Abstract
A series of cytochrome P450 (P450) inhibition experiments were conducted with four hepatic uptake substrates (AZ3, AZ25, atorvastatin, and pitavastatin) using hepatocytes and recombinant P450s. The uptake was shown to be temperature-dependent and was inhibited by estrone sulfate, signifying an active component. At the lowest concentrations tested, the inhibitors concentrated up to 1000-fold in rat hepatocytes, but demonstrated only 5-fold greater P450 inhibition relative to recombinant rat P450s, indicating high intracellular binding. Inhibitor accumulation was considerably lower in human hepatocytes and an increase in inhibitory potency relative to recombinant human P450s was not obvious. This study highlights several technical and conceptual issues in the study of P450 inhibition mediated by compounds actively transported across the basolateral hepatocyte membrane. Primarily, the incubation medium concentration once the inhibitor has fully accumulated into the hepatocytes rather than the starting medium concentration, along with the extent of intracellular binding, must be considered as a foundation for in vitro-in vivo extrapolations. Additionally, it is suggested that if the Km value for the active uptake process is close to the P450 inhibition Ki, hepatocytes may be used only to establish the free drug accumulation ratio at a clinically relevant drug concentration, and this information, along with the (recombinant P450) Ki value, may be used to simulate the likely impact of active hepatic uptake on P450 inhibition in vivo.
Drug-drug interactions (DDIs) that result in a high number of hospital admissions and even deaths (Lazarou et al., 1998; Köhler et al., 2000) are in many cases caused by drug cotherapy resulting in plasma concentrations of one of the coadministered drugs being elevated to toxic levels (Hollenberg, 2002). The mechanism is frequently competition at the active site of drug-metabolizing enzymes. Great importance is now placed on in vitro studies as tools for predicting in vivo DDIs, particularly those resulting from cytochrome P450 (P450) inhibition (Lin and Lu, 1998), because the metabolic elimination of a large number of drugs is dependent on the P450 family of enzymes. In vitro screening for potential DDIs is therefore a crucial part of the drug discovery and development paradigm, exemplified by P450 inhibition assays typically performed with recombinant human P450s or human liver microsomes. Over the last decade considerable progress has been made in the mechanistic understanding of such DDIs, with increased accuracy of quantitative predictions made from in vitro data (Brown et al., 2006; Grime and Riley, 2006; Obach et al., 2006). Such predictions require an accurate prediction of the free drug concentration at the enzyme site.
Primary hepatocytes provide the in vitro model closest to the intact liver, because they contain a full complement of drug-metabolizing enzymes. Additionally, unbound drug has to cross the hepatocyte membrane, which not only provides a permeability barrier but also contains active drug transporter proteins. Drug transporters, such as members of the OATP family, located on the basolateral membrane can result in unbound drug concentrations in excess of those found in the blood, which may have an impact on the pharmacological effect (Sai and Tsuji, 2004), toxicity (Mizuno and Sugiyama, 2002; Nozawa et al., 2004), and drug elimination (Webborn et al., 2007).
Despite an awareness that active uptake into hepatocytes should be taken into consideration in the prediction of DDIs (Ito et al., 1998), comparatively few in vitro studies using isolated hepatocytes have been published (Cohen et al., 2000; Di Marco et al., 2003; Jones et al., 2004; McGinnity et al., 2005, 2006; Oleson et al., 2005; Zhao et al., 2005; Brown et al., 2007; Mano et al., 2007). Of these studies, only a small subset have compared inhibition constants generated using recombinant P450s or hepatic microsomes with those generated using isolated hepatocytes for inhibitors that are actively concentrated into hepatocytes. This is surprising, given that several HMG-CoA reductase inhibitors (statins) are not only P450 inhibitors but also actively accumulate in hepatocytes (Ito et al., 1998). Of particular interest is the fact that these studies found little difference between hepatocyte and microsomal P450 inhibition (Cohen et al., 2000).
The aim of this study was to investigate the functional consequences of active transport on P450 inhibition in hepatocyes. Four compounds were chosen: AZ3 (a moderately lipophilic carboxylic acid, logD7.4 of 1.8) and AZ25 (a moderately lipophilic carboxylic acid, logD7.4 of 2.3) from a previous study investigating the impact of hepatic uptake on drug clearance (Soars et al., 2007), atorvastatin, and pitavastatin. Atorvastatin is a substrate for (human) OATP1B1 (Shitara et al., 2005; Lau et al., 2006) and (rat) oatp1a4 and oatp1b2 (Lau et al., 2006). The profile for pitavastatin is similar (Shitara et al., 2005, 2006). In addition, the metabolic intrinsic clearance of all of the compounds is low in rat (12, 4, 14, and 7 μl/min/million viable cells) and human (4, 4, 3, and 15 μl/min/million viable cells) hepatocyte incubations (in-house data), ensuring that rapidly changing inhibitor concentrations due to metabolism should not lead to inaccuracies in data interpretation. Consideration was also given to several other important factors that can confound interpretation of the data and possibly explain the paucity of information on this subject in the scientific literature.
Materials and Methods
Chemicals. All chemicals and reagents used were of the highest available grade. Pitavastatin and atorvastatin were obtained from Sequoia Research Products Ltd. (Oxford, UK). All other chemicals were purchased from Sigma-Aldrich (Poole, Dorset, UK). AZ compounds were synthesized at AstraZeneca R&D Charnwood (Loughborough, UK).
Recombinant P450. Bactosomes (membranes prepared from Escherichia coli coexpressing human CYP2C9 and human P450 reductase or human CYP3A4 and human P450 reductase) were purchased from Cypex (Dundee, UK). Supersomes (membranes prepared from baculovirus-infected insect cells coexpressing rat P450s and P450 reductase and cytochrome b5) were purchased from BD Gentest (Woburn, MA).
Human Hepatocytes. Freshly isolated human hepatocytes were obtained from the UK Human Tissue Bank (Leicester, UK) after appropriate consent and ethical approval. Immediately after isolation, hepatocyte viability was >80% (assessed by the trypan blue exclusion test).
Preparation of Rat Hepatocytes. Isolation of rat hepatocytes was performed essentially using the two-step in situ collagenase perfusion method of Seglen (1976), as previously described (Soars et al., 2007). The cells were finally suspended in ice-cold hepatocyte suspension buffer (pH 7.4 at 37°C) comprising sodium HEPES (2.34 g/l), d-fructose (0.4 g/l), and Dulbecco's modified Eagle's medium (1-liter powder; Sigma, Gillingham, UK). Immediately after isolation, hepatocyte viability was >80% (assessed by the trypan blue exclusion test).
Measurement of Drug Uptake into Rat Hepatocytes. The method was adapted from the centrifugal filtration technique of Petzinger and Fuckel (1992) as follows. A vial containing hepatocytes (0.5 ml at a concentration of 2 million viable cells/ml) was preincubated for 5 min in a waterbath (37°C). For the experiments pertaining to the results presented in Table 1, the dose solutions (1 ml) were prepared by addition of DMSO stock solutions (20 μl of atorvastatin, pitavastatin, AZ3, or AZ25 at 0.02, 0.06, 0.2, 0.6, 1.7, and 5 mM) to incubation buffer. The incubation buffer used was the hepatocyte suspension buffer described above. For the experiments pertaining to the results presented in Fig. 1, the dose solutions contained 2 μM atorvastatin or pitavastatin (5 μl of the 5 mM DMSO stock solution was diluted in 12.5 ml of incubation buffer) plus DMSO (20 μl/ml dose solution) or 1000 μM estrone sulfate (20 μl of a 50 mM DMSO stock solution/ml dose solution). Additional experiments were performed to determine whether diclofenac, midazolam, or salicylamide inhibited the hepatocyte accumulation of atorvastatin. The dose solutions contained 2 μM atorvastatin plus DMSO (20 μl/ml dose solution) or 2 μM diclofenac or 2 μM midazolam (20 μl of a 100 μM DMSO stock solution/ml dose solution) or salicylamide (20 μl of a 150 mM DMSO stock solution/ml dose solution). Reactions were initiated by the addition of prewarmed (37°C) dose solution (0.5 ml) to the hepatocyte suspension (0.5 ml) so that incubations contained 1 μM atorvastatin or pitavastatin plus 500 μM estrone sulfate or 1 μM diclofenac or 1 μM midazolam or 1.5 mM salicylamide. Aliquots (100 μl) were removed at 10, 20, 30, 40, and 60 s, for the experiments relating to the results described in Fig. 1 or at 0 s and 10 min for the experiments relating to the results described in Table 1. The aliquots were immediately centrifuged at 7000g for 30 s in microtubes containing 15 μl of cesium chloride (lower layer) and 150 μl of silicone oil (upper layer) using a MiniSpin centrifuge (Eppendorf, Cambridge, UK). During this process the hepatocytes pass through the oil into the cesium chloride solution. Each centrifuge tube was then frozen in liquid nitrogen and cut, with collection of the cell pellet in a plastic tube. After addition of methanol (200 μl) and water (100 μl), the samples were vortexed and centrifuged. Supernatants were analyzed as described below. The data referred to as “4°C” in Fig. 1 were generated using the same experimental protocol except that the preincubations and the incubations were performed on ice.
Determination of the Appearance Rate of [3H]Estrone Sulfate into Human Hepatocytes. The experimental procedure was essentially as described above for rat hepatocyte uptake, with the exception that radiolabeled estrone sulfate was used as the substrate (final concentration of 3 μM, 2120 mBq/mmol). The frozen cell pellets were cut from the oil tubes into scintillation vials. After the addition of scintillation cocktail, the amount of radioactivity in the cells was determined using a Packard 2200CA Tri-Carb liquid scintillation counter (PerkinElmer Life and Analytical Sciences, High Wycombe, UK).
Measurement of Drug Loss from the Incubation Medium. P450 inhibitor stocks were prepared in dimethyl sulfoxide at 100-fold incubation concentration and aliquots of these stocks (10 μl) were added to vials containing hepatocyte suspension buffer (490 μl). A vial containing either rat or human hepatocytes at a concentration of 2 million viable cells/ml was preincubated for 5 min in a shaking (80 oscillations/min) waterbath at 37°C along with the vial containing the drug-buffer mix. Reactions were initiated by adding hepatocyte suspension (500 μl) to the drug/buffer mix (500 μl). Aliquots (80 μl) were removed at 0 and 10 min and placed into centrifuge tubes. These aliquots were immediately centrifuged (7000g, 30 s) using a MSE MicroCentaur centrifuge (Fisher Scientific, Loughborough, UK), and an aliquot of the supernatant (40 μl) was pipetted into ice-cold methanol (120 μl). Samples were then frozen for 1 h at -20°C and centrifuged (2000g for 20 min at 4°C). The supernatants were removed and analyzed as described below.
Determination of Rat P450 Selectivity for the Diclofenac 4-Hydroylation and Midazolam 1′-Hydroxylation Reactions. Incubations were performed with six rat P450s (CYP1A2, CYP2B1, CYP2C6, CYP2C11, CYP2D1, and CYP3A2) chosen because of their relevance to drug metabolism in male rats (Mugford and Kedderis, 1998; Kobayashi et al., 2002; Caron et al., 2005). Diclofenac (1 μM incubation concentration) or midazolam (1 μM incubation concentration) was incubated with each rat P450 (individually) for 10 min. The incubation buffer was the same as that used for the hepatocyte incubations described above and incubations were initiated by the addition of NADPH (50 μl of a 10 mM stock solution to a total incubation volume of 0.5 ml). Reactions were terminated by the removal of aliquots (100 μl) into ice-cold methanol (200 μl). Samples were then frozen for 1 h at -20°C and centrifuged (2000g for 20 min at 4°C). The supernatants were removed and analyzed as described below.
Enzyme Kinetics.Time and enzyme concentration linearity. Incubations were performed for 2, 5, and 10 min using 0.1 μM diclofenac and concentrations of CYP2C6 and (separately) CYP2C11 of 5, 10, 20, and 50 pmol/ml. Similarly, incubations were performed for 2, 5, and 10 min using 0.1 μM midazolam and CYP3A2 concentrations of 5, 10, 20, and 50 pmol/ml. In addition, incubations were performed with rat hepatocytes (1 million viable cells/ml) for 5 and 10 min. The incubation buffer was the same as that used for the hepatocyte incubations described above, and incubations were initiated by the addition of NADPH (50 μl of a 10 mM stock solution to a total incubation volume of 0.5 ml). Reactions were terminated by the removal of aliquots (100 μl) into ice-cold methanol (200 μl). Samples were then frozen for 1 h at -20°C and centrifuged (2000g for 20 min at 4°C). The supernatants were removed and analyzed for 4-hydroxydiclofenac and 1′-hydroxymidazolam as described below.
Km determination. Incubations were performed for 5 min with 20 pmol/ml CYP2C6 and (separately) CYP2C11 and a range of concentrations of diclofenac (0.1, 0.2, 0.6, 1.9, 5.6, 13, 25, and 50 μM). Similarly, incubations were performed for 5 min with 20 pmol/ml CYP3A2 and a range of concentrations of midazolam (0.1, 0.2, 0.6, 1.9, 5.6, 13, 25, and 50 μM). The incubation buffer was the same as that used for the hepatocyte incubations described above, and incubations were initiated by the addition of NADPH (50 μl of a 10 mM stock solution to a total incubation volume of 0.5 ml). Reactions were terminated by the removal of aliquots (100 μl) into ice-cold methanol (200 μl). Incubations were also performed in a shaking (80 oscillations/min) water bath for 5 min with isolated rat hepatocytes (1 million viable cells/ml) and the same concentrations of diclofenac and midazolam listed above. Reactions were terminated by the removal of aliquots (100 μl) into ice-cold methanol (200 μl). Samples were then frozen for 1 h at -20°C and centrifuged (2000g for 20 min at 4°C). The supernatants were removed and analyzed for 4-hydroxydiclofenac and 1′-hydroxymidazolam as described below.
Assessment of P450 Inhibition Using Recombinant Rat and Human P450. Incubation mixtures comprised CYP2C6 or CYP2C11 or CYP2C9 (5 pmol/ml), substrate (1 μM diclofenac), and test inhibitor (atorvastatin, pitavastatin, AZ3, or AZ25 at incubation starting concentrations of 0.2, 0.6, 1.9, 5.6, 17, and 50 μM). The incubation buffer was the same as that used for the hepatocyte incubations described above. The diclofenac was added as an aliquot (5 μl) from a methanolic stock solution (100 μM) to give an incubation concentration of 1 μM. The inhibitors were added as aliquots (5 μl) from DMSO stock solutions (prepared as 100× the required incubation concentration). In control incubations, DMSO (5 μl) was added in place of the test inhibitor. The final incubation volume was 0.5 ml, and the incubations were initiated by the addition of NADPH (50 μl of a 10 mM stock solution to the 0.45-ml incubation mixture). Reactions were terminated after 5 min by the removal of aliquots (100 μl) into ice-cold methanol (200 μl). Samples were then frozen for 1 h at -20°C and centrifuged (2000g for 20 min at 4°C). The supernatants were removed and analyzed for 4-hydroxydiclofenac and 1′-hydroxymidazolam as described below. Experiments to determine the extent of CYP3A2 and CYP3A4 inhibition were performed in exactly the same way with the one exception that the substrate was midazolam, added as an aliquot (5 μl) from a methanolic stock solution (100 μM) to give an incubation concentration of 1 μM.
Assessment of P450 Inhibition Using Isolated Rat and Human Hepatocytes. Rat or human hepatocytes (2 million viable cells/ml) were preincubated for 5 min at 37°C. The test P450 inhibitor (atorvastatin, pitavastatin, AZ3, or AZ25) dose solutions were prepared to be double the incubation concentrations of 0.2, 0.6, 1.9, 5.6, 17, and 50 μM as follows. Aliquots (10 μl) of DMSO stock solutions (0.02, 0.06, 0.19, 0.56, 1.7, and 5 mM) were added to incubation buffer (0.5 ml). In control incubations, DMSO (5 μl) was added in place of the test inhibitor. The dose solutions were prewarmed to 37°C. To initiate the “uptake period” for the test inhibitors, dose solutions (0.5 ml) were added to hepatocyte suspensions (0.5 ml). The incubations were allowed to proceed in a shaking (80 oscillations/min) water bath for 10 min before addition of P450 substrate (10 μl of 100 μM methanolic stock solutions to give final incubation concentrations of 1 μM diclofenac or midazolam). The incubations were allowed to proceed for a further 5 min before being terminated by the removal of aliquots (100 μl) into ice-cold methanol (200 μl). Samples were then frozen for 1 h at -20°C and centrifuged (2000g for 20 min at 4°C). The supernatants were removed and analyzed for 4-hydroxydiclofenac and 1′-hydroxymidazolam as described below.
Simulation of Intracellular Unbound Atorvastatin Concentrations. As described previously, a physiologically based model was used to simulate the in vitro kinetics of atorvastatin (Paine et al., 2008). The model includes terms for membrane and intracellular binding and for passive and active uptake into isolated rat hepatocytes. All the parameters were the same as those defined previously, including intracellular atorvastatin binding of 99% (Paine et al., 2008), except that the model was refined to incorporate saturation of the uptake process. A Km of 20 μM was used for Fig. 2A, to concur with the findings of Lau et al. (2006) and an uptake intrinsic clearance of 132 μl/min/million hepatocytes (using data from Fig. 1), such that the Vmax was estimated as 2.64 nmol/min/million cells. For Fig. 4, Km values of 300, 30, 3, and 1 μM were used. Simulations were performed using Berkeley Madonna version 8.3 (http://www.berkeleymadonna.com), and media and unbound intracellular concentrations at 10 min were obtained. For Fig. 2A, the simulated unbound intracellular concentrations were used to estimate the percentage of P450 activity remaining by reference to the recombinant CYP2C11 inhibition curve. The x-axis was the simulated media concentration at 10 min.
Analysis of Samples Containing Atorvastatin, Pitavastatin, AZ3, and AZ25. Mass spectrometry was performed on a Micromass Quattro Ultima Platinum triple quadrupole mass spectrometer (Waters, Manchester, UK) using a Hewlett Packard 1100 high-performance liquid chromatography system (Hewlett Packard, Palo Alto, CA) for analyte separation. The analysis mode was multiple reaction monitoring (positive ionization). Cone voltage and collision energy were optimized for each compound. Samples (10 μl) were injected on to an HPLC column (Hypersil Gold C18, 2.1 × 30 mm, 3 μm; Thermo Electron Corporation, Basingstoke, UK) and chromatographic separation was achieved with a mobile phase gradient as follows: t = 0 min % organic = 5, t = 0.5 min % organic = 5, t = 1.5 min % organic = 100, t = 2.25 min % organic = 100, t = 2.3 min % organic 5; total run time = 2.5 min). The mobile phase consisted of water with formic acid (0.1% v/v) and methanol with formic acid (0.1% v/v) and the HPLC flow rate was set at 1 ml/min, introduced into the mass spectrometer source at 0.2 ml/min.
Analysis of Samples Containing Diclofenac and Midazolam. Mass spectrometry was performed on a TSQ Quantum Ultra triple quadrupole mass spectrometer (Thermo Scientific, San Jose, CA) using a Surveyor MS high-performance liquid chromatography system (Thermo Scientific) for analyte separation. The analysis mode was multiple reaction monitoring (positive ionization). Tube lens and collision energy were optimized for each compound. Samples (10 μl) were injected on to an HPLC column (Hypersil Gold C18, 2.1 × 100 mm, 3 μm; Thermo Electron Corporation) and chromatographic separation was achieved with a mobile phase gradient as follows: t = 0 min % organic = 5, t = 0.5 min % organic = 5, t = 2.5 min % organic = 100, t = 3.5min % organic = 100, t = 3.6 min % organic 5; total run time = 5.0 min). The mobile phase consisted of water with formic acid (0.1% v/v) and methanol with formic acid (0.1% v/v) and the HPLC flow rate was set at 0.4 ml/min introduced into the mass spectrometer source.
Results
Measurement of Drug Uptake into Hepatocytes. Atorvastatin and pitavastatin displayed time-dependent accumulation into isolated rat hepatocytes when incubated at 37°C. The rate of accumulation was greatly reduced when the incubations were performed on ice (nominally denoted as 4°C in Fig. 1). Coincubation with estrone sulfate (500 μM) also decreased the rate of hepatocyte uptake by approximately 75% (Fig. 1). Similar data were obtained for AZ3 and AZ25 (data not shown). Neither diclofenac, midazolam, or salicylamide inhibited the rate of atorvastatin accumulation into rat hepatocytes, showing that the two P450 ligands and the glucuronidation inhibitor would not impact on the assessment of the potency of the actively transported P450 inhibitors in hepatocytes. Because the mass balance calculations revealed a close agreement between appearance of drug into rat hepatocytes and drug loss from the incubation medium (see below), incubations to assess the intracellular drug concentrations were not repeated with human hepatocytes. Instead, the uptake rate for estrone sulfate was determined so that the OATP activity in isolated human hepatocytes could be compared with values quoted in the relevant scientific literature. Rates of uptake assessed at 37°C for the three human hepatocyte preparations used were 128, 225, and 144 pmol/min/million viable hepatocytes. These values agreed well with data available in the scientific literature (Yamashiro et al., 2006; Soars et al., 2007).
Measurement of Drug Loss from the Incubation Medium. When the data presented in Table 1 are scrutinized, it is clear that there is close agreement between the “drug loss from the medium” data and the “drug appearance in the rat hepatocytes” data. For example, in the atorvastatin rat hepatocyte incubations, the media concentrations dropped from 1.8 to 0.4 μM during the 10-min uptake period, when the starting concentration was 2 μM. So, from Fig. 2A, 0.4 μM results in 30% inhibition of diclofenac 4-hydroxylation, not 2 μM, as viewed from the medium concentration. Clearly at this starting concentration, 1.4 μM (1.4 nmol in 1 ml of incubation medium) atorvastatin has accumulated in 4 μl (the intracellular volume of 1 million hepatocytes) (Reinoso et al., 2001). This gives an estimated intracellular concentration of 350 μM atorvastatin, which agrees well with the measured concentration of 390 μM (Table 1). However, this high concentration only resulted in the same percentage inhibition of diclofenac 4-hydroxylation as 2 μM atorvastatin in recombinant rat CYP2C11 incubations, indicating extensive intracellular binding of atorvastatin. Similar calculations using the remaining data in Table 1 reveal good mass balance between the two assays.
The “% Drop” column in Table 1 gives a simple way to visualize the extent of uptake. For example, when the starting incubation concentration of pitavastatin was 0.2 μM, 80% of the drug accumulated in the rat hepatocytes after 10 min of incubation at 37°C. If we assume the intracellular volume of 1 million hepatocytes to be 4 μl (Reinoso et al., 2001), this represents a concentration of 1000-fold from a 1-ml incubation containing 1 million hepatocytes. (That is, by 10 min, 80% of the 0.2 nmol of pitavastatin had accumulated into 4 μl, giving an intracellular concentration of 40 μM. At this point the medium concentration was 20% of the initial 0.2 μM. The concentration ratio was, therefore, 40/0.04 or 1000). These “% Drop” data provide evidence that for each of the four compounds, the uptake into rat hepatocytes was more extensive than that into human hepatocytes at the corresponding incubation concentrations. It is also clear that the extent of uptake into rat and human hepatocytes decreases as the starting incubation media concentrations rise. This is additional proof that an active process is responsible for the uptake of the compounds and suggests that the Km values for the active uptake process for all four of the compounds are close to the range of the incubation concentrations used in the study.
Determination of Rat P450 Selectivity for the Diclofenac 4-Hydroylation and Midazolam 1′-Hydroxylation Reactions and Enzyme Kinetics. Of six rat P450s selected for study, only CYP3A2 was found to metabolize midazolam to the 1′-hydroxy product. Both CYP2C6 and CYP2C11 metabolized diclofenac to the 4-hydroxy metabolite, with CYP3A2 also performing the reaction to a minor extent (less than 10% when compared with CYP2C6 and CYP2C11). The amount of 1′-hydroxy midazolam formed was linear to the 5-min incubation time and 20 pmol of CYP3A2/ml incubation. The same was true for the formation of 4-hydroxydiclofenac by CYP2C6 and CYP2C11. The two reactions were linear to 10 min in incubations with rat hepatocytes (1 million viable cells/ml). The Km estimates for recombinant CYP3A2 and rat hepatocyte-dependent midazolam 1′-hydroxylation were identical (2 μM). The Km estimates for CYP2C6, CYP2C11, and rat hepatocyte-dependent diclofenac 4-hydroxylation were 2, 7, and 2 μM, respectively.
Assessment of P450 Inhibition Using Recombinant Rat and Human P450 and Isolated Rat and Human Hepatocytes. Of the four compounds being tested, only AZ25 was found to be a potent inhibitor of CYP3A2 (IC50 value of 4 μM). The compound was a less potent inhibitor of CYP3A4 (22 μM). Therefore, the majority of the work presented and discussed refers to inhibition of diclofenac 4-hydroxylation. Inhibition profiles were generated using the starting incubation concentration of inhibitor for the recombinant P450 reactions. For rat and human hepatocyte experiments, both the starting incubation medium concentration and the medium concentration after 10 min of incubation with the hepatocytes and inhibitor (before the addition of P450 substrate) were used to assess the inhibition profiles (Fig. 2). When the starting incubation concentrations were used, there was effectively no difference between the rat hepatocyte and recombinant rat P450 inhibition profiles. However, when the incubation medium concentration after a 10-min preincubation with the hepatocytes and P450 inhibitor were used, there was a left shift in the rat hepatocyte inhibition profiles that was more pronounced at the lower inhibitor concentrations (Figs. 2 and 3). This was not obvious in the human hepatocyte inhibition profiles, with only a slight shift being detectable at the low inhibitor concentrations.
Simulation of Intracellular Unbound Atorvastatin Concentrations.Figure 2A shows good agreement between the observed and simulated shift in rat hepatocyte and recombinant P450 inhibition profile, suggesting that the model and the input parameters are at least plausible. Figure 4 demonstrates that when the Km for the drug uptake transporter impinges on the P450 inhibition IC50, the shape of the hepatocyte inhibition profile changes, potentially leading to an inaccurate estimation of the IC50 from hepatocyte experiments.
Analysis of Samples Containing Diclofenac and Midazolam. The retention time of the 4-hydroxydiclofenac standard (multiple reaction monitoring transition of 312 > 230) was 3.53 min and the corresponding single peak was present in the mass chromatograms for the recombinant CYP2C11 samples. However, the mass chromatograms for the rat hepatocyte samples contained three peaks for the multiple reaction monitoring transition 312 > 230 (Fig. 5). Further analysis using the transition 488 > 230 demonstrated that the two additional peaks were glucuronide metabolites of 4-hydroxydiclofenac (Fig. 5). A mass chromatogram from a sample of a rat hepatocyte incubate containing 1 μM diclofenac and 1.5 mM salicylamide contained only the peak with a retention time of 3.53 min (4-hydroxydiclofenac). No glucuronide metabolites of 1′-hydroxy midazolam were detected.
Discussion
The aim of this study was to investigate the functional consequences of active transport on P450 inhibition and to examine the impact of any cross-species differences between rat and human hepatocytes. AZ3 and AZ25 were chosen for the experiments along with atorvastatin and pitavastatin, two drugs known to be actively transported into hepatocytes by (human) OATP1B1 and (rat) oatp1a4 and oatp1b2 (Shitara et al., 2005, 2006; Lau et al., 2006). All four compounds accumulated into isolated rat hepatocytes at 37°C at a much greater rate than when the experiments were performed on ice (Fig. 1). This finding coupled with the facts that estrone sulfate, a known (rat) oatp and (human) OATP substrate (Meier et al., 1997; Hirano et al., 2004; Noé et al., 2007), greatly reduced the uptake rate of the compounds and that the extent of uptake at 10 min was shown to be concentration-dependent, substantiated the role of active transport in this intracellular accumulation.
There is the potential to regard P450 inhibition from a consideration of incubation medium or intracellular free drug concentration. However, because clinical DDI risk is best evaluated from the perspective of free blood concentration (Grime and Riley, 2006), the incubation medium concentration should be used as a foundation for in vitro extrapolations. In this study, incubations were conducted for 10 min before the addition of the P450 substrates to allow time for the inhibitors to accumulate to pseudo-steady state in the hepatocytes. This time period has previously been shown to allow near-maximal accumulation into the hepatocytes (Soars et al., 2007). Incubations were performed to determine the concentration of the P450 inhibitors in the hepatocytes and in the incubation medium (by rapidly centrifuging the cells and sampling the supernatant) at 0 and 10 min. Six concentrations of inhibitors were used, and in each case the time 0 medium concentration was lower than the starting incubation concentration (Table 1). This difference is probably due to the instantaneous binding of the inhibitors to the hepatocyte membrane, as described previously (Reinoso et al., 2001). For example, the difference between the total incubation concentration and the medium concentration at time 0 in the 2 μM atorvastatin-rat hepatocyte incubation was 0.24 μM (12% incubational binding at a rat hepatocyte concentration of 1 million/ml). Membrane binding of 9% is predicted from the logD7.4 of 1 (Austin et al., 2005) and from the compartmental model described by Paine et al. (2008).
Given suitable accuracy of parameter estimation, physiologically based pharmacokinetic modeling should provide the most thorough method for predicting clinical DDIs, as changing drug concentrations can be simulated in all relevant compartments, including the liver (Kanamitsu et al., 2000). On the other hand, a pragmatic approach using a single inhibitor concentration in the blood can be acceptable in some circumstances (Grime and Riley, 2006). The added complexity of active drug transport necessitates the use of a compartmental modeling approach incorporating all of the relevant parameters, including active transport Km and Vmax, passive permeability, intra- and extracellular binding, elimination Km and Vmax for the P450 inhibitor, and Ki for the inhibitor-P450 interaction defined using recombinant P450s. To clarify the last statement, it is noteworthy that it may be inappropriate to determine kinetic inhibition constants (IC50 or Ki) from hepatocyte incubations, as the Km for any inhibitor-drug transporter interaction may be similar to the P450 inhibition IC50 or Ki. Under these circumstances the inhibition curves will be affected (as shown in Fig. 4), and this effect can lead to difficulties in inaccurately estimating IC50 or Ki. Such a problem is evident from the data presented in Table 1, which shows that the intracellular concentrations do not increase linearly with the starting incubation concentrations (the extent of cellular uptake slows with increasing drug concentration, suggesting that the Km for the transporter is in this concentration region). Indeed, the Km values reported for the atorvastatin oatp1a4, oatp1b2, and OATP1B1 interactions are 22, 7, and 12 μM respectively (Kameyama et al., 2005; Lau et al., 2006). Therefore, it is recommended that, rather than establishing a full inhibition profile, hepatocyte experiments should ascertain the extent of inhibition at relevant drug concentrations (that is, equivalent to the steady-state unbound blood concentrations observed clinically) and use this information along with the Ki value determined using recombinant P450s to simulate the likely impact of active hepatic uptake on P450 inhibition in vivo.
Recently, a physiologically based compartmental model has been used to describe the disposition of drugs (including atorvastatin) as a result of basolateral hepatic drug transport (Paine et al., 2008). When this model was used to predict the inhibition profile of diclofenac 4-hydroxylation by atorvastatin, using a Km for the active uptake process of 20 μM and intracellular binding of 99%, the model well predicted the extent and shape of the observed inhibition profile (Fig. 2A). Further subtleties can be incorporated into this mathematical modeling approach where appropriate. These include the well documented polymorphisms in the OATP family (Mizuno and Sugiyama, 2002; Shitara et al., 2006) and canalicular efflux (atorvastatin and pitavastatin are known to be substrates of P-glycoprotein and breast cancer resistance protein, respectively) (Zhao et al., 2005; Ieiri et al., 2007).
In the compartmental model described by Paine et al., 2008), the fraction of atorvastatin unbound inside the hepatocytes is predicted to be 0.01 (99% bound)., 2008). In the present study, an intracellular atorvastatin concentration of 39 μM gave the same percentage of inhibition as 0.2 μM in the recombinant CYP2C11 incubation, confirming the fact that the intracellular binding was approximately 99%. The same exercise performed for the other three compounds revealed a similar extent of binding inside the hepatocytes. Appreciation that binding may remove a large proportion of the free drug concentrated inside the hepatocyte is not novel (Di Marco et al., 2003; Jones et al., 2004), but it is a concept that is occasionally overlooked. For drugs not actively transported into hepatocytes, extensive intracellular binding still occurs, but the impact is relatively minor because the hepatocytes are a part of the whole incubation system (typically 1 ml, because there is no compartmentalization) and the free drug concentration therefore refers to the whole incubation. Under conditions of free drug accumulation inside the hepatocytes, intracellular drug binding is specifically relevant to the small intracellular volume (typically 4 μl for an incubation of 1 million hepatocytes/ml), and the impact on intracellular free drug concentration is more obvious. In this study the data clearly show that for low concentrations of atorvastatin, pitavastatin, AZ3, and AZ25, free drug accumulates approximately only 5-fold despite a cell-medium concentration factor of approximately 1000 (Table 1). This finding has potential implications for pharmacological, toxicological, and DDI predictions. In other words, the active transport process works to achieve a particular unbound drug concentration ratio, and the extent of intracellular binding dictates, at least in part, the incubation medium (or blood) concentration required to achieve this ratio at equilibrium.
The data submitted here, indicating at most only a 2-fold concentration of unbound drug inside human hepatocytes, initially appear perplexing: statins such as atorvastatin exert their pharmacological selectivity because of hepatic targeting (van Vliet et al., 1995; Shitara and Sugiyama, 2006). There are only a limited amount of data in the literature comparing cell-free and rat and human hepatocyte HMG-CoA reductase activity, summarized in a recent review (Shitara and Sugiyama, 2006). In one report the ratio of HMG-CoA reductase inhibitory potency of atorvastatin in rat hepatocytes was found to be approximately 2 orders of magnitude greater than that in other cell types (Chapman and McTaggart, 2002) and yet here we report an estimated cell/media unbound drug accumulation ratio of only 5-fold. However, akin to the data presented here, other researchers have also failed to observe a human liver microsomal to hepatocyte increase in P450 inhibitory potency for a range of statins (Cohen et al., 2000). An explanation for this puzzle may lie in the fact that 5-fold unbound drug accumulation (observed in this study) is less accumulation than it is possible to obtain with this type of experiment (Paine et al., 2008) and that in vitro experiments may slightly underestimate in vivo active hepatic uptake (Soars et al., 2007). If one assumes similar issues with the estimation of in vivo human hepatic free drug concentration, the hepatic selectivity of the statins is more intelligible. In recognition of these issues, some caution should be given to in vitro-in vivo DDI extrapolations when OATP-dependent active transport is involved. One solution would be to include an additional 5- to 10-fold safety factor if the experimental data for marker compounds such as atorvastatin and pitavastatin are similar to those reported in this study. Nevertheless, the concentration effect due to rat oatp is likely to be greater than that due to human OATP (Sandker et al., 1994; Soars et al., 2007).
In executing these experiments, consideration was given to each part of the process such as allowing time for the P450 inhibitor to accumulate inside the hepatocytes before performing the P450 inhibition assay, potential interaction of P450 inhibitor and P450 substrate with drug transporters and conjugation of the P450-produced metabolite from which the inhibition was monitored. The data showed midazolam to be primarily metabolized to 1′-hydroxymidazolam by rat CYP3A2, confirming previous findings (Kobayashi et al., 2002). Thus, male rats show close similarity with humans for this route of metabolism, in which CYP3A4 is the dominant enzyme responsible (Weaver et al., 2003). As previously demonstrated (Caron et al., 2005), the current study showed that both rat CYP2C6 and CYP2C11 metabolized diclofenac to the 4-hydroxy metabolite, whereas in humans only CYP2C9 is primarily responsible for this route of metabolism (Weaver et al., 2003). This result could have confounded the interpretation of the data, but several observations strongly support the view that CYP2C11 is the dominant male-specific CYP2C isoform in rats (Mugford and Kedderis., 1998), and it was therefore appropriate to compare CYP2C11, rather than CYP2C6, inhibition profiles with those generated using rat hepatocytes. The Km values determined for the biotransformation performed with rat hepatocytes and recombinant CYP2C11 were identical (2 μM) and lower than the value estimated for the recombinant CYP2C6-dependent reaction (7 μM). In addition, atorvastatin was a more potent inhibitor of CYP2C6 than CYP2C11-dependent diclofenac 4-hydroxylation (IC50 values of 10 and 26 μM, respectively) and yet the IC50 value for the inhibition of the reaction catalyzed by rat hepatocytes was found to be 10 μM; thus, the intracellular binding would have to be effectively complete for it to be plausible that CYP2C6 plays a dominant role in rat hepatocyte diclofenac 4-hydroxylation. Finally, the shift in CYP3A2 to the rat hepatocyte inhibition profile for 1′-hydroxymidazolam was very similar to that observed for 4-hydroxydiclofenac inhibition.
An extra possible complication was further metabolism of the 4-hydroxydiclofenac and 1′-hydroxymidazolam. Two glucuronide metabolites of 4-hydroxydiclofenac were detected in the rat (Fig. 5) but not the human hepatocyte incubates. A high concentration of salicylamide (1.5 mM) inhibited the formation of the glucuronide metabolites and did not inhibit the active transport of the four P450 inhibitors or the formation of 4-hydroxydiclofenac by recombinant CYP2C11. However, inclusion of salicylamide (1.5 mM) did not alter the inhibition profile of atorvastatin in rat hepatocytes, suggesting negligible impact of glucuronidation during the 5-minute incubation. For simplicity, therefore, salicylamide was not included in the experiments used to generate Fig. 2. Another factor requiring preliminary investigation was the extent to which diclofenac or midazolam inhibited the active transport of the four P450 inhibitors. Fortunately, neither P450 substrate inhibited the active transport process.
In summary, this study highlights the in vitro experimental complications encountered in the attempt to predict the impact of active drug transporters on DDIs. As a result of the experimental findings, it is suggested that a comparison of inhibition data obtained with hepatocytes and recombinant P450 should be used to assess the enzyme inhibition of compounds that undergo significant cellular uptake. The P450 inhibition Ki should be obtained from the recombinant data and the relevant unbound concentration should be that measured in the hepatocyte incubation medium after uptake is expected to be complete, which in turn should reflect the circulating unbound concentration in vivo. An additional point of interest from the data presented here is that for atorvastatin, pitavastatin, AZ3, and AZ25, the concentration effect due to rat oatp is greater than that due to human OATP. This is not an uncommon observation (Sandker et al., 1994; Soars et al., 2007) but clearly has potential implications for the interpretation of hepatic toxicology in preclinical rat safety studies and estimation of human safety margins.
Acknowledgments
We thank Peter White for the analytical work in identifying and separating the diclofenac glucuronides.
Footnotes
-
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
-
doi:10.1124/dmd.108.021055.
-
ABBREVIATIONS: DDI, drug-drug interaction; P450, cytochrome P450; OATP/oatp, organic anion transporter peptide; DMSO, dimethyl sulfoxide; HPLC, high-performance liquid chromatography.
- Received February 15, 2008.
- Accepted May 9, 2008.
- The American Society for Pharmacology and Experimental Therapeutics