Abstract
Time-dependent inhibition (TDI) of CYP3A is an important mechanism underlying numerous drug-drug interactions (DDIs), and assays to measure this are done to support early drug research efforts. However, measuring TDI of CYP3A in human liver microsomes (HLMs) frequently yields overestimations of clinical DDIs and thus can lead to the erroneous elimination of many viable drug candidates from further development. In this investigation, 50 drugs were evaluated for TDI in HLMs and suspended human hepatocytes (HHEPs) to define appropriate boundary lines for the TDI parameter rate constant for inhibition (kobs) at a concentration of 30 µM. In HLMs, a kobs value of 0.002 minute−1 was statistically distinguishable from control; however, many drugs show kobs greater than this but do not cause DDI. A boundary line defined by the drug with the lowest kobs that causes a DDI (diltiazem) was established at 0.01 minute−1. Even with this boundary, of the 33 drugs above this value, only 61% cause a DDI (true positive rate). A corresponding analysis was done using HHEPs; kobs of 0.0015 minute−1 was statistically distinguishable from control, and the boundary was established at 0.006 minute−1. Values of kobs in HHEPs were almost always lower than those in HLMs. These findings offer a practical guide to the use of TDI data for CYP3A in early drug-discovery research.
Significance Statement Time-dependent inhibition of CYP3A is responsible for many drug interactions. In vitro assays are employed in early drug research to identify and remove CYP3A time-dependent inhibitors from further consideration. This analysis demonstrates suitable boundaries for inactivation rates to better delineate drug candidates for their potential to cause clinically significant drug interactions.
Introduction
CYP3A is the major drug-metabolizing enzyme in humans and is expressed in the liver and intestine. It contributes to the metabolic clearance and first-pass extraction of a vast array of drugs and, as such, is also the target for numerous pharmacokinetic drug-drug interactions (DDIs). Interactions arise via coadministration of CYP3A-cleared drugs with those that cause inhibition (e.g., itraconazole), inactivation (e.g., clarithromycin), or induction (e.g., rifampin) of the enzyme. Depending on the rate and fraction of clearance and/or first-pass extraction occurring via CYP3A-mediated metabolism, the magnitude of drug interactions can be high. For example, the antianxiety agent buspirone, which has high clearance and a high fraction of its metabolism catalyzed by CYP3A, has been reported to have increases in exposure of up to 19-fold with coadministration of itraconazole (Kivistö et al., 1997). Time-dependent inactivators (TDIs) also elicit significant DDIs when coadministered with the sensitive CYP3A marker substrate midazolam (e.g., ritonavir increased midazolam exposure 26-fold) (Greenblatt et al., 2009). Moreover, combination therapy with ritonavir and indinavir (not a known TDI) increased alfentanil exposure 37-fold (Kharasch et al., 2009).
Over the past 2 decades, considerable effort has been made to develop in vitro assays for cytochrome P450 activity that can be used to evaluate drugs and other xenobiotics as inhibitors and time-dependent inactivators of this important enzyme. Complex assays to measure KI and kinact are used in drug development, and these parameters can be used in combination with other input values to make estimates of DDI using mechanistic and physiologically based pharmacokinetic modeling. Such data can also be used in calculating the “R2” value described by the US Food and Drug Administration to make a first decision regarding whether a new drug candidate needs further evaluation (Food and Drug Administration, 2020). In earlier phases of drug research, assays to measure TDI that are pared down from the labor-intensive full determination of KI and kinact have been described, such as the IC50 shift or 2 + 2 assays (Grimm et al., 2009). Midazolam 1′-hydroxylase and testosterone 6β-hydroxylase activities have been most frequently employed as marker activities for CYP3A4 (and the closely related enzyme CYP3A5 for midazolam) in human liver microsomal (HLM) preparations (Kawano et al., 1987; Waxman et al., 1988; Gorski et al., 1994; Kokudai et al., 2009). Data obtained from such assays can be combined with other input parameters (in vivo concentration of the inhibitor, fraction of the substrate that is cleared by CYP3A in vivo) and used to make comparisons among compounds and predictions of the magnitude of DDI in vivo. Methods used to correlate in vitro inhibition data to in vivo DDI range from the use of simple relationships derived from the Rowland-Matin equation (Rowland and Matin, 1973) to more sophisticated physiologically based pharmacokinetic modeling approaches (Jones and Rowland Yeo, 2013).
Because of the high number of drugs cleared by CYP3A-mediated metabolism, in vitro assays to measure inhibition of this enzyme are frequently employed early in the drug research process to avoid developing new drugs that could cause DDI with CYP3A-cleared drugs. The data from these assays can be used in a simple screening-type strategy (i.e., qualitatively categorizing as an inhibitor versus noninhibitor) or in developing structure-activity relationships that can be used in designing away from CYP3A inhibition. As drug research projects approach the selection of the best candidate compound to bring forth into the development phase, in vitro CYP3A inhibition data are used in the prediction of DDI when combined with other important input parameters, such as dose, free fraction in plasma, and projected clinical exposure. Assays used include tests for simple reversible inhibition as well as time-dependent irreversible inactivation. Reversible inhibition data for CYP3A have been generally found to be reliable in the prediction of DDI (Obach et al., 2006; Einolf, 2007; Vieira et al., 2014); however, time-dependent inhibition frequently overestimates the magnitude of DDI (Obach et al., 2007; Kenny et al., 2012; Vieira et al., 2014). The reasons for overprediction for TDI are not known.
Time-dependent inhibition of cytochrome P450 enzymes can occur via a few different mechanisms. Metabolism of certain chemical substituents (e.g., alkylamines, methylenedioxyphenyl) can yield intermediate metabolites that can form a tight noncovalent interaction with the heme iron [referred to as metabolite-intermediate complex formation (Franklin, 1977)]. Examples of drugs that inactivate CYP3A by this mechanism include verapamil, lapatinib, and erythromycin (McConn et al., 2004; Wang et al., 2005; Barbara et al., 2013). Metabolism of other chemical substituents can lead to the generation of reactive electrophilic intermediates that can form covalent bonds with the enzyme or the heme. Drugs that inactivate CYP3A in this manner include ethinyl estradiol, raloxifene, and ritonavir (Lin et al., 2002; Baer et al., 2007; Rock et al., 2014). For these mechanisms of inactivation, the relationship between chemical structure, metabolism/bioactivation, and inactivation is well understood. However, in the routine employment of a liver microsomal CYP3A time-dependent inhibition screening assay in early drug discovery [i.e., measurement of inactivation rate constant (kobs) at a test concentration of 30 µM], we have found that 75% (unpublished observations) of new compounds synthesized exhibited time-dependent inhibition. An observation similar to this was reported by Zimmerlin et al. (2011) using a test concentration of 10 µM. The vast majority of compounds showing kobs values that are statistically different from solvent control possess no obvious chemical structure associated with the aforementioned inactivation mechanisms. Thus, drug-discovery project teams are faced with the dilemma of whether the assay findings for a compound of interest truly portend a DDI risk, which will require further design efforts to mitigate that risk, or whether the assay is either oversensitive or generates artifactual findings for the compound. One hypothesis for this is whether liver microsomes artificially lack some unknown factor present in vivo that protects against the inactivation of CYP3A. Early efforts we made to address this by adding factors, such as ascorbate, superoxide dismutase, glutathione, and other agents, to microsomal TDI assays had no impact on reducing TDI rates (unpublished observations).
The use of suspended human hepatocytes (HHEPs) as an alternate system to measure CYP3A TDI has been proposed because they are more physiologically relevant than HLMs. HHEPs have an intact cell membrane, a full complement of drug-metabolizing enzymes, and physiologic cofactor concentrations. Chen et al. (2011) compared inactivation kinetics for four CYP3A TDIs and showed that KI was greater in hepatocytes than in microsomes with kinact mostly unchanged. Conversely, some authors have shown comparable KI values, whereas HHEP kinact was significantly reduced relative to that of HLMs (Mao et al., 2013; Kimoto et al., 2019). More recently, it has been suggested that HLM inactivation parameters corrected for free cytosolic inhibitor concentrations estimated in HHEPs could improve DDI predictions (Filppula et al., 2019). Mao and coworkers (2011, 2012, 2016) have evaluated HHEPs as a system for measuring TDI, also adding plasma to the incubation medium to include the impact of protein binding on projection of in vivo DDI with improved success.
Based on our experiences with overprojection of DDI from liver microsomal TDI data and the high rates of positive TDI without obvious structural alerts among thousands of new chemical entities in drug-discovery programs, we endeavored to make a systematic comparison of liver microsomes and suspended HHEPs for CYP3A TDI. Fifty drugs for which clinical DDI data for CYP3A exist (both positive and negative) were evaluated for TDI in liver microsomes and hepatocytes. The overall objectives were 1) to compare TDI kinetics in the two systems to determine whether there is a systematic trend of higher inactivation in microsomes and 2) to develop an empirically driven cutoff value for kobs (at [I] = 30 µM) under which drug design teams can be more assured that DDI will not be a clinical problem.
Materials and Methods
Materials
Research was conducted on human tissue acquired from a third party that has been verified as compliant with Pfizer policies, including institutional review board/independent ethic committee approval. Human liver microsomes pooled from 36 male and 14 female donors were purchased from Sekisui XenoTech (Kansas City, KS). Cryopreserved HHEPs pooled from four male and six female donors were purchased from BioIVT (Westbury, NY). Monobasic and dibasic potassium phosphate buffers, magnesium chloride, NADPH, HEPES, and DMSO were purchased from Sigma (St. Louis, MO). Midazolam was purchased from U.S. Pharmacopeia (Rockville, MD). 1′-Hydroxymidazolam and [2H4]-1′-hydroxymidazolam were synthesized at Pfizer (Groton, CT). William’s medium E was purchased from Gibco (Dublin, Ireland). Commercially obtained chemicals and solvents were of high-performance liquid chromatography or analytical grade. Tested drugs were purchased from one of the following sources: Sigma-Aldrich, Toronto Research Chemicals (North York, ON, Canada), or MedChemExpress (Monmouth Junction, NJ).
Identification of Test Drugs
A list of drugs for which clinical CYP3A interaction studies have been conducted was compiled using the University of Washington Drug Interaction Database (https://www.druginteractionsolutions.org). For drugs in which more than one interaction study was published, the study that had the greatest magnitude of interaction was selected. Oral midazolam was the preferred probe substrate, but in a few cases, studies using intravenous midazolam or alternate CYP3A probe substrates were referenced. The magnitude of DDI was based on area-under-the-plasma-concentration-time-curve ratios (AUCRs).
Time-Dependent Inhibition in Human Liver Microsomes
Time-dependent inhibition of CYP3A was measured in human liver microsomes (0.3 mg/ml) supplemented with MgCl2 (3.3 mM) and NADPH (1.3 mM) in potassium phosphate buffer (100 mM, pH 7.4). Drug stock solutions prepared at 100 times the incubation concentration in 50% acetonitrile and 50% water, or control solvent was added to this incubation mixture to initiate the reaction. Final incubation concentration was typically 30 µM. At various time points (1, 5, 10, 20, 30, and 40 minutes), an aliquot of this mixture was transferred to an activity incubation mixture containing midazolam (20.9 µM, 10-fold KM), MgCl2 (3.3 mM), and NADPH (1.3 mM) in potassium phosphate buffer (100 mM, pH 7.4), resulting in a 20-fold dilution. After 6 minutes, the activity reaction was terminated by the addition of two volumes of acetonitrile containing internal standard (100 ng/ml [2H4]-1′-hydroxymidazolam). All reactions were carried out at 37°C at a final volume of 200 µl and done in duplicate. Samples were vortexed and centrifuged for 5 minutes at approximately 2300g at room temperature. The supernatant was mixed with an equal volume of water containing 0.2% formic acid and analyzed directly by tandem liquid chromatography–mass spectrometry (LC-MS/MS).
For several drugs, assay conditions had to be modified to accommodate potent inhibition at the initial time point and rapid inactivation. A 50-fold dilution was used for mibefradil, propiverine, saquinavir, and simvastatin. Lower concentrations were tested: 1 µM for itraconazole; 3 µM for cobicistat, conivaptan, mibefradil, nelfinavir, and tofisopam; and 10 µM for nilotinib. An alternate stock solvent was used to prepare clarithromycin (6% DMSO, 94% acetonitrile), nefazodone (20% DMSO, 80% acetonitrile), pimavanserin (6% DMSO, 20% water, 74% acetonitrile), ritonavir (64% acetonitrile, 36% water), sorafenib (90% ethanol, 10% water), terbinafine (44% acetonitrile, 16% methanol, 40% water), terfenadine (65% acetonitrile, 15% methanol, 17% water), alectinib (12% DMSO, 42% acetonitrile, 38% methanol, 8% water), carfilzomib (11% DMSO, 22% water, 67% acetonitrile), midostaurin (9% DMSO, 74% acetonitrile, 17% water), and telaprevir (10% DMSO, 49% acetonitrile, 20.5% methanol, 20.5% water). The final total solvent in the primary incubations was ≤1%.
Time-Dependent Inhibition in Suspension Human Hepatocytes
Time-dependent inhibition of CYP3A was measured in HHEPs (0.45 million hepatocytes/ml) suspended in William’s medium E supplemented with l-glutamine and HEPES (50 mM). Drug stock solutions prepared at 10 times the incubation concentration in 90% media, 5% acetonitrile, and 5% water were added to this incubation mixture to initiate the reaction. The final incubation concentration was typically 30 µM (unless otherwise stated) in a volume of 50 µl. At various time points (typically 5, 10, 30, 60, 90, and 120 minutes unless otherwise stated), a 200-µl aliquot of the activity incubation mixture consisting of midazolam (final concentration 80 µM, approximately 5-fold KM) in media was added to the incubation wells, resulting in a 5-fold dilution. After 20 minutes (which had previously been demonstrated to show linear product formation with time), the activity reaction was terminated by the addition of two volumes of acetonitrile containing internal standard (100 ng/ml [2H4]-1′-hydroxymidazolam). All reactions were carried out at 37°C in a humidified incubator (75% relative humidity, 5% CO2) in duplicate. Samples were vortexed and centrifuged for 5 minutes at approximately 2300g at room temperature. The supernatant was mixed with an equal volume of water containing 0.2% formic acid and analyzed directly by LC-MS/MS.
For several drugs, the assay conditions had to be modified to accommodate potent inhibition observed at the initial time point, rapid inactivation, or insolubility. Drugs with shortened primary incubation times (3, 5, 10, 15, 20, and 30 minutes) due to rapid inactivation include troleandomycin, tabimorelin, telithromycin, boceprevir, erythromycin, amlodipine, imatinib, saquinavir, cobicistat (3 µM), conivaptan (3 µM), nelfinavir (3 µM), and mibefradil (3 µM). The drugs with solubility issues when prepared in media required dilution directly from organic stocks (prepared at 100 times the final incubation concentration) into suspension hepatocytes: simvastatin, alectinib, carfilzomib, midostaurin, sorafenib, tadalafil, and mibefradil (3 µM); these drugs were tested at a final density of 0.5 million hepatocytes/ml.
LC-MS/MS Methodology for the Quantitation of 1′-Hydroxymidazolam
LC-MS/MS analysis was conducted on a Sciex 6500 triple quadrupole mass spectrometer (Framingham, MA) fitted with an electrospray ion source operated in positive ion mode using multiple reaction monitoring. An Agilent 1290 binary pump (Santa Clara, CA) with a CTC Leap autosampler (Leap Technology, Carrboro, NC) was programmed to inject 10 µl of sample on a Halo 2.7 µm C18 2.1 × 30 mm column (Advanced Materials Technology, Wilmington, DE). A binary gradient was employed using 0.1% (v/v) formic acid in water (mobile phase A) and 0.1% (v/v) formic acid in acetonitrile (mobile phase B) at a flow rate of 0.5 ml/min. Mass-to-charge transitions for analytes 1′-hydroxymidazolam and [2H4]-1′-hydroxymidazolam were 342.2 → 324.2 and 346.2 → 328.2, respectively. Analytes were quantified using Analyst software (Sciex). The peak area ratio of analyte to internal standard was used to determine inactivation rates.
Estimation of Observed Inactivation Rate
Data analysis methods have been previously described (Yates et al., 2012). Briefly, kobs was determined by normalizing the peak area ratio in each sample to that of the mean solvent control area ratio in the initial time point, plotting the natural log of percent remaining activity versus preincubation time, and then calculating the slope of the line (−kobs) using the initial linear portion of the curve. A statistical test was done for each drug to determine whether kobs was statistically different from the within-experiment solvent control (i.e., a parallel lines test), as shown in eq. 1.
Here kobs[I] and kobs[0µM] represent the inactivation rate for an inhibitor at a single concentration and inactivation rate with solvent control, respectively. When P < 0.05, there is statistically significant or measurable TDI. Analyses were performed using Microsoft Excel (Redmond, WA) and GraphPad Prism (La Jolla, CA). To account for between-assay variability, kobs of the within-experiment solvent control was subtracted from each drug kobs. (In hepatocytes, any subsequent formation of glucuronide conjugates of 1′-hydroxymidazolam was not measured but would be expected to be formed at small levels, and furthermore, kobs values are calculated relative to the solvent control in which this secondary metabolism would also be occurring.)
Confusion Matrix Analyses
This analysis was done separately for the HLM and HHEP results. True positives were defined as compounds that were above the kobs boundary line and had an observed AUCR ≥2. True negatives were compounds that were below the kobs boundary line and had an observed AUCR <2. False positives were compounds that were above the kobs boundary line but did not exhibit DDI (AUCR < 2), whereas false negatives were below the kobs boundary line and did exhibit DDI (AUCR ≥ 2). The boundary line was set at the lowest kobs associated with a clinical DDI of >2-fold. Positive predictive value describes the chance of an in vitro positive result exhibiting a clinical DDI (eq. 2).
The analysis was repeated, but rather than using the kobs boundary line, the parallel lines test kobs P value was used to define the in vitro TDI positives (P < 0.05) and negatives (P ≥ 0.05).
Results
Clinical drug interactions (AUCRs) for CYP3A DDI studies for 50 drugs are summarized in Table 1. The majority of studies (43) used midazolam as the CYP3A substrate, whereas the remaining studies employed alprazolam (1), triazolam (2), simvastatin (2), terfenadine (1), or buspirone (1). Of the drugs listed in Table 1, 26 drugs have no clinical drug-drug interaction (AUCR < 1.25), 4 have weak interactions (AUCR 1.25–2), and 20 have moderate [AUCR 2–5 (6)] or strong [AUCR > 5 (14)] interactions.
Representative examples of plots of natural log of percent activity remaining versus time determined in HLMs are presented in Fig. 1. These plots show CYP3A activity lost over time for NADPH-supplemented incubations containing solvent (solvent control), a drug that is statistically different from solvent control (eplerenone), a moderate inactivator (imatinib), and a fast inactivator (disulfiram). For most drugs, a 30 µM test concentration was used. However, lower concentrations were used for some drugs because of limits on aqueous solubility, potent reversible inhibition, or rapid time-dependent inhibition. A summary of kobs determined in HLMs and HHEPs are presented in Supplemental Table 1 and Table 2. Rates of inactivation (solvent control–corrected) from the HLM assay ranged from −0.001 (terbinafine) to 0.2333 minute−1 (0.3 µM ritonavir). By comparison, the average HLM solvent control kobs value was found to be 0.0056 ± 0.0014 minute−1 (S.D.). HHEP kobs values ranged from −0.0003 (citalopram, flumazenil) to 0.2826 minute−1 (troleandomycin). By comparison, the average HHEP solvent control kobs value was found to be −0.0010 ± 0.0030 minute−1 (S.D.). It should be noted that the solvent control kobs value for HLM was measurable, whereas the solvent control kobs in HHEP was essentially zero Fig. 2.
In general, drugs exhibit lower kobs values in the HHEP assay compared with that of HLM. Plots of HHEP versus HLM kobs for the 43 drugs determined in both systems at the same concentration are presented in Fig. 3. No apparent correlation is observed in the linear-scaled plot (Fig. 3A). However, for the vast majority of drugs, the HLM kobs was markedly higher than the HHEP kobs. A log-scaled plot (Fig. 3B) shows this trend. In contrast, three drugs—troleandomycin, saquinavir, and cobicistat—appear to have significantly higher HHEP kobs compared with their HLM kobs values.
A total of 50 drugs were evaluated for CYP3A TDI in HLMs and ranked in order of increasing kobs in Fig. 4A. Arranging the data from low to high, diltiazem (kobs of 0.011 minute−1) is the first drug for which a clinical DDI magnitude exceeds the 2-fold boundary line. A boundary line corresponding to the kobs of 0.01 minute−1 is established, above which a 2-fold clinical interaction is more likely. Similarly, Fig. 4B shows data for 44 drugs evaluated for CYP3A in the HHEP TDI assay. A boundary of 0.006 minute−1 for the HHEPs is proposed. The drug with the lowest kobs value at 30 µM that shows a 2-fold DDI in vivo is diltiazem, and thus a boundary for kobs set from that drug would be 0.006 minute−1. (It should be noted mibefradil, which yielded a kobs of 0.0045 minute−1, had to be tested at 0.3 µM and not the standard 30 µM concentration used for other drugs because inactivation was too rapid to be accurately measured at the 100-fold higher concentration. At 30 µM, mibefradil would have shown a kobs much greater than the 0.006 minute−1 cutoff.) Diltiazem serving as the boundary would be consistent for both in vitro systems. The data were further evaluated in a confusion matrix for sensitivity and specificity of these assays (Fig. 5). For DDI, a binary categorization for DDI is defined by a boundary of 2-fold, which for making decisions early in the drug design process is deemed adequate, as opposed to the strict bioequivalence cutoff values for DDI used in drug regulatory definitions (see explanation in the Discussion below). For the in vitro data, two different criteria for cutoffs were evaluated: (a) kobs values deemed statistically different from solvent control (by P < 0.05) and (b) the aforementioned cutoff values for kobs derived empirically from the in vivo DDI data (i.e., for diltiazem; 0.01 minute−1 for HLMs and 0.006 minute−1 for HHEPs). In all cases, there were no instances of false negatives, and all positives were correctly identified. [Using the (b) criteria, this is obviously the case since the boundary was defined by the lowest kobs value for a drug known to cause a DDI.] Thus, these assays are highly sensitive to classify molecules exhibiting in vitro TDI. However, the false positive rates translating kobs directly to clinical DDI are as high as 40% when using the criteria of kobs statistically different from solvent control and 28% when using the kobs cutoff criteria. Thus, these assays, although sensitive, have unsatisfactory specificity. The highest positive predictive value was 65%, which was obtained using data from HHEPs and a kobs cutoff value of 0.006 minute−1. When using these assays in a prospective manner, as in early drug discovery, compounds not demonstrating TDI can be progressed with confidence that they will not cause DDI, but those demonstrating in vitro TDI may or may not cause DDI.
Discussion
Drugs exhibiting time-dependent inhibition of CYP3A is one of the main causes of drug interactions by virtue of the large number of drugs for which CYP3A-catalyzed metabolism is the main clearance mechanism. Because so many drugs are cleared by CYP3A, it is desired that new drug candidates lack the ability to cause TDI for CYP3A. TDI assays are employed early in the drug research process, sometimes as one of the earliest absorption, distribution, metabolism, and excretion assays performed along with other in vitro assays (e.g., metabolic lability, membrane permeability, etc.) that yield a fundamental knowledge of potential dispositional properties. Compounds exhibiting TDI for CYP3A are undesired, and early TDI data are used either in a simple binary filter fashion, or they are used to develop structure-activity relationships to aid medicinal chemists in the design of alternate compounds that lack this property. However, we and others (Zimmerlin et al., 2011) have observed generally high hit rates in CYP3A TDI assays using HLMs, and frequently the compounds exhibiting TDI possess none of the well known structural entities that have been associated with mechanism-based inactivation of cytochrome P450 enzymes (Orr et al., 2012; Kalgutkar, 2017). Furthermore, utilization of CYP3A in vitro TDI parameters (kinact/KI) generated in HLMs in DDI prediction algorithms frequently led to overestimations of in vivo DDI (Rowland Yeo et al., 2011; Vieira et al., 2014). This leaves drug design teams with the ambiguity of whether the CYP3A TDI data generated is relevant to clinical DDI for their newly synthesized compounds and useful in decision-making of compound progression or for use in development of structure-activity relationships. Thus, the objective of the present study was to address three questions: 1) Do drugs that do not cause DDI with CYP3A yield measurable TDI data in HLM?; 2) If so, is there an empirical boundary line for in vitro TDI values under which the TDI data can be disregarded as an important design attribute?; and 3) Is there a difference in this relationship for TDI data generated in human liver microsomes versus human hepatocytes?
In HLMs, the drugs evaluated in this study can be considered in three groups: 1) those that demonstrate a measurable kobs values and are known to cause a DDI (e.g., clarithromycin, verapamil, etc.), 2) drugs that show no detectable kobs (i.e., no statistical difference from solvent control) and also did not cause a DDI for a CYP3A-cleared drug (e.g., fluoxetine, terbinafine), and 3) drugs that show measurable and statistically significant kobs values but do not cause a meaningful DDI (AUCR < 2) on a CYP3A-cleared drug (e.g., propranolol, paroxetine, erlotinib) (Fig. 4A). The latter group can be considered false positives; TDI is observed in vitro but DDI is not observed in vivo (Fig. 5). It is certainly the case that there are other very important considerations when attempting to relate in vitro TDI data to in vivo DDI, most important being the dose and exposure of the time-dependent inhibitor. However, in early drug research efforts, the potential dose that will be needed for efficacy is not yet known, and accordingly, design teams are faced with uncertainty when new compounds exhibit TDI. Thus, using the set of drugs in this study, boundary lines for kobs values were established for decision-making. For HLMs, a kobs value of 0.01 minute−1 was established as a boundary under which there are no drugs known to cause a clinical DDI with a CYP3A-cleared drug. Thus, even when kobs values can be statistically different from solvent control, these compounds are highly unlikely to be of any concern for a DDI liability. Above this cutoff value, there are still many drugs that do not cause a clinical DDI, whereas others do. Drug design teams can continue to pursue compounds in this category, but the risk of ultimately pursuing a compound that will cause DDI is greater, especially if the dose ultimately needed for clinical efficacy is high.
Although cytochrome P450 TDI assays are well established in HLMs, reports have been emerging on the use of suspended HHEPs for this measurement (Zhao et al., 2005; Xu et al., 2009; Chen et al., 2011; Mao et al., 2011, 2012, 2013, 2016; Kimoto et al., 2019). Thus, an evaluation was undertaken in which nearly the same set of drugs were tested as TDI in HHEP suspension assays. It was noteworthy that with the exception of three drugs, the kobs values in hepatocytes were much lower than those in liver microsomes (Fig. 3). Interestingly, the relationship did not exhibit a simple linear shift, which is indicative that the difference in TDI observed in HLMs versus HHEPs is unlikely to be related to a single mechanism. The statistical difference from solvent control was observable at lower kobs values because the rate of loss of CYP3A activity in hepatocytes was much slower than that for liver microsomes (Fig. 2). The same interrogation of the data for cutoff boundaries was done for the hepatocyte data, and a similar observation could be made as for microsomes (Fig. 4B), albeit the absolute value for the boundary line was lower for hepatocytes than that for microsomes. Hepatocytes offer additional complexity that may better represent the actual in vivo situation relative to liver microsomes. Reactive intermediates generated by CYP3A that may be the cause of TDI in liver microsomes could be quenched by further metabolism by conjugating enzymes active in hepatocytes (e.g., glutathione transferase, glucuronosyl transferase, and others) but absent in microsomes. In hepatocytes, there is a membrane barrier between the medium to which the test compound is dosed and the cytochrome P450 enzymes inside the cells, and test compounds that have low membrane permeability could be expected to cause lower inactivation rates. Other unknown mechanisms could be operating in the intact cellular milieu that protect CYP3A from inactivation that are not operative in liver microsomes. Currently, a reason for the differences between TDI data generated in microsomes and hepatocytes is unknown.
Overall, generation of this in vitro TDI dataset with drugs that have been evaluated for CYP3A DDI in the clinic has shown that many drugs will exhibit TDI in vitro but show no clinically meaningful DDI in vivo. In some instances, this may be because the clinical dose given and/or free exposure is low (e.g., ethinyl estradiol), but in others the reason is unknown. Boundaries are established such that when below a kobs value (at a test concentration of 30 µM generally used in this investigation) there is little concern that the test compound will be a perpetrator of a clinically meaningful DDI, even when the rate of loss in CYP3A activity is statistically distinguishable from the solvent control. Furthermore, when a test compound demonstrates a kobs value above the boundary it does not necessarily mean that the compound will cause DDI, only that the likelihood of this is greater. The boundary kobs values reported here, 0.01 minute−1 for liver microsomes and 0.006 minute−1 for suspended hepatocytes, were determined for the pooled lots used in this investigation. Specific values may vary with different preparations. Such simple boundaries can be of use when TDI data are generated early in a drug research process and decisions need to be made regarding whether the TDI data need to be taken seriously or whether the risk of DDI may be exaggerated. This can be used in conjunction with knowledge of the chemical structure (i.e., whether the compound showing TDI possesses a structure known to cause TDI) as well as follow-up experiments to uncover a mechanism for the TDI (e.g., bioactivation to reactive intermediates, generation of a metabolite-intermediate complex, adduction to protein or heme) (Hollenberg et al., 2008; Orr et al., 2012). These latter experiments can confirm the kinetic observation of TDI with mechanistic evidence.
It is important to note that these experiments were designed with early drug research decision-making processes in mind. For these purposes, we considered a clinical DDI magnitude of 2-fold to be relevant for decision-making at this point. Most pharmacotherapies can withstand 2-fold changes in exposure without deleterious outcomes, and it is only for those drugs with extremely narrow therapeutic indices (e.g., alfentanyl, cisapride, cyclosporine, fentanyl, terfenadine, quinidine, or tacrolimus) in which a doubling of exposure can be harmful. This is consistent with current regulatory guidance regarding the preferred approach to define no-effect DDI boundaries for substrate drugs, which should be based on concentration-response relationships derived from pharmacokinetic and pharmacodynamic analyses as well as other information, such as the maximum-tolerated dose (Food and Drug Administration, 2020). In contrast, when no-effect boundaries are not clearly defined by the former approach, drug regulatory agencies define increases in exposure of a mere 1.25-fold to be a cutoff for DDI, which is recognized as a very conservative standard for drugs that have wide safety margins. Nevertheless, in early drug research, multiple parameters and properties undergo simultaneous optimization (e.g., potency, clearance, and oral absorption), which is a difficult task, and thus driving drug design to avoid nominal DDIs (i.e., 1.25–2.0-fold) is of lower priority. As drug candidates progress into the development phase, more elaborate TDI experiments (e.g., KI/kinact) can be done to generate the data needed for projecting clinical DDI through physiologically based pharmacokinetic modeling, and through this process, more refined predictions can be made and include more subtle DDIs. The reported findings of the present work must be viewed as for the purposes of defining simple cutoff values for early decision-making.
In summary, this report established that many drugs that do not cause DDI with CYP3A yield measurable TDI data in human liver microsomes and hepatocytes. Empirical boundary lines for in vitro TDI values under which the TDI data can be disregarded as an important design attribute were established, with approximately 1.7-fold–lower boundary conditions in hepatocytes relative to liver microsomes. Future endeavors of this research include understanding why the different values are observed between liver microsomes and hepatocytes, why some compounds show TDI without possessing one of the known structural elements associated with this phenomenon, and how TDI data can be better used in predicting clinical DDI.
Acknowledgments
The authors would like to thank Lauren Horlbogen for conducting several in vitro studies.
Authorship Contributions
Participated in research design: Eng, Tseng, Cerny, Goosen, Obach.
Conducted experiments: Eng, Tseng.
Performed data analysis: Eng, Tseng, Cerny, Obach.
Wrote or contributed to the writing of the manuscript: Eng, Tseng, Cerny, Goosen, Obach.
Footnotes
- Received January 1, 2021.
- Accepted March 18, 2021.
This work received no external funding.
Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
↵This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- AUCR
- area-under-the-plasma-concentration-time-curve ratio
- DDI
- drug-drug interaction
- HHEP
- human hepatocyte
- HLM
- human liver microsome
- KI
- inhibition constant
- kinact
- maximal rate of enzyme inactivation
- kobs
- rate constant for inactivation
- LC-MS/MS
- tandem liquid chromatography–mass spectrometry
- TDI
- time-dependent inhibition
- Copyright © 2021 by The American Society for Pharmacology and Experimental Therapeutics