Abstract
The utility of chemical inhibitors in cytochrome P450 (CYP) reaction phenotyping is highly dependent on their selectivity and potency for their target CYP isoforms. In the present study, 17 inhibitors of CYP1A2, 2B6, 2C8, 2C9, 2C19, 2D6, and 3A4/5 commonly used in reaction phenotyping were evaluated for their cross-enzyme selectivity in pooled human liver microsomes. The data were evaluated using a statistical desirability analysis to identify (1) inhibitors of superior selectivity for reaction phenotyping and (2) optimal concentrations for each. Among the inhibitors evaluated, α-naphthoflavone, furafylline, sulfaphenazole, tienilic acid, N-benzylnirvanol, and quinidine were most selective, such that their respective target enzymes were inhibited by ∼95% without inhibiting any other CYP enzyme by more than 10%. Other commonly employed inhibitors, such as ketoconazole and montelukast, among others, were of insufficient selectivity to yield a concentration that could adequately inhibit their target enzymes without affecting other CYP enzymes. To overcome these shortcomings, an experimental design was developed wherein dose response data from a densely sampled multi-concentration inhibition curve are analyzed by a six-parameter inhibition curve function, allowing accounting of the inhibition of off-target CYP isoforms inhibition and more reliable determination of maximum targeted enzyme inhibition. The approach was exemplified using rosiglitazone N-demethylation, catalyzed by both CYP2C8 and 3A4, and was able to discern the off-target inhibition by ketoconazole and montelukast from the inhibition of the targeted enzyme. This methodology yields more accurate estimates of CYP contributions in reaction phenotyping.
SIGNIFICANCE STATEMENT Isoform-selective chemical inhibitors are important tools for identifying and quantifying enzyme contributions as part of a cytochrome P450 (CYP) reaction phenotyping assessment for projecting drug-drug interactions. However, currently employed practices fail to adequately compensate for shortcomings in inhibitor selectivity and the resulting confounding impact on estimates of the CYP enzyme contribution to drug clearance. In this report, we describe a detailed half maximal inhibitory concentration (IC50) study design with 6-parameter modeling approach that yields more accurate estimates of enzyme contribution.
Introduction
Interpatient variability in drug exposure is an important reason underlying variable drug response. Exposure is a function of the drug clearance rate and, if administered orally, the extent of drug absorption. Mechanisms of drug clearance include direct excretion and/or metabolism, with the latter being most prevalent. Among drug metabolizing enzymes, the cytochrome P450 (CYP) family is most frequently involved in metabolic drug clearance (Cerny, 2016). Activities of individual CYP enzymes can vary among individuals due to pharmacogenetic differences, drug interactions, and disease state, as well as other intrinsic and environmental factors. Thus, identification of specific CYP enzymes involved in metabolism and their respective quantitative contributions can offer great insight into interpatient variability in drug exposure and, hence, pharmacological response. The in vitro experiments conducted to quantitatively identify individual CYP enzymes in the metabolism of drugs have been termed “CYP reaction phenotyping” (Rodrigues, 1999; Zhang et al., 2007; Lu et al., 2003; Zientek et al., 2015). The data generated from these experiments comprise an important part of drug development required for understanding interpatient variability in exposure and drug-drug interactions.
CYP reaction phenotyping is conducted using the following experiment types, with intention of general agreement between two orthogonal methods employed: (Bjornsson et al., 2003; Bohnert et al., 2016; US FDA, 2020; EMA, 2012): (1) effect of CYP selective inhibitors or inhibitory antibodies on drug metabolism in a human-derived hepatic in vitro system; (2) measurement of the drug metabolism rates in individual CYP enzymes and scaling to reflect the whole liver; (3) comparison of drug metabolism rates to those of a set of CYP selective substrates across a panel of individual liver microsomes with specific polymorphic CYP enzyme expression (e.g., “correlation analysis”). While these methods have been employed for well over 20 years, they each suffer from some shortcomings. (1) Substrates and chemical inhibitors have imperfect selectivity for their intended CYP enzymes (Nirogi et al., 2015; Khojasteh et al., 2011; Lu et al., 2003). (2) Inhibitory antibodies fail to yield maximal inhibition (Shou et al., 2000; Polsky-Fisher et al., 2006). (3) Scaling factors needed for rate data obtained from individual enzymes can vary with the marker substrates used in their derivation (Dantonio et al., 2022; Lindmark et al., 2018; Wang et al., 2019; Siu and Lai, 2017). (4) Correlation analysis is not a quantitative method and generally requires one dominant CYP enzyme to observe a strong correlation (Ogilvie et al., 2008; Wienkers et al., 2002). (5) Genotyped reagents prepared from individuals may not reflect a global average population. (6) CYP reaction phenotyping is further confounded when the drug is metabolized slowly, requiring the measurement of metabolite as opposed to the simpler approach of measuring substrate consumption.
In our own laboratories, when using a panel of chemical inhibitors to quantitatively identify CYP enzymes involved in the metabolism of a new drug candidate, we frequently observe the sum of inhibition across the chemical inhibition panel exceeding 100%. This outcome confounds subsequent fraction metabolized (fm) calculations by overestimating minor isoforms (due to non-selectivity) and diluting contribution of major isoforms (when normalized to 100%), thereby skewing the isoform profile when the underlying limitations of inhibitor selectivity are not considered. If using standard binding theory, the inhibitor selectivity must be 1000-fold to yield 99% inhibition of a target CYP and not cause more than 10% inhibition of off-target CYPs. The inhibitors commonly employed for CYP reaction phenotyping do not possess this degree of selectivity, and the resulting spillover inhibition when using single concentrations leads to incorrect assignment of minor contributing isoforms. Furthermore, even for the most widely employed CYP inhibitors, there is a wide range of reported potency values (Fig. 1), and these can be confounded by experimental conditions, such as variance in non-specific binding with microsomal concentration and marker substrate concentration used relative to its Michaelis-Menten parameter.
Summary of Potency Values Reported in the Scientific Literature for Common Cytochrome P450 Inhibitors. Data were obtained from the University of Washington Drug Interaction Database (https://www.druginteractionsolutions.org/). IC50, Ki, and KI values were restricted to those measured in human liver microsomes and recombinant P450 enzymes. Reactions were restricted to phenacetin O-deethylase and tacrine 1'-hydroxylase, bupropion hydroxylase, amodiaquine N-deethylase and paclitaxel 6α-hydroxylase, diclofenac 4'-hydroxylase, (S)-warfarin 7-hydroxylase, and tolbutamide 4-hydroxylase, (S)-mephenytoin 4'-hydroxylase and omeprazole 5-hydroxylase, bufuralol 1'-hydroxylase and dextromethorphan O-demethylase, and midazolam 1'-hydroxylase and testosterone 6β-hydroxylase. Horizontal bars and displayed numeric values represent the median values.
To address this issue, 17 CYP-selective inhibitors were thoroughly characterized for quantitation of on- and off-target selectivity. Additionally, a multiple-concentration inhibitor study design employing a six-parameter curve-fitting approach is proposed to exclude off-target effects and lead to improved quantitation of fm. The application of combined inhibitor selectivity knowledge with a data-rich 22-point concentration experimental design is exemplified using rosiglitazone N-demethylation activity for estimation of CYP3A and CYP2C8 fm. A comparison of single- and multiple-concentration approaches highlights the pitfalls of previous practices. Adoption of this methodology and incorporation into a broader integrated, sequential two-step approach, as described in the accompanying report in this journal (Dantonio et al., 2022), should offer substantial improvement to the current CYP reaction phenotyping practices.
Materials and Methods
Materials
Experiments were conducted on human tissue acquired from a vendor that was verified as compliant with internal policies, including institutional review board/independent ethics committee approval. Pooled human liver microsomes of low CYP3A5 expression, prepared from 50 mixed-sex donors, were purchased from Sekisui XenoTech (Kansas City, KS). cDNA expressed human CYP3A4 Supersomes and untransfected baculosome control were purchased from Corning (Corning, NY).
Chemicals and solvents of high-performance liquid chromatography or analytical grade were purchased from Fisher Scientific (Waltham, MA). Monobasic and dibasic potassium phosphate buffers, MgCl2, NADPH, and DMSO were purchased from Sigma-Aldrich (St. Louis, MO). Substrates were purchased from either Sigma-Aldrich (St. Louis, MO), US Pharmacopeia (Rockville, MD), or Toronto Research Chemicals (North York, Ontario, Canada), except for rosiglitazone, which was synthesized internally at Pfizer (Groton, CT). The 17 chemical inhibitors tested (typical purity >95%) were either synthesized internally at Pfizer, Groton, CT or purchased from one of the following sources: Sigma-Aldrich (St. Louis, MO) (ex. CYP3cide), Toronto Research Chemicals (Ontario, Canada), MedChemExpress (Monmouth Junction, NJ), US Pharmacopeia (Rockville, MD), or Enzo Life Sciences (Ann Arbor, MI). Deuterated analytes used as internal standards were synthesized internally at Pfizer (Groton, CT).
The equilibrium dialysis apparatus and Spectra Por (MWCO 12–14K) cellulose dialysis membranes for microsomal binding experiments were obtained from Repligen (formerly Spectrum, Boston, MA).
Inhibitor Selectivity in CYP2B6 and CYP-Selective Probe Cocktail Assays
The effects of selective chemical inhibitors on the CYP-mediated metabolism of probe substrates were characterized using pooled HLM under verified conditions of linear product formation. Incubations (200 µl) were conducted in 100 mM of potassium phosphate buffer (pH 7.4) containing 3.3 mM of MgCl2, 1.2 mM of NADPH, 0.03 mg/ml microsomal protein, and probe substrates at 37°C in a heat block. A series of 100X concentrated inhibitor stock solutions were prepared by serial dilution for testing across a broad concentration range. For incubations containing competitive inhibitors, microsomes were prewarmed with inhibitors for 5 minutes prior to initiating the reactions by the sequential addition of NADPH followed immediately by substrate (see below). Incubations containing time dependent inhibitors (TDI) were preincubated with HLM, the inactivator, and NADPH for 10 minutes to achieve complete inactivation of the respective CYP isoform. Reactions were initiated by the addition of 10X concentrated substrate stock solution containing a cocktail of the CYP-specific probe substrates, phenacetin (6 µM, CYP1A2), amodiaquine (0.33 µM, CYP2C8), diclofenac (1.3 µM, CYP2C9), mephenytoin (7.9 µM, CYP2C19), dextromethorphan (0.36 µM, CYP2D6), and midazolam (0.42 µM, CYP3A4), prepared in a mixture of aqueous and organic solvents (acetonitrile or methanol) to achieve final incubation concentrations of approximately 1/5 Michaelis-Menten parameter based on in-house determined kinetics (Supplemental Table 1) (Walsky and Obach, 2004). The final solvent concentration in the incubation was <1%. Reactions were terminated after 14 minutes by quenching 125-µl aliquots of incubation mixture into 200 µl of acetonitrile containing internal standard. Following vortex mixing and centrifugation at 1700 x g for 5 minutes, the resulting supernatants (225 µl) were transferred to clean 96-well plates, concentrated under a stream of warm nitrogen, and reconstituted in 100 µl of 90/10 water/acetonitrile for compatibility with initial liquid chromatography with tandem mass spectrometer (LC-MS/MS) conditions.
Separate incubations for the assessment of CYP2B6 activity were conducted in a similar fashion as described above with the following modifications: the substrate stock solutions contained bupropion (18 µM) alone, reactions were terminated after 20 minutes by quenching of 100-µl aliquots with 200 µl of acetonitrile containing internal standard. All HLM selective chemical inhibition incubations were conducted in duplicate. Detailed study conditions are described in the Supplemental Table 1.
Ketoconazole and Montelukast Inhibition of Rosiglitazone Metabolism in HLM
Using similar incubation conditions described for the cocktail substrate assay above, reactions (400 µl) with 0.1 mg/ml of HLM protein and inhibitors were initiated by addition of the substrate rosiglitazone (1 µM) and quenched at 30 minutes by protein precipitation with acetonitrile containing internal standard. Incubations were conducted in triplicate.
LC-MS/MS Methodology for Quantitation of Probe Substrates and Rosiglitazone
All microsomal incubation samples were quantitated by LC-MS/MS bioanalysis conducted using a Sciex Triple Quad 6500 mass spectrometer (Framingham, MA) with an electrospray ion source, Agilent 1290 binary pump (Santa Clara, CA), and CTC Analytics autoinjector (Zwingen, Switzerland).
Cocktail Assay
The aqueous mobile phase (A) was comprised of 0.1% acetic acid in water, and the organic mobile phase (B) consisted of acetonitrile. Samples (10 µl) from the in vitro incubations were injected onto an Acquity BEH C18 (2.5 × 50 mm, 1.7 µm) (Waters, Milford, MA) column at room temperature with a flow rate of 0.5 ml/min. The gradient program began with 2% initial mobile phase B with a linear gradient to 45% B over 2.2 minutes, then to 90% over 0.2 minutes, held at 90% B for 0.5 minutes, and followed by re-equilibration to initial conditions for 0.6 minutes. The mass spectrometer was operated in multiple reaction monitoring mode (MRM) while monitoring positive ion detection with the following mass transitions (first quadrupole of triple quadrupole mass spectrophotometer [Q1]/third quadrupole of triple quadrupole mass spectrophotometer [Q3]) and collision energies (in parenthesis): 4'-hydroxymephenytoin 235/150 (26), [2H3]-4'-hydroxymephenytoin 238/150 (26), N-desmethylamodiaquine 328/283 (23), [2H5]-N-desmethylamodiaquine 333/283 (23), dextrorphan 258/201 (31), [2H3]-dextrorphan 261/201 (34), 1'-hydroxymidazolam 342/324 (30), [2H4]-1'-hydroxymidazolam 346/328 (30), acetaminophen 152/110 (22), [2H7]-acetaminophen 159/115 (22), hydroxydiclofenac 312/230 (29), and [13C6]-hydroxydiclofenac 318/236 (20). It is noted that during bioanalytical development, the removal of formic acid from all samples and mobile phases was necessary to eliminate background interferences in the 4'-hydroxymephenytoin MRM channel and monthly replacement of Mobile Phase A was required to maintain overall chromatographic performance. Additionally, extra effort was taken to ensure chromatographic separation of acetaminophen from a peak arising from in-source fragmentation of phenacetin.
CYP2B6 Assay
Chromatographic separation of the CYP2B6 incubation samples was achieved following injection (10 µl) onto a Halo C18 (2.1 × 10 mm, 2.7 µm) (Advanced Materials Technology, Wilmington, DE) column at room temperature with a flow rate of 0.5 ml/min. The aqueous mobile phase (A) was comprised of 0.1% formic in water and organic mobile phase (B) consisted of 0.1% formic acid in acetonitrile. The gradient program began with 5% initial mobile phase B with a linear gradient to 95% B over 0.9 minutes, held at 95% B for 0.3 minutes, and followed by re-equilibration to initial conditions for 0.6 minutes. The mass spectrometer was operated in MRM mode while monitoring positive ion detection and collision energy 36, following mass transitions (Q1/Q3) for hydroxybupropion 256/130, and [2H6]-hydroxybupropion 262/130.
Rosiglitazone Assay
Chromatographic analysis of N-desmethylrosiglitazone and internal standard was achieved using aqueous mobile phase (A) comprised of 0.1% formic in water and organic mobile phase (B) consisting of 0.1% formic acid in acetonitrile. Samples (10 µl) from the in vitro incubations were injected onto an Acquity BEH C18 (2.1 × 50 mm, 1.7 µm) (Waters, Milford, MA) column at room temperature with a flow rate of 0.5 ml/min. The gradient program began with 2% initial mobile phase B for 0.5 minute with a linear gradient to 35% B over 1.1 minutes, stepped immediately to 95% B after 0.01 minutes, held at 95% B for 0.3 minutes, and followed by re-equilibration to initial conditions for 0.6 minutes. The mass spectrometer was operated in MRM mode, while monitoring positive ion detection and collision energy (35), with the following mass transitions (Q1/Q3) for rosiglitazone 358/135, N-desmethylrosiglitazone 344/121, and [2H4]-N-desmethylrosiglitazone 348/125.
Microsomal Binding of CYP-Selective Chemical Inhibitors
Human liver microsomes were diluted to 0.03 or 0.1 mg/ml of protein in 100 mM of potassium phosphate buffer (pH 7.45) and spiked with 1000X concentrated inhibitor stock solution, prepared in DMSO, to achieve a final drug concentration of 1.0 µM. Aliquots (1 ml) of the prepared microsome solution were prewarmed to 37°C and loaded into one side of a Spectrum equilibrium dialysis cell prepared with a SpectraPor (Repligen, Boston, MA) 12-14 kDA cellulose membrane, prepared per manufacturer’s instructions. The microsomes were dialyzed against an equal volume of 100 mM of potassium phosphate buffer of pH 7.45. The entire apparatus was incubated for 18 hours under constant rotation while submerged in a 37°C water bath. Samples were processed using a mixed-matrix protein precipitation method and analyzed by LC-MS/MS. To each sample, an equal volume of opposing control buffer or microsome solution was added resulting in common matrix composition across all donor and receiver samples. Aliquots (0.6 ml) of the mixed matrix samples were precipitated with a 5X volume of acetonitrile and 0.1 ml of internal standard (1 µM of clozapine), vortexed, then centrifuged at 1800 x g for 5 minutes. The supernatants were evaporated to dryness under vacuum centrifuge and residues reconstituted in 0.1 ml of 20/80 acetonitrile/water containing 1% formic acid, for analysis by LC-MS/MS.
The LC-MS/MS system used for quantitation of nonspecific binding samples consisted of an Orbitrap Elite mass spectrometer with ion spray source, Accela Quanternary UHPLC pump (Thermo-Fisher; Waltham, MA), and CTC Analytics autoinjector. Aqueous mobile phase (A) was comprised of 0.1% formic acid in water and organic mobile phase (B) consisted of acetonitrile. Samples (10 µl) were injected onto a Phenomenex Kinetex C18 (2.1 × 50 mm, 1.7 µm) (Phenomenex, Torrance, CA) column at room temperature with a flow rate of 0.4 mL/min. The gradient program began with 2% initial mobile phase B with a linear gradient to 15% B over 3.5 minutes, then to 50% over 3 minutes, to 80% over 0.9 minutes, held at 95% B for 0.5 minutes, and followed by re-equilibration to initial conditions for 1 minute. Slight modifications to the bioanalytical method for analysis of gemfibrozil glucuronide binding samples included aqueous mobile phase A comprised of 10 mM of ammonium acetate, organic mobile phase B consisting of acetonitrile, and the gradient program beginning with 5% B, increased linearly to 95% B over 3.5 minutes, then held for 1 minute at 95% B, and re-equilibrated to initial conditions for 0.9 minutes.
The mass spectrometer was set to MS1 scan mode (m/z 100–900) at a resolution setting of 30,000 and operated in positive ion mode for all analytes except gemfibrozil glucuronide. Response for each analyte was determined by post-analysis construction of the extracted ion chromatograms of protonated molecular ion (deprotonated for gemfibrozil glucuronide) to a high-resolution mass accuracy of 5 ppm.
Due to challenges with nonspecific binding and/or chemical instability observed during equilibrium dialysis, the microsomal binding of troleandomycin was assessed by measuring inhibition of midazolam 1’-hydroxylase marker activity in incubations of spanning inactive protein concentrations while maintaining a consistent active protein concentration. The effect of troleandomycin nonspecific binding was assessed in incubations (150 µl) with recombinant human CYP3A4 (0.01 mg/ml) diluted in 100 mM potassium phosphate buffer pH 7.4 containing 3.3 mM of MgCl2, 1.2 mM of NADPH, and 0.43 µM of midazolam. Manipulation of the background protein concentration was achieved by supplementation with untransfected baculosome, resulting in a range of protein concentrations from 0.01 to 3 mg/ml. Troleandomcyin was prepared as a series of 100X concentrated stock solutions in 50/50 methanol/water, to enable testing across a range of inhibitor concentrations for IC50 determination. Reactions were initiated by addition of NADPH and incubated for 4 minutes at 37°C in a heat block.
Microsomal samples for determination of troleandomycin nonspecific binding were analyzed by LC MS/MS analysis. Chromatographic separation was achieved following injection (10 µl) onto a Halo C18 (2.1 × 30 mm, 2.7 µm) column at room temperature with a flow rate of 0.6 ml/min. The aqueous mobile phase (A) was comprised of 0.1% formic acid in water and organic mobile phase (B) consisted of 0.1% formic acid in acetonitrile. The gradient program began with 10% initial mobile phase B with a linear gradient to 90% B over 0.5 minutes, held at 90% B for 0.5 minutes, followed by re-equilibration to initial conditions for 0.55 minutes. The mass spectrometer operated in MRM mode while monitoring positive ion detection and collision energy 30, with the following mass transitions (Q1/Q3) for 1’-hydroxymidazolam of 342/324 and [2H4]-1’-hydroxymidazolam, 346/328.
Data Analysis
Inhibitor selectivity
The inhibitory profiles of each inhibitor were generated from the combined activity of 2–3 experiments using GraphPad Prism for Windows (version 9). The dose-response data were modeled by nonlinear regression using the 4-parameter logistic equation (referred to as the “4-parameter fit”):
where Y is the percent of control activity remaining and x is the inhibitor concentration. The four parameters in the model are the IC50, which represents the inhibitor concentration that yields a response halfway between the upper and lower asymptotes, the hill slope (h), which represents the steepness of the curve, and the upper and lower asymptotes (Top and Bottom, respectively), which represent the maximum and minimum possible responses. The maximal contribution of a CYP was determined by the “span” or the difference between the fitted upper and lower asymptotes. Absolute IC50 values were reported whenever an inhibition of 50% or greater was observed within the inhibitor concentration range tested. The bottom asymptote was constrained to values greater than zero. Curve fitting parameters were denoted by subscripted letter u when adjusted for nonspecific microsomal binding (i.e., unbound inhibitor concentration).
Desirability Scoring
Optimal single inhibitor concentrations were determined using a desirability scoring approach (Derringer and Suich, 1980). Desirability functions map each endpoint (% activity remaining for on- and off-target CYPs, in this case) to a number between zero and one. These individual desirability scores were combined using a geometric mean to create a single desirability score for each concentration of an inhibitor. For target CYPs, lower % activity remaining was preferred, while higher % activity remaining was preferred for off-target CYPs (Supplemental Fig. 1). Specifically, the desirability score for the most inhibited off-target CYP started decreasing linearly from a score of 1 at 95% activity remaining to a score of 0 at 50% activity remaining. This was balanced with the desirability score for the target CYP, which started increasing linearly from a score of 0 at 50% relative activity remaining to a score of 1 at 0% relative activity remaining, where relative percent activity remaining refers to the percent activity remaining normalized to the upper and lower asymptote of the fitted 4-parameter logistic curve. This design minimizes the penalty to inhibitors that never fully inhibited the target enzyme, such as α-naphthoflavone or CYP3cide. The overall desirability scores were used to rank the inhibitors, with more selective inhibitors having higher desirability scores.
Simulated Inhibition Profiles when Using 6-Parameter Model
The ability of a 6-parameter model to quantitate both target and off-target isoform contribution from a multiple concentration study design was investigated. Comparison of the theoretical concentration response profiles resulting from fitting to 6-parameter inhibition model (Fig. 4) was conducted under varying conditions of inhibitor potency, selectivity, and enzyme contribution,
where Y is the percent control activity remaining, and x is the inhibitor concentration. The six parameters in the model are MAXA and MAXB, which represent the maximum contribution of enzyme A and B, respectively, IC50A and IC50B, which are the inflection points for the inhibitor on enzymes A and B, respectively, the hill slope (h), which represents the steepness of the curve for enzyme A, and Imax, which represents the maximum possible response. The simulations were based on an inhibitor with a target potency of 0.001 µM IC50 and a second off-target IC50 of 10 µM, representing a selectivity of 10,000-fold. The effect of the inhibitor selectivity was modeled under assumptions of a single-enzyme contribution of 100%, or two enzymes contributing proportions of 60/40 or 70/30.
Rosiglitazone Dose Response Analysis
Data from test compounds (rosiglitazone) were fit using either a traditional 4-parameter dose-response curve (eq. 1) or a 6-parameter dose-response curve (eq. 2). Statistical comparison and model selection was conducted by analysis of an extra sum-of-squares F-test.
After fitting the dose-response curves to the test compound data, a two-one-sided t-tests (TOST) equivalence procedure was used to determine if the test compound IC50 was significantly within 5-fold of the probe substrate IC50 (Schuirmann, 1987; Walker and Nowacki, 2011). When the 6-parameter model was selected, the IC50 of the first phase (IC50A) of the curve was compared with the IC50 from the probe substrate. If a significant 5-fold equivalence in IC50 values between the test compound and probe was established, the reduction in % activity of the test compound was compared with zero. For the 4-parameter model, the span parameter (difference between upper asymptote and lower asymptote) must be significantly greater than zero, while the MAXA parameter must be significantly greater than zero for the 6-parameter model. If both the equivalence tests were passed and the decrease in activity was significantly greater than zero, then the inhibition of the target enzyme was deemed significant and reported.
Rosiglitazone Single Concentration Analysis
Statistical significance of inhibition at a single inhibitor concentration was determined using Welch’s unpaired two-sample t-test with unequal variances (Welch, 1947) to compare the peak area ratios without any inhibitor to the peak area ratios with the inhibitor.
Results
Microsomal Binding
The extent of non-specific binding in vitro can be important for some CYP inhibitors requiring that inhibition potency values be corrected for this phenomenon. The fraction unbound (fumic) of 17 inhibitors were determined by equilibrium dialysis at 0.03 mg/ml of human liver microsomal protein, consistent with conditions used for measurement of CYP inhibition selectivity (vide infra). Values ranged from 0.00691 to unity (Table 1). Consistent with previous literature reports, montelukast exhibited the highest nonspecific binding among the inhibitors, with an fumic of 0.00691. Among the remaining inhibitors tested, dasotraline, ketoconazole, and α-naphthoflavone also demonstrated relevant nonspecific binding to microsomes with moderate fumic values of 0.218, 0.391, and 0.390, respectively. Nearly half (8 of 17) the inhibitors showed low affinity for microsomes as indicated by fumic values >0.9. Among these, only one inhibitor, troleandomycin, was characterized as having no nonspecific binding as indicated by fumic equivalent to 1 across all protein concentrations tested. Therefore, the nominal and unbound concentrations for this subset of select inhibitors were considered equivalent. For all other inhibitors, a correction for nonspecific binding during the incubations was applied during data analysis. Additional fumic values of 0.336 for ketoconazole and 0.00193 for montelukast were determined for a microsomal protein concentration of 0.1 mg/ml, the condition used during assessments of rosiglitazone metabolism (vide infra).
Nonspecific Binding of CYP-Selective Inhibitors to Microsomal Protein
Inhibition Profiles
The selectivity profiles of 17 chemical inhibitors were determined against CYP-selective probe activity using a substrate cocktail assay in HLM for simultaneous characterization of inhibitory potency against the major human CYP isoforms CYP1A2, CYP2C8, CYP2C9, CYP2D6, and CYP3A4, as well as a separate assay of CYP2B6 activity. Inhibition plots are presented in Figs. 3 and 4 and absolute IC50 values, corrected for non-specific binding as appropriate, are reported in Table 2. Full inhibition parameters from curve fitting of cocktail and CYP2B6 inhibition are provided in Supplemental Table 2.
Absolute Unbound IC50 Values from Inhibitor Selectivity Assessment by Cocktail Assay in Human Liver Microsomes
Inhibitors Targeting CYP1A2
α-Naphthoflavone was a highly potent inhibitor of CYP1A2 mediated phenacetin O-deethylase activity (IC50,u of 0.728 nM) with good selectivity versus the remaining CYP isoforms. However, stimulation of CYP2C9-mediated diclofenac hydroxylation activity was observed with α-naphthoflavone at concentrations greater than 0.1 µM. Furafylline inhibited CYP1A2 with an IC50,u of 0.139 µM and was selective, demonstrating off-target inhibition only at concentrations greater than 10 µM, resulting in >100-fold selectivity. Furafylline was tested as a TDI of CYP1A2 (Fairman et al., 2007; Racha et al., 1998) following preincubation with HLM; however, it should be considered that selectivity will be even greater with dilution of the inactivation mixture prior to activity measurement with substrate. The potency of methoxsalen for CYP1A2 inhibition under the current experimental conditions was an order of magnitude greater for CYP1A2 (IC50,u < 10 nM) than for other CYP isoforms, of which the next most potent IC50,u was 3.83 µM against CYP2B6. However, under these conditions methoxsalen is expected to inhibit CYP2A6, an isoform not included in our investigations, based on previously reported KI values ranging from 0.33–1.9 µM (Draper et al., 1997, Koenigs et al., 1998; Kharasch et al., 2000). Significant inhibition (>50%) of the remaining CYP isoforms by methoxsalen concentrations greater than 10 µM was consistent with its utility as a pan-CYP inhibitor as suggested by Palarcharla et al. (2019).
Inhibitors Targeting CYP2B6
The IC50 of the TDI 2-phenyl-2-1(1-piperidinyl)propane (PPP) against target isoform CYP2B6 and the next most potently inhibited off-target isoform (CYP2D6) represented ∼54-fold selectivity for CYP2B6. PPP is a TDI of CYP2B6 (Chun et al., 2000; Walsky et al., 2007), and it should be considered that selectivity could be even greater when the pre-incubation mixture is diluted before addition of the substrate. In HLM, dasotraline was a potent inhibitor of CYP2B6 (IC50,u of 1.43 nM); however, selectivity over the next reported isoform, CYP2D6, was only 27-fold. These findings are similar to the inhibitory selectivity of dasotraline reported in human hepatocytes (Yang et al., 2019) and to our knowledge represent the first published report of dasotraline inhibition activity in HLMs. Ticlopidine was a potent inhibitor of CYP2B6 (IC50,u 25.9 nM) with poor selectivity of approximately only ninefold versus CYP2C19 (IC50,u 22.2 nM), and ∼82- to 94-fold versus CYP1A2 and CYP2D6, respectively.
Inhibitors Targeting CYP2C8
Montelukast was a highly potent inhibitor of CYP2C8 with an IC50,u of 93.1 pM. It should be noted that montelukast demonstrates high non-specific binding to microsomes and will exhibit nominal IC50 values that are dependent on the microsomal concentration employed. It was 77-fold selective, with CYP2C9 as the next most potently inhibited enzyme. Gemfibrozil glucuronide was also evaluated and showed 74-fold selectivity versus CYP2C9; however, it also showed inhibition against other CYPs. Gemfibrozil glucuronide is a TDI of CYP2C8 (Baer et al., 2009), and selectivity may be even greater when the pre-incubation mixture is diluted before addition of substrate.
Inhibitors Targeting CYP2C9
Sulfaphenazole and tienilic acid were among the most potent and selective of all the inhibitors tested with sulfaphenazole having >500-fold selectivity for CYP2C9 (IC50,u 0.120 µM) versus the next isoform (CYP3A4). Tienilic acid was potent for CYP2C9 (IC50,u 61.1 nM) even without a dilution step to leverage the mechanism-based inhibition property known for this compound.
Inhibitors Targeting CYP2C19
N-Benzylnirvanol was a highly selective inhibitor for CYP2C19 (IC50,u 51.1 nM) with >300-fold selectivity over CYP3A4 and CYP2C8. When used as a reversible inhibitor, esomeprazole demonstrated only 9-fold selectivity for CYP2C19 inhibition versus CYP3A4 as well as noteworthy inhibition of CYP1A2, CYP2B6, CYP2C8, and CYP2D6. However, it should be noted that esomeprazole is reported as a TDI of CYP2C19 (Ogilvie et al., 2011) and that selectivity is likely improved when subject to preincubation with microsomes and NADPH prior to addition of substrate.
Inhibitors Targeting CYP2D6
Quinidine was a potent and selective inhibitor of CYP2D6 (IC50,u 15.0 nM), with negligible effects (>2000-fold selectivity) on other off-target isoforms. Paroxetine demonstrated similar potency for CYP2D6 but lower selectivity of only 13-fold over the next most potently inhibited enzyme CYP2B6. Paroxetine is a known TDI (Bertelsen et al., 2003), and its true selectivity requires evaluation with a dilution step that follows preincubation. However, when employing a protocol wherein inhibitor and test substrate are co-incubated, quinidine offers a superior choice over paroxetine.
Inhibitors Targeting CYP3A4
The most potent inhibitors of human liver microsomal midazolam 1’-hydroxylase activity were ketoconazole (IC50,u 2.95 nM), CYP3cide (IC50,u 9.72 nM) and troleandomycin (IC50,u 2.71 µM). Surprisingly, off target inhibition of all six of the other isoforms by ketoconazole was significant based on absolute IC50,u values ranging from the most potent, 0.316 µM (CYP2B6) and 0.617 µM (CYP2C8), to 1.80 µM for CYP2D6. In contrast, both CYP3cide and troleandomycin showed little inhibition of other CYP isoform activity within the concentration ranges tested, even when these known TDIs (Walsky et al, 2012; Pessayre et al., 1982) were not subject to a dilution step suggesting improved selectivity profiles relative to ketoconazole. However, using any of these three inhibitors at single concentrations can cause complications (see below).
Desirability Scoring for Inhibitor Comparisons and Optimized Test Concentrations
CYP reaction phenotyping experiments are frequently conducted using single concentrations of inhibitors, for reasons of simplicity and sparing resources. However, as described above, many of the inhibitors employed for this purpose lack enough selectivity to avoid inhibition of off-target CYP enzymes when used at a concentration that effects maximal inhibition of the target CYP. To evaluate this, a desirability score algorithm (Harrington, 1965) was employed to compare the selectivity of the inhibitors and identify optimal single concentrations for use in CYP reaction phenotyping (Table 3). Optimal single concentrations for several of the most selective inhibitors (i.e., α-naphthoflavone, tienilic acid, quinidine, sulfaphenazole, furafylline, and N-benzylnirvanol) were readily identified as achieving >90% inhibition of their respective target CYP enzymes with negligible (<10%) inhibition of any other isoforms (Table 3). For these inhibitors, the reported optimum condition simultaneously achieves maximal inhibition of targeted isoforms with negligible penalties of off-target inhibition.
Recommended Single-Concentration, Desirability Score, and Selectivity Profiles of Optimum Inhibitor Selectivity Conditions
Data for 16 inhibitors ranked by desirability scoring of individual curve analysis of target- versus most-inhibited off-target inhibition curves. Values in bold-face type represent the inhibition of the target enzyme for each inhibitor. Dashed entries indicate less than 1% inhibition. Selective inhibitors achieving >90% target inhibition with <10% off target activity are positioned above the bold line. An analysis of the optimal target condition for methoxsalen was not conducted given its evaluation within these studies as a pan-CYP inhibitor.
The inhibitors listed in Table 3 that reside below the solid line lack the selectivity to deliver >90% inhibition of their target enzymes without causing problematic off-target inhibition. Thus, the optimal concentrations for these reflected a compromise of efficacy and selectivity whereby the magnitude of target inhibition was decreased to minimize inhibition of other isoforms. However, even when the target isoform inhibition was less than ideal (<90%) some degree of off-target inhibition was still projected, and it was often the case that one other CYP enzyme was inhibited by at least 10%. If these inhibitors are used at concentrations that yield 95% inhibition of their target enzymes, then substantial off-target inhibition will occur (Table 4). It was noteworthy that a single concentration of the widely used CYP3A4 inhibitor ketoconazole could not be identified that achieves 95% inhibition of CYP3A4 without causing measurable inhibition of several other isoforms. An optimal unbound concentration of 0.030 µM of ketoconazole yields about 88% inhibition of CYP3A while inhibiting CYP2B6 by 14%. However, to achieve 95% inhibition of CYP3A4 with ketoconazole (unbound concentration of 95% inhibition of 0.09 µM), concomitant inhibition of CYP2B6 increased to nearly 30% and that of CYP2C8, CYP2C9, and CYP2C19 all exceed 10%.
Recommended IC95 and Inhibitory Activity Associated with Maximal Target Inhibition by Selected Inhibitors
Conditions of maximal inhibition are provided for inhibitors requiring conditions beyond recommended optimum guidance (Table 3) to achieve approximately 95% loss of target isoform activity. The increased concentrations required to achieve full target isoform inhibition incur larger off-target penalties leading to artificially increased inhibition values. CYP3cide (not shown) was unable to achieve 95% inhibition during the selectivity evaluation. Values in bold-face type represent the inhibition of the target enzymes for each inhibitor.
Simulated Inhibition Curves from Six-Parameter Curve Fitting
For purposes of illustration, the representative profiles in Fig. 4 were generated to simulate the effects of an inhibitor with 10,000-fold target selectivity on a single enzyme of varying contribution to the overall activity. When off-target inhibition occurs, the 6-parameter model attempts to quantify inhibition parameters for both on- and off-target inhibition and therefore can dissect the maximal contribution of the target enzyme and eliminate confounding data from off-target inhibition. In panel A, the characteristic profile represents the simplest scenario of a single enzyme-catalyzed reaction, such as for well-established CYP marker reactions. In this case, the nominal and absolute IC50 are the same, the maximum inhibition value is 100%, and the data are best fit to a simple 4-parameter model. In panel B, the contribution of the enzyme is partial (60%), and the inhibitor is highly selective; therefore, the lower asymptote is flat at high inhibitor concentrations and reflects the maximum inhibition (MAXA). When these types of data were fit to the 4-parameter model, the MAXA was less than 100% and the IC50A represented the inflection point midway between the upper and lower asymptotes and not a concentration that yields 50% nominal inhibition. The situation becomes more complex in Panels C and D with profiles representative of frequently encountered drug metabolism reactions comprised of multiple enzymes catalyzing the same reaction combined with inhibitors of suboptimal selectivity. These data require fitting to the 6-parameter model to delineate the important MAXA values that reflect fm. The first phase represents metabolism by the enzyme that is targeted by the inhibitor being employed and the second is metabolism catalyzed by a second enzyme that is an off-target enzyme of the inhibitor. The shape of the % activity versus log[I] curve will depend on the relative contributions of the two enzymes (i.e., MAXA and MAXB), the span of the inhibitor potencies for the two enzymes (IC50A versus IC50B), and the range of inhibitor concentrations tested. Clearly, fitting the complex 6-parameter function successfully requires a greater number of datapoints than typically employed and a wide span of test concentrations; in the data reported in this study there were >20 concentrations evaluated in replicates that spanned over five orders of magnitude. Furthermore, successful fitting of data to the 6-parameter model is reliant on the precision of replicates and statistical identification of outlier points.
Demonstration of the Utility of Full Inhibition Curves and Six-Parameter Curve Fitting for Reaction Phenotyping: Example of Rosiglitazone N-demethylase
Contributions of CYP2C8 and CYP3A4 to rosiglitazone N-demethylase activity in HLM were evaluated by comparing results from single- versus multiple-inhibitor concentration approaches (Table 5). Inhibition of rosiglitazone N-demethylase was 68.0% when using the optimum, single montelukast concentration from Table 3. This increased to 81.1% (+13%) under the maximal-montelukast concentration tested, 0.00312 µM, a difference consistent with the observed magnitude of CYP3A4 off-target (18%) inhibition associated with the unbound concentration of 95% inhibition of montelukast from selectivity experiments (Table 4). However, employing multiple concentrations of montelukast and using a detailed 6-parameter fitting of the multiple concentration inhibition curve yielded a CYP2C8 contribution of 70.8%. The montelukast unbound IC50A value of 0.0468 nM for rosiglitazone N-demethylation (Table 5) was validated by the equivalence test comparison with the positive control activity amodiaquine N-deethylase (IC50,u 0.0483 nM; Table 6, Fig. 5B). Visible in the inhibition curve (Fig. 5A) was the off-target inhibition of CYP3A at concentrations of montelukast greater than 0.001 µM. Thus, the estimation of fm for CYP2C8 in rosiglitazone N-demethylation from the multiple concentration data (fm = 0.71) was sandwiched between the projections from the less efficacious optimum and the less selective maximal conditions for single-inhibitor concentration tests.
Quantitation of Isoform Contribution to Rosiglitazone N-demethylase Activity in Human Liver Microsomes from Multiple- Versus Single-Inhibitor Concentration Analysis
When necessary for single concentration approach, values were reported for the tested concentration nearest to the targeted optimal condition (within 20%). IC50 equivalence from multiple concentration analysis assessed by TOST equivalence procedure and Welch’s unpaired two-sample t- test with unequal variances of inhibited versus solvent control applied to single concentration analysis. Data represent mean and 95% confidence interval of N=3 replicates.
Statistical Equivalence Testing of IC50 Determinations From Rosiglitazone Inhibition Experiments
Comparison of intraexperimental inhibitor potency parameters was conducted by TOST equivalence procedure for verification of inhibitor selectivity. IC50 Data represent mean and 95% confidence interval of N=3 replicates.
The other enzyme that contributes to rosiglitazone N-demethylation was shown to be CYP3A and both ketoconazole and troleandomycin were evaluated as inhibitors for comparison (Fig. 5A). CYP3A4 contributions of 18.3% and 22.2% were projected from optimal and maximal single ketoconazole concentration analysis, respectively, and these were greater than the estimate of 14.3% made from a full inhibition curve with 6-parameter fitting. Overestimation by the single concentration conditions likely reflected non-selective inhibition of CYP2C8 by ketoconazole under the single concentration conditions. The necessity of the 6-parameter fit was demonstrated by the downward slope of the inhibition curve at concentrations of ketoconazole > 0.1 µM presumably due to off-target inhibition of CYP2C8 activity, the major contributing enzyme in this pathway (Fig. 5A). Estimation of CYP3A4 contribution by a single optimal concentration of troleandomycin was slightly lower than that from curve fitting of the multiple concentration inhibition profile (14.9% versus 22.1%). A simple 4-parameter curve sufficiently fit the troleandomycin inhibition curve. The 2-fold difference in troleandomycin IC50,u values for rosiglitazone versus positive control (Table 6, Fig. 5B) failed the equivalence test, and this was most likely due to variability from the low response magnitude and not a true difference in selectivity.
Discussion
Accurate estimation of contributions to metabolism by individual CYP enzymes by reaction phenotyping methods is integral to a robust prediction of in vivo drug-drug interactions and interindividual variability in pharmacokinetics due to genetic polymorphisms of drug metabolizing enzymes. Although multiple reaction phenotyping tools have been developed over the past decades, isoform-selective chemical inhibition methods are most commonly used and recommended by drug regulatory authorities in combination with a parallel approach, such as recombinant human CYP enzymes. The utility of chemical inhibitors is highly dependent on their selectivity for a target isoform. Successful use of an inhibitor requires knowledge of both its broad ranging selectivity among isoforms and the quantitative potency of those interactions.
In the present study, the selectivity of several CYP inhibitors was evaluated, and results were analyzed using a desirability scoring statistical approach (Derringer and Suich, 1980) to provide an optimal, single-concentration recommendation for use during reaction phenotyping. The resulting conditions targeted optimum >90% inhibition of the targeted enzyme while affording <10% inhibition of the next most potently inhibited enzyme (Table 3). The results indicated significant spillover of inhibition to non-targeted CYP isoforms for most of the inhibitors tested. However, the selectivity of some inhibitors was good enough for use as reliable tools for reaction phenotyping. Among the more selective inhibitors were α-naphthoflavone and furafylline for CYP1A2, sulfaphenazole and tienilic acid for CYP2C9, N-benzylnirvanol for CYP2C19, and quinidine for CYP2D6. Even under the suggested conditions, these inhibitors are not entirely devoid of off-target effects but rather achieve a delicate balance; true selectivity would require a ≥1000-fold spread in target and off-target enzymes.
For the remaining inhibitors, the optimal concentration yielded insufficient inhibition of the target isoform before simultaneously incurring off target enzyme inhibition greater than 10%. Conditions under which these less selective inhibitors could achieve maximal target inhibition were also considered; however, using the IC95 resulted in marked inhibition of multiple off-target CYP enzymes (Table 4). Notable among this group of inhibitors is the presence of the widely used CYP3A inhibitor, ketoconazole. As described above, the optimum ketoconazole condition identified, 0.03 µM, unbound, yielded only 88% inhibition of CYP3A4, while concurrently inhibiting CYP2B6 by 14%. The use of ketoconazole is reportedly further complicated by specific and non-specific binding and stoichiometry relative to the amount of CYP3A4 enzyme in incubations (Tran et al., 2002). The other CYP3A inhibitors tested, troleandomycin and CYP3cide, demonstrated similar selectivity profiles as ketoconazole thereby leaving few options for accurately testing CYP3A4 using a single concentration condition (i.e., typical reported ketoconazole test concentration of 1 µM total), especially concerning given CYP3A is a major contributing DME to many marketed drugs, accounting for up to two thirds of the reported DDIs among drugs approved between 2013–2016 (Yu et al., 2018). For this cohort of inhibitors, the result of discounted off target activity could lead to overestimation of isoform contribution and ultimately to unnecessary conduct of costly victim DDI studies. Although not evaluated in these studies, the inclusion of a dilution step following preincubation would enhance the selectivity of inactivators tested.
To address these shortcomings, it was hypothesized that a more comprehensive study design would improve estimations of fm values by addressing the often-discounted selectivity problems of single concentration inhibitor usage that leads to cumulative inhibition exceeding 100% across a panel of inhibitors. When generating full inhibition curves with dense datasets over a wide concentration range, it was possible to distinguish between the inhibitor activity toward target and off-target isoforms (of sufficiently separated potency values) using a complex 6-parameter data fitting model. When off-target inhibition occurs, the 6-parameter model quantifies parameters for both on- and off-target inhibition and therefore can dissect the maximal contribution of the target (MAXA) to eliminate confounding off-target inhibition.
To experimentally demonstrate the approach, rosiglitazone N-demethylation was selected as a model drug metabolism reaction since it is known to be catalyzed by multiple CYP enzymes (Bazargan et al., 2017; Baldwin et al., 1999; Park et al., 2004). Montelukast and ketoconazole were employed as typical inhibitors of CYP2C8 and CYP3A4, and troleandomycin was also evaluated for comparison with the ketoconazole data. CYP2C8 was the largest contributing isoform to the reaction as shown by montelukast inhibition in HLM (Fig. 5A). Visual assessment of the multiphasic concentration response profile was consistent with the selectivity of montelukast (Table 2) and indicative of off-target inhibition at the higher concentrations. This was also evident when comparing the rosiglitazone inhibition data to the inhibition for the CYP2C8 selective marker activity amodiaquine N-deethylase, which yields a complete inhibition curve. Statistical analysis confirmed a best-fit of the data with the 6-parameter model in which it was calculated that CYP2C8 contributes 71% of this reaction in HLM (Table 5). Confirmation of the assignment to CYP2C8 was made through demonstration of the statistical equivalence of the observed montelukast IC50 for rosiglitazone N-demethylation to that of the within-experiment positive control amodiaquine N-deethylase reaction (Table 6 and Fig. 5B). The integration of a microsomal binding correction for unbound inhibitor concentration is critical to conducting the statistical determination of equivalence. The inhibition of rosiglitazone formation by montelukast at unbound concentrations greater than 0.001 µM (Fig. 2) resulted in an estimated off-target IC50B value within the range of montelukast off target IC50 values for CYP3A4 and CYP2C9 (Table 3) from selectivity studies. These secondary parameters can sometimes provide clues as to the identity of other contributing isoforms, but themselves are not definitive enough for quantitation; use of the inhibitors that target these other enzymes is necessary. A smaller contribution by CYP3A4 to rosiglitazone N-demethylation formation was determined from the ketoconazole inhibition profile, which was consistent with multiple phases and best fit to the 6-parameter model. The initial shallow plateau was consistent with a small but significant CYP3A4 contribution of 14% (Fig. 5A; Table 5) and the second phase of inhibition at high inhibitor concentrations was likely spillover to CYP2C8. In contrast, the troleandomycin inhibition profile readily fit to the simpler 4-parameter model due to its improved selectivity versus ketoconazole. It should be noted that the IC50 values for troleandomycin inhibition of midazolam 1’-hydroxylase activity in Table 3 and Table 5 differed by 2.7X, and this exemplifies inter-experimental variability that shows the necessity of the use of a positive control marker substrate activity within each experiment and not merely making comparisons to previously obtained data.
Inhibition Curves for Commonly Used Inhibitors. Data represent line of best fit (solid line) and 95% confidence interval (shaded) from N = 2–3 experiments. Abbreviations: PPP, 2-phenyl-2-(1-piperidinyl)propane.
Inhibition Curves for Alternate Inhibitors. Data represent line of best fit (solid line) and 95% confidence interval (shaded) from N = 2–3 experiments. Abbreviations: α-NF, α-naphthoflavone; GG: gemfibrozil glucuronide; TAO: troleandomycin.
Theoretical Inhibition Profiles Resulting from 6-Parameter Modeling. Profiles demonstrating inhibitor activity effecting a single target enzyme contributing 100% of reaction activity with IC50A (A), complete inhibition of a single target enzyme contributing 60% of reaction activity with IC50A (B), complete inhibition of target enzyme contributing 70% of reaction activity by inhibitor with IC50A for target isoform and IC50B for off-target isoform (C), inhibition of target enzyme contributing 30% of reaction activity by inhibitor with IC50A for target isoform and IC50B for off-target isoform.
(A) Rosiglitazone N-demethylation in Human Liver Microsomes: Inhibition by Montelukast, Ketoconazole, and Troleandomycin. Inhibition of CYP3A and CYP2C8 activity to rosiglitazone N-demethylation (red circles) and probe substrate reaction (black squares) in HLM by inhibitors montelukast (top panel), ketoconazole (middle panel), and troleandomycin (bottom panel). Nonlinear regression analysis of montelukast and ketoconazole inhibition of N-desmethylrosiglitazone formation represent best fit from 6 parameter model. Data points represent mean and SD (N = 3). (B) Comparison of intra- and inter- experimental IC50 values from montelukast, ketoconazole, and troleandomycin inhibition of rosiglitazone N-demethylation and probe substrates.
From these findings, recommendations on the use of chemical inhibitors for quantitative reaction phenotyping are summarized (Table 7). First, the data clearly show that most CYP inhibitors used in reaction phenotyping have selectivity profiles which may impact the quality of prediction when unaccounted under commonly used single-concentration conditions. Limiting inhibitor concentrations to the optimal condition described will minimize confounding effects due to selectivity. Second, inhibitor evaluations conducted across a wide concentration range and combined with modeling and statistical analysis of the full inhibition curves offers a superior approach to the use of chemical inhibitors in CYP reaction phenotyping, as compared with the more typically used single concentration data. Finally, it must be appreciated that chemical inhibitor studies are one component of a larger phenotyping evaluation. The study designs should be fit-for-purpose with more complex resource-intensive full inhibition curves considered appropriate to influence decision making on clinical pharmacology studies for advanced development candidates for example, whereas the abbreviated single-inhibitor concentration experimental design may suffice to preliminarily assess enzyme involvement when designing optimal drug candidates. Additionally, an informed selection of chemical inhibitors for reaction phenotyping can be guided by foreknowledge of contributing isoforms provided from supporting assays, such as recombinant CYP. The manner in which chemical inhibition data can be best integrated into CYP reaction phenotyping is described in the accompanying report (Doran et al., manuscript submitted). The integrated understanding of chemical inhibitor selectivity should result in improved experimental design, in vitro inhibitor identification, and ultimately improvements in delineation of CYP contribution to drug metabolism.
Summary of Recommended Inhibitor Conditions for Use in Reaction Phenotyping Experiments
Recommended conditions for both single and multiple concentration use of chemical inhibitors based on desirability scoring of inhibitor selectivity analysis. The recommended conditions are reported as unbound concentrations along with nominal concentration associated with a 0.1 mg/ml protein condition. Nominal inhibitor concentrations for alternate incubation conditions should be derived by scaling of the recommended unbound concentration by the appropriate microsomal unbound fractions listed in Table 1.
Acknowledgments
The authors would like to acknowledge Carolyn Richardson for assistance with selectivity experiments.
Authorship Contributions
Participated in research design: Doran, Burchett, Goosen, Obach.
Conducted experiments: Landers, Gualtieri, Balesano, Obach.
Performed data analysis: Doran, Burchett, Landers, Gualtieri, Balesano, Dantonio, Obach.
Wrote or contributed to the writing of the manuscript: Doran, Burchett, Landers, Gualtieri, Dantonio, Goosen, Obach.
Footnotes
- Received March 2, 2022.
- Accepted June 9, 2022.
This work received no external funding.
No author has an actual or perceived conflict of interest with the contents of this article.
↵
This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- CYP
- cytochrome P450
- fm
- fraction metabolized
- fu, mic
- microsomal unbound fraction
- HLM
- human liver microsome
- IC50
- half-maximal inhibitory concentration
- IC50,u
- unbound half-maximal inhibitory concentration
- IC95
- concentration of 95% inhibition
- LC-MS/MS
- liquid chromatography with tandem mass spectrometer
- MAX
- maximal percent inhibition
- MRM
- multiple reaction monitoring
- PPP
- 2-phenyl-2-1(1-piperidinyl)propane
- Q1
- first quadrupole of triple quadrupole mass spectrophotometer
- Q3
- third quadrupole of triple quadrupole mass spectrophotometer
- TDI
- time-dependent inhibitor
- Copyright © 2022 by The American Society for Pharmacology and Experimental Therapeutics
