (±)-4′-hydroxymephenytoin, 6′,7′-dihydroxybergamottin, and bergamottin were purchased from Ultrafine (Manchester, UK). Midazolam maleate salt, diclofenac sodium salt, dextromethorphan hydrobromide monohydrate salt, testosterone, quinidine, troleandomycin, sulfaphenazole, furafylline, α-naphthoflavone, ticlopidine, 17α-ethynylestradiol, mifepristone, glucose 6-phosphate, NADP+, glucose-6-phosphate dehydrogenase, potassium ferricyanide, raloxifene hydrochloride, diltiazem hydrochloride, nicardipine, (±)-verapamil hydrochloride, resveratrol, isoniazid, suprofen, tacrine, and carbamazepine were obtained from Sigma (St. Louis, MO). 7-Ethoxyresorufin, resorufin sodium salt, and ketoconazole were obtained from MP Biomedicals (Aurora, OH). Paroxetine, troglitazone, rosiglitazone maleate, and pioglitazone hydrochloride were obtained from Toronto Research Chemicals, Inc. (North York, ON, Canada). 1-[(2-Ethyl-4-methyl-1H-imidazol-5-yl)methyl]-4-[4-(trifluoromethyl)-2-pyridinyl]piperazine (EMTPP) was from Ryan Scientific, Inc. (Mt. Pleasant, SC). Tienilic acid and mibefradil were obtained from the compound room of Aventis (Bridgewater, NJ). Pooled, mixed gender HLMs were obtained from Xenotech LLC (Lenexa, KS). Azamulin, S-(+)-N-3-benzyl-nirvanol and Supersomes were obtained from BD Biosciences Discovery Labware (Bedford, MA). Single 96-well DispoDIALYZER (10,000 mol. wt. cut-off) was purchased from Harvard Apparatus Inc. (Holliston, MA). Reagents and materials for Western immunoblotting were obtained from Invitrogen (Carlsbad, CA). Aquasil C18 columns and other chromatographic supplies were obtained from Thermo Electron Corporation (Waltham, MA). Fisher Scientific Co. (Fairlawn, NJ) was the supplier of other reagent grade chemicals, HPLC grade solvents, and 96-well plates.
Characterization of Microsomes. The Km value of each P450 isozyme for each batch of microsomes was characterized as published previously (Dierks et al., 2001). The content of individual P450 isozymes in microsomes was quantified by densitometry after separation by SDS-polyacrylamide gel electrophoresis, and identified on Western immunoblots using standard protocols recommended by the vendors.
APR Screens. Incubations in the presence or absence of an NADPH-regenerating system were conducted in 2-ml 96-well plates containing individual test compounds and a known MBI as a positive control. The primary incubations, in total volumes of 200 μl per well, consisted of 2 mg/ml HLM, 0 to 50 μM test compound, 10 mM MgCl2, 2 mM EDTA, 100 mM potassium phosphate buffer, pH 7.4, 1 mM NADP+, 10 mM glucose 6-phosphate, and 2 U/ml glucose-6-phosphate dehydrogenase. The organic solvent was typically dimethyl sulfoxide unless otherwise indicated, at a final concentration of 0.5% (v/v). The reactions were initiated by addition of the NADPH-regenerating system and incubated for 1 h at 37°C to ensure complete inactivation. Residual catalytic activities were assayed by transferring 10-μl aliquots (20 μl for CYP1A2) to preincubated secondary assay plates, which were identical to primary plates except that they contained a saturated concentration (5× Km) of marker substrate (Table 1) instead of test compound.
Incubations were allowed to proceed for another 20 min and then quenched with 100 μl of acetonitrile containing 2.5 pg/μl dextromethorphan as an internal standard. Microsomal proteins were pelleted by centrifugation at 1968g for 20 min before transfer of the supernatants to new 2-ml 96-well plates for LC/MS/MS analysis. Samples were pooled for analysis of testosterone 6β-hydroxylase and midazolam 1′-hydroxylase activities. Mean values from duplicate analyses were used to calculate the percentage of activity remaining, based on values from respective control incubations with solvent alone as 100%. The APRs were then calculated by plotting percent activity remaining as a function of the molar ratio of test compound to P450 isozyme (Silverman, 1996).
Automation of APR Screen. The APR screen was automated using a Genesis Workstation 200 (Tecan, Research Triangle Park, NC) retrofitted with a Te-MO 96-multichannel pipette head. Automation was validated using mifepristone for CYP3A4 isozyme, with testosterone as the marker substrate. The validation run was conducted using three separate 96-well plates of primary incubation, to simulate screening of five compounds and one positive control for each P450 isozyme analyzed. The throughput was six compounds per P450 isozyme, in duplicate, per day. LC/MS/MS analysis, data processing and calculation required an additional day.
Time- and Concentration-Dependent Inactivation Assays. All incubations were performed in triplicate in 2-ml 96-well plates with appropriate MBIs as positive controls. The conditions and procedures for the primary and secondary incubations were as described for the screening procedure. Concentration ranges for compounds tested in the primary incubation were established from the results of the APR screen. Typically, the selected concentrations corresponded to the molar ratio of test compound to P450 isozyme that resulted in inactivation greater than 30%. Residual catalytic P450 isozyme activities were assayed at 0, 7.5, 15, and 30 min. Mean values from the triplicate analyses were used to calculate the percentage of activity remaining, using as 100% the respective control incubation with solvent at time 0.
Reversibility of Inactivation. The reversibility of inactivation of P450 isozyme was investigated by oxidation with potassium ferricyanide and also by dialysis, at single concentrations (n = 8) that resulted in more than 50% inactivation. Troleandomycin and mifepristone were included as positive and negative controls respectively; an appropriate quasi-irreversible inhibitor and an MBI for whichever P450 isozyme is under investigation may replace them.
For oxidation with potassium ferricyanide, the primary incubates were diluted 2-fold with 50 μl of either 2 mM potassium ferricyanide or 0.1 M potassium phosphate buffer, pH 7.4. Each mixture was incubated for 15 min at 37°C, before another 20-fold dilution (10-fold for CYP1A2) for a secondary incubation designed to assess residual catalytic P450 isozyme activities.
For dialysis, the primary incubates were diluted 10-fold by transferring 20-μl aliquots to 180 μl of 0.1 M potassium phosphate buffer, pH 7.4, in each well of a 96-well DispoDIALYZER. Dialysis was conducted with 4 liters of 0.1 M potassium phosphate buffer, pH 7.4. The buffer was changed after 3 h, and dialysis was continued for an additional 21 h. Catalytic P450 isozyme activities remaining after dialysis were assayed after 2-fold dilution by transferring 100-μl aliquots to secondary incubations. The protein concentration in each dialysis well was determined in a 96-well plate format using a BCA protein assay kit from Pierce Chemical (Rockford, IL). The primary and secondary incubations were conducted as described in the APR screen. Both reversibility experiments were conducted on the same day, and aliquots of primary incubates were removed for assessment of residual P450 isozyme activities without either oxidation by potassium ferricyanide or by dialysis.
Mean values were used to calculate the percentage of activity remaining, using the solvent control as 100% activity. The reversibility of P450 isozyme inactivation was determined by comparing the enzyme activity of samples before and after oxidation by potassium ferricyanide or dialysis. The enzyme activities of dialyzed samples were normalized to protein concentrations.
LC/MS/MS Analysis. All samples were analyzed by LC/MS/MS in the positive-ion electrospray ionization mode. The LC/MS system consisted of a Gilson 306 HPLC equipped with a Gilson 215 autosampler coupled to a Sciex API III+ (Applied Biosystems, Foster City, CA) or a Surveyor HPLC and an autosampler coupled to a Quantum AM (Thermo Electron Corporation).
All chromatographic separations were carried out on an Aquasil C18 column (50 × 2 mm i.d., 5 μm) thermostated at 50°C, using a linear 1.35-min gradient from 95:5 to 5:95 (water/acetonitrile with 0.05% formic acid) for analysis of samples from all marker substrates, except for testosterone and midazolam, where the starting gradient consisted of 19% acetonitrile. The initial organic content was adjusted for optimum separation of multiple hydroxy metabolites for each marker substrate. A 20-μl aliquot of sample was typically injected into the column and eluted using a programmed flow rate from 1 to 1.5 ml/min over a 1.35-min gradient. The total cycle time for each analysis was 3 min.
The liquid chromatography eluant was split to about 250 μl/min before being sprayed into the mass spectrometer at 4.5 kV. Nebulization of the liquid chromatography eluant on the API III+ was achieved using nebulizer and auxiliary gases (nitrogen) set at 1.1 l/min and 8000 cc/min, respectively. The sheath and auxiliary gases were set at 35 and 10 arbitrary units, respectively, on the Quantum AM. Desolvation of the solvent droplets on the API III+ was aided by heating the nebulizer to 500°C and the interface to 60°C, with a declustering potential of 35 V and curtain gas set at 1.8 l/min. Desolvation on the Quantum AM relied on a heated capillary set at 350°C.
Metabolites and internal standard were detected by multiple reaction monitoring of precursor-product ion pairs (Table 1) at a speed of 2 scans/s with Q2 settling to reduce cross-talk for API III+, or 0.1 s/scan on the Quantum AM. Precursor ions were selected using Q1 at a resolution of 1.0 Da at half-height for collision-induced dissociation in Q2, and the product ions were mass analyzed with Q3 at unit resolution. The collision gas (argon) was set at 250 × 1013 atoms/cm2 for the API III+ or 1.5 mTorr on the Quantum AM. The metabolite and the internal standard were monitored using the same collision energy on the API III+; optimum collision energy for each ion pair was used on the Quantum AM. Product ions were detected with the electron multiplier set at saturation (4100 V) on the API III+ or 888 V with a gain of 2.7 × 105 on the Quantum AM. Data acquisition and reduction for the API III+ were achieved using RAD 2.6 and MacQuan 1.3, whereas Xcalibur 1.3 was used on the Quantum AM.
Data Analysis. Data corresponding to percentage of activity remaining in the presence or absence of a NADPH-regenerating system were plotted as a function of the molar ratio of test compound to P450 isozyme, using Sigma Plot (SPSS Inc., Chicago, IL). Curve fitting was carried out using single exponential decay with r2 of at least 0.9. The APR was calculated by a linear regression analysis of the linear-decay portion of the curve and the intercept, with the x-axis corresponding to the APR + 1 (Silverman, 1996). The molar ratio of test compound to P450 isozyme that resulted in maximum loss of activity was the lowest molar ratio on the linear regression line that corresponded to the saturated concentrations of test compound.
Apparent inactivation kinetic constants were determined as described previously (Kitz and Wilson, 1962). Briefly, the natural logarithm of percentage of activity remaining was plotted against the preincubation time for each concentration of test compound investigated. The slope from the linear regression analysis gave the observed inactivation rate constant (kobs) for each concentration. A double-reciprocal plot (also known as the Kitz-Wilson plot) of kobs (y-axis) and the inactivator concentrations (x-axis) produced the apparent kinact (minute-1) from the reciprocal of the y-axis intercept and KI (micromolar) from the negative reciprocal of the x-axis intercept.
Results
Mechanism-Based Inactivation by APR Screen. APRs were obtained for 19 known MBIs (Table 2). The most potent MBIs detected by this screen were ritonavir, mibefradil, and azamulin for CYP3A4, with APRs ranging from 1 to 2 depending on the marker substrate. Other potent compounds (APRs less than 50) inactivated CYP3A4 (raloxifene, bergamottin, and 6′,7″-dihydroxybergamottin), CYP2C9 (tienilic acid), CYP2D6 (paroxetine), and CYP1A2 (furafylline). Competitive inhibitors and substrates of each of the P450 isozymes consistently tested negative by this screen, showing little or no separation of the plus- and minus-NADPH titration curves (data not shown).
Figure 1, A to L, depict titration curves for five MBIs and five competitive inhibitors of P450 isozymes in the presence and absence of NADPH. Separation of the plus- and minus-NADPH titration curves occurred for known MBIs such as mifepristone (CYP3A4), tienilic acid (CYP2C9), ticlopidine (CYP2C19), paroxetine (CYP2D6), and furafylline (CYP1A2). Plus-NADPH titration curves were associated with residual functional isozyme activity after oxidative metabolism. Minus-NADPH titration curves represented competitive inhibition by residual test compounds. MBIs showed lower residual functional activities in the plus-NADPH titration curves than in the minus-NADPH titration curves. Conversely, there was little or no separation of the plus- and minus-NADPH titration curves for competitive inhibitors such as ketoconazole (CYP3A4), sulfaphenazole (CYP2C9), S-(+)-N-3-benzyl-nirvanol (CYP2C19), quinidine (CYP2D6), and α-naphthoflavone (CYP1A2). Any separation of the plus- and minus-NADPH titration curves observed for competitive inhibitors was usually associated with the plus-NADPH titration curves being only slightly higher than those without NADPH.
Figure 1, A to L, also provide information on the APR, the residual functional activity, the maximum loss of activity, and molar ratio at which maximum loss of activity occurred. For example, the APR (residual functional activity) obtained for mifepristone, using testosterone 6β-hydroxylation and midazolam 1′-hydroxylation, was 5.7 (6%) and 7.2 (14%), respectively. The maximum losses of CYP3A4 activity were 94 and 86% as measured by testosterone 6β-hydroxylation and midazolam 1′-hydroxylation, respectively, at a preincubation concentration of 10 μM mifepristone (mifepristone/CYP3A4 ratio of 38). This loss of CYP3A4 activity was comparable with that previously reported using a reconstituted system (He et al., 1999). Losses of CYP3A4 activity were due to inactivation as a result of the higher residual CYP3A4 activity observed at each concentration in the absence of NADPH (Fig. 1, A and C, open circles). Competitive inhibition contributed to about 11 and 30% of CYP3A4 activity measured by testosterone 6β-hydroxylation and midazolam 1′-hydroxylation, respectively, at the highest preincubation concentration of 50 μM. This corresponds to a residual concentration of 2.5 μM, assuming no metabolism. The APRs obtained for ticlopidine, tienilic acid, paroxetine, and furafylline were 70.3, 13.2, 6.6, and 28.5, respectively. The maximum losses of enzyme activity for ticlopidine, tienilic acid, paroxetine, and furafylline were 85, 89, 73, and 72% at molar ratios of compound to P450 isozyme of 156, 61, 16, and 142, respectively.
The APR screen also detected quasi-irreversible inhibitors, including troleandomycin, diltiazem, nicardipine and verapamil (Table 2), which formed a nitrosoalkane metabolite-intermediate (MI) complex with the enzyme. Troleandomycin was the most potent quasi-irreversible inhibitor investigated, with APRs of 4.5 and 4.0 using testosterone and midazolam, respectively, as marker substrates.
Accuracy and Precision of the APR Screen. Data concerning intra- and interday precision and accuracy for determination of each MBI by the APR screen are presented in Table 3. Intraday precision ranged from 6.7 to 20.3% coefficient of variation (CV) for MBIs of the five major P450 isozymes; interday CVs ranged from 1.8 to 22.4%. Measured APRs of 12.6, 78.1, and 35.3 for tienilic acid, ticlopidine, and furafylline were, respectively, 1.1-, 3.0-, and 5.6-fold greater than previously reported values [tienilic acid: CYP2C9, r = 12.0 (Lopez-Garcia et al., 1994); ticlopidine: CYP2C19, r = 26.0 (Ha-Duong et al., 2001); and furafylline: CYP1A2, r = 5.6 (Kunze and Trager, 1993)].
Automation. The precision obtained for the automated APR screen using 96-multichannels Te-MO was 7.8% CV. The APR for mifepristone from the automated screen was 6.4 ± 0.5 compared with an APR of 6.0 ± 0.4 obtained by the manual method.
Time- and Concentration-Dependent Inactivation. Kitz-Wilson plots in Figure SI1 showed that inactivations of CYP3A4, CYP2C9, CYP2C19, CYP2D6, and CYP1A2 by mifepristone, tienilic acid, ticlopidine, paroxetine, and furafylline were dependent on time and concentration, as indicated by a linear relationship with an r2 of at least 0.9. Competitive inhibitors did not display time- and concentration-dependent P450 isozyme inactivation (data not shown).
Intraday precision for kinact ranged from 3.1 to 31.4 and KI ranged from 8.4 to 44.4% CV (Table SI1). Interday precision was slightly higher for kinact (31.0 to 57.2% CV), but the precision from interday determination of KI (7.2–45.3% CV) was comparable with that of intraday. The intraday precision obtained for kinact and KI of each MBI was comparable with the corresponding precision values reported previously for CYP2C9 (Lopez-Garcia et al., 1994), CYP2C19 (Ha-Duong et al., 2001), CYP2D6 (Bertelsen et al., 2003), and CYP1A2 (Clarke et al., 1994). The differences between our observed mean values of kinact and KI and those previously reported ranged from –4.5- to +23.7 and from –16.0- to +2.0-fold.
Reversibility of P450 Isozyme Inactivation. Data from assessment of reversibility of P450 isozyme inactivation by oxidation with potassium ferricyanide and by dialysis are tabulated in Table SI2. Only CYP3A4, when inactivated by either troleandomycin or verapamil, was reversible by oxidation with potassium ferricyanide. However, neither CYP2D6 inactivated by paroxetine nor CYP3A4 inactivated by mifepristone or mibefradil was reversible by oxidation with potassium ferricyanide. The intra- and interday reproducibility of the experiments for assessing reversibility of P450 isozyme inactivation by oxidation with potassium ferricyanide had a CV of <11%.
Activity of CYP3A4 inactivated by azamulin, mibefradil or mifepristone was not completely recovered by dialysis. The irreversibility of azamulin-inactivated CYP3A4 was consistent with mechanism-based inactivation as reported previously (Stresser et al., 2004). However, the activity of CYP3A4 inactivated by troleandomycin, verapamil, or nicardipine was completely restored by dialysis. Interestingly, only about 40% of bergamottin-inactivated CYP3A4 activity was restored by dialysis, and partial reversibility after dialysis was also observed for paroxetine-inactivated CYP2D6, as indicated by restoration of about 42 and 21% activity of enzyme that had been inactivated by 0.625 and 20 μM paroxetine, respectively. The intra- and interday precision of dialysis in 96-well format was <12% CV. Agreement of the current data with those previously reported suggested that dialysis in 96-well plates was just as effective as using single dialysis cartridges. In addition to positive and negative controls, one can use restoration of enzyme activity in samples lacking NADPH to check for effectiveness of dialysis.
Screening of Unknowns. Data from profiling troglitazone, rosiglitazone, and pioglitazone for mechanism-based inactivation of CYP3A4 by the APR screen are presented in Fig. 2. The separation of the titration curves of each compound, with and without NADPH, using either of two marker substrates of CYP3A4 suggested that the inactivation of CYP3A4 by troglitazone, rosiglitazone, and pioglitazone was mechanism-based, with APRs (residual functional isozyme activity) of 45.7 (4%, testosterone) and 51.0 (5%, midazolam) for troglitazone, 185.2 (12%, testosterone) and 226.5 (21%, midazolam) for rosiglitazone, and 277.3 (28%, testosterone) and 407.5 (36%, midazolam) for pioglitazone. Hence, troglitazone was 4.1-fold (testosterone) and 4.4-fold (midazolam) more potent than rosiglitazone, and 6.1-fold (testosterone) and 8.0-fold (midazolam) more potent than pioglitazone. Rosiglitazone was 1.5-fold (testosterone) and 1.8-fold (midazolam) more potent than pioglitazone.
Time and concentration dependence of inactivation of CYP3A4 by troglitazone, rosiglitazone, and pioglitazone was confirmed by the Kitz-Wilson plots shown in Fig. 3. The inactivation kinetic constants were kinact = 0.0335 min-1 and KI = 5.0 μM for troglitazone, kinact = 0.0195 min-1 and KI = 11.9 μM for rosiglitazone, and kinact = 0.0112 min-1 and KI = 10.4 μM for pioglitazone. Inactivation clearance data indicated that troglitazone was 4.2-fold more potent than rosiglitazone and 6.1-fold more potent than pioglitazone; rosiglitazone was 1.5-fold more potent than pioglitazone. Overnight dialysis did not restore the activity of CYP3A4 inactivated by these compounds (Table 4), evidence that inactivation of CYP3A4 by any of the three glitazones was irreversible.
Discussion
Development and Validation of an APR Screen. The titration method commonly used to determine partition ratios (Silverman, 1996) was recently applied to human P450 isozyme (Lin et al., 2002). We report here for the first time the development of an APR screen by a titration method using HLMs.
Our APR screen for MBIs was validated by correct identification of 19 known MBIs, with no false positive results. The detection of MBIs with approximately 7000-fold differences in APRs (Table 2) indicates the wide dynamic range of the screen. This wide range may require retesting of some compounds at different concentration ranges depending on their inactivation potency, and, if present, any competitive inhibition potency as well. This issue is illustrated by our identification of ritonavir as an MBI of CYP3A4 from the APR screen at a concentration only 10-fold lower than the default concentration range (Figure SI2). The sensitivity of the screen was underscored by the detection of weak MBIs with kinact/KI < 1 min-1mM-1, such as isoniazid for CYP3A4, CYP2C19, and CYP1A2 (Wen et al., 2002), and diclofenac for CYP3A4 (Masubuchi et al., 2002).
Intra- and interday precision of <25% CV achieved for APRs of five major P450 isozymes (Table 3) are acceptable for ranking compounds in discovery. We have no explanation for lower intraday precision for all the isozymes except CYP1A2. Better precision might be obtained by replacing current internal standard for bioanalysis with corresponding stable isotope-labeled metabolite of each marker substrate (Walsky and Obach, 2004). The accuracy of the method is difficult to demonstrate because no APR using HLMs has been reported to date. Despite experimental differences, reasonably good agreement is indicated by less than 6-fold differences from reported intrinsic partition ratios of tienilic acid, ticlopidine, furafylline, 17α-ethynylestradiol, EMTPP, and suprofen. However, agreement is closer if the intrinsic partition ratios of 17α-ethynylestradiol (Lin et al., 2002), EMTPP (Hutzler et al., 2004), and suprofen (O'Donnell et al., 2003) are estimated from intercepts with the x-axis as originally defined (Silverman, 1996), and applied to other enzymes (Pochet et al., 2000). The reduction in potency after moving from a recombinant to a microsomal system could be explained by a different degree of protein binding and/or reduction of substrate due to metabolism by other P450 isozymes.
All CYP3A4 MBIs (except isoniazid) we detected using testosterone and midazolam had the same potency ranking and differences of less than 2-fold in their APRs. Interestingly, the APRs and residual CYP3A4 activities measured using midazolam were consistently higher than the results with testosterone, suggesting that the contribution of midazolam to inactivation is negligible. The small difference in the APRs obtained with either marker substrate suggests that covalent modification of testosterone- and midazolam-binding sites of CYP3A4 by each inactivator is comparable. However, loss of enzyme activity associated with covalent modification is greater for the testosterone-binding site, as indicated by lower % residual CYP3A4 activity (Table 2).
Inactivation Kinetic Constants. To assess the potential magnitude of pharmacokinetic DDI by potent MBIs with APRs less than 50, inactivation kinetic constants must be determined before a compound can be brought into development. The challenge for measuring kinact and KI is to select a concentration range for the experiment (Silverman, 1996); that choice is made easier by data from the APR screen.
In general, higher variability is associated with this assay, an observation consistent with at least one literature report (Voorman et al., 1998). The precision of the assay might be improved by using a stable isotope-labeled metabolite of each marker substrate because of similarity of mass spectrometric and chromatographic behaviors (Watson, 1997).
Reversibility of P450 Inactivation. In addition to data for time- and concentration-dependent inactivation, reversibility of inactivation must be assessed for confirming any potent inactivator indicated by the APR screen. MBIs carry a risk of idiosyncratic toxicity (Tucker et al., 2001) in addition to the pharmacokinetic DDI liability observed with competitive or quasi-irreversible inhibitors. Hence, an ability to differentiate an MBI from a quasi-irreversible inhibitor will allow better decision-making in nominating a compound for development. The reversibility of the nitrosoalkane-MI P450 complex by either oxidation with potassium ferricyanide (Franklin, 1991; Ma et al., 2000) or dialysis (Ma et al., 2000), suggests that an MBI can be differentiated from this type of quasi-irreversible inhibitor. Hence, we have validated the reversibility of P450 isozyme inactivation in our experiments by oxidation with potassium ferricyanide and dialysis, with an acceptable precision of <12% CV (Table SI2). Our data indicate that oxidation with potassium ferricyanide and dialysis is equally effective in dissociating nitrosoalkane-MI complexes, consistent with that reported elsewhere for nitrosoalkane-MI complexes (Ma et al., 2000). Alternatively, nitrosoalkane-MI complexes can be detected by analysis of substrate-induced difference spectra (Franklin, 1991).
Chemical structures responsible for quasi-irreversible P450 inhibition have included dialkylamino (Bensoussan et al., 1995) and 3,4-methylenedioxyphenyl (Murray, 2000) functionalities. Because the dialkylamino functionality is often used to improve the druggable property of a compound, a compound with such functionality must be checked for potential to cause quasi-irreversible inhibition if we want to design out this liability. For example, inactivation of CYP3A4 by mifepristone has been associated with formation of nitrosoalkane-MI complexes through oxidation of the dimethylamino moiety (Jang and Benet, 1998) or through covalent binding to apoCYP3A4 by the reactive ketene metabolite produced from oxidation of pro-1-ynyl moiety (He et al., 1999). Our data indicate that inactivation of CYP3A4 by mifepristone is not reversible (Table SI2), confirming the involvement of the prop-1-ynyl moiety as reported previously (He et al., 1999). This correct identification of functionality on a compound is necessary for designing out mechanism-based inactivation liability during lead optimization.
3,4-Methylenedioxyphenyl compounds, via a carbene-MI complex or through covalent binding by ortho quinone, can inactivate P450 isozymes. In our experiments, the carbene-MI CYP2D6 complex resulting from paroxetine (Bertelsen et al., 2003) was not reversible by oxidation with potassium ferricyanide, despite a report that the carbene-MI CYP3A4 complex resulting from dimethyl-4,4′-dimethoxy-5,6,5′,6′-dimethylene-dioxybiphenyl-2,2′-dicarboxylate was completely reversible (Kim et al., 2001). However, the concentration-dependent partial reversibility of paroxetine-inactivated CYP2D6 by dialysis we have reported here is consistent with that observed for DDP-inactivated CYP3A4 (Kim et al., 2001). Whether inactivation of CYP2D6 activity by 20 μM paroxetine is reversible by dialysis is unclear, however, because of a comparable gain in activity for the dialyzed sample in the absence of NADPH. Therefore, identification of carbene-MI complex in paroxetine-inactivated CYP2D6 has to rely on substrate-induced difference in spectral analysis (Bertelsen et al., 2003) or on electrospray ionization mass spectrometry, for detection of covalent binding by ortho quinone (Hutzler et al., 2004).
Screening of Unknowns. Compounds presenting a structural alert for reactive metabolites and/or formation of MI complexes are candidates for screening for potential mechanism-based inactivation. For example, the three glitazones (troglitazone, rosiglitazone and pioglitazone) are able to generate reactive metabolites from oxidation of chromane and/or 2,4-thiazolidinedione moieties (Kassahun et al., 2001).
In our experiments, troglitazone was a potent MBI for CYP3A4, with an APR of about 50. Inactivation was dependent on time and concentration and was not reversible by dialysis. However, it is not clear which functionality on troglitazone is responsible for inactivation. Our data on APRs and inactivation clearance numbers indicate that all three glitazones are MBIs of CYP3A4, and their order of potency for inactivation is troglitazone > rosiglitazone > pioglitazone. Structurally, the three glitazones share a 2,4-thiazolidinedione functionality. Reactive metabolites from bioactivation of 2,4-thiazolidinedione moiety can inactivate CYP3A4. However, troglitazone is the only one containing a chromane moiety; instead, rosiglitazone has a dialkylamino-pyridine and pioglitazone has a dialkylpyridine group. Formation of quinone methide from chromane might contribute to the greater potency of troglitazone for inactivating CYP3A4. The less effective formation of covalent adducts in CYP3A4 by rosiglitazone and pioglitazone, combined with the much lower doses generally prescribed (less than 10 mg/day; Uetrecht, 2002) may explain the lacking of idiosyncratic hepatotoxicity and pharmacokinetic DDI of those drugs, compared with troglitazone, in clinical settings. This example clearly illustrates that the APR screen has adequate resolution for identifying problematic functionalities for lead optimization.
In conclusion, we have developed and validated an automated screen for determining APRs to detect mechanism-based inactivation of CYP3A4, CYP2C19, CYP2C9, CYP2D6, and CYP1A2. The APR screen has the resolution and sensitivity for detecting MBIs and quasi-irreversible inhibitors. Potent inactivators identified in APR screens are then confirmed by time- and concentration-dependent inactivation experiments. Differentiation of mechanism-based inactivation from quasi-irreversible inhibition involving the nitrosoalkane-MI complex can be achieved by oxidation with potassium ferricyanide or by dialysis. A strategy based on structural alerts for detecting MBI successfully identified troglitazone, rosiglitazone, and pioglitazone as CYP3A4 MBIs.
Acknowledgments
We thank Ruth Foltz and Dr. Rodger Foltz (University of Utah, Salt Lake City, UT), and Dawn Baumgardner and Kent Stevens (Johnson & Johnson Pharmaceutical Research & Development, Raritan, NJ) for editing the manuscript.
Footnotes
-
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
-
doi:10.1124/dmd.104.003475.
-
ABBREVIATIONS: DDI, drug-drug interaction; P450, cytochrome P450; HLM, human liver microsome; MBI, mechanism-based inactivator; LC/MS/MS, liquid chromatography/tandem mass spectrometry; APR, apparent partition ratio; EMTPP, 1-[(2-ethyl-4-methyl-1H-imidazol-5-yl)methyl]-4-[4-(trifluoromethyl)-2-pyridinyl]piperazine; HPLC, high-performance liquid chromatography; MI, metabolite-intermediate.
-
↵ The online version of this article (available at http://dmd.aspetjournals.org) contains supplemental material.
- Received December 23, 2004.
- Accepted April 27, 2005.
- The American Society for Pharmacology and Experimental Therapeutics